UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Self-directed tablet-based naming therapy in chronic aphasia Rowe, Jacob 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_september_rowe_jacob.pdf [ 1.48MB ]
Metadata
JSON: 24-1.0305685.json
JSON-LD: 24-1.0305685-ld.json
RDF/XML (Pretty): 24-1.0305685-rdf.xml
RDF/JSON: 24-1.0305685-rdf.json
Turtle: 24-1.0305685-turtle.txt
N-Triples: 24-1.0305685-rdf-ntriples.txt
Original Record: 24-1.0305685-source.json
Full Text
24-1.0305685-fulltext.txt
Citation
24-1.0305685.ris

Full Text

SELF-DIRECTED TABLET-BASED NAMING THERAPY IN CHRONIC APHASIA  by Jacob Rowe  B.A., The University of British Columbia, 2013  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Audiology and Speech Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  June 2016  © Jacob Rowe, 2016 ii  Abstract Background: The recent interest in the use of mobile devices (like iPads and their application programs; apps) in aphasia therapy has been motivated in part by a belief that they could facilitate self-directed home practice with minimal therapist supervision. Such practice in turn could allow people with aphasia to receive more therapy, without taxing current clinical resources even further. However, it is unknown whether mobile technology can feasibly facilitate self-directed practice. Furthermore, it is unclear whether this kind of therapy can improve outcomes, and there are doubts about the quality of the therapy process when therapy is provided without the expertise of a therapist. Aims: This study aimed to explore the feasibility, therapeutic effect, and therapy process of a naming therapy using an iPad-based therapy app for minimally-supervised self-directed home practice in chronic post-stroke aphasia. Method: A single-subject experiment using a multiple baseline design was replicated across three participants with chronic aphasia, who independently used an iPad-based therapy app to practice naming pictures with the support of cues. Outcome measures included extensive usage and accuracy data unobtrusively collected using the app’s internal logging system, pre- and post-therapy language measures, and qualitative semi-structured interviews to explore participants’ experiences of therapy and perceptions of therapy effects. Results: Therapy was found to be accessible and acceptable for participants with aphasia; however, problems were identified with the therapist’s administrative role in developing and managing word sets, and monitoring therapy progress. Participants demonstrated medium to large naming accuracy gains for practiced words, but minimal gains for unpracticed words. iii  Enabling participants to direct their own therapy process led them to spontaneously enact therapy in ways generally similar to how a therapist might. Conclusion: Despite limitations, self-directed practice using mobile technology may have the potential to improve outcomes by making ongoing long-term gains feasible, and to foster greater collaboration and shared expertise between therapists and people with aphasia. Findings were discussed in terms of suggestions for future research and therapy app development that could allow mobile technology to deliver on this potential. iv  Preface The University of British Columbia Behavioural Research Board reviewed and approved this study: UBC Ethics Certificate number H14-02433.    All aspects of the development of the research design, collection and analysis of the research data, and preparation of this thesis were completed primarily by the author, Jacob Rowe, with input from the researcher supervisors, Drs. Barbara Purves and Jeff Small, and thesis committee member, Dr. Paola Colozzo. Although the author collaborated with Tactus Therapy Solutions Ltd. to develop a special research version of one of their commercially available therapy application programs for use solely during this study, Tactus Therapy Solutions Ltd. provided this service pro bono, the author is not an employee of Tactus Therapy Solutions Ltd., and he is not receiving financial compensation from Tactus Therapy Solutions Ltd. v  Table of Contents  Abstract .......................................................................................................................................... ii!Preface ........................................................................................................................................... iv!Table of Contents .......................................................................................................................... v!List of Tables ................................................................................................................................ xi!List of Figures ............................................................................................................................. xiii!List of Abbreviations ................................................................................................................. xiv!Glossary ....................................................................................................................................... xv!Acknowledgements ..................................................................................................................... xx!Chapter 1: Introduction ............................................................................................................... 1!1.1! Aphasia and its treatment ................................................................................................... 1!1.2! The accessibility of mobile technology for people with aphasia ....................................... 3!1.3! A framework for quantifying therapy ................................................................................ 6!1.4! Quantity of therapy: Filling a dosage gap with mobile technology-based therapy ........... 7!1.4.1! Dosage in neuroscience, learning, and aphasia research ............................................ 8!1.4.2! Therapy services for aphasia in clinical practice ...................................................... 10!1.5! Quality of therapy: Learning through mobile technology ............................................... 13!1.5.1! Dose form: Personal relevance and self-directed practice ........................................ 13!1.5.2! Therapeutic procedures: Cueing, feedback, and performance-contingency ............. 16!1.6! Self-directed computer- and mobile technology-based aphasia therapy ......................... 20!1.7! Research questions ........................................................................................................... 25!Chapter 2: Method ...................................................................................................................... 27!vi  2.1! Overview of methods ....................................................................................................... 27!2.2! Design .............................................................................................................................. 30!2.3! Participants ....................................................................................................................... 31!2.3.1! Screening and clinical profile measures ................................................................... 31!2.3.2! Recruitment ............................................................................................................... 35!2.3.3! Demographics and clinical profiles .......................................................................... 35!2.3.4! Experience and perceptions of technology and therapy ........................................... 40!2.3.5! Summary of participant characteristics ..................................................................... 44!2.4! Materials .......................................................................................................................... 44!2.4.1! Tablet ........................................................................................................................ 45!2.4.2! Naming Therapy© .................................................................................................... 45!2.4.3! Word set stimuli ........................................................................................................ 47!2.4.3.1! Intuition-based nomination ................................................................................ 47!2.4.3.2! Frequency-based nomination ............................................................................. 48!2.4.3.3! Participant nomination ....................................................................................... 49!2.4.4! Outcome measures .................................................................................................... 50!2.4.4.1! Naming Therapy© data logging system ............................................................. 51!2.4.4.2! Probe picture naming ......................................................................................... 52!2.4.4.3! Secondary outcome measures ............................................................................ 55!2.5! Procedure ......................................................................................................................... 55!2.5.1! Phase 1: Pre-therapy screening and clinical profile evaluation ................................ 56!2.5.2! Phase 2: Word set development and baseline ........................................................... 56!2.5.3! Phase 3: Therapy ....................................................................................................... 59!vii  2.5.3.1! Minimally-supervised self-directed naming therapy ......................................... 59!2.5.3.2! Word set management ........................................................................................ 60!2.5.4! Phase 4: Post-therapy outcome evaluation ............................................................... 62!2.6! Analysis ............................................................................................................................ 62!2.6.1! Feasibility .................................................................................................................. 63!2.6.2! Therapeutic effect ..................................................................................................... 64!2.6.3! Therapy process ........................................................................................................ 68!2.7! Summary of methods ....................................................................................................... 68!Chapter 3: Results ....................................................................................................................... 70!3.1! Feasibility ......................................................................................................................... 70!3.1.1! Word set development .............................................................................................. 70!3.1.2! Adherence ................................................................................................................. 75!3.1.3! Accessibility .............................................................................................................. 79!3.1.4! Remote progress monitoring ..................................................................................... 83!3.1.5! Acceptability ............................................................................................................. 86!3.1.6! Summary of findings for the feasibility analysis ...................................................... 90!3.2! Therapeutic effect ............................................................................................................ 92!3.2.1! Naming accuracy and naming acquisition ................................................................ 92!3.2.2! Secondary outcomes ............................................................................................... 100!3.2.3! Summary of findings for the therapeutic effect analysis ........................................ 104!3.3! Therapy process ............................................................................................................. 105!3.4! Summary of results ........................................................................................................ 107!Chapter 4: Discussion ............................................................................................................... 109!viii  4.1! Summary and interpretation of findings ........................................................................ 110!4.1.1! Feasibility ................................................................................................................ 110!4.1.1.1! Word set development ..................................................................................... 110!4.1.1.1.1! Efficiency of nomination methods ............................................................ 110!4.1.1.1.2! Functional relevance of nominated words ................................................ 112!4.1.1.2! Adherence ........................................................................................................ 116!4.1.1.3! Accessibility ..................................................................................................... 118!4.1.1.4! Remote progress monitoring ............................................................................ 121!4.1.1.5! Acceptability .................................................................................................... 123!4.1.2! Therapeutic effects .................................................................................................. 125!4.1.2.1! Naming accuracy and naming acquisition ....................................................... 125!4.1.2.2! Secondary effects and clinical significance ..................................................... 127!4.1.3! Therapy process ...................................................................................................... 131!4.1.3.1! Dosage .............................................................................................................. 131!4.1.3.2! Therapeutic procedures .................................................................................... 135!4.2! Limitations ..................................................................................................................... 141!4.2.1! Research design and internal validity ..................................................................... 141!4.2.2! External validity ...................................................................................................... 144!4.2.3! Measurement ........................................................................................................... 146!4.2.4! Analysis ................................................................................................................... 149!4.3! Conclusion ..................................................................................................................... 150!References .................................................................................................................................. 153!Appendices ................................................................................................................................. 171!ix  Appendix A Summary of variables in feasibility analysis ...................................................... 171!Appendix B Summary of variables in therapeutic effect analysis .......................................... 173!Appendix C Summary of variables in therapy process analysis ............................................. 174!Appendix D Pre-therapy participant-perspective interview guide .......................................... 175!Appendix E Quotes from pre-therapy participant-perspective interviews ............................. 176!Appendix F iPad-Naming Therapy© navigation path for therapy picture naming ................. 178!Appendix G Optional Naming Therapy© cueing system ....................................................... 179!Appendix H Five-number summaries of lexical characteristics for the NT sets, the BOSS sets, and the participant-nominated word sets ................................................................................ 180!Appendix I Development of the NT and BOSS word sets ...................................................... 181!Appendix J Data collected by the Naming Therapy© data logging system ........................... 182!Appendix K Naming accuracy and naming acquisition scoring criteria for probe picture naming ..................................................................................................................................... 183!Appendix L Post-therapy participant-perspective interview guide ........................................ 184!Appendix M Variables used for word set matching ............................................................... 185!Appendix N Five-number summaries of the characteristics of matched therapy and control word sets for each participant ................................................................................................. 186!Appendix O Participants’ dosage adherence per therapy week for mean dose per word, mean session frequency per word, and mean cumulative dose ........................................................ 187!Appendix P Quotes from post-therapy participant-perspective interviews ............................ 188!Appendix Q Alternative and maladaptive button selections across participants, organized by step in the iPad-Naming Therapy© navigation path ............................................................... 192!x  Appendix R Five-number summaries of participants’ inter-selection intervals during therapy for each step in the iPad-Naming Therapy© navigation path ................................................. 193!Appendix S Five-number summaries of the duration of time participants took in therapy for various time scales .................................................................................................................. 194! xi  List of Tables Table 1. Overview of the study procedure. ................................................................................... 28!Table 2. Overview of the research questions and their sub-components. ..................................... 29!Table 3. Participants’ demographics and clinical profiles. ........................................................... 37!Table 4. Participants’ results on standardized tests contributing to their clinical profiles. .......... 38!Table 5. Intra- and inter-rater reliability of naming accuracy for each participant. ...................... 55!Table 6. Overview of research questions and their sub-components. ........................................... 63!Table 7. Overview of word set development variables and data sources. .................................... 70!Table 8. Number of participant-nominated words for each participant, the nomination strategies used to collect them, and their composition. ................................................................................. 71!Table 9. Attrition rates of, and the proportion of words in the finalized study word sets for, the nomination methods for each participant. ..................................................................................... 73!Table 10. Overview of adherence variables and data sources. ..................................................... 75!Table 11. Overview of accessibility variables and data sources. .................................................. 79!Table 12. Participants’ navigation scoring during therapy and probe sessions for each step in the basic iPad-Naming Therapy© navigation path. ............................................................................ 80!Table 13. Overview of the remote progress monitoring variable and data sources. ..................... 83!Table 14. Diagnostic 2x2 tables for each participant’s self-scoring categorization accuracy during probe sessions. .............................................................................................................................. 85!Table 15. Overview of acceptability variables and data sources. ................................................. 86!Table 16. Overview of naming accuracy and naming acquisition variables and data sources. .... 92!Table 17. Effect of therapy in absolute terms for therapy set words during the therapy phase for each participant. ............................................................................................................................ 93!xii  Table 18. Effect size estimates for probe picture-naming outcomes. ........................................... 95!Table 19. Overview of secondary outcome variables and data sources. .................................... 100!Table 20. Pre and post scores for secondary impairment-level outcome measures. ................... 101!Table 21. Overview of therapeutic procedure variables and data sources. ................................. 105!Table 22. Overview of research questions and their subcomponents. ........................................ 109! xiii  List of Figures Figure 1. P1’s dose per therapy day during self-directed practice and probe sessions. ................ 77!Figure 2. P2’s dose per therapy day during self-directed practice and probe sessions. ................ 77!Figure 3. P3’s dose per therapy day during self-directed practice and probe sessions. ................ 77!Figure 4. Distribution of therapy session start times for each participant during therapy. ........... 87!Figure 5. P1’s percentage of the therapy sets assigned to the practice set and accurate trials per trial series for therapy and control sets. ........................................................................................ 97!Figure 6. P2’s percentage of the therapy sets assigned to the practice set and accurate trials per trial series for therapy and control sets. ........................................................................................ 98!Figure 7. P3’s percentage of the therapy sets assigned to the practice set and accurate trials per trial series for therapy and control sets. ........................................................................................ 99!Figure 8. Distribution of cues used by each participant during self-directed practice. .............. 106! xiv  List of Abbreviations AAC Alternative and augmentative communication App Application program BNT-2 Boston Naming Test–Second Edition BOSS Bank of Standardized Stimuli CLQT Cognitive Linguistic Quick Test CT Computer-based therapy dB HL Decibel Hearing Level kHz Kilohertz LR+ Positive likelihood ratio LR- Negative likelihood ratio MTT Mobile technology-based therapy NT Naming Therapy© OR Odds ratio PN Participant-nominated SLP Speech-language pathologist SMD Standardized mean difference SMG Standardized mean gain Tablet Tablet personal computer WAB-R Western Aphasia Battery–Revised   xv  Glossary Computer-based therapy Therapy that is carried out on a personal computer.  Control word set The control set consisted of words that were never made available for practice, but rather were only presented before and after therapy. One of the three study word sets created during the word set development process was randomly assigned the role of control set. The others were randomly assigned as the therapy sets. Cueing  Broadly speaking, a behaviour intended to facilitate a client’s responses (Horton & Byng, 2000). Cues were operationalized in this study as the selection of a set of buttons on the Naming Therapy© Trial Screen (see Appendix G).  Cumulative dose  The sum of learning episodes across therapy. This term was originally labeled as ‘cumulative intervention intensity’ in Baker (2012) and Warren, Fey and Yoder’s (2007) framework; however, the term ‘intensity’ was avoided in this thesis in order to reduce confusion related to variations in its definition within and across literatures. Dosage  An umbrella term for the quantity of therapy and representing one component of the therapy process (the other being therapeutic procedures). It subsumes six variables: dose form (the activity or context within which learning episodes occur), dose (learning episodes per session), session duration, session frequency, total therapy duration, and cumulative dose (total number of learning episodes across therapy). This term was originally labeled as ‘intensity’ in xvi  Baker (2012) and Warren et al.’s (2007) framework; however, the term ‘intensity’ was avoided in this thesis in order to reduce confusion related to variations in its definition within and across literatures. Dose  The number of learning episodes per session of practice (Baker, 2012; Warren et al., 2007); also known as ‘repetition’ in the neuroscience literature (Kleim & Jones, 2008). Dose form  The context within which learning episodes occur, such as a therapy task. Dose form and learning episodes together define the quality of therapy in Baker (2012) and Warren et al.’s (2007) dosage framework. Feedback  Broadly speaking, a behaviour intended to modify a client’s response (Horton & Byng, 2000) that was produced spontaneously or facilitated via cueing. For example, the provision of accuracy information or explanation of a client’s behaviour are considered types of feedback. Learning episode  An individual or unique combination of therapeutic procedures. This term was originally labeled as ‘teaching episode’ in Baker (2012) and Warren et al.’s (2007) framework. However, ‘learning’ was substituted as a more generic term, as ‘teaching’ may suggest the presence of a therapist, which is not applicable to self-directed practice. Dose form and learning episodes together define the quality of therapy in Baker (2012) and Warren et al.’s (2007) dosage framework.  xvii   Mobile technology-based therapy  Therapy that is carried out on a mobile device. Mobile devices are wireless and/or handheld devices with touchscreen interfaces, such as smartphones or tablet personal computers (Brandenburg, Worrall, Rodriguez, & Copland, 2013). Monitor word set  One of the sets of words (the other being the practice set) presented to participants during probe sessions in the therapy phase. The monitor set consisted of words that were not currently available for practice, but were still presented for probe picture naming during probe sessions. Therapy set 2 was initially assigned to the monitor set; however, the monitor set’s composition was dynamic, changing in accordance with the process of word set management (see section 2.5.3.2). Performance-contingency The state of a behaviour as being systematically dependent on past behaviour (Keetch & Lee, 2007). Performance-independence refers to the state of a behaviour as not being systematically dependent on past behaviour.  Practice word set  A set of words that were made available to participants for practice during self-directed practice during the therapy phase. The practice set was also one of two word sets (the other being the monitor set) that was presented to participants during probe sessions in the therapy phase. Therapy set 1 was initially assigned to the practice set; however, the monitor set’s composition was dynamic, changing in accordance with the process of word set management (see section 2.5.3.2).  xviii  Self-directed practice  Practice in which the client has control over some or all aspects of the therapy process. It may be facilitated through the use of accessible therapy apps (mobile or computer technology), home practice, reduced therapist supervision, and/or the use of unfixed therapeutic procedures. SLP-based therapy Traditional therapy carried out by a speech-language pathologist. Study word set  An umbrella term for the therapy and control sets. Therapeutic procedure  A behaviour during therapy that is hypothesized to cause an improvement in a client’s targeted behavior, including therapeutic inputs and client acts (Baker, 2012; Warren et al., 2007). Therapeutic inputs in this study were operationalized as cueing and performance-contingency and client acts were operationalized as naming attempts. Therapist supervision  The degree to which a therapist is involved in carrying out the therapy process, ranging from no supervision (i.e., no therapist is involved in clients’ therapy process) to minimal supervision (i.e., the therapist does not actively enact the therapy process but is involved in monitoring progress and modifying therapy goals) to complete supervision (i.e., the therapist is physically present and actively enacts the therapy process; Zheng, Lynch, & Taylor, 2016). Therapy word set  One of the three study word sets created during the word set development phase that was randomly assigned the role of one of the two therapy sets. The others were randomly assigned as xix  the other therapy set and as the control set. Therapy set 1 was initially assigned to the practice set, whereas therapy set 2 was initially assigned to the monitor set. xx  Acknowledgements Above all, thank you to the participants of this study. It can sound like an empty platitude to say that the achievement of completing this research was really yours, rather than mine, but in this case there is a great deal of truth to it. Ultimately, my role in this study was auxiliary. You, the participants did most of the real work: you not only made most of the decisions regarding what, where, when, how, and how much to practice, but you then also went and actually carried out all of that work. Whether it is explicitly acknowledged or not, clients are the ultimate therapeutic agent of change; the technology used in this study just helped highlight that fact. Thank you also to participants’ families. Although not even officially participants, you were invaluable to this study through your support for your loved ones and your unwavering kindness to me. Thank you to the dedicated people at Tactus Therapy Solutions for developing the special research version of the application program used in this study, for your open-mindedness, and for putting up with all my endless questions, concerns, and suggestions. Without your willingness to support research, despite no guarantee of benefit to you or your company, much of this study would not have been possible.  Thank you to Dr. Barbara Purves for originally suggesting the idea of looking at using mobile devices in aphasia therapy and for the welcomed wealth of theoretical, clinical, and personal guidance and knowledge that you offered freely from the very beginning of this two-year project. Thank you to Dr. Jeff Small for encouraging me to think deeply and for willingly spending hours with me to discuss the finer points of methodology and comb through all the data. Thank you to Dr. Paola Colozzo for your perceptive second set of eyes, your earnest thoroughness, and your insightful questions. Thank you all for your persistent patience and xxi  steadfast support, and for offering me the freedom and space to see how far I could take my ideas. Finally, thank you to my family for your loving support and encouragement during this long and maddening journey, and for offering to read every word of my thesis out of solidarity with me. You probably did not realize the full (215-page) weight of that commitment when you made it, so reading it ‘in spirit’ would be fine too.1  Chapter 1: Introduction 1.1 Aphasia and its treatment Aphasia is a disorder of language use following brain damage. Beyond this generic description, no definition of aphasia is widely accepted (Code & Petheram, 2011). However, aphasia has been more narrowly defined as a multimodal impairment of language production and comprehension at the phonological, morphological, lexical semantic and/or syntactic level that typically results from a lesion to the left cerebral hemisphere (Code & Petheram, 2011; McNeil & Pratt, 2001). Available data on the incidence of aphasia after stroke ranges from 0.02% to 0.06% with a prevalence from 0.1% to 0.4% of the population in the developed world (Code & Petheram, 2011). In the U.S. alone, this translates to an estimated 83,000 new cases of aphasia per year, with a prevalence of one million people with some degree of aphasia – a proportion similar to Parkinson disease (Code & Petheram, 2011; Elman, Ogar, & Elman, 2000). These values are likely an underestimate due to variation in stroke and aphasia reporting, and will likely increase as stroke survival rates continue to improve and more people live longer with aphasia (Code & Petheram, 2011; Lakshminarayan, Anderson, Jacobs, Barber, & Luepker, 2009). Spontaneous recovery from aphasia usually occurs to some degree: large language improvements in the weeks and months after stroke are mainly related to reduced edema, reperfusion, and the resolution of diaschisis. In contrast, ongoing longer-term improvements primarily relate to neural network reorganization resulting from experience-dependent neuroplasticity, the mechanism by which the brain encodes experience and learns new behaviours (Kiran, 2012; Kleim & Jones, 2008; Salter, Teasell, Foley, & Allen, 2013).  In spite of spontaneous recovery, aphasia frequently results in a chronic communication disability. Aphasic impairments result in functional limitations and restricted participation in 2  many communication-related social, vocational, and recreational activities (Davidson, Worrall, & Hickson, 2003). These can ultimately have significant negative consequences on the quality of life of people with aphasia, often in the form of social isolation, disrupted relationships, stigmatization, loneliness, depression, loss of autonomy, role changes, and an identity of incompetence (Cherney, Patterson, & Raymer, 2011; Davidson, Howe, Worrall, Hickson, & Togher, 2008; Shaddon, 2005). Research has suggested consequences of aphasia can be severe, with Lam and Wodchis (2010), for example, finding that, out of 60 diseases, aphasia had the largest negative relationship with health-related quality of life among long-term care residents.  Considering this profound impact, it is perhaps not surprising that a collaborative priority-setting project – with equal participation of stroke survivors, caregivers, and health professionals – recently identified the treatment of aphasia as the third highest research priority of 226 distinct unanswered research questions related to life after stroke (Pollock, St. George, & Firkins, 2012). Although research on adjuvant pharmacological and electrophysiological treatments is emerging, behavioural aphasia therapy continues to be the standard and best-supported treatment for aphasia (Zumbansen & Thiel, 2014). Meta-analyses and systematic reviews, varying in methodology and adequacy, have mostly supported the efficacy of aphasia therapy in general (Brady, Kelly, Godwin, & Enderby, 2012; Cicerone et al., 2011; Robey, 1998; Robey, Schultz, Crawford, & Sinner, 1999; Salter et al., 2013). However, more specific evidence related to the effects of different kinds of aphasia therapy, the kinds of clients for whom they are appropriate, their active ingredients and optimal dosages, and their utility in clinical practice is currently limited (Brady et al., 2012; Ratner, 2006; Raymer et al., 2008). Accordingly, the present exploratory study examined such issues in relation to a very new kind of therapy, mobile 3  technology-based therapy (MTT), by evaluating a tablet-based naming therapy app in the context of self-directed home practice with minimal therapist supervision. The recent rapid development and adoption of mobile technologies, including the iPad (Apple Inc., Cupertino, CA) and other smartphone and tablet devices, has introduced potentially valuable new tools in the management of communication disorders (Brandenburg, Worrall, Rodriguez, & Copland, 2013; McNaughton & Light, 2013). These technologies have been received with a “barely restrained excitement” in clinical settings (Kurland, 2014, p. 3). This interest has been motivated primarily by the argument that mobile devices have the potential to boost the effects of traditional aphasia therapy carried out by a speech-language pathologist (SLP-based therapy), by allowing people with aphasia to independently achieve higher therapy dosages than is currently feasible in clinical practice (Brandenburg et al., 2013; Code & Petheram, 2011; van de Sandt-Koenderman, 2011). However, this argument rests on a number of suppositions: (a) mobile technologies are sufficiently accessible for people with aphasia to facilitate self-directed practice; (b) there is a therapy dosage level at which people with aphasia can make optimal improvements; (c) this optimum dosage is greater than what SLP-based therapy services can currently provide; and (d) the quality of the learning experience in MTT is sufficient such that it at least does not undermine potential therapy effects. Each of these assumptions is in dispute, and the relevant evidence to substantiate them is lacking, equivocal, or preliminary. The following sections detail the arguments for and against the use of mobile technology in aphasia therapy and evaluate the available evidence relevant to these arguments. 1.2 The accessibility of mobile technology for people with aphasia If people with aphasia are to benefit from practicing on their own by using mobile technology, they must first be able to access and operate mobile devices independently. Mobile 4  technologies have been argued to have unique characteristics that could increase the accessibility of SLP services for people with aphasia, compared to personal computers, traditional high-tech communication aids, and SLP-based services (Brandenburg et al., 2013; McNaughton & Light, 2013). However, a number of possible accessibility barriers to using mobile technologies have also been proposed that could limit their potential. The potential accessibility advantages of mobile technologies like iPads include their adaptability, wide popularity and availability, and affordability (Brandenburg et al., 2013; McNaughton & Light, 2013). The portable size, customizable touchscreen interface, and multimedia capabilities of mobile devices could allow them to be adapted to some extent for mobility, dexterity, language, cognitive, and sensory perceptual issues commonly implicated in aphasia, stroke, and aging (Brandenburg et al., 2013). Furthermore, the popularity of mobile technologies means more people with aphasia may already be open to their use in therapy and be familiar with their operational requirements, which could reduce the learning demands of new application programs (apps; McNaughton & Light, 2013). The wide availability of mobile devices, the ease of purchasing and downloading their wide range of apps, and their relatively low cost have made speech-language pathology interventions accessible to the general public (McNaughton & Light, 2013). Not only could this offer people with aphasia more options for, and control over, their healthcare decisions, but it could also open up the field of speech-language pathology to a wide range of programmers to potentially allow for faster, cutting-edge technical development (McNaughton & Light, 2013). However, the wide availability and popularity of mobile technologies are also their potential drawbacks. There is concern that the excitement for these new technologies could result in devices being purchased and used without adequate attention to whether such solutions are 5  most appropriate for the client’s abilities and needs (Kurland, 2014; McNaughton & Light, 2013). Because mobile technologies are marketed to the general public, they are designed with the average user in mind, so there are currently few access options available for clients with more complex motor, language, cognitive, or perceptual issues (Brandenburg et al., 2013; McNaughton & Light, 2013). Besides the device itself, the ever-increasing number of available apps can make it difficult to select ones that are appropriate. Moreover, most of the programmers developing apps have little knowledge of aphasia, so relatively few apps have been designed to be accessible for people with aphasia or to use evidence-based therapy approaches, further complicating the selection of appropriate apps (Brandenburg et al., 2013; McNaughton & Light, 2013). All of these issues appear particularly problematic in light of the fact that the availability and affordability of mobile technology mean that clients can now purchase therapy apps on their own without input from knowledgeable professionals who could help guide them through the process (McNaughton & Light, 2013). Very little research has directly evaluated mobile technologies’ accessibility for people with aphasia, and none has evaluated their proposed advantages and drawbacks relative to other service delivery options. However, by supplementing the few available studies with findings from related research, Brandenburg et al. (2013) were able to provide preliminary guidelines for accessible mobile technology for aphasia. The only other published journal articles on app use with aphasia are a number of reports based on clinicians’ personal experiences that contain suggestions for selecting appropriate apps, using apps in rehabilitation, and facilitating their use by clients (Holland, Weinberg, & Dittelman, 2012; McCall, 2012; Ramsberger & Messamer, 2014; Szabo & Dittelman, 2014). Thus, while the potential advantages of mobile technology may make it an appealing alternative in aphasia therapy and preliminary research is promising, a 6  number of outstanding issues must be addressed and further research on how people with aphasia can access mobile technology is needed (Brandenburg et al., 2013; McNaughton & Light, 2013). 1.3 A framework for quantifying therapy As a number of the arguments for and against self-directed MTT discussed in sections 1.4 and 1.5 below revolve around dosage and active ingredients of therapy, a useful framework is first introduced to help structure and clarify the review. Dosage in SLP research has been operationalized in a variety of ways, but commonly it involves the concept of repeated, spaced ‘units’ of therapy over a period of time (Baker, 2012; Warren, Fey & Yoder, 2007). These ‘units’ are typically defined in terms of session duration (e.g., 1 hour of therapy per unit time). However, using duration as the unit of analysis is insufficient for determining an optimal dosage, as it provides no information about the quality and quantity of the presumed active ingredients of a therapy (Baker, 2012; Warren et al., 2007). To do this instead, it is necessary to operationalize therapy in terms of learning episodes, which can then be counted to determine their quantity and frequency over a specified period of time (Baker, 2012; Warren et al., 2007). A learning episode is an individual or unique combination of therapeutic procedures, defined in observable behavioural terms, that are hypothesized to cause an improvement in a client’s targeted behavior (i.e., active ingredients). In Baker’s (2012) expanded version of the Warren et al. (2007) original framework, such procedures consist of therapeutic inputs (e.g., therapist actions) and client acts. Their framework also offers six variables that both give a more precise specification of dosage and could affect outcomes independently of one another: dose form (the activity or context within which learning episodes occur), dose (learning episodes per session), session duration, session frequency, total therapy duration, and cumulative dose (total number of learning episodes across therapy). 7  There is great potential in quantifying therapy this way. It could result in more consistent and precise reporting of dosage variables in the literature (Baker, 2012; Warren et al., 2007), which would permit a more thorough examination of dosage in meta-analyses and systematic reviews. Furthermore, when learning episodes are disregarded, the number of learning episodes per session (i.e., dose) is allowed to vary as a function of their rate of occurrence, creating heterogeneity within and between participants even when the other dosage variables are held constant (Warren et al., 2007). This is especially important when trying to control for dosage to evaluate the relative efficacy of different therapy approaches or when comparing different participants. Moreover, the unit of analysis of this framework requires a more explicit examination of how therapy is hypothesized to relate to outcomes (Warren et al., 2007). Testing such hypotheses, in combination with relevant moderator variables (e.g., client, therapist, or service-related factors), could help lead to a more explicit and refined theory of therapy, which has been argued to be an important goal in the development and adoption of evidence-based practice (Baker, 2012; Ratner, 2006; Warren et al., 2007). 1.4 Quantity of therapy: Filling a dosage gap with mobile technology-based therapy There is a commonly held belief that the dosage currently available through therapy services is insufficient for people with aphasia to make optimal gains, and that this problem can be corrected simply by increasing the quantity of therapy they receive (Cherney et al., 2011; Code & Petheram, 2011). Although this belief has motivated much research on ways to increase therapy dosage through trained volunteers, personal computers, and now mobile technologies (Brady et al., 2012; Brandenburg et al., 2013; Code & Petheram, 2011; van de Sandt-Koenderman, 2011), it has not actually received strong empirical support. Available evidence does suggest aphasia therapy services are relatively limited and underutilized; however, basic 8  and applied research from a variety of fields examining dosage is mixed and complex, and no optimal dosage of aphasia therapy has yet been established. 1.4.1 Dosage in neuroscience, learning, and aphasia research A large body of evidence from basic research on experience-dependent neuroplasticity suggests that high doses for long durations may be needed to induce and maintain neural changes and their corresponding functional benefits (Kleim & Jones, 2008). This could suggest that considerable, long-term use of a language behavior, past the point of acquisition (i.e., overlearning), may be needed to maintain gains in therapy (Raymer et al., 2008). However, this conclusion is primarily derived from animal models of motor and sensory learning, so it is unclear to what extend these principles of recovery can be applied to human language (Raymer et al., 2008). Moreover, even if it is applicable, this literature suggests that the relationship between dosage variables and neural and behavioural outcomes is more complex than just ‘more is better’, with quality and timing of practice, attentional salience, age, etiology, and impairment factors moderating the relationship (Kleim & Jones, 2008). Research on learning and memory in nonclinical populations may also be relevant to neurorehabilitation (Raymer et al., 2008). Evidence from this literature on the value of overlearning for long-term retention is equivocal (Peladeau, Forget, & Gagne, 2003; Rohrer & Taylor, 2006; Rohrer, Taylor, Pahler, Wixted, & Cepeda, 2005), and a meta-analysis of the substantial body of literature on the effects of manipulating practice frequency (i.e., distributed versus massed practice) in healthy individuals demonstrated that retention is actually greater and decreases at a slower rate after lower-frequency (i.e., distributed) practice than after higher-frequency (i.e., massed) practice (Cepeda, Pasler, Vul, Wixted, & Rohrer, 2006). However, like 9  the neuroplasticity literature, evidence from learning in healthy individuals can be extended to people with aphasia only by inference.  Similar to other areas of speech-language pathology research, dosage in aphasia therapy has been examined and reported haphazardly (Baker, 2012; Cherney, 2012). Instead of learning episodes, studies reporting dosage have used some measure of duration as the unit of analysis, such as session duration or hours per week (the latter of which conflates session duration and session frequency). The literature has also conflated the other dosage variables, especially equating session frequency to dosage in general, and experimental studies examining dosage effects have failed to control for multiple dosage variables, therapy type, and other moderators (Cherney et al., 2011). Such inconsistencies and imprecision in the definition of dosage and the effects of unmeasured moderators may contribute to the overall unclear findings of systematic reviews and meta-analyses regarding dosage-outcome relationships in aphasia therapy (Cherney, 2012; Cherney et al., 2011; Teasell, Foley, Hussein, & Speechley, 2013). For example, Robey (1998) reported large effect sizes associated with therapy provided for ≥2 hours per week; and Bhogal, Teasell, and Speechley (2003) found that studies with statistically significant therapy effects had average dosage schedules of 8.8 hours per week for 11.2 weeks, whereas those without significant effects had schedules of ~2 hours per week for 22.9 weeks. In contrast, systematic reviews of the few experiments directly manipulating dosage variables (primarily hours per week) have been mixed. Cherney et al.’s (2011) results were equivocal for both acute and chronic aphasia, with no clear difference between high- and low-dosage therapy across studies. In contrast, a Cochrane review (Brady et al., 2012) found some differential benefits of high-dosage therapy, but this conclusion was based on a small number of studies and was confounded 10  by differential attrition, perhaps suggesting that such high dosage levels (i.e., 7-20 hours per week) may not be tolerable for all clients. Thus, although the commonly held belief that ‘more therapy is better’ is arguably the primary rationale for using mobile technology in therapy (Cherney et al., 2011; Code & Petheram, 2011), it is clear from the literature that dosage-outcome relationships are not straightforward and further research is needed to critically examine the role of dosage in aphasia therapy. 1.4.2 Therapy services for aphasia in clinical practice Most of the research on the continuum of care that people with aphasia receive from acute onset and admission to hospital through to the chronic stage has been studied cross-sectionally (i.e., by examining different people across time points), but some limited longitudinal research (i.e., examining the same people across time points) also exists. Findings from both types of research suggest that there is great variability in the pathway people with aphasia take through the healthcare system and in the nature and extent of therapy services they receive along the way. This situation is likely related to underfunding, resource and staffing restrictions, and competing clinical demands in contemporary healthcare systems (Code & Heron, 2003; Code & Petheram, 2011; Foster, Worrall, Rose, & O’Halloran, 2015; Lalor & Cranfield, 2004). Cross-sectional research on therapy services for aphasia has used surveys of SLPs to examine the therapy dosages available across the various healthcare settings making up the continuum of care for aphasia. Findings from these surveys generally suggest that most therapy services in English-speaking countries are focused on the acute and subacute stages, with much less available for people with chronic aphasia. Although dose has not been reported, available evidence suggests that session duration is relatively consistent across regions: the most common session duration reported in Australia, Canada, the US and UK was 30 minutes in acute inpatient 11  settings and one hour in chronic outpatient settings (Katz et al., 2000). Session frequency appears to peak in the acute or subacute stage, and is lowest in the chronic stage. An Australian study found that daily sessions (amounting to ~4 hours per week) were most frequently reported in inpatient rehabilitation settings, 2-3 sessions per week (~2 hours per week) in inpatient acute and outpatient rehabilitation settings, and ≤1 session per week (<2 hours per week) in community, university, long-term care, and private practice settings (Verna, Davidson, & Rose, 2009). Total therapy duration appears to generally be short in the acute stage and become longer from subacute to chronic stages, consistent with a transition from relatively more massed to relatively more distributed practice. Total therapy duration may often be determined more by service constraints than client needs. Katz et al. (2000) found the average total number of therapy sessions (a measure conflating session frequency and total therapy duration) to be 1-5 sessions in acute settings in Australia and the UK, and 16-20 sessions in Canada and the US. Especially in the US private sector, limits on the number of therapy sessions provided – imposed primarily by insurance companies – ranged from 1-20 sessions with an average limit of nine, which therapists reported usually resulted in insufficient therapy (Katz et al., 2000). Verna et al. (2009) found that therapy duration ranged from 1 day-12 weeks in inpatient acute settings, 2 weeks-6 months in inpatient rehabilitation settings, 4 weeks-18 months (or up to several years for clients receiving group therapy) in outpatient rehabilitation settings, 1 session-2 years in community and long-term care settings, and 1-6 years in university clinics. It is worth noting that all the statistics discussed above are likely not representative of direct therapy time, as they probably also include assessment, counseling, and other clinical activities (Code & Petheram, 2011). Likely owing to the practical challenges of conducting longitudinal research, the few available longitudinal studies on the continuum of care for people with aphasia have been 12  cursory in scope and limited to information available from their stay in the inpatient acute setting. Nonetheless, while cross-sectional research suggests that the therapy services available across the various healthcare settings are relatively restricted, longitudinal research suggests that some people with aphasia may not be given access to these services in the first place. For example, Dickey et al. (2010) found that 35% of people with aphasia admitted to acute hospitals in Ontario, Canada were never seen by a SLP in any capacity before they were discharged. Moreover, Lalor and Cranfield’s (2004) Australian study found that even when people with aphasia were referred for SLP services and were considered alert and medically stable enough to receive therapy, 75% of them still did not receive any therapy before discharge, due to caseload demands on SLPs. Due in part to such demands, it has been argued that aphasia therapy is not feasible in acute care hospitals and should instead be reserved for rehabilitation settings where people with aphasia can receive high-dosage therapy (Foster et al., 2015). However, Dickey et al. (2010) found that only 34% of people with aphasia were discharged to rehabilitation and that nearly half were instead discharged directly home (33%) or to long-term care (14%), where the therapy dosages available are negligible (Verna et al., 2009). Thus, the patchwork of available evidence about the continuum of care for people with aphasia raises questions about the equity of contemporary service provision and suggests that therapy dosages provided in current clinical practice, when available at all, are limited, especially in the chronic stage. Despite inconclusive findings from research regarding the optimal dosage of therapy for people with aphasia (see section 1.4.1), the state of contemporary therapy services has led some to conclude that there is probably a significant gap between client needs and service realities that is unlikely to ever be filled if all therapy is provided solely by a qualified SLP – a 13  valuable, yet expensive and limited resource (Code & Heron, 2003; Code & Petheram, 2011; Katz et al., 2000). 1.5 Quality of therapy: Learning through mobile technology The potential dosage gap in clinical practice, and the need to minimize it without taxing current clinical resources even further, has been the major push for alternative approaches to providing aphasia therapy, like computer-based therapy or mobile technology-based therapy (MTT). Less attention has been given to the quality of the learning experience when therapy is provided using technology. This is consistent with a general trend in the aphasia literature, where descriptions of the therapy process are commonly prescriptive and underspecified or ambiguous, resulting in substantial variability in how therapy is enacted in research and clinical practice (Byng, 1995; Hinckley & Douglas, 2013; Horton, 2006; Horton & Byng, 2000). Nevertheless, assumptions about what constitutes effective therapy are widely held and have guided debates about the adequacy of MTT. As such, this study defined therapy quality in terms of Baker (2012) and Warren et al.’s (2007) concept of the learning episode, specified in terms of its dose form (the context within which learning is intended to occur) and its therapeutic procedures (the actions that are intended to foster learning). The relative importance ascribed to the nature of these components of the therapy process has depended on one’s theoretical orientation (e.g., operant conditioning versus psycholinguistic approaches; Byng, 1995; Horton & Byng, 2000). 1.5.1 Dose form: Personal relevance and self-directed practice Discussions of dose form in aphasia therapy have typically concentrated on the tasks or activities within which therapy occurs. Since mobile technology is not a therapy task per se, but rather a medium through which tasks and procedures can be enacted, discussions of the dose form of self-directed MTT have focused on the two contextual factors distinguishing it most 14  from SLP-based therapy: the presence of (potentially) accessible multifunctional technology and the absence of a therapist during therapy. It has primarily been argued that both factors could augment the therapy process and the clinical significance of outcomes by tapping into principles of neuroplasticity and the life participation model of aphasia. The adaptability of mobile devices, with their range of built-in sensors, could give people with aphasia the opportunity to take their own pictures and videos to create personally relevant and contextualized therapy materials (Brandenburg et al., 2013). While traditional SLP-based therapy often uses generic materials, it has been argued that personalized materials could not only increase the practical importance of therapy effects, but also the attentional and motivational salience of therapy, which neuroscience research suggests is important for inducing neuroplasticity (Kleim & Jones, 2008; Horton, 2006; Nickels, 2002a; Raymer et al., 2008; Renvall, Nickels, & Davidson, 2013a; van de Sandt-Koenderman, 2011). However, personal relevance has been understudied in the aphasia literature; there is little clarity regarding how personal or functional relevance and salience should be operationalized; and the little research directly manipulating personal relevance has not evaluated the practical impact of effects, but rather has shown mixed results only on constrained impairment-level tasks (e.g., spoken word-picture matching; Cherney, Kaye, Lee, & van Vuuren, 2015; McKelvey, Hux, Dietz, & Beukelman, 2010; Renvall et al., 2013a). While limited therapist supervision could have a negative impact on outcomes by reducing clients’ motivation to adhere to a therapy regimen (Zheng, Lynch, & Taylor, 2016), much of the discussion about the effects of self-directed practice using mobile technology has focused on its potential benefits: not only the opportunity it could provide for increased quantity of practice (see section 1.4), but also its potential to improve the learning process and have 15  secondary effects on quality of life. By allowing people with aphasia to practice independently, accessible mobile technology could provide them the opportunity to direct their own therapy process. It has been proposed that giving clients this kind of control could improve outcomes by allowing clients to customize their learning strategies based on their past performance and by increasing attentional salience through clients’ active engagement in therapy (Doesborgh et al., 2004; Hickin, Best, Herbert, Howard, & Osborne, 2002; Leonard et al., 2015). Moreover, self-directed practice, facilitated through accessible computer or mobile technology, could also help promote clients’ autonomy (Katz, 2010; Kurland, Wilkins, & Stokes, 2014; McNaughton & Light, 2013; van de Sandt-Koenderman, 2011). This may be an important goal for aphasia therapy, as promoting the autonomy of people with aphasia could in turn impact on their quality of life, via the interaction between autonomy, communication confidence, and life participation (Babbitt & Cherney, 2010). Furthermore, some research has suggested that, in the traditional pedagogical approach commonly found in SLP-based aphasia therapy, therapists appear to have a tendency to take control over the therapy process, allowing clients to participate only passively (Horton, 2007; Simmons-Mackie & Damico, 1999). Just as promoting autonomy may improve confidence and life participation, it has been suggested that instead promoting a passive and dependent role in traditional SLP-based therapy could exacerbate the level of disability of people with aphasia (Babbitt & Cherney, 2010; Horton, 2007; Kimbarow, 2007; Simmons-Mackie & Damico, 1999). Although developing self-reflexive practices among therapists (cf. Hersh & Cruice, 2010; Horton, Byng, Bunning, & Pring, 2004) may be a way of addressing the possible negative effects of SLP-based therapy, self-directed practice – whether via accessible mobile technology or other means – may represent an alternative method of doing so. 16  Unfortunately there is little empirical evidence to support any of these possibilities. Therapy adherence has been grossly underreported in the aphasia literature in general (Hinckley & Douglas, 2013), so the importance of adherence to therapy outcomes, and the role therapists might play therein, has not been systematically examined nor experimentally manipulated. While research on motor learning in nonclinical populations does suggest that giving learners control over aspects of the dosage schedule, cueing, and feedback can augment learning retention (Keetch & Lee, 2007), the few counterpart studies from the aphasia literature have shown no advantage for client control on language outcomes (Doesborgh et al., 2004; Hickin et al., 2002; Leonard et al., 2015). Furthermore, evidence for therapists’ excessive control in SLP-based therapy has been based primarily on qualitative analyses of small samples of dyads (Horton, 2007), and the possible differential effect of therapist and client control on clients’ psychological state and life participation has not been experimentally compared, nor measured directly with objective performance-based or client-reported measures. Thus, although providing therapy in the context of self-directed practice using mobile technology has been proposed as a way to create a more customized, engaging, meaningful, and client-centred learning environment for people with aphasia, more research is needed to examine the effects of mobile technology and self-directed practice on the therapy process and the clinical significance of outcomes. 1.5.2 Therapeutic procedures: Cueing, feedback, and performance-contingency While discussions of dose form have largely emphasized the potential advantages of MTT, much of the discussion in the aphasia literature about therapeutic procedures has focused on what is presumed to be lacking when therapy is provided without the expertise of a therapist. Therapeutic procedures include therapeutic inputs and client acts (Baker, 2012). In the pedagogical tradition of aphasia therapy, therapeutic inputs consist of cueing and feedback – 17  therapists’ attempts to, broadly speaking, promote and modify clients’ responses, respectively – both of which are commonly considered to be routine and essential parts of aphasia therapy (Horton & Byng, 2000). Client acts have traditionally been conceptualized as responses therapists passively elicit from clients, with little acknowledgment given to the active role that clients may play in their own learning (Byng, 1995; Horton, 2008). Although the dynamic interactive aspects of aphasia therapy have similarly often been neglected, discussions of cueing and feedback typically assume that they cannot be enacted at random, but rather, in order to promote success, their pattern of presentation must be systematically modified in a manner that is contingent on clients’ past performance (i.e., performance-contingency; Byng, 1995; Horton, 2008; Katz, 2010; Zheng et al., 2016). The assumed importance of these therapeutic procedures, and observation of the limited capacity of current technologies to replicate them, has led to concerns about the feasibility of using computers or mobile technology for self-directed practice. In SLP-based therapy, there are a great variety of cueing strategies that a therapist could call upon to support a given client at a given moment (Horton, 2006). In contrast, all cues available to a client on a computer or mobile device must be determined ahead of time by the program developer (Katz, 2010). While efficient multimodal stimulus presentation may be a particular strength of mobile devices due to their multimedia capabilities, the cueing repertoire of MTT will likely always be comparatively limited, potentially resulting in therapy that is too simplistic or inflexible to promote success in all the clients who may use it (Brandenburg et al., 2013; Katz, 2010). More problematic is the lack of robust automatic speech recognition software, which means that current aphasia therapy apps cannot monitor clients’ performance when the task requires a verbal response (Brandenburg et al., 2013). The resulting inability of current therapy apps for verbal production to provide performance-contingent cueing and feedback has 18  been argued to make them inappropriate for self-directed practice, on the argument that these limitations could lead clients to practice and learn their own errors (Brandenburg et al., 2013; Kurland et al., 2014). This would severely limit the potential value of MTT as an alternative to SLP-based therapy, as most aphasia therapy research has been directed at improving verbal language production deficits (Cherney & Robey, 2008). However, the validity of such concerns has not been established. The relative importance attributed to cueing and feedback has depended greatly on the therapy approach, and contradictory advice about their utility has sometimes been offered even within the same theoretical orientation (Fillingham, Sage, Lambon Ralph, 2005a; Horton & Byng, 2000). Moreover, the motor learning literature suggests that, at least in nonclinical populations, limited external control over performance-contingency may actually be advantageous because it provides learners the opportunity to contingently organize their learning environment in a manner that maximizes their success (Keetch & Lee, 2007). Furthermore, some evidence from the aphasia literature suggests that therapists may not enact therapeutic procedures, including performance-contingency, consistently in actual clinical practice (Horton, 2006; 2008). Intricately predetermined therapeutic procedures are more often described in the research literature; however, the underreporting of therapy adherence means it is unknown how reliably therapeutic procedures are administered (Hinckley & Douglas, 2013; Horton, 2008). Nonetheless, even just in the naming therapy literature – the most developed line of aphasia therapy research (Cherney & Robey, 2008; Wisenburn & Mahoney, 2009) – definitions of what constitutes cueing and feedback have varied widely, and the empirical evidence has been inconclusive or indicated that all cueing, feedback, and performance-contingency procedures that have been studied are generally equivalently efficacious. 19  The literature evaluating the effects of cueing type – including the heterogeneous and somewhat contrived categories of semantic, phonological, orthographic, and mixed cueing – has been based on the premise that outcomes would be improved if cueing type were matched to a client’s impairment type. However, literature reviews and meta-analysis have not supported this hypothesis, instead finding equivalent positive effects across impairment types (Nickels, 2002a; Wisenburn & Mahoney, 2009). Recent interest in errorless and errorful learning has drawn attention to the diversity of cueing techniques used in naming therapy studies. These include a number of techniques based on performance-contingent psychophysical methods (e.g., increasing or decreasing/vanishing cueing), but also others involving the static, performance-independent presentation of high- or low-support cues, or even no cues (Cherney, Kaye, & van Vuuren, 2014; Fillingham, Hodgson, Sage, & Lambon Ralph, 2003; Treutwein, 1995; Middleton & Schwartz, 2012). Contradictory findings regarding the relative benefits of errorless and errorful learning have each been obtained across clinical and nonclinical research in other fields, and reviews of the two learning approaches when applied to aphasia have found no essential difference between them (Fillingham et al., 2003; Middleton & Schwartz, 2012). Very little research has examined the role of feedback in the aphasia therapy literature (Fillingham et al., 2005a). As suggested by the name, feedback has usually been performance-contingent, taking the form of explicitly or implicitly presented accuracy information, further cueing, or descriptions and explanations of the client’s performance. In some studies, however, feedback has also been omitted altogether, negating the need for performance-contingency (Breitenstein, Kamping, Jansen, Schomacher, & Knecht, 2004; Fillingham et al., 2005a, 2005b; Horton, 2006, 2008; McKissock & Ward, 2007; Nickels, 2002b). The few studies that have manipulated feedback experimentally (in terms of the presence versus absence of accuracy 20  information and/or a model of the correct target) have found conflicting results. Thus, on the basis of all of the evidence reviewed above, it is not clear to what degree therapy effects are dependent on the specific individual or combined therapeutic procedures used in sessions, including those that may or may not be present in self-directed MTT. More research in general is needed on the therapeutic procedures of aphasia therapy and their impact on outcomes, as well as research specifically evaluating concerns regarding restrictions on the therapy process of self-directed MTT that may be imposed by the technical limitations of mobile devices. 1.6 Self-directed computer- and mobile technology-based aphasia therapy While no clear conclusions can be drawn regarding the arguments for and against self-directed MTT on the basis of the evidence reviewed in preceding sections, preliminary studies on the effects of MTT, as well as the related computer-based therapy (CT) literature, have shown promising results or provide indirect support. Reviews of the evidence base for CT suggest that it is effective in general when compared to no therapy and that it may be as effective as SLP-based therapy (Cicerone et al., 2011; Brady et al., 2012; Wertz & Katz, 2004; Zheng et al., 2016). However, the majority of this literature has not examined CT in the context of self-directed practice, meaning that one of the primary advantages proposed for technology-based therapy (see section 1.5.1) has received little attention. Instead, in most studies, therapy was carried out in a clinical setting rather than at home, the therapist provided all or a portion of the therapy, and/or cueing and feedback were fixed, all of which limited how much control participants could have over the therapy process (Katz, 2010; Routhier, Bier, & Macoir, 2016; Ramsberger & Marie, 2007; Wertz & Katz, 2004; Zheng et al., 2016). In fact, it appears that only six outcome studies have examined self-directed CT in the home environment, with therapist supervision of the therapy process limited to troubleshooting and monitoring of progress (Fridriksson et al., 2009; 21  Mason et al., 2011; Mortley, Wade, & Enderby, 2004; Palmer et al., 2012; Pedersen, Vinter, & Olsen, 2001; Ramsberger & Marie, 2007). In comparison, a total of six small, highly heterogeneous outcome studies examining MTT have so far been published (Fridriksson et al., 2012; Hoover & Carney, 2014; Kiran, Des Roches, Balachandran, & Ascenso, 2014; Kurland et al., 2014; Lavoie, Routhier, Légaré, & Macoir, 2016; Routhier et al., 2016). However, unlike the CT literature, all but one of these MTT studies (Hoover & Carney, 2014) evaluated self-directed home practice with minimal therapist supervision.  Both sets of studies suggest that self-directed practice using computer or mobile technology can lead to improvements in the specific language behaviours trained; however, outcomes have varied across participants, and the evidence for maintenance of these effects post-therapy and for generalization to unpracticed behaviours or tasks is less consistent. Some studies have also noted positive effects of self-directed CT or MTT on participants’ technology skills, autonomy, functional communication, participation, and/or confidence. Although most studies only reported these outcomes anecdotally (Fridriksson et al., 2009; Kurland et al., 2014; Mason et al., 2011; Ramsberger & Marie, 2007), two CT studies also used semi-structured interviews to more precisely document participant and family perspectives of outcomes (Palmer, Enderby, & Paterson, 2013; Wade, Mortley, & Enderby, 2003). These two studies also reported possible negative effects of CT in a minority of participants: lost time for everyday life activities due to time-consuming practice; worries about withdrawal of therapy and loss of gains; and negative psychological responses to therapy, including frustration, fatigue, boredom, and anxiety (Palmer et al. 2013; Wade et al., 2003). Like the broader CT literature (Zheng et al., 2016), the majority of participants across all of the studies examining self-directed practice were younger (under 65 years old) and had mild or moderate non-fluent aphasia with no comorbid speech or 22  nonlinguistic cognitive deficits, so it is unclear how generalizable findings are to participants with other combinations of characteristics. Results of some of the studies suggested that participants with more severe aphasia, apraxia of speech, or semantic impairments tend to show less or no benefit from self-directed practice using technology (Fridriksson et al., 2009, 2012; Palmer et al., 2012; Routhier et al., 2016). Regarding feasibility, a number of studies also reported – either anecdotally (Fridriksson et al., 2009; Ramsberger & Marie, 2007; Kurland et al., 2014) or on the basis of a self-report questionnaire that had not been psychometrically tested (Routhier et al., 2016) – that participants found self-directed home practice satisfactory, acceptable, and/or that they valued the independence it afforded them. Palmer et al. (2013) and Wade et al.’s (2003) interviews additionally indicated that most participants would use CT again, and valued the opportunity for unlimited practice, flexible scheduling, and its affordability relative to SLP-based therapy. Some participants nonetheless preferred SLP-based therapy for the opportunity for social contact and conversation. Furthermore, there is some indication that self-directed CT can also be an efficient use of therapist time and a cost-effective means of therapy delivery (Latimer, Dixon, & Palmer, 2013; Mortley et al., 2004; Palmer et al., 2012). Positive therapy outcomes across the studies might be taken as indirect evidence for the accessibility of computer or mobile technology for people with aphasia (Brandenburg et al., 2013). However, Palmer et al. (2013) and Wade et al.’s (2003) interviews indicated that this might not always be a valid inference. They found that substantial caregiver support for technical issues and therapeutic procedures was at times required to achieve success, suggesting that their CT was not sufficiently accessible to facilitate self-directed practice for some participants. Unfortunately, like most of the broader CT literature, none of the studies discussed here directly examined how participants used technology, so little 23  information is available about what factors increase or decrease the accessibility of computer or mobile technology for people with aphasia (Brandenburg et al., 2013). Although dosage schedules were prescribed in all studies but one (Mortley et al., 2004), no controls were in place to enforce the therapy protocol, meaning participants were free to practice wherever, whenever, and however much they wanted during self-directed home practice. Results from a number of the studies suggested that participants are usually adherent to the dosage regimen, and sometimes will choose without prompting to practice for long durations at high frequencies when given the opportunity (Kiran et al., 2014; Kurland et al., 2014; Mason et al., 2011; Mortley et al., 2004; Palmer et al., 2012; Routhier et al., 2016). However, dosage adherence has not been perfect (Mason et al., 2011; Palmer et al., 2012), and there is some evidence to suggest that external supervision may play an important role in promoting therapy adherence. Palmer et al. (2012) completed the only randomized controlled trial among the studies and reported that three of the four unsupervised participants in their therapy group did not adhere to the prescribed dosage schedule, whereas only one of the 11 supervised participants did not. Moreover, participants indicated during interviews that therapist/volunteer support was important for providing both motivation and reassurance to continue practicing (Palmer et al., 2013; Wade et al., 2003). Interestingly, despite giving participants control over most dosage variables, few studies gave the same opportunity in terms of therapeutic procedures: almost all reported using predetermined tasks and fixed cueing and feedback, meaning participants had relatively little control over the learning process of the therapy they received. The majority of the studies from both the MTT and CT literatures targeted spoken naming, despite limitations in automatic speech recognition. To circumvent this issue, there was a bias towards using tasks requiring a 24  written/pointing response (with the expectations that effects would generalize to spoken naming) so that fixed performance-contingent cueing and accuracy feedback could be provided, and/or towards using verbal response tasks with fixed performance-independent cueing and no feedback. Only Mortley et al. (2004) reported attempting to address this issue by giving participants control over cueing and feedback in their picture-naming task. This task was the only one of its kind in their study, among many other tasks that were fixed, so it is not possible to isolate the effect of participant-controlled therapeutic procedures. In fact, it appears that only one study in the CT literature has specifically examined giving participants control over some aspect of therapeutic procedures: Doesborgh et al.’s (2004) CT was conducted in a clinical setting with considerable therapist supervision and used a written response task with fixed performance-contingent feedback; however, it intentionally provided an unfixed cueing repertoire in order to facilitate participants’ “discovery-based learning” (Doesborgh et al., 2004, p. 215). This concept appears to be largely equivalent to the performance-contingency account for the selective advantage of self-directed practice on learning retention in the nonclinical motor learning literature (Keetch & Lee, 2007). However, Doesborgh et al.’s (2004) results did not support this hypothesis, as they found no statistically significant difference between their CT and a no-therapy control group, suggesting that there may be limits to the value of self-directed practice in aphasia therapy. In summary, the evidence reviewed here suggests that self-directed home practice, whether provided via a computer or a mobile device, may improve language performance for at least some people with aphasia. It also provides some preliminary insight on other positive and negative effects of self-directed therapy using technology, and on the acceptability of this form of therapy for people with aphasia and their families. However, the literature is small, especially 25  for MTT, and more research is needed to confirm and further explore the inconsistent findings of the extant literature. In fact, on the basis of much of the same evidence reviewed here, the most recent reviews of mobile technology in aphasia and CT (Brandenburg et al., 2013; Zheng et al., 2016) came to the conclusion that the accessibility of technology for people with aphasia, and the roles of client characteristics, therapy quantity, and therapy quality in technology-based therapy are areas in particular need of further research. 1.7 Research questions Proponents for MTT suggest that incorporating mobile technology into aphasia therapy could increase the accessibility of therapy for people with aphasia and thereby facilitate self-directed home practice with minimal therapist supervision. Such practice is then expected to augment the therapy process in terms of therapy quantity (and quality), resulting in improved therapy outcomes relative to traditional SLP-based therapy. However, although research has provided encouraging results about the acceptability and therapeutic effects of CT, little is known about the accessibility of mobile technology for people with aphasia or its ability to facilitate self-directed home practice. Moreover, it is not known whether self-directed practice can actually increase therapy quantity or quality, and it is not even clear how quantity and quality are related to therapy outcomes in the first place. Thus, each of the causal steps supposedly linking MTT to improved outcomes is contentious, and decisive conclusions about its potential to deliver on them cannot be drawn based on the available empirical evidence. Given these circumstances, the purpose of this study was to explore MTT’s potential by evaluating the (a) feasibility, (b) therapeutic effect, and (c) therapy process of a minimally-supervised self-directed home practice program using a tablet-based aphasia therapy app called Naming 26  Therapy© (Tactus Therapy Solutions Ltd., Vancouver, Canada). Specifically, the research questions were: 1. What is the feasibility of enacting minimally-supervised self-directed MTT in terms of: a. The efficiency of developing word sets and their functional relevance to participants with aphasia; b. Participants’ adherence to the study protocol; c. The accessibility of the tablet computer and therapy app both for participants and for the researcher; d. The potential clinical utility of the therapy app for remote monitoring of therapy progress; and e. The acceptability of minimally-supervised self-directed MTT for participants? 2. What are the therapeutic effects of minimally-supervised self-directed MTT, both in terms of: a. The naming accuracy of practiced and unpracticed words; and b. Language, technology use, communication activity, participation, and confidence/well-being? 3. How is the therapy process of minimally-supervised self-directed MTT characterized in terms of: a. Dosage (the quantity, frequency, and duration of practice); and b.  Therapeutic procedures (the use of cueing strategies and performance-contingency)?  27  Chapter 2: Method This section details the methodology of the study’s single-subject experiment, including its multiple baseline design, participants recruited and their characteristics, materials, procedures, and analysis. As the methodology was complex, an overview is first provided to help orient the reader. 2.1 Overview of methods The study procedure included four phases: pre-therapy screening and clinical profile evaluation (Phase 1); word set development and baseline (Phase 2); therapy (Phase 3); and post-therapy outcome evaluation (Phase 4; Table 1). All sessions were conducted in participants’ homes, a designated public place, or in a research lab depending on their preferences. These sessions were audio recorded and detailed notes about their contents and the researcher’s impressions were made for later reference. All interactions with potential and recruited participants during the study were facilitated via partner-supported augmentative and alternative communication (AAC) strategies that were adjusted to the individual’s communicative abilities: e.g., supplementing short simple phrases with drawing, writing, gestures, and/or objects (e.g., demonstrating the tablet computer); attending to the individual’s nonverbal communication (e.g., facial expressions, gestures, postures, and vocalizations); and checking mutual understanding via yes/no questions, repetition, and paraphrasing (Dalemans, Wade, van den Heuvel, & de Witte, 2009; Luck & Rose, 2007). During Phase 1, multiple standardized and informal measures were used to screen potential participants for their study eligibility and to document their clinical profiles. These measures could be administered over multiple days if needed to avoid fatiguing or overwhelming participants. Phase 2 involved a multistage process to develop three study word sets (two therapy 28  sets and a control set) and collect initial baseline data for the multiple baseline design. During Phase 3, the therapy set initially assigned to the practice set was made available for participants to practice independently during self-directed therapy using Naming Therapy© (NT) on an iPad. As it was not possible to remotely monitor participants’ progress with sufficient detail or remotely manage participants’ word sets at all using NT, a multiple-probe technique was used (Barlow, Nock, & Hersen, 2009). A probe session was conducted at approximately the end of every therapy week, during which the researcher assessed participants’ naming accuracy without Table 1. Overview of the study procedure. Phase 1: Pre-therapy screening and clinical profile evaluation  • Screening/clinical profile/secondary outcomes measures: 1 or 2 sessions 1 day to 1 week Phase 2: Word set development and baseline  Nomination stage 1 week Evaluation stage/Baseline:  • 3 naming accuracy probe sessions: 1 trial per word per session  Selection stage 2 to 4 weeks Confirmation stage/Baseline:  • ≥1 naming accuracy probe session(s): 1 trial per word per session  Matching and random assignment stage 1 week Phase 3: Therapy     4 weeks to ~6 weeks Minimally-supervised self-directed naming therapy: • ≥2 trials per word per day, ≥4 days per week • Independent practice in Naming Practice mode of NT, no instructions regarding cue use, self-scoring at end of each trial Word set management: • 1 naming accuracy probe session per therapy week: 2 trials per word per session • Manipulating word set presentation, check-in interviews, troubleshooting Phase 4: Post-therapy outcome evaluation  Immediately post: 1 week • 3 naming accuracy probe sessions: 2 trials per word per session  One month post: 1 day to 1 week • 1 naming accuracy probe session: 2 trials per word per session • Secondary outcome measures: 1 or 2 sessions  Note. NT = Naming Therapy©. 29  cueing support for the practice set as well as the monitor set, which consisted of therapy set words not currently available for practice. Based on participants’ performance, the researcher manipulated which words would be assigned to the practice and monitor sets for the coming therapy week in order to enact the multiple baseline design. These probe sessions were also used to check in with participants regarding their experiences during self-directed practice and to troubleshoot any technical issues encountered. During Phase 4, participants’ uncued naming accuracy for both therapy sets and the unpracticed control set were assessed during multiple probe sessions immediately after therapy and one month later in order to evaluate the effect of therapy on naming accuracy outcomes. Some of the measures used during Phase 1 were also readministered during Phase 4 in order to evaluate secondary therapy outcomes. The data collected during these various phases were used to address the sub-components of each of the study’s research questions (Table 2). The feasibility question was evaluated with a diverse set of variables derived primarily from usage data automatically logged by NT, the researcher’s session notes, and qualitative data regarding participants’ perspectives (see Appendix A for further information). These data were collected during Phases 2 through 4. Questions regarding therapeutic effect were addressed primarily through the study’s multiple baseline design enacted during Phases 2 through 4, based on naming accuracy data collected  Table 2. Overview of the research questions and their sub-components. Feasibility Therapeutic effect Therapy process 1. Word set development 2. Adherence 3. Accessibility 4. Remote progress monitoring 5. Acceptability 1. Naming accuracy 2. Secondary outcomes 1. Therapy dosage 2. Therapeutic procedures (cueing and performance-contingency) 30  during probe sessions. Secondary effects of therapy on language, technology use, communication activity/participation, and self-confidence were also evaluated using quantitative and qualitative data collected during Phases 1 and 4 (Appendix B). The therapy process question was addressed with variables derived from NT’s data logging system during Phase 3 as well as qualitative data collected from participant-perspective interviews during Phase 4 (Appendix C). 2.2 Design A multiple baseline across-behaviours design was carried out during Phases 2 through 4 using dynamic therapy sets of practiced words and a control set of unpracticed words. The core features of this design could permit evaluation of causality within a single participant (Barlow et al., 2009). The longitudinal phase structure (i.e., baseline and therapy phases) helped to establish the first two criteria for causality, the association and the time order of independent and dependent variables. To establish the third criterion, nonspuriousness, repeated measurement of behaviour helped to control for confounding variables related to endogenous change (i.e., testing, maturation, and statistical regression effects) and instrumentation effects, and the staggered introduction of therapy to multiple different but related behaviours (i.e., different words) helped to control for history effects. Pre- and post-testing of the control set provided added evidence for establishing causality, similar to a no-therapy control group in group experimental designs (Barlow et al., 2009). The dynamic word set approach was a variant of the criterion-referenced multiple baseline across-lists approach common to aphasia naming therapy research, in which the introduction and withdrawal of therapy was based on a predetermined criterion (i.e., acquisition). However, the difference was that this criterion was applied to individual words, rather than aggregate word lists, resulting in incremental changes in the composition of the practice set (cf. Conroy & Scowcroft, 2012; Fridriksson, Holland, Beeson, & Morrow, 2005; 31  Fridriksson et al., 2009; Weill-Chounlamountry, Capelle, Tessier, & Pradat-Diehl, 2013). This staggering of the introduction and withdrawal of therapy was accomplished during therapy probe sessions in Phase 3 by manipulating which words were presented to participants for self-directed practice. 2.3 Participants 2.3.1 Screening and clinical profile measures This section details the measures used during Phase 1 to screen individuals interested in the study for their eligibility to participate and to document the clinical profiles of participants successfully enrolled. The Western Aphasia Battery–Revised (WAB-R; Kertesz, 2007) was used to establish a diagnosis of aphasia and to provide a comprehensive evaluation of language function. The WAB-R consists of tasks to measure speech information content, fluency, auditory comprehension, repetition, naming, reading, and writing. It also includes optional tasks to assess nonlinguistic cognitive functions; however, these were not administered in this study in favour of the comparatively more comprehensive Cognitive Linguistic Quick Test (see below). The WAB-R’s measure of aphasia severity, the aphasia quotient, is a composite of the oral production and comprehension tasks, whereas the language quotient also includes the reading and writing sub-tests. The test also includes an algorithm for classifying people with aphasia into one of eight aphasia types based on a neoclassical taxonomy. The Boston Naming Test–Second Edition (BNT-2; Goodglass, Kaplan, & Baressi, 2001; Kaplan, Goodglass, & Weintraub, 2001) was used to screen for anomia and to characterize its severity for participants’ clinical profile. It is a spoken picture-naming task in which the participant is required to name a series of 60 line-drawn pictures intended to represent an increasing hierarchy of difficulty. In this study, a rounded raw score equivalent to a z score two standard deviation units below the mean of participants’ age 32  group was used as a rough cutoff for identifying the presence of anomia. Although the WAB-R and the BNT-2 have been criticized in recent years for having poor psychometric properties (Harry & Crowe, 2014; Hula, Donovan, Kendall, & Gonzalez-Rothi, 2010), they were included in this study because their historical widespread use allowed the results of this study to be cautiously interpreted in relation to the broader aphasia naming therapy literature. The Cookie Theft Picture (Goodglass et al., 2001) was used as a discourse measure of syntactic complexity and informativeness for participants’ clinical profiles. It is a standardized picture-description task designed to evaluate connected speech. It involves asking participants to describe everything they see going on in a picture of a kitchen, in which a woman is drying dishes at a sink overflowing with water, while a boy standing on a toppling stool and a girl next to him attempt to steal from a cookie jar in the cupboard. Scoring systems have been developed to evaluate the discourse samples produced by this task in terms of syntactic complexity (Goodglass et a., 2001), the informativeness and efficiency of connected speech (i.e., correct information units; Nicholas & Brookshire, 1993), and discourse structure (i.e., main concepts; Nicholas & Brookshire, 1995). The informativeness variables for the Cookie Theft Picture have demonstrated relatively good inter-rater and test-retest reliability and concurrent criterion validity when they are computed on the basis of multiple discourse samples. As only one discourse sample was collected per time point in this study, results from this measure were interpreted cautiously. Semi-structured participant-perspective interviews were conducted to document participants’ perceptions of technology and their expectations of the process and outcomes of minimally-supervised self-directed therapy. An interview guide (Appendix D) was adapted from Wade et al. (2003) and Palmer et al. (2013) for this purpose. The interview was also used as an 33  opportunity to collect self-reported data regarding demographic information, participants’ previous experience with technology, and eligibility criteria. In addition to general partner-supported AAC strategies (see section 2.1), a funneling question approach, adjusted to participants’ communicative abilities, was used during the interview to accommodate production difficulties (Luck & Rose, 2007): participants were first given the opportunity to respond to open-ended questions on a broad topic, then, as needed, to probe questions on specific areas of interest, and finally to specific ideas tentatively offered by the researcher in the form of yes-no questions (Appendix D). While the WAB-R, BNT-2, Cookie Theft Picture, and interview were additionally conducted during Phase 4 for use as secondary outcome measures, a number of other measures were used solely for screening and clinical profiles during Phase 1 to evaluate participants’ sensory, motor, and cognitive abilities, as well as their current and potential iPad proficiency. The Cognitive Linguistic Quick Test (CLQT; Helm-Estabrooks, 2001) was used as a measure of overall cognitive function and contributed to screening of the adequacy of participants’ cognitive and visual perceptual abilities for tablet use. A cognitive test was included because cognitive abilities appear to contribute significantly to the ability of people with aphasia to benefit from therapy in general (Lambon Ralph et al., 2010) and to use technology independently (Nicholas, Sinotte, & Helm-Estabrooks, 2005). The CLQT consists of 10 tasks designed to measure five cognitive domains: attention, memory, language, executive functions, and visuospatial skills. Five of these tasks were intentionally created to involve minimal language demands to permit a more valid evaluation of nonlinguistic cognitive deficits in aphasia. The test includes cutoff scores for each task, as well as severity ratings for the five cognitive domains and an overall severity rating. According to the manual (Helm-Estabrooks, 2001), the CLQT has demonstrated 34  acceptable test-retest reliability for the cognitive domains, good inter-rater reliability, and good content and construct validity. Participants also underwent a hearing acuity screening to contribute to evaluation of the adequacy of their sensory abilities for tablet use. They were required to pass at 1 and 2 kHz at 30 dB HL aided in at least one ear. An informal iPad proficiency screening/training was also conducted to evaluate participants’ current knowledge of, and learning potential for, operational functions relevant to the iPad navigation path they were instructed to follow during therapy, including: (a) basic functions of the tablet (i.e., turning it on and off, putting it to sleep and waking it up, locating and controlling the volume and home buttons, charging it, placing it in a stand), (b) interacting with the touchscreen interface (i.e., tapping, swiping, and using a stylus if needed), (c) operating Naming Therapy© (NT; i.e., starting the app, navigating the menu, selecting cues, and scoring performance), and (d) troubleshooting common problems (i.e., navigational errors, iPad being out of power, deleting the app, setting the volume too low, and the iPad going to sleep; Kurland et al., 2014; Ramsberger & Messamer, 2014; Szabo & Dittelman, 2014). Screening involved asking participants to demonstrate their current knowledge of the steps in the iPad navigation path, followed by accuracy feedback, modeling, and repetition of the correct action when errors occurred. Five high-frequency words from those native to NT were used during screening to simulate the navigational steps of a therapy session while reducing demands on participants’ naming abilities. The screening was then conducted one or more additional times within the same session (as long as participants showed continued improvement) until participants were able to perform all steps correctly without support. Because the screening also involved iPad operational instruction and multiple opportunities to learn, iPad screening also doubled as iPad training, and the provision of further training after the initial screening session, during the Phases 35  2 and 3, depended on participants’ performance and level of comfort operating the iPad and NT. Whether participants received additional training or not, the researcher supervised their first therapy session to evaluate their iPad proficiency with the full task demands of therapy, and to provide further training as needed. 2.3.2 Recruitment All participants were recruited from the same local aphasia support group. Inclusion criteria for the study included: a self-reported history of a single left hemisphere stroke; a diagnosis of aphasia as assessed by the WAB-R; prominent anomia on the BNT-2; at least one year post onset of signs/symptoms (self-reported); adequate sensory, motor, language, and cognitive abilities for tablet use, as assessed by the combination of the CLQT, the hearing acuity screening, the iPad proficiency screening, and self-report; self-reported English as the primary language; and self-reported residence within the Lower Mainland of British Columbia, Canada. The exclusion criteria included a severe motor speech impairment significantly impairing intelligibility as evaluated through informal clinical observation, and self-reported participation in individual SLP-based therapy totaling either more than two hours per week or more than three sessions per week. The first three participants that met inclusion/exclusion criteria and self-selected to participate were included in this study. No interested individuals were excluded and there was no participant attrition during the study. 2.3.3 Demographics and clinical profiles Participants’ demographic information and clinical profiles are presented in Tables 3 and 4. The following section highlights key information for each participant in turn. At the beginning of the study, P1 was a 55 year-old male with moderate-severe Broca’s aphasia. He had reduced nonfluent speech output, poor repetition abilities, severe anomia, and relatively good auditory 36  comprehension. His speech output was limited primarily to single content words and frequent, often propositional stereotypic phrases (e.g., “because it’s…”, “this one here”). Although he showed relatively good auditory comprehension on the WAB-R, the researcher had to use AAC strategies to facilitate P1’s comprehension in interaction more frequently than he did with any other participant. P1’s written production was more severely impaired than his spoken production, but he did show good reading comprehension for single words and some longer passages. On the CLQT, P1 overall showed moderate cognitive deficits. However, his scores may have been negatively biased by his prominent language deficits. He showed severe impairments in language and memory, due to an expected floor effect for the language domain and the high language demands for most of the memory-related tasks. To compare, for the only memory task with minimal language demands (visual design recognition), P1 showed perfect performance, which, though not definitive, was more consistent with clinical impressions of P1’s memory abilities in everyday life. P1 also showed a moderate attention deficit and mild deficits in executive functions and visuospatial skills. The errors that contributed primarily to these domains were on multiple nonlinguistic tasks suggestive of deficits in task-switching, self-monitoring, and/or inattention. During the iPad screening, P1 demonstrated basic knowledge and skill using the iPad without any instruction. His initial problems related primarily to lack of knowledge of how to navigate Naming Therapy© (NT); however, he quickly caught on after instruction, felt comfortable enough by the end of the screening session to opt out of receiving further iPad training sessions, and maintained his high performance during the initial therapy session.  37  Table 3. Participants’ demographics and clinical profiles.  P1 P2 P3 Demographics    Age at baseline (y) 55 59 68 Sex Male Male Male Handedness Right Right Right Languages English English English Education (y) 14 14 19 Occupation Importer/exporter Financier Psychologist Clinical profiles    Etiology Single stroke Single stroke Single stroke Lesion lateralization Left hemisphere Left hemisphere Left hemisphere Time post-onset at baseline (y) 7.08 5.08 7.50 Hearing screening Passed Passed Passed in left ear Vision Uncorrected Uncorrected but needed prescription Corrected, left visual field cut Hemiparesis Right-sided Right-sided Right-sided Motor speech impairment No No Yes Technology experience:    Mobile technology/ Computer Yes Yes Yes Naming Therapy© Yes No Yes iPad screening:    Initial accuracy for operational functions 86% 73% 76% Training trials 2 2 1 Training sessions 1 1 1 Concurrent SLP-based therapy (hr/wk) 1.0 during baseline and therapy phases, 0.0 during post-therapy phase 1.5 throughout study 1.0 throughout study    38  Table 4. Participants’ results on standardized tests contributing to their clinical profiles.  Max P1 P2 P3 BNT-2 60 8 40 25 WAB-R     Spontaneous speech 20 11 15 11 Auditory comprehension 10 7.95 7.95 8.4 Repetition 10 2.7 7.9 9 Naming & word finding 10 3.6 8.0 6.3 Reading 20 8.4 16.4 11.2 Writing 20 5.0 16.0 9.8 Aphasia Quotient 100 50.5† 77.7* 69.4† Language Quotient 100 46.6 79.2 64.1 Aphasia type  Broca’s Anomic Transcortical motor Cookie Theft     Subclausal utterances (%)  18 8 27 Agrammatic deletions (%)  0 21 0 Complexity index  0.0 1.1 0.0 Words/min  77.0 53.1 6.0 CIUs (%)  9 37 3 CIUs/min  7.0 19.8 0.2 AC concepts  0 5 0 AI + IN concepts  1 1 0 AB concepts  6 0 7 CLQT     Attention 215 93† 84† 192o Memory 185 81# 149* 90# Executive functions 40 21* 25o 25o Language 37 11# 26* 16# Visuospatial skills 105 68* 70* 94o Composite severity 4.0 2.0† 3.0* 2.8* Clock drawing 13 10* 10* 10* Note. BNT-2 = Boston Naming Test–Second Edition; WAB-R = Western Aphasia Battery–Revised; complexity index = mean number of clauses per utterance; CIUs = correct information units (i.e., words that are accurate, relevant, and informative relative to the Cookie Theft Picture); main concepts = presence, completeness, and accuracy of speech content without considering its form, AC = accurate/complete, AI = accurate/incomplete, IN = inaccurate, AB = absent; CLQT = Cognitive Linguistic Quick Test; o = within normal limits, * = mild, † = moderate, # = severe.   39  At the beginning of the study, P2 was a 59 year-old male with mild anomic aphasia. His speech output consisted primarily of single-clause and occasional multiclause sentences during assessment; however, outside of the testing situation he would revert to using more single words and short sentences. His spontaneous speech output was diminished and halting due to frequent word-finding difficulty, but also by the added demands of producing grammatically complete sentences. He showed relatively good auditory comprehension and repetition abilities. P2 performed well on the BNT-2 compared to other participants, but still scored 3.85 SD below the mean for his age range. His production and comprehension in the written modality mirrored his abilities in the spoken modality. On the CLQT, P2 overall showed mild deficits. Despite only mild deficits in language, P2 showed moderate attention and mild visuospatial-skill deficits resulting from significant errors on multiple nonlinguistic tasks. His errors were consistent with clinical impressions of P2’s attention in everyday interactions. During the iPad screening, P2 showed basic knowledge and skill using the iPad without any instruction. His initial problems related to navigating NT, but also troubleshooting common problems like waking the iPad up or activating the option to delete the app. However, P2 quickly caught on after instruction, felt comfortable enough by the end of the screening session to opt out of receiving further iPad training sessions, and maintained his high performance during the initial therapy session. At the beginning of the study, P3 was a 68 year-old male with moderate transcortical motor aphasia. P3 had very little substantive speech, but good auditory comprehension and excellent repetition. His spontaneous speech was slow and limited to single content words, infrequent simple sentences, and frequent stereotypic phrases that were often not clearly propositional (e.g., “round thing, a round thing, another thing to do”, “a little bit different”). P3 was the only participant with clear motor speech impairment. This, along with his already limited 40  output, resulted in heavy use of AAC strategies for facilitating his production and the researcher’s comprehension. Interestingly, in contrast to his spoken output, P3’s spontaneous written output consisted of fluently-written neologistic and paragrammatic sentences that were largely incomprehensible, of which P3 showed little awareness. He also demonstrated limited reading comprehension of longer sentences, but his oral reading mirrored his excellent repetition abilities, and he showed good reading comprehension for single words and short sentences. On the CLQT, P3 overall showed mild deficits. However, like P1, P3’s composite severity and memory scores may have been negatively biased by his prominent language deficits. Aside from errors that could be explained by language deficits, P3 did not demonstrate any clear signs of nonlinguistic cognitive deficits. During the iPad screening, P3 showed basic knowledge and skill using the iPad without any instruction. His initial problems related to navigating NT, but also troubleshooting common problems like waking the iPad up or increasing the volume when it was too low. However, P3 quickly caught on after instruction, felt comfortable enough by the end of the screening session to opt out of receiving further iPad training sessions, and maintained his high performance during the initial therapy session. 2.3.4 Experience and perceptions of technology and therapy P1 used to use computers extensively for work before his stroke and has become an adept mobile technology user since. He primarily used his iPhone but also owned an iPad. His iPad was loaded with a variety of therapy apps, which he reported using at least weekly, but he had not been using NT recently before the study. P1 also had a dedicated AAC app on his iPhone and iPad, which contained primarily stored messages regarding basic personal information. He reported using the AAC app only for the occasional time when he needed to convey this information accurately and efficiently to people who did not know him well. Despite his 41  language deficits, P1 was an effective communicator due to his use of AAC strategies. These included heavily supplementing his limited language with prosody, gestures, and facial expressions, but also by using apps on his mobile devices (especially the Photos, Calendar, and Maps apps) as communication supports during conversation (Szabo & Dittelman, 2014). He also reported extensively using both his devices as external memory aids, for entertainment, for accessing information, and for maintaining social connections with family and friends through texting and, to a lesser extent, the telephone. P1 accessed these functions via extensive use of strategies like typing single words, copying and pasting text, liberal and well-placed use of emoticons, and text-to-speech software. Given his extensive use of it, P1 was unsurprisingly very appreciative of mobile technology; however, he also expressed frustration about having to be dependent on it in the first place (Appendix E). When comparing and contrasting computers, iPhones, and iPads, P1 noted advantages for mobile technology in terms of accessibility and affordability, but also noted disadvantages related to screen size, technical problems, and the risk of losing devices. However, he emphasized that the differences did not matter to him, and that his iPhone and iPad were both highly valuable to him and essentially interchangeable. P1 was overall uncertain what to expect from self-directed mobile technology-based therapy (MTT). He did emphasize flexibility and guessed that he would probably practice as much as possible (up to seven days per week) by fitting therapy into his daily routine, so that he could balance it with his other time commitments. P1 reported no concern practicing by himself since he already did this with other therapy apps. P1 was uncertain how self-directed MTT would compare to SLP-based therapy, but he reported liking the idea of practicing whenever he wanted and challenging himself through repeated practice. When asked about the potential effects of therapy, P1 was not sure what specific effects NT would have, and was uncertain whether it might prove too easy for 42  him, but was hopeful that therapy would improve his language in general, citing his past improvement as evidence. P2 used computers and mobile devices extensively for business before his stroke, but reported currently using his laptop computer only minimally (less than once per month). Although his iPad was loaded with many apps, P2 reported having never used or having only briefly used most of them, including NT. He reported using his iPad once per day primarily for entertainment and to a lesser extent for therapy, primarily to practice writing. P2 reported using his Android (Google Inc., Mountain View, CA) smartphone at least once per day, but only for making phone calls, as an external memory aid, and for entertainment and accessing information. P2 could provide relatively little explanation for his patterns of technology use or his perception of mobile devices (Appendix E). However, he did attribute his use of his smartphone and iPad only for their basic functions to being an effect of his age. He further acknowledged that he preferred to use the iPad, rather than his smartphone, for reading because its bigger screen made the process easier. P2 was overall uncertain what to expect from self-directed MTT. Like P1, he guessed that he would probably practice as much as possible. P2 was confident he could practice by himself, but could provide no reasons for this. P2 intended to use the iPad to augment the SLP-based therapy he was already receiving, and assumed it would be good to be able to practice according to his own schedule. P2 was hopeful that self-directed MTT would help him learn to, and motivate him to start to, use his iPad more, but was very uncertain whether therapy would have specific effects on naming and his broader language abilities. P3 used computers extensively before his stroke, and still did at the time of the study. He had four computers at home and an iPad, but he no longer owned a cell phone. P3 had extensive computer knowledge and skills, even devoting part of his pre-stroke research endeavors to 43  information technology and telecommunications. He reported using the computer everyday, primarily for emails and entertainment/accessing information relevant to his scientific and technological interests. P3 even maintained a blog, where he would post, multiple times per day, media articles related to these interests. He reported using the iPad every day, primarily for entertainment/accessing information and email, but also once a week for therapy, primarily to practice reading and writing. His iPad was also loaded with NT, which he reported using in Flashcards mode, but only rarely. Open-ended questions resulted in P3 generating very little information regarding his reasons for his patterns of technology use or his perceptions of mobile technology; however, probe questions and researcher-generated ideas indicated that he had a very positive opinion of iPads/mobile devices, particularly valuing them for their technological innovations. He also noted that the touchscreen interface and the portability of iPads made them easier to use than computers (Appendix E). P3 identified nothing bad about mobile devices, except for acknowledging that the small screen size of smartphones could be inconvenient, but he did not consider this an important drawback. Like the other participants, P3 was overall uncertain what to expect from self-directed MTT. However, he reported no concern about practicing by himself since he felt he knew how to use iPads already. P3 reported that he would actually prefer self-directed MTT to SLP-based therapy for the opportunity for increased practice, and did not consider face-to-face interaction an important aspect of therapy. When asked about the potential effects of therapy, P3 expected that he would become better at naming the pictures, and was hopeful that practicing words he nominated for use in therapy would help in conversation with his spouse. 44  2.3.5 Summary of participant characteristics All participants were males in their mid 50s to late 60s at the beginning of the study. They were all monolingual English speakers, had many years of education, and previously had relatively high socioeconomic-status occupations. All participants received concurrent SLP-based therapy during the study. Clinical observation and the WAB-R confirmed the presence of aphasia, which all participants reported had resulted from a single left-hemisphere stroke ranging from 5 to 7.5 years ago at the beginning of the study, placing them all within the chronic stage of recovery. All participants had non-fluent aphasia with relatively good comprehension. P1 and P3’s aphasia was moderately severe, whereas P2’s was mild. P1 and P2 showed clear signs of nonlinguistic cognitive deficits, whereas P3 did not. P3 was the only participant with motor speech impairment. All participants reported extensive past or current experience with computers or mobile technology, and at least two of them (P1 and P3) had very positive impressions of mobile technology. All participants were overall uncertain what to expect from self-directed MTT. Nonetheless, they were all confident they would be able to carry it out successfully. They noted the value of flexible scheduling and of the opportunity for frequent repetition. Participants were overall uncertain about what the outcomes of therapy would be, but were hopeful for positive effects on language generally, naming, or technology use. None of the participants could identify negative effects of therapy. 2.4 Materials This section describes the materials for self-directed naming therapy and measures for outcome evaluation that were used during Phases 2 through 4. 45  2.4.1 Tablet Three iPad Air tablet computers (model A1474; Apple, Inc., Cupertino) were acquired for use during this study. To increase visual and navigational simplicity and interface stability (Brandenburg et al., 2013), all apps native to the iPad were placed in the same folder on a second page of the iPad Home Screen, so that they were not immediately accessible (Kurland et al., 2014); Naming Therapy© was downloaded onto each iPad and placed in the app dock so that it was visible from any page; and the iPad’s various popup screens and bars (e.g., notifications) were all disabled. A carrying case and stylus were also acquired for each iPad, for added protection and dexterity issues, respectively. 2.4.2 Naming Therapy© Naming Therapy© (NT; Tactus Therapy Solutions Ltd., Vancouver, Canada) is an aphasia therapy app aimed at improving spoken naming. On the face of it, NT appears to have been designed with many of the accessibility features for people with aphasia identified by Brandenburg et al. (2013), including aphasia-friendly text, large buttons (at least when presented on a tablet), a stable interface, simple navigation (i.e., a two- or three-tier menu hierarchy), visual simplicity, and to some extent multimodality presentation (e.g., use of symbols for navigation buttons). The app also has a modern appealing user interface (van de Sandt-Koenderman, 2011) designed to be appropriate for adult users. NT is a spoken picture-naming task split up into four modes. One of these modes, Naming Practice, was used to facilitate both picture-naming tasks used in this study: probe picture naming, which was used during probe sessions and is discussed in section 2.4.4.2 with the other outcome measures; and therapy picture naming, which was used by participants during self-directed practice in Phase 3. 46  During therapy picture naming, participants were instructed to navigate from the iPad Home Screen through to NT’s Trial Screen in Naming Practice mode without changing any settings or using other modes (Appendix F). On the Trial Screen, a photograph for one of the target words in participants’ practice set first faded in over 1.2 seconds in the centre of the screen. Along the bottom of the Trial Screen were buttons for six optional cues (Appendix G). Like other aphasia therapy apps for verbal production, NT had no speech recognition software, so it could not provide performance-contingent cueing or feedback (see sections 1.5.2 and 1.6 for review). Instead, NT left control over cueing up to the user and simply omitted accuracy feedback. Thus, although Naming Practice was, according to the instructions, intended to be used with a performance-contingent increasing cueing hierarchy, the app had the capacity to deliver, from trial to trial, any combination of cueing types and techniques used in the literature (see section 1.5.2) or any other idiosyncratic strategies the user might devise. As this study aimed to explore these self-devised strategies as part of the therapy process research question, participants were instructed to use cues however and whenever they felt it would be helpful and to make as many or as few naming attempts as they preferred, so long as they said their responses out loud during each trial. The Trial Screen also included an optional accuracy self-scoring system in the form of Correct and Incorrect Buttons in the upper right corner. Participants were instructed to score the accuracy of their last naming attempt using these buttons when they were done practicing a word, in order to evaluate the self-scoring system’s potential for remote progress monitoring (i.e., part of the feasibility question of this study). Participants were instructed to repeat this process on the Trial Screen for each of the remaining words in their practice set, presented in random order each time a trial series was started. At the end of the trial series, a Popup Screen appeared, on which the number of Correct and Incorrect Button selections were 47  summed and from which the user could then optionally email these results to a pre-established address. From the Popup Screen, participants were instructed to either start another trial series or return to NT’s Home Screen. 2.4.3 Word set stimuli This section details the nomination methods and sources of photograph and cue stimuli used in Phase 2 to develop participants’ therapy and control sets. The process of word set development is discussed in section 2.5.2, and the outcomes of this process are discussed in section 3.1.1 since they relate to the feasibility question of this study. Nomination methods in this study included intuition-based nomination, frequency-based nomination, and participant nomination. See Appendix H for the lexical characteristics (i.e., word frequency and word length) of the words nominated according to each of these methods (see section 2.5.2 for clarification about the lexical variables). 2.4.3.1 Intuition-based nomination Intuition-based nomination is the traditional method of nominating words for aphasia therapy and is characteristic of many commercially available materials for therapists (Renvall et al., 2013a). In this method, therapists nominate words according to semantic categories (e.g., food, clothing, sports, furniture, etc.) that intuitively seem the most common for people in general. This tends to result in an overrepresentation of concrete nouns among the nominated words. The words native to NT appear to represent this method, being restricted primarily to concrete nouns from categories like animals, body parts, food, objects, furniture, and sports. It was not necessary to collect stimuli for these words, because NT included colour photographs and cues for each of them by default. However, it was necessary to ensure the NT photographs matched the needs of the study. As such, photographs were included if: they accurately and 48  unambiguously represented the target word; target word had word frequency data (required for matching word sets during Phase 2) in the SUBTLEX-US database (Brysbaert & New, 2009); and they had not been used during the iPad proficiency screening (Appendix I). Screening for the representational adequacy of photographs was based on the results of informal pilot testing with 3 healthy adults. To reduce risk of fatiguing or overwhelming participants and to economize the number of words presented, the NT photographs were split into two sets (NT set 1 and NT set 2) intended to represent varying levels of difficulty: NT set 1 consisted of the first 150 NT words with the highest frequency of occurrence as nouns in the SUBTLEX-US database (Brysbaert, New, & Keuleers, 2012), and NT set 2 consisted of the remaining eligible NT words. In addition to the words native to NT, NT included functionality allowing the user to add custom words with photographs and cues, which permitted frequency-based and participant nomination, discussed below, to occur in this study. To align with intuition-based nomination, words identified through these methods had to be concrete nouns. 2.4.3.2 Frequency-based nomination In frequency-based nomination, words are nominated according to objective data on their frequency of use in nonclinical populations (Renvall et al., 2013a). Typically, high-frequency words are nominated in an attempt to increase their functional relevance to participants. In contrast, in this study a difficult low-frequency word set was intentionally created in response to an insufficient number of words being identified for P2 on the basis of NT sets and participant-nominated words alone. Words and photographs were collected from the Bank of Standardized Stimuli (BOSS; Brodeur, Dionne-Dostie, Montreuil, & Lepage, 2010; Brodeur, Guérard, & Bouras, 2014). The BOSS is a free database of 1,468 standardized photographs of objects providing norms for more than 15 dimensions based on 141 healthy English-speaking adults. 49  Photographs with high proportions of the normative sample who reported knowing the name but being unable to retrieve it (i.e., a tip-of-the-tongue moment) and/or knowing the object but not knowing the name were selected with the expectation these objects were low-frequency and had a high propensity to produce naming problems even in nonclinical populations. Target words for photographs, based on the name most commonly reported in the normative data, had to have word frequency data in the SUBTLEX-US database (Brysbaert & New, 2009). Like the NT sets, photographs were excluded if they inaccurately or ambiguously represented the target word on the basis of pilot testing with two healthy adults (Appendix I). The resulting BOSS set was added to the iPads using NT’s Custom Categories Screen. Whereas words and photographs were collected from the BOSS, cues had to be developed by the researcher. 2.4.3.3 Participant nomination Participant nomination refers to strategies for collecting words offered up by people with aphasia themselves (Renvall et al., 2013a). Participants were encouraged to nominate words they could not consistently say but considered important for their communicative success. These words could include proper names. Multiple strategies were used to collect participant-nominated (PN) words from participants, including spontaneous nomination, informal Internet checklists, modified versions of Bryen’s (2008) checklists, collaboration with informants (family and friends), and any other resources on hand, such as personal diaries (Renvall et al., 2013a). Spontaneous nomination and checklists were used during nomination interviews with the researcher, whereas collaboration with family and friends and any other resources were used on the participants’ own time. Spontaneous nomination referred to words that participants directly nominated without external aids during nomination interviews regarding their hobbies, interests, and common activities. The Internet checklists were word lists, related to topics of interest 50  identified by participants during nomination interviews, that were found via entering topic words into an Internet search engine. Participants nominated words from these lists with the support of the researcher for navigating the Internet and for reading words out loud when requested. The Bryen checklists were constructed from Bryen’s (2008) word lists for six socially valued adult roles: college life; sexuality, sex, and intimate relationships; crime reporting (by victims or witnesses of crime); managing personal assistant services; managing health care; and managing transportation. The almost 3,000 words collected in Bryen’s study were nominated by heterogeneous email focus groups made up of AAC users, university students and professionals with relevant experience, and family. In order to be included in the current study, words had to be non-duplicate nouns, resulting in 940 words remaining from Bryen’s original lists. To reduce risk of fatiguing or overwhelming participants, 50 words from each checklist were randomly selected to form abbreviated checklists. The remaining words were included in supplemental checklists available for participants if they were interested. Photographs for PN word sets had to be collected from a variety of sources, including participants’ or the researcher’s personal collections, BOSS photographs, adapted NT photographs, and the Internet. All cues had to be developed by the researcher, and semantic information for PN words particular to the participant’s life had to be collected from participants during the nomination interviews for later development of individualized cues. A participant’s PN words were added to their iPad using NT’s Custom Categories Screen. 2.4.4 Outcome measures This section details the primary and secondary outcome measures used to address the study’s research questions regarding the feasibility, therapeutic effect, and therapy process of self-directed mobile technology-based therapy with minimal supervision (see section 1.7). 51  2.4.4.1 Naming Therapy© data logging system A special research version of NT with a data logging system was used in this study for enhanced data collection. As the only measure of participants’ behaviour during self-directed practice, the data logging system collected data that contributed to every aspect of the research questions. However, its largest independent contributions were to the feasibility research question, in particular adherence and accessibility (Appendix A). To this end, the logging system was programmed to automatically and unobtrusively record the NT Screens the participant was viewing and the button selections they made on those screens, including related temporal variables, starting from when the NT icon was selected on the iPad Home Screen through to when the iPad Home Button was selected (Appendix J). The set of button selections and screens logged were those relevant to the navigation path participants were trained to follow during therapy picture naming (Appendix F). As the logging system only collected raw data, it was necessary for the researcher to complete navigation scoring – a measure of the appropriateness of each button selection in relation to the current step within the prescribed navigation path, which was used to in part evaluate both accessibility and adherence. The data logging system also collected raw data for probe picture naming (see section 2.4.4.2), thereby also indirectly contributing to all research questions addressed by this measure. The logging system used the iPad’s built-in microphones to audio record each trial’s duration, from when each Trial Screen started until it ended, or until three minutes passed. The system created two copies of each trial’s audio recording, one labeled with the trial’s date, time, and target word, and another with only its target word and a randomly generated 8-digit number. The latter recordings were used to facilitate blinded naming accuracy scoring and reliability evaluation (see section 2.4.4.2). The 52  system recorded each trial’s target word and random number in the data logs so that blinded accuracy scorings could later be matched up with their date and time for analysis. 2.4.4.2 Probe picture naming Probe picture naming was one of the two picture-naming tasks used in this study (the other being therapy picture naming, which was discussed in detail in section 2.4.2) and it was the study’s primary outcome measure, contributing primarily to the evaluation of therapeutic effect (Appendix B). Many aspects of the study procedure were also dependent on probe picture naming, in particular word set development and management for the multiple baseline design during Phases 2 through 4. The procedure for probe picture naming was the same as for therapy picture naming, except that, in the former, the researcher was present and participants were responsible only for navigating on the Trial Screen of NT’s Naming Practice mode (i.e., not for navigating to and from the Trial Screen) and were instructed to not select cues while doing so. Participants were instructed to name the target word for each photograph out loud, without selecting cues, but making as many or as few naming attempts as they preferred in the process. Like therapy picture naming, participants were instructed to score the accuracy of their last naming attempt using NT’s self-scoring system when they were done attempting to name the word, and to repeat this whole process on the Trial Screen for each of the remaining words in the trial series, which was presented in random order each time a trial series was started. This kind of task (i.e., uncued confrontation naming using picture stimuli) is universally used as an outcome measure in the impairment-based aphasia therapy literature (Wisenburn & Mahoney, 2009). This popularity is related to several advantages resulting from its high contrivance: simple administration and relatively objective scoring, very high reliability (i.e., test-retest correlations over 0.90), and high sensitivity to small changes over a large severity range (Herbert, Best, 53  Hickin, Howard, & Osbourne, 2003; Herbert, Hickin, Howard, & Best, 2008). Confrontation naming accuracy is highly correlated with aphasia severity (Mayer & Murray, 2003); however, there have been conflicting findings regarding whether picture-naming tasks can predict performance in more ecologically valid contexts like connected speech or everyday conversation (Carragher, Conroy, Sage, & Wilkinson, 2012; Conroy, Sage, & Lambon Ralph, 2009c; Best et al., 2011; Carragher, Sage, & Conroy, 2013). Two variables were scored from probe picture naming: naming accuracy and naming acquisition. Scoring criteria for naming accuracy were modeled on the dichotomous scoring system of the BNT-2 (Kaplan et al., 2001), which requires that participants’ final naming attempt be accurate in terms of latency, intelligibility, lexical-semantics, and phonology (Appendix K). Naming acquisition was derived from naming accuracy as a dichotomous measure of the stability of naming accuracy over time. The scoring criterion for acquisition is typically applied to the naming accuracy of aggregate word lists over one or two time points in the aphasia naming therapy literature; however, due to the design of this study (see section 2.2), it was necessary that it be applied to individual words. Because probe picture naming filled dual roles in enacting the study procedure and evaluating outcomes, naming accuracy and acquisition were scored twice during the study. Both variables were first scored online by the researcher during probe sessions so that word sets could be developed and managed as Phases 2 and 3 were carried out. They were then scored offline by the research after Phase 4 when all probe picture-naming data had been collected, so that, for the purposes of outcome evaluation, naming accuracy scoring could be completed blinded to time point using the trial audio recordings collected by NT’s data logging system. This additional offline blinded scoring was completed because the multifactorial nature of the scoring system made the validity of online scoring questionable, and more 54  importantly, because knowledge of the study time point and words’ therapy status could have biased the researcher’s online scoring (Dollaghan, 2007). Naming accuracy was the only outcome variable in this study for which reliability was evaluated. The fact that the same researcher completed both online and offline naming accuracy scoring allowed intra-rater reliability to be examined. For inter-rater reliability, the researcher and a second rater, an experienced SLP, first scored a small sample of the probe picture-naming audio recordings for each participant and discussed their results in order to clarify their understanding of the scoring criteria. Then the researcher scored all remaining audio recordings in random order for each participant, whereas the second rater scored a randomly selected ~10% of all the audio recordings for each participant, in order to evaluate inter-rater reliability of naming accuracy. Intra- and inter-rater reliability for probe naming accuracy scoring was measured with point-to-point agreement as well as Cohen’s kappa statistic in order to take into account the extent of agreement expected by chance (Viera & Garrett, 2005). Reliability values were excellent for all participants, exceeding 80% for point-to-point agreement and 0.60 for Cohen’s kappa in all cases (Table 5). High intra-rater reliability indicated that the online naming accuracy scores the researcher used in order to make decisions about developing and managing the word sets were consistent with the offline scores used in final analyses. In addition to providing a check on scoring consistency, high inter-rater reliability minimized the possibility that scoring was biased by the researcher’s prior knowledge of the study time point and words’ therapy status. Reliability values were consistently lower for P3 than P1 and P2, due to the added scoring complexity introduced by the presence of P3’s motor speech impairment. Disagreements for both intra- and inter-rater reliability resulted primarily from the researcher using a more conservative threshold for scoring attempts as intelligible.  55  Table 5. Intra- and inter-rater reliability of naming accuracy for each participant.  P1 P2 P3 Intra-rater reliability    Point-to-point agreement 99% 97% 90% Cohen’s kappa 0.97 0.94 0.78 Inter-rater reliability    Point-to-point agreement 96% 96% 83% Cohen’s kappa 0.92 0.93 0.66  2.4.4.3 Secondary outcome measures Three of the standardized tests administered during Phase 1 – the WAB-R, BNT-2, and the Cookie Theft Picture – were readministered again during Phase 4. These measures were intended as global language generalization measures to evaluate the therapeutic effect research question (Appendix B). Participant-perspective interviews (Appendix L) were also conducted with participants again during Phase 4, in order to explore participants’ perceptions of the acceptability and accessibility of self-directed mobile technology-based therapy, their process of engaging in such therapy, and any positive or negative effects it had for them. As such, data from the interviews contributed to all three research questions; however, it contributed most saliently to the feasibility question (in terms of word set development and the acceptability of therapy; Appendix A) and the therapy process question (Appendix C). The researcher’s observations and participants’ feedback from informal check-in interviews in probe sessions during Phases 3 and 4 were also used to address these issues. 2.5 Procedure This section describes in detail how the study was conducted step by step across the four phases of the study. 56  2.5.1 Phase 1: Pre-therapy screening and clinical profile evaluation Potential participants who gave informed consent were evaluated with multiple standardized and informal measures to determine their eligibility (see section 2.3.1). Measures that contributed only to screening and forming a clinical profile were administered once only before therapy during Phase 1, whereas measures also used to evaluate secondary outcomes (see section 2.4.4.3) were administered once during both Phase 1 and Phase 4. All of these measures were administered on one day or in two sessions spread over a week, if needed, to avoid fatiguing or overwhelming potential participants. Those meeting all inclusion/exclusion criteria were enrolled in the study, and their results were used for their clinical profile and/or pre-therapy secondary outcome measures. 2.5.2 Phase 2: Word set development and baseline Word set development was a five-stage process involving the nomination, evaluation, selection, confirmation, and matching and random assignment of word sets. The stages were intended to be a balanced compromise between fostering participants’ self-determination in selecting therapy goals and fulfilling the research goals and methodological requirements of the study. Data on naming accuracy collected during the evaluation and confirmation stages formed the baseline data of therapy and control sets. The nomination stage was completed with participants in 1-2 sessions, and on the researcher’s own time, over a week. It involved the collection of potential words, photographs, and information relevant to creating cues. Nomination interviews were conducted with participants using participant-nomination strategies to collect information regarding their hobbies, interests, and common activities (see section 2.4.3.3). A goal of 50 participant-nominated (PN) words was initially suggested (Mason et al., 2011); however, participants could nominate more or fewer words. Photograph and cue stimuli 57  were generally collected on the researcher’s own time, except for photographs and semantic information for PN words that were particular to the participant. The evaluation stage involved reducing the pool of nominated words to those most in need of therapy (i.e., words that participants’ could not reliably name accurately) on the basis of online naming accuracy scoring of probe picture naming, completed once on three separate days. All words that participants accurately named more than once out of the three occasions were excluded. The NT and BOSS sets were presented in a stepped approach: only NT set 1 was initially presented with the PN set, then NT set 2 was added, followed by the BOSS set, if online accuracy scoring suggested that insufficient numbers of words would be identified on the basis of the previous sets. The goal was to have three sets of at least ~20 words during the matching and random assignment stage. The purpose of the subsequent selection stage was to give participants as much control as possible over the words they practiced in therapy. Additionally, it was to ensure that both the researcher and participants understood and agreed on each photograph’s intended target word, so that appropriate cues could then be developed and naming accuracy could be validly measured. During one session, participants were presented with written lists of the eligible words from the evaluation stage – along with their corresponding photographs if participants wished to clarify which photograph represented a given word – and they were encouraged to cross out any words they would prefer not to practice. The confirmation stage involved online scoring of probe picture naming at least one additional time after the selection stage and before the beginning of Phase 3. Its purpose was to evaluate whether any of the remaining nominated words had been compromised by the selection stage. Without including this stage, changes in naming accuracy between the last baseline probe session and the first probe session during Phase 3 could have been interpreted as resulting from a 58  number of factors other than the introduction of therapy: (a) the transition would have resulted in a large change in task demands related to set size (i.e., from >90 to ~40 words); (b) establishing the intended targets of photographs could have corrected any perceptual misidentifications; and (c) the target words were stimulated during the selection stage, in the form of the written list of target words, paired with their photographs and accompanying discussion. The confirmation stage was intended to help control somewhat for these confounding variables by extending the baseline phase in order to screen for changes in accuracy with an intermediate-sized (~60 words) set. Any words named accurately were excluded. The evaluation, selection, and confirmation stages together spanned 2-4 weeks. During the week-long matching and random assignment stage, the researcher, in the absence of the participant, divided the remaining nominated words into three approximately equally sized sets of words, aggregately matched as closely as possible for baseline naming accuracy, word frequency (indexed by Lg10CD from the SUBTLEX-US database; Brysbaert & New, 2009), word length (indexed by NPHON from the MRC Psycholinguistic database; Coltheart, 1981; Wilson, 1988), and proportion of PN words (Appendix M). The purpose of matching these study word sets was to ensure the therapy sets and the control sets were comparable for evaluating therapeutic effect and to ensure that participants would be exposed to words representing a range of difficulty levels and personal relevance. The three matched study word sets were then randomly assigned to fill the roles of therapy set 1, therapy set 2, and the control set (Dollaghan, 2007). See Appendix N for the lexical characteristics of participants’ study word sets; however, note that P3’s therapy sets had to be revised in response to a clerical error that disrupted the original matching (see section 3.1.3). Therapy set 1 was initially assigned to the practice set, whereas therapy set 2 was initially assigned to the monitor set. The practice 59  set consisted of words that were made available for the participants to practice during self-directed practice in Phase 3; the monitor set consisted of words that were not currently available for practice, but were still presented for probe picture naming during probe sessions in Phase 3; and the control set consisted of words that were never made available for practice, but rather were only presented during Phases 2 and 4. As this study compared a behavioural therapy to no-therapy conditions, it was not possible to blind participants to the intervention provided (Dollaghan, 2007). Cues for all BOSS and PN words in the therapy sets were developed and modified based on feedback from two researchers experienced in aphasia therapy and psycholinguistics. Cues were then added to the existing entries for BOSS and PN words using NT’s Custom Categories Screen.  2.5.3 Phase 3: Therapy 2.5.3.1 Minimally-supervised self-directed naming therapy Based on Baker (2012) and Warren et al.’s (2007) dosage framework (see section 1.3), therapy picture-naming trials were considered to be the learning episodes of NT in this study. Participants were instructed to complete at least two trial series of therapy picture naming (see section 2.4.2) per day for their individualized practice set (i.e., a prescribed dose per practiced word of at least two trials per day). The prescribed session frequency was at least four days per week and the prescribed total therapy duration was a minimum of four weeks and a tentative maximum of six weeks, depending on participants’ performance during therapy (see section 2.5.3.2). No maximum limit was placed on participants’ practice, and session duration was permitted to vary. During the first therapy session of Phase 3, the researcher delivered the equipment for self-directed practice – a loaned iPad with carrying case, charger, and stylus (if needed) – to 60  participants and observed participants complete therapy picture naming in order to ensure they had understood the therapy regimen and were comfortable using the iPad and NT. The researcher was not present for any of participants’ other self-directed therapy sessions. Participants were instructed to take care of the equipment, and to return it at the completion of Phase 3. They were also instructed to complete therapy picture naming solely on their own. If help was needed for troubleshooting issues, participants were instructed to contact the researcher at any time via the medium most accessible to them (e.g., telephone, texting, email, or in person), or to raise issues during the next scheduled probe session. Participants were permitted to practice wherever they wished, but it was requested that they try to practice somewhere quiet for the sake of the audio recordings. They were assured that there would be no penalties for not meeting any of these expectations, but were informed that the iPad would record their actions when it was used. It was requested that participants also inform the researcher of any changes in their concurrent SLP-based therapy routine during the study. 2.5.3.2 Word set management Participants were instructed to meet in person with the researcher at a previously agreed-upon time and place at approximately the end of every therapy week for probe sessions. Probe sessions included three components: probe picture naming, word set management, and check-in interviews. Probe picture naming was discussed in detail in section 2.4.4.2. Word set management referred to the process of manipulating which words were presented to participants in NT for probe and therapy picture naming. The design of NT required that word set management be accomplished by selecting all the Category Checkboxes on NT’s Category Screen and individually selecting only the appropriate words from the Customize Word List window on NT’s Settings Screen. This process was not only used to manipulate the practice and 61  monitor sets during Phase 3, but also to manage nominated words during Phase 2 and to merge study word sets for probe picture naming during Phase 4. Check-in interviews during probe sessions involved asking participants about changes in their routine and their views about the study, troubleshooting any issues raised, retraining iPad/NT use as necessary, and informal social interaction. At the beginning of each probe session during Phase 3, the researcher merged both therapy sets together. Participants were then instructed to complete probe picture naming twice, once right after the therapy sets were merged and then again after a short break. This break was dedicated to checking in with participants and/or resting. After the second trial series of probe picture naming, the researcher transferred data logs and trial audio recordings (collected by NT’s data logging system during the preceding week) from the iPad’s hard drive onto a laptop for temporary storage until data could be securely stored at the research lab. Then, based on online naming accuracy scoring of probe picture naming, the researcher reassigned therapy set words to the practice and monitor sets in NT for therapy picture naming during the coming therapy week. Individual therapy set 1 words in the practice set that achieved the acquisition criterion were exchanged with a randomly-selected equal number of unpracticed therapy set 2 words from the monitor set. Therapy set 2 words in the practice set that were then acquired were similarly exchanged with a randomly-selected equal number of, or the remaining, unpracticed therapy set 2 words in the monitor set. Thus, the staggered introduction and withdrawal of therapy characterizing multiple baseline designs was achieved (see section 2.2). Therapy set 2 words in the monitor set that achieved the acquisition criterion without being practiced were excluded from the practice set to avoid compromising evaluation of therapeutic effect. The therapy phase continued: (a) until all words from the therapy sets had been acquired or excluded after four 62  weeks; (b) until participants’ had at least had the opportunity to practice all the eligible therapy set words (i.e., all therapy set 2 words had been transferred to the practice set or excluded) after five or after six weeks; or (c) at participants’ discretion and as time constraints permitted, if not all therapy set 2 words had been transferred to the practice set or excluded after six weeks. 2.5.4 Phase 4: Post-therapy outcome evaluation The researcher met in person with participants for three probe session within approximately one week after the end of therapy, and then again for one probe session approximately one month after the end of therapy, depending on scheduling logistics. The format of these probe sessions was the same as during Phase 3, except that the researcher had to manipulate the word sets only once, at the beginning of the first probe session, in order to merge the therapy and control sets together for all probe picture-naming trials during Phase 4. Additionally, during the last probe session, after the trial two series of probe picture naming and the logging system data had been transferred, secondary outcome measures were reassessed (see section 2.4.4.3). These measures could also be administered in two sessions spread over a week, if needed, to avoid fatiguing or overwhelming participants. 2.6 Analysis This section discusses the analytic techniques used to evaluate each of the study’s research questions and their sub-components (Table 2 is reproduced below in Table 6 for the reader’s convenience), along with the sources of data employed to address each question. Each research question sub-component was evaluated using multiple variables derived from multiple measures. Given the exploratory nature of the study, only descriptive and qualitative analyses were conducted. Inferential statistics were not used. All quantitative analyses were conducted in 63  Excel. Qualitative analyses involved synthesis of participants’ responses during interviews and/or the researcher’s observations in session notes. Table 6. Overview of research questions and their sub-components. Feasibility Therapeutic effect Therapy process 1. Word set development 2. Adherence 3. Accessibility 4. Remote progress monitoring 5. Acceptability 1. Naming accuracy 2. Secondary outcomes 1. Therapy dosage 2. Therapeutic procedures (cueing and performance-contingency)  2.6.1 Feasibility The feasibility of self-directed mobile technology-based therapy (MTT) was evaluated in terms of word set development, adherence, accessibility, NT’s potential for remote progress monitoring, and acceptability. The quantitative data automatically collected by the NT data logging system permitted a thorough descriptive analysis of variables relevant to each of these areas. These data were supported and extended by qualitative analysis of the participant-perspective interviews, the probe session check-in interviews, and the researcher’s session notes (Appendix A). The efficiency of word set development during Phase 2 was evaluated in terms of the demands it placed on participants and the researcher, the effectiveness of different nomination methods for identifying words that ‘survived’ until the matching stage (i.e., attrition rate), and the proportion of the resulting therapy and control word sets represented by each nomination method. The functional relevance of the word sets was evaluated qualitatively on the basis of the topics represented by the words participants nominated and participants’ reported perceptions of the usefulness of their therapy set words. Participants’ adherence to the study protocol was examined in terms of the broad expectations placed on participants during self-directed practice in Phase 3 (i.e., in terms of maintaining equipment, meeting the dosage 64  expectations, attending probe sessions, and following the instructions for probe and therapy picture naming). The iPad and NT’s accessibility was evaluated both for participants and for the researcher. Accessibility for participants was examined in terms of how much their navigation patterns interfered with therapy practice during Phase 3 (i.e., navigation scoring), how quickly they navigated NT (i.e., inter-selection intervals), and their perceptions of its ease of use. Accessibility for the researcher was examined in terms of his perceptions of the ease of use of NT’s word management system during Phases 2 through 4. The potential to use NT’s optional self-scoring system for remote monitoring of progress was evaluated in terms of the accuracy of the system for distinguishing between accurate and inaccurate naming responses. To do this, positive and negative likelihood ratios (LRs) were calculated from data comparing participants’ self-scoring of their accuracy to naming accuracy scoring of probe picture naming during Phases 2 through 4 (Dollaghan, 2007; Grimes & Schulz, 2005). Evaluation of acceptability was naturally primarily based on participants’ reported perceptions of therapy; however, usage data on their patterns of practice also contributed. 2.6.2 Therapeutic effect The therapeutic effect of self-directed MTT was evaluated in terms of naming accuracy, naming acquisition, and secondary outcomes. Picture-naming outcomes were analyzed in both absolute and relative terms. The former involved examining participants’ performance during Phase 3 (e.g., the number of words practiced or excluded during therapy) and Phase 4 (e.g., the number of words acquired at post-therapy; Lipsey & Wilson, 2001). The latter involved five comparisons of naming accuracy and acquisition outcomes for different word sets across different phases of the multiple baseline design (Appendix B). While comparing therapy set data from Phase 2 to Phase 4 (i.e., therapy sets pre-to-post gain) represented the direct benefit of 65  therapy to participants and the actual goal of the therapy, this comparison only indexed the association and time order relationship between therapy and outcomes (Barlow et al., 2009). To rule out confounding variables and demonstrate a causal link between therapy and outcomes, it was necessary to make three additional comparisons for outcomes of unpracticed words: comparing Phase 2 to Phase 4 for participants’ control sets (i.e., control set pre-to-post gain), comparing practice set and unpracticed monitor set words during Phase 3 (i.e., practice set-unpracticed monitor set difference), and comparing therapy and control sets during Phase 4 (i.e., therapy sets-control set post difference). Phase 3 and Phase 4 therapy set data (i.e., therapy sets therapy-to-post gain) were also compared to gain an impression of the retention of therapeutic effects. The traditional approach to analyzing relative outcome comparisons from single-subject experiments is visual analysis; however, this approach has fallen out of favour due its poor reliability and has been replaced by more objective effect size statistics (Robey et al., 1999). Therefore, while the relative comparisons of naming accuracy and acquisition outcomes were in part evaluated through visual analysis in this study, results were primarily analyzed in terms of effect sizes. Because the independent variable (i.e., therapy versus no therapy), naming accuracy, and naming acquisition were all naturally dichotomous variables, the statistical properties of the odds ratio (OR) made it the most appropriate effect size statistic for evaluating the effects of therapy in this study (Durlak, 2009; Lipsey & Wilson, 2001). Because of its flexibility, the OR could be used to calculate effect size estimates for all five of the relative outcome comparisons for both naming accuracy and naming acquisition. The OR involved comparing the odds of a probe picture-naming trial being accurate or a word being acquired when practiced to the odds of the same situation occurring when not practiced. Following recommendations (Durlak, 2009; 66  Lipsey & Wilson, 2001), 0.5 was added to each OR cell frequency when any of them equaled zero so that it was still possible to calculate an OR effect size estimate. To maintain comparability with calculation of standard mean gain and difference statistics (see below), extended baseline and withdrawal periods were excluded for OR calculations.  Interestingly, even though naturally dichotomous outcome variables, presented in the form of relative frequencies or proportions, are common in the aphasia therapy literature, standardized mean gain and difference effect sizes (SMG and SMD; e.g., Cohen’s d1 or d2, and Cohen’s d, respectively) have become the preferred statistics, despite their being appropriate only for naturally continuous variables (Durlak, 2009; Lipsey & Wilson, 2001; Beeson & Robey, 2006). Nonetheless, both SMG and SMD statistics were calculated so that outcomes could be compared with the broader aphasia naming therapy literature. Each of these statistics is restricted in terms of the relative outcome comparisons for which it can be calculated. SMG statistics can only be used to estimate gains across phases for the same word set (i.e., therapy set and control set pre-to-post gains). Cohen’s d1 is the preferred SMG statistic for single-subject designs (Beeson & Robey, 2006) and is calculated by dividing the gain from pre- to post-therapy means (i.e., Phases 2 and 4) by the standard deviation at pre-therapy. In cases when there is no variance in the pre-therapy data, it is recommended that Cohen’s d2 instead be used so that it is still possible to calculate an SMG effect size estimate (Beeson & Robey, 2006). Cohen’s d2 is calculated the same as d1, except that it substitutes the pre-therapy standard deviation with the pooled standard deviation, which is equal to the square root of the weighted mean of the variances for pre- and post-therapy phases. SMD statistics can only be used to estimate differences between word sets at the same phase (i.e., therapy sets-control set difference). Cohen’s d was calculated by dividing the difference between the means of the therapy and 67  control sets at post-therapy by their pooled standard deviation, which was equal to the square root of the weighted mean of the variances for both word sets at post-therapy. Furthermore, the properties of SMG and SMD statistics, being incompatible with nominal naming accuracy data (Lipsey & Wilson, 2001), precluded calculating effect sizes at the individual word level. This meant they could only be applied to naming accuracy for word sets as a whole, similar to the traditional across-word lists approach (see section 2.1; Beeson & Robey, 2006). However, because the staggered introduction and withdrawal of therapy was applied to individual words, rather than the composite word sets used for the SMG and SMD, it was not possible to use Beeson and Robey’s (2006) observation sample-weighting approach in their calculation. Instead, SMG and SMD statistics were weighted for word set size and calculated only on the pre- and post-therapy phase data, excluding the word-specific extended baseline and withdrawal periods for monitor set words. Although not ideal because it meant effect size estimates would represent heterogeneous retention intervals, this approach would tend to create conservative downward-biased estimates, which was deemed tolerable for this exploratory study. As SMG and SMD statistics are not directly interpretable (Durlak, 2009; Lipsey & Wilson, 2001), Robey and Beeson’s (2005) preliminary SMG (d1 and d2) benchmarks for lexical retrieval therapies were used to interpret results: 4.0, 7.0, and 10.1 for small, medium, and large effect sizes, respectively (cited in Beeson & Robey, 2006). These benchmarks were also used for Cohen’s d, as no equivalent benchmarks have been created for them. In addition to naming accuracy and acquisition, secondary outcome variables (Appendix B) were evaluated descriptively based on changes in standardized language tests, but also qualitatively using data collected from participant-perspective interviews and probe session 68  check-ins. Secondary outcomes noted in the self-directed computer-based therapy and MTT literatures were explored in particular during these interviews. 2.6.3 Therapy process In addition to participants’ dosage process, which was examined incidentally when evaluating dosage adherence for the feasibility research question, analysis of the therapy process of self-directed MTT was explored in terms of participants’ process of enacting therapeutic procedures (Appendix C). Qualitative analysis of the post-therapy participant-perspective interview and the probe session check-in interviews during Phases 3 and 4 explored both participants’ preferred cueing strategies and the extent to which they might reflect participants’ sensitivity to their own past performance (i.e., performance contingency). This analysis was supported and extended by quantitative data automatically collected during self-directed practice in Phase 3 by the NT data logging system regarding participants’ frequency of use of different cues. 2.7 Summary of methods In this four-phase study, a single-subject experiment using a multiple baseline design was conducted with three participants with chronic poststroke aphasia. Participants were first administered multiple measures to develop a clinical profile of their characteristics during Phase 1, after which, during Phase 2, the participants and researcher took steps to set up the experiment, including developing word set materials and collecting baseline data. In Phase 3, participants independently used an iPad-based therapy app called Naming Therapy© to practice naming pictures with the support of cues for words that had been assigned to therapy during Phase 2. Meanwhile, at the end of each week, the researcher manipulated which words participants would practice next, in order to enact the multiple baseline design. During Phase 4, 69  the researcher continued to monitor participants’ performance over one month in order to evaluate the outcomes of therapy. Outcome measures included extensive usage data unobtrusively collected using Naming Therapy©’s internal logging system, naming accuracy data, and pre- and post-therapy language outcomes and interviews exploring participants’ perceptions of therapy and therapy effects. Using this data, the study aimed to explore the feasibility, therapeutic effect, and therapy process of minimally-supervised self-directed MTT. 70  Chapter 3: Results This section presents findings relevant to the feasibility, therapeutic effect, and therapy process of self-directed MTT for each participant in turn. Tables 7, 10, 11, 13, 15, 16, 19, and 21 (adapted from Appendices A through C) are presented at the beginning of each sub-section in order to orient the reader to the information reported for each research question sub-component.  3.1 Feasibility 3.1.1 Word set development Table 7. Overview of word set development variables and data sources. Variables Data sources Efficiency: 1. Time and effort to nominate words, collect photographs, and develop cues using different nomination methods 2. Attrition rate of different nomination methods across word set development stages 3. Proportion of study word sets from different nomination methods Functional relevance: 1. Composition of PN word sets 2. Participants’ perception of the usefulness of word sets 1. Researcher session notes 2. Probe picture naming 3. Participant-perspective interview and check-ins Note. PN = participant-nominated.  The process of developing word sets was highly time-consuming both for participants and for the researcher. For participants, it required at least six separate sessions in which they had to go through hundreds of photographs at a time. For the researcher, completing sessions, scoring, and developing word sets were all time-consuming. However, the efficiency of different nomination methods varied within and between stages of the development process. At the nomination stage, the NT sets, simply by virtue of being native to NT, required no participant involvement and came preloaded with words, photographs, and cues, making them the most 71  efficient of the nomination methods. The BOSS set similarly did not require participants and included photographs by default, but these still had to be manually added to NT by the researcher and novel cues had to be developed for each. Participant-nominated (PN) sets were the least efficient, requiring time-consuming participant nomination, the collection of novel photographs, and the development of novel cues. In response to lack of success using spontaneous nomination, P1 and the researcher reviewed stored messages regarding basic personal information in his AAC app, from which P1 selected key words. He then repeated this process using the Bryen checklists, resulting in roughly  half of his PN set coming from each source (Table 8). The words he nominated related to his daily activities (technology, chores/errands, money, food, and transportation), his interests (music, sports, and fishing), and his children, including their names. Although certainly more time-consuming than the NT sets, collecting photographs was not especially difficult to do for P1 since most came from the Internet or photographs stored on P1’s iPad. While phonological and orthographic cues were simply derived from the word form of the intended target word, semantic  Table 8. Number of participant-nominated words for each participant, the nomination strategies used to collect them, and their composition.  P1 P2 P3 PN words 30 47 53 Strategies    Spontaneous nomination 0% 4% 0% Internet checklists 0% 96% 0% Bryen checklists 43% 0% 0% Other AAC device: 57% 0% P4: 57%, Spouse: 43% Composition of PN sets Daily activities: 53% Interests: 33% Children: 13% Politics: 55% Sports: 40% Leisure activities: 4% Proper names: 65% Daily activities: 35% Note. PN = participant-nominated. See section 2.4.3.3 for description of the nomination strategies. 72  cues required additional time and consideration to develop. However, this process was aided in P1’s case by the ready availability of semantic information contained in the AAC app stored messages and the highly idiosyncratic semantic features of the personalized words he chose, which made it relatively easy to narrow the field of semantic competitors.  During the evaluation through confirmation stages, the NT and PN sets for P1 varied in their attrition rates. A large portion of both sets survived the evaluation stage (Table 9); however, P1 disproportionately chose not to practice more words from the NT set than his PN set during the selection stage, which he reported was because he found that many of the NT set words were not useful/relevant to him. Both sets showed further, proportionally equivalent reductions at the confirmation stage. Although the NT set showed a higher attrition rate, because of the small initial sizes of P1’s PN set, NT words still made up the majority of his final study word sets. Neither nomination method was able to identify a sufficient number of study words independently. During the post-therapy interview, P1 reported that some of the words he practiced were not particularly useful, including some of the PN words. He noted that some NT words were things he could see and use without needing to say them, while some PN words were relevant to his past interests, but not to his current everyday life. P2 was able to spontaneously nominate only two words related to his leisure activities during the nomination stage. Instead, the vast majority of his PN set was collected using the Internet checklist (Table 8). He predominantly chose words related to his interests in sports and American politics, including the names of American senators and famous tennis and football athletes as well as specialized terms related to each of these topics. As most of his PN words were public figures, photographs were readily available from the Internet. In contrast, developing cues for P2’s BOSS and PN sets was highly time-consuming. Their highly specialized but 73  impersonal terms and proper names, from very narrowly defined semantic categories, made it difficult to develop semantic cues that presented information concisely, while still narrowing the field of competitors sufficiently to elicit only the target word. The researcher had to consult encyclopedias to develop many of these cues (adding to their time demands) since he lacked adequate content knowledge to develop them independently. The resulting Definition and Phrase Completion cues were of variable cueing adequacy, in that they were often long and usually unlikely to elicit the target word in likely all but the few people with relevant expert knowledge. This gave them the distinct impression of the arcane kinds of questions found in trivia games. Table 9. Attrition rates of, and the proportion of words in the finalized study word sets for, the nomination methods for each participant.  Nomination Evaluation Selection Confirmation Study sets P1      NT set 1 150 106 66 53 76% Attrition -0% -29% -56% -65%  PN set 30 23 20 17 24% Attrition -0% -23% -33% -43%  P2      NT set 1 150 5 4 0 0% Attrition -0% -97% -97% -100%  NT set 2 172 28 25 5 8% Attrition -0% -84% -85% -97 %  BOSS set 83 30 30 30 47% Attrition -0% -64% -64% -64%  PN set 47 34 34 29 45% Attrition -0% -28% -28% -38%  P3      NT set 1 150 29 29 19 32% Attrition (%) -0% -81% -81% -87%  NT set 2 172 47 46 30 50% Attrition (%) -0% -73% -73% -83%  PN set 53 15 15 11 18% Attrition (%) -0% -72% 72% -79%  Note. NT = Naming Therapy©, BOSS = Bank of Standardized Stimuli, PN = participant-nominated. See section 2.5.2 for description of the word set development stages.  74  While P2’s PN set performed relatively well during the evaluation stage, the NT sets showed such high attrition rates that the high-difficulty BOSS set had to be developed to try to make up the difference; however, even the BOSS set showed high attrition at this stage (Table 9). In contrast, P2 chose to eliminate only a few words, all from the NT sets, during the selection stage, which he reported was motivated by an awareness that he was approaching the study’s 60-word minimum and by a desire to practice as many words as possible. By the end of the confirmation stage, virtually all of the words from the NT sets had been eliminated, leaving P2’s finalized study words roughly evenly split between BOSS words and his PN words. None of the nomination methods were able to independently identify a sufficient number of study words. During the interview, P2 reported being satisfied with the particular words he had practiced. P3 preferred to nominate words on his own time, so it was not possible to monitor his nomination strategies directly during the nomination stage. He reported collaborating with his spouse, with each of them identifying roughly half of P3’s PN set (Table 8). He reported that they both nominated words using multiple methods. They identified words related to recurrent topics in P3’s personal diary of his past everyday activities. They also kept a running list of potentially useful words collected throughout the day while carrying out P3’s current daily activities (i.e., a communication diary; Renvall et al., 2013a). Although they did not select any words from the Bryen checklists, P3 reported that they also looked through these lists to give them ideas for potentially relevant words. The words he and his spouse nominated included names of friends, neighbours, family, and pets as well as words related to P3’s daily activities (food, chores/errands, and health). Photographs for some of P3’s PN words could be collected from his iPad or scanned from hardcopies; however, due to the large number of highly personalized words among his PN words, it was also necessary for P3 and his spouse to 75  collaborate in photographing people, places, and things in the course of their everyday routine. Similar collaboration was required to identify semantic information for cue development. The highly time-consuming process of cue development was somewhat alleviated for P3 by the idiosyncratic semantic features of the personalized words he chose. The NT and PN sets for P3 showed uniformly high attrition rates during the evaluation stage. The addition of NT set 2 did little to correct the situation (Table 9). P3 chose to eliminate only one word during the selection stage, reportedly out of a concern about going under the 60-word minimum. All of the word sets showed further proportionally equivalent reductions at the confirmation stage. Because all sets showed similar attrition rates and his PN set started out smaller than the NT sets, NT words still made up the majority of P3’s finalized study word sets. All three word sets were required to identify a sufficient number of study words for P3. During the interview, P3 acknowledged having found some of his therapy set words unuseful.  3.1.2 Adherence Table 10. Overview of adherence variables and data sources. Variables Data sources 1. Equipment maintained and returned 2. Therapy picture-naming trials practiced independently 3. Proportion of probe picture-naming trials cued 4. Attendance at probe sessions Dosage: 1. Visual inspection of dose-per-therapy day distribution 2. Actual-to-expected ratios of mean dose, frequency, and cumulative dose per therapy word per week 3. Participants’ reported therapy process 1. NT data logging system 2. Researcher session notes 3. Participant-perspective interview and check-ins Note. NT = Naming Therapy©. 76  P1 generally closely followed the expectations of this study. He returned the loaned equipment in good working condition. A sample of recordings from each therapy session for P1 revealed that he completed one trial series in the presence of someone he personally knew. He did not mention this occasion during the post-therapy interview. His verbal behaviour during the recordings indicated that he was using the iPad to support discussion about the iPad and NT and about his subjective experience of anomia, rather than to practice per se. P1 also selected cues during five (0.3%) of all his probe picture-naming trials (which were meant to be uncued). Three of these five cases were the result of accidentally selecting cues or lapses in memory; however, in two instances he selected cues intentionally as a communication support to explain to the researcher that he was frustrated about not being allowed to use them during probe sessions (see section 3.2.2). Figures 1 through 3 show each participant’s mean dose (i.e., trials per day) per word plotted against therapy phase day, and Appendix O describes their dosage schedules relative to the values expected according to their prescribed schedule. P1 showed a very methodical pattern of practice (Figure 1), typically practicing each word twice a day, 4-6 days per week, for a mean of roughly 45 therapy picture-naming trials per word. His mean dose per word always hovered around the prescribed value, whereas he typically exceeded expectations in terms of session frequency and therefore cumulative dose as well (Appendix O). P1’s reported adherence was largely consistent with the observed patterns discussed above. During the post-therapy interview, P1 reported that while he had appreciated the researcher’s involvement, he practiced above the prescribed session frequency because he was motivated to do well.  77   Figure 1. P1’s dose per therapy day during self-directed practice and probe sessions.  Figure 2. P2’s dose per therapy day during self-directed practice and probe sessions.  Figure 3. P3’s dose per therapy day during self-directed practice and probe sessions.0"2"4"6"8"10"12"1" 2" 3" 4" 5" 6" 7" 8" 9" 10"11"12"13"14"15"16"17"18"19"20"21"22"23"24"25"26"27"28"29"30"31"32"33"34"35"36"37"38"39"40"41"42"43"44"45"46"47"48"49"50"51"Mean%dose%per%word%Therapy%phase%day%Self0directed"prac8ce"Therapy"probe"sessions"0"2"4"6"8"10"12"1" 2" 3" 4" 5" 6" 7" 8" 9" 10"11"12"13"14"15"16"17"18"19"20"21"22"23"24"25"26"27"28"29"30"31"32"33"34"35"36"37"38"39"40"41"42"43"Mean%dose%per%word%Therapy%phase%day%Self0directed"prac8ce"Therapy"probe"sessions"0"2"4"6"8"10"12"1" 2" 3" 4" 5" 6" 7" 8" 9" 10"11"12"13"14"15"16"17"18"19"20"21"22"23"24"25"26"27"28"29"30"31"32"33"34"35"36"37"38"39"40"41"42"43"44"45"Mean%dose%per%word%Therapy%phase%day%Self0directed"prac8ce"Therapy"probe"sessions"78  P2’s adherence to the therapy protocol was generally good, except in terms of dosage. He returned the study equipment in good working condition. A sample of recordings from each therapy session for P2 revealed that he completed one trial series in the presence of his SLP. He did not mention this occasion during post-therapy interviews; however, his verbal behaviour during these recordings indicated that he was using the iPad to support discussion about the iPad and NT. P2 selected cues during four probe picture-naming trials (0.4% of all his probe picture-naming trials), all of which were the result of accidentally selecting cues or lapses in memory. P2’s pattern of practice was very variable and largely non-adherent until primarily the last two weeks of the therapy phase. He chose to practice for only a few days scattered throughout the first few weeks of therapy, with over half of this practice attributable to only one day when he practiced each word roughly ten times before a therapy probe session (Figure 2). Then P2 went on a last-minute vacation straddling therapy weeks 3 and 4, which resulted in the loss of a probe session and very little practice on his part. Although he practiced near or far above the prescribed dose after therapy week 1 (i.e., up to three times the prescribed dose; Appendix O), his cumulative dose lagged far behind the expected value until after he returned from his trip, when he started practicing at high doses again and increased his session frequency to 5-6 days per week in therapy weeks 5 and 6. By the end of therapy, his cumulative dose was a mean of roughly 44 therapy picture-naming trials per word. The dosage schedule that P2 reported both during probe sessions and after therapy (i.e., roughly 20-30 minutes/day for 4-5 days/week) was broadly consistent with his behaviour only during the last two therapy weeks. P2 reported that he practiced as much as he did because he wanted to fulfill the dosage expectations. His intended meaning was not entirely clear (Appendix P), but he appeared to suggest that supervision of self-directed practice was important to maintaining his motivation. 79  P3 was highly adherent to the therapy protocol. He returned the study equipment in good working condition, he always practiced independently, and he selected cues during only three (0.3%) of all his probe picture-naming trials, all of which were the result of accidentally selecting cues. P3 showed a methodical pattern of practice (Figure 3). He typically practiced each word 2-4 times per day, for 6-7 days per week, for a mean of roughly 63 therapy picture-naming trials per word. His mean dose and session frequency per word always exceeded expectations, leading his cumulative dose to consistently be 2-3 times the expected value (Appendix O). P3’s reported adherence was largely consistent with the observed patterns discussed above. He reported practicing as much as he did because he wanted to fulfill the dosage expectations, and acknowledged that supervision contributed to his motivation to continue practicing. 3.1.3 Accessibility Table 11. Overview of accessibility variables and data sources. Variables Data sources 1. Navigation scoring per navigation step 2. Five-number summaries of inter-selection interval distributions per navigation step 3. Participants’ reported accessibility of therapy 4. Accessibility of NT’s word management system 1. NT data logging system 2. Participant-perspective interview and check-ins 3. Researcher sessions notes Note. NT = Naming Therapy©.  Overall, participants made few deviations from the navigation path they were instructed to follow (i.e., alternative button selections) and even fewer selections that interfered with completing therapy (i.e., maladaptive button selections; see navigation scoring in Appendix J). Because of this, their patterns of navigation are described only generally here (see Appendix Q for a full list of alternative and maladaptive button selections). P1 deviated occasionally from the 80  path by navigating from the Home Screen to the settings screens and the other practice modes (Table 12). However, he never changed any settings, and his use of the other practice modes consisted of quickly moving through a few trials without selecting cues and then prematurely ending the trial series. Self-scoring caused relatively more problems for P1, mostly in terms of skipping trials, but these were still rare. When exiting NT, P1 would sometimes use an alternative method than the one trained, or would sometimes prematurely end a trial series by selecting the iPad Home Button. P1 navigated through NT quite quickly, dwelling for relatively little time on steps that did not involve direct practice with cues (Appendix R). The 75th percentile for his inter-selection interval distribution (see inter-selection interval in Appendix J) indicated that most of his selections were completed in less than ~5 s, depending on the step. P1 reported during the post-therapy interview that he had no difficulty completing therapy by himself and  Table 12. Participants’ navigation scoring during therapy and probe sessions for each step in the basic iPad-Naming Therapy© navigation path.   Navigation step  NT navigation scoring (%) Home Screen Category Screen Therapy Trial Screen Probe Trial Screen Popup Screen Exiting NT P1 Adherent 83.5 98.7 99.2 99.7 100.0 79.6  Alternative 16.5 1.3 0.2 0.0 0.0 10.2  Maladaptive 0.0 0.0 0.5 0.3 0.0 10.2 P2 Adherent 77.6 100.0 14.2 93.7 97.6 55.3  Alternative 22.4 0.0 84.7 4.5 2.4 6.7  Maladaptive 0.0 0.0 1.1 1.8 0.0 40.0 P3 Adherent 100.0 100.0 99.4 98.0 100.0 5.6  Alternative 0.0 0.0 0.1 0.8 0.0 94.4  Maladaptive 0.0 0.0 0.5 1.2 0.0 0.0 Note. All values are in percentage points of the total button selections per navigation step. Scoring is relative to the basic navigation path that participants were instructed to follow and the functionality of NT. Adherent = consistent with this path, alternative = not consistent with this path but not interfering with carrying out therapy practice, maladaptive = interfering with carrying out therapy practice. See Appendix F: Home Screen = Step 5, Category Screen = Step 6, Therapy Trial Screen = Steps 8-9 for therapy picture naming, Probe Trial Screen = Steps 8-9 for probe picture naming, Popup Screen = Step 10, Exiting NT = Step 11. NT = Naming Therapy©. 81  that he found the iPad and NT easy to use (Appendix P). He also reported that he tried out other practice modes and explored other apps on the iPad out of curiosity, but made sure to only do so after completing the prescribed dose. He also reported that he did not mind self-scoring, and in  fact felt it did not matter that NT could not automatically provide accuracy feedback, since he felt he was able to judge his accuracy for himself. P2 generally showed high adherence to the navigation path, except in terms of self-scoring and exiting NT. P2 deviated occasionally from the path by navigating from the Home Screen to the settings screens and the other practice modes (Table 12). However, he never changed any settings, and his use of the other practice modes consisted of quickly moving through a few trials without selecting cues and then prematurely ending the trial series. Most of the time P2 chose to select the Next Button rather than self-score each trial during self-directed practice. This greatly contrasted with his behaviour during probe sessions, in which he relatively consistently self-scored each trial without needing the researcher to step in. On the Popup Screen, P2 more often selected the Popup Try Again button to start another trial series immediately, especially during his long high-dose sessions (see section 3.1.2). P2 also commonly ended his therapy sessions prematurely in the middle of a trial series by simply allowing the iPad to time out and go to sleep. P2 navigated through NT quite quickly: the 75th percentile for his inter-selection interval distribution indicated that most of his selections were completed in less than ~3.5 s, depending on the step (Appendix R). The only exception to this pattern was P2’s lengthy intervals for exiting the app due to his maladaptive choice of strategies. P2 reported having no difficulty completing therapy by himself and that he found the iPad and NT easy to use. P2 reported during the post-therapy interview that he usually did not self-score when he was by himself because he found it tedious and not useful. 82  P3 was highly adherent to the navigation path he was instructed to follow, except primarily when it came to exiting the app (Table 12). His typical method of exiting NT was to simply let it time out after completing the trial series. His only maladaptive selections related to self-scoring, primarily in the form of accidentally skipping trials by double tapping the Correct Button. P3 was also very quick when navigating NT, with the 75th percentiles for his inter-selection interval distribution indicating most of his selections were completed in less than ~5 s, depending on the step (Appendix R). The biggest exception to this pattern was P3’s lengthy intervals for exiting the app, which resulted from his tendency to allow the iPad to time out. P3 reported that he had no difficulty completing therapy by himself and that he found the iPad and NT easy to use. P3 also found self-scoring somewhat bothersome, but reported completing self-scoring every trial because he wanted to fulfill the researcher’s expectations. Word set management (i.e., manipulating which words were presented to participants during each study phase) proved cumbersome for the researcher. As it is not possible to create customized sets of NT words, the researcher had to use the Customize Word List window on the Settings Screen each time he had to manipulate the word sets. This entailed meticulously scanning through a lengthy list of the hundreds of words native to NT in order to find and select each desired word individually. Selection errors were thus inevitable; however, as there was no easy way of reviewing which words had been selected in order to detect such errors, the entire list had to be repeatedly checked to identify and correct them. Moreover, editing the custom PN word sets on the Categories Screen automatically selected all of these words in the Customize Word Lists window, meaning that the entire list of words had to be reviewed once again to deselect all the undesired words. 83  Manually reconstructing the study word sets word by word at least twice during each session (for completing probe picture naming and reassigning words to the practice and monitor sets) not only resulted in lengthy delays for participants, but also compromised the experimental design for P3. Due to a clerical error on a reference list the researcher used to help him reconstruct the practice and monitor sets, P3’s therapy word sets were randomly assigned to the practice and monitor sets at the end of the therapy week 1 probe session, thus compromising their original matching and the staggering of therapy. As this oversight was detected only weeks later, the researcher proceeded to manage the word sets as if the therapy words were still assigned to the original sets, resulting in 13 of P3’s therapy set words being removed from the practice set after only one week (instead of the minimum two), with three of these then returned to the practice set several weeks later. Although highly inconvenient, the oversight thankfully did not catastrophically compromise the experimental design. The randomly assigned therapy sets that were accidentally formed (i.e., revised therapy sets), rather than P3’s original matched therapy sets, could be used for therapeutic effect analyses, as the unintentional random assignment produced revised therapy sets for P3 that were still relatively similar in composition to one another and the control set (Appendix H). 3.1.4 Remote progress monitoring Table 13. Overview of the remote progress monitoring variable and data sources. Variable Data sources 1. Categorization accuracy of NT’s self-scoring system in terms of positive and negative likelihood ratios 1. NT data logging system 2. Probe picture naming Note. NT = Naming Therapy©.  Analysis of self-scoring categorization accuracy during probe picture naming showed the same pattern across all three participants. Therefore, these results are summarized collectively, 84  rather than sequentially. Results indicated a dissociation between the ability of Correct and Incorrect Button selections to predict participants’ naming accuracy during probe picture naming (Table 14). Comparison of participants’ self-scoring during probe sessions with the corresponding results of naming accuracy scoring indicated that, across participants, an Incorrect Button was extremely likely (i.e., LR+ >10.00) to be preceded by an inaccurate attempt, whereas a Correct Button was only marginally more likely (i.e., LR- ~0.30) to follow an accurate attempt than an inaccurate one, meaning that a Correct Button selection was suggestive but insufficient to rule out the possibility that it was preceded by an inaccurate attempt (Dollaghan, 2007; Grimes & Schulz, 2005). This pattern of findings resulted from the self-scoring system having very high specificity for accurate attempts. Participants’ tendency to have relatively higher false negative/miss rates resulted in the self-scoring system having low to moderate sensitivity to inaccurate attempts; however, because participants also tended to have low false positive/false alarm rates (i.e., high specificity), an Incorrect Button selection was very likely to represent an inaccurate attempt. Although perhaps counter-intuitive at first, the logic here is that if a participant had perfect self-scoring specificity, meaning they never mistakenly scored an accurate attempt with the Incorrect Button, then if they did select the Incorrect Button, it must have followed an inaccurate attempt (Dollaghan, 2007).   85  Table 14. Diagnostic 2x2 tables for each participant’s self-scoring categorization accuracy during probe sessions. P1 Probe picture naming Inaccurate Accurate Self-scoring Incorrect 47% 0% Correct 14% 39% Base rate 61% 39% Sensitivity 0.77 Specificity 0.99 LR+ 95.43 LR- 0.23 P2 Probe picture naming Inaccurate Accurate Self-scoring Incorrect 38% 1% Correct 14% 47% Base rate 52% 48% Sensitivity 0.72 Specificity 0.98 LR+ 41.20 LR- 0.28 P3 Probe picture naming Inaccurate Accurate Self-scoring Incorrect 35% 0% Correct 25% 40% Base rate 60% 40% Sensitivity 0.58 Specificity 0.99 LR+ 61.30 LR- 0.42 Note. All true and false positive and negative values are percentages of the whole sample. True positive (upper left cell) = trials scored as inaccurate by the researcher that the participant scored as incorrect, true negative (lower right cell) = trials scored as accurate by the researcher that the participant scored as correct, false positive (upper right cell) = trials scored as accurate by the researcher that the participant scored as incorrect, false negative (lower left cell) = trials scored as inaccurate by the researcher that the participant scored as correct. LR+ = positive likelihood ratio, LR- = negative likelihood ratio. LR+ benchmarks: ≥10 = very positive, 3 = moderately positive, 1 = neutral; LR- benchmarks: ≤0.10 = extremely negative, ≤0.30 = moderately negative, 1 = neutral (Dollaghan, 2007).   86  3.1.5 Acceptability Table 15. Overview of acceptability variables and data sources. Variables Data sources 1. Visual inspection of therapy session start time distribution 2. Five-number summaries of trial, trial series, session duration, and weekly time commitment distributions 3. Total time commitment 3. Participants’ reported acceptability of therapy 1. NT data logging system 2. Participant-perspective interview and check-ins Note. NT = Naming Therapy©.  The NT data logging system indicated that P1 primarily preferred to practice at a relatively set time in the late morning (Figure 4). Very little time was required on P1’s part to fulfill or exceed the prescribed dosage schedule (Appendix S): using the 75th percentile of the distributions, P1 most often completed a trial in under ~12 seconds, a trial series in under ~4 minutes, a session in under ~8 minutes, the weekly time commitment in under ~40 minutes, and the entire therapy phase took him under four hours and ~11 minutes. During the post-therapy interview, P1 reported being satisfied with self-directed mobile technology-based therapy (MTT) overall (Appendix P). He reported he would use NT again if he had the chance and was interested in having the word sets created for the study loaded onto his personal iPad. P1 reported preferring self-directed MTT to traditional SLP-based therapy for the flexibility and convenience that came with self-directed practice, as the opportunity to choose where and when to practice made it easy for him to fit therapy into his schedule without disrupting his other daily activities.  87   Figure 4. Distribution of therapy session start times for each participant during therapy. However, P1 explained that what he valued most about self-directed practice with minimal supervision was the control it gave him over the therapy process. This stood in contrast to working with therapists, who he perceived as tending to take too much control over his learning. To explain his perspective, P1 enacted a satirical skit in which he role-played both a therapist and a client interacting in therapy (Appendix P). In it, he presented a client motivated to learn who pleads to practice more with a highhanded therapist, who in turn makes an appeal to their own expert status when questioned about their actions. Although his skit reflected poorly on therapists, it was intended in good humour, and P1 was quick to repeatedly emphasize that any therapy is good therapy, whether it comes from a therapist, an iPad, an aphasia support group, or anyone else. Although he appreciated the flexibility, the independence, the researcher’s involvement, and the opportunity for repetition that self-directed MTT gave him, none of that mattered to him in the end as much as getting therapy in the first place. P1 also noted some dissatisfying aspects of self-directed MTT. P1 reported finding NT boring and overly repetitive; however, he was quick to counter this with his previous rationale that any therapy is good therapy. He also complained that the practice word set used in the study 0"10"20"30"40"50"60"70"Percentage)of)therapy)sessions)Therapy)session)start)2me)of)day)P1"P2"P3"88  was too small, and that he should have practiced all the study word sets at once to make even more progress. As previously mentioned, P1 also reported that some of the words he practiced were not particularly useful; however, he pointed out that he did not really mind what words he practiced, since he wanted to be able to say everything again. P2 preferred to practice at a relatively set time in the late morning (Figure 4). Therapy entailed a relatively small time commitment for P2 (Appendix S): his trials were most often completed in under ~11 seconds, his trial series in under ~4 minutes, his sessions in under ~14 minutes, his weekly time commitment in under one hour and ~13 minutes, and the entire therapy phase took him under ~5 hours. P2 reported being satisfied with therapy overall because he felt it helped him make progress (Appendix P). He indicated that he would use NT again and was interested in having the word sets created for the study loaded onto his personal iPad. P2 valued both self-directed practice and SLP-based therapy. He valued the flexibility and convenience of self-directed practice; however, he reported preferring SLP-based therapy because of the social interaction it provided, and, more importantly to him, because of the comprehensive therapy, tailored to his own needs, that a therapist could provide. Regardless of P2’s intended meaning about the importance of supervision (Appendix P), it was clear that he enjoyed the social interaction that supervision provided and was a little dejected that it would be coming to an end. Although P2 preferred SLP-based therapy, he also valued self-directed MTT as part of a holistic approach to recovery. As such, his ideal was to combine both together, and in fact he reported planning on proposing exactly that idea to his SLP during their next session. P2 was careful to temper his criticism by reiterating the benefits of iPads throughout the interview. However, he wished he could have practiced all of the study word sets, as he was disappointed that he did not perform as well on the control set words at post-therapy. As previously mentioned, P2 reported 89  finding the self-scoring system tedious and not useful. He also complained that the audio recordings for his semantic cues were overly long and wordy. Finally, during the post-therapy interview, P2 also admitted that he found NT was not difficult enough for him to make substantial gains (Appendix P). P3 chose to practice following a much more dispersed daily schedule, practicing at least once across the study during every hour of the 17-hour window between 8:00 AM and 12:00 AM (Figure 4). NT’s data logging system indicated that it took P3 little time to meet the study’s prescribed dosage (Appendix S): he most often completed trials in under ~12 seconds, completed trial series in under ~3 minutes, and most often finished his therapy sessions in less than ~8 minutes. However, because P3 often completed multiple sessions per day for most days of the week, his weekly time commitment of therapy was relatively high, being most often under one hour and ~40 minutes. By the end of therapy, his total time commitment had been ~8.5 hours. During the post-therapy interview, P3 reported being satisfied with therapy overall because he found it helpful (Appendix P). Moreover, he reported he would use NT again if given the chance and was interested in having the word sets created for the study loaded onto his personal iPad. P3 reported preferring self-directed MTT to traditional SLP-based therapy, but found it difficult to explain why. He appreciated the flexibility of self-directed practice and appreciated the researcher’s involvement in therapy, reporting that he would miss meeting with him after the end of the study. Although P3 acknowledged that supervision contributed to his motivation to continue practicing, he reported that he would still prefer self-directed practice if there were no supervision, because even if he practiced less, he would still be practicing and that would help. The only dissatisfactions P3 reported about self-directed MTT was his 90  acknowledgement that some of the words he practiced had not been useful and that self-scoring had been somewhat tedious. 3.1.6 Summary of findings for the feasibility analysis To gain an impression of the feasibility of enacting self-directed MTT with minimal supervision, this study examined the efficiency of developing word sets for the study and their functional relevance to participants, participants’ adherence to the study protocol, the accessibility of NT both for participants and for the researcher, the potential of using NT’s self-scoring system for remote progress monitoring, and the acceptability of this form of therapy to participants. Word set development was found to be time-consuming for all parties involved. NT sets were the most efficient and PN sets the least at the nomination stage; however, NT sets generally had higher attrition rates at subsequent stages. None of the nomination methods were able to identify a sufficient number of study words independently. P1 and P3’s PN sets consisted of words related to their everyday lives; however, at post-therapy both participants reported being dissatisfied with the usefulness of some of the words they had practiced. Adherence to the study protocol was generally high for all participants, with a few exceptions for P2. P1 and P3 showed a methodical routine, in which they consistently practiced at or above the prescribed dosage, whereas P2 practiced inconsistently, demonstrating poor dosage adherence until the last two weeks of therapy. P1 reported that he had been motivated to practice by a desire to do well, whereas P2 and P3 reported wishing to fulfill the researcher’s expectations and at least P3 reported that the researcher’s supervision had affected his motivation to practice. In terms of accessibility for participants, all of them were largely able to consistently follow the navigation path they had been instructed to follow, and they were able to do so very quickly. They all reported finding NT easy to use, except that P2 and P3 found self-scoring to be tedious. For the 91  researcher, word set management using NT was cumbersome, time-consuming, and resulted in P3’s experimental design being compromised. In terms of remote progress monitoring, NT’s self-scoring system demonstrated partial value in differentiating accurate and inaccurate naming attempts during probe sessions. Participants reported finding self-directed MTT acceptable and showed unique daily patterns of practice, with the therapy protocol requiring little time commitment from any of them. P1 and P3 reported preferring self-directed MTT to SLP-based therapy, whereas P2 valued both. Participants valued the scheduling flexibility, autonomy, and the opportunity for repetition that came with self-directed MTT; but they also reported dissatisfaction with the words they had practiced, and reported finding NT boring or insufficiently challenging.  92  3.2 Therapeutic effect 3.2.1 Naming accuracy and naming acquisition Table 16. Overview of naming accuracy and naming acquisition variables and data sources. Variables Data sources Absolute effects: 1. Number of words acquired, excluded, practiced, unpracticed during therapy 2. Number of therapy staggerings 3. Number of words acquired at post-therapy, including number of PN words among them Relative effect comparisons: 1. Visual inspection of multiple-baseline figures 2. Therapy sets pre-to-post gain (SMG/OR) 3. Control set pre-to-post gain (SMG/OR) 4. Therapy sets therapy-to-post gain (OR) 5. Practice set-unpracticed monitor set difference (OR) 6. Therapy sets-control set difference (SMD/OR) 1. Probe picture naming 2. Researcher session notes Note. SMG = standard mean gain statistic, SMD = standard mean difference statistic, OR = odds ratio.  Results regarding the relative effect of therapy on naming accuracy are depicted in multiple-baseline format for each participant in Figures 5-7. Note that the figures do not accurately represent the experimental design employed in this study. Illustrating the dynamic word set approach used in this study would require graphing word-level data. As such data cannot be manageably depicted in graphical format, the figures depict only aggregate naming accuracy data for each study word set. Although this makes the figures similar to how the more the traditional across-word lists approach (see section 2.2) would be depicted graphically, it obscures the incremental staggering of therapy as enacted in this study. In order to give some impression of the fact that therapy set 1 words in the practice set were incrementally swapped as 93  they were acquired with words from the monitor set, the percentage of the therapy sets that had been assigned to the practice set for a given therapy week is also depicted in the figures so that the general temporal relationship between the staggering of therapy and changes in naming accuracy can be seen. The reader can also gain an impression of how the dynamic word set approach was enacted for each participant from the summary statistics reported in Table 17. Table 18 contains relative estimates of effect size for naming accuracy and acquisition for each participant. Table 17. Effect of therapy in absolute terms for therapy set words during the therapy phase for each participant.  P1 P2 P3  n % n % n % Acquired words 14 30.4 24 57.1 18 45.0 Excluded words 3 6.5 4 9.5 1 2.5 Practiced words 38 82.6 37 88.1 39 97.5 Unpracticed words 5 10.9 1 2.4 0 0.0 Therapy staggerings 5  4  5  Note. Acquired = words that achieved the acquisition criterion (see Appendix K) while in the practice set; excluded = words that achieved the acquisition criterion without ever being transferred to the practice set; practiced = eligible words (i.e., not excluded) that started out in, or were eventually transferred to, the practice set; unpracticed = eligible words (i.e., not excluded) that were never transferred to the practice set before the end of the therapy phase; therapy staggerings = occasions when therapy was introduced and withdrawn from words (see section 2.5.3.2).  Over five staggerings of the introduction and withdrawal of therapy to individual words, P1 had the opportunity to practice the majority of his therapy set words by the end of therapy (Table 17). Only three of his therapy set words in the monitor set had to be excluded due to being acquired without therapy. However, because P1 was able to acquire only about a third of his therapy set words during therapy, five words could not be transferred to the practice set before P1 chose to end therapy at the end of therapy week 7 (see sections 2.2 and 2.5.3.2 for 94  clarification of the experimental design). Moreover, only nine (20%) of P1’s total therapy set words remained acquired at post-therapy, only one of which was a participant-nominated (PN) word. Effect size estimates for P1 indicated that self-directed MTT resulted in large improvements in naming accuracy and acquisition relative to both baseline and unpracticed words. The pre-to-post standard mean gain (SMG) effect size for P1’s therapy sets indicated a large improvement from baseline to post-therapy, relative to the naming therapy literature (Table 18). The pre-to-post SMG effect size for P1’s control set showed a small naming accuracy gain from baseline to post-therapy. Moreover, P1’s therapy sets showed a medium standard mean difference (SMD) effect size relative to his control set. Examining the odds ratio (OR) for this same comparison indicated that the odds of a given probe picture-naming trial being accurate post-therapy were more than eight times higher for P1’s therapy sets than his control set. More importantly, the odds of a word being acquired (i.e., consistently accurate) post-therapy were similarly more than seven times higher for therapy sets than control sets. Moreover, the odds of a word being acquired during the therapy phase were ~4 times higher if it was receiving therapy in the practice set than if it was still awaiting therapy in the monitor set. Although such acquisition was the ultimate goal of therapy, naming accuracy provides a more conservative estimate of the causal link between therapy and outcomes, as it is more sensitive than acquisition to small changes in performance. Nonetheless, the OR indicated that the odds of an accurate probe picture-naming trial occurring were almost five times higher for P1’s words in the practice set than unpracticed words in the monitor set. However, the OR analysis of the change in performance from therapy to post-therapy phases indicated that P1’s odds of acquisition were almost two times lower after therapy than they were during therapy. 95  Table 18. Effect size estimates for probe picture-naming outcomes.  P1 P2 P3  d/d1/d2 OR d/d1/d2 OR d/d1/d2 OR Uncued naming accuracy       Therapy sets pre-to-post gain 13.63 34.64 29.66 113.79 7.15 19.06 Control set pre-to-post gain 3.93 4.20 2.65 2.76 1.36 2.39 Therapy sets-control set post difference 5.97 8.69 10.59 23.83 5.45 6.01 Practice set-unpracticed monitor set difference  4.79  3.66  3.93 Acquisition       Therapy sets pre-to-post gain  24.80  79.05  21.32 Control set pre-to-post gain  3.13  3.14  1.00 Therapy sets therapy-to-post gain  0.53  0.57  0.30 Therapy sets-control set post difference  7.14  22.17  11.06 Practice set-unpracticed monitor set difference  3.89  7.85  16.29 Note. d = Cohen’s d (a standardized mean difference statistic); d1 = Cohen’s d1 and d2 = Cohen’s d2 (standardized mean gain statistics); OR = odds ratio. Standard mean gain benchmarks: 4.00 = small, 7.00 = medium, 10.10 = large (Robey & Beeson, 2005; cited in Beeson & Robey, 2006). Blank cells indicate comparisons that cannot be calculated using standardized mean differences or gains.  For P2, based on his naming accuracy performance, it was possible to stagger therapy four times over the therapy phase, allowing him to practice all of his eligible (i.e., not excluded) therapy set words (Table 17; note that one of P2’s words was incorrectly excluded due to a mismatch between online and offline naming accuracy scoring of probe picture naming). During the therapy phase, P2 acquired roughly 60% of his therapy sets through self-directed practice, whereas only 10% of the sets were excluded due to being acquired without practice. Nineteen (45%) of P2’s total therapy set words remained acquired at post-therapy, among which nine were PN words. The pre-to-post SMG effect size for P2’s therapy sets indicated a very large gain in naming accuracy from baseline to post-therapy (Table 18). The equivalent statistic for P2’s control set showed a negligible gain. Moreover, P2’s therapy sets showed a large SMD effect size relative to his control set. The value of the OR between P2’s therapy and control sets at post-96  therapy indicated the odds of one of P2’s words being acquired post-therapy were roughly 22 times higher for his therapy sets than control set. Moreover, the odds of an accurate trial occurring in P2’s practice set were ~3.5 times higher than among unpracticed words in the monitor set. However, P2 also showed noteworthy loss of retention, with the odds of acquisition being ~1.75 times lower after therapy than during therapy. P3 had the opportunity to practice all the words in his therapy sets over five staggerings, except for one word that was excluded for being acquired without therapy (Table 17). However, P3 was able to acquire less than half of his therapy set words during therapy, and only eight (20%) of P3’s total therapy set words remained acquired at post-therapy. Only one of these words was a PN word. P3 showed a medium naming accuracy gain from pre- to post-therapy (Table 18). P3’s control set showed negligible change from pre- to post-therapy, and his therapy sets showed a medium SMD effect size relative to his control set. The odds of a word being acquired post-therapy were similarly roughly 11 times higher for his therapy sets than control set. Furthermore, the odds of an accurate trial occurring during therapy were roughly four times higher for words in P3’s practice set than unpracticed words in his monitor set. The odds of a word being acquired were roughly three times lower at post-therapy than during therapy.97   Figure 5. P1’s percentage of the therapy sets assigned to the practice set and accurate trials per trial series for therapy and control sets. 4" 9" 0" 4" 0"65" 61"48" 48" 78"57"70"57" 52"43"52" 48"39"30"52"43" 43" 48"61"52" 52" 48"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22" 23" 24" 25" 26" 27"%"of"therapy"set"1"Baseline"phase" Therapy"phase" Post4therapy"phase"0" 0" 0"7"0"13"7"0"7"20"27"33" 33"20"33" 33" 33" 33" 33"53"47"27"47"60"47" 47"40"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22" 23" 24" 25" 26" 27"%"of"therapy"set"2"4" 0" 4" 4" 0"8" 8" 8" 4"17" 13" 17" 17"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22" 23" 24" 25" 26" 27"%"of"control"set"Probe"picture4naming"trial"series"Naming"accuracy"Prac7ce"set"98   Figure 6. P2’s percentage of the therapy sets assigned to the practice set and accurate trials per trial series for therapy and control sets. 5" 0" 0" 0" 0" 5" 0" 0"48"57"48"71" 76" 71" 71"62" 67"71" 67"76"62"76"62" 62"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22" 23" 24" 25" 26"%"of"therapy"set"1"Baseline"phase" """"""Therapy"phase" Post4therapy"phase"5" 5" 0" 0" 0"9"0" 0" 5"9" 9"18"5"14" 14" 18"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22" 23" 24" 25" 26"%"of"control"set"Probe"picture4naming"trial"series"Naming"accuracy"Prac7ce"set"0" 0" 0" 0"6" 6"0" 0"6"13"6"0"25"44"81"75"100"88"94"75"94"88"63"81"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22" 23" 24" 25" 26"%"of"therapy"set"2"99    Figure 7. P3’s percentage of the therapy sets assigned to the practice set and accurate trials per trial series for therapy and control sets. 15" 10"0" 0"60" 60"45"65"50" 45"60" 55" 55"65"55"45"55"65"50" 50" 50"60"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22"%"of"revised"therapy"set"1" Baseline"phase" Therapy"phase" Post4therapy"phase"15" 10"0" 5"10"25" 20" 20"10" 10"20" 15"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22"%"of"control"set"Probe"picture4naming"trial"series"Naming"accuracy"Prac7ce"set"11" 11"0" 0"16"32"63" 63"42" 37"47"58"47"74"37"58"47"68" 63" 63"37"58"0"20"40"60"80"100"1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14" 15" 16" 17" 18" 19" 20" 21" 22"%"of"revised"therapy"set"2"100  3.2.2 Secondary outcomes Table 19. Overview of secondary outcome variables and data sources. Variables Data sources Pre-to-post changes in: 1. Language impairment 2. Technology use 3. Communication activity/participation 4. Confidence/well-being 1. WAB-R 2. BNT-2 3. Cookie Theft 4. Participant-perspective interview and check-ins 5. Researcher session notes Note. WAB-R = Western Aphasia Battery–Revised, BNT-2 = Boston Naming Test–Second Edition.  No clear patterns of change were apparent on the BNT-2, WAB-R, or Cookie Theft Picture for any participants (Table 20). Post-therapy participant-perspective interviews, with reference to qualitative data from other time points where relevant, documented participants’ reported perceptions of the effects of therapy (Appendix P). P1 reported that therapy helped improve his picture naming for the words that he practiced and reported feeling that therapy may have also improved some of the words he did not practice. P1 also reported using his practiced words more often in everyday life, but he could not provide a concrete example of when this had happened. He noted that the gains he made were inconsistent in that he reported being able to say a practiced word at one moment, only to forget it later on. He was confident he would not lose the gains he made. P1 reported no changes in his perceptions or use of mobile technology as a result of the study, since he continued to have very positive impressions and high use of technology in his everyday life. P1 reported that self-directed MTT helped improve his self-confidence a little at the beginning, but this effect was lost once he started struggling more and became frustrated.   101  Table 20. Pre and post scores for secondary impairment-level outcome measures.   P1 P2 P3  Max Pre Post Pre Post Pre Post BNT-2 60 8 13 40 45 25 27 WAB-R        Spontaneous speech 20 11 11 15 16 11 11 Auditory comprehension 10 7.95 7.55 7.95 8.45 8.4 8.15 Repetition 10 2.7 2.6 7.9 8.6 9 9.2 Naming & word finding 10 3.6 3.6 8.0 8.0 6.3 6.9 Reading 20 8.4 7.8 16.4 17.6 11.2 14.00 Writing 20 5.0 5.9 16.0 12.2 9.8 9.2 Aphasia Quotient 100 50.5 49.5 77.7 82.1 69.4 70.5 Language Quotient 100 46.6 46 79.2 79.3 64.1 66.6 Aphasia type  Broca’s Anomic Transcortical motor Cookie Theft        Subclausal utterances (%)  18 33 8 7 27 44 Agrammatic deletions (%)  0 0 21 7 0 6 Complexity index  0.0 0.0 1.1 1.1 0.0 0.4 Words/min  77.0 34.9 53.1 54.7 6.0 34.7 CIUs (%)  9 9 37 54 3 19 CIUs/min  7.00 3.08 19.77 29.53 0.15 6.43 AC concepts  0 0 5 3 0 0 AI + IN concepts  1 2 1 1 0 4 AB concepts  6 5 0 3 7 3 Note. BNT-2 = Boston Naming Test–Second Edition, WAB-R = Western Aphasia Battery–Revised, complexity index = mean number of clauses per utterance, CIUs = correct information units (i.e., words that are accurate, relevant, and informative relative to the Cookie Theft Picture), main concepts = presence, completeness, and accuracy of speech content without considering its form, AC = accurate/complete, AI = accurate/incomplete, IN = inaccurate, AB = absent.  P1 reported feeling very frustrated about not making as much progress in therapy as he wanted (Appendix P), and he demonstrated these signs of frustration very early during therapy. P1 tended to react negatively whenever he experienced struggle, resulting in the audio recordings of his self-directed therapy sessions being dotted with swearing, frequent angry exclamations, deep sighs, and other verbal behaviour suggestive of frustration. However, he tended to become particularly frustrated when he perceived his performance as being worse than what it had been 102  previously. As such, during the therapy week 2 probe session, P1 became so frustrated by an apparent dip in his performance from the previous week’s probe session that he protested at length with the researcher about not being able to use cues, indicating that he felt probe picture naming unfairly masked his potential. P1’s frustration may have negatively affected his naming accuracy. In the last three weeks of therapy, an initial dip in performance during therapy week 5, which P1 attributed to a loss of confidence resulting from deficit testing he reported recently having undergone outside of the study, preceded a spiraling cycle of greater frustration and poorer performance. This situation culminated in P1 becoming so frustrated and disheartened by not meeting his expectations for himself that he chose to discontinue therapy at the end of week 7, despite being eligible for further practice. During the post-therapy interview one month later, P1 emphasized that he still felt positively about NT and self-directed practice, but simply needed a break from it before returning to it to practice more. He blamed and was frustrated with himself, rather than NT or the iPad, for what he perceived as his lack of success. P2 reported that his ability to name the particular words he practiced had improved as a result of therapy and that he was using his practiced words more often in everyday life, but he could not provide a concrete example of when this had happened. He noted that the gains he made were inconsistent across time, but he was nonetheless confident he would not lose the gains he made. He reported also noticing small improvements in his reading and writing that he attributed to the effects of NT. While P2 reported no changes in his self-confidence as a result of therapy, he repeatedly discussed at length changes in how he used and felt about iPads as a result of the study, and was able to provide specific concrete examples to demonstrate (Appendix P). He reported changes both in what he used his iPad for and in how frequently he used it. However, the examples he gave of the new ways he used the iPad did not notably differ from the 103  accessing information and entertainment functions P2 reported in the pre-therapy interview. P2 reported using his iPad once a month before the study, and explained that he had essentially given up on the iPad years previously, but was now using it once or twice per day. However, the pre-therapy frequency of use he reported at post-therapy contradicted his earlier report during the pre-therapy interview, and it was unclear how long this period of nonuse had lasted. He reported that he had almost completely stopped using his laptop now that he was better at using his iPad and found it easy. Unlike during the pre-therapy interview, P2 reported finding his laptop difficult to use and that he would become frustrated when trying to use it to access information. P2 was confident that his new enthusiasm for his iPad and his increased use of it would persist into the future. Although P2 did not specifically report any negative effect of self-directed MTT, he did report feeling disappointed and frustrated that he still could not name all of the words he practiced during therapy, and reported further disappointment after the first probe session after therapy when he realized he still could not name the control set words.  P3 felt that self-directed MTT helped him make progress in terms of improving his picture naming for the particular words practiced. He also reported feeling that therapy had possibly improved some other words he did not practice; however, most of the benefits were restricted to practiced words. He was confident he would not lose the gains he made. P3 reported no changes in his perceptions or use of mobile technology or computers as a result of the study, since he continued to have very positive impressions and high use of technology in his everyday life. Although P3 similarly reported no change in his self-confidence level, he did report using the practiced words more often in everyday life; however, he could not provide a concrete example of when this had happened. P3 acknowledged that he was disappointed whenever he did 104  not do as well as he wanted to during probe picture naming, as his performance during self-directed practice led him to expect to perform perfectly during the probe sessions. 3.2.3 Summary of findings for the therapeutic effect analysis In order to evaluate the therapeutic effect of minimally-supervised self-directed MTT, this study used a single-subject experiment to examine the effect of therapy on participants’ picture naming ability for practiced and unpracticed words as well as changes in their performance for secondary outcomes. All participants were able to practice the majority of their therapy sets during the therapy phase; however, each also had a small number of words that were excluded for being acquired without ever being practiced. Participants demonstrated medium to large naming accuracy gains for their therapy sets from pre- to post-therapy, but negligible to small gains for their control sets. Moreover, the gains were medium to large (or, in terms of ORs, 7-22 times higher) relative to the gains for the control sets. During the therapy phase, there were 4-5 opportunities to stagger the introduction and withdrawal of therapy per participant in order to evaluate the likelihood that confounding variables could account for the results. Comparing the practice set words and unpracticed monitor set words during these staggerings showed that the odds of a trial being accurate were 4-5 times higher for the practice set. However, participants also demonstrated poor retention, with the odds of acquisition becoming 2-3 times lower from the therapy to post-therapy phase. This reflected the fact that participants acquired roughly 30-60% of their therapy sets during therapy, but only 20-45% of them remained acquired after therapy. Only a minority of these acquired words were PN words. In line with these findings, participants reported picture-naming gains for the words they practiced and also possible improvements for other words or behaviours. Regarding secondary outcomes, there were no clear changes on language measures for any of the participants; however, they all reported that 105  the gains they had made on picture naming generalized to their everyday use. P1 also reported a short-lived boost in his self-confidence and P2 reported improved technology use, although there were inconsistencies in his reporting. P1 reported and demonstrated substantial frustration during self-directed therapy and the others reported disappointment about their less-than-desired progress. 3.3 Therapy process Table 21. Overview of therapeutic procedure variables and data sources. Variables Data sources 1. Proportion of use of each cue 2. Participants’ reported therapy process 1. NT data logging system 2. Participant-perspective interview and check-ins Note. NT = Naming Therapy©.  As participants’ dosage patterns have already been reported elsewhere (see section 3.1.2), only their enactment of therapeutic procedures, specifically cueing and performance-contingency, are explored here. P1 primarily used only two of NT’s optional cues during self-directed practice (Figure 8; see Appendix G for clarification of NT’s cues): approximately 91% of the cues he selected were phonological, roughly equally split between the First Sound and Spoken Word cues. Of the remaining cues, the Written Word cue was most dominant; however, it still only represented ~4% of the data. The cueing preferences P1 reported during the post-therapy interview provided additional information about his cueing techniques and the intentions behind his use of cues. He reported preferring the First Sound and First Letter cues the most because he found them the most helpful for naming the target word. He reported also liking them because they only gave him a hint, requiring him to come up with the rest of the word by himself. P1 reported disliking the Spoken Word and Written Word cues, feeling they were too easy because they gave him the answer, so he reported using them only as a last resort when the 106  other cues failed. Consistent with his limited use of them, P1 reported particularly disliking the semantic Definition and Phrase Completion cues because he felt they did not help him name the word.  Figure 8. Distribution of cues used by each participant during self-directed practice. P2 had a strong tendency to use the Spoken Word cue (~71%) during self-directed practice, followed by the Written Word cue in a distant second place (~12%; Figure 8), with the remaining ~17% of cue use equally split across the remaining cues. During the post-therapy interview, P2 reported finding most of the cues helpful to some extent, but that the Written Word cue was his go-to cue because he found it helped but not too much. He reported not liking the Spoken Word cue very much because he considered using it to be giving away the answer. As such, he reported preferring to use it only early on in the therapy phase of a new word when he needed a lot of help, and then to move to using the Written Word cue as soon as possible to increase the difficulty level. P2 also reported disliking the Definition and Phrase Completion cues because he felt they did not provide him new information and because they were tediously long and wordy. 0"10"20"30"40"50"60"70"80"Defini0on" First"le7er" Wri7en"word" Phrase"comple0on"First"sound" Spoken"word"Percentage)of)cue)use)Naming)Therapy)cues)P1"P2"P3"107  P3 almost exclusively used the orthographic First Letter and Written Word cues during self-directed practice, which together accounted for ~93% of his cue selections (Figure 8). P3 reported preferring the First Letter cue the most because he found it the most helpful. He was uncertain about why he preferred it relative to the Written Word cue. However, he explained that he prefers orthography in general because he finds things easier to say when they are written, which was why he reported not liking the First Sound or Spoken Word cues. He also reported disliking the semantic cues because he found them unhelpful. P3 was unable to provide a clear explanation of his preferred cueing techniques even with partner-supported AAC strategies. In summary, all participants showed a strong preference for word-form cueing types (phonological or orthographic) over semantic cues. Regarding cueing techniques, P1 reported preferring to increase the level of cueing support within trials, whereas P2 reported preferring to decrease the level of cueing support across trials. In terms of performance-contingency, all participants reported that their decisions to use certain cues over others were based on the probability of success they perceived they had when using those cues in the past. 3.4 Summary of results This study explored the feasibility, therapeutic effect, and therapy process of minimally-supervised self-directed MTT. Regarding feasibility, developing word sets for participants was found to be time-consuming. Different nomination methods were found to be more or less efficient at different stages of the development process; however, none of them were sufficient in isolation to develop the full set of words used in the study. Despite this process, some participants expressed dissatisfaction with the usefulness of the words they had practiced. Adherence during self-directed practice was generally high, but varied across participants, with P1 and P3 highly adherent and P2 less so for certain aspects of the study protocol. Practice-108  related aspects of NT were found to be accessible for participants, but NT’s system for managing word sets was found to be cumbersome for the researcher and NT’s self-scoring system was found to have only partial value for remote progress monitoring. Participants found self-directed MTT using NT acceptable, requiring little time commitment from them to complete, but they also expressed dissatisfaction with the tedium and challenge level of NT. Regarding the therapeutic effect of therapy, participants showed medium to large naming accuracy gains for therapy word sets, but minimal gains for unpracticed monitor and control word sets. Participants also showed poor retention of their gains, with only a modest portion of the words they practiced remaining acquired after therapy. Regarding therapy process, participants showed varied dosage and therapeutic procedure processes. P1 and P3 chose to follow a methodical dosage routine, whereas P2 practiced variably until the last two weeks of therapy when he dramatically increased his dose and frequency of practice. Participants used different cueing types and techniques; however, they all demonstrated and reported preferring to use word-form cues and their self-reported processes reflected sensitivity to their own past performance.  109  Chapter 4: Discussion The purpose of this study was to explore the potential of self-directed MTT with minimal therapist supervision. To this end, the study examined the feasibility, therapeutic effects, and therapy process of such a therapy when participants practiced using NT (Table 2 is reproduced below in Table 22 for the reader’s convenience). Questions about the feasibility of this therapy were explored in terms of the efficiency and functional relevance of word set development, participants’ adherence to a study protocol, the accessibility of NT both for participants and for the researcher, the potential to use NT for remote therapy progress monitoring, and the acceptability of this form of therapy to people with aphasia. Questions about the therapeutic effect of the therapy were addressed using a single-subject multiple baseline design to evaluate its effects on naming accuracy and acquisition as well as on secondary language and functional domains. Finally, questions about the process of self-directed therapy were explored in terms of both the quantity of practice and its quality. In this chapter, a summary of findings relevant to each of the research questions is first presented, along with interpretation of the findings in light of the study’s other findings, theory, other studies, and clinical practice. Suggestions for future research, future app development, and clinical practice are then discussed as part of the broader goal of optimizing this form of therapy for people with aphasia. Finally, the limitations of this study are discussed and suggestions for improvement in future research are presented. Table 22. Overview of research questions and their subcomponents. Feasibility Therapeutic effect Therapy process 1. Word set development 2. Adherence 3. Accessibility 4. Remote progress monitoring 5. Acceptability 1. Naming accuracy 2. Secondary outcomes 1. Therapy dosage 2. Therapeutic procedures (cueing and performance-contingency) 110   4.1 Summary and interpretation of findings 4.1.1 Feasibility 4.1.1.1 Word set development 4.1.1.1.1 Efficiency of nomination methods The process of developing word sets, as enacted in this study, was highly time-consuming both for participants and for the researcher. However, the efficiency of different nomination methods varied between the different stages of the development process. While the NT sets (i.e., the words native to NT) required the least amount of time and effort to collect and develop, they had the highest attrition rate, producing the fewest eligible study set words. Conversely, the PN sets (i.e., the words participants nominated themselves) required the most time to collect and develop, but generally had the lowest attrition rate. The BOSS set (i.e., words from the BOSS database) was presented only to P2 and demonstrated an intermediate level of efficiency in terms of both development of materials and attrition rate. The relative inefficiency of the BOSS and PN sets when developing materials related primarily to the time demands of using participant-nomination strategies and of developing novel cues. However, some strategies and cue types were still more efficient than others. Almost all PN words were elicited via supported checklist or diary methods for all participants, with very few identified through spontaneous nomination. Among cue types, semantic cues were the least efficient to develop, especially for P2. Although results varied across participants, in general, the higher attrition rates of the NT and BOSS sets were due to more words from these sets being inadequately difficult to name or being perceived as unuseful/irrelevant. 111  Although no studies in the aphasia therapy literature appear to have empirically evaluated how best to identify therapy words, the findings of this study are consistent with the speculations of Renvall et al.’s (2013a; 2013b) recent literature review on the topic. Namely, participant nomination was far more time-consuming than intuition-based nomination (represented by the PN and NT sets, respectively) when developing materials; however, in general, participant nomination proved to be a much more efficient way of identifying appropriate therapy targets that were also perceived as useful by participants. Regarding the merits of different participant-nomination strategies, the findings supported Renvall et al.’s (2013a) hypothesis about the limited value of spontaneous nomination in light of the language deficits inherent to aphasia.  Variations in the efficiency of nomination methods across participants may have in part related to participant characteristics and study methodology. The fact that P1 required only one NT set to identify a sufficient number of therapy words, P3 required two, and P2 required all three of the NT and BOSS sets was generally consistent with their aphasia/anomia severity on the WAB-R/BNT-2. Although supported strategies were able to facilitate participants’ nominations, none of the methods were able to elicit sufficient numbers of therapy words on their own, with the fewest elicited from P1 and the most from P2. In line with Renvall et al. (2013a), this demonstrates the challenges of participant nomination in this population and may have been influenced by differences in aphasia severity and/or the availability of informants across participants. The efficiency advantage for PN sets was absent only for P3, who showed roughly equivalent low-efficiency for both NT and PN sets. It is possible that this finding related to the fact that P3 chose to use different participant-nomination strategies than the other participants. This would be inconsistent with expectations, given that diary methods are considered relatively more objective ways of identifying candidate words (Renvall et al., 2013a). Alternatively, this 112  finding may have related to the apparent discrepancy between P3’s picture naming and his spontaneous speech in everyday life. While P3 had similarly impaired spontaneous speech output to P1, he showed notably better picture naming on both the WAB-R and the BNT-2. Thus, as P3 and his spouse likely based their nomination decisions on P3’s relatively poorer spontaneous speech, they may have been misled to nominate words that were insufficiently difficult for him in picture naming. Only P1 eliminated a substantial number of words at the selection stage, citing their irrelevance to his life, whereas P2 and P3 eliminated almost no words at the same stage. Their stated reasons for not doing so related to their awareness that they were nearing the study’s 60-word minimum limit. It is unclear whether they would have been more discriminating at this stage if the study design had allowed them more leeway. 4.1.1.1.2 Functional relevance of nominated words Despite time-consuming attempts to make the word set development process inclusive, it became clear during the post-therapy interviews that P1 and P3 felt that some of the words they practiced were not actually useful to them. This was in spite of their PN words’ intuitive functional relevance, being related to their everyday personal lives. Moreover, it was in spite of the fact that all of the words participants practiced, regardless of which nomination method had been used to identify them, were ones they had personally chosen to practice during the selection stage. In contrast, P2, whose PN words were less intuitively functionally relevant, in that they were limited to specialized jargon and strangers’ names, was satisfied at post-therapy with the words he had practiced. This pattern of findings may have related to participant characteristics, participants’ chosen therapy process, and/or the composition of their study word sets. P1 and P3 had more severe naming deficits than P2, and they practiced more frequently and at a higher cumulative 113  dose than P2. Thus, greater difficulty naming words and/or the greater investment P1 and P3 put into therapy may have led them to be more discerning at post-therapy regarding the practical value of what they had gained. Although P2’s nominations may not immediately appear functionally relevant in general, they could be relevant specifically for him. Evidence has suggested that people with aphasia often wish to be able to express opinions and ideas, but frequently have difficulty doing so (see Renvall et al., 2013a, for review). Thus, to P2, given his special interest in politics and current events, difficulty expressing specific opinions clearly and efficiently may be a particularly salient way in which his anomia impacts on his everyday life. In fact it could represent one of his few unaddressed daily communication needs, given his mild aphasia (Garrett & Lasker, 2013). Moreover, as Renvall et al. (2013a) pointed out, people with aphasia may often wish to practice proper names because of the unique emotional and relational value they carry in interaction. P2’s choice to nominate names and terms related to his interests could thus represent his valid attempt to reduce the degree of his disability and reveal his competence as an informed and intelligent individual when interacting with others. In this light, it would be unsurprising that P2 voiced no concerns about the value of his therapy set words post-therapy, since the majority of them came from his PN set. In contrast, the majority of P1 and P3’s came from the NT sets. As such, the perceived low usefulness of some of the words they practiced may support Renvall et al.’s (2013a) contention that therapist intuition-based nomination is unlikely to produce functionally relevant therapy targets. In summary, findings from the process of developing word sets for this study indicated a number of issues in need of improvement that can be evaluated in future research and that may inform therapists’ decisions regarding the use of NT and similar apps for particular clients. Specifically, the results suggested that the words included in NT by default, although convenient 114  in terms of developing therapy materials, are likely insufficiently difficult for people with mild aphasia/anomia and are possibly of limited functional relevance even to people with moderate impairments. To use NT with clients like this, it would likely be necessary to supplement the words native to NT with words personally nominated by participants themselves. However, participant nomination entailed time-consuming collection and development of therapy materials and, just like the words native to NT, using them did not necessarily ensure that the words clients selected would benefit their everyday communication. Moreover, even when all this work did result in a set of possibly functionally relevant therapy words for the mildest participant in this study (P2), he still found NT to be insufficiently challenging. Thus, NT in its current form and as enacted in this study may simply not be appropriate for clients with clinical profiles similar to P2’s. Despite these somber implications, the results of this study and past research also suggested how these issues might be addressed in future research, in future therapy app development, and in clinical practice. Specifically, results suggested that having participants nominate words from pre-made checklists may be an efficient way of facilitating participant nomination and of reducing the time demands involved in developing therapy materials. Given that cue development also contributed notably to these time demands, methods of automating aspects of the process, perhaps like using text-to-speech software (Beukelman & Mirenda, 2013) for phonological cues, could help increase the time efficiency of participant nomination in future therapy apps. In light of the negative findings for semantic cues in this study (see section 4.1.3.2), it may also be possible for therapists to greatly increase the efficiency of the cue development process by simply forgoing semantic cues for participant-nominated words without compromising therapy effects. 115  Since typical intuitions about the commonness of different words often do not align with reality (Renvall et al., 2013a), researchers and therapists might obtain more positive results than the current study regarding the functional relevance of therapy words if nominated words are collected using frequency- intuition-based nomination. The development of future therapy apps may similarly benefit if their native words come from such sources. To this end, Renvall et al. (2013b) have developed and published word lists of high-frequency words derived from objective data for the population in general and for older adults in particular that may be relevant to people with aphasia. Therapists may supplement the words native to NT by using Renvall et al.’s (2013b) lists as checklists to increase the functional relevance of therapy targets. Given that many abstract nouns, verbs, adjectives, and adverbs are represented in these lists, therapy methods to address these targets are needed (Renvall et al., 2013a). The multimedia capabilities of mobile technology may make it particularly well suited to targeting these less concrete words, so future research and therapy app development might consider tapping into this potential for multimodal stimulation through use of not only traditional picture stimuli, but also video and audio stimuli (cf. Lavoie et al., 2016; Routhier et al., 2016). However, although frequency-based nomination might be more likely to represent functionally relevant words than intuition-based nomination, results suggested that they may be similarly insufficiently difficult for people with mild aphasia/anomia, at least in the context of picture naming. Since the BOSS set for P2 consisted of very low-frequency words, but still had a relatively high attrition rate, it seems likely that the attrition rate for Renvall et al.’s (2013b) lists would be even higher. It may be that only participant nomination is an appropriate method of identifying picture-naming therapy targets for people with mild aphasia/anomia, perhaps because frequency-based nomination is too broad-based for the relatively specific communication needs 116  that people with mild impairments are likely to have (Garrett & Lasker, 2013). The Internet checklists used in this study were intended to identify these kinds of idiosyncratic words, specifically in terms of special interests; however, they were informal and unvalidated. Systematic approaches like data-mining the Internet for topic-specific vocabulary, as preliminary work in the AAC literature has begun to do (Luo, Higginbotham, & Lesher, 2007; cited in Beukelman & Mirenda, 2013), may be more likely to identify words representative of particular special interests. Communication diaries are likely an additional strategy for identifying idiosyncratic words more generally (Renvall et al., 2013a; Beukelman & Mirenda, 2013); however, research is needed to examine how they can be made accessible for people with aphasia, even when they do not have a caregiver to support them. Ultimately, the ability of various nomination methods to predict the eventual functional relevance of therapy targets is an empirical question that should be evaluated in future research that directly compares the clinical significance of therapy outcomes obtained using different methods. Until that time, the results of this and past research (Renvall et al., 2013a) suggest that therapists should identify potential functionally relevant words using multiple nomination methods and should confirm their appropriateness through periodic re-evaluation on an ongoing basis. 4.1.1.2 Adherence Participants were adherent to most aspects of the study protocol, with a few exceptions for P2. P1 and P3 consistently met or exceeded expectations in terms of probe session attendance, dosage, and self-scoring, whereas P2 showed lower adherence for these aspects of the protocol. He missed one out of his six scheduled probe sessions during the therapy phase, he only started to meet (and exceed) the prescribed dosage schedule during the last two weeks of 117  therapy, and he rarely scored the accuracy of his naming attempts during therapy sessions, but did so consistently during probe sessions. These findings are consistent with recent studies on self-directed CT or MTT indicating that participants are usually, but not always, adherent to the therapy regimen, and will even sometimes practice much more than they were asked to (e.g., Kiran et al., 2014; Mason et al., 2011; Mortley et al., 2004; Palmer et al., 2012). Also in line with past research (e.g., Palmer et al., 2012; 2013; Wade et al., 2003), several pieces of evidence in the current study suggested that the researcher’s supervision through periodic probe sessions with participants influenced their adherence. At least one participant (P3) reported during post-therapy interviews that the researcher’s involvement contributed to motivation to continue practicing. Furthermore, the fact that P2 primarily only self-scored when in the presence of the researcher, but not when he was on his own, could indicate an effect of supervision. Finally, P2’s chosen dosage schedule appeared to be associated with the occurrence of probe sessions and the ending of the therapy phase. Specifically, for much of the therapy phase, P2 practiced very little; however, some of his highest doses occurred within hours before a probe session with the researcher and the bulk of his practice occurred shortly before the end of the therapy phase. This may suggest that P2 was motivated to practice by a feeling of obligation to the researcher – also perhaps reflected in his comments during the post-therapy interview – and an awareness that his opportunity to practice was soon coming to an end. Patterns of practice like this have been documented previously in people with aphasia during self-directed practice and have been explained similarly (Petheram, 1996). In summary, the current study and past research suggest that people with aphasia often carry out therapy as instructed, at least when they are minimally supervised in terms of technical support and progress monitoring. Although observational data exist to suggest that this 118  supervision may contribute to continued practice, supervision cannot be considered necessary to adherence or therapy outcomes because no studies have experimentally manipulated supervision (Zheng et al., 2016). Moreover, strict adherence did not clearly relate to outcomes for any aspect of the study protocol in this study, as P2, who showed the poorest adherence, also showed the largest naming accuracy gains. Given that no particular aspect of the aphasia therapy process has been demonstrated to be necessary (see sections 1.4.1 and 1.5 for elaboration), future research might consider manipulating and reporting more refined measures of the therapy process, rather than only arbitrarily-defined dichotomous adherence variables, so that it would be possible to evaluate which aspects of the therapy protocol are necessary to therapy outcomes (and thus need to be strictly adhered to) and which aspects can be permitted to vary. 4.1.1.3 Accessibility Overall, all participants were able to navigate NT during self-directed practice in ways highly consistent with the navigational path they were trained and instructed to follow, and were able to do so very quickly. Variations in the navigational path taken were restricted primarily to self-scoring and ending the app, and those that interfered with carrying out therapy practice were rare overall for all participants. Moreover, all participants reported finding NT easy to use and that no design feature interfered with their completing the study protocol independently. In contrast, NT’s system for managing word sets, which was the responsibility of the researcher, was perceived as unwieldy, due to the need for extensive scanning and manual reconstruction of word sets whenever it was necessary to change which words were presented to participants. These issues resulted in lengthy delays for participants during probe sessions, and contributed to the experimental design for P3 becoming compromised. 119  These findings demonstrate that an aphasia therapy app ostensibly designed with many of the accessibility features reviewed by Brandenburg et al. (2013) can in fact be accessible for people with aphasia. This study appears to be the first to objectively evaluate the accessibility of mobile technology in aphasia therapy through in-vivo data collection. Most past research on mobile or computer technology has not examined accessibility, and the little that has done so primarily used self-report methods, in some cases finding that, unlike the current study, the therapy app evaluated was not accessible for all participants, instead requiring significant caregiver support (Brandenburg et al., 2013; Palmer et al., 2013; Wade et al., 2003). It appears that only one other study (Allen, McGrenere, & Purves, 2008), examining a mobile technology-based AAC app for aphasia, has supplemented qualitative interview data with quantitative data of participants’ actual device usage in everyday settings, as the current study did. Despite large technological discrepancies between Allen et al. (2008) and the current study, they similarly found that participants were able to use mobile technology independently and did so for most days of the study’s duration. It is also possible that participants in the current study showed such high performance during self-directed practice in part because of the inclusion of an iPad proficiency screening/training session prior to the start of therapy. However, as it was only one session, and participants were already performing well without instruction, their success may have related more to the fact that they were all experienced current or past technology users. It may be that participants would have adapted fairly easily to using NT independently even if they had not received this session. Thus, although past self-directed CT or MTT studies have tended to provide all participants with technology training by default, the findings of this study suggested that such training may not always be necessary for people with aphasia who have similar characteristics to 120  this study’s participants. This would likely increase the feasibility of completely unsupervised self-directed practice. However, this possibility does not necessarily extend to people with different characteristics (e.g., people with more complex motor, language, cognitive, or perceptual issues, or less technology experience) and/or to apps with higher operational demands. Given these uncertainties, it may be most profitable for therapists to assess clients’ technology proficiency on a case-by-case basis by using dynamic methods similar to this study. Therapists could then use this information to evaluate individual clients’ candidacy for self-directed MTT and whether and how much additional technology training may be necessary. Future research might attempt to clarify these uncertainties by systematically evaluating the interaction between participant characteristics, the operational demands of technology, and the provision of technology training, in order to determine which design and training factors affect the accessibility of self-directed MTT for different people with aphasia. In contrast, the findings of this study regarding NT’s administrative functions suggested that NT in its current form may be insufficiently accessible for most applications. NT’s word management system was perceived to be unwieldy even for the researcher, a person without any of the motor, sensory perceptual, linguistic, or cognitive issues commonly experienced by people with aphasia. Thus, although the study did not evaluate participants’ potential to use these functions, the findings suggested that it would likely have been very difficult, perhaps even after thorough training, for participants to have independently managed their word sets using NT. As a result, it might not be advisable for people with aphasia to take on administrative roles when using NT in its current form, as they might do in completely unsupervised self-directed practice. Similarly, NT may be too cumbersome for both research and clinical practice, given the time demands and propensity for human error entailed by its current word management system. 121  Future app development could attempt to improve the accessibility and clinical utility of word management functions by automating systems and reducing the need for scanning, such as by allowing the user to create customized word sets that can be easily created, selected, and deselected. Accessibility and utility could be further improved by including functionality to allow word management (and progress monitoring; see section 4.1.1.4) to be conducted remotely over the Internet. This would increase the accessibility of minimally-supervised self-directed practice for clients living in remote locations and would reduce the time demands of supervision for the therapist even further (Mortley et al., 2004). 4.1.1.4 Remote progress monitoring All participants tended to be more likely to mistakenly score inaccurate attempts as being correct than they were to do the reverse during probe sessions. In other words, they were more likely to miss their errors than to raise a false alarm. This resulted in the self-scoring system demonstrating only partial predictive value: when participants selected the Incorrect Button, it was extremely likely to represent an inaccurate attempt, but when they selected the Correct Button, it was unclear what their naming accuracy had been. This study only evaluated participants’ naming accuracy during probe picture naming, so it was not possible to examine participants’ self-scoring categorization accuracy under the more ecologically valid conditions of self-directed practice. Nonetheless, both P2 and P3 reported finding self-scoring tedious and the fact that P2 usually did not self-score during self-directed practice in the first place meant that it would not have been possible to evaluate his categorization accuracy for the majority of his therapy trials anyways.  In summary, the available evidence regarding the utility of NT’s optional self-scoring system for remote progress monitoring is not promising, as participants were not able to score 122  themselves with consistent accuracy, and some of them disliked and/or were not willing to do so during self-directed practice. As such, therapists hoping to use NT for minimally-supervised self-directed practice may have no reliable, valid, and acceptable way of monitoring clients’ progress through NT in order to modify therapy accordingly (e.g., swapping acquired words with unpracticed words). On the other hand, given that NT in its current form already necessitates the use of periodic in-person probe sessions due to its inability to facilitate remote word set management (see section 4.1.1.3), therapists can likely gain some impression of clients’ progress during self-directed practice based solely on their performance during these probe sessions. Indeed, participants’ probe session performance was used in exactly this way during the current study. The modified version of NT used in this study also collected data regarding participants’ use of cues and audio recordings of their therapy trials during self-directed practice. While analyzing detailed usage data in clinical practice would be highly impractical, future research on therapy process-outcome relationships (see section 4.1.3.2) might be able to identify summary statistics for use in therapy apps that could usefully inform therapists’ decisions about modifying therapy. Moreover, if audio recordings of participants’ self-directed practice were available to therapists, therapists might be able to evaluate clients’ therapy process and progress by analyzing a sample of them, similar to the technique used by Mortley et al. (2004). This could inform their management decisions while decreasing the time demands of analyzing audio recordings of self-directed practice. As such, future development of apps targeting verbal production might consider using automatic audio recording of therapy trials to facilitate remote progress monitoring, rather than using unreliable surrogate markers of progress like NT’s optional self-scoring system. 123  4.1.1.5 Acceptability All participants considered self-directed MTT with minimal supervision to be an acceptable form of therapy, either on its own or in combination with SLP-based therapy. Moreover, they were all able to integrate the therapy into their everyday routines, practicing at various times during the day, including late at night when SLP-based therapy would likely be unavailable (Petheram, 1996). The study protocol was found to require very little time commitment to complete. However, participants also experienced dissatisfaction with the number and quality of the words they practiced, the monotony or tedium of practice, and/or the challenge level of NT. These findings add to previous CT and MTT studies indicating that people with aphasia can find technology-based therapy an acceptable alternative to SLP-based therapy (e.g., Palmer et al., 2013; Routhier et al., 2016; Wade et al., 2003). Moreover, participants highlighted many of the same themes reported by participants in past studies regarding the relative merits of self-directed MTT and SLP-based therapy, including valuing the scheduling flexibility, opportunity for repetition, and independence facilitated by self-directed MTT as well as the social interaction provided by SLP-based therapy. P2’s comments regarding the greater tailoring of therapy possible in SLP-based therapy are consistent with concerns about the disadvantages of MTT highlighted in the literature (e.g., Brandenburg et al., 2013; Katz, 2010), and indeed, his dissatisfaction with the challenge level of NT may have in part motivated these comments and his preference for SLP-based therapy. Likewise, P1’s dissatisfaction with the power relations between therapists and clients in SLP-based therapy, which motivated his preference for self-directed practice, was reminiscent of the concerns about authoritarian and paternalistic therapist behaviour raised in the literature (e.g., Horton, 2007; Simmons-Mackie & Damico, 1999). In 124  light of these issues, it is perhaps interesting that the two participants with more severe aphasia preferred self-directed home practice. It may be P1 and P3 preferred self-directed MTT because NT was at a more appropriate difficulty level for them and/or because the severity of their aphasia makes it harder to exert their autonomy in interaction with therapists whereas self-directed practice helps facilitate autonomy. Participants may have had such positive appraisals of the therapy and may have been able to accommodate it so easily, because of the very low time demands involved, how easy it was for them to use NT, the positive therapy effects they experienced, and/or the fact that all of them already had positive technology perceptions/experience before therapy. In summary, the current study and past research suggest that self-directed MTT may be generally tolerable to people with aphasia, providing one piece of the evidence (also see section 4.1.2) for the evaluation of its safety – a minimum requirement of all treatments (Robey & Schultz, 1998). Given that the findings of this and past studies on self-directed MTT and CT have repeatedly highlighted the importance of scheduling flexibility, repetition, autonomy, and social interaction to people with aphasia, future research may examine ways of maximizing the advantages of self-directed MTT and overcoming the reduced social interaction that comes with it so that this form of therapy can be optimally delivered. The findings of this and past research have also highlighted the potential negative perceptions of self-directed technology-based therapy that people with aphasia may hold. Dissatisfaction with self-directed MTT is important because it could negatively bias people with aphasia against NT, technology, or even aphasia therapy in general, and ultimately lead them to abandon therapy. As such, therapists should remain aware of these issues when deciding whether therapy apps are appropriate for particular clients. It is also important for researchers, app developers, and therapists to try to find ways of 125  overcoming technological limitations, such as the apparent inflexibility of NT to meet the challenge level needed by P2, in order to maximize the potential of self-directed MTT. For example, although it was not clear what exactly caused the challenge level of NT to be inappropriate for P2, altering the task demands might have improved the mismatch. Increasing the word set size or practicing abstract nouns, verbs, adjectives, or adverbs (Renvall et al., 2013a) could increase the difficulty level of practice for clients like P2 while maintaining or even increasing its functional relevance. It could also be that cueing hierarchy tasks are simply too easy for clients with mild aphasia/anomia, so including a range of therapy tasks for clients to choose from could increase the chances that therapy is perceived as adequately challenging and could also impact positively on interest and motivation (Horton, 2008; Wade et al., 2003). This range of tasks could be embedded in either separate apps or within the same app – for example, in one of its other picture-naming modes (Describe mode), NT also includes the option for Semantic Feature Analysis and/or Phonological Components Analysis (van Hees, McMahon, & Copland, 2013). 4.1.2 Therapeutic effects 4.1.2.1 Naming accuracy and naming acquisition All participants demonstrated improved picture naming for the words that they practiced during therapy. These improvements could not be attributed to the influence of confounding variables, as participants demonstrated stable baselines prior to therapy, there was a substantial advantage for practiced therapy set words over unpracticed control and monitor set words, and no clear changes were observed on secondary language outcome measures. Furthermore, blinding and high inter-rater reliability of scoring decreased the likelihood that these results were influenced by subjective bias. Thus, it can be concluded that self-directed, minimally supervised, 126  tablet-based naming therapy caused participants’ improved picture naming. However, participants also demonstrated decreased naming acquisition retention at post-therapy relative to the level they achieved during therapy. In line with quantitative results, participants reported perceiving improvement in their naming of practiced words, but P1 and P2 also noted persistent difficulties with the consistency of their naming. These findings are consistent with recent studies on the effectiveness of self-directed MTT indicating that participants can improve their naming of specific words using mobile technology with minimal therapist supervision (e.g., Kurland et al., 2014; Lavoie et al., 2016; Routhier et al., 2016). Naming accuracy outcomes varied across participants in the current study, with P2 showing the largest effect sizes and P3 the smallest. Studies from the MTT and CT literatures suggested that variation in the magnitude of therapy outcomes might relate to participant characteristics (e.g., Fridriksson et al., 2009; Palmer et al., 2012; Zheng et al., 2016). In the current study, there were no clear relationships across participants between the participant characteristics measured at baseline and the eventual therapy outcomes. P2 and P1’s comparatively better and poorer naming performance post-therapy, respectively, was generally consistent with their aphasia severity and/or composite cognitive deficit severity; however, P3 broke this trend, as he had less severe aphasia than P1 and no clear signs of nonlinguistic cognitive deficits, but still showed the least improvement of all the participants. It is possible that P3’s performance relative to the other participants may have related to the compounding effect of his motor speech impairment on his naming accuracy, or to an inadequate dosing of a large portion of his therapy set words as a result of the clerical error that randomized their assignment to the practice set. 127  The evidence of poor retention and lack of generalization obtained in this study is representative of the limitations of naming therapy more generally. Poor intrinsic generalization to unpracticed words is actually expected in naming therapy on theoretical grounds, due to the idiosyncratic and arbitrary relationship between word meaning and word form (Nadeau, Gonzalez Rothi, & Rosenbek, 2008). This prediction has been substantiated empirically: unless systematic steps are taken to promote such generalization (e.g., manipulating the semantic relatedness of therapy targets; Kiran & Thompson, 2003), the therapeutic effects of naming therapy are typically much larger for practiced words than unpracticed ones (Nickels, 2002a; Wisenburn & Mahoney, 2009). Moreover, Wisenburn and Mahoney’s (2009) meta-analysis of 44 naming therapy studies found that accuracy gains made in naming therapy tend to sharply decrease with each subsequent month post-therapy. Thus, the poor retention and generalization shown in this study, and indeed, as shown in other self-directed MTT or CT studies using naming therapy, may reflect the inefficiency of naming therapy in general, rather than the potential of self-directed technology-based therapy per se. 4.1.2.2 Secondary effects and clinical significance Despite no clear evidence of change on quantitative language measures, all participants reported perceiving improvements in their functional communication as a result of therapy. P2 also specifically reported increased use of and a more positive attitude towards iPads, and P1 reported a short-lived boost in his self-confidence. However, participants also reported and demonstrated negative psychological responses to therapy, including boredom as well as disappointment or significant frustration about not meeting their own performance expectations. These findings are in line with the participant-reported results of CT studies suggestive of positive effects of self-directed practice using technology on activity/participation and/or general 128  well-being, as well as such therapy’s potential risks in terms of negative psychological responses (e.g., Palmer et al. 2013; Wade et al., 2003). Moreover, participants reported these changes in spite of the fact that this study was not their only means of receiving therapy, which as Wade et al. (2003) pointed out, might have biased them to perceive self-directed therapy in a positive light. Indeed, the fact that all participants were receiving therapy outside of the study, or soon would be, at the time of the post-therapy interviews may account for why, in contrast to Wade et al.’s (2003) findings, participants reported no worries about the end of the study or losing the gains they made. Some of the participants in Wade et al.’s (2003) study were able to support their assertions about improved communication participation with specific examples of how communication had changed as a result of therapy, whereas none of the participants in the current study were able to do so. This discrepancy might relate to the inclusion of participants’ partners in Wade et al.’s (2003) study, who could provide specific corroboration. The difference may have also related to the 6-month therapy phase and 3-month post-therapy phase in Wade et al.’s (2003) study, which could have given participants more time to develop and reflect on new participation behaviours than the 6-7 week therapy phase and 1-month post-therapy phase in the current study. The only specific examples provided by any of the participants in the current study related to P2’s reported increased use of his iPad. However, it is not clear how these changes should be interpreted in light of the fact they were either not clearly quantitatively or qualitatively different from what P2 reported at pre-therapy, or they directly contradicted what he previously reported. Given that P2 reported expecting improvements in his technology use before therapy, it is possible that he was biased to perceive improvements after therapy that confirmed his expectations. Contrary to Wade et al. (2003) and Palmer et al. (2013), participants 129  in this study also did not report that therapy interfered with their everyday life activities, likely because of the very low time demands of this therapy. Thus, despite promising results for picture naming, there was little evidence that self-directed practice, as employed in this study, had a clinically significant effect on participants’ lives. Although the magnitude of this therapy’s direct effect on picture naming was medium to large relative to the effects typically obtained in the literature for SLP-based naming therapy (Beeson & Robey, 2006), these effect size estimates cannot necessarily be taken to indicate their clinical significance (Durlak, 2009). In reality, these seemingly impressive estimates represented the acquisition of only a relatively small set of 8-19 words (20-45% of participants’ therapy sets) – certainly insufficient to represent a full working vocabulary. On the other hand, even a small number of reliably accurate words could arguably represent a meaningful change in participants’ lives if those words were highly functionally relevant and participants were able to generalize their use to everyday contexts, thereby reducing the degree of their disability (Renvall et al., 2013a; Carragher et al., 2012). However, only a minority of participants’ acquired words were ones they nominated, and even then, both P1 and P3 acknowledged that participant nomination did not guarantee that identified words were particularly relevant to their everyday lives. Moreover, quantitative positive effects were observed only on a highly contrived picture-naming task, so there is no guarantee that the obtained effects generalized to more ecologically valid contexts (Carragher et al., 2012). The only evidence of clinically significant improvements came from participant-perspective interviews. However, the fact that participants’ claims could not be supported by specific, unambiguous examples necessarily reduced their persuasiveness. In fact, the clearest indication of a clinically significant effect in this study was the substantial 130  frustration P1 experienced during therapy, which was not only reported by P1 but was also observed behaviourally. In summary, the results of this exploratory study provided preliminary evidence that self-directed MTT using NT, as enacted in this study, is therapeutically active. Thus, along with the results of other studies on self-directed technology-based therapy, the findings indicate that self-directed MTT warrants further research and suggest that therapists are not wholly misdirected in using NT or self-directed MTT in general. However, evidence of poor retention and generalization in the absence of a clinically significant effect suggest that it may not be advisable for people with aphasia or therapists to use NT in isolation. Future studies might examine whether outcomes can be improved by supplementing self-directed MTT using NT with other self-directed or SLP-based therapy tasks, as part of a comprehensive therapy approach directed toward promoting retention and generalization to functional communication. Moreover, self-directed treatments for chronic illnesses (e.g., diabetes and rheumatoid arthritis) in the broader healthcare literature are often complex, with components addressing not only the impairment itself, but also education, the training of problem solving, goal setting, and coping skills, and psychological counseling (Newman, Steed, & Mulligan, 2004). As the research on self-directed treatment is much more developed for these other chronic illnesses, self-directed aphasia therapy and the clinical significance of its outcomes might be improved if therapy addresses more than just language and communication, by incorporating similar components into self-directed aphasia therapy (Holland, 2007). Findings regarding negative psychological effects experienced by participants in this and past studies necessarily negatively impact appraisal of the safety of NT and self-directed MTT for people with aphasia. Indeed, these findings are particularly concerning in light of theory and 131  evidence suggesting that negative emotionality or attitudes can significantly reduce the functional communication, participation, and response to therapy of people with aphasia (e.g., Babbitt & Cherney, 2010; Code & Herrmann, 2003; Fucetola et al., 2006; Votruba, Rapport, Whitman, Johnson, & Langenecker, 2013). As such, the psychological consequences not only of self-directed MTT, but also of aphasia and aphasia therapy in general are deserving of greater attention in the literature. Further research should determine the prevalence of negative psychological responses to therapy, factors affecting their occurrence, their possible immediate and long-term impacts on outcomes, and strategies to mitigate them. In the meantime, therapists should monitor clients’ psychological state and attempt to address any negative side effects of therapy, perhaps by modifying task demands or providing education or counseling. 4.1.3 Therapy process 4.1.3.1 Dosage The temporal distribution of participants’ self-directed practice varied across time and across participants on multiple dosage dimensions. For the most part, P1 and P3 followed a methodical routine, practicing at about the same dose (i.e., number of trials per day) and same session frequency (i.e., number of days per week) throughout the therapy phase. Both consistently practiced most days of the week, but P3 practiced at a higher dose than P1. P2’s dosage schedule was far more variable. With one notable very high-dose exception, P2 generally practiced at a low dose and session frequency, often lower than P1 and P3, for the majority of the therapy phase. However, in the last two weeks of therapy, P2 dramatically increased both his dose and session frequency to levels that were comparable to or much higher than P1 and P3’s dosage schedules. In the end, P3’s moderate-dose, high-frequency routine resulted in the highest 132  cumulative dose (i.e., total number of trials across therapy) of the participants, with P1 and P2 following with similar lower values. It is difficult to discuss participants’ dosage schedules in relation to the aphasia therapy literature, as the current study chose to control dose and allow session duration to vary, whereas the majority of the literature has opted to control session duration and leave dose unreported (Cherney, 2012). In principle, it would still be possible to extract dose information from these studies for comparison with the current one, since dose (i.e., trials per session) and session duration (i.e., time per session) are directly related to one another by dose rate (i.e., trials per unit time; Warren et al., 2007). However, in practice, this is not possible since dose rate has similarly been unreported in the literature and cannot be assumed to be invariant across sessions, across participants, or across therapy approaches. Moreover, even if it were possible to compare participants’ dosage schedules to those in other aphasia therapy studies, the variability within and between participants in the current study on multiple dosage variables makes their dosage schedules difficult to compare to the static unidimensional conceptualization of dosage found in the literature (Cherney, 2012; Cherney et al., 2011). To illustrate, following traditional definitions of the terms (Nadeau et al., 2008), P1 and P3’s schedules could both be considered relatively more distributed practice (i.e., lower-dose, higher-frequency), whereas P2’s dynamic schedule defies a simple label, but might be considered to most closely represent relatively more massed practice (i.e., higher-dose, lower-frequency). Despite this, because dose and session frequency are reciprocal according to these definitions, they cannot easily reconcile P3’s intermediate dose and high session frequency with P1 and P2’s. Another option more commonly employed in the aphasia therapy literature is to ignore dose and consider only session frequency when characterizing dosage (Cherney, 2012; 133  Cherney et al., 2011). Using this definition, P1 and P3’s schedules could instead be considered high-dosage practice and P2’s might be considered lower-dosage practice. Yet another common option is to ignore both dose and session frequency and instead use some variant of session duration as the unit of analysis. Using this definition of dosage, no matter whether duration is measured per day, per week, or summed across the therapy phase, it is clear that all participants’ patterns would mostly be considered to represent very low-dosage practice, since they practiced for much less time than what has typically been considered high-dosage therapy in the literature (Brady et al., 2012; Cherney et al., 2011). As proposed by Mortley et al. (2004) and Palmer et al. (2012), variations in participants’ dosage schedules may have related to the moderation of unmeasured personality, learning style, or lifestyle variables. They could have also related to the severity of participants’ aphasia/anomia deficits: P2’s more mild impairments may have given him more leeway to adapt his dosage schedule flexibly to his preferences or lifestyle and still benefit from therapy, whereas P1 and P3, being more severely impaired, may have required a more methodical routine to make gains and maintain them. The fact that P2 reported finding therapy inadequately challenging could also suggest an interaction between aphasia/anomia severity and task demands on motivation to practice: namely, P2’s perceived mismatch between the challenge level of NT and his needs could have reduced his motivation to practice regularly (Wade et al., 2003). Given the complexity of participants’ dosage schedules, it is unsurprising that there were no clear relationships between dosage and naming accuracy outcomes across participants. There may be hints of a temporal relationship between dosage and outcomes in P2’s usage and probe session data (cf. Figures 2 and 6): namely, P2’s sudden increase in dosage in his last two weeks of 134  therapy was associated with rapid gains in probe naming accuracy for his therapy set 2, the majority of which was being practiced at the time.  In summary, each participant chose to follow an idiosyncratic dosage schedule and no clear relationships between dosage and outcomes emerged in the current study. These findings can be taken as an illustration of the complexity of dosage and of why a simple unidimensional conceptualization of it is insufficient. Greater clarity about dosage and its relationship to outcomes may be gleaned if the aphasia therapy literature adopts a more refined model of dosage like Baker (2012) and Warren et al.’s (2007), and routinely measures, manipulates, and reports dosage accordingly (see section 1.3 for further discussion). Findings regarding the time commitment of self-directed MTT using NT have implications for future research and clinical practice. Namely, disregarding clinical significance for a moment, it is impressive that participants demonstrated such large effect sizes considering the fact that the total amount of time they had to practice was far less than even the weekly time commitment people with aphasia have been able to tolerate in many high-dosage therapy studies (Brady et al., 2012; Cherney et al., 2011). This suggests that therapists and researchers may have great leeway in increasing the dose and/or the number of words practiced without compromising acceptability. Some research suggests that both of these options are tolerable to people with aphasia, that increasing dose can improve naming accuracy outcomes, and that people with aphasia can make equivalent proportional improvements for larger therapy word sets as smaller ones (i.e., in absolute terms, they relearn more words with larger sets; Fillingham et al., 2005a; Laganaro, Di Pietro, & Schnider, 2006; Snell et al., 2010). Furthermore, the latitude to practice larger word set sizes could increase the feasibility of including a maintenance mechanism in future studies, whereby acquired words continue to be practiced (but perhaps at a lower dosage) while new 135  words are added to the practice set over time. This could improve the clinical significance of therapy by countering the poor retention rate obtained in this study and naming therapy generally while enabling the gradual expansion of acquired words into a full working vocabulary (Pedersen et al., 2001; Nadeau et al., 2008; cf. Conroy & Scowcroft, 2012). 4.1.3.2 Therapeutic procedures Results suggested that each participant had a strong bias to use only a small subset of all the possible cueing strategies that NT permits. They showed a strong preference for word-form cues over semantic cues. P1’s reported pattern of cueing was consistent with errorful increasing cueing techniques found in the literature (Fillingham et al., 2003; Middleton & Schwartz, 2012). In contrast, the cueing technique P2 reported using was similar to an across-session variant of the decreasing cueing techniques described in the literature (e.g., Conroy et al., 2009a). Quantitative data was not available to confirm participants’ self-report or establish P3’s cueing techniques; however, participants’ predominant use of only one or two cues and the short duration of most of their therapy trials necessarily restricted the range of cueing strategies they could have been using, suggesting that they were using relatively unelaborated variants of the cueing techniques found in the literature. Participants provided reasonable explanations for their reported patterns of cue use, suggesting that they selected cues purposefully and systematically, contingent on their prior performance. Thus, despite receiving no direction from the researcher regarding cue use, participants appeared to use a small set of simple yet theoretically-motivated cueing strategies based on a reasoned approach to the therapy process similar to that found in the SLP-based therapy literature. The cueing preferences participants reported and/or demonstrated are interesting when interpreted in relation to the literature. Despite variation in the operationalization of errorless and 136  errorful learning, participants with aphasia in studies on errorless learning have often reported preferring errorless to errorful learning, because they experienced more success when using it and found it less frustrating and more rewarding (Cherney et al., 2014; Conroy et al., 2009a, 2009b; Fillingham et al., 2005a, 2005b, 2006; Raymer, Strobel, Prokup, Thomason, & Reff, 2010). Despite this, when given free rein over therapeutic procedures during self-directed practice, P1 instead spontaneously chose to use an apparent errorful learning approach, using the Spoken Word cue (the highest-support cue) only after lack of success with lower-support cues. Furthermore, P1’s perception of using the Spoken Word cue as a sign of failure was inconsistent with the principles of errorless learning, which typically uses high-support cues to reduce the likelihood of errors (Middleton & Schwartz, 2012). P1’s preference for errorful learning might have related to participant characteristics, his past exposure to different kinds of cueing techniques, the study design, and/or his apparent high valuing of effortful practice. Like dosage schedules, participants’ chosen cueing strategies may reflect the influence of unmeasured personality, learning style, or lifestyle variables (Mortley et al., 2004; Palmer et al., 2012). Some errorless learning studies have also found that participants with more severe aphasia were more likely to prefer errorless learning, possibly because it was the only approach likely to facilitate their immediate success, whereas milder participants had more variable preferences or preferred errorful learning, perhaps because they were likely to benefit from either approach or because they found it more challenging and less intrusive (Cherney et al., 2014; Conroy et al., 2009a). As P1 had moderate aphasia, it is possible that he perceived errorless and errorful learning similarly to the milder participants in these studies. Furthermore, errorful increasing cueing techniques like those reported by P1 are common in SLP-based therapy, and both therapists and clients in such contexts demonstrably value effortful 137  practice (Horton, 2008). As P1 has a long history of SLP-based therapy and was engaged in concurrent SLP-based therapy during much of the study, it is possible that P1 had had little exposure to errorless cueing techniques, and therefore was simply replicating the techniques and values he had learned from SLP-based therapy in his self-directed practice. Exposure to words without cueing during probe picture naming, which is essentially an errorful cueing technique, may have similarly biased P1 to favour errorful strategies during self-directed practice. Moreover, as McKissock and Ward (2007) pointed out, beliefs about the importance of effortful practice may lead people with aphasia like P1 to reject errorless cueing techniques, despite available evidence from the aphasia literature suggesting errorless and errorful learning have essentially equivalent effects (Fillingham et al., 2003; Middleton & Schwartz, 2012). In contrast, participants in studies comparing these two approaches may have gained sufficient experience with both to appreciate the merits of errorless learning.   In contrast to P1, P2 reported using a decreasing cueing technique, but still valued effortful practice. This may appear contradictory given that decreasing cueing is typically considered a type of errorless learning and errorless learning has often been assumed to involve low effort (Conroy & Lambon Ralph, 2012). However, these assumptions are controversial, as others have proposed classifying decreasing cueing as a type of errorful learning due to its greater tolerance for error production than the more common high-support static cueing variant of errorless learning (Middleton & Schwartz, 2012), or have argued that it instead represents a compromise between the two approaches, balancing error rates and effort (Conroy et al., 2009a; Conroy & Lambon Ralph, 2012). P2’s self-report may be consistent with this latter interpretation, in that he appeared to be attempting to modify his cue use in response to his progress in therapy in order to maintain high effort with low error rates as naming became easier 138  for him. In addition to their reported reasons, participants may have preferred unelaborated word form-based cueing strategies because they permitted a relatively fast pace of naming practice. This was likely possible because, as Best, Herbert, Hickin, Osborne, and Howard (2002) noted, word-form cues are presented much more concisely than semantic cues. P2’s comments about the length and wordiness of his semantic cues may indicate his sensitivity to this fact. Since anomia is frequently highly frustrating for people with aphasia, as it was particularly for P1, participants may have chosen a fast pace of practice as an attempt to limit the uncomfortable symptoms of anomia.  There appears to be little precedent in the aphasia therapy literature for the therapeutic procedure analyses conducted in this study. Few studies have empirically described any aspect of the therapy process (Hinckley & Douglas, 2013; Horton & Byng, 2000). Doesborgh et al.’s (2004) therapy study is likely the most similar to the current study in terms of giving participants some control over cueing, except that they did not attempt to describe participants’ chosen cueing strategies. They did, however, report null findings regarding the therapeutic effect of their therapy, which appears to conflict with the findings of the current study. Aside from an unfixed cueing hierarchy, Doesborgh et al. (2004) and the current study differed greatly on multiple aspects of methodology. Although any of these differences may have contributed to the discrepancy between studies, the difference in outcomes measures was particularly salient. Doesborgh et al. (2004) used the Boston Naming Test (BNT) as their primary and only outcome measure. Because their participants did not practice the specific words used in the BNT, the BNT was acting as an intrinsic generalization measure. Given that naming therapy in general has shown poor intrinsic generalization (Nickels, 2002a; Wisenburn & Mahoney, 2009), no changes on the BNT would be anticipated. Moreover, the BNT, as a standardized norm-referenced test, 139  was not designed to be sensitive to the small, word-specific changes in accuracy targeted in naming therapy in the first place (Dollaghan, 2008; Harry & Crowe, 2014). In contrast, the primary outcome measure of the current study, picture naming of the specific words practiced in therapy, represented a direct and very sensitive measure of change (Herbert et al., 2003).  In line with this account, participants in the current study similarly did not demonstrate any clear change on the BNT-2. Mortley et al.’s (2004) study was similar to the current study by evaluating self-directed minimally-supervised CT home practice, and even included some participant-controlled therapeutic procedures. However, although the authors noted individual differences in participants’ patterns of practice and suggested that they may relate to personality or learning style variables, they did not describe these patterns in detail. In contrast, Fink, Brecher, Schwartz, and Robey (2002) evaluated an in-clinic computer-based naming therapy, in which a portion of the therapy sessions were unsupervised and participants were given control over the cueing hierarchy, making it somewhat similar to the current study. Moreover, Fink et al. (2002) did briefly compare two of their participants’ use of specific therapeutic procedures and speculate about how they may relate to outcomes. They noted their participants preferred to use simple cueing strategies, one using an increasing and the other a decreasing cueing technique, much like in the current study. In light of equivalent therapeutic effects despite these individual differences in cueing strategies, Fink et al. (2002) hypothesized that particular cueing strategies per se may have not been critical to the outcomes. The findings of the current study are similar, in that there was no clear relationship between participants’ chosen cueing strategies and their naming accuracy outcomes across participants. 140  In summary, these findings suggested that self-directed practice using NT enabled participants with aphasia to spontaneously enact therapy in ways similar to how a therapist might. This may indicate that concerns regarding the negative effects of the lack of performance-contingent cueing and feedback in therapy apps may be overstated (Brandenburg et al., 2013; Kurland et al., 2014). Participants appeared to be able to overcome these technological limitations when given the freedom to do so, independently using systematic and reasoned cueing strategies with positive outcomes. Instead of specific cueing strategies being important, Fink et al. (2002) postulated that the extent to which cues were able to successfully and repeatedly elicit accurate responses could have been the active ingredient of their naming therapy. As naming accuracy during self-directed practice was not examined, this study could not evaluate this possibility. Future research may be able to provide insights into the possible active ingredients of aphasia therapy by examining its process in detail and relating it to outcomes (i.e., process-outcome research). This could help direct researchers and therapists towards ways to optimize aphasia therapy by maximizing the provision of active ingredients and minimizing or eliminating inert or burdensome components of therapy. In order to accomplish this goal, the aphasia therapy literature might benefit from clinical psychology, which has a process-outcome psychotherapy literature with well-developed methodology and quantitative analytic methods (e.g., Crits-Christoph, Gibbons, & Mukherjee, 2013). The aphasia therapy literature may nonetheless also benefit from quantitative and qualitative research examining only the therapy process, without directly relating it to outcomes (i.e., process research), as the current study did. For example, Horton (2008) examined the process of SLP-based therapy using data from naturalistic therapy sessions, without relating the process findings to the outcomes of therapy. Instead, he reported detailed analyses of the 141  dynamics of therapeutic procedures using qualitative methodologies, and identified themes overlapping with the current study, such as the valuing of effortful practice and a tendency to use increasing cueing techniques, as well as the interactional dynamics between therapists and clients and the subtle yet active role clients played in the therapy process in spite of therapists’ overt control over it. Furthermore, the therapy process analysis of this study, despite not directly relating process to outcomes, provided some insight into ways of optimizing self-directed practice MTT using NT. Results revealed that participants both disliked and chose to not use the semantic cues during self-directed practice. Combined with the finding that the semantic cues were time-consuming to develop, these findings suggest that the particular semantic cues included in NT may have limited clinical utility for some people with aphasia and might be profitably replaced or supplemented in future app development with different semantic cues or other kinds of cues (e.g., articulatory-kinematic or rate/rhythm-control cues for people with aphasia and apraxia of speech; Wambaugh & Shuster, 2008) that may increase the efficiency of aphasia therapy while maximizing the potential for all people with aphasia to respond to therapy (cf. Wambaugh, Wright, Nessler, & Mauszychi, 2014). 4.2 Limitations 4.2.1 Research design and internal validity This study employed a multiple baseline across-behaviours design using a dynamic set of words. While this design satisfied the minimum requirements of an experimental design allowing causal conclusions to be drawn regarding the study’s naming accuracy results (see section 2.2 for further discussion), it also had a number of other advantages and some disadvantages associated with the dynamic word set approach. Because the acquisition criterion was applied to individual words rather than composite word lists, the dynamic therapy sets provided many more 142  opportunities for staggering the introduction and withdrawal of therapy, and thus stronger evidence of the internal validity of findings. In fact, whereas the design of this study resulted in four or five therapy staggerings across participants, a more traditional word-list acquisition criterion (e.g., 80% of the whole word list accurate across two occasions) would have resulted in a total of only one staggering across the participants: for P2, at the beginning of his last week of therapy. Furthermore, as discussed by Conroy and Scowcroft (2012), a dynamic word set is more clinically practical than the traditional across-lists design, as it could allow participants’ progress to potentially go on indefinitely by avoiding a performance ceiling. In fact, if therapy had not been withdrawn from acquired words, but rather they had been practiced alongside new words (similar to Conroy & Scowcroft, 2012), then the design of this study would have approximated the maintenance mechanism discussed earlier for augmenting the clinical significance of naming therapy (see section 4.1.3.1).  On the other hand, the dynamic word set approach using in this study had some disadvantages. Words responding well to therapy were acquired and removed from the practice set, whereas words responding poorly were kept on for more practice. This meant that the approach carried the risk of causing a gradual accumulation of poorly-responding words in the practice set (i.e., oversaturation; Conroy & Scowcroft, 2012), resulting in the stalling of progress as the frequency of inaccurate naming attempts limited the opportunity for new words to be introduced to practice. This situation appeared to be the case for P1, as the number of unpracticed words being transferred to the practice set reached a plateau towards the end of his therapy phase, resulting in a subset of them never being practiced before the end of the therapy phase. This situation, combined with loss of gains on previously acquired words that were no longer being practiced, may have contributed to the substantial frustration P1 experienced during 143  therapy. Thus, P1’s frustration, the major negative outcome found in this study, may have in part been an artifact of the design used, rather than a negative side effect of self-directed MTT per se. Future research should examine how these potential consequences of the dynamic word set approach can be mitigated so that continued progress can be maintained. For example, continuing to practice acquired words alongside unacquired ones not only could improve maintenance of gains, but experiencing continued success in therapy could help maintain clients’ morale. Moreover, adding an error criterion to future versions of this design – whereby words showing frequent errors are removed from therapy – could help reduce frustration and avoid the potential for word-specific difficulty dictating therapy progress (Conroy & Scowcroft, 2012).  The dynamic word set approach also had practical limitations in terms of the amount of effort required of the researcher and the number of opportunities allowed for human error. The dynamic word set approach required that participants’ performance for individual words be scored online during probe sessions, placing high demands on the researcher’s attention. Although intra-rater reliability suggested this did not substantially affect the researcher’s scoring performance, it nonetheless required greater effort than staggering therapy based on an interval criterion (i.e., an a priori decision to switch words or lists after a set amount of time) or an across-lists acquisition criterion. An interval criterion would have required no online monitoring by the researcher, whereas an across-lists acquisition criterion would have required tracking accuracy only for the composite list rather than for each individual word. Moreover, application of the acquisition criterion to individual words meant that keeping track of the composition of the practice set was challenging, as it tended to change frequently in this approach. This clearly contributed in part to the clerical error that compromised P3’s design, as word set management would have been comparatively much simpler if an interval or across-lists acquisition criterion 144  were used. Thus, for P3, the therapy staggering advantage for internal validity afforded by the dynamic word set approach was undermined by its higher opportunity for human error. Future research should evaluate whether the advantages of this approach are practically worthwhile for research or clinical purposes in light of the increased researcher/therapist effort it can require. 4.2.2 External validity Similar to the broader MTT and CT literatures (see section 1.6 for further discussion), participants in this study were relatively young, had attained a high level of education and formerly had relatively high socioeconomic status occupations, had mild or moderate non-fluent aphasia with good comprehension, none to moderate cognitive deficits, and possessed extensive past technology experience. Furthermore, participants self-selected to participate in this study and all reported positive perceptions of mobile technology at baseline. As such, it is unclear how generalizable findings are to participants with other combinations of characteristics. For example, accessibility or acceptability may have been lower for people with aphasia who have more complex motor, language, cognitive, or perceptual issues, or who have less experience or less positive impressions of mobile technology (Brandenburg et al., 2013; McNaughton & Light, 2013). Given that the sample size of this study was small, the findings are limited in the extent to which they can be applied to other people with aphasia. This is a limitation common to all single-subject research (Barlow et al., 2009). In this methodology, confidence in the reliability and generalizability of findings is increased through direct and systematic replication. Although staggering therapy as part of the multiple baseline design of this study limited the possibility that history effects compromised the study’s internal validity, the fact that all participants were engaged in concurrent SLP-based therapy may nevertheless limit the study’s external validity. That is, in principle it is possible that the concurrent SLP-based therapy 145  influenced participants’ behaviour during, and their response to, therapy. For example, therapy outside of the study could have primed or augmented the naming accuracy gains participants made during self-directed practice. Furthermore, engagement in SLP-based therapy may have helped maintain participants’ motivation to practice on their own, or conversely, reduced their investment in the success of self-directed practice as it was not their only opportunity for therapy. It could have also altered their therapy process, for example, leading them to modify their practice schedule to accommodate the demands of SLP-based therapy or to use therapy techniques similar to those experienced in their SLP-based therapy. Although it is important to consider these possibilities when interpreting the results of this study, for practical purposes this may not be a significant limitation, as in most clinical applications, minimally-supervised self-directed home practice may likely be used in conjunction with SLP-based therapy.  As picture naming with a cueing hierarchy is a relatively straightforward task, positive findings regarding giving participants substantial control over the therapy process may not generalize to therapy approaches that are more complex (e.g., involving numerous steps). Similarly, the findings cannot be assumed to be representative of the feasibility of self-directed MTT using NT if participants had been completely unsupervised. However, findings of this study suggested that therapist involvement in self-directed NT home practice may be important for the therapy process in terms of creating, tailoring, and managing word sets, as well as maintaining motivation to continue practicing.   Reactivity may have also been an issue for this study. In principle, participants’ awareness of being monitored by NT may have altered how they acted during minimally-supervised self-directed practice, compared to how they may have if they had not been monitored. However, since the NT data logging system was very unobtrusive, this may not be a 146  significant limitation. The participant-perspective interviews may have been more significantly susceptible to reactivity, as they were conducted by the researcher. Participants were aware that the researcher was an SLP graduate student who was personally invested in the study as a part of his coursework. This knowledge could have biased participants to report their experiences of the study and self-directed MTT in a more positive light and to withhold information they felt the researcher would not want to hear (Wade et al., 2003). Moreover, P1 and P2’s comments indicating that any therapy is good therapy could suggest that an additional general positive bias for therapy was operating in this study: that is, the hope resulting from any attempt to improve may lead people with aphasia to be less discriminating regarding the advantages and disadvantages of different kinds of therapy (Wade et al., 2003). Reactivity might be predicted to have particularly influenced responses regarding the importance of the researcher’s role in the process and outcomes of therapy and regarding dissatisfaction with therapy or negative side effects. As participants did report some dissatisfaction and negative side effects, it is possible that this was not in the end a significant limitation of this study. Nonetheless, reactivity may be reduced in future studies by having a third party conduct all interviews with participants (Wade et al., 2003).  4.2.3 Measurement A number of measures that were not included or were inadequately measured in this study may have been helpful for interpreting results. No formal measures of apraxia of speech, dysarthria, or speech intelligibility were included to help characterize P3’s motor speech impairment, which might have aided in the interpretation of his results and contributed to an evaluation of suitable candidates for this therapy. Moreover, quantitative data regarding the researcher’s time commitment in conducting the study – including time taken to develop word 147  sets and commute to participants’ homes – was not consistently measured. Given that minimally-supervised, self-directed MTT has been argued to be a time-efficient therapy approach (Mortley et al., 2004; Palmer et al., 2012), this data may have aided in evaluating whether this therapy was an efficient use of therapist time. Lastly, the modified version of NT used in this study was not utilized to its full potential, as the audio recordings of participants’ therapy trials were not used to score participants’ naming accuracy during self-directed practice. As a result, it was not possible to evaluate the adequacy of the self-scoring system under the conditions where remote progress monitoring would actually occur, which necessarily limited the value of the study’s findings regarding this sub-component of the feasibility research question. Moreover, the fact that the therapy trial audio recordings were not analyzed meant that the therapeutic procedures sub-component of the therapy process research question had to be evaluated largely on the basis of participants’ self-report. Despite the use of partner-supported AAC strategies (including demonstration using NT), participants were unable to supply sufficiently detailed information to adequately evaluate key concerns highlighted in the literature regarding the appropriateness of therapy apps like NT for self-directed practice (specifically, their inflexibility and the potential for error learning; see section 1.5.2). Given that cueing techniques and performance-contingency are complex topics, these concerns might have been more fruitfully evaluated on the basis of quantitative data derived from the extensive audio data collected by the NT data logging system.   Inter-rater reliability was evaluated only for naming accuracy scoring of probe picture naming. Although many of the variables relevant to the feasibility and therapy process research questions were automatically collected by the NT data logging system, navigation scoring, all standardized tests and other clinical profile measures, and the participant-perspective interviews were administered and/or scored only by the researcher. As a result, an estimate of measurement 148  error is unavailable for many of the variables in this study, and their values may have been influenced by the subjective biases of the researcher.  There were also issues with the secondary outcomes measures used that limited their usefulness. Because, despite their frequent use, both the WAB-R and BNT-2 have poor psychometric properties and are inappropriate for measuring small changes targeted in therapy (Dollaghan, 2008; Harry & Crowe, 2014; Hula et al., 2010), they could only be used to coarsely control for history or maturational effects, rather than used according to their intended purpose as generalization measures. Results on the Cookie Theft Picture, the only quantitative discourse/activity-level measure used in this study, were largely uninterpretable because of inadequate sampling (i.e., the measure was assessed only once pre- and post-therapy), making it impossible to differentiate therapy effects from the effects of confounding variables related to endogenous change (i.e., testing, maturation, and statistical regression effects) and instrumentation effects. Thus, the participant-perspective interviews were the de facto secondary outcome measure in this study. However, although clients’ perceptions of the effects of therapy are important in ensuring quality care, using qualitative interviews to assess them is susceptible to reactivity and subjective bias in administration and analysis. As such, it is important that future self-directed MTT studies also include quantitative outcome measures at the activity- and participation-levels in order to evaluate the clinical significance of therapy outcomes. In particular, multiple authors have called for more systematic and routine evaluation of everyday conversation, tested over multiple days per time point, in order to understand the functional impacts of therapy (Boyle, 2011; Carragher et al., 2012). Moreover, little information could be collected from P3 on the basis of the interviews, despite the fact that they were facilitated using partner-supported AAC strategies. This led the researcher to have to make heavy use the more 149  constrained probe questions and researcher-generated ideas for P3 (see section 2.3.1 for clarification of the interview methodology), which increased the risk of the researcher’s subjective biases impacting P3’s responses (Luck & Rose, 2007). As there are clearly limits to the use of verbal questioning for collecting participant-report data for some people with more severe aphasia, it may be advisable for future studies to include higher-support AAC strategies during interviews, like the visual analog scale and picture-selection task used in Palmer et al.’s (2013) study. 4.2.4 Analysis Blinded analysis was conducted only for naming accuracy scoring of probe picture naming, and all results of this study were analyzed descriptively, so the researcher’s subjective biases likely influenced the analysis of all other findings. Moreover, because the experimental design targeted individual words, but Cohen’s d and d1 could not be used for word-level nominal accuracy data (Lipsey & Wilson, 2001), it was not possible to calculate these standard mean gain effect size statistics according to Beeson and Robey’s (2006) recommended method (see section 2.6.2 for discussion), which affected their interpretation relative to Beeson and Robey’s benchmarks. However, as previously discussed (see section 2.6.2), standard mean gain or difference effect sizes are inappropriate in the first place for data that is naturally dichotomous like naming accuracy. The odds ratio is instead most appropriate for such data, and this study demonstrated the flexibility and greater ease of interpretation of the odds ratio. As such, future studies might consider reporting odds ratios alongside standard mean gains and differences in order to provide a more valid measure of effect size while still allowing comparison to established benchmarks for the standard mean gain. 150  4.3 Conclusion In summary, minimally-supervised self-directed home practice using Naming Therapy© was found to be accessible and acceptable for participants with aphasia. Participants were overall mostly adherent to the study protocol, which required little time commitment from them to complete. Despite the short duration and the low time demands of the therapy for participants, therapy was found to be therapeutically active, with participants demonstrating medium to large naming accuracy gains for practiced words, but minimal gains for unpracticed words. Enabling participants to direct their own therapy process led them to show individualized patterns of practice. Participants chose to spontaneously practice much more than expected in some cases and all of them enacted simple yet systematic therapeutic procedures, similar to how a therapist might. These results suggest that self-directed mobile technology-based therapy with minimal therapist supervision may be a viable service delivery option for people with aphasia and thus warrants further investigation. However, problems were identified with the researcher’s administrative role in developing and managing word sets, and monitoring therapy progress. The words native to Naming Therapy© were found to be insufficient for use with people with mild impairments, the app’s system for manipulating word sets was insufficiently accessible for the researcher, and its system for remote progress monitoring was found likely to have little clinical utility. These issues limit the feasibility of using Naming Therapy© for both research and clinical practice. Moreover, there was no evidence of therapy causing a clinically significant improvement in outcomes. Finally, the study documented participants’ general satisfaction with the therapy program, despite some dissatisfaction with aspects of the therapy they received, as well as, for one participant, negative side effects on his psychological state. It should be acknowledged, however, that this was an exploratory study with a small and non-representative 151  sample of participants. Moreover, some of the secondary outcome measures were noted to be inadequate, reactivity to assessment and experimental design were flagged as issues, and there was the potential for subjective bias to influence the measurement and analysis of some of the findings. Although the concept of self-directed practice is relatively new to aphasia therapy, analogous treatments that aim to increase clients’ involvement and control in their treatment and its effects on their lives are not new to healthcare, especially in the context of chronic illnesses (Newman et al., 2004). The forces driving the development and application of these self-directed treatments in medicine and other healthcare fields are the same as those driving the nascent interest in the use of mobile technology in aphasia therapy: namely, limited resources in the face of growing demand as the population ages, and a philosophical shift from the traditional biomedical model of care to a more collaborative model with shared expertise between clients and healthcare professionals (Newman et al., 2004). The fact that these driving forces represent a broader trend in healthcare suggests that speech-language pathology will not be exempt from their influence. Indeed, it may be necessary for therapists and researchers to find ways of adapting many aphasia therapy approaches (not only naming therapy) to be self-directed, if aphasia therapy is to remain a viable treatment option for people with aphasia in the changing healthcare system. Fortunately, the preliminary evidence from this study and relevant past research suggests that accessible mobile technology may be one way of addressing these challenges. In any case, the mobile technology revolution currently underway will likely cement mobile technology’s place in speech-language pathology, as it has already in the larger society (Brandenburg et al., 2013). As such, future research should explore ways of optimizing self-directed mobile 152  technology-based therapy with minimal supervision. The findings of this study and past research suggest that one potential way of doing this for naming therapy might be to provide self-directed naming therapy in the context of a longer-term study, with an ongoing dynamic word set development process targeting functionally relevant words, and with mechanisms in place to maintain gains while expanding a working vocabulary (Nadeau et al., 2008). Self-directed therapy could be remotely monitored and managed, allowing the therapist to then take on more of a coaching and counseling role in therapy (Holland, 2007; Mortley et al., 2003), in order to promote generalization to functional communication as well as life participation and psychological well-being. 153  References Allen, M., McGrenere, J., & Purves, B. (2008). The field evaluation of a mobile digital image communication application designed for people with aphasia. ACM Transactions on Accessible Computing, 1(1), 1-26. Babbitt, E. M., & Cherney, L. R. (2010). Communication confidence in persons with aphasia. Topics in Stroke Rehabilitation, 17(3), 214-223. Baker, E. (2012). Optimal intervention intensity. International Journal of Speech-Language Pathology, 14(5), 401-409. Barlow, D. H., Nock, M. K., & Hersen, M. (2009). Single case experimental designs: Strategies for studying behavior change (3rd ed.). Boston, MA: Pearson Education, Inc. Beeson, P. M., & Robey, R. R. (2006). Evaluating single-subject treatment research: Lessons learned from the aphasia literature. Neuropsychological Review, 16: 161-169. Beukelman, D. R., & Mirenda, P. (2013). Augmentative & alternative communication: Supporting children & adults with complex communication needs (4th ed.). Baltimore, MA: Paul H. Bookes Publishing Co. Best, W., Herbert, R., Hickin, J., Osborne, F., & Howard, D. (2002). Phonological and orthographic facilitation of word-retrieval in aphasia: Immediate and delayed effects.  Aphasiology, 16(1/2), 151-168. Best, W., Grassly, J., Greenwood, A., Herbert, R., Hickin, J., & Howard, D. (2011). A controlled study of changes in conversation following aphasia therapy for anomia. Disability and Rehabilitation, 33(3), 229-242. Bhogal, S. K., Teasell, R., & Speechley, M. (2003). Intensity of aphasia therapy, impact on recovery. Stroke, 34, 987-993. 154  Boyle, M. (2011). Discourse treatment for word retrieval impairment in aphasia: The story so far. Aphasiology, 25(11), 1308-1326. Brady, M. C., Kelly, H., Godwin, J., & Enderby, P. (2012). Speech and language therapy for aphasia following stroke. The Cochrane Database for Systematic Reviews, 5, CD000425. Brandenburg, C., Worrall, L., Rodriguez, A. D., & Copland, D. (2013). Mobile computing technology and aphasia: An integrated review of accessibility and potential uses. Aphasiology, 27(4), 444-461. Breitenstein, C., Kamping, S., Jansen, A., Schomacher, M., & Knecht, S. (2004). Word learning can be achieved without feedback: Implications for aphasia therapy. Restorative Neurology and Neuroscience, 22, 445-458. Brodeur, M. B., Dionne-Dostie, E., Montreuil, T., & Lepage, M. (2010). The Bank of Standardized Stimuli (BOSS), a new set of 480 normative photos of objects to be used as visual stimuli in cognitive research. PLoS ONE, 5(5), 1-13. Brodeur, M. B., Guérard, K., & Bouras, M. (2014). Bank of Standardized Stimuli (BOSS) phase II: 930 new normative photos. PLoS ONE, 9(9), 1-10. Bryen, D. N. (2008). Vocabulary to support socially-valued adult roles. Augmentative and Alternative Communication, 24(4), 294-301. Brysbaert, M., & New, B. (2009). Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41(4), 977-990. Brysbaert, M., New, B., & Keuleers, E. (2012). Adding part-of-speech information to the SUBTLEX-US word frequencies. Behavior Research Methods, 44: 991-997. 155  Byng, S. (1995). What is aphasia therapy? In C. Code & D. Müller (Eds.), The treatment of aphasia: From theory to practice (pp. 3-17). London: Whurr. Carragher, M., Conroy, P., Sage, K., & Wilkinson, R. (2012). Can impairment-focused therapy change the everyday conversations of people with aphasia? A review of the literature and future directions. Aphasiology, 26(7), 895-916. Carragher, M., Sage, K., & Conroy, P. (2013). The effects of verb retrieval therapy for people with non-fluent aphasia: Evidence from assessment tasks and conversation. Neuropsychological Rehabilitation, 23(6), 846-887. Cepeda, N. J., Pasler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. Cherney, L. R. (2012). Aphasia treatment: Intensity, dose parameters, and script training. International Journal of Speech-Language Pathology, 14(5), 424-431. Cherney, L. R., Kaye, R., & van Vuuren, S. (2014). Acquisition and maintenance of scripts in aphasia: A comparison of two cuing conditions. American Journal of Speech-Language Pathology, 23, S343-S360. Cherney, L. R., Kaye, R. C, Lee, J. B., van Vuuren, S. (2015). Impact of personal relevance on acquisition and generalization of script training for aphasia: A preliminary analysis. American Journal of Speech-Language Pathology, 24, S913-S922. Cherney, L. R., Patterson, J. P., Raymer, A. M. (2011). Intensity of aphasia therapy: Evidence and efficacy. Current Neurology and Neuroscience Reports, 11, 560-569. Cherney, L. R., & Robey, R. R. (2008). Aphasia treatment Recovery, prognosis, and clinical effectiveness, in R. Chapey (Ed.), Language intervention strategies in aphasia and 156  related neurogenic communication disorders (5th ed.) (pp. 186-202). Baltimore, MA: Lippincott Williams & Wilkins. Cicerone, K. D., Langenbahn, D. M., Braden, C., Malec, J. F., Kalmar, K., Fraas, M.,… Ashman, T. (2011). Evidence-based cognitive rehabilitation: Updated review of the literature from 2003 through 2008. Archives of Physical Medical Rehabilitation, 92, 1681-1692. Code, C., & Heron, C. (2003). Services for aphasia, other acquired adult neurogenic communication disorders and swallowing disorders in the United Kingdom, 2000. Disability and Rehabilitation, 25(21), 1231-1237. Code, C., & Herrmann, M. (2003). The relevance of emotional and psychosocial factors in aphasia to rehabilitation. Neuropsychological Rehabilitation, 13(1/2), 109-132. Code, C., & Petheram, B. (2011). Delivering for aphasia. International Journal of Speech-Language Pathology, 13(1), 3-10. Coltheart, M. (1981). The MRC psycholinguistic database. The Quarterly Journal of Experimental Psychology Section A, 33(4), 497-505. Conroy, P., & Lambon Ralph, M. A. (2012). Overview and ways forward for future research. Neuropsychological Rehabilitation, 22(2), 319-328. Conroy, P., & Scowcroft (2012). Decreasing cues for a dynamic list of noun and verb naming targets: A case-series aphasia therapy study. Neuropsychological Rehabilitation, 22(2), 295-318. Conroy, P., Sage, K., & Lambon Ralph, M. A. (2009a). The effects of decreasing and increasing cue therapy on improving naming speech and accuracy for verbs and nouns in aphasia. Aphasiology, 23(6), 707-730. 157  Conroy, P., Sage, K., & Lambon Ralph, M. A. (2009b). Errorless and errorful therapy for verb and noun naming in aphasia. Aphasiology, 23(11), 1311-1337. Conroy, P., Sage, K., & Lambon Ralph, M. A. (2009c). Improved vocabulary production after naming therapy in aphasia: Can gains in picture naming generalize to connected speech? International Journal of Language & Communication Disorders, 44(6), 1036-1062. Crits-Christoph, P., Gibbons, M. B. C., & Mukherjee, D. (2013). Psychotherapy process-outcome research. In Lambert, M. J. (Ed.), Bergin and Garfield’s handbook of psychotherapy and behaviour change (6th ed.) (pp. 298-340). Hoboken, NJ: John Wiley & Sons, Inc. Dalemans, R., Wade, D. T., van den Heuvel, W. J. A., & de Witte, L. P. (2009). Facilitating the participation of people with aphasia in research: A description of strategies. Clinical Rehabilitation, 23, 948-959. Davidson, B., Howe, T., Worrall, L., Hickson, L., & Togher, L. (2008). Social participation for older people with aphasia: The impact of communication disability on friendships. Topics in Stroke Rehabilitation, 15(4), 325-340. Davidson, B., Worrall, L., & Hickson, L. (2003). Identifying the communication activities of older people with aphasia: Evidence from naturalistic observation. Aphasiology, 17(3), 243-264. Dickey, L., Kagan, A., Lindsay, P., Fang, J., Rowland, A., & Black, S. (2010). Incidence and profile of inpatient stroke-induced aphasia in Ontario, Canada. Archives of Physical Medicine and Rehabilitation, 91, 196-202. 158  Doesborgh, S. J. C., van de Sandt-Koenderman, M. W. M. E., Dippel, D. W. J., van Harskamp, F., Koudstaal, P. J., & Visch-Brink, E. G. (2004). Cues on request: The efficacy of Multicue, a computer program for wordfinding therapy. Aphasiology, 18(3), 213-222. Dollaghan, C. A. (2007). The handbook for evidence-based practice in communication disorders. Baltimore, MD: Paul H. Brookes Publishing Co. Durlak, J. A. (2009). How to select, calculate, and interpret effect sizes. Journal of Pediatric Psychology, 34(9), 917-928. Elman, R. J., Ogar, J., & Elman, S. H. (2000). Aphasia: Awareness, advocacy, and activism. Aphasiology, 14(5/6), 455-459. Fillingham, J. K., Hodgson, C., Sage, K., & Lambon Ralph, M. A. (2003). The application of errorless learning to aphasic disorders: A review of theory and practice. Neuropsychological Rehabilitation, 13(3), 337-363. Fillingham, J. K., Sage, K., & Lambon Ralph, M. A. (2005a). Treatment of anomia using errorless versus errorful learning: Are frontal executive skills and feedback important? International Journal of Language & Communication Disorders, 40(4), 505-523. Fillingham, J. K., Sage, K., & Lambon Ralph, M. A. (2005b). Further explorations and an overview of errorless and errorful therapy for aphasic word-finding difficulties: The number of naming attempts during therapy affects outcome. Aphasiology, 19(7), 597-614. Fillingham, J. K., Sage, K., & Lambon Ralph, M. A. (2006). The treatment of anomia using errorless learning. Neuropsychological Rehabilitation, 16(2), 129-154. Fink, R. B., Brecher, A., Schwartz, M. F., Randall, R. R. (2002). A computer-implemented protocol for treatment of naming disorders: Evaluation of clinician-guided and partially self-guided instruction. Aphasiology, 16(10/11), 1061-1086. 159  Foster, A. M., Worrall, L. E., Rose, M. L., & O’Halloran, R. (2015). ‘I do the best I can’: An in-depth exploration of the aphasia management pathway in the acute hospital setting. Disability and Rehabilitation, 1-15. Fridriksson, J., Holland, A. L., Beeson, P., & Marrow, L. (2005). Spaced retrieval treatment of anomia. Aphasiology, 19(2), 99-109. Fridriksson, J., Baker, J. M., Whiteside, J., Eoute, D., Moser, D., Vesselinov, R., Rorden, C. (2009). Treating visual speech perception to improve speech production in nonfluent aphasia. Stroke, 40, 853-858. Fridriksson, J., Hubbard, H. I., Hudspeth, S. G., Holland, A. L., Bonilha, L., Fromm, D., & Rorden, C. (2012). Speech entrainment enables patients with Broca’s aphasia to produce fluent speech. Brain, 3815-3829. Fucetola, R., Connor, L. T., Perry, J., Leo, P., Tucker, F. M., & Corbetta, M. (2006). Aphasia severity, semantics, and depression predict functional communication in acquired aphasia. Aphasiology, 20(5), 449-461. Garrett, K. L., & Lasker, J. P. (2013). Adults with severe aphasia and apraxia of speech, in D. R. Beukelman & P. Mirenda (Eds.), Augmentative & alternative communication: Supporting children & adults with complex communication needs (4th ed.) (pp. 405-445). Baltimore, MA: Paul H. Bookes Publishing Co. Grimes, D. A., & Schulz, K. F. (2005). Refining clinical diagnosis with likelihood ratios. Lancet, 365, 1500-1505. Goodglass, H., Kaplan, E., & Baressi, B. (2001). Boston Diagnostic Aphasia Examination (3rd ed.). Philadelphia: Lippincott Williams & Wilkins. 160  Harry, A., & Crowe, S. F. (2014). Is the Boston Naming Test still fir for purpose? The Clinical Neuropsychologist, 28(3), 486-504. Helm-Estabrooks, N. (2001). Cognitive Linguistic Quick Test. London, England: Harcourt Assessment. Herbert, R., Best, W., Hickin, J., Howard, D., & Osborne, F. (2003). Combining lexical and interactional approaches to therapy for word finding deficits in aphasia. Aphasiology, 17(12), 1163-1186. Herbert, R., Hickin, J., Howard, D., Osborne, F., & Best, W. (2008). Do picture-naming tests provide a valid assessment of lexical retrieval in conversation in aphasia? Aphasiology, 22, 184-203. Hersh, D., & Cruice, M. (2010). Beginning to teach the end: The importance of including discharge from aphasia therapy in the curriculum. International Journal of Language & Communication Disorders, 45(3), 263-274. Hickin, J., Best, W., Herbert, R., Howard, D., & Osborne, F. (2002). Phonological therapy for word-finding difficulties: A re-evaluation. Aphasiology, 16(10/11), 981-999. Hinckley, J. J., & Douglas, N. F. (2013). Treatment fidelity: Its importance and reported frequency in aphasia treatment studies. American Journal of Speech-Language Pathology, 22(2), pS279. Holland, A. (2007). Counseling/coaching in chronic aphasia: Getting on with life. Topics in Language Disorders, 27(4), 339-350. Holland, A. L., Weinberg, P., & Dittelman, J. (2012). How to use apps clinically in the treatment of aphasia. Seminars in Speech and Language, 33(3), 223-233. 161  Hoover, E. L., & Carney, A. (2014). Integrating the iPad into an intensive, comprehensive aphasia program. Seminars in Speech and Language, 35(1), 25-37. Horton, S. (2006). A framework for description and analysis of therapy for language impairment in aphasia. Aphasiology, 20(6), 528-564. Horton, S. (2007). Topic generation in aphasia language therapy sessions: Issues of identity. Aphasiology, 21(3/4), 283-298. Horton, S. (2008). Learning-in-interaction: Resourceful work by people with aphasia and therapists in the course of language impairment therapy. Aphasiology, 22(9), 985-1014. Horton, S, & Byng, S. (2000). Examining interaction in language therapy. International Journal of Language and Communication Disorders, 35(3), 355-375. Horton, S., Byng, S., Bunning, K., & Pring, T. (2004). Teaching and learning speech and language therapy skills: The effectiveness of classroom as clinic in speech and language therapy student education. International Journal of Language & Communication Disorders, 39(3), 365-390.  Hula, W., Donovan, N. J., Kendall, D. L., & Gonzalez-Rothi, L. J. (2010). Item response theory analysis of the Western Aphasia Battery. Aphasiology, 24(11), 1326-1341. Kaplan, E. Goodglass, H, & Weintraub, S. (2001). Boston Naming Test (2nd ed.). Philadelphia, PA: Lippincott Williams & Wilkins. Katz, R. C. (2010). Computers in the treatment of chronic aphasia. Seminars in Speech and Language, 31(1), 34-41. Katz, R. C., Hallowell, B., Code, C., Armstrong, E., Roberts, P., Pound, C., & Katz, L. (2000). A multinational comparison of aphasia management practices. International Journal of Language and Communication Disorders, 35(2), 303-314. 162  Keetch, K. M., & Lee, T. D. (2007). The effect of self-regulated and experimenter-imposed practice schedules on motor learning for tasks of varying difficulty. Research Quarterly for Exercise and Sport, 78(5), 476-486. Kertesz, A. (2007). The Western Aphasia Battery–Revised. San Antonio, TX: Psych Corp. Kimabrow, M. L. (2007). Integrating life participation approaches to aphasia treatment with adult learning theory. Topics in Language Disorders, 27(4), 318-323. Kiran, S. (2012). What is the nature of poststroke language recovery and reorganization? ISRN Neurology, 1-13. Kiran, S., Des Roches, C., Balachandran, I, & Ascenso, E. (2014). Individualized treatment workflow using an iPad-based software platform. Seminars in Speech and Language, 35(1), 38-50. Kiran, S., & Thompson, C. K. (2003). The role of semantic complexity in treatment of naming deficits: Training semantic categories in fluent aphasia by controlling exemplar typicality. Journal of Speech, Language, and Hearing Research, 46, 608-622. Kleim, J. A., & Jones, T. A. (2008). Principles of experience-dependent neural plasticity: Implications for rehabilitation after brain damage. Journal of Speech, Language and Hearing Research, 51, S225-S239. Kurland, J. (2014). Preface: iRehab in aphasia treatment. Seminars in Speech and Language, 35(1), 3-4. Kurland, J., Wilkins, A. R., & Stokes, P. (2014). iPractice: Piloting the effectiveness of a tablet-based home practice program in aphasia treatment. Seminars in Speech and Language, 35(1), 38-50. 163  Laganaro, M., Di Pietro, M, & Schnider, A. (2006). Computerised treatment of anomia in acute aphasia: Treatment intensity and training size. Neuropsychological Rehabilitation, 16, 630-640. Lakshminarayan, K., Anderson, D. C., Jacobs, D. R., Barber, C. A., & Luepker, R. V. (2009). Stroke rates: 1980-2000: The Minnesota stroke survey. American Journal of Epidemiology, 169(9), 1070-1078. Lalor, E., & Cranfield, E. (2004). Aphasia: A description of the incidence and management in the acute hospital setting. Asia Pacific Journal of Speech, Language and Hearing, 9, 129-136. Lam, J. M. C. & Wodchis, W. P. (2010). The relationship of 60 disease diagnoses and 15 conditions to preference-based health-related quality of life in Ontario hospital-based long-term care residents. Medical Care, 48(4), 380-387. Lambon Ralph, M. A., Snell, C., Fillingham, J. K., Conroy, P., & Sage, K. (2010). Predicting the outcome of anomia therapy for people with aphasia post CVA: Both language and cognitive status are key predictors. Neuropsychological Rehabilitation, 20(2), 289-305. Latimer, N. R., Dixon, S., & Palmer, R. (2013). Cost-utility of self-managed computer therapy for people with aphasia. International Journal of Technology Assessment in Health Care, 29(4), 402-409. Lavoie, M., Routhier, S., Légaré, A., & Macoir, J. (2016). Treatment of verb anomia in aphasia: Efficacy of self-administered therapy using a smart tablet. Neurocase, 22(1), 109-118. Leonard, C., Laird, L., Burianová, H., Graham, S., Grady, C., Simic, T., & Rochon, E. (2015). Behavioural and neural changes after a “choice” therapy for naming deficits in aphasia: Preliminary findings. Aphasiology, 29(4), 506-525. 164  Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage Inc. Luck, A. M., & Rose, M. (2007). Interviewing people with aphasia: Insights into method adjustments from a pilot study. Aphasiology, 21(2), 208-224. Mason, C., Nickels, L., McDonald, B., Moses, M., Makin, K., & Taylor, M. (2011). Treatment of word retrieval impairments in aphasia: Evaluation of a self-administered home programme using personally chosen words. Aphasiology, 25(2), 245-268. Mayer, J. F., & Murray, L. L. (2003). Functional measures of naming in aphasia: Word retrieval in confrontation naming versus connected speech. Aphasiology, 17(5), 481-497. McCall, D. (2012). Steps to success with technology for individuals with aphasia. Seminars in Speech and Language, 33(3), 234-242. McKelvey, M. L., Hux, K., Dietz, A., & Beukelman, D. R. (2010). Impact of personal relevance and contextualization on word-picture matching by people with aphasia. American Journal of Speech-Language Pathology, 19, 22-33. McKissock, S., & Ward, J. (2007). Do errors matter? Errorless and errorful learning in anomic picture naming. Neuropsychological Rehabilitation, 17(3), 355-373. McNaughton, D., & Light, J. (2013). The iPad and mobile technology revolution: Benefits and challenges for individuals who require augmentative and alternative communication. Augmentative and Alternative Communication, 29(2), 107-116. McNeil, M. R., & Pratt, S. R. (2001). Defining aphasia: Some theoretical and clinical implications of operating from a formal definition. Aphasiology, 15(10/11), 901-911. Middleton, E. L., & Schwartz, M. F. (2012). Errorless learning in cognitive rehabilitation: A critical review. Neuropsychological Rehabilitation, 22(2), 138-168. 165  Newman, S., Steed, L., & Mulligan, K. (2004). Self-management interventions for chronic illness. Lancet, 364, 1523-1537. Mortley, J., Wade, J., & Enderby, P. (2004). Superhighway to promoting a client-therapist partnership? Using the internet to deliver word-retrieval computer therapy, monitored remotely with minimal speech and language therapy input. Aphasiology, 18(3), 193-211. Nadeau, S. E., Gonzalez Rothi, L. J., & Rosenbek, J. (2008). Language rehabilitation from a neural perspective, in R. Chapey (Ed.), Language intervention strategies in aphasia and related neurogenic communication disorders (5th ed.) (pp. 689-735). Baltimore, MA: Lippincott Williams & Wilkins. Nickels, L. (2002a). Therapy for naming disorders. Revisiting, revising, and reviewing. Aphasiology, 16(10/11), 935-979. Nickels, L. (2002b). Improving word-finding: Practice makes (closer to) perfect? Aphasiology, 16(10/11), 1047-1060. Nicholas, L. E., & Brookshire, R. H. (1993). A system for quantifying the informativeness and efficiency of the connected speech of adults with aphasia. Journal of Speech and Hearing Research, 36, 338-350. Nicholas, L. E., & Brookshire, R. H. (1995).Presence, completeness, and accuracy of main concepts in the connected speech of non-brain-damaged adults and adults with aphasai. Journal of Speech and Hearing Research, 38, 145-156. Nicholas, M., Sinotte, M., & Helm-Estabrooks, N. (2005). Using a computer to communicate: Effect of executive function impairment in people with severe aphasia. Aphasiology, 19(10/11), 1052-1065. 166  Palmer, R., Enderby, P., Cooper, C. C., et al. (2012). Computer therapy compared with usual care for people with long-standing aphasia poststroke: a pilot randomized controlled trial. Stroke, 43, 1904-1911. Palmer, R., Enderby, P., & Paterson, G. (2013). Using computers to enable self-management of aphasia therapy exercises for word finding: The patient and carer perspective. International Journal of Language & Communication Disorders, 48(5), 508-521. Pedersen, P. M., Vinter, K., & Olsen, T. S. (2001). Improvement of oral naming by unsupervised computerized rehabilitation. Aphasilogy, 15(2), 151-169. Peladeau, N., Forget, J., & Gagne, F. (2003). Effect of paced and unpaced practice on skill application and retention: How much is enough? American Educational Research Journal, 40(3), 769-801. Petheram, B. (1996). The behaviour of stroke patients in unsupervised computer-administered aphasia therapy. Dsiability and Rehabilitation, 18(1), 21-26. Pollock, A., St. George, B., Fenton, M., & Firkins, L. (2012). Top ten research priorities relating to life after stroke. The Lancet Neurology, 11, 209. Ramsberger, G., & Marie, B. (2007). Self-administered cued naming therapy: A single-participant investigation of a computer-based therapy program replicated in four cases. American Journal of Speech-Language Pathology, 16, 343-358. Ramsberger, G., & Massamer, P. (2014). Best practices for incorporating non-aphasia-specific apps into therapy. Seminars in Speech and Language, 35(1), 17-24. Ratner, N. B. (2006). Evidence-based practice: An examination of its ramifications for the practice of speech-language pathology, Language, Speech, and Hearing Services in Schools, 37, 257-267.  167  Raymer, A. M., Beeson, P., Holland, A., Kendall, D., Maher, L. M., Martin, N., . . . Gonzalez Rothi, L. J. (2008). Translational research in aphasia: From neuroscience to neurorehabilitation. Journal of Speech, Language, and Hearing Research, 51, S259-S275. Raymer, A., Strobel, J., Prokup, T., Thomason, B., & Reff, Kelli-Lynn (2010). Errorless versus errorful training of spelling in individuals with acquired dysgraphia. Neuropsychological Rehabilitation, 20(1), 1-15. Renvall, K., Nickels, L., & Davidson, B. (2013a). Functionally relevant items in the treatment of aphasia (part I): Challenges for current practice. Aphasiology, 27(6), 636-650. Renvall, K., Nickels, L., & Davidson, B. (2013b). Functionally relevant items in the treatment of aphasia (part II): Further perspectives and specific tools. Aphasiology, 27(6), 651-677. Robey, R. R. (1998). A meta-analysis of clinical outcomes in the treatment of aphasia. Journal of Speech and Hearing Research, 41(1), 172-187. Robey, R. R., & Schultz, M. C. (1998). A model of conducting clinical-outcome research: An adaptation of the standard protocol for use in aphasiology. Aphasiology, 12(9), 787-810. Robey, R. R., Schultz, M. C., Crawford, A. B., & Sinner, C. A. (1999). Single-subject clinical-outcome research: Designs, data, effect sizes, and analysis. Aphasiology, 13(6), 445-473. Rohrer, D., Taylor, K., Pahler, H., Wixted, J. T., & Cepeda, N. J. (2005). The effect of overlearning on long-term retention. Applied Cognitive Psychology, 19, 361-374. Rohrer, D., & Taylor, K. (2006). The effects of overlearning and distributed practice on the retention of mathematics knowledge. Applied Cognitive Psychology. 20, 1209-1224. Routhier, S., Bier, N., & Macoir, J. (2016). Smart tablet for smart self-administered treatment of verb anomia: two single-case studies in aphasia. Aphasiology, 30(2-3), 269-289. 168  Salter, K., Teasell, R., Foley, N., & Allen, L. (2013). Evidence-based review of stroke rehabilitation. Aphasia. Version 16. Canadian Stroke Network 9. Retrieved from www.ebrsr.com (accessed June 21, 2016). Shaddon, B. B. (2005). Aphasia as identity theft: Theory and practice. Aphasiology, 19(3/4/5), 211-223. Simmons-Mackie, N., & Damico, J. (1999). Social role negotiation in aphasia therapy: Competence, incompetence, and conflict. In D. Kavorsky, J. Duchan, & M. Maxwell (Eds.), Constructing (in)competence: Disabling evaluations in clinical and social interaction (pp. 313-342. Mahwah, NJ: Lawrence Erlbaum Associates Inc. Snell, C., Sage, K., & Lambon Ralph, M. A. (2010). How many words should we provide in anomia therapy? A meta-analysis and a case series study. Aphasiology, 24(9), 1064-1094. Szabo, G., & Dittelman, J. (2014). Using mobile technology with individuals with aphasia: Native iPad features and everyday apps. Seminars in Speech and Language, 35(1), 5-16. Teasell, R., Foley, N., Hussein, N., & Speechley, M. (2013). Evidence-based review of stroke rehabilitation. The elements of stroke rehabilitation. Version 16. Canadian Stroke Network 9. Retrieved from www.ebrsr.com (accessed June 21, 2016). Treutwein, B. (1995). Minireview: Adaptive psychophysical procedures. Vision Research, 35(17), 2503-2522. van de Sandt-Koenderman, M. (2011). Aphasia rehabilitation and the role of computer technology: Can we keep up with modern times? International Journal of Speech-Language Pathology, 13(1), 21-27. 169  van Hees, S., Angwin, A., NcMahon, K., & Copland, D. (2013). A comparison of semantic feature analysis and phonological components analysis for the treatment of naming impairments in aphasia. Neuropsychological Rehabilitation, 23(1), 102-132. Verna, A., Davidson, B., & Rose, T. (2009). Speech-language pathology services for people with aphasia: A survey of current practice in Australia. International Journal of Speech-Language Pathology, 11(3), 191-205. Viera, A. J., & Garrett, J. M. (2005). Understanding interobserver agreement: The Kappa statistic. Family Medicine, 37(5), 360-363. Votruba, K. L., Rapport, L. J., Whitman, D., Johnson, A., & Langenecker, S. (2013). Personality differences among patients with chronic aphasia predict improvement in speech-language therapy. Topics in Stroke Rehabilitation, 20(5), 421-431. Wade, J., Mortley, J., & Enderby, P. (2003). Talk about IT: Views of people with aphasia and their partners on receiving remotely monitored computer-based word finding therapy. Aphasiology, 17(11), 1031-1056. Wambaugh, J., & Shuster, L. (2008).The nature and management of neuromotor speech disorders accompanying aphasia, in R. Chapey (Ed.), Language intervention strategies in aphasia and related neurogenic communication disorders (5th ed.) (pp. 1009-1042). Baltimore, MA: Lippincott Williams & Wilkins. Wambaugh, J. L., Wright, S., Nessler, C., & Mauszycki, S. C. (2014). Combined aphasia and apraxia of speech treatment (CAAST): Effects on a novel therapy. Journal of Speech, Language, and Hearing Research, 57, 2191-2207. 170  Warren, S. F., Fey, M. E., & Yoder, P. J. (2007). Differential treatment intensity research: A missing link to creating optimally effective communication interventions. Mental Retardation and Developmental Disabilities, 13, 70-77. Wertz, & Katz, (2004). Outcomes of computer-provided treatment for aphasia. Aphasology, 18(3), 229-244. Weill-Chounlamountry, A., Capelle, N., Tessier, C., & Pradat-Diehl, P. (2013). Multimodal therapy of word retrieval disorder due to phonological encoding dysfunction. Brain injury, 27(5), 620-631. Wilson, M. (1988). MRC psycholinguistic database: Machine-usable dictionary, version 2.00. Behavior Research Methods, Instruments, & Computers, 20(1), 6-10. Wisenburn, B., & Mahoney, K. (2009). A meta-analysis of word-finding treatments for aphasia. Aphasiology, 23(11), 1338-1352. Zambansen, A., & Thiel, A. (2014). Recent advances in the treatment of post-stroke aphasia. Neural Regeneration Research, 9(7), 703-706. Zheng, C., Lynch, L., & Taylor, N. (2016). Effect of computer therapy in aphasia: A systematic review. Aphasiology, 30, 2-3, 211-244. 171  Appendices Appendix A  Summary of variables in feasibility analysis Research question   sub-components Variables Measures 1. Word set development 1. Time and effort to nominate words, collect photographs, and develop cues using different nomination methods 2. Attrition rate of different nomination methods across word set development stages 3. Proportion of study word sets from different nomination methods 4. Composition of PN word sets 5. Participants’ perception of the usefulness of word sets Researcher session notes   Probe picture naming   Researcher session notes  Researcher session notes Participant-perspective interview and check-ins 2. Adherence 1. Equipment returned, operational, and undamaged at end of study 2. Therapy picture-naming trials practiced independently 3. Proportion of probe picture-naming trials cued 4. Attendance at probe sessions 5. Visual inspection of dose-per-therapy day distribution 6. Actual-to-expected ratios of mean dose, frequency, and cumulative dose per therapy word per week 7. Participants’ reported therapy process Researcher session notes  NT data logging system, therapy audio recordings NT data logging system  Researcher session notes NT data logging system  NT data logging system   Participant-perspective interview and check-ins 3. Accessibility 1. Navigation scoring per navigation step 2. Five-number summaries of inter-selection interval distributions per navigation step 3. Participants’ reported accessibility of therapy 4. Accessibility of NT’s word management system NT data logging system  NT data logging system   Participant-perspective interview and check-ins Researcher session notes 4. Remote progress monitoring 1. Categorization accuracy of NT’s self-scoring system in terms of positive and negative likelihood ratios NT data logging system, probe picture naming Note. PN = participant-nominated, NT = Naming Therapy©.   172  Research question   sub-components Variables Measures 5. Acceptability 1. Visual inspection of therapy session start time distribution 2. Five-number summaries of trial, trial series, session duration, and weekly time commitment distributions 3. Total time commitment 4. Participants’ reported acceptability of therapy NT data logging system  NT data logging system   NT data logging system Participant-perspective interview and check-ins Note. PN = participant-nominated, NT = Naming Therapy©.   173  Appendix B  Summary of variables in therapeutic effect analysis Research question   sub-components Variables Measures 1. Absolute effects 1. Number of words acquired, excluded, practiced, unpracticed during therapy 2. Number of therapy staggerings 3. Number of words acquired at post-therapy, including number of PN words among them  Researcher session notes  Researcher session notes Researcher session notes, probe picture naming  2. Relative effect comparisons: Naming accuracy   1. Visual inspection of multiple-baseline figures 2. Therapy sets pre-to-post SMG and OR 3. Control set pre-to-post SMG and OR 4. Therapy vs. control sets post SMD and OR 5. Practiced vs. unpracticed monitor set OR   Probe picture naming  Probe picture naming  Probe picture naming Probe picture naming  Probe picture naming Naming acquisition 1. Therapy sets pre-to-post OR 2. Control set pre-to-post OR 3. Therapy vs. control sets post OR 4. Practiced vs. unpracticed monitor set OR 5. Therapy sets therapy-to-post OR Probe picture naming  Probe picture naming  Probe picture naming Probe picture naming  Probe picture naming 3. Secondary outcomes 1. Pre-to-post changes in language impairment    2. Pre-to-post changes in technology use   3. Pre-to-post changes in communication activity/participation  4. Pre-to-post changes in confidence/well-being WAB-R, BNT-2, Cookie Theft Picture, participant-perspective interview and check-ins, researcher session notes Participant-perspective interview and check-ins, researcher session notes Participant-perspective interview and check-ins, researcher session notes Participant-perspective interview and check-ins, researcher session notes Note. SMG = standard mean gain statistic, SMD = standard mean difference statistic, OR = odds ratio, WAB-R = Western Aphasia Battery–Revised, BNT-2 = Boston Naming Test–Second Edition. 174  Appendix C  Summary of variables in therapy process analysis Research question   sub-components Variables Measures 1. Dosage process See Appendix A: 1. Visual inspection of dose-per-therapy day distribution 2. Participants’ reported therapy process See Appendix A: NT data logging system  Participant-perspective interview and check-ins 2. Therapeutic procedure process 1. Proportion of use of each cue 2. Participants’ reported therapy process NT data logging system Participant-perspective interview and check-ins Note. NT = Naming Therapy©.   175  Appendix D  Pre-therapy participant-perspective interview guide Acceptability of technology and therapy: 1. Open-ended: How do you feel about using [specific technology]? a. Probes: What kinds of things do you use [specific technology] for? How often do you use [specific technology]? Have you had any challenges using [specific technology]? 2. Open-ended: How do you think using the iPad/Naming Therapy© to practice on your own will go? a. Probes: How easy or hard do you think it will be to use the iPad/Naming Therapy©/to practice on your own? Can you think of any good things or bad things about using the iPad/Naming Therapy©/practicing on your own? Why? b. Ideas: No face-to-face SLP contact, home, amount of time/practice 3. Open-ended: How do you think therapy with the iPad will compare to your previous SLP therapy/other things you do alone/previous technology use? a. Probes: Do you prefer face-to-face therapy with an SLP or practicing by yourself with an iPad? Why? b. Ideas: Autonomy, therapy dosage, nature of recovery, therapy timing  Process of therapy: 1. Open-ended: How do you think you will fit practicing into your day? a. Probes: When? How often? Time of day? Why?  Effects of therapy: 1. Open-ended: How do you think things will be the same or different after therapy? a. Probes: What do you think this therapy might or might not do for you? Any negative effects? Examples? Why? b. Ideas: naming words in Naming Therapy©, using the iPad/technology generally, communication in daily life, independence in communication, confidence, lost  time, lost gains, negative psychological effects  176  Appendix E  Quotes from pre-therapy participant-perspective interviews Perceptions of technology P1 Very appreciative of mobile technology: “Thank you, thank you, oh God! (left hand on chest, looking up) Huge, huge.”  Frustrated by dependence on it: “But it’s hard because I can’t…speak, so I can’t- (shrugs, points around room and out window)…you know?...(sighs, points to iPad and iPhone) It’s weird because I can’t…ugh frustrating! (laughs).”  Hard to type on computer: “Um, yes…well (hesitant)…yeah, I’m fine. But it’s bad because it’s frustrating because I can’t- (lifts right arm, laughs).”  Computers are too expensive: “Well, money so…”  Doesn’t want computer because needs are met by iPhone/iPad: “Um…well, I’m…I don’t think so because it’s fine, because it’s- (pointing to iPhone/iPad) yeah, because it’s why [would I need one]? Why?! (laughs).”  iPad’s big screen helps; iPhone’s screen is too small: “This [iPad] is really good though actually, it’s fine, because it’s huge (expands arm out, laughs)…Well, because [iPhone] it’s…p-puny, but this is- (expands arm out).”  iPhone is more portable than iPad: “Because it’s…um…I didn’t- oh don’t worry about that (waves iPad away) because I’m…I’mma going there, here, here (points in different directions).”  Difficulty managing technical problems, such as with his Wifi, because of impairments: “Well, because I can’t speak so I’m not-…You know it’s uh…um…you know, it’s weird (mock confusion). Like, I didn’t know (emphasized), so I’m not sure.” P2 Used only basic functions of mobile technology because of his age: “No, I 60 years old (laughs)…No, minimal, minimal uh…uh…uh iPad is uh…texting, everything, it uh…smartphone uh…young people is…talking on the telephone, texting, everything…iPad, smartphone, computer uh, it’s minimal [for me].”  Bought iPad for entertainment and accessing information: “I don’t know…I-I-I…got an iPad and it’s…I don’t know it’s…I want to get…uh iPad…I want to get- it’s um…good uh reasoning, good uh…it’s a good thing…uh…I don’t know (laughs) but uh…entertainment, uh news, and…that’s it.”  iPad’s big screen makes it easier to read than smart phone: “I don’t know…it’s…easier [to read].” Note. Parenthetical text represent relevant non-linguistic information, whereas bracketed text is intended to clarify the intention of participants’ responses when it is ambiguous from their words alone. - = an abrupt stop without falling intonation that would indicate the end of a sentence. … = a notable silent interval. X = unintelligible speech.   177  Perceptions of technology P3 Very positive perception of mobile technology: Researcher: “What do you like about the iPad?” P3: “Everything. (nods head)”  Found iPad easier to use than computer: “A little bit not…a little bit not-, no…not…X iPad was…instrumental but-…I…pad…and another ping-…I…pad…and another thing to do…right there (taps on iPad repeatedly)…iPad…is easi-no…I…iPad is…easier than computer…I-I don’t know [why]…it…another thing to do…” Expectations of self-directed practice P1 Uncertain what to expect: “But I’m not sure because I’m not sure.”  Confident that can practice by himself because he already does so: “That’s fine because it’s- (demonstrates using various therapy apps on iPad).”  Liked the idea of challenging himself by practicing over and over: “But’s it’s…Me? Like, this one here? (pointing to researcher’s iPad) Oh perfect! That’s better because it’s hard, hard, hard, because it’s I’m trying, I’m trying, I’m trying.”  Hopeful that therapy would generally improve his language; cited past improvement, from how he was to how he is now, as evidence: “Oh yeah, I’m sure, that’s better…it should- zzzzzzzzz (traces increasing slope through air with finger)…oh yeah, easy. Well, because I’m not…look, this (points to himself)…I didn’t, you know (shakes head)…But it’s ‘Bluh bluh bluh bluh’ (pointing at mouth, mock muffled speech), so…but it’s ‘Hi, how are you, boom-ba-boom!’ So it’s better, huge…It better [improve]! (laughs)” P2 Confident that can do therapy/practice by himself: “Oh I…going ok, yeah…I don’t know [why]! (laughs)…but it’s Naming [Therapy]… (shrugs) I can do it…it’s hard, but it’s…doable (laughs).”  Hopeful that using iPad in therapy would help him learn to use it more in the future: “Uh…iPad is technology…and…uh…uh it’s…uh…I can more use the iPad in the future…In the future…I will use it more…I assume so…I can use it [therapy] for uh…uh learning…about…the iPad.”  Felt therapy might also help with naming and have broader effects, but very tentative: “Different [after therapy]…I don’t know, but it’s…It will be different…I assume so, the better…I assume so uh…it will be better, it will be better uh…naming um…naming, describing and um…I assume (emphasized) so…reading, writing uh…um…I assume so (laughs)…reading, writing and um…uh…conversation abilities.” Note. Parenthetical text represent relevant non-linguistic information, whereas bracketed text is intended to clarify the intention of participants’ responses when it is ambiguous from their words alone. - = an abrupt stop without falling intonation that would indicate the end of a sentence. … = a notable silent interval. X = unintelligible speech.   178  Appendix F  iPad-Naming Therapy© navigation path for therapy picture naming Step User actions 1 Open cover/place iPad in standa 2 Push iPad on/off buttona (to wake or turn on iPad) 3 Swipe to the right on iPad lock screen  4 Select NT icon on iPad Home Screen 5 Select Naming Practice on Home Screen 6 Select Category Right Arrow Button on Category Screen 7 Select Cuesa on Trial Screen 8 Select Correct Button or Incorrect Button + Next Button on Trial Screen 9 Repeat 7-8 for entire trial series. In last trial, select Correct Button or Incorrect Button + Home Button on Trial Screen 10 Select Try Again (repeat steps 7-9) or Done on Popup Screen 11 Push iPad Home Button on Home Screen 12 Push iPad on/off buttona (to put iPad to sleep or turn it off) 13 Close covera 14 Charge iPada Note. Bolded selections and screens are those recorded by the data logging system. Italicized parenthetical remarks provide additional information about functionality. NT = Naming Therapy©, a = Optional/as needed.   179  Appendix G  Optional Naming Therapy© cueing system Cue Cueing type Cueing-support hierarchy No cue N/A 1 Definition Semantic 2 First Letter Orthographic 3 Written Word Orthographic 4 Phrase Completion Semantic 5 First Sound Phonological 6 Spoken Word Phonological 7 Note. No cue = The user attempts to name the photograph without first selecting a cue from the Trial Screen; Definition = an audio recording of a phrase or sentences describing the target word is presented; First Letter = the initial letter of the target word with blanks for the remaining letters is presented above the photograph, and remains there until Written Word is selected or the trial ends; Written Word = the full written form of the target word is presented above the photograph, and remains there until First Letter is selected or the trial ends. Phrase Completion = an audio recording of a phrase or sentence fragment with the last word missing is presented, in which the target word would be appropriate for finishing the fragment; First Sound = an audio recording of the initial onset phoneme(s) or initial onset plus nucleus phonemes of the target word is/are presented; Spoken Word = an audio recording of the full phonological form of the target word is presented. Cueing-support hierarchy = the hypothesized ordinal relationship between the cues in terms of how easily they elicit the target word (i.e., 1 = least supportive, 7 = most supportive), according to the instructions provided on Naming Therapy©’s Info Screen.   180  Appendix H  Five-number summaries of lexical characteristics for the NT sets, the BOSS sets, and the participant-nominated word sets   NT set 1 NT set 2 BOSS P1’s PN P2’s PN P3’s PN  n 150 172 83 30 47 53 Characteristics Percentiles       Lg10CD Minimum 2.58 1.46 0.48 N/A N/A N/A  First quartile 2.88 2.12 1.13 N/A N/A N/A  Median 3.09 2.40 1.51 N/A N/A N/A  Third quartile 3.27 2.57 1.78 N/A N/A N/A  Maximum 3.92 3.81 3.11 N/A N/A N/A NPHON Minimum 1 2 2 3 4 2  First quartile 3 4 4 4 7 4  Median 4 4 6 5 8 4  Third quartile 4 6 7 6 11 6  Maximum 9 11 9 11 19 11 Note. NT and BOSS sets were designed to differ from one another in terms of word frequency, with NT set 1 having the highest values and the BOSS set the lowest values. Lg10CD = logarithmic proportion of films/episodes containing a word (i.e., word frequency); Lg10CD benchmarks: 0.95 = 0.1%, 1.93 = 1.0%, 2.92 = 10.0%, 3.92 = 100% (Brysbaert & New, 2009). NPHON = number of phonemes in a word (i.e., word length). NT = Naming Therapy©, BOSS = Bank of Standardized Stimuli, PN = participant-nominated.   181  Appendix I  Development of the NT and BOSS word sets Step NT sets 1 & 2 % BOSS set % Collection 418 100 95 100 Pilot testing 333 80 84 88 SUBTLEX-US entry 327 78 N/A N/A Not used in iPad training 322 77 N/A N/A Note. NT = Naming Therapy©, BOSS = Bank of Standardized Stimuli.   182  Appendix J  Data collected by the Naming Therapy© data logging system Data Description NT Screens and button selections Screens: Home Screen  Category Screen  Trial Screen   Popup Screen  N/A N/A Buttons: Naming Practice, Describe, Naming Test, Flashcards; Info Button, Settings Button, Connect Button, Other Apps Button Category Right Arrow Button, Category Back Button, Category Add/Edit Custom Button Correct Button, Incorrect Button; Definition, First Letter, Written Word, Phrase Completion, First Sound, Spoken Word; Next Button, Previous Button, Home Button Popup Continue, Popup Try Again, Popup E-mail Results, Popup Done Start App iPad Home Button Date The day, month, and year that a screen starting/ending or a user selection was logged on. Time The hour, minute, and second (rounded to six decimal places) that a screen starting/ending or a user selection was logged at. App start latency The time (in seconds rounded to six decimal places) that passed between the app start being logged and the current screen/selection being logged. Trial series start latency The time (in seconds rounded to six decimal places) that passed between the trial series start being logged and the current screen/selection being logged. Trial start latency The time (in seconds rounded to six decimal places) that passed between the trial start being logged and the current screen/selection being logged. Inter-selection interval The time (in seconds rounded to six decimal places) that past between the previous screen/selection being logged and the current screen/selection being logged. Navigation scoring A nominal variable of the appropriateness of each button selection in relation to the current step within the trained navigation path (Appendix F). Button selections that were consistent with this path were rated as ‘adherent’, those that were not consistent with this path but did not interfere with carrying out therapy practice were rated as ‘alternative’, and those that did interfere with carrying out therapy practice were rated as ‘maladaptive’. Note. NT = Naming Therapy©.   183  Appendix K  Naming accuracy and naming acquisition scoring criteria for probe picture naming Variables Description Naming accuracy A dichotomous variable of the accuracy of participants’ uncued attempts to name photographs. Online scoring was completed in person during Phases 2 through 4. Offline scoring was completed blinded to time point using the trial audio recordings collected by the NT data logging system after all probe picture-naming trials had been collected.  Probe picture-naming trials as a whole were scored as accurate on the basis of five criteria: 1. Order: Scoring was based on the final response produced. 2. Latency: Scoring was restricted to responses with onset latencies less than 20 seconds (estimated during online scoring and measured from the start of the audio recording during offline scoring). 3. Intelligibility: The listener was able to recognize a word, however degraded, based on the acoustic signal of the response. 4. Lexical-semantics: The response was lexically the target word established during the selection stage of the word set development phase. 5. Phonology: The response contained no phonological errors (i.e., phonemic substitutions, additions, or deletions). Phonetic distortions that did not cross a phoneme boundary were not considered to be phonological errors. Probe picture-naming trials as a whole were scored as inaccurate if they did not meet all of these criteria.  Circumlocutions and other commentary that did not include a naming response were not scored as naming attempts and thus impacted on this variable only in terms of latency.  Naming acquisition A dichotomous variable of the temporal stability of participants’ uncued attempts to name photographs. Acquisition was scored twice, once during therapy on the basis of the online naming accuracy scoring and once after all probe picture-naming trials had been collected on the basis of offline naming accuracy scoring.  Words were scored as acquired if they demonstrated ≥80% naming accuracy across consecutive probe sessions. In practice, this meant that a word was scored as acquired only if naming accuracy was 100% (i.e., 4/4 probes) across two consecutive probe sessions during the therapy phase, ≥83.33% (i.e., ≥5/6 probes) across three consecutive probe sessions during the therapy phase, or ≥87.5% (i.e., ≥7/8 probes) across the four probe sessions during the post-therapy phase.   184  Appendix L  Post-therapy participant-perspective interview guide Acceptability of technology therapy: 1. Open-ended: How did you find using the iPad/Naming Therapy© to practice on your own? a. Probes: How easy or hard did you find using the iPad/Naming Therapy©/practicing on your own? Can you think of any good things or bad things about using the iPad/Naming Therapy©/practicing on your own? Anything that was missing/could be improved? Would you use an iPad/Naming Therapy© again for therapy? Why? Can you show me on the iPad? b. Ideas: No face-to-face SLP contact, home, amount of time/practice, features of the iPad/Naming Therapy© 2. Open-ended: How did therapy with the iPad compare to your previous SLP therapy/other things you do alone/previous technology use? a. Probes: Do you prefer face-to-face therapy with an SLP or practicing by yourself with an iPad? Why? b. Ideas: Autonomy, therapy dosage, nature of recovery, therapy timing 3. Open-ended: How do you feel about therapy ending? a. Probes: What do you think will happen now that therapy has ended? Why?  Process of therapy: 1. Open-ended: How did you practice therapy on the iPad? a. Probes: When? How often? Time of day? Who with? How did you use cues? Why? Can you show me on the iPad?  Effects of therapy: 1. Open-ended: How are things the same or different now compared to before therapy? a. Probes: What do you think this therapy did or did not do for you? Any negative effects? Examples? Why?  b. Ideas: naming words in Naming Therapy©, using the iPad/technology generally, communication in daily life, independence in communication, confidence, lost  time, lost gains, negative psychological effects  185  Appendix M  Variables used for word set matching Variable Description Baseline naming accuracy Percentage of all the probe picture-naming trials during the baseline phase (i.e., the evaluation and confirmation stages; 2.5.2) that were accurate according to online naming accuracy scoring. Lg10CD Word frequency data was obtained for each NT and BOSS set word in participants’ word sets from the SUBTLEX-US database (Brysbaert & New, 2009). SUBTLEX-US is a free database of lexical variables based on subtitles from American films and television programs and contains 51 million tokens of 74,281 English words coming from 8,388 different films and television episodes. Lg10CD, a logarithmic measure of the proportion of films and episodes containing a word, was used to index word frequency, since Brysbaert and New’s (2009) analyses indicated that it was the best variable in their database to use for matching. The following benchmarks indicate the relationship between Lg10CD and the percentage of the sample containing a word: 0.95 = 0.1%, 1.93 = 1.0%, 2.92 = 10.0%, 3.92 = 100% (Brysbaert & New, 2009). Lg10CD values were not obtained for PN words as many did not have entries in the SUBTLEX-US database and even if they did, these values would likely not accurately represent the familiarity of PN words to participants. NPHON Word length data was obtained for each word in participants’ word sets from the Medical Research Council Psycholinguistic Database, a free database of lexical variables containing 150,837 words (Coltheart, 1981; Wilson, 1988). NPHON, a measure of the number of phonemes in a word based on the phonetic transcriptions from Daniel Jones’ Pronouncing Dictionary of the English Language (Guierre, 1966; cited in Coltheart, 1981), was used to index word length. NPHON values were not available for many words, so missing values were manually entered using the database’s conventions. Participant-nominated words Percentage of the words in a word set that were nominated by participants during the nomination stage (see section 2.5.2) Note. NT = Naming Therapy©, BOSS = Bank of Standardized Stimuli, PN = participant-nominated.   186  Appendix N  Five-number summaries of the characteristics of matched therapy and control word sets for each participant  P1 P2 P3  TS1 TS2 CS TS1 TS2 CS TS1-R TS2-R CS n 23 23 24 21 21 22 20 20 20 PN words (%) 26 26 21 48 43 45 20 15 20 Baseline accurate responses (%)  3  2  3  2  4  4  6  6  8 Lg10CD          Minimum 2.82 2.65 2.65 0.60 0.60 0.60 1.65 1.71 1.86 First quartile 2.86 2.87 2.90 1.23 1.31 1.29 2.21 1.87 2.29 Median 3.08 3.08 3.12 1.59 1.59 1.50 2.57 2.65 2.57 Third quartile 3.31 3.26 3.29 1.81 1.68 1.85 2.85 2.89 2.70 Maximum 3.66 3.90 3.92 2.74 3.10 2.53 3.90 3.35 3.37 NPHON          Minimum 2.00 1.00 2.00 3.00 4.00 4.00 3.00 2.00 2.00 First quartile 3.00 3.00 3.75 5.00 5.00 6.25 4.00 4.00 4.00 Median 4.00 4.00 4.00 7.00 7.00 7.00 5.00 4.50 4.00 Third quartile 5.00 5.00 6.00 8.00 9.00 8.00 6.00 6.25 5.25 Maximum 8.00 11.00 8.00 13.00 13.00 13.00 9.00 10 9.00 Note. PN = participant-nominated. Lg10CD = logarithmic proportion of films/episodes containing a word (i.e., word frequency); Lg10CD benchmarks: 0.95 = 0.1%, 1.93 = 1.0%, 2.92 = 10.0%, 3.92 = 100% (Brysbaert & New, 2009). NPHON = number of phonemes in a word (i.e., word length). TS1 = therapy set 1, TS2 = therapy set 2, CS = control set, TS1-R = revised therapy set 1, TS2 = revised therapy set 2. P3’s therapy word sets were revised due to a clerical error (see section 3.1.3).    187  Appendix O  Participants’ dosage adherence per therapy week for mean dose per word, mean session frequency per word, and mean cumulative dose P1 Mean dose Mean session frequency Mean cumulative dose Week Actual Expected Ratio Actual Expected Ratio Actual Expected Ratio 1 2.03 2.00 1.02 7.00 4.00 1.75 8.61 4.84 1.78 2 1.99 2.00 0.99 3.04 4.00 0.76 12.26 9.68 1.27 3 2.09 2.00 1.04 6.00 4.00 1.50 19.84 14.53 1.37 4 2.00 2.00 1.00 5.00 4.00 1.25 25.89 19.37 1.34 5 2.00 2.00 1.00 6.00 4.00 1.50 33.16 24.21 1.37 6 2.28 2.00 1.13 4.00 4.00 1.00 38.68 29.05 1.33 7 2.03 2.00 1.02 5.00 4.00 1.25 44.84 33.90 1.32 P2 Mean dose Mean session frequency Mean cumulative dose Week Actual Expected Ratio Actual Expected Ratio Actual Expected Ratio 1 1.00 2.00 0.50 1.24 4.00 0.31 0.70 4.54 0.15 2 6.19 2.00 3.10 2.00 4.00 0.50 7.73 9.08 0.85 3 2.43 2.00 1.21 2.00 4.00 0.50 10.49 13.62 0.77 4 1.84 2.00 0.92 1.19 4.00 0.30 11.73 18.16 0.65 5 5.46 2.00 2.73 6.00 4.00 1.50 30.32 24.22 1.25 6 5.92 2.00 2.96 5.00 4.00 1.25 44.14 29.19 1.51 P3 Mean dose Mean session frequency Mean cumulative dose Week Actual Expected Ratio Actual Expected Ratio Actual Expected Ratio 1 3.41 2.00 1.71 7.00 4.00 1.75 12.26 4.10 2.99 2 2.33 2.00 1.17 6.00 4.00 1.50 19.44 8.21 2.37 3 3.33 2.00 1.66 6.00 4.00 1.50 29.67 13.33 2.26 4 4.03 2.00 2.02 6.00 4.00 1.50 42.49 18.46 2.30 5 3.12 2.00 1.56 7.00 4.00 1.75 52.05 23.18 2.25 6 3.56 2.00 1.78 9.00a 5.71a 1.58 63.03 26.46 2.38 Note. Ratio = the actual dosage value relative to the expected value. Bolded values are non-adherent. a = due to a delay in scheduling the final therapy probe session, P3 had the opportunity to practice 10 days instead of 7 days during his final ‘week’. The expected value is adjusted accordingly.   188  Appendix P  Quotes from post-therapy participant-perspective interviews Accessibility/Acceptability of self-directed therapy P1 Found iPad and NT easy to use: “Oh, I’m, no, I’m fine…because it’s boom-boom-boom-boom (rapidly tapping iPad).”  Overall satisfied with therapy: “Um, no it’s really good! Uh…it’s hard um, but it’s uh…but it’s fine, yeah. Yeah, it’s really good.”  Preferred self-directed therapy to SLP-based therapy, but not against it either: “Uh, I don’t think so, no [would not prefer working with a therapist]. Because it’s me and-and you and that’s it (shrugs)…I mean it’s-it’s…it’s uh…like, I can do it [work with a therapist] though, that’s whatever, doesn’t matter, but you (emphasized)…can be this one here (points to iPad)…and then, me (emphasized)…Perfect, that’s it.”  Appreciated working by himself, with only minimal supervision, because it gave him independence: “Um…well, because it’s me (emphasized), it’s myself (emphasized, left hand on chest). That’s it (emphasized, cutting motion with hand). So…um…‘Ok, ok perfect, ok per-, ok per- bye!’ (mock probe session, swiping through photographs, waves goodbye to researcher, laughs) because it’s me (emphasized), because it’s I’m trying, so…”  Disliked SLP-based therapy because therapists take too much control; role-played a therapist-client interaction to illustrate: “Oh, this one! Like this? This one here? (puts hand on iPad): [Therapist:] ‘Oh, sorry, no, uh don-don’t worry, don-here! (rushed, pulls iPad away from researcher, closes cover) [Client:] ‘Oh come on! We can do it!’ (frustrated, mock reaching for iPad) [Therapist:] ‘Oh no, sorry no, sorry.’ (shaking head, gestures ‘stop’ at researcher) [Client:] ‘Wh-why, why?! Come on, please!’ (exasperated, laughs) [Client:] ‘Well, no it’s-’ (points at iPad) [Therapist:] ‘No sorry, can’t.’ [Client:] ‘Why not?!’ [Therapist:] ‘Well, because it’s oo-doo-doo-doo.’ (mockingly pompous voice and hand waving, laughs) [Client:] ‘Oh ok, whatever’ (sceptical, rolls eyes, laughs) [Ends skit] Sorry! (extends hand out to researcher, laughs)”  Does not matter who he gets therapy from, just as long as he gets therapy at all: “Boy, well…hm, I don’t care, really…you know, anything, really…any any anybody (emphasized), it’s fine…so I don’t care really…I don’t care, really, it’s whatever. I mean it’s whatever (emphasized, expands arm out, laughs).”  NT is boring and repetitive, but it does not matter as long as he gets therapy: “Well, no it’s bad, it’s boring…but it’s I don’t care, it’s whatever. Anything, anything, anything.” Note. Parenthetical text represents relevant non-linguistic information, whereas bracketed text is intended to clarify the intention of participants’ responses when it is ambiguous from their words alone. - = an abrupt stop without falling intonation that would indicate the end of a sentence. … = a notable silent interval. X = unintelligible speech. 189  Accessibility/Acceptability of self-directed therapy P2 Satisfied with therapy because he made progress: “It’s great, that’s great! Um…it’s progressing, it’s uh…app, it’s uh…it’s a great app…but uh…um…um…uh…(demonstrates in NT how he has improved on therapy words).”  NT was minimally adequate for making some degree of progress, but he felt he needs a higher level of difficulty to make more progress; however, using the iPad more was progress in itself: “Yeah…uh more intense? No…It’s- it’s ok (emphasized, shrugs), but it’s- I feel…it’s ok, but it’s…ha-harder with uh…uh medium setting [i.e., not hard enough]…and I say ‘Ok’ (shrugs)…and um…the harder I’m going…greater progress I’m going to-…[therapy was] light…I’m…going-, no. I going to-, no, I am…striding, I am…making progress, and I think…it’s…ok…but it’s-it’s…it’s uh…it’s harder and harder, with me…it’s pushing and pushing and pushing and pushing, but uh this way [NT] is…ok (shrugs)…But the iPad is…harder and harder and harder and harder, it’s…it’s…back there (points over shoulder) [in the past] uh…the iPad is being used a lot now, it’s- and that’s progressing.”  Unclear: made analogy to him and researcher in student and teacher roles with therapy as homework; appeared to suggest that he was motivated to do therapy by having supervision to check in on him, but P2 denied this interpretation: “I was…I was goin- I was going to do-, no…It’s ok, and I was saying ‘It’s ok’ and uh…dutiful and…ok, ‘write ‘em down’ (imitates researcher taking notes)…The iPad is-, no, the app is…great, but it’s uh…I’m doing it because you’re (emphasized) doing it…The iPad is…no…uh…I can’t say it today…the dutiful student, me…It’s the dutiful student, ‘Ok, ok, sounds great!’…It was a dutiful student, it- uh, my was a dutiful student…‘Ok, sounds great!’, and going ok (mock swiping through NT photographs)…’Goodbye’ (waves goodbye at researcher) and I say…‘Ok!’ (pulls iPad towards himself, mock swiping through photographs) And then uh…a week gone by, ‘Ok, it’s’ (waves at researcher, laughs).”  Preferred SLP-based therapy over self-directed MTT because of social interaction and because offers comprehensive therapy that can be tailored to his individual level, but also thinks iPad is useful: “uh…uh it was a uh…SLP uh…it was a-, it was a um…(sighs) I can’t say it, but I know it…uh, relationship and studying…uh…the iPad is great…and I use it all the time…but the SLP is…uh…writing and learning and everything…And uh…(shows researcher paper-based homework on sentence processing from SLP-based therapy)…But it’s a tool, too, it’s a-, it’s an i-uh…the iPad is great and it’s a uh…It’s revolutionary the industry.”  iPad is one part of his larger therapy program, each part of which has helped him slowly improve, from how he was when he had his stroke to how he is now: “This [iPad] was-is-…this was a…good thing, but…uh…I’ve gotten better still…with this [iPad]…I-I believe it’s…uh five years ago…and today…uh this one is useful…and uh and um…[his aphasia support group] is useful too and uh…SLP or [his SLP’s name], it’s a useful too, and…it’s progressing…inch by inch by inch…little by little (laughs).” Note. Parenthetical text represents relevant non-linguistic information, whereas bracketed text is intended to clarify the intention of participants’ responses when it is ambiguous from their words alone. - = an abrupt stop without falling intonation that would indicate the end of a sentence. … = a notable silent interval. X = unintelligible speech. 190  Accessibility/Acceptability of self-directed therapy P2 A little dejected that study is coming to an end, because therapy was helpful and he enjoyed interacting with the researcher: “The learning was good…and…you (laughs)…demeanour (laughs)…and that’s it…every week…is…is uh…supposedly ‘working’ with you and I…Supposedly (emphasized)…end of the study, and…that’s it, see you later? (shrugs, frowns)…in-interaction…uh people skills or people-…I’m a people person.” P3 Appreciated therapy/the researcher’s involvement: “Thank you.”  Liked therapy overall because found it helpful: “Yeah, good…a little bit helps.” Perceived effects of therapy P1 Motivated to improve but became frustrated when did not make as much progress as he wanted:  “I’d wanna do it and then…and then you [would come], because it’s better, but it’s (shrugs)…(deep sigh) hard, eh?”  Angry at himself for not improving as much as he wanted: “Oh uh, angry…me (points at himself)…um…frustrating (laughs) so…you know, whatever (shrugs).” P2 Used to use his iPad, but then other things got in the way, so he stopped; thought about using it again, but did not: “I was doing this…iPad…I was doing it a long time ago…I…uh…riding my bike and…uh friends and family and everything…and then back- (closes iPad cover)…packed away…and…oh, uh…three years ago, I think…uh…open the box, and ‘Ok, pff’ (waves iPad away), close the box, and then (laughs) I left it…didn’t use it.”  Had more positive opinion of iPad and using his own more often because of therapy: “Uh, iPad is…working…and um…i-it- iPad is working…greatly uh…I can’t s-uh…The I-, the uh iPad is…uh oh, four or five months ago, the iPad is fine but today it’s great. Uh, it’s easy…uh, it’s fun (laughs)…and uh…the iPad is uh…computer is nonexistent now, it’s (taps iPad repeatedly)…it’s the iPad now. It’s a great thing, it’s-…[used to do only] Basic, a long time ago…This one [taps on iPad] is…this one is right there, it’s…it’s handy…[use it] more in general…iPads, I’m a great advocate…now (emphasized, laughs) and uh…and uh…and that’s it.”  Before therapy, would become frustrated trying to use the computer to access information, but now that he is better at using iPad, uses the computer almost never, because switched to iPad: “Uh…Wikipedia…oh, six months or-…(shrugs) when you started the study…once a-…computer, once a…computer and…the laptock…every six months I…kipedia- I… (mock having difficulty trying to use computer and getting frustrated) And then I…oh, 45 minutes to an hour, it’s…Aw man! It’s forget it! (mock exhaustion)…the iPad uh…the iPad uh…the laptop is basically nonexistent.” Note. Parenthetical text represents relevant non-linguistic information, whereas bracketed text is intended to clarify the intention of participants’ responses when it is ambiguous from their words alone. - = an abrupt stop without falling intonation that would indicate the end of a sentence. … = a notable silent interval. X = unintelligible speech. 191  Perceived effects of therapy P3 Felt that practiced words improved somewhat: “Words…a little bit more words.”  Not sure if therapy resulted in him being able to communicate better because changes are incremental and difficult to see: “I don’t know (shrugs)…it’s a little bit, a little bit, but not…enough…a little bit, a little bit.” Note. Parenthetical text represents relevant non-linguistic information, whereas bracketed text is intended to clarify the intention of participants’ responses when it is ambiguous from their words alone. - = an abrupt stop without falling intonation that would indicate the end of a sentence. … = a notable silent interval. X = unintelligible speech.   192  Appendix Q  Alternative and maladaptive button selections across participants, organized by step in the iPad-Naming Therapy© navigation path  Alternative button selections Maladaptive button selections All Trial Screen: • Selecting the Next Button without scoring • Using the Correct Button in place of the Next Button when scoring a trial as incorrect • Selecting the Incorrect Button before the last response of the trial Trial Screen: • Double-tapping the Correct Button or Next Button (i.e., skipping the next trial in the series) • Ending a trial series prematurely by selecting the Home Button before the last trial P1 Home Screen: • Navigating to Settings or other NT modes Exiting NT: • Ending therapy session by selecting the Other Apps Button on the Home Screen Exiting NT: • Selecting the iPad Home Button before completing the trial series P2 Home Screen: • Navigating to Settings or other NT modes Trial Screen: • Selecting the Prev Button to move back to earlier trials in the series Exiting NT: • Ending therapy session by selecting the Other Apps Button on the Home Screen Exiting NT: • Selecting the iPad Home Button before completing the trial series • Allowing the iPad to time out and go to sleep prematurely in the middle of a trial series P3 Exiting NT: • Allowing the iPad to time out and go to sleep after completing a trial series  Note. NT = Naming Therapy©.   193  Appendix R  Five-number summaries of participants’ inter-selection intervals during therapy for each step in the iPad-Naming Therapy© navigation path Step Percentiles P1 P2 P3 Home Screen Minimum 0.82 0.71 0.90 First quartile 1.32 1.37 1.96  Median 1.87 1.70 5.31  Third quartile 2.89 2.84 10.87  Maximum 32.66 124.29 110.05 Category Screen Minimum 1.22 1.19 1.77 First quartile 1.73 1.55 2.40  Median 1.97 1.74 2.85  Third quartile 2.20 2.18 4.13  Maximum 16.55 14.53 47.20 Two-step scoring Minimum 0.42 0.68 0.72 First quartile 0.48 1.15 1.18  Median 0.50 1.52 1.37  Third quartile 0.50 2.25 1.63  Maximum 0.52 12.94 24.04 Popup Screen Minimum 2.10 1.43 2.36 First quartile 4.20 2.08 2.87  Median 5.35 2.40 3.10  Third quartile 7.39 3.50 4.12  Maximum 14.33 17.14 18.01 Exiting NT Minimum 2.32 1.97 5.84  First quartile 2.80 8.36 121.09  Median 3.69 104.59 121.20  Third quartile 4.83 121.73 121.28  Maximum 16.55 232.50 121.34 Note. All values are in seconds. See Appendix F: Home Screen = Step 5, Category Screen = Step 6, Two-step scoring = Incorrect Button to Next Button interval and Incorrect Button to Home Button interval (Steps 8-9), Popup Screen = Step 10, Exiting NT = Step 11. NT = Naming Therapy©.   194  Appendix S  Five-number summaries of the duration of time participants took in therapy for various time scales Duration Percentiles P1 P2 P3 Trial Minimum 00:00:01.68 00:00:01.53 00:00:01.78  First quartile 00:00:02.77 00:00:02.60 00:00:02.90  Median 00:00:06.17 00:00:04.83 00:00:04.99  Third quartile 00:00:11.50 00:00:10.55 00:00:11.48  Maximum 00:01:02.58 00:04:45.10 00:02:19.96 Trial series Minimum 00:01:38.33 00:00:10.06 00:01:14.65  First quartile 00:02:29.16 00:01:34.98 00:01:51.60  Median 00:03:03.68 00:02:20.13 00:02:14.07  Third quartile 00:03:31.78 00:03:29.76 00:02:40.92  Maximum 00:06:40.19 00:12:25.33 00:05:27.68 Session Minimum 00:02:37.88 00:01:41.88 00:03:49.88  First quartile 00:05:41.15 00:04:38.94 00:06:25.99  Median 00:06:30.70 00:07:44.74 00:07:09.22  Third quartile 00:07:32.46 00:13:33.26 00:08:03.28  Maximum 00:13:37.33 00:35:54.88 00:15:20.36 Weekly time commitment Minimum 00:23:17.67 00:07:30.36 00:47:03.24 First quartile 00:25:11.50 00:15:40.08 01:14:34.49  Median 00:36:42.60 00:39:39.66 01:32:50.38  Third quartile 00:38:20.80 01:13:20.41 01:39:00.78  Maximum 01:03:28.28 01:56:52.74 01:44:46.33 Total time commitment  04:10:33.14 04:55:56.64 08:27:03.77 Note. All values are in hours:minutes:seconds.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0305685/manifest

Comment

Related Items