UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Optimal outcomes of nonlinear phonological intervention Edwards, Susan Melinda 1995

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

Item Metadata

Download

Media
831-ubc_1995-0540.pdf [ 9.91MB ]
Metadata
JSON: 831-1.0087117.json
JSON-LD: 831-1.0087117-ld.json
RDF/XML (Pretty): 831-1.0087117-rdf.xml
RDF/JSON: 831-1.0087117-rdf.json
Turtle: 831-1.0087117-turtle.txt
N-Triples: 831-1.0087117-rdf-ntriples.txt
Original Record: 831-1.0087117-source.json
Full Text
831-1.0087117-fulltext.txt
Citation
831-1.0087117.ris

Full Text

OPTIMAL OUTCOMES OF NONLINEAR PHONOLOGICAL INTERVENTION by SUSAN MELINDA EDWARDS B.Sc. (CD.) University of Western Ontario, 1988 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF MEDICINE School of Audiology and Speech Sciences We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA September 1995 ©Susan Melinda Edwards, 1995 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of fiuDioLo&y + SpeeCU ^c^t^C£S The University of British Columbia Vancouver, Canada DE-6 (2/88) Abstract Efficacy research has been the focus of many phonological intervention projects recently (Bernhardt, 1990,1994; Von Bremen, 1990; Howell & Dean, 1995). The importance of conducting efficacy research is not in dispute; rather, researchers are focusing on how outcomes can be optimized in everyday clinical situations with typical clinicians given time and financial constraints (Fletcher, Fletcher & Wagner, 1988). The purpose of the present study was to optimize and evaluate the effectiveness of nonlinear phonological theory in typical clinical practise. Intervention took place in bi-weekly sessions over a period of 16 weeks. Data were collected to measure intermediate stages of generalization at six points and generalization to conversational speech was measured at three data points. Relative progress/overall gains in terms of Percent Consonant Correct (PCC) and Wordshape Match (WSM) were compared to subjects from Bernhardt (1990,1993a). Results are discussed in terms of nonlinear phonological theory and more recent developments in the area of constraint theory. Further discussion involves the interaction of phonetics and phonology, especially where physical limitations are concerned. Finally, the process of optimization is discussed in terms of effectiveness and research design. Suggestions for future phonology research based on everyday practise are provided. ii Table of Contents Abstract ii Table of Contents iii List of Figures : vi List of Tables vii Acknowledgments : viii 1 Introduction ; 1 1.1 Purpose and Overview 1 1.2 Phonological and Phonetic Disorders 3 1.2.1 Phonological Disorders 3 1.2.2 Phonetic Disorders 5 1.3 Theoretical Foundations 5 1.3.1 Nonlinear Phonological Theory 6 1.3.2 Theoretical Issues in Phonological Development 24 1.4 Efficacy Studies Using Nonlinear Phonology 30 1.4.1 Previous Studies 30 1.5 Moving from Efficacy to Effectiveness 33 1.5.1 Variations In Therapy Research Design 36 1.6 Summary 37 2 Methods 39 2.1 Overview 39 2.2 Subject Description 39 2.2.1 Subject 1: Shaun (Age 3;5 to 4;1) 39 2.2.2 Subject 2: Eddie (Age 3;7 to 4;2) 40 2.3 Research Methods 42 2.3.1 Experimental Design 42 2.3.2 Data Collection 44 2.3.3 Transcription Reliability 44 2.3.4 Analysis 45 2.3.5 Intervention 46 3 Results: Shaun ,.49 3.1 Initial Phonological System (T1) 49 3.1.1 Segmental Analysis 49 3.1.2 Default Use 52 3.1.3 Prosodic Analysis 54 3.2 Block I Goals, Rationale And Progress 55 3.2.1 Goal #1: Root ([+continuant], [+consonantal]) 55 3.2.2 Goal #2: Root [+lateral] ...57 3.2.3 Goal #3: Place (Dorsal) 58 3.2.4 Goal #4: Glide and Liquid Clusters 60 3.2.5 Summary of Block I Progress 62 iii 3.3 Block II Progress on Block I Goals 64 3.3.1 Goal #1: Indirectly Treated 65 3.3.2 Goal #2: Untreated in Singletons/Treated in Clusters 66 3.3.3 Goal #3: Untreated in Singletons/Treated in Clusters 67 3.3.4 Goal #4: Treated -.68 3.4 Block II Goals, Rationale and Progress 70 3.4.1 Goal #5: Word-Initial If I and Nl 70 3.4.2 Goal #6: WI and WF /s/-Clusters 72 3.4.3 Summary of Block II Progress (T2-T3) 76 3.5 General Observation Goals, Rationale and Progress 78 3.5.1 Vowel Observation Goal 78 3.5.2 Wordshape Observation Goal 79 3.6 Summary of Progress T1-T3 81 3.7 Comparison to Original Efficacy Study 82 3.7.1 Percent Consonant Correct: Relative progress 82 3.7.2 Wordshape match: Relative progress 83 4 Eddie Results 85 4.1 Initial Phonological System (T1) 85 4.1.1 Segmental Analysis 85 4.1.2 Default Use 88 4.1.3 Prosodic Analysis 90 4.2 Block I: Goals, Rationale and Progress 92 4.2.1 Goal#l: WI and WF/t/and/d/.... 92 4.2.2 Goal #2: Root [-(-lateral] 94 4.2.3 Goal #3: WI /s/-Clusters 95 4.2.4 Goal #4: Link (Coronal, [+continuant]) 97 4.2.5 Summary of Block I Progress 98 4.3 Block II Progress on Block I Goals 100 4.3.1 Goal #1: Treated (One Session) 101 4.3.2 Goal #2: Treated (One Session) 102 4.3.3 Goal #3: Treated (One Session) 103 4.3.4 Goal #4: Treated (One Session) 104 4.4 Block II Goals, Rationale and Progress 105 4.4.1 Goal #5: Place (Coronal, [+distributed]) 105 4.4.2 Goal #6: Recycle Goals #1-5. 107 4.4.3 Summary of Block II Progress (T2-T3) ; 107 4.5 Intervention Session Data: Block I and II 109 4.5.1 Breakthrough Moments 109 4.6 Summary of Progress T1-T3 110 4.7 Comparison to Original Efficacy Study I l l 4.7.1 Percent Consonant Correct: Relative Progress 112 4.7.2 Wordshape match: Relative progress 112 5 Discussion 113 5.1 Original Questions 113 5.2 Outcomes of NLP Intervention 115 5.2.1 Quantitative Progress on Trained Targets 115 5.2.2 Qualitative Progress on Trained Targets 117 iv 5.2.3 Observation Targets 118 5.2.4 Imitated Versus Spontaneous Productions 120 5.3 Optimization of Intervention... 121 5.3.1 Twice Versus Three Times Weekly Intervention 121 5.3.2 Optimization Issues 122 5.3.3 Summary of Optimization Questions 125 5.4 Informing Phonological Theory. 125 5.4.1 Underlying Specification 126 5.4.2 Segmental and Sequence Constraints 132 5.4.3 Consonant-Vowel Interactions: Shaun 149 5.4.4 Interactions Between Phonology and Phonetics 153 5.4.5 Summary of Theoretical Questions 155 5.4.6 Limitations of the Present Study 155 5.4.7 Directions for Future Research 158 5.5 Conclusions 160 References : 162 A Shaun's Data 170 B Eddie's Data 177 C Intervention strategies 185 D Syllable structure intervention strategies 186 List of Figures Figure 1-1 Hierarchical representation of phonological tiers 7 Figure 1-2 Onset-rime representations 14 Figure 1-3 Moraic representations 14 Figure 1-4 Feature geometry for english consonant system 16 Figure 1-5 Consonant and vowel Place representation 18 Figure 3-1 Feature geometry: Shaun T1 50 Figure 3-2 Examples of Coronal and [+voice] default representations 53 Figure 3-3 Block I progress: Goal #1 56 Figure 3-4 Block I progress: Goal #2 58 Figure 3-5 Block I progress: Goal #3 60 Figure 3-6 Block I progress: Goal #4 62 Figure 3-7 Feature geometry: Shaun T2 64 Figure 3-8 Block II: Goal #4 70 Figure 3-9 Block II: Goal #5 72 Figure 3-10 Block II: Goal #6 75 Figure 3-11 Feature geometry: Shaun T3 77 Figure 3-12 Summary of progress: T1-T3 81 Figure 3-13 PCC progress relative to original efficacy study 83 Figure 3-14 WSM progress relative to original efficacy studies 84 Figure 4-1 Feature geometry: Eddie T l •. 87 Figure 4-2 Block I progress: Goal #1 94 Figure 4-3 Block I progress: Goal #3 96 Figure 4-4 Block I progress: Goal #4 98 Figure 4-5 Feature geometry T2: Eddie 100 Figure 4-6 Block II: goal #5 106 Figure 4-7 Feature geometry T3: Eddie 108 Figure 4-8 Summary of progress T1-T3: Eddie 110 Figure 5-1 Coronal feature spreading 129 Figure 5-2 Adult form versus Shaun's surface production 134 Figure 5-3 [-(-continuant] delinked from Root 134 Figure 5-4 Feature cooccurrence constraints 136 Figure 5-5 Delink-relink repair 136 Figure 5-6 Timing tier constraint 137 Figure 5-7 Double linking repair Shaun 146 vi List of Tables Table 1-1 Proposed consonant specifications for adult English 17 Table 1-2 Proposed vowel specifications for adult English 20 Table 2-1 Background information: Shaun 40 Table 2-2 Background information: Eddie 41 Table 2-3 Intervention plan: Shaun 47 Table 2-4 Intervention plan: Eddie 48 Table 3-1 Summary of segmental analysis: Shaun T1 51 Table 3-2 Default diagrams 52 Table 3-3 Goal #1 example data T1-T3: Shaun 66 Table 3-4 Goal #2 example data T1-T3: Shaun 67 Table 3-5 Goal #3 example data T1-T3: Shaun 68 Table 3-6 Goal #4 example data T1-T3: Shaun 69 Table 3-7 Goal #5 example data T1-T3: Shaun 71 Table 3-8 Goal #6 example data T1-T3: Shaun 74 Table 3-9 Vowel example data T1-T3: Shaun 78 Table 3-10 Vowel development: Shaun 79 Table 3-11 CV, CVC, CVCV and CVCVC gain T1-T3: Shaun 80 Table 3-12 Study comparison data 82 Table 3-13 PCC and WSM comparison scores with original studies 84 Table 4-1 Summary of segmental analysis:Eddie T l 88 Table 4-2 Dorsal default diagrams: Eddie T1 89 Table 4-3 Labial default diagrams: Eddie T l 90 Table 4-4 Goal #1 example data T1-T3: Eddie 102 Table 4-5 Goal #2 example data T1-T3: Eddie 103 Table 4-6 Goal #3 example data T1-T3: Eddie 103 Table 4-7 Goal #4 example data T1-T3: Eddie 104 Table 4-8 Goal #5 example data T1-T3: Eddie 106 Table 4-9 CV, CVC, CVCV and CVCVC gain T1-T3: Eddie 111 Table 5-1 Coronal versus Dorsal productions 128 Table 5-2 Deletion of CI and C2 in clusters 139 Table 5-3 Shaun's /si/ and /sw/ cluster data 140 Table 5^ -4 Eddie's /si/ and /sw/ cluster data 140 Table 5-5 Default value repairs 143 Table 5-6 Double linking repairs: Shaun 145 Table 5-7 Consonant-vowel interactions 150 Table 5-8 Facilitative vowel contexts for Dorsal Place 151 Table 5-9 Cluster development: Eddie T1-T3 183 vii Acknowledgments This thesis represents many people's work and I would like to thank them: • The children and their families, who taught me more about where I thrive and where I need to grow as a clinician. • Dr. Barbara Bernhardt, who gave generously of her time and self. Her vast knowledge of the field of nonlinear phonology was a unique resource. • Lisa Avery, who was a positive influence as colleague and friend at UBC and the Richmond Health Department. Your painstaking editing was a tremendous service. • Dr. John Gilbert, who provided helpful editorial comments and lab space. • Richmond Health Department and the clinicians in the Speech and Hearing Clinic, who shared space and ideas. • The British Columbia Health Research Foundation whose generous studentship funded this research. Thanks also to those who supported my application. A Master's degree doesn't seem like much when everyone around you is doing one, but I thank my family and friends for their enthusiasm and encouragement. I dedicate this work to my mother, Linda Jane Edwards (1940-1993), life-long learner and teacher. Finally, I thank my husband, Andrew Cripps who cajoled and inspired me throughout this lengthy process. viii 1 Introduction "Research on treatment efficacy may save the day because it makes a natural bridge between the requirements of careful research and the needs of clinical practice" (Siegel, 1993, p. 37). 1.1 Purpose and Overview This thesis presents a study designed to optimize and evaluate the effectiveness of nonlinear phonological (NLP) theory in typical clinical practise. The general design follows that of Bernhardt (1990, 1993a) and Bernhardt, Miller, Barton, MacAulay, MacKenzie and Wastie (1992) whose investigations have successfully used NLP theory in intervention studies with children demonstratingsevere phonological disorders. The present study elaborates on Bernhardt's foundational work and applies it within the constraints of typical clinical practise. Two subjects with moderately severe phonological disorders participated and the choice of intervention goals was guided by and measured against claims made in current phonological theory. This study is also explorative. Siegel (1993) commented that efficacy studies in speech-language pathology need to ask not only, "Will intervention work?" but also "Why does it work?" and "How does it work?" In answering the question "Why does it work?", special topics in NLP such as sequential and complexity constraints will be explored with reference to the performance of the subjects during the period of intervention. Further, the interaction between phonology and phonetics was investigated, both in terms of theory and its effect on progress in intervention. The question "How does it work?" can be broken down into questions of technique and design. Techniques used in this study are discussed in detail by Bernhardt (1992b, 1994b) and described briefly in Appendix C and D. 1 Fletcher, Fletcher and Wagner (1988) differentiate between efficacy studies and effectiveness studies in the epidemiological literature. Efficacy researchis conducted under tightly controlled circumstances by highly trained individuals while effectiveness research is conducted by typical clinicians with typical clients under everyday circumstances. Given that NLP theory is effective for intervention in formal efficacy studies, how effective will it be for a typical speech-language pathologist (SLP) working in a clinical setting considering the demands of: 1. An SLP's available time for individual clients (i.e. for transcription, goal selection, intervention preparation, etc.) 2. Clinic policy concerning length and types (e.g., individual or groups) of intervention available due to waiting lists 3. Expanding the SLP's knowledge of NLP theory and intervention applications? This study is a pilot for future studies that will place increasing responsibility on the clinician for assessment, goal selection and intervention based on NLP theory. The goal was to evaluate whether a practising SLP would be successful in the independent application of NLP with support available from an expert in NLP, in this case from Dr. Bernhardt at The University of British Columbia, on an "on-call basis." Section 1.2 provides background information related to phonological and phonetic disorders while Section 1.3 reviews the main tenets of NLP. Section 1.4 and Section 1.5 outline previous efficacy research and issues in designing an effectiveness study, culminating in the proposal of questions related to the present study. 2 1.2 Phonological and Phonetic Disorders 1.2.1 Phonological Disorders A child who experiences difficulty acquiring the sound patterns of the ambient language in the absence of a hearing impairment, structural defects of the oral mechanism or neurological impairment, is usually described as having a phonological delay or disorder (Stoel-Gammon & Dunn, 1985). Children with phonological disorders can exhibit phonetic inventory developmental patterns similar to those of typically developing age peers but may also produce errors expected of younger children (Hodson & Paden, 1991; Grunwell, 1985; Stoel-Gammon & Dunn, 1985). Stoel-Gammon and Dunn (1985) compared children with phonological disorders (ages 2; 7-8; 0) with three typically-developing children (ages 0; 11-1; 5) and noted parallels in segmental development including inventories consisting of stops, nasals, glides and a few fricatives, produced at all places of articulation. Segments appeared in word-initial position more frequently than in word-medial or word-final positions. Stoel-Gammon and Dunn also examined similarities in syllable shapes of children with phonological disorders and typically developing children at the one-word stage of production and reported that Consonant + Vowel (CV) and CVC were the most common ones. Both groups used non-complex (i.e. C vs. CC) and [-continuant] onsets and neither group used developed liquids. Leonard (1992) also found similarities when matching subjects from the two groups for consonant inventory or mean length of utterance. Several authors have described phonological phenomena not usually observed in normal development. These include: 1. Uneven or unusual/idiosyncratic development such that one or more later-developing phonemes may be produced whilst a problem with an earlier-developing phoneme persists 3 2. Unusual patterns/processes such as atypical cluster reduction, initial consonant deletion, glottal replacement, backing, overuse of one segment lor many others, replacing earlier developing sound classes with later-developing ones such as producing fricatives for stops and using epenthetic vowels (Grunwell, 1985; Stoel-Gammon & Dunn, 1985; McCauley, 1993) 3. Variability in the production of targets without indication of the development of a new contrast (Edwards & Bernhardt, 1973; Grunwell, 1985) A child's use of phonetic inventories within the syllable structure of the language may also be restricted (Grunwell, 1985; Stoel-Gammon & Dunn, 1985; Bernhardt, 1992a; Bernhardt & Stoel-Gammon, 1994). Phonologically disordered children sharing similar phonetic inventories may master sounds in different word positions or in sound sequence combinations of varying complexity. Observation of allowed segments and available consonant slots word-internally can provide information concerning constraints on multisyllabic words, intervocalic consonants and onset-coda asymmetries. Stoel-Gammon and Dunn (1985) and McCauley (1993) report that children with phonological disorders are at risk for language disorders. Louko, Edwards and Conture (1990) and Nippold (1990) observed that children with dysfluent speech are more likely to have a phonological disorder than control subjects with no dysfluencies. Whether the child's phonology is an isolated disorder or occurring with another problem, early assessment and intervention are important to reduce the child's risk for negative social experiences or impeded academic development, especially in the area of literacy (Bryant, Bradley, MacLean & Crossland, 1989; Magnusson & Naucler, 1990; Blachman, 1991). 1. "Processes" refer to the Stampean (1969,1973) theory of phonological processes used in addition to pho-netic inventories by speech-language pathologists in the past ten years (Shriberg & Kwiatkowski, 1980; Grunwell, 1985; Stoel-Gammon & Dunn, 1985; Hodson & Paden, 1989) 4 1.2.2 Phonetic Disorders If the child has the phonemic distinction between two sounds (i.e. competence) but is unable to produce that distinction, usually due to an abnormal oral mechanism or other reasons (i.e. performance), this difficulty is generally described as a phonetic control or artieulatory problem rather than a phonological disorder. Ohde and Sharf (1992) discuss this distinction with the example of the substitution of /w/ for /J / versus a variant production of b I (e.g., derhoticized) saying that the first is a linguistic process (i.e. phonological) error while the second is an artieulatory adjustment (i.e. phonetic) error. Phonetic errors may be related to abnormalities in the speech production mechanism (Stengelhofen, 1989) although Dworkin and Culatta (1985) found that comparisons of tongue strength, diadochokinetic rate and oral mechanisms between control subjects and articulation-disordered subjects revealed no significant differences between the two groups! They suggested that no one factor could explain the developmental cause of disordered articulation and that other factors such as auditory discrimination, attention span, memory, intelligence, and possible environmental influences should always be considered. Phonetic limitations on phonological output will be discussed in detail in Section 1.3.2.4. 1.3 Theoretical Foundations This section provides a comprehensive review of the theoretical foundations upon which this thesis is based. Nonlinear Phonological Theory will be reviewed first followed by a more detailed exploration of developmental phonological phenomena and other theoretical issues including Optimality Theory (OT), sequence and complexity constraints, and consonant-vowel interactions. The interactions between phonology and phonetics will also be discussed. The key 5 theoretical and developmental issues in phonology and phonetics as they relate to typically developing and phonologically disordered children will be summarized at the end of the section. 1.3.1 Nonlinear Phonological Theory Explanations and applications of NLP theory, its developmental implications and applications to severely disordered child phonology can be found in several publications (cf. Bernhardt, 1990; Von Bremen, 1990; Goldsmith, 1990; Bernhardt & Gilbert, 1992; Bernhardt, 1992a; Bernhardt, 1992b, Bernhardt & Stoel-Gammon, 1994; Kenstowicz, 1994). The major elements of NLP theory are: 1. Hierarchical representation 2. Autonomy of tiers 3. Associations between tiers (rules, constraints and repairs) 4. Underlying representation and feature specification. 1.3.1.1 Hierarchical Representation NLP theory emphasizes multi-tiered hierarchical representation of words, syllables, segments and features rather than rules and processes of earlier "linear" generative phonology (Chomsky & Halle, 1968; Kenstowicz & Kisseberth, 1979). Within a nonlinear framework, a child's phonology can be analysed in terms of hierarchically ordered phonological tiers. The focus on representations does not eliminate the need for phonological rules (but see discussion of rules in Section 1.3.1.3); however, it does reduce the number of rules needed to describe sound patterns in any language. Separate levels of organization have been put forward for the segmental and prosodic information which may be linked by the skeletal (CV) tier as in Figure 1-1 below. 6 Word tier Foot (F) tier Syllable (o) tier (Onset-Rime tier) Skeletal (CV) tier Segmental tier 'monkey' WORD (CVC.CV) (Strong) cr O R cr (Weak) O R Root Root Root Root Root and other eto features lm if Figure 1-1 Hierarchical representation of phonological tiers 1.3.1.2 Autonomy of Tiers Goldsmith (1976) introduced the autonomy of the various tiers as it relates to discussion of tone in African languages. The tiers discussed most frequently in this thesis are the prosodic tiers (word and syllable structure) and the segmental (feature) tier. The features of a segment "may interact in rules as a whole, or independently" (Bernhardt & Stoel-Gammon, 1994, p. 127). For example, a segment such as HI (specified for Labial, [+cOntinuant]) might surface as [p] if [-(-continuant] is not linked at Root. IfRoot is delinked from the timing tier, the segment will not surface at all. Linking of features through Root to the timing tiers is also autonomous. For example, [+voice] may apply only in onset position rather than in the onset and the coda of syllables. At the prosodic level, there may be a restricted use of syllable structures (e:g., no word initial clusters) even in the presence of fully developed segmental features (Bernhardt, 1992b). Consequently, intervention for children with severely disordered phonologies can selectively target the prosodic tier, the segmental tier or the interaction of the two where necessary (Bernhardt, 1990,1992a, 1992b, 1994c; Bernhardt & Gilbert, 1992, Bernhardt, MacNeill, & Bohlen, 1994). 1.3.1.3 Association Between Tiers: Rules, Constraints and Repairs In NLP theory, the focus is on representations, and as described above, separate levels of organization have been hypothesized for segmental and prosodic information. Representations are hierarchically ordered in tiers and linked formally by association lines to indicate simultaneity of occurrence in real time and allowing several possibilities for association. These include one-to-one associations where each C or V slot links to one segment, one to many associations (e.g., spreading of the features from one segment to more than one C or V slot), and many to one associations (e.g., long segments, diphthongs and geminates) between timing units and segments. Optimally, there should be as few association lines as possible, leading to feature groupings and dependency relations (Bernhardt & Stemberger [in prep]). Principles govern association lines between tiers, constraining and explaining rules that occur. For instance, Goldsmith (1976) stated that the association lines may not cross. In order to discuss feature grouping and associations between features, rules, processes, constraints and repairs must be reviewed. Early generative phonology (Chomsky & Halle, 1968) formulated linear rules of the type "A —> B / C D", where A is the focus of the rule, B is the resulting change, and C and D are the environment in which the process occurs. Although many processes could be represented in this manner, there was no way to explain universal processes or changes to multiple segments; several operations were needed to explain phenomena such as compensatory lengthening which relates to the syllable structure of the language. SPE analyses appealed to rule-ordering that needed to be 8 stated extrinsically rather than recognizing the syllable as a unit of analysis resulting in disjunction in the phonological representation (Durand, 1990). Linear rules have, for the most part, been abandoned by phonologists. A "constraints-plus-rules" approach was proposed by Menn (1978) as a theory of child phonological development. Later Macken and Ferguson (1983) and Macken (1987) suggested that children begin the process of learning language with severe constraints (e.g., sequence constraints) and that individual children use different repair strategies in order to pronounce words. These repair strategies are sometimes referred to within the framework of Stampe's (1969,1973) theory of phonological processes. Macken (1978) discussed sequence constraints in terms of phonetic context and word position, demonstrating with one Mexican-Spanish child that previously allowed segments assimilated when syllable length and complexity were increased. Macken explained this phenomenon with reference to the interaction of cooccurrence constraints and syllable structure as a trade-off between the two tiers. In summary, a constraints-plus-rules approach suggests that surface phonetic constraints (Goldsmith, 1976) and syllable structure constraints (Clements & Keyser, 1983) trigger rules to ocCur as repair strategies. At present, the view first proposed by Paradis (1988,1993) that all phonological rules are repairs motivated by constraints is most widely agreed upon. Optimality Theory (OT) is a relatively new development in the area of phonological theory (McCarthy & Prince, 1993; Prince & Smolensky, 1993; Bernhardt & Stemberger, in prep), and is related to work in the fields of neural networks and connectionist modeling. OT challenges previous theories that posit a "rules only" approach or a "constraints plus rules" approach. Bernhardt and Stemberger (in prep) argue that both child and adult phonological data can be motivated using families of constraints and that this theoretical explanation is less complex than ordered rules. They describe how constraints are ranked, violated and reranked during child 9 phonological development such that the lowest ranked constraint violation is the optimal output candidate. While the constraints as outlined in Optimality Theory may be difficult to determine initially, Bernhardt and Stemberger (in prep) assert that the analysis is most beneficial in the description of the phonology. The focus of OT is on a large, universal set of simultaneously effective constraints rather than rule-ordered processes. These constraints act on output concurrently, with faithfulness to the underlying representation being the most important constraint of all. OT's "constraints only" approach eliminates the need for repairs. The surface representation is as faithful as possible to the underlying representation in the absence of other constraints. Deletion and insertion are no longer due to a rule or process (i.e. a repair) but are consequences of violations of faithfulness constraints. Constraints represent pronunciations that would be difficult but not impossible for a speaker to produce if the constraint were to be violated. OT rank orders the constraints. Highest-ranked constraints are the least likely to be violated, leading to the optimal pronunciation and faithfulness to the underlying representation. The representations of NLP such as feature geometry are preserved in current versions of OT (e.g., McCarthy & Prince, 1993; Prince & Smolensky, 1993; Bernhardt & Stemberger in prep). OT constraints used in the present thesis (based on Bernhardt & Stemberger, in prep) are segmental and complexity constraints. Segmental constraints refer to features or groups of features that may not be permitted in the phonological representation. The following segmental constraints are relevant to the subjects' data in the present study: 1. "Not"1: A family of constraints that can apply to both underlying and default features at Root, Laryngeal and Place (e.g., Not ([+consonantal]) means no true 1. OT constraints are usually represented in bold-face type. In order to comply with manuscript standards pro-hibiting bold-face type, OT constraints will be represented in quotation marks in this thesis. 10 consonants allowed). These constraints are usually pervasive in the system but can apply to certain word positions. 2. "NotCooccurring": A family of constraints dealing with the combinations of features allowed at Root, Laryngeal andPlace(e.g., NotCooccurring ([+consonantal], Place) would result in glottals being produced for consonants with place features. Complexity constraints are sensitive to the sequence of elements and may prohibit sequences at a variety of levels of adjacency. The following complexity constraints (taken from Bernhardt & Stemberger, in prep) are discussed later in terms of Shaun and Eddie's data: 1. "NotTwice": An element may not occur twice if the two tokens are adjacent. 2. "NoSequence (A...B)": Given two segments containing elements A and B, A cannot be the first segment if B is in the second segment. The "NotTwice" constraint may apply at different levels of adjacency such as the level of the Roots (two identical segments not allowed in sequence) or the feature tiers (two identical place features not allowed in sequence). The "NoSequence" constraint is most likely to be active between two adjacent segments that may or may not be surface contiguous. Surface-contiguous sequences include both neighbouring segments and neighbouring prosodic units (e.g., consonant clusters, diphthongs, or CV units at the timing tier). It is important to look at the child's entire system when trying to understand sequence constraints. Apparent variability in the realization of phonological elements may be explained by appealing to sequence constraints but also by the uneven nature of child development. Specified elements are more likely to be implicated in sequence constraints while unspecified elements will 11 be affected by repairs (Stemberger & Stoel-Gammon, 1991). Relative sonority can also influence which elements are affected and what kind of repair is used (Chin, 1994; Bernhardt, 1995). Phonologists and speech-language pathologists rely on the child's output to make hypotheses about how the child's system deals with constraints on that output. A sequence constraint can be seen as applying at the segmental level whereas a complexity constraint applies at the timing tier. For example, if a child has difficulty producing some clusters but not all, it would be likely that his problem was in producing feature sequences at the level of Root or Place. However, if the child could produce no clusters, the problem would be one of a complexity constraint in that two consonant slots could not occur beside each other at the prosodic level regardless of their segmental content. Given that the majority of phonologists are presently using a constraints plus rules approach to OT, Bernhardt and Stemberger' (in prep) framework of constraints plus "repairs" has been adopted for the purposes of this thesis. Bernhardt and Stemberger summarize the rules that were feature changing in earlier process phonology as follows: (1) the addition of an association line to an existing element (i.e. spreading /all assimilations) or inserting an element and then adding an association line, and (2) the deletion of an association line from an existing element (i.e. delinking). Delinking may lead to the presence of a floating feature which is interpreted as violating the Well-Formedness Condition of autosegmental phonology (Goldsmith, 1976). Since all elements have to be linked, this can be accomplished by linking the floating feature to another existing element, inserting a new element for the floating feature to attach to, or deleting the floating element (known as "stray erasure"). The most common process is deletion when an element is delinked. Repairs can also involve a combination of addition and deletion which will be 12 referred to as "delink-relink" repairs. In general, Bernhardt & Stemberger (in prep) state that specified elements in the underlying representation are often subject to constraints while unspecified elements are subject to repairs. 1.3.1.4 Underlying Representation and Feature Specification McCarthy's (1988, p.84) remark "If the representations are right, then the rules will follow," emphasizes the representation-over-rules focus of NLP theory. Many authors have pointed out that, while processes describe what is happening on the surface of an utterance, they do not explain the underlying representations involved (Smith, 1973; Spencer, 1984; Stoel-Gammon & Dunn, 1985). Theories of sub s^yllabic structure representation will now be discussed followed by a presentation of feature specification for consonants and vowels used in the present study. 1.3 1.4.1 Representation of Sub-syllabic Structures Theories pertaining to the description of timing unit organization differ as to whether or not there is an intermediate organizing level between the syllable tier and the segmental tier. Two theories that will be referred to during this thesis are "onset-rime" theory and "moraic" theory. Onset-rime theory was first introduced by Pike and Pike (1947). Others who support it include Halle and Vergnaud (1980), Steriade (1988), and Kaye, Lowenstamm and Vergnaud (1990). According to onset-rime theory (Figure 1-2), the onset (O) dominates the rime (R), which in turn dominates the nucleus (N). The syllable thus branches into a tree structure in which the increasing number of branches represents increasingly complex syllable structure. Branching in the rime is also related to stress assignment in many languages. Branching rimes may attract stress (e.g., CVC) and non-branching rimes (e.g., CV, CVV) may not attract stress. 13 O CT O-cOk N V CV unit \ \ C V c CVC unit N A C W unit Figure 1-2 Onset-rime representations1 Moraic theory also focuses on the "rime" but is concerned with units of weight that attract stress to the syllable (see Figure 1-3). Moras are thought to be the critical constituents of syllables for stress assignment and other phonological phenomena (Hyman,1985; McCarthy and Prince, 1986; Hayes, 1989). Those prevocalic and postvocalic consonants that do not contribute to the weight are simply adjoined to the edges of the syllable. The onset is not considered a constituent as in onset-rime theory but merely a consonant adjoined to the mora node. CT Jv l CV unit vl M M A /\ V V or C CVC unit C W or CVC unit, C2=weightless when C2 has weight Figure 1-3 Moraic representations2 1. According to Kaye, et al. (1992) 2. According to Hayes (1989) 14 1.3.1.4.2 Feature Representation: Consonants Segmental feature organization has been described in terms of "feature geometry"1 (see Figure 1-4) rather than the unorganized feature bundles of earlier Chomsky and Halle (1968) distinctive features. The major organizing nodes in the feature geometry are Root, Laryngeal and Place. Root links the feature geometry to the skeletal (CV) tier. Major class features such as [+sonorant], [+nasal], [-(-consonantal] and [-(-continuant] are linked to Root which dominates the other nodes: Laryngeal and Place (Bernhardt, 1992 a,b). Laryngeal designates glottal characteristics such as [+/-voice] and [-(-/-spread glottis] while Place designates oral cavity characteristics of the segment such as Labial (lips), Coronal (tongue tip and blade) and Dorsal (tongue dorsum/body). Although a topic of some discussion (e.g., Rice & Avery, 1993) , Root, Laryngeal'and Place (including Labial, Coronal and Dorsal) are all treated here as privative features meaning that they are either present or absent. The features that they dominate will be treated as binary meaning that they are either on or off (Bernhardt & Stemberger, in prep). According to early generative theory (Chomsky & Halle, 1968), phonological rules operate on fully specified, binary (i.e. + and -) features. Underspecification means that only feature value information which is unpredictable (i.e. specified) is present in underlying representation. This implies that no feature is specified for both values (+ and -) underlyingly. Instead, predictable (i.e. unspecified) information is left blank in underlying representations and unpredictable (i.e. specified/marked) information is insertedlater in the derivation, either through the assimilation of a feature from a nearby segment or through the operation of a feature-filling rule (Kiparsky, 1985; Archangeli, 1988). For example, all sonorants are spontaneously voiced. Since all nasals are 1. The feature geometry used in the present study was based on Bernhardt's (1990; 1992a,b) modification of McCarthy's (1988) proposal. 2. Rice and Avery (in press) treat all features as monovalent. 15 sonorant, [+voice] is a redundant specification and does not need to be included in the underlying representation of a nasal consonant. Root I / \ A\ [+md] [+dist] [-ant] [-bk] [-hi] [+Io]: Figure 1-4 Feature geometry for english consonant system The theory of underspecification, especially radical underspecification (Archangeli, 1988)] has been used to analyze phonological data in many languages and is also relevant to child phonological acquisition (Stoel-Gammon & Stemberger, 1994). Table 1-1 (based on Bernhardt, 1992a) summarizes the assumed underspecified representations for discussion in this paper. 1. Archangeli & Pulleyblank (1994), in Iheir presentation of combinatorial specification (discussed in Section 1.3.1.4.5 below), refute earlier work on radical underspecification using examples from Barrow Inupiaq and Ainu that are incompatible with the elimination of redundancy from underlying representa-tions. 16 Table 1-1 Proposed consonant specifications for adult English2 Segment Root Laryngeal Place Iml [+consonantal], t+nasal] Labial lal [+cons], [+nasal] [+cons], [+nasal] Dorsal V [+cons] Labial • Ibl [+cons] [+voice] Labial Ixl [+cons] .Idl • [+cons] [+voice] ' Ikl t+cons] Dorsal ••Igl t+cons] [+voice] Dorsal .ni [+cons], [+continuant] ; Labial Ivl [+cons], [+cont] [+voice] Labial lei [+cons], [+cont] Coronal: redistributed] •ibl [+cons], [+cont] [+voice] Cor [+distributed] .1st [+cons], [+cont] • / z / [+cons], [+cont] [+voice] •IV [+cons], [+cont] Cor [-anterior] • w [+cons], [+cont] [+voice] Cor: [-anterior] . itpc [+cons], Branching [cont] Cor [-anterior] /*/ r+cons], Branching [cont] [+voice] Cor: [-anterior] /w/ [+sonorant] Labial Ihl [+cons], [+cont]d Laryngeal •1)1 [+son] 'ill [+cons], [+son]e ' Ill- [+cons], [+son] Lab + Cor, or [-ant] a. Only features presumed to be specified in underlying representation are indicated. b. Allophonic specification of [+spread glottis] is assumed for the word-initial voiceless stops. c. Bernhardt and Stemberger (in prep) suggested that [+grooved] is the most plausible representation for affricates; however, they also provided developmental evidence to support a branching [-cont], [+cont] representation. d. Bernhardt and Stemberger (in prep) treat [h] and [?] as [+sonorant] e. Bernhardt and Stemberger (in prep) treat [1] as [+continuant] and [+lateral] and [J] as [-consonantal] 17 1.3.1.4.3 Feature Representation: Vowels Theories of vowel representation in NLP are more disputed than consonant representation. Much of the work on Vowel-Place (V -Place) description has centered on explaining the spreading of vocalic features to other neighbouring and distant consonants and vowels. In this respect, the representation of vowels on the segmental tier has been in relation to, if not dependent on, the representation of consonants. Sagey's (1986) model presents V-Place as a dependent of Dorsal Place for the features [high], [low] and [back] and assigns the feature [round] to Labial Place as presented in Figure 1-4 and used as the basis of analysis and intervention in the present thesis. Clements (1985) argues that the artieulatory principles that led phonologists to positing Labial, Coronal and Dorsal Place for consonants should be applied to vowels such that front vowels are Coronal and back vowels are Dorsal. Place features for consonants and vowels occupy different tiers in the Clements model as shown in Figure 1-5 below. Place C-Place V-Place Figure 1-5 Consonant and vowel Place representation l. Based on Clements (1985) 18 These are just two of the many geometries put forward for vowel representation. It would appear that for any feature geometry to adequately represent consonants and vowels, C-Place and V -Place need to occupy separate planes in the representation (see Figure 1-4) so that the variety of spreading and failure to spread observed between vowels and consonants can be taken into account. e.g., "money" C C 1 h\oot | Root 1 [A] L [i] 1 I Root 1 Root 1 1 1 Figure 1-4 Multi-planar representation of consonants and vowels Table 1-2 (based on Bernhardt, 1992a) summarizes the assumed underspecified representations for discussion in this paper. 19 Table 1-2 Proposed vowel specifications for adult English Segment Root Laryngeal Place [+sonorant] III [+son] Dorsal: [-bk] ml [+son] Dorsal: [-bk], [+lo] lul [+son] Labial: [+rd] /{o/a}/b [+son] Dorsal: [+lo] Lax vowels: ll, z, o, ol [+son] Dorsal: As Ii, ei, oo, ul but with [-tensef Diphthongized vowels: leil* tool [+son]; [+son] [+son]; [+son] Dorsal: [-bk], [-hi]; [-bk] Dorsal: [+rd], [-hi]; [+rd] Diphthongs: laifml /oil Ida, ao/e [+son]; [+son] [+son]; [+son] [+son]; [+son] Dorsal: [+lo]; [-bk] Dorsal: [+rd], [-hi]; [-bk] Dorsal: [+lo]; [+rd] a. The least specified vowel is considered to be /a/ (/A/ is the stressed counterpart of lal). For vowels, it implies that [+sonorant] is the markedifoof feature, as are the Vowel-Place features [+round], [-back], and |+lo|. b. There appears to be some non-significant phonetic shifting in the low vowel series in the Vancouver dialect of English between lal and lol..It is acknowledged, however, that lal is further front phonet-ically than the Id and sometimes appears to be more closely related to the front than to the back series of vowels. c. The feature [tense] is designated as a Dorsal feature in this paper, since it refers to lingual tension. Diphthongs have [-tense] second components as a redundant feature in the dialect area. Tense vow-els are slightly diphthongized as well in the dialect area. Originally these were the long vowels of English. d. Diphthongs are treated as having two Roots. e. Canadian 'raising' results in I Ail for / ai and I Mil for laol in closed syllables with final voiceless con-sonants. 13.1.4.5 Feature Specification and Grounded Phonology One assumption of NLP is that once the input representation is established, the appropriate output representation is determined automatically, as the result of convention. Archangeli and Pulleyblank (1994) discussed problems related to the phonetic realization of phonological 20 organization and introduced Grounded Phonology, with its two grounding conditions, in an attempt to demonstrate the relationship of phonological representation to phonetic content. Grounding Condition I makes the absolute claim that well-formed feature paths (under Combinatorial Specification) in any language must be phonetically grounded (i.e. all terminal nodes must correspond to phonetic content). Under Grounding Condition I, ungrounded conditions are absolutely ruled out while permissible cooccurrence conditions are drawn from a phonetically motivated set. Grounding Condition II is more language specific and addresses markedness such that the feature paths allowed by this condition will tend to be invoked or not invoked as a function of their strength. Archangeli and Pulleyblank (1994) argue that the presence of a strong second condition is beneficial to the grammar of a language because the unmarked case is imposed maximally and enjoys a wide distribution. Archangeli and Pulleyblank (1994) argue against the use of segments as the primitive unit in phonological analysis citing evidence that feature representation involves overlapping artieulatory gestures (Clements, 1985; Sagey, 1986). They prefer a hierarchical structure that directly links the Root to the syllable constituents, bypassing a prosodic equivalent of the segment (i.e. CV-slots mentioned above). They suggest that the term "segment" be replaced with the term "path" and posit Combinatorial Specification to licence which features associate together on a path. Combinatorial Specification as proposed by Archangeli and Pulleyblank (1994) includes F-elements and their association status. F-elements are the binary features and class nodes (Place, Laryngeal, and Root) that do not carry any intrinsic feature content. Association status refers to whether the F-element is free (unassociated) or linked (associated), with combination of F-elements occurring at any given moment during the phonological derivation. Combinations are 21 restricted by cooccurrence restrictions which prevent or require F-elements to be on paths with each other. These paths "characterize the domain of paradigmatic Unkings of F-elements that occur either underlyingly or as a result of association or spreading" (Archangeli & Pulleyblank, 1994, p. 49). Features that are allowed to join paths are governed by a restricted domain such that only linked features are allowed on paths, whereas floating features are not. Further, there can be no more than one instance of each node or feature or prosodic category in a single path resulting in the restriction of implausible combinations such as [+nasal] being [+voice] and [-voice]. Archangeli and Pulleyblank (1994) propose "Grounded Phonology" wherein phonological features are restricted to F-element combinations that satisfy the principle of Recoverability. This states that phonological representations and phonetic content are related. It is not necessary to include all of the phonetic content in the representation because much of it is redundant. However, the authors demonstrate the tension between phonological and phonetic representation as a balance between providing phonological representations that are "as impoverished as possible" while enabling an unambiguous representation of phonetic content. As an example, for a language that does not distinguish between [a] and [ae], the representation [+low] would be sufficient. But, in a system that does make this contrast, it is necessary to provide a representation that will distinguish them phonologically (i.e. [a]: [+low] and [»]:..[+1QW], [-back]). How these F-elements are interpreted phonetically by the speaker (i.e. relative "lowness" and "backness") is assumed to be language-specific. Radical Underspecificatioh (Archangeli, 1988), used in the present analyses, is incompatible with Combinatorial Specification in theory because its basis is the identification of segments and the specification of such segments with the fewest possible feature values. In this way all redundant features are eliminated from the underlying representation. Combinatorial Specification deals with F-elements and ""prefers such elimination [in order to achieve 22 representational simplicity] but counteracts the tendency toward'such elimination by the principle of Recoverability [which holds that phonological representations and phonetic content are related]." (Archangeli & Pulleyblank, 1994, p. 110). Although Radical Underspecification theory and Combinatorial Specification theory disagree about the nature of the primitive, they share assumptions of cooccurrence constraints and associations (linking and delinking) with NLP theory. Since speech-language pathologists currently use segmental analyses in their assessment of phonological abilities in children, the relative benefits of each theory needs to be reviewed. 1.3.1.4.6 Phonological and Phonetic Interactions in Representation Grounded phonology (Archangeli & Pulleyblank, 1994) was developed as an attempt to link phonological organization with phonetic reality. Before further discussing how phonology and phonetics are related, it might be useful to describe how they are distinct. Phonetics is the study of the sounds of language in terms of their physical and artieulatory characteristics1, while phonology is the study of how sounds signify meaning in a particular language (Ingram, 1981). Relationally then, the phonology relates the phonetic events of speech to the grammatical units operating at the morphological, lexical, syntactic and semantic levels of language. A one-to-one relationship between phonology and phonetics cannot be assumed in that a phonemic distinction such as [+/-voice] in English has several phonetic implications. Beyond the presence or absence of voice, the voice onset time for initial plosives determines the presence or absence of aspiration and in post-vocalic plosives, the presence versus absence of preglottalization, rate of closure onset, and durational differences in preceding vowels (Vennemann & Ladefoged, 1973; Ladefoged, 1980; Harris & Cottam, 1985). Goldsmith (1990) comments that phonetic reality may motivate a phonological representation, but argues that it neither justifies nor ultimately explains it; "Phonetic reality provides the stuff of which phonological theory provides the organization" (p. 10). 1. Perceptual characteristics are also important. 23 Even so, it appears that the phonetic reality sometimes dictates the phonological organization. For example, onset-coda asymmetry is manifested phonetically when a voiced stop is attempted in coda position. It is articulatorily more difficult to produce an acceptable-sounding voiced stop in codas because voicing must be maintained for a longer period of time to be perceived in a coda than in an onset. An exception to this argument would be nasals which are redundantly specified as [+voice] underlyingly but are easily prolonged in coda position. In the case of a voiced coda, the phonetic complexity may be related directly to the phonological organization. The relationship between phonology and phonetics cannot and should not be separated. As Laver (1994) states, the descriptions of phonology and phonetics are independent yet intimately connected: "indeed it is not unreasonable to suggest that neither good phonology nor good phonetics is feasible without an adequate understanding of the other" (p.30). 1.3.2 Theoretical Issues in Phonological Development Issues in phonological development relevant to this thesis relate to early-developing wordshapes and segments, sequence and complexity constraints and phonetic limitations on phonological output. The phonologist or speech-language pathologist studying phonological development in children may hypothesize about how the phonological elements are represented underlyingly but descriptions of constraints, or rules, must be based on any given child's phonological output. Sequence constraints play a prominent role in early phonological development, and in the protracted phonological development of children with phonological disorders. These constraints may apply at the prosodic level (complexity constraints) or the segmental level (feature cooccurrence constraints) of representation, reinforcing the notion of tier autonomy. Repairs used to overcome sequence constraints provide a window into the child's 24 development of feature specification in underlying representation. Ultimately, output is governed by anatomical, phonetic abilities and that is where phonology and phonetics meet as discussed in Grounded Phonology (Archangeli & Pulleyblank, 1994). 1.3.2.1 Segmental Development Segmental development in early speech is continuous with babbling development (deBoysson-Bardies & Vihman, 1991). Children generally learn stops, nasals, glides and certain fricatives prior to learning liquids, affricates, or clusters (Smit, 1993). Although consonant acquisition cannot be predicted in an invariant order, Stoel-Gammon and Stemberger (1994) made predictions concerning the relevance of underspecification for aspects of phonological acquisition. Following an examination of their data, they concluded that less specified segments tended to be mastered first and that underspecified segments tend to serve as substitutes for more highly specified segments (e.g., stopping or velar fronting explained by not specifying [+continuant] or [Dorsal] respectively, resulting in production Of default stops or alveolars). Bernhardt and Stoel-Gammon (in press) described the consistent use of less specified features for more highly specified target features as child default production. They suggested that the following universal default features (i.e. the least marked features cross-linguistically) could be expected in early child phonology: 1. Root: [-continuant], [-nasal], with early specification starting with [+nasal] then including[+ continuant] before [+lateral] 2. Laryngeal: [-voice], [-spread glottis], with gradual specification of [+voice]/Onset prior to [+voice]/Coda 3. Place: Coronal [+anterior], with development of Labial prior to Dorsal 25 Given all of the above feature specifications, III should emerge as the default consonant. This concurs with Paradis & Prunet (1991) who argue for Coronal as the unmarked Place of articulation. Bernhardt and Stemberger (in prep) discuss the common phenomenon of "fronting" in child phonology in relation to the coronal default. When this happens, the specified feature such as Labial or Dorsal is delinked and the default Coronal Place inserted. Bernhardt and Stemberger (in prep) suggest that feature geometry expands in an incremental manner with multiple influences of complexity, underspecification and hierarchy. The main factor in segmental development appears to be the autonomy of the features. When these features become specified and what types of restrictions for their cooccurrence apply, is highly individual. 1.3.2.2 Prosodic Development Prosodic organization has beert described in terms of development as beginning with the universal canonical structure consonant plus vowel (CV), also referred to as the "core syllable" (Jakobson, 1968). Development of the prosodic tier is independent of segmental development. In early development, only one segment is produced in onset, nucleus and, later on, in the coda (Lleo & Prinz, 1994). There is a paucity of information in the literature concerning complex wordshapes. Information concerning cluster development is presented in the next section. 1.3.2.3 Sequence and Complexity Constraints Cooccurrence of features and prosodic structures are subject to sequence and complexity constraints. Sequence constraints play a prominent role in early phonological development, and in the protracted phonological development of children with phonological disorders. The distinction 26 between sequence constraints and complexity constraints is related to the level of representation at which each applies and the adjacency of the elements involved. Non-adjacent elements cannot affect each other but adjacency can be defined at several levels. Odden (1994) referred to tier, syllable and Root adjacency and Bernhardt and Stemberger (in prep) added timing unit adjacency. Timing unit adjacency accounts for CV interactions since C and V timing units can be adjacent. Conflicts between elements can occur at different levels such as the feature (e.g., place, manner and voicing) or the prosodic unit (e.g., heads, margins). Similar constraints and repairs occur for adjacent elements that are surface contiguous or non-surface contiguous although certain developmental effects of constraints such as consonant Place harmony have the most impact on noncontiguous sequences. Smit, Hand, Freilinger, Bernthal, & Bird (1990) defined order of acquisition for word-initial clusters as follows: 1. Stop-plus-/w/-clusters first emerged as obstruents (e.g., 'queen' /kwin/ -> [kin]). 2. All /l/-clusters emerged except for /si/. Obstruent + /l/-clusters emerged first as an obstruent (target or substitution), and except for IW, reduction to [w] was relatively uncommon. Clusters with dorsal stops generated more errors than clusters with labial stops. Labial variants (e.g., [fw]) were used extensively even for clusters without a labial consonant. 3. All /r/-clusters emerged except for76r/. Obstruent +/r/-clusters emerged as obstruents (target or substitution). In very young children reduction to a liquid or glide was occasionally observed. Again, [fw] was used even for those clusters that have neither a labial or a fricative component. 27 4. All /s/-clusters, including /svv/ emerged. Smit (1993) reported that /s/+consonant clusters emerge as [w], nasal or stop. Labial variants occurred only for those clusters with a labial target (i.e. /sw/, /sm/, /sp/). Asymmetry was noted for 1st, sk, sn/: the most common surface form was usually a stop or nasal. Reduction to the [s] occurred in the youngest age groups for these three clusters and they were rarely labialized. 5. The /si/ and /6r/-clusters were the last two-element clusters to emerge. 6; The three-element cluster/skw/emerged. 7. Other three-element clusters emerged last. Other studies demonstrated different patterns of development for the /s/-clusters. Smith (1973) observed /si/ as the earliest /s/-cluster. Ringo (1985) stated that /st/, /sn/ and /si/ were the earliest developing of the Isl clusters. It was predicted that the subjects in the present study would develop clusters in a more or less typical fashion. Where this did not happen, the interaction of constraints with the child's entire phonological system is discussed. 1.3.24 Phonetic Limitations on Phonological Output The phonological development of children with oral-motor limitations or unusual vocal cavities can be affected by their inability to articulate or coarticulate at the babbling stage of development and beyond. Stoel-Gammon & Cooper (1984) discuss how children's favourite babble sounds and shapes often form the basis for their early words and are instrumental in later linguistic preferences. The close relationship between the child's exploration of the articulators and the development of the neurological network that governs their behaviour and eventually allows for phonological representations were discussed by de Boysson-Bardies & Vihman (1991). Cleft 28 palate children are often cited as an example of those who lack the typical phonetic experience of babbling and early speech (Stengelhofen, 1989). Limited oral-motor and perceptual experience and the absence of normal articulators with which to produce the phonetic distinctions of the ambient language may result in the child's inability to develop meaningful contrasts. Further, children with cleft palate and other oral structure abnormalities may use compensatory strategies to produce speech using the structurally stable areas of their mouths, leading to articulatory or resonatory preferences such as dorsal and pharyngeal phones or persistent hypemasality. Stengelhof en (1989) suggests that children with high, narrowly arched hard palates in the absence of cleft palate (e.g., Eddie, in this study) are at risk for articulatory, phonatory and resonatory disturbances, similar to their cleft palate counterparts due to the reduced surface area of the palate and the tongue's restricted ability to either reach or make proper contact with the palate. Additionally, if the alveolar ridge is underdeveloped or elevation of the tongue to the ridge is prohibited or imprecise, "the articulation of a number of fricatives and plosives may be affected" (Stengelhofen, 1989, p. 15). Hewlett (1985) argues that structural abnormalities and the compensatory articulation required to adapt to them may affect the development of the child's speech-related neurological development and, ultimately, representation of phonological information. He states that the use of compensatory articulation may lead to patterns resistant to change, even following the alleviation of the Original physical limitations. Laver (1994) described the "tendency towards phonetic similarity in the segmental realizations that make up a given stretch of speech" (p.394) as a 'phonetic setting' (see also Weiner, 1981: 'systematic sound preference'; Edwards & Shriberg, 1983; Bernhardt & Stoel-Gammon, 1994: 'child default'). Although there is not a direct relationship between unusual oral cavities and difficulty with speech sound production, some children who have abnormal mechanisms also present with a phonetic or phonological disorder. It is difficult to determine whether or not the abnormal 29 mechanism is contributing to the delayed phonological development of the child. For example, children with severe athetoid cerebral palsy learn to read and write with limited control over their oral mechanisms, indicating that they have mastered the phonology of their language even in the presence of impaired biofeedback (Bishop, 1988). One must consider children who consistently produce default features or segments but may not have oral-motor difficulties (e.g., Colin: Bernhardt, MacNeill & Bohlen, 1994). Grunwell (1985) concludes that "even a child without a strictly visible organic disorder may need help in extending his phonetic inventory in order to realize his developing phonological system" (p.33). These contrasting examples make it clear that the relationship between phonetics and phonology needs further exploration. Researchers such as Kent and Hodge (1991) and Kent (1992) would argue that the physical or biological basis for speech is central. Autosegmental phonologists have acknowledged the physical and phonetic underpinnings of speech production by referring to the Place features as "Labial", "Coronal", etc., yet conclude that "the anatomic system proposes, while the phonology disposes" (Goldsmith, 1990, p.277). Issues relating to phonological development and disorders in the presence and absence of oral structural abnormalities are discussed further in Chapter 5. 1.4 Efficacy Studies Using Nonlinear Phonology 1.4.1 Previous Studies The Bernhardt (1990,1993a) and Von Bremen (1990) studies explored the developmental implications of NLP as they related to description of a child's disordered phonological system and intervention programming. Aspects of nonlinear phonological theory such as the autonomy of the prosodic and segmental tiers were examined with respect to the researcher's ability to conduct intervention separately at each tier and the effects were measured in terms of rate of acquisition. 30 Another autonomy issue related to whether the higher level features in the feature geometry hierarchy would be established more quickly than the lower level features. Results indicated that: 1. Independent development occurred at the prosodic and the segmental levels. 2. Development through intervention was more rapid for the prosodic tier. 3. Moraic and onset-rime approaches to prosodic tier development had equivalent results. 4. Higher level features had a faster rate of acquisition, although the difference was not significantly different except for the notably slow rate of Dorsal Place consonant acquisition. The Bernhardt (1993a) study used a combined series approach employing multiple baselines and an alternating treatment design (ATD) for segmental and prosodic conditions. The segmental conditions included feature establishment (SE) and feature linkage (SL). The Prosodic structure conditions included structure establishment (PE) and structure mapping (PM). 1. Feature establishment (SE) refers to training a new feature. If the child uses only [-continuant] segments, the feature [+continuant] could be taught. 2. Feature linkage (SL) refers to feature cooccurrence. For example, if the child says Ipl (Labial) and /hi ([+continuant]) but not /f/ (Labial, [+continuant]), then the combination of [+continuant] and [Labial] is the goal. 3. Structure establishment (PE) refers to adding new syllable shapes to the existing inventory (e.g., adding CVCV or CC VC to existing wordshapes CV or CVC. 31 4. Structure mapping (PM) refers to moving features or segments from established syllable positions to unestablished syllable positions (e.g., moving [+continuant] from word-final to word-initial position). Subjects were assigned semi-randomly to treatment conditions. Efforts to avoid multiple treatment interference effects were made by controlling sequential confounds, carryover effects, and alternation effects. Multiple baselines were taken in the form of three major probes and four minor probes. In the Bernhardt (1993a) study, two blocks of therapy over 12 weeks on the controlled treatment design were followed by a four week "ethical block" of therapy optimized for individual needs and learning style. The Bernhardt (1990) and Von Bremen (1990) studies consisted of a total of 18 weeks of intervention. All of the children achieved age-appropriate intelligible speech by the end of the Bernhardt (1990) study and 19/20 made notable gains in the Bernhardt (1993a) study. Fewer subjects achieved age-appropriate speech by the end of the 1994 study (40% instead of 100%) conducted by 20 speech-language pathologists under the supervision of Bernhardt. Severity levels of the subjects, familial factors, the variety of experimenters, and a shorter time frame (i.e. 16 vs. 18 weeks) contributed to reduced overall efficacy. Bernhardt (1990,1993a) reported that the rate of phonological change and the types of generalization which occurred were indications of the efficacy of the phonological intervention. As predicted, more rapid gains were made for syllable/word structure goals than for the introduction of new segments across subjects, regardless of treatment strategy used.1 There was no 1. Treatment strategies including "onset-rime", "mora-edges", "awareness" and "oral-motor" approaches are described in Appendix D. 32 difference between the rate of gain for establishment of new features versus combination (i.e. linking) of features already present. Individuals varied in which goal and segment demonstrated faster gains, implying that both types of segmental goals should be included early in intervention cycles. There was no difference between relative gain for new word shape establishment versus mapping of features to new positions. Individuals varied in their acquisition of new word shapes or movement of features within words, implying that both types of prosodic goals should be attempted early in intervention cycles. Preliminary results indicate high individual variability implying that both establishment and linking type goals should be attempted during the early stages of therapy to gain understanding of the individual's strategies for learning. Further, techniques such as oral motor versus awareness approaches and onset-rime versus mora/edges approaches should be tested for their facilitative value with any given child. All were equally facilitative across children even though the degree of responsiveness differed for each individual. Efficacy studies performed by Bernhardt (1990,1993a) and Von Bremen (1990) demonstrated that NLP theory provided a viable theoretical framework for a comprehensive analysis and intervention procedure. The theory's focus on hierarchical representation, underspecification, autonomy and association allowed for goals that were specific to the individual's segmental or prosodic phonological system. Since the efficacy of NLP has been demonstrated in controlled studies, the next question is its effectiveness in everyday practise. 1.5 Moving from Efficacy to Effectiveness The importance of conducting outcomes research is not in dispute. Hodson and Scudder (1990) pointed out the potential problems that occur in the absence of intervention efficacy including negative effects for the child educationally and emotionally and the cost of extended 33 intervention to parents (financial and emotional), employers and taxpayers. Goldberg (1993) summarizes success as the client getting better and functioning at a higher level than before receiving treatment. Hodson and Scudder (1990) comment that efficacy reports need to go beyond the "it works" level of reporting. They call for information detailing original severity levels, intelligibility gains, and hours, months or years of contact time for the treatment program and suggest that assessments or probes occur at the beginning of a treatment session so that there will be no opportunity for the child to have a "warm-up" period prior to being tested. Olswang (1993) emphasizes the difficulty of defining variables and their interactions when examining the treatment efficacy for developmental speech and language disorders. The broad-based nature of the papers published by Shriberg and Kwiatkowski (1994), Shriberg, Kwiatkowski and Gruber (1994), and Shriberg, Gruber and Kwiatkowski (1994) provide an excellent basis for identifying characteristics of children with phonological disorders and information on both short-term and long-term effects of intervention. As mentioned earlier, Fletcher et al. (1988) distinguish between efficacy and effectiveness. Efficacy is a measure resulting from a highly controlled experiment conducted by highly trained individuals. Effectiveness measures are the result of interventions being carried out by the people who usually do the intervention in the places that they usually do it. Blockberger (1993) points out that in the field of augmentative and alternative communication, many efficacy studies and few effectiveness studies are conducted. She questions the value of information derived from highly controlled studies given the reality of non-compliance, equipment breakdown and staffing changes in everyday user situations and emphasizes the need for measures of effectiveness of speech-language intervention under typical conditions. How can theory-driven therapy be achievable within the confines and realities of daily practise and how can issues of non-compliance and family dynamics be factored into the design of 34 the experiment? One factor known to all clinicians and emphasized by Olswang and Bain (1994) is that progress in therapy is individual: 'The manner in which development progresses naturally, and with treatment, clearly will vary according to disorder type and etiology. For example, children with speech/language deficits originating from sensory-motor problems versus perceptual problems will respond differently to particular treatment techniques. Even populations of children who appear similar...are showing different responsiveness to treatment perhaps because of different levels of readiness for treatment (maturation), or different underlying causes for the disor-der" (p.43). In large studies such as Bernhardt (1990, 1993a), individual subjects' performance is shown to be the most interesting area of study, a fact that is discussed also in the aphasia literature (see Chapey, 1994). Similar cases are compared in discussions of the effectiveness of treatments. The portion of the Bernhardt (1993a) study described above had the goal of testing NLP theoretical principles applied to therapy. These were tested in Blocks I and II. Moreover, the design allowed for an "ethical block" in which therapy was tailored to the child's remaining needs as well. This customizing of intervention in the clinic is usually considered desirable from the outset of the intervention process because "the needs of an individual client [may] not match the requirements for reliable and replicable data collection" (Siegel, 1993, p: 37). Research can also be conducted with adaptation to the client's needs in mind, especially if the goal is to do an effectiveness study. Single-subject research design, reflective of daily intervention programs in the typical health clinic, needs to be dynamic in nature given the subject's variability in performance (Holcombe, Wolery and Gast, 1994). Following the assessment, goals are chosen and ordered to match the needs of the child and their appropriateness is monitored as therapy progresses. Olswang and Bain (1994) emphasize the value of using ongoing clinical data to make decisions about the course of intervention and its effectiveness. In the case of phonological intervention, the child might experience "breakthrough" moments in the learning of new syllable structures or segmental features that affect the direction in which they are ready to progress. While Siegel(1993) feels "that 35 most useful clinical insights come from work that [is] not designed to be treatment research (i.e. 'basic research') and that the biggest impact of research is exploring the underlying nature of a communication disorder" (p.36), the type of data that one collects during intervention research is "real-time" information about the development of the child's phonology and how it is changing. Intervention research provides data with which past and current hypotheses are supported or refuted. Whether or not treatment efficacy research addresses the "underlying nature " of a problem or not is a moot question. It addresses the question of how phonology "works" in a variety of individuals and sometimes leads to an answer of why. One cannot simply modify a design at will. Researchers wanting to maintain the integrity of studies with the goal of replicating findings over multiple case studies must be careful with their research design. Further, in order to make a generalization about the effectiveness of a particular intervention, in this case therapy based on NLP theory, interventions must be implemented with precision (Wolery and Ezell, 1993) and as planned (LeLaurin and Wolery, 1992), something that Johnston (1988) maintains rarely occurs. 1.5.1 Variations In Therapy Research Design The present study is an effectiveness study designed to optimize and evaluate the application of NLP theory in typical clinical practise. Manipulable optimization factors in the research design include time allotments for goals, goal order and selection, family priorities and strategies which accommodate the individual characteristics of the children. Modifications of time factors include lengthening or shortening the treatment block and protracting or retracting the total time span over which intervention occurred. Flexibility with respect to goals allows for: 1. The focus of intervention to be more on either segments or word shapes 2. Adjustment of the time framework to spend more or less time on goals 36 3. Some goals to be targeted simultaneously 4. The use of new segments with new structures. Factors that were not manipulable were: 1. The phonological principles on which intervention is based (e.g., using the child's existing word shapes, such as CV, to add new segments or adding new word shapes using existing segments) 2. Goal selection based on the theory 3. The overall time framework for therapy and assessments at the onset of the project. 1.6 Summary Bernhardt's (1990,1993a) investigations have attempted to use NLP theory as the basis for accounts of outcomes in intervention studies with children demonstrating severe phonological disorders. The present study is similar, building on Bernhardt's foundational work and applying it independently in everyday clinical practice. The analyses of the data and the choice of intervention goals were guided by and measured against claims made by current phonological theory. The nature of this study is also explorative, the original question being: will phonological intervention based on principles of NLP work? Given that the preliminary results of the Bernhardt (1993a, 1994c) research were positive and that an optimization of this approach has not previously been attempted, the following research questions related to optimization and theory were posed: 37 1. Optimization Questions: a. Will therapy derived from nonlinear phonological theory result in positive developmental changes for these subjects when carried out independently by a speech-language pathologist in a typical clinical setting? b. Given the typical health unit setting in which therapy is available once or twice weekly, will twice weekly intervention be as effective for these subjects as three times per week therapy for selectively matched subjects in the Bernhardt (1990; 1993) studies? 2. Theoretical Questions: a. How do the subjects' initial and intermediate phonological systems inform us about issues related to phonological development within nonlinear phonological theory? b. What is the nature of the interaction between phonology and phonetics in the phonologies of the two subjects? Chapter Two outlines the subjects and methods used to collect and analyze data. Chapters Three and Four describe the intervention plans and provide the results of major and minor assessments for Shaun and Eddie, respectively. A discussion of how the results relate to NLP theory and the optimization process is found in Chapter Five. 38 2 Methods 2.1 Overview The present study examined the effectiveness of phonological intervention in typical clinical practise. Two subjects, Shaun and Eddie, participated in a 16-week intervention program to improve their phonological skills. The investigation was based on the original question: "Will phonological intervention based on principles of NLP work in a typical health care setting?" The particular phonologies of the two subjects led to exploration of NLP theory including consonant-vowel interactions, sequence constraints and the relationship between phonological and phonetic development. 2.2 Subject Description 2.2.1 Subject 1: Shaun (Age 3;5 to 4;1) Shaun was three years, five months old and living with his parents and older brother at the time of the initial assessment CT {f. His hearing and language comprehension were average and his oral structures and voice were unremarkable. Table 2-1 provides a summary of background information related to Shaun. Low intelligibility of his speech was the main concern at the time of Shaun's initial screening although a language delay was also suspected. Shaun's mother understood him approximately 80% of the time while other family members understood 50% or less. 39 Table 2-1 Background information: Shaun Medical • Unremarkable pregnancy and birth history • Developmental milestones achieved within normal limits Social • Lives with mother, father and older brother by two years • Parental education: father (grade 10), mother (grade 12) • No daycare experience prior to final month of intervention Hearing • Two unilateral ear infections reported (7m, 22m) • Assessment June 22, 1992 indicated normal hearing and impedance measures bilaterally Language and Vocabulary • Average score on EOWPVT-R. a test of productive vocabulary • Simple sentence structures used. Limited overall language output. • Average to above average sentence comprehension on the TACL-R • Good phonemic awareness skills on minimal pairs task Oral Examination • Oral structures unremarkable • Rapid syllable production adequate Voice and Fluency • Voice unremarkable • Cyclical periods of dysfluent speech During the initial assessment, Shaun produced word-initial sound prolongations, infrequent syllable repetitions and hard glottal attacks on sentence-initial "I". Many of the dysfluencies occurred when Shaun and an adult were speaking simultaneously. Shaun's dysfluency did not appear to be a major factor in inhibiting speech sound production. Delayed morphosyntactic skills were related to but not fully explained by his phonological disorder. Language comprehension and phonemic awareness skills were relative strengths for Shaun. 2.2.2 Subject 2: Eddie (Age 3;7 to 4;2) Eddie was three years, seven months and living with his parents and younger sister at the time of the initial assessment (Ti). His hearing and language comprehension were average and he produced many complex sentences in appropriate narrative structures. Fluency and voice were unremarkable at T^ Table 2-2 provides a summary of background information related to Eddie. 40 Table 2-2 Background information: Eddie Medical • Caesarean section delivery; Early difficulty breast-feeding resolved within two weeks • Diagnosed with insulin dependent diabetes at age 1; l l a • Motor development milestones within normal age limits but had not yet completed toilet training by the end of the intervention project (age 4-2). • Oral pacifier used at bedtime Social • Lives with mother, father and one younger sister • History of language processing difficulties in family (uncle-maternal side/paternal history unknown) • Both parents university educated • Attended playschool twice weekly during the intervention project Hearing • Normal hearing and impedance bilaterally on an audiological assess-ment Language and Vocabulary • Produced many complex sentences • Appropriate narrative (i.e. storytelling and conversation) skills • Age-appropriate development of vocabulary, basic concepts and understanding sentences of increasing length and complexity • Poor phonemic awareness skills on minimal pairs taskb Oral Examination Structure • Relatively high-arched palate • Anterior open bite • Nasal air emission during production of nasal clusters (e.g. /sn/, /sm/) • Suspected hairline sub-mucous cleft seen by transillumination at speech and hearing clinic: not verified at cleft palate clinic Function • Reduced tongue tip control, especially on elevation and protrusion to the left of mouth • Jaw assisted articulation of phonemes, with lateral release of frica-tives • Rapid syllable production within normal limits (Robbins & Klee, 1987) for repeated /pA/ and lk\l. Reduced "buttercup" to lb^k\l and "pattycake" to /paeikeik/ Voice and Fluency • Fluency and voice initially unremarkable • Increasingly hoarseness and breathy quality over intervention period0 a. Family systems effects of childhood diabetes are well-documented in Baum, J.D., & Kinmonth, A.-L. (1985) and Raymond, M. (1992). b. Eddie was unable to perform beyond a chance level on the minimal pair task presented. c. Muscle misuse voice behaviours (e.g. shouting, using loud intensity over a long period of time and overtaxing the vocal system when fatigued) were reported and observed outside the clinic context. 41 Low intelligibility of his speech was the main concern at the time of screening although a language delay was also suspected. Eddie's mother reported that he demonstrated frustration and anger when not understood. She understood him approximately 80% of the time while other family members understood 50% or less. Eddie's oral structures appeared to be a major factor in his inability to produce speech sounds accurately although they did not affect his language production. Many of his [+consonantal] productions had a Palatal or Dorsal Place, a pattern similar to that found in children with cleft palate (Stengelhofen, 1989). Poor tongue control and a substantial open bite resulted in the production of sibilants with a lateral lisp. Further, it was apparent that medical factors and implications such as low energy related to blood sugar levels needed to be accounted for and integrated into the overall therapy plan. In summary, Shaun and Eddie presented with moderately-severe phonological disorders that were different in nature. Shaun's need for intervention was primarily linguistic with phonological and language delays and a secondary factor of fluency. Eddie required therapy to address both phonological organization and phonetic implementation related to atypical oral structure and function. 2.3 Research Methods 2.3.1 Experimental Design This study employed a multiple baseline design over a 16-week period of twice-weekly intervention: Goals for therapy blocks and treatment techniques were assigned prior to the initiation of intervention and changed in accordance with the optimal design. Observation goals were assigned to each subject with respect to individual phonological systems and goals, with the 42 prediction that untargetedgoals would show progress unrelated to therapy (i.e. maturation). Goals and intervention plans are detailed in Chapters Three and Four. Two evening caregiver meetings were arranged to share information regarding the intervention process, therapy goals and techniques. This was also an opportunity for caregivers to share their impressions of therapy and their child's progress and comfort level with the process. Subject performance during each session was monitored using audio and/or video taped recordings. Sessional data were used to address questions related to the subject's developing phonological knowledge as measured breakthrough moments1 in learning that may or may not have been directly related to the session's goal. Data related to the proportion of match/nonmatch productions and proportion of spontaneity and imitation were also collected. Intermediate stages of generalization were assessed using trained and untrained target words referred to as minor probes, which occurred at six points (ti-%) during the period of intervention. These data were used to inform decisions concerning the optimization of therapy, such as adding goals to the intervention program or reassigning intervention goals to observation status. Generalization to conversational use was assessed at week 1 (Tj), week 8 (T2), and week 16 (T3). Words used to collect phonological data during the major probes were not used in any other probes or during intervention. Overall gains in terms of Percent Consonant Correct (PCC) and Wordshape Match (WSM) were compared to six selectively matched subjects from Bernhardt (1993). Matches were chosen on the basis of initial PCC and WSM scores which had to be between 20 and 50 percent. These cutoff scores were based on the severity ratings of Edwards and Shriberg (1983). 1. A breakthrough moment is defined as a moment at which previously unobserved progress occurs on present, previous or untargeted goals. 2. Lower case is used to distinguish the minor probes tj-i6 from the major probes T1-T3. 43 2.3.2 Data Collection The two subjects described above had received speech-language screenings three months prior to the study and were awaiting full assessment for suspected severe phonological impairment and related language delays. Neither had received phonological therapy before the onset of the present intervention project. A sample of at least 150 words (ranges164-824, x=395) balanced for phonemes across word positions (Bernhardt, 1990) wascollected at each major probe time. A free play situation was used in an attempt to obtain spontaneous productions of the probe words. Elieitation and imitation were used, where necessary, and are indicated on the transcripts in Appendix B and C. Sessions were recorded on Fuji FR-II Pro60 tapes using a Marantz PMD221 with a Sound Grabber microphone (Model 12SG) and narrowly transcribed according to the IPA (1989) using a Yamaha KX-530 tape recorder and K240 Monitor (600 ohms) headphones. 2.3.3 Transcription Reliability Transcription was completed by the author and another reliable transcriber on ten percent of the initial and final assessment tapes. To calculate intertranscriber reliability, each phonetic slot counted as one item. Where there was disagreement between the two transcribers, a joint transcription was made until agreement was reached. Items were discarded in the cases where no agreement was reached. For Shaun, the two transcribers were in agreement for 79% of the items prior to discussion, and 94% agreement following the joint transcription and discussion. Disagreements primarily concerned vowels (e.g. Ill was transcribed as Id by one and /ae/ [raised] by the other) and the presence of a final consonant in two instances. For Eddie, the two transcribers were in agreement for 82% of the consonants prior to discussion, and 96% agreement following the joint transcription 44 and discussion. Disagreements primarily concerned the degree of palatalization on Inl and the presence or absence of voicing on final consonants. None of these disagreements affected the intervention plan or outcomes. 2.3.4 Analysis Phonological systems were analyzed in terms of prosodic and segmental inventories using Speech Application (Bernhardt & Cam, 1994), a computerized analysis that provided the following details: 1. Total child segmental inventory, substitution patterns, and proportion of match with adult targets in the following word positions: a. Word-initial position (WI) b. Syllable-initial position within word (SI) c. Syllable-final position within word (SF) d. Word-final position (WF)1. 2. Total child prosodic inventory, substitution patterns and percent match to adult prosodic targets (e.g. CV, CVC). The numbers obtained from the Speech Application program were categorized into different degrees of establishment based on Bernhardt (1990, 1993a) with a corresponding bracketing system including: (1) Marginal (((1-10%))); (2) Emerging ((11-40%)); (3) Developing (41-74%); (4) Established 75-100%, and (5) One token { }. In calculating the degree of establishment, zero values were not included in the total number. For example, if [+nasal] was counted in all positions at T l for Shaun as 59/80 (74%), then 1. Grunwell (1985) abbreviated these four word positions as SIWI, SIWW, SFWW, and SFWF respectively. 45 his degree of establishment would be "developing". However, in four productions, the final C was absent (zero value). Since it is impossible to know whether these four omissions were because of segmental rather than prosodic restrictions, they are removed from the count, resulting in a total of 59/76 (78%) and indicating that the feature [+nasal] was "established". Further, in counting features such as [-(-continuant], both substitutions and matches were counted if the feature was present. In Shaun's data, matched segments with the feature [+continuant] were present in 7/134 (5%) of his attempts suggesting that [+continuant] was a marginally established feature. If substitutions and matches were counted, his score increased to 55/134 (41%), demonstrating that [+continuant] was a developing feature, which was a more accurate analysis. Data were interpreted within a NLP theoretical framework that emphasized hierarchical representations, tier autonomy, underspecified representations and tier associations. Intervention goals were selected based on the NLP analysis of the subjects' segmental and prosodic phonological skills. Intervention data were used to examine relative rates of change for the various feature and stuctural goals and to provide rationale for alterations to the initial course of therapy. 2.3.5 Intervention Intervention was conducted over a 16-week period for a total of 28 sessions. Therapy was conducted in 60 (Shaun) and 45 (Eddie) minute sessions, twice weekly. The multiple baseline design was based on Bernhardt's (1990,1993a) studies and involved similar goals (i.e. segmental establishment, segmental linking, prosodic establishment and prosodic mapping) and strategies for attaining these goals (see Appendix D). Although the intervention sequence is described more fully in Chapters 3 and 4, summaries of the plans are included here to demonstrate the optimization process. 46 2.3.5.1 Shaun's Intervention Plan Table 2-3 Intervention plan: Shaun Session Numbers Goal (Observation Goal) 1-4 Establish the contrast/link between default segments /h,w,j/ and other consonants word-initially using onset-rime theory and an awareness approach. (Observe 16/ and/&/ in word-initial position.) 5-8 Establish Liquid III word-initially using onset-rime theory and an oral motor approach. (Observe/j/ in word-initial and syllable-initial posi-tions.) 9-12 Establish Dorsal stops word-finally. Map production to word-initial position using onset-rime theory and a combined awareness/oral-motor approach. (Observe word-final and syl-lable-final /rj /.) 13-16 Establish CCVC word shape using glides and liquids for C2. 17-20 Map word-final HI and 'NI to word-initial position using onset-rime the-ory and awareness/oral motor combined approach. (Observe 16/ and Ibl in word-initial position.) 21-24 Establish sCVC and CVCs using mora "edge" theory and an oral motor approach. (ObserveIJIclusters.) 25-30 Review all cluster goals. Changes to Shaun's original intervention plan included reordering goals, combining therapy approaches (e.g. combining a segmental establishment goal with a prosodic mapping goal in sessions nine to twelve) and introducing a highly specified segment in clusters (i.e. /J/). Shaun's intervention goals, their rationale and strategies used to achieve these goals are described more fully in Chapter Three. 47 2.3.5.2 Eddie's Intervention Plan Table 2-4 Intervention plan: Eddie Session Numbers Goal (Observation Goal) 1-4 Establish Coronal Place stops using onset-rime theory and an aware-ness approach. (Observe In/, Is/.) 5-8 Establish [+lateral] III word-initially using onset-rime theory and an oral motor approach. (Observe IJI in and word-initial and word-medial positions.) 9-12 Establish SCV word-initially using mora "edge" theory and an oral motor approach. (Observe clusters of the form stop + glide.) 13-18 Link Coronal Place and [+continuant] in the context of the high vowels l\l and /u/ using onset-rime theory and a combined awareness and oral motor approach. (Observe /J7.) 19-26 Establish coronal [-(-distributed] using onset-rime theory and a com-bined awareness/oral motor approach. (Observe complex segments with branching lt[t oV) 27-30 Recycle each goal prior to terminating therapy Changes to Eddie's original intervention plan included extending the number of sessions for an intervention target (e.g. Coronal [-(-distributed] increased from four to eight sessions), focusing all goals on Coronal Place and reducing session length from 60 minutes to 45 minutes. Each session also began with a warm-up period of oral-motor exercises. Eddie's intervention goals, their rationale and strategies used to achieve these goals are described in greater detail in Chapter Four. 48 3 Results: Shaun 3.1 Initial Phonological System (Tt) 3.1.1 Segmental Analysis During the initial assessment, Shaun used short, simplified productions of words and sentences and was reluctant to imitate the examiner. The following specified features were established at Tj: 1. Root: [+nasal], [+consonantal] and [+sonorant]1 2. Laryngeal: [+voice] 3. Place: Labial Place2, [+round] as a terminal feature. Partially established features were: 1. Root: [+continuant] was developing (53/134=40%) 2. Laryngeal: [+spread glottis] developing (24/78=31%); [h] was also used as a default 3. Place: Coronal Place was used as default Place and [-anterior] was emergent (4/28=14%). 1. It was unclear as to whether or not Shaun made a clear distinction between [+consonantal] and [+sonorant] atT.!. 2. Coronal is unspecified at Place and appeared to act as a default in Shaun's phonology. That is, Coronal was only partially established as a Place and Idl acted as a default in Shaun's system. 49 Shaun's surface output feature geometry at T t is summarized in Figure 3-1. Refer to Appendix Table A-If or a summary of his phonetic inventory at the time of the initial assessment ( T O . Figure 3-1 Feature .geometry'1: Shaun T j Segments produced in onsets had more matches (WI =48%, SI=32%) than segments produced in coda position (SF=19%, WF=27%) demonstrating onset-coda asymmetry. A further preference for [+voice] in onsets was noted in Shaun's use of E+voice] defaults in this context. 1. In the feature geometry figure, dashed lines represent partially-specified features and dashed circles repre-sent default features. Feature geometries in this thesis represent surface output specified features of the child. 2. WI=word-initial, SI=syllable-initial, SF=syllable-final, WF=word-final 50 Given that three of Shaun's four default segments /w,j,h,d/ are phonotactically constrained to onset position (i.e. /w,j,h/) and three are also voiced (i.e. /w,j,d/), these observations should not be surprising. The following features were not established at Tj: 1. Root: a. Unspecified for [+lateral] III -> [j] WI and [w] SI b. Branching of [continuant] for /tf, dj/ unspecified 2. Laryngeal: [+spread glottis] developing specification (24/78 = 31•%), although [h] used as a default 3. Place: a. Coronal: unspecified for [+distributed] fricatives IQ, & /; and Labial-Coronal III -> [w]1 b. Dorsal Place: nasal /rj/ and stops /k,g/ The degree of feature establishment is summarized in Table 3-1. Table 3-1 Summary of segmental analysis: Shaun Tj Tier Established Partially Established Unestablished Root [+nasal], [+consonantal] [+sonorant] [+continuant] Liquid Complex branching of/tLdy Laryngeal [+voice] [+spread glottis] Place Labial Place [+round] Coronal Place [-anterior] Dorsal Place [+distributed] I. hi —> [w] (Place delinked) was treated separately from III --> [w] and [j] (sonorant defaults). 51 3.1.2 Default Use The following table represent Shaun's use of default features for underspecified segments in word-initial (WI) and syllable-initial (SI) position at Tj. The relative strength of that default is represented by the length of the line connecting it to each segment and the brackets enclosing the phones1. Examples of words that resulted from the use of defaults are provided below each diagram. Table 3-2 Default diagrams Word-initial position Syllable-initial position ((/v/)) \ (((#/))) e.g. 'leaf -> fjis], 'five' -> [jar], 'sides' -> fjai] e.g. 'screwdriver' -> [wuwaijo1] Iw] (HI)/ ^ «*">> pi/) ^ \ s [ w ] / e.g. 'read' -> [wid], Tour' -> [woo] e.g. 'starry' -> [dowi], 'toothpaste' -> [tuwei?], 'Molly's' -> [mAowi] 1. ((())): marginal 1-10%; (()): emergent 11-40%; (): developing 41-74%; {..}: one instance only 52 Table 3-2 Default diagrams Word-initial position Syllable-initial position { / k h / } / ^ / e / } / V i } e.g. Van' -> [haen], 'zipper' -> [hsbo:], 'six' -> [hi?], 'shower' -> [haowo], v m ^ / y V/s/) e.g. 'dresses' -> [wedia], 'waving' -> [weihi], 'feather' -> [wsho:], 'mouthy' -> [maohi], Coronal defaults (e.g. 'coffee' -> [dati], 'chickenpox' -> [dAZsnbap], 'laughing' -> [J33s)in]) and [+voice] defaults in word-initial position (e.g. 'page' -> |bei], time' -> [dAim], 'candle' -> [daendoo], 'thumb' -> [vAm], etc.) were also present as shown in Figure 3-2 below. Target word: 'coffee' /kafi/ -> [dati] /(t) C Root o Shaun's UR a V Shaun' s Production f C i/ V Id C [+son] Root A ^',[+cont] * ° £ ft V-Place O C-Place O Dors! |(Default Coronal) +cons] Root C-Place >rsal O [+bk] [+lo] L a b i a l V-Place [+voice] O Dorsal / f+hi] [+son] a v Root A il V t c Root Root J+cont] C-Place Q V-Place O D o ^ a vf+cons] C-Pface 0 o Dorsal a, [ + ^ L a b i a l y (Default Coronal) Figure 3-2 Examples of Coronal and [+voice] default representations 53 3.1.3 Prosodic Analysis Shaun's syllable structure at was limited by the small number of established segments available for use in coda position. Analysis at Tj revealed the following details: 1. Open vs. closed syllables a. Monosyllabic words were produced as open syllables (e.g. CV: 53/60=88%) more frequently than as closed syllables (e.g. CVC: 50/127=39%). b. Bisyllabic words were less divergent in the proportion of open CVCV (8/18=44%) and closed CVCVC (7/29 = 24%) wordshapes. 2. Cluster production a. One token was produced in WF position (e.g. 'jump') and two tokens were produced in WI position consonant plus glide sequences (e.g. [mjao]). b. Some clusters were observed over syllable boundaries (e.g. [pcnhoo] for 'pencils', [wailuendoo] for 'triangle', [wEndoo] for 'finger' and [handaja] for 'Santa Claus'). 3. Total Wordshape Match (WSM): 135/368 (37%). 4. Maximum word length demonstrated: Four syllables (i.e. 'sewing machine')1 He generally matched the number of syllables in the target word. 5. Typical word length: One syllable (284/369=77%). 6. Maximum word shape: CVV.CVC+V.CVC (i.e. 'sewing machine')2 7. Developing wordshapes: CVC, CVCV, CVCVC, CGV, etc. 1. The maximum non-compound word length in syllables was three (i.e.' triangle'). 2. The maximum non-compound word shape was CVV.CVC.CVV (i.e. 'triangle', [wAiruendoo] ) . 54 3.2 Block I Goals, Rationale And Progress Intervention goals were chosen to optimize Shaun's phonological development rather than to follow an alternating treatment design strictly. Each goal, the rationale for its choice and progress achieved during the first intervention period, Block I, is listed below. Measures of progress included the major probe at T 2 , minor probes at t1? t2 and t3 and intervention session data. 3.2.1 Goal #1: Root ([-fcontinuant], [+consonantal]) Establish the contrast/link between default segments lh,w,jl and other consonants word-initially using onset-rime theory and an awareness approach. In onset, if Place was redundant (Coronal) Shaun used a [+consonantal] substitution for a [+continuant] segment but did not always do so (e.g. 'sun' / s A n / - > [ d A n ] and [hAn]). If the target was [+consonantal] with specified Placefeature Labial, he did not combine [+consonantal], [+labial] and [+continuant] to produce an [fj. Given that [+continuant] was not a major problem in and for itself (40% match), the linking of [+continuant] and [+consonantal] was considered a n important contrast with /h j,w/ since the [+consonantal] feature was less well-established than the [+sonorant] feature at Root. Onset-rime approaches were chosen for two reasons. Firstly, the default segments [h,w,j] can only appear in onset position and had to be contrasted with other segments in that position. Secondly, although codas were generally weak for Shaun, he did not demonstrate the cooccurrence constraint to the same degree in coda position. That is, [+continuant] segments specified for Place were allowed in coda more often than in onset position. Shaun was very reluctant to speak in the early stages of assessment and therapy. Given this disposition and his cyclical stuttering, a low demand introduction to phonological intervention was 55 chosen as optimal for him. Using the awareness approach allowed Shaun to participate in therapy without being required to produce any speech. Observation targets included IQI and Ibl in word-initial position. The major probe taken at T 2 demonstrated that [+continuant] onsets (19/36= 53%) and codas (15/27= 56%)were developing, with a proportional gain of 34% and 14% respectively. There was no change in the observation targets IQI and Ibl following intervention on this goal. A minor probe for goal #1 taken at t{ indicated that [+continuant] cooccurred with [-(-consonantal] preceding high vowels for Coronal Place only. This high vowel context did not facilitate the production of ([+continuant] Labial) combinations (e.g. /fufu/ -> [huhu]). Shaun's progress on goal #1 during Block I is summarized in Figure 3-3 below. Relatively higher scores for untrained words are possibly the result of a delayed echoic response mode. 60 Major and Minor Probes • T I B tl • T2 Figure 3-3 Block I progress: Goal #1 56 3.2.2 Goal #2: Root [+lateral] Establish [^-lateral] III word-initially using onset-rime theory and an oral motor approach. Establishing IM would provide a building block for clusters not present in Shaun's phonology, plus combine [+consonantal] and [+sonorant] productively with the feature [+lateral]. During a stimulability probe for Dorsal Place stops (i.e. /k,g/), Shaun inadvertently produced an IM when attempting Igl. This led to consideration of III as the next target. Onset-rime approaches were used because Shaun's discovery of the IM phoneme occurred in onset position (i.e. during the initial oral mechanism assessment I attempted to stimulate /g/ using 'gun, guh, gun' as a model to which he replied 'nduh, uluh, uluh'). Prompting for the imitation of IM in the context of touching the tip of his tongue to his lip and dropping it to produce Hal was also successful so that a oral-motor approach was viable. The opportunity to further establish this phoneme was optimized by reordering the segmental goals such that Dorsal Place stops followed the segment IM. SI position was not targeted because Shaun was able to imitate it with 81% accuracy during a minor probe. Instead, IM in this word position was designated as a post hoc observation goal. Since IM in word-initial and initial-word position (onsets) is qualitatively different than/1/ in coda position, coda IM was not chosen as a target or an observation goal. The phoneme Ii /was an observation target in both word-positions. Shaun quickly learned and generalized the production of WI IM (3/3=100%) to spontaneous conversation by 7% He overgeneralized its use in the case of WI /w/ and If I for two words at T 2 (e.g. 'waving' -> [leihin], 'feather' -> [khoo]). By T 2 , Shaun's production of SI IM was developing (3/5=60%) and this gain corresponded with an improvement in C 2 in C^VC^V production (11/13=85%). No change in the official observation goal lal in onset position (Tj: 0/18; T 2 : 0/8) was observed. Minor probe data taken at t2 indicated that Shaun was able to produce trained words 57 with (14/14=100%) accuracy and untrained words with (5/6=83%) accuracy. Progress on goal #2 is summarized in Figure 3-4. „ 100 Major and Minor Probes 0T1 Ot2 "T2 Figure 3-4 Block I progress: Goal #2 3.2.3 Goal #3: Place (Dorsal) Establish Dorsal stops word-finally and map production to word-initial position using onset-rime theory and a combined awareness/oral-motor approach. Dorsal Place was not established nor stimulable at the time of the initial assessment. Following the first block of therapy, no dorsals were observed. However, during a minor probe for III, Shaun was able to imitate word-final /ltd in the words 'lock' and 'dock'. Further probing indicated that he could only produce word-final lid (but not /g/) in the context of the low back vowel lal. Shaun's difficulty producing the Igl may have been related to coda preference for 58 voiceless sounds (Laver, 1994) or feature cooccurrence constraints that disallowed the cospecification of Dorsal and [+voice]. Production of Dorsal segments in children who stutter may induce blocking given the pharyngeal constriction involved in their production. Thus the first activity was an awareness activity focusing on the7k/ phoneme in the context of the vowel Id. Eventually, Shaun was encouraged to participate in the activity by repeating 'awk', 'awkawkawk-ka' after the clinician. Production was considered a necessary part of this activity given that Bernhardt's (1990-1994) preliminary results indicated that Dorsal Place therapy was less successful with an awareness only approach. The nasal /rj/ (Dorsal, [+nasal]) was produced as a Coronal or Palatal nasal and was chosen as an observation goal. Dorsal stops were established word-finally in the context of the low, back vowel /a/ in the words 'awk' and 'ahg' prior to being mapped by the onset-rime method to word-initial position. Major probe data are less informative than minor probe and intervention data for this goal. At T 1 ? Dorsals were not produced in any context. AtT 2 , following intervention, Dorsals segments Ikl and /g/ were produced with in onsets (11/18= 61%) and codas (5/29= 17%). The observation goal /IJ/ was emergent with a small amount of improvement between T^O/20) and T2(4/29=14%). Minor probe data taken at t3 showed that Dorsal segments in WI position were developing (trained: 5/8=63%, untrained: 2/7=30%) while Dorsal segments in codas were absent for both trained and untrained targets. Progress on goal #3 during Block I is summarized in Figure 3-5. 59 70 -+•> <D Ut i _ 60 -ra H 4-1 50 -3 40 -< JO •w $ 30 -JC u 20 -eg Z 10 -$ J J V, cn +j D) CD j) cn o> 0} CO ( - h-o cn T3 D) ra \ cn -a 7 3 >k ra ^  4-» L i -5 5 a> . +-> cn cn O Major and Minor Probes • Tl • t3 • T2 Figure 3-5 Block I progress: Goal #3 3.2.4 Goal #4: Glide and Liquid Clusters Establish CCVC word shape using glides and liquids for C2 using onset-rime theory and an oral motor approach. Shaun produced only one token of CCV at Tj (i.e.[mjao]). The glides and liquids were chosen as C2 because they occur as C2 in word-initial clusters and were relatively strong segments in his system. CCVC wordshapes were chosen over CVCC wordshapes because the target glides and [+lateral] segments are only available for C2 in onset and not in coda position. Onset-rime theory served as a basis for treatment. C2 was a [+sonorant] consonant and could be epenthesized or lengthened to form a long onset (or syllable) to which the rime could be attached (e.g. [b9l] 'bull' + [u] 'oo' = 'blue'; [hi] 'he' + [u] 'oo' = 'Hugh'). Imitation and ehcitation 60 were optimal approaches because Shaun could produce all of the sounds easily. All Is/- clusters in word-initial position served as observation targets. The examiner was interested as to whether Is/ + sonorant clusters (i.e.. /si/, /sw/) would develop prior to other /s/-clusters given the above intervention. CG-clusters (i.e. Cw, Cj) were more accurate (2/9=22%) at Tj than [+lateral] clusters (i.e. CL: 0/11= 0%). A t T 2 CL-clusters were marginal (1/13=8%) in conversation,and CG-clusters were absent (0/4). Shaun used [1] for /fl/ (e.g.[lai] for 'fly'; [laowo] for 'flower') while he used [fl] for III (e.g. /fliv/ for 'leave'). These isolated productions were thought to be idiosyncratic and were not analyzed with the rest of the data. The observation goal of /s/-clusters demonstrated no improvement between Tj and T 2 . Minor probe data indicated that immediate learning following intervention was taking place for CG-clusters (trained: 3/4=75%) and CL-clusters (trained: 2/4=50%, untrained: 1/3=33%). Shaun's progress on goal #4 during Block I is summarized in Figure 3-6. 61 80 +-> D>70 4 CO H 60 4 -g 50 < x 40 ^  % 30 | 20 1 ra 2 10 ^ n _ l o 1 U <u i CP +-» CD V) CD V) o > c n 3 L_ CO o CO o 1 - 1 -U to k_ CU i te i "O cu +-< 10 in cu to CU CO X ! cr 3 3 o O "co U CO (J Tr CG-Tr CL-T3 CU cu 4-> I i id _ J o 1 v_ u i— cu cu • M 4-> cu CO 05 CO D) 3 CO 3 i ~ ro u 1— ro o ( - h -cn tn X I 3 O o Major and Minor Probes 1T1 B t 3 "T2 Figure 3-6 Block I progress: Goal #4 3.2.5 Summary of Block I Progress Most of the therapy goals for Block I were segmental in nature. Changes to Shaun's surface output feature geometry (Figure 3-7) were evident in the major probe at T 2 , following the first block of therapy. At the end of therapy for Block I goals, the following changes at Root, Laryngeal and Place were observed: 1. Root a. [+continuant] changed from emergent (55/134=40%) to developing (84/116=72%). b. [+lateral] was established (6/8=75%) in obligatory contexts and overgeneralized to some /w/-initial words (e.g. [lsrvin] vs. [wcrvin]). 62 2. Laryngeal a. Shaun demonstrated an increased control of voicing for onsets and [+voice] was no longer used as a default onset feature. b. Aspiration was established phonetically in combination with Coronal Place (9/10=90%) but not in combination with Labial (0/17) or Dorsal (0/16). 3. Place a. Dorsal was developing (33/66=50%) with an emergent link between Dorsal Place and [+nasal] (4/29=14%). b. Coronal was no longer a major default Place and [-anterior] was established (24/31=77%). c. The affricates [tf ] (8/8=100%) and [d3] (6/6=100%) were established. d. No generalization was noted for observation targets IQ, 6, J /. 63 ([+cont]) [+nasal] (([+spread glottis])) Labial ([+round]) +voice] [+son] ([+cons]) lateral v (Dorsal) (([+ant])) Figure 3-7 Feature geometry: Shaun T 2 Progress on the prosodic goal of establishing the clusters CG and CL word-initial clusters was minimal according to the major probe taken at the end of Block I (10% gain). However, minor probe data indicated short-term learning gains. The /s/-cluster observation goal showed no gains in the absence of direct intervention. 3.3 Block II Progress on Block I Goals Some of the goals targeted in Block I continued to demonstrate progress with and without intervention during Block II. Other goals were indirectly targeted when clusters containing 64 intervention target segments were trained. Finally, the CG and CL-clusters were reviewed during the final four sessions of the intervention period. Progress and data from these Block I goals during the Block II intervention period will be reviewed prior to the presentation of Block II goals and data in Section 3.4. 3.3 1 Goal #1: Indirectly Treated Establish the contrast/link between default segments lh,w,jl and other consonants word-initially using onset-rime theory and an awareness approach. Although therapy following T 2 was not specifically designed to target this goal, other goals such as mapping /f,v/ to word-initial position and word-initial Isl clusters were directly relevant. At T 3 , further improvement in onset production of [+continuant] consonants (114/126=90%) was observed and default production in onset position was correspondingly reduced (13/187=7%). Codas were established (54/61=89%), demonstrating a T 2 to T 3 proportional gain of 37% and 33% for onsets and codas respectively. There was no progress for the observation goals / #(0%) and / d (0%) from T 2 to T 3 . The data in Table 3-3 represent a subset of probe results between T t and T 3 : 65 Table 3-3 Goal #1 example data T r T 3 : Shaun 'gloss' /target/ - T 2 T 3 ' f i v e ' /fare/ tjai] [jar] [fail 'four' /foj/ [hoo] [hoo] [foo],but'for'-> [hoo] ' v a n ' /vaen/ [haen] [haen] [vaen] 'sun' / s A n / [cKn], [ h A n ] [ s )An] a , [ s ) a n ] [s>An], [ S A n ] 'side(s)' /said(z)/ [jai] [saitti] 'zipper' /zip / [hcbo:] feipoo], [z)ipo], [d3ipo] [zipo] 'shoe' /fu/ Lfu] [fu] [s)fu]b, [s)ju], [s)u] 'shower' /f aow / [haowo] ftfaows] [faowo] 'thing' /eirj/ [ h m ] [hm], [fm] 'that' /&33t/ [dae?], [daeth] [dart] [daen] a. [s>] refers to a retracted/s/ b. [s>/] refers to a phone that is initiated as an [s] and ends in the production of an [f] 3.3.2 Goal #2: Untreated in Singletons/Treated in Clusters Establish Liquid III word-initially using onset-rime theory and an oral motor approach. Shaun maintained his mastery of WI III at T 3 (28/28=100%). He overgeneralized its use in the case of WI /w/ and /f/ for two words at T 2 and for /w/ only at T 3 as seen in examples of his productions over time in Table 3-4. The level of establishment of SI III was maintained at T 3 (9/16=56%). No change in the observation goal lit in WI or SI position was observed (T2: 0/8; T 3 : 0/39). The data in Table 3-4 represent a subset of probe results between Tj and T 3 . 66 Table 3-4 Goal #2 example data T rT3: Shaun ' g l o s s ' / target / T i T 2 T 3 Ta' ./la/ Pja] [ia] [la] ' l e a f ' / l i f / Gis>] [lif] ' l a u g h i n g ' / lcefirj/ tJ33S)m]] [fefm] ' c a l l i n g ' /kohq/, / k a l i n / [ tDl in] [ k a l m ] ( r s d [ i ] 3 ) ' w a v i n g ' / w e r v i r j / [ w e i h i n ] [ l e i h i n ] [ l e r v m ] / s d [ v w e i b i n ] , r s d , [ w e i b i n ] ' f e a t h e r ' /fs6/ [wcho: ] [tejo], [ k h o o ] [fc?ho:] a Shaun's production of the vowel [1] was slightly raised (rsd). 3.3.3 Goal #3: Untreated in Singletons/Treated in Clusters Establish Dorsal stops word-finally and map production to word-initial position using onset-rime theory and a combined awareness/oral-motor approach (combined seg-mental-prosodic goal). The major probe at T 3 followed a period of no intervention for Dorsal segments /k/ and /g/ although some /sk/ clusters were targeted between T 2 and T 3 . Final results indicated that Dorsal onsets (54/57=95%) remained stronger than codas (31/44=70%) although codas demonstrated a larger proportional gain between T 2 and T 3 (53%) than did onsets (34%). In summary, the feature Dorsal was absent at Tj and became established for non-nasal stops at 129/143 (90%) in all word positions by T^ When Dorsal nasals were included, the accuracy rate dropped to 145/216 (67%).The observation goal It]/ was emergent with marginal improvement (3%) between T 2 and T 3 . The data in Table 3-5 represent a subset of probe results between T1 and T 3 . 67 Table 3-5 Goal #3 example data T r T 3 : Shaun 'gloss' /target/ T i T 2 T 3 'coffee' /kafi/ [dati] [tafi] [kafi] 'cooking' /kokin/ [dodin] [kotin] [kokin] 'going' / goowirj/ [doowm] [goo win] [goowin] 'chicken' /tfikon/ [dsetin] [tfcdsn] [tfikm] 'again' /sgsn/ [oden] [agen] 'snake' /sneik/ [neit] [neik] [sneik] 'Pig' /pig/ [bah] [peg] [pig] 'thing' /Gin/ [hm] [hirj], [flTfl 3 3 4 Goal #4: Treated Establish CCVC word shape using glides and [+lateral] for C2. At T 3 , the CG clusters were accurate 4/13=31% of the time and the CL clusters in 6/42=12% productions with a total of 10/55=18%. There were 25/42=60% CL clusters containing epenthetic vowels atT3 (e.g. [g3laes], [k°lak]). If these are added to the total, Shaun achieved a CG and CL cluster accuracy rate of 35/55=64% at the end of the intervention period. Minor probe data taken at tg demonstrated development or Cl-type clusters (trained: 11/13=85%; untrained: 5/8=63%), Cw-type clusters (trained: 3/3=100%; untrained 1/1=100%) but no spontaneous productions of Cr-type clusters. Two of eleven echoic productions of Cr-type clusters were correctly produced. Some examples of Shaun's productions are provided in Table 3-6 below: 68 Table 3-6 Goal #4 example data T r T 3 : Shaun 'gloss' /target/ T i T 2 T 3 'music' /mjuzibaks/ [nruEt] [muhE?] [nruzik] 'plum' / p l A m / [ w A m ] [pkm] [pL\m] 'black' /bisk/ [was?] [week] [b 3tek] 'quiet' /kNvAI9t/ [WAIJEt] [ k W A l j l t ] 'glasses' /gtessz] fjetiae] [lzesiz] 'glove' / g l A v / [ W A V ] [lAV] [g 3lAv] 'flower(s)' /flaow (z)/ [ w a n ] [laowo] [vaowo], [laowo], [flaowoz], [faok>z] 'sleeping' /slipirj/ [ w i p m ] [wipin] [s:wipin] 'sweater' /sWEt / [ w E d g ] [ w E d o o ] [ s w E d o ] , [s'Wdo] 69 Progress on goal #4 during Block II is summarized in Figure 3-8. Figure 3 -8 Block II: Goal #4 3.4 Block II Goals, Rationale and Progress 3.4.1 Goal #5: Word-Initial HI and NI Map word-final Ifl and Ivl to word-initial position using onset-rime theory and both awareness and oral motor approaches. 70 This goal was a partial "recycling" of the first goal of linking [+continuant] to [-(-consonantal]. While this link occurred for HI and Nl in word-final position, they were not established in word-initial position. Onset-rime theory was chosen to map (also 'copy', Bernhardt, 1994) HI and Nl from coda to onset. A mainly awareness approach was used because Shaun's cyclical dysfluency was active at the beginning of Block II. The observation targets IQI and 16/ in word-initial position were chosen as in goal #1. The reader is reminded that at T l 5 the production of HI and Nl in onset position in (3/22=14%) was relatively weaker than in coda position (2/8=25%). At T 2 , codas (6/8=75%) were strongly established while onset production of HI and -Nl (4/19=21%) was emergent. At T 3 , a substantial improvement for HI and Nl was observed in onset position (56/70=80%). Minor probe data taken at t4 indicated that (Labial, [+continuant]) was developing (trained: 8/12=67%; untrained: 5/7=70%) but the voicing contrast was unstable (e.g. /fit/ -> /vit/). There was no change in the observation targets IQI and 16/ during Block II. Examples of Shaun's use of HI and Nl over the course of the entire intervention program are found in Table 3-7. Table 3-7 Goal #5 example data T r T 3 : Shaun 'gloss' /target/ Ti T 2 T 3 'fish' Ifif/ [wath], [fif] (E) [ f : w e ] (E) [WAS) ] , [ W A f ] , [ W E T ] [fis] 'five' /farv/ fjai] m [fai] 'four' /fw/ [hoo] [hoo] [foo],but'for'->[hoo] 'van' /vaen/ [Iran] [haen] [vaen] 'thing' /em/ [hm] [tun], [fnj] 'that' /&33t/ [dae?], [dae t h ] [da3?] [cten] 71 Shaun's progress on goal #5 during Block II is summarized in Figure 3-9. Major and Minor Probes O J2 O t4 • T3 Figure 3-9 Block II: Goal #5 3.4.2 Goal #6: WI and WF /sACIusters Establish sCVC and CVCs using mora " 'edge' theory and an oral motor approach. Shaun was able to produce CG-clusters and CL-clusters but had not yet used /s/-clusters. Mora 'edge' theory was used because of the extrametricality of Isl before stop consonants in a cluster (Kiparsky, 1979; Bernhardt, 1994). Further, voiceless consonants were developing word-initially and Is/ was being used more accurately as a singleton, it was decided to target /s/-clusters in onset and coda positions. In the present case, Shaun was asked to produce Is/ + 'stop consonant-initial' word (e.g. 'bin'), with an decreasing length of time between the Is/ and the consonant (C) until the two met as the cluster 'sC' (e.g. sCVC: 'spin'). The observation targets were /J / clusters 72 (although, Ii I clusters were to be introduced during the final session) and the Is/ + sonorant clusters (i.e. /si, sw/). The /s/-clusters were not targeted for intervention and none were observed until T3, following a period of intervention. Results at T 3 were as follows: 1. Word-initial clusters were developing (22/45=49%) accuracy. 2. When productions involving initial I si lengthening (e.g. [s:wipin] for 'sweeping'; [s:too] for 'star') and epenthesis (e.g:[samal] for 'small') were included in the total, Shaun's accuracy for word-initial /s/-clusters increased (36/45= 80%). 3. Word-final /s/-clusters were established (24/29=83%) by T 3 , making word-final/s/-clusters more accurate than word-initial/s/-clusters. 4. Observation goals included word-initial7r/-clusters (Cr) and word-final clusters (CC). a. The Cr-clusters demonstrated a marginal increase in accuracy (1/47=2%). b. A CG-initial cluster substitution was made for /r/-cluster target words (e.g. [bwin] for 'bring'). e. Use of epenthetic vowels (e.g. [b^WA/m] for 'brushing'; [d°WAm] for 'drum') improved Shaun's score (13/47= 28%). d. The CC clusters, produced as homorganic stops by Shaun, were emerging (2/12=17%) atT^e.g. [wEnt] for 'present'). Intervention for word-final /s/-clusters provided between T 2 and T 3 , resulted in the establishment (36/43=84%) of word-final stop clusters. During minor probe #5 (15), /sC/-clusters 73 appeared to be developing but many were produced in delayed echoic response (8/14) or with initial I si lengthening (3/14=21%). Examples of Shaun's use of clusters over the course of the entire intervention program are found in Table 3-8. Table 3-8 Goal #6 example data T yT^: Shaun 'gloss' /target/ T - i T 2 T 3 'sleeping' / slipirj/ [wipm] [wipm]. [s:wipin] 'sweater' / SWEt / [WE d a ] [wEdoo] [ s w E d o ] , [s°WEdo] 'snake' /sneik/ [neit] [neik] [sneik] 'star' /staj/ [da:], [dab] [doo] [s:too], [s:to:] 'musicbox' /mjuzibaks/ [mu?Ebat"i] [muhsba?] [muzibaks] 'mask' /maesk/ [mae?] [mask] [maesk] 'six' /siks/ [hi?] [s>i] •[s>iks>] 'pink' /phrnk] [pwmt], [pi?] [pink] [piijk^ ] Shaun's progress on goal #6 during Block II is summarized in Figure 3-10. 74 100 T 90 CU 00 I— 3 < 5 ro 80 70 r 60 50 40 30 20 4 10 ^» CO S IS O ) <J 03 U O CO ers WF ster 3 CO CO 3 _Q o O CO ^ CO -6 § c u — £ ° 'ro co 5 CO 5 b CO T3 c u ' r o 5 CO 1— CO -6 c u *-> 1— c u ne CO 3 CO ' r o o 3 -o 5 CO CO i l CP — CO <-> 1 -ro U cu c n th CO ™ ^ ' ^ CO c L L 5 c u • w CO 3 = 2 i c CU c n (J c u ro j — a CD 1 — ' u o Major and Minor Probes T2 Bt5"T3 Figure 3-10 Block II: Goal #6 75 3.4.3 Summary of Block II Progress (T 2-T 3) At the end of therapy for Block II goals, the following changes at Root, Laryngeal and Place were observed: 1. Root a. The feature [+continuant] was established (336/375=90%). 2. Laryngeal a. The segment [h] remained as a default but with reduced frequency of use. A more established default pattern emerged in which [h] appeared for the [-voice] segments /f,s,6/ in word-initial position and for [+voice] segments lv,bl in medial position. b. Aspiration appeared occasionally on Dorsal stops (2/42=5%). 3. Place a. Dorsal Place was established (129/143=90%). b. A small increase (4%) in linkage between Dorsal Place and [+nasal] was noted. In summary, by T 3 Shaun's surface production feature geometry was almost fully specified with the exception of ffspread glottis] which was developing (54/140=39%), and [+distributed], which was unestablished. Links between Dorsal Place and [+nasal] were emerging: 12/67=18%. The low, back vowel lal was developing: 58/65=65%. Figure 3-11 represents Shaun's T 3 surface production feature geometry. 76 r+nasal] [+cont]-[+lateral] Laryngeal ([+branching]) ([+spread glottis]) Labial [+round] Coronal Dorsal [+hi] [+lo] [+back] [-ant] Figure 3-11 Feature geometry: Shaun T 3 Shaun's progress on prosodic targets during Block II was as follows: 1. Mapping of WF HI and NI to WI position: onsets developing (78/117=67%) 2. WI sC-clusters: developing (22/45=49%) 3. WF Cs-clusters: established (24/29=83%) 4. Observation goals: a. Cr-clusters: marginal (1/47=2%) b. CVCC-type clusters: developing (4/6=67%) 77 c. Closed syllables: CVC established (78%), CVCVC developing (69%) 3.5 General Observation Goals, Rationale and Progress Each goal targeted for intervention in this study had a related observation goal. The following two observation goals acted as general indicators of change in Shaun's phonological system. Intervention would have been provided at some point during or after the study had there been no indicators of progress from the major and minor probe data points. However, both the vowel and wordshape observation targets were established by the end of the second block of intervention. 3.5.1 Vowel Observation Goal To observe the development of the vowels /V, /Wand I d in the absence of intervention. At Tj, the vowels and diphthongs in Shaun's phonological system were established (See Appendix Table A-5) with the exception of a few irregularities which appeared to be contextually conditioned. Some examples of Shaun's vowel use are shown in Table 3-9. Table 3-9 Vowel example data Tj-T 3: Shaun 'Gloss' Target Ti T 2 T 3 'lid' /lid/ (jet] [lid] 'zipper' /zip / [hebo:] [zipo] 'fish' /fiJV [wath] fwef] [fis] 'chicken' /tfikan/ [daetin] [t/Edcn] [t/ikin] 'doll' 7dal/,/dol/ [doo] [do°] [dal] 'call' /khal/,/khDV [kD] [ko:l] 'cold' /khDld/ [ko:d] 78 As this goal was not to be a focus of intervention but was of some interest, the development of the vowels III, lot and Id was monitored as summarized in Table 3-10. Table 3-10 Vowel development: Shaun Vowel/ Diphthong Ti T 2 T 3 % gain T r T 3 x/y % x/y % x/y % III 23/53 43% 21/54 39% 143/171 84% 41% hi 3/3 100% 4/6 . 67% 15/17 88% (-12%) Id 19/35 54% 14/22 64% 58/89 65% 11% All vowels were established without any direct intervention by T 3 with the exception of /at. 3.5.2 Wordshape Observation Goal To observe the development of the CVC, CVCV, and CVCVC wordshapes in the absence of intervention. Although Shaun used a wide variety of wordshapes at Tj (See Appendix Table A-6) only basic CV wordshapes were established. Given Shaun's emerging ability to produce many different wordshapes and his limited segmental repertoire, CVC, CVCV and CVCVC wordshapes were observed without intervention for the period of Tj -T 2 to determine whether an increase in the variety of segments available would enable a larger wordshape repetoire to become established. No direct intervention for C2 in CVC or for C3 in CVCVC words was provided in either Block I or II. At T 2 , CVC and CVCVC were developing (34/60=57% and 12/26=46% respectively) and CVCV was established (11/13=85%). By T 3 , CVCVC was developing (38/55=69%) and CVC was established (143/183=78%) at a level comparable with CVCV. The development of these wordshapes was monitored as summarized in Table 3-11. 79 Table 3-11 CV, CVC, CVCV and CVCVC gain T r T 3 : Shaun Word Shape T i T 2 % Gain T 3 % Gain X/Y % X/Y % X/Y % CV 53/60 88% 23/24 96% 8% 56/59 95% (-1%) CVC 50/127 39% 34/75 45% 6% 143/211 68% 23% CVC (-WF/l,r/) 50/116 43% 34/60 57% 14% 143/183 78% 21% CVCV 8/18 44% 11/13 85% 41% 37/47 79% (-6%) CVCVC 7/29 24% 12/13 36% 12% 47/88 53% 17% CVCVC '(-WF /l,r/) 7/21 33% 12/26 46% 13% 38/55 69% 23% 80 3.6 Summary of Progress T-1-T3 Figure 3-2 summarizes Shaun's initial performance CT{), progress during the intervention period (T2), and his final results (T3) on targeted therapy goals. The data used to generate Figure 3-12 are found in Table A-4 on page 173. T l T2 T3 Major Probes Over Time — • — — [+consonantal]vs. [+sonorant] • w i / I / — i t — "Dorsa ls / k , g / ~ ~X~ CG and CL clusters — * - - W I / f , v / * ! * ! * ! * ! = ! : ! Wi / s / clusters — 1 — " W F / s / clusters Figure 3-12 Summary of progress: T r T 3 81 3.7 Comparison to Original Efficacy Study Comparisons involving small numbers of subjects do not yield substantive conclusions. The following results are intended to demonstrate the progress and the relative proportional gains of Shaun, Eddie and the six matched subjects from the original efficacy studies. Table 3-12 summarizes the timing details of intervention provided to the subjects and the matched controls. Table 3-12 Study compari son data Study This Study Bernhardt (1990) Bernhardt (1993a) Subjects SI, S2a M l , M2 b M3-M6C Length of sessions (minutes) 45-60 30-60 30-60 Frequency of sessions 2x/weekly 3x/weekly 3x/weekly Total # of sessions 30 48 48 Range of total intervention time (in minutes) 1350-1800 1440-2880 1440-2880 Total # of weeks 16 18 16 Total # of months 5-6 5 5-9 Total # of probes 3 4 4 a. Sl=Shaun, S2=Eddie b. Ml=Jeremy, M2=Sean c. M3=Miles, M4=Roger, M5=Faith, M6=Serena 3.7.1 Percent Consonant Correct: Relative progress Figure 3-13 summarizes the relative progress Of the two subjects from this study with six matched control subjects from the Bernhardt (1990-94) studies. Shaun made proportional PCC gains of 28.5% which placed him 8.25% above the mean of the comparison group (x=20.25; s.d.= 12.18). Proportional gain on PCC was smaller for Shaun (SI) than for M l and M2 (Bernhardt, 1990) who received between 1440 and 2880 minutes of intervention three times weekly over a 82 period of five months. However, Shaun's relative progress on PCC exceeded that of M3, M4, M5 and M6 (Bernhardt, 1994) who received between 1440 and 2880 of intervention three times weekly over a period of five to nine months. o 4-1 ra 70 i 60 50 40 30 20 10 0 m 1 S1 1 S2 M l M2 M3 S u b j e c t s M4 M5 M6 PCC Initial 1 = 1 PCC Final • % Change Group x=20.25 Figure 3-13 PCC progress relative to original efficacy study 3.7.2 Wordshape match: Relative progress Figure 3-14 summarizes the relative progress of the two subjects from this study with six matched control subjects from the Bernhardt (1990-94) studies. Shaun made smaller proportional gains for WSM than for PCC. Shaun's WSM proportional change of 12% placed him 4.63% below the comparison group's mean score of 16.63% (s.d.=13.45). Shaun's WSM progress was substantially lower than that of M l and M2 but his score was comparable to the average of WSM scores (10.5%, s.d.=7.72) achieved by M3, M4, M5 and M6. Figure 3-13 and Figure 3-14 were constructed using the data in Table 3-13. 83 u ra 70 60 50 40 30 20 10 0 I I SI 1 I i i i I I i . I I I i S2 Ml M2 M3 Subjects M4 M5 I P M6 1^23 WSM Initial L = l WSM Final • % Change Group x=16.63 Figure 3-14 WSM progress relative to original efficacy studies Table 3-13 PCC and WSM comparison scores with original studies Subject PCC Initial PCC Final % Change %WSM Initial %WSM Final % Change Shaun 40.00 68.50 28.50 42.00 54.00 12.00 Eddie 44.00 60.50 16.50 39.00 46.00 7.00 M l 29.00 65.00 36.00 35.00 65.00 30.00 M2 31.00 68.00 37.00 21.00 63.00 42.00 M3 45.00 49.00 400 31.00 47.00 16.00 M4 36.00 48.00 12.00 32.00 34.00 2.00 M5 29.00 40.00 11.00 39.00 57.00 18.00 M6 22.00 39.00 17.00 34.00 40.00 6.00 "x 34.50 54.75 20.25 34.13 50.75 16.63 S.D. 8.12 12.23 12.18 6.51 10.93 13.45 84 4 Eddie Results 4.1 Initial Phonological System (T-,) 4.1.1 Segmental Analysis At the initial assessment, Eddie used complex sentences and demonstrated no apparent reluctance to speak. Eddie's phonetic inventory at the time of the initial assessment (T^ is summarized in Appendix Table B-1. Established categories of the segmental system (at least 75% matches with the adult model) were: 1. Nasals and glides 2. Labial and dorsal stops 3. Labiodental fricatives. In terms of the feature hierarchy, the following specified features were established: 1. Root: r+nasal], [+consonantal], and [+sonorant] 2. Laryngeal: [+voice] 3. Place: Labial, [+round] as a terminal feature, Dorsal. Default use will be discussed in Section 4.1.2. The specified feature geometry for Eddie's phonology at T1 is summarized in Figure 4-1. 85 Partially established features of Eddie's phonology were: 1. Root: a. [+continuant] was developing in WI position (50/95=53%) for segments HI and NI. b. [+continuant] was emergent in WF position (44/154=28%) with Coronal-Dorsal default Place. 2: Laryngeal: [+spread glottis] was emergent (15/43=35%) with aspiration being used occasionally in WF position. 3. Place: Coronal was developing (44/154=28%) for /t.dj/1. The following features were not established at T^ 1. Root: [-(-lateral] III 2. Branching of [continuant] for /tf, dP,/ 3. Coronal Place /t,d,s,z/, including [-anterior] Is, f/ and the [-(-distributed] interdental fricatives 70,6/. Segments produced in onsets had more matches (WI: 103/202=51%; SI: 48/66=73%) than those produced in coda position (SF: 6/21=29%; WF: 60/165=36%)2. In terms of features, [+continuant] was 50% accurate in WI position but only 39% in WF position. 1. The productions of /t,d/ as Jjt, dj were included in this total. Sometimes /j/ is considered as Coronal-Dorsal Place. 2. WI=word-initial; SI=syllable-initial; SF=syllable-final; WF=word-final 86 ([+cont]) [+son] I ((Coronal)) ([+round]) Figure 4-1 Feature geometry : Eddie T j 1. In the feature geometry figure, dashed lines represent partially-specified features and dashed circles repre-sent default features. Feature geometries in this thesis represent surface output specified features of the child's phonological system. Table 4-1 provides a summary of Eddie's segmental analysis. Table 4-1 Summary of segmental analysis:Eddie T l Tier Established Partially Established Absent Root [+nasal] [+consonantal] [+sonorant] [•continuant] [+lateral] Complex branching of [continuant] Laryngeal [+voice] [+spread glottis] Place Labial [+round] Dorsal Coronal [-anterior] [•distributed] 4 .1.2 Default Use Eddie used both place and manner default features in default segment production. Table 4-2 represent Eddie's use of default segments for underspecified segments in word-initial and initial-word position at Tj. The relative strength of that default is represented by the length of the line connecting it to each segment and the brackets enclosing the phonemes.1 Examples of words that resulted from the use of defaults are provided below each diagram. In word-initial position, Eddie used Dorsal default Place with [-continuant] manner. In word-final position, his use of [k] for IQI and Is/ indicated Place (i.e. Dorsal) and manner (i.e.[-continuant]) default features; his use of \x\ represented a Dorsal Place default only as the [•continuant] feature of Is/ and IQI was preserved. 1. ((())): marginal 1-10%; (()): emergent 11-40%; (): developing 41-74%; { }: one instance only 88 Table 4-2 Dorsal default diagrams: Eddie T i Word-initial position Word-final position [k]^ w ((/&/)) e.g. 'tub' -> Tk/vbJ, 'chair' -> [khej], 'this' -> [gik], 'dolly' -> [gaji] [k ] ^ ^ _ ( / e / ) ^^^-^i /s /} e.g. 'watch' -> [wakh], 'mouth' -> [mAok], 'bus' -> [bak], 'teeth' -> [kix], Eddie's non-nasal coronal [-continuant] segments were retracted or palatalized (28/94=30%). For example, /too/ -> [t)oo], /dasoV -> [d)aed], /gAnV -> [JAm]. Otherwise they were produced as Dorsal stops. His n a s a l segments were palatalized (34/51=67%). For example,/nAIn/ -> [pAin]). Eddie's Coronal [-(-continuant] segments were palatalized (19/39=49%) For example, /nooz/ -> [noo?]). Finally, in SI position, 5/12=42% of coronal stops and nasals were produced as the glide [j] (e.g. / m A n i / -> [mAJi]; /daedi/ -> [daeji:]; /fe6/ -> [fsjs]). Eddie also appeared to be using Labial defaults other than the expected/LJ/ -> [w] as summarized in Table 4-3. 89 Table 4-3 Labial default diagrams: Eddie Tj Word-initial position Word-final position [p] {/e/} ^ ^ ( ( / J / ) ) e.g. 'thumb'>[pAm], 'van'>[baen], 'these'>[biz], 'read'>[bid] [p] IW [b]—(/f/) e.g. 'black'>[waep], 'leaf'>[wib] In summary, 134/421=32% of Eddie's segments were produced with default features. Palatalized (Dorsal-Coronal range) segments accounted for the majority (87%) while Dorsal and Labial features appeared in 35% and 12% of the default productions, respectively. 4.1.3 Prosodic Analysis Eddie's syllable structure analysis revealed the following details: 1. Open vs. closed syllables: a. Monosyllabic words were correctly produced in open (CV: 22/27=81%) and closed (CVC: 64/83=77%) syllables. b. Bisyllabic words were produced less accurately than monosyllabic words in open (CVCV: 8/19=41%) and closed (CVCVC: 10/25=40%) wordshapes. 90 2. Cluster production: a. Some clusters were produced using a glide in the C2 position including: 'music' [mjugik], 'flower' [fwAoa], 'three' [fwi], and 'sweater' [fwcjo]. b. Although 'thumb' does not contain a cluster, Eddie produced it as [fwAm]. c. Word-final cluster production was limited to one instance of CVCC (e.g. 'jump'). d. Consonant sequences were also demonstrated over syllable boundaries (e.g. CVC.CV: 'finger' [fmda], 'candle' [k^ndooJiCVCCVC: 'jumping' [dAmpm], and CVC+CVG: 'drumstick' [gAmkik]). With the exception of 'drumstick', consonant sequences across syllable boundaries were homorganic. 3. Total Wordshape Match (TWM) was 94/258 (36%). 4. Maximum word length demonstrated was four syllables (i.e. 'sewing machine')1 and he generally matched the number of syllables in the target word. 5. Typical word length was one syllable (157/236=67%). 6. Maximum word shape was CVV.CV+CV.CVC (i.e. 'sewing machine')2 although he had established the use of CV and CVC wordshapes only at Tj. 7. Developing wordshapes included: CVCV, CVCVC, and CGV. The wordshape CVCC was only marginally established. 1. The maximum word length for a non-compound word was three syllables (e.g. 'television' and 'crocodile') 2. The maximum word shape for a non-compound word was CV:.GV?.CVV (i.e. 'crocodile') 91 4.2 Block I: Goals, Rationale and Progress Intervention goals were chosen to optimize Eddie's phonological development rather than strictly following an alternating treatment design. All of Eddie's segmental intervention goals were focused on increasing use of the 'front' articulators (i.e. alveolar ridge, tongue tip, teeth and lips) in the production of speech sounds. In addition to the phonological goals, oral-motor exercises (e.g. licking peanut butter off of alveolar ridge or top lip, tongue clicking, etc.) were practised during the first five minutes of Block I sessions. Each goal, the rationale for its choice and progress achieved during the first intervention period is listed below. Measures of progress included the major probe at T 2 , minor probes at tj, t2 and t3, and intervention session data. 4.2.1 Goal #1: WI and WF /t/ and 161 Establish specification of Coronal Place stops using onset-rime theory and an aware-ness approach. Eddie was using Dorsal as his default Place, possibly due to his highly arched palate and underdeveloped alveolar ridge. The goal was to establish a 'true' default value (i.e. the universal Coronal default) and to develop experience with the most frequently occurring place of articulation in the English language. The Coronal stops /t,d/ were chosen because default productions for [-continuant] were less variable than for [+continuant] Coronals. Moreover, concentrating on [+anterior] was a way to draw attention away from Eddie's Coronal-Dorsal artieulatory setting. Onset-rime approaches were chosen in order to focus on the contrast between Coronal and Dorsal at the beginning and ends of words. Rhyming (e.g. 'moat' and 'boat') and alliterative (e.g. 'dough' and 'dam') words were incorporated into play activities. An awareness activity was chosen for two reasons: 92 1. Eddie was resistant initially to prompted production activities, possibly related to the difficulty with which he performed the oral motor exercises. 2. The rhyming and alliterative contrasts were expected to be more noticeable if uninterrupted. The observation goals chosen were /n/, because of its Coronal [-continuant] specification, and /si, because of its Coronal [-t-eontinuant] specification. The major probe taken at T 2 demonstrated that (Coronal, [-continuant] onsets (8/21=38%) and codas (7/19=37%) were developing, with a proportional gain of 17% and 1%, respectively. Increased accuracy of onset production coincided with a 36% increase in the accurate production of word-initial7t/ and decreased use of Dorsal and Labial defaults word-initially and glide defaults in syllable-initial position. The observation goals demonstrated different rates of progress during the period of intervention. The Coronal, [-continuant] nasal segment (i.e. /n/) was produced with 8/15=53% accuracy, demonstrating an improvement of 53% between Ti and T 2 . The Coronal, [+continuant] segment (i.e. Is/) was produced with 2/12=22% accuracy, resulting in an improvement of 17% between T 1 and T 2 . Minor probe data taken at tj indicated improvement on trained words (3/7=43%) and marginal improvement on untrained words (1/7=14%). Eddie's progress on goal #1 during Block I is summarized in Figure 4-2 below. 93 s 6 0 l Major and Minor Probes • Tl H t1 • T2 Figure 4-2 Block I progress: Goal #1 4.2.2 Goal #2: Root [+lateral] Establish [+lateral] III word-initially using onset-rime theory and an oral motor approach. Addition of the feature [-(-lateral] at the Root was the second goal in Block I of therapy. Eddie was also more comfortable being prompted to move his articulators so an imitative component was combined with his exercises mentioned above. Eddie used a tongue tip to upper lip strategy to produce III during the first block. The use of peanut butter and soft cheese on his alveolar ridge and having him lick it off with the tip and blade of his tongue was helpful as a tactile cue and motivation. The observation goal Ii I was chosen. No change was observed in Eddie's production of [-(-lateral] during the T 2 generalization probe (0%). Similarly, the minor probe at t 2 demonstrated that Eddie had not made any short-term 94 gains in learning the [-(-lateral] feature. Intervention data were more informative in that Eddie was able to produce WI III in words with 16/36=44% accuracy following a period of intensive cueing and feedback. 4.2.3 Goal #3: WI /sACIusters Establish SCVword-initially using mora 'edge' theory and a combined awareness and oral motor approach. Eddie was producing WI clusters with [-continuant] segments and needed to expand these to include a [+continuant] segment. The mora approach to therapy was used because of the extrametricality of 1st before stop consonants (Kiparsky, 1979; Bernhardt, 1994). Eddie was able to imitate homorganic clusters /sn/, /st/ and these were used as well as the Is/ + Labial clusters (i.e. /sp/ and /sm/). The /sk/ clusters were not targeted in keeping with the focus on articulators at the front of Eddie's mouth. Awareness activities were used to demonstrate how words changed when the /s/ was added to their edges. Eddie was prompted and cued in order to provide practise producing the transition from a [-(-continuant] segment to a [-continuant] segment. Clusters of the form stop + glide which were chosen as the observation goal. No WI /s/-cluster productions were observed at T 2 . Eddie used nasal air emission, tongue protrusion and other [+continuant] segments to mark his knowledge of the mora edge (e.g. 'snakes': [%jieik)^ ], [fwin]). In terms of the observation goal, Eddie produced stop + glide clusters (3/11=27%) demonstrating a 17% increase between Tj at T 2 . The minor probe and intervention data were the most informative concerning Eddie's progress on this goal. At Eddie indicated the C j-slot of the /s/-cluster by either releasing the air 95 nasally or protruding his tongue (without blowing) prior to producing C 2 of the cluster. Intervention session data were similar. Further, Eddie's production of /s/-clusters improved substantially as the session progressed. For example, in session 12 Eddie's production of /s/-clusters in the first 10 minutes of the session (8/28=29%) was much less accurate than his production of /s/-clusters during the second 10 minutes of the session (29.36=81%). Progress goal #3 during Block I is summarized in Figure 4-3. 60 o) 50 ra a 40 < JZ 30 4 Major and Minor Probes M Tl H t3 • T2 Figure 4-3 Block I progress: Goal #3 4.2.4 Goal #4: Link (Coronal, [+continuant]) Link Coronal Place and [^continuant] preceding high vowels HI and lul using onset-rime theory and a combined awareness and oral motor approach. Coronal stops were produced correctly in the context of high vowels. Since Eddie appeared to have the most phonetic control of coronal fricatives in this context, the goal was to establish awareness of the link between Coronal Place and [+continuant] and to work concurrently on oral motor practise of phonetic control. The coronal [-anterior] [+continuant] segment /{/ was chosen as an observation goal given that its phonological patterns had been similar to Is/ and Izl in the context of high vowels at T^ The major probe taken at T 2 demonstrated that Is/ and Izl were produced with 2/12=17% accuracy, indicating a proportional gain of 13%. These outcomes are for Is/ and Izl in all vowel contexts, given that there were few major probe words specifically chosen for the context 'preceding high vowels'. No progress was noted for the observation goal ///. The minor probe taken at t3 showed that Eddie's productions were imprecise, with lateral release, even though they were no longer produced with a Coronal-Dorsal default setting (trained: 1/12=8%; untrained: 2/8=25%). When lateralized productions were included, his accuracy rate improved for both trained (5/12=42%) and untrained (6/10=60%) targets. Intervention data indicated that Eddie's productions were most accurate in the SI word position. Eddie's progress on goal #4 during Block I is summarized in Figure 4-4. 97 50 -4-- 45 -v S 4 0 -h-- 35 -§ 30 -< £ 25 4-> '5 20 -Major and Minor Probes Figure 4-4 Block I progress: Goal #4 4.2.5 Summary of Block I Progress Most of the therapy goals for Block I were segmental. Changes to Eddie's surface output feature geometry (Figure 4-1) were evident in the major probe at T 2 . At the end of therapy for Block I goals, the following changes at Root, Laryngeal and Place were observed: J. Root a. Further establishment of [+continuant] from 67/125=53% to 59/94=63%. b. Use of SI [+sonorant] default was reduced slightly from 9/22=41% to 4/11=36%. 98 2. Laryngeal a. Aspiration was developing (18/27=67%) in obligatory contexts, up from 8/36= 22% at Ti . 3. Place a. Coronal had improved from emergent (57/166= 34%) to developing (61/117=52%) status. b. Coronal-Dorsal (i.e. palatal) Place defaults for Coronal Place segments were reduced from 84/199=42% to 30/117=26%. The major probe at T 2 demonstrated that there was limited progress for the Isl clusters, the only prosodic goal in Block I, following therapy. Eddie indicated awareness of two consonant slots by using nasal air emission and tongue protrusion in place of the [+continuant]. Although Eddie produced none of the stop + glide onset clusters chosen for the observation goal, he did produce the following clusters:?VVCC (100%), CVCC (17%). 99 ([+cont]) [+nasal] Laryngeal / / / , ' [+voice] ([-(-spread glottis]) [+son] > [+cons] Place Labia [+round] . / Dorsal "\ v / (Cor6nal) f ((Coronal- \ N Dorsal)) / Figure 4-5 Feature geometry T 2: Eddie 4.3 Block II Progress on Block I Goals All of the goals from Block I received a second cycle (Hodson & Paden, 1991) of intervention during Block II. Progress on Block I goals during Block II will be reviewed prior to the presentation of progress on Block II goals in Section 4.5. 100 4.3.1 Goal #1: Treated (One Session) Establish specification of Coronal Place stops using onset-rime theory and an aware-ness approach. At T 3 , Coronal Place stops were produced with 40/43=84% accuracy and onsets (24/26=93%) were somewhat stronger than codas (21/30=67%); demonstrating a gain of 11% and 16% T 2 to T 3 , respectively. The observation goals demonstrated different rates of progress during the period of intervention as summarized below: 1. Coronal nasal (i.e. [-continuant]) improved by 53% between Tj^  and T2and by 41% between T2 and T3 with an overall proportional gain of 94%. 2. The [+continuant] Coronal (i.e. Is/1) improved by 17% between Tj and T 2 and by 8% between T 2 and T 3 with an overall proportional gain of 25%. In summary, [-continuant] Coronal consonants were 69% more accurate than [+continuant] Coronal consonants at T 3 . Default production was correspondingly reduced by 15% for onsets and 36% for codas. In summary, from T r T 3 , proportional gains of 72% were observed in onset position and 34% in coda position. Default onset and coda productions were reduced by 53% and 60% respectively. The data in Table 4-4 represent a subset of probe results between Ti and T 3 . 1. Only the [-voice] [+continuant] Coronal stop was used as an observation goal as the [+nasal] Coronal stop was redundantly [+voice]. 101 Table 4-4 Goal #1 example data T r T 3 : Eddie ' g loss ' /target/ T 2 . T 3 'bootie' /budi / [buwi] [bu w i ] ' teeth' /tie/ [ k h i x ] [t h i ] , [t4t] ' toes' HSxaJ [ k ^ o ] [ ^ o o z 1 ] [t'kxjz] 't.v.' / t hibi/ ttbi] [tibi] [thibi] ' d o l l y ' / d a l i / [gaji] [daji] [da^i] 'do ' /du/ [gu] [du] 'dumptruck' /dAmptoAk/ [gAkAk] [dAmta:] ' nothing ' /naGirj/ [iiA:pin]/rnApiji] [nApiji] 'soap' /scop/ [hoob] [woop] [foop] ' some ' /sAm/ [^Am] lat. [c^m] [s)Am] 4.3.2 Goal #2: Treated (One Session) Establish [+lateral] III word-initially using onset-rime theory and an oral motor approach. At T 3 , Eddie produced 2/16 (13%) untrained generalization targets correctly. There was no change for the observation goal /a/. Intervention data indicated that Eddie's ability to produce III spontaneously in CV syllables was developing (10/16=63%). The difference between early (15/33=45%) and later (15/22=68%) productions in the session was reduced. The data in Table 4-5 represent a subset of probe results between Tl and T 3 102 Table 4-5 Goal #2 example data T r T 3 : Eddie 'gloss' /target/ Ti T 2 T 3 'laughing' / tofiq/ rjBBbiji] tjasprp] [jaefin] 'leaves' /livz/ [wifri] [jivs] 'lion' / k i j a n / [JAIJEJI] [lAijin] 'rabbit' /oaebat/ [WAbicn], [WAbit] [W33blt h ] [waebit] 'reading' /jidin/ [bijm] [wihn] [wijin] 4.3.3 Goal #3: Treated (One Session) Establish 1st-clusters word-initially using mora 'edge' theory and an oral motor approach. At T 3 , /s/-clusters were marginal (1/13=8%) in Eddie's general production (see Table 4-6). The observation goal, stop + glide clusters, were produced (4/30=13%). Table 4-6 Goal #3 example data T r T 3 : Eddie ' g l o s s ' /target/ T 2 T 3 'snake' /sneik/ [%nelk]a [cpei] [ x n e i k ] b 'star' /steal [kA:] [goJ] [kec?] ' s k y ' /skai/ IgAl] [(s)kai] 'smile' /smAil/ [%mAi] [f rriAijoo] 'music' /mjuzik/ [mjugik] [mijudik] 'quiet' / k w A i j a t / [kAl j l t h ] [k hAijst] 'snakes' /sneiks/ [%nei?] [%jieik>^ ] [nelks] 'jump' /djAmp/ [dAtnp] [gAmp] [dAtnp] a. %=nasal air emission b. x=tongue protruded in interdental position without air release. 103 4.3.4 Goal #4: Treated (One Session) Link Coronal Place and [+continuant] preceding the high vowels HI and Jul using onset-rime theory and a combined awareness and oral motor approach. The following outcomes are for Isl and Izl in all vowel contexts, given that there were few major probe words specifically chosen for this context. At T 3 , Isl was produced with 10/31=32% accuracy and Izl was produced once in the word 'buzzing' (1/8=13%), indicating a proportional gain of 8% between T 2 and T 3 and 21% gain in total. Defaults consisted of Coronal Place stops and glides. No progress was noted for the observation goal /J7 until T 3 when Eddie produced it correctly (3/5=60%). In summary, total proportional gains of 28% and 13% for Isl and Izl respectively were observed at the end of the intervention period. Examples of Eddie's production of major probe target words related to the intervention goal are provided Figure 4-7. Table 4-7 Goal #4 example data T r T 3 : Eddie 'gloss' /target/ Ti T 2 T 3 'see' /si/ [c'i] [s)i] 'zoo' /zu/ [gJu], [Ju]a Du] 'shoe' /fii/ [s'u] [?lu] LfJu] a. J=palatal fricative 104 4.4 Block II Goals, Rationale and Progress 4.4.1 Goal #5: Place (Coronal, [+distributed]) Establish coronal [+distributed] using onset-rime theory and a combined aware-ness/oral motor approach. This goal provided what Elbert & Geirut (1986) call a maximal contrast to the developing Coronal Place [-continuant] and to Dorsal. It also further strengthened Coronal Place link to [+continuant] by developing a [^ -distributed] branch. Finally, the target was visible and less dependent on oral structures, although his open bite may have made the interdental placement more difficult to master. Therapy began with an awareness approach to allow Eddie to observe tongue placement and to hear contrasts visually demonstrated in stories and pictures. Oral motor activities allowed for a variety of cueing and prompting methods given the visibility of the targets IQI and /&/. The observation goals were affricates with branching /tf, oty. Major probe data for this goal, which indicate 0% production of either IQI or Ibl are insufficient for describing the progress Eddie made on this goal. There was no progress on the observation targets until T 3 when Eddie produced /tf / (5/11=45%) and Id&l (0/4) in WI position. Default productions that were Palatal or Dorsal accounted for the unmatched segments (6/9 =67%). Minor probe data for IQI and Ibl indicated progress only for delayed echoic responses (1/10=10%). Eddie's progress on goal #5 during Block II is summarized in Figure 4-6. More information was available from the intervention data in which Eddie's pattern of 'warming up the system' resurfaced. Productions of IQI and Ibl early (4/9=44%) in the first sessions were inferior to those produced later (15/20=75%). 105 * 35 cu CD ro 30 I-X 25 3 TJ < 20 1 1 5 ro 2 5 0 CD <P i5 ^ I— O CU -co c T3 cu \ 'ro -3 i vCCJ 4-1 K CU >c CD <D Major and Minor Probes B T2 H t5 • T3 CO X I O Figure 4-6 Block II: goal #5 By the end of eight sessions targeting this goal, Eddie produced IQI and /&/ in syllables in non-elicited contexts (5/19=26%). Substitutions for the target sounds were typical in word-initial position (i.e. /6/ -> /f/; /6/ -> Idl) but were more unusual in syllable-initial position (i.e. /0/ -> /f,p,b,k/; /&/ -> 1)1). Examples of Eddie's productions are shown in Table 4-8 below. Table 4-8 Goal #5 example data T r T 3 : Eddie 'gloss' /target/ T i T 2 T 3 'thumb' IQ\ml [ p A m ] , [ f w A m ] [fAHl] [ f A m ] 'there' /&EJ] [dc:] [ d e 9 ] 'cherries' /tfcjiz/ [khei. is1)*] [tzs. iz] [cs. iz1] 'jumping' /d3Ampirj/ [ d A m p m ] [ d A m p m ] [ d A m p m ] 106 4.4.2 Goal #6: Recycle Goals #1-5. Recycle each goal prior to terminating therapy. A review of all goals seemed appropriate because therapy focused on [+anterior] and Eddie was to have a break from therapy over the summer. It also provided an opportunity for him to make the connection that all of the goals involved Coronal Place. The same approaches to therapy were used as in the first cycle. Information concerning progress on this goal consists of the summary of overall progress. Intervention session data can be found in Appendix Table B-6 on page 183. 4.4.3 Summary of Block II Progress (T 2-T 3) At the end of therapy for Block II goals, the following changes at Root, Laryngeal and Place were observed: 1. Root: Marginal appearance (2/60=3%) of [+lateral]. 2. Laryngeal: [+spread glottis] was established (85/87=98%). 3. Place a. Coronal Place was almost established (208/293=71%), demonstrating a proportional gain of 19% since T 2 . b. [-anterior] emerged (10/33=30%) as a segment rather than just as a default Place. c. Dorsal and palatal defaults remained but were both marginal at 7%. Features that had not yet emerged included [-(-distributed] and [+grooved] from Coronal. Figure 4-7 represents Eddie's surface production feature geometry at T 3 . 107 ([•fcont]) [+ s o n] Figure 4-7 Feature geometry T 3: Eddie No new prosodic goals were targeted in Block II. Eddie made progress on Goal #3 (/s/-clusters) which was reviewed in session 29. Prosodic gains were as follows: 1. /s/-clusters: marginal (1/13=8%) 2. Onset stop + glide clusters: CGVCVC emergent (17%) 3. Word-final clusters: CVCC emergent (21%). 108 4.5 Intervention Session Data: Block I and II Eddie's intervention session data provided the examiner with the most insight into his overall performance which varied greatly from session to session. His refusal to imitate was generally observed in the introductory session for new goals (see sessions) and his performance following breaks (see session 21) was usually poorer indicating a loss of progress. Eddie's performance also varied within sessions in that the first 10 minutes were usually characterized by fewer overall responses and fewer correct responses than the last 10 minutes (seen in sessions 11 and 19, not seen in session 24). Later in the session, following many prompted imitations, he often produced targets in delayed imitation with reasonable accuracy. Most of his spontaneous productions were produced at the single word level. Sessions including periods of non-compliance were usually accompanied by parental explanations of changed routines or low blood sugar. 4.5.1 Breakthrough Moments Three breakthrough moments were notable: 1. In session 10, Eddie named the letter ' C with only minimal lateral release. Given his success with [-continuant] Coronals in this context, an optimization goal was inserted into his overall plan to take advantage of the superior control he appeared to have in the context of [+hi] vowels. 2. In session 18, following two weeks away from the /s/-cluster goal, Eddie produced the word 'snowman' twice in spontaneous speech as [Gnoomazen], clearly marking a two element Isl cluster. 3. In session 25, Eddie was practising 'that' on his own, producing it correctly three times in a row. His mother had not noticed 'that' at home. 109 4.6 Summary of Progress TVTg Figure 4-8 provides a summary of progress on intervention goals. 100 Major Probes Over Time — - « • — ~ Coronals / t , d / "*""" a - ^ - w i / | / * — W I /s/-clusters - -X- - / s , z / \ _ V [+hi] a«*«vvi /e, "oV Figure 4-8 Summary of progress Tj-T 3 : Eddie In summary, both intervention and observation goals demonstrated progress with the exceptions of /Q,6I and /j/.The Coronal intervention targets /t,d/ made a T r T 3 proportional gain of 57% while the observation targets In! and /s/ showed proportional gains of 94% and 25%, respectively. 110 Progress on /s/-clusters occurred following the second block of intervention. Although it appears that word-final /s/-clusters developed more quickly, results were based on only one production at each major probe. Stop + glide clusters fell in accuracy by 14% between T 2 and T 3 but experienced an overall reduction of only 6%. Untargeted goals such as CVC, CVCV and CVCVC (see Table 4-9) all demonstrated proportional gains of similar magnitude to the intervention goal over the period of intervention, especially between T 2and T 3 . Table 4-9 CV, CVC, CVCV and CVCVC gain T r T 3 : Eddie Word Shape Ti T 2 T 3 % Gain x/y % x/y % x/y % CV 22/27 81% 15/19 79% 44/55 80 (-1%) CVC 27/57 47% 12/25 58% 47/83 57% 10% CVC (-WF /l,r/) 27/51 53% 12/22 55% 47/71 66% 13% CVCV 8/19 42% 5/10 50% 14/28 58% 16% CVCVC 3/15 20% 3/13 23% 12/39 31% 11% CVCVC (-WF/l,r/) 3/11 27% 3/9 33% 12/33 36% 9% 4.7 Comparison to Original Efficacy Study As mentioned in the previous chapter, comparisons involving small numbers of subjects do not yield substantive conclusions. The following results are intended to demonstrate relative proportional gains between Eddie and the subjects of the original efficacy study. Reference will be made to the group data found in Figure 3-13, Figure 3-14 and Table 3-13 in Section 3.7. A summary of the lengths of the comparison studies in terms of number of sessions, weeks, months and probes taken can be reviewed in Table 3-12 on page 82. Ill 4.7.1 Percent Consonant Correct: Relative Progress Eddie's PCC gain of 16.50 placed him 3.75% below the mean of the comparison group (x=20.25; s.d.=12.18) as shown in Figure 3-13 on page 83. Proportional gain on PCC was smaller for Eddie (SI) than for M l and M2 (Bernhardt, 1990) who received between 1440 and 2880 minutes of intervention three times weekly over a period of five months. However, Eddie's relative progress on PCC exceeded that of M3, M4, and M5 but not M6 (Bernhardt, 1994) who received between 1440 and 2880 of intervention three times weekly over a period of five to nine months. 4.7.2 Wordshape match: Relative progress Eddie made smaller proportional gains for WSM than for PCC. Eddie's WSM proportional change placed him at 9.63% below the group mean of 16.63% (s.d.=13.45). Eddie's WSM progress was substantially lower than that of M l and M2 but his score was comparable to the average of WSM scores (10.5%, s.d.=7.72) achieved by M3, M4, M5 and M6. 112 5 Discussion The purpose of this study was to investigate the effectiveness of nonlinear phonological theory as applied to assessment and intervention for two subjects in a typical clinical setting. The study addressed questions of clinical and theoretical importance. Outcomes will be discussed in terms of quantitative and qualitative measures taken within the multiple baseline design. The relative progress of the two subjects in the present study and six subjects from the Bernhardt (1990, 1993a) studies will be considered, especially in terms of the frequency of intervention and proportional gains in PCC and WSM. Theoretical implications of subject progress such as changes to underlying representation, output constraints and the interaction between phonological representation and phonetic implementation will then be discussed. Concluding remarks concerning the limitations of the present study and suggestions for future clinical research will be provided at the end of this chapter. 5.1 Original Questions The original questions addressed the optimization of intervention and the theoretical underpinnings of clinical research. In review, they were: 1. Optimization Questions: a. Will therapy derived from nonlinear phonological theory result in positive developmental changes for these subjects when carried out independently by a speech-language pathologist in a typical clinical setting? 113 Both subjects made phonological gains as a result of intervention based on nonlinear phonological theory. Shaun (SI) was dismissed from the direct service caseload and Eddie (S2) received another short block of intervention six months after the study was completed. b. Given the typical health unit setting in which therapy is available once or twice weekly, will twice weekly intervention be as effective for these subjects as three times per week therapy for selectively matched subjects in the Bernhardt (1990, 1993a) studies? Twice weekly intervention was at least as effective as intervention provided three times weekly. Shaun and Eddie's proportional gains on PCC and WSM were within the average group range of improvement. 2. Theoretical Questions: a. How do the subjects' initial and intermediate phonological systems inform us about issues related to phonological development within nonlinear phonological theory? Data collected before and during the intervention program provided additional evidence for nonlinear phonological theory in the areas of feature specification, feature cooccurrence, constraints and repair strategies. b. What is the nature of the interaction between phonology and phonetics in the phonologies of the two subjects? The nature of the interaction between phonology and phonetics was child-specific. For Shaun, consonant-vowel feature cooccurrence restrictions appeared to be phonetically motivated. 114 For Eddie, who presented with a Coronal-Dorsal phonetic setting, reduced overall progress in phonological therapy appeared to be related to his unusual oral structures. 5.2 Outcomes of NLP Intervention 5.2.1 Quantitative Progress on Trained Targets Shaun demonstrated no progress in his phonological skills between his screening assessment (age 33) and the initiation of intervention (age 3;5). Following the first block of intervention, Shaun's production of the trained, mostly segmental, targets made the expected improvement while his production of the /CG/ and /CL/-clusters appeared to decrease in accuracy. This apparent lack of progress may have been attributable to an inability to produce the recently-acquired phoneme (i.e., Ill) in a new wordshape (i.e., clusters). Minor probe data indicated that immediate learning of the /CG/ and /CL/-clusters had occurred. Trained targets demonstrated the most progress in the minor probe data with the exception of the first goal. Given Shaun's low expressive vocabulary at the beginning of intervention, some untrained items had to be elicited in a delayed echoic format. Shaun was an excellent imitator and his ability to use this strategy inflated his minor probe data for the first goal. Shaun continued to demonstrate progress on Block I trained targets during Block II, achieving conversational mastery of WI and WF ([-(-consonantal], [-(-continuant]) segments, WI and SI III; and WI lid and /g/. Word-final Dorsals were developing. Progress may have been related to the indirect treatment of segmental targets due to their inclusion in prosodic cluster targets during Block II. Given that this was an effectiveness study and that progress during Block I was so clearly related to intervention, the experimenter decided to optimize Shaun's ability to learn wordshapes by including many exemplars of clusters in the cluster treatment goals. Shaun made no progress on Block II goals during Block I but mastered WI If I and Nl and WF clusters following 115 one block of intervention. If epenthesis was not counted as an error, WI clusters were also established by the end of Block II. The /CG/ and /CL/-clusters failed to show generalization to conversational speech following two blocks of intervention. This result may have been due to the low frequency of /CG/-cluster vocabulary items and to Shaun's idiosyncratic reduction of /CL7-clusters to III singletons (e.g., 'flower' -> fjaowo]). Shaun was dismissed from phonology intervention following Block II and he remained on the caseload at the Richmond Speech and Hearing Clinic only for monitoring of his fluency skills. Eddie demonstrated no progress in his phonological skills between his screening assessment (age 3;5) and the initiation of intervention (age 3;7). Performance on the major generalization probes at the end of Block I indicated that Eddie made more progress on WI III and Idl (i.e. Coronal, [-continuant]) than on the goal that targeted the linking of the features (Coronal, [+continuant]) preceding high vowels. Major probes indicated no progress on WI III following Blocks I and II, and no progress for IQI or /&/ following Block II intervention and WI /s/-clusters showed limited progress by the end of Block II. Minor probe results for Blocks I and II also showed limited progress, which was demonstrated mostly for delayed echoic responses. The facility with which Eddie imitated target segments improved. The value of such an improvement will be discussed further in Section 5.2.4. Eddie's limited progress during Block I may have been attributable to several factors. There was a strong emphasis on Coronal Place which was difficult for him to produce. Given Bernhardt's (1994) findings that changes to the prosodictier are learned more quickly, a more optimal course of intervention may have been to choose a developing wordshape and train it for increased success. Further, it may be more difficult to fine-tune phonetic production than to learn major phonological category distinctions, especially given his oral-motor limitations and difficulty with discrimination tasks at the beginning of therapy. Finally, quantitative measures were not 116 sensitive to the subtle shift that was occurring from the Dorsal default to more Coronal-Dorsal (i.e. palatal) default productions which made a perceptual difference in intelligibility but did not appear as a change on the major generalization probes. One might conclude that Eddie made very limited progress during the course of the intervention program if quantitative data were analyzed alone. Given that quantitative data are not always sensitive to subtle changes in developing phonological systems, qualitative data were also examined. 5.2.2 Qualitative Progress on Trained Targets Both Shaun and Eddie demonstrated qualitative differences in their phonologies that were not measured by a strict child-to-adult target match count. Family members and neighbours reported an increase in speech intelligibility following the first block of intervention. For both of them, epenthesis and initial consonant lengthening in clusters emerged prior to mastery of cluster wordshapes. These strategies may have been related to the approaches taken in intervention such as using an epenthetic vowel for onset-rime condition stimuli (e.g., [pal] for /pi/) or lengthening of the extrametrical Isl in mora condition stimuli (e.g., [s:] for Isl). Eddie's strategy of protruding his tongue to 'mark' the Isl in /s/-clusters was self-initiated, possibly indicating that he understood that Is/ shared a similar Place feature with III which he also produced with a slightly protruded tongue. Nasal air emission on the /sn/ and /sm/ clusters indicated that Eddie was aware of the [-(-continuant] status of CI but was unable to provide a Place feature for both elements in the cluster. In general, it appeared that Eddie's phonological setting was shifting gradually from a 'back of the mouth' setting to a 'front of the mouth' setting. Combined quantitative and qualitative data relating to this phenomena are discussed in more detail in Section 5.4.4. 117 5.2.3 Observation Targets 5.2.3.1 Related Observation Targets The following generalization of trained features to untrained targets occurred in Shaun's productions: 1. After training of ([-(-consonantal], [+continuant]) versus [sonorant] in awareness tasks, the complex branching /tJ7 and /dj/ were established. 2. After training WI IM, SI and WF IM appeared. 3. The improved ease with which [J] was produced imitatively may have been related to his mastery of IM and the alveopalatals, since [J] requires the specification of ([-(-consonantal], [+sonorant]) as well as the [-anterior] tongue position. Productions of [J] in clusters were observed to be marginal at T 3 . 4. Vowel productions were mastered. If vowels share features with consonants, then this is not a surprising result. 5. The wordshapes CVC and CVCVC were developing. The wordshape CVCV was mastered although Shaun's proportional accuracy at T 3 was slightly lower than at T 2 . This could be attributable to a larger sample with more unfamiliar vocabulary taken at T 3 . The following generalization of trained features to untrained targets occurred in Eddie's productions: 1. The segments HI and Ivl generalized to SI word position. 2. The use of [j] as a SI default decreased with a concomitant increase in accurate production of segments in SI word position. 118 .2 Unrelated Observation Targets The following unpredicted changes occurred in Shaun's productions: 1. Shaun established the [+voice] feature. Also at Laryngeal, aspiration on Coronals but not Labials or Dorsals was established at T2 and aspiration on Dorsals was established at T3. Aspiration on Labial place segments was still developing at T3. Shaun's development of aspiration demonstrated a complexity constraint restriction in that it was established first with default place (i.e. Coronal), then with Dorsal (which is phonetically proximal to the larynx) and, finally, with Labial (which is phonetically distal from the larynx.) 2. No change on IQI and 161 which was related to the [-(-continuant] goal but no change was expected. 3. The wordshapes CVC and CVCVC were developing by the end of the intervention program. The wordshape CVCV was mastered although Shaun's proportional accuracy at T 3 was slightly lower than at T 2 . This could be attributable to a larger sample with more unfamiliar vocabulary taken at T 3 . The following unpredicted changes occurred in Eddie's productions: 1. Aspiration was established for Labial, Coronal and Dorsal segments. Given that aspiration is a phonetically-motivated feature, Eddie's improved articulatory control may have contributed to the establishment of [+spread glottis] in the absence of intervention. 2. [-anterior] (i.e. ///) emerged as a phonological category rather than a default production of Place. Intervention focused on [-(-anterior] and Eddie may have 119 been able to use [-anterior] phonemically following his phonetic mastery of [-t-anterior] segments. In summary, very little untrained progress occurred for Eddie which may be attributable to the nature (i.e. phonological and phonetic) of his disorder. 5.2.4 Imitated Versus Spontaneous Productions Shaun's imitation of words proved to be excellent and provided an indicator of his stimulability for new goals. Shaun's willingness to imitate and perform repetitive production activities may have contributed to his rapid progress in therapy. Eddie imitated few words at the time of his initial assessment. Further, he was reluctant to imitate modeled targets early in the intervention program. This reluctance to imitate may have been related to his low success rate for imitated responses or, perhaps, a strategy for controlling the activity. Improvements in Eddie's ability to imitate adult models of targets may have been related to (1) improved control of his tongue and lips in relation to his palate, (2) improved ability to discriminate differences between segments, and also to produce them, and (3) an increased level of comfort concerning the intervention session routines and expectations. Both Shaun and Eddie demonstrated quantitative and qualitative gains in phonological development as a result of intervention based on nonlinear phonological principles. Shaun, who presented with a primarily phonological disorder, made larger and faster gains than Eddie whose unusual oral structures and general health may have impeded his progress in intervention. 120 5.3 Optimization of Intervention 5.3.1 Twice Versus Three Times Weekly Intervention The question of frequency of intervention is a common one, and is raised by a variety of people including: (1) parents, who want the best for their children, (2) health administrators, who want to see resources used in the most cost-effective way; and, (3) clinicians who are trying to service many clients and who may have long waiting lists for these services (Hodson & Scudder, 1990). Effective intervention relies on the caregivers to provide follow-up practise at home to enhance the skills taught during the sessions with the speech-language pathologist, although this does not always happen. The rate of improvement can also vary with severity level of the disorder and characteristics of the individual(s) involved (Olswang, 1993; Holcombe, Wolery & Gast, 1994). Fraser and Avery (1985), reported on fluency intervention data based on caseloads during a two-year period. Their findings suggest that intensive fluency programs may not be as time-efficient in establishing new fluency skills as less frequent intervention over a longer period of time.1 Fraser and Avery concluded that much of the time spent in intensive intervention was devoted to stabilizing newly learned skills rather than learning new skills. Shaun (SI) and Eddie (S2), who received intervention twice weekly for a period of 16 weeks, demonstrated PCC and WSM gains which placed them within the standard range of the comparison group. Both Shaun and Eddie made larger proportional gains for PCC rather than WSM. Further, both made smaller proportional gains in PCC and WSM than M l and M2 (Bernhardt, 1990) and their progress on PCC and WSM exceeded that of M3, M4, M5 and M6 1. Fraser and Avery (1985) reported that the length of the establishment phase for fluency intervention varied with the intensity of the service provided. The intensive program took 40 clinic hours to complete over a three week time period. Intervention provided twice weekly for 10 weeks took a total of 20 clinic hours before skills were established while intervention provided once weekly for 15 weeks took a total of 15 clinic hours to attain the same goal. (Bernhardt, 1994) who received approximately the same amount of intervention as M l and M2. Shaun's poorer performance on WSM scores as compared to M l and M2 may be attributable to the complexity of the wordshape goals (i.e. clusters) and the recently established segments that were targeted in the clusters. Likewise, Eddie's WSM goals were also with clusters which may have been more difficult to establish than other wordshapes. Shaun and Eddie performed within the group standard deviation for both PCC and WSM measures. Their performances clustered with the matched subjects from Bernhardt (1993a) who received intervention from a variety of clinicians in the province of British Columbia for whom nonlinear phonological intervention was a new theoretical approach. In contrast, the matched subjects M l and M2 from Bernhardt (1990) received intervention from Bernhardt herself, who developed the application of the theory for phonologically disordered children and who was highly motivated to achieve positive results to report in her dissertation. In the present study, with a clinician who has an intermediate knowledge of the theory and its applications, intervention performed on a twice weekly basis was at least as effective as intervention performed three times weekly. 5.3.2 Optimization Issues 5.3.2.1 Faithfulness to Design Johnston (1988) argued that clinical research rarely adheres closely enough to the original design to make valid conclusions concerning its efficacy. In the case of the present optimization study, consultation with Bernhardt was conducted once or twice weekly and included questioning theory, discussing activities and making decisions about goals and timing choices that deviated from the original plan. Given this frequent consultation, it is very likely that the present study was 122 conducted with the precision with which interventions were originally operationalized (LeLaurin & Wolery, 1992; Wolery & Ezell, 1993). 5.3.2.2 Flexibility of Design Faithfulness to the research design is essential in efficacy studies in order to make scientific conclusions. However, the design must be flexible to be effective in daily clinical practise. Connell and Thompson (1986) discussed the use of flexibility to design or modify experiments. As far as changes to the research design, these were mainly timing and combination changes. Timing changes involved lengthening or shortening the number of sessions for which a particular goal was targeted and were based on intervention session data and clinical judgement. Combination changes involved using two therapy methods (e.g. awareness and oral motor) that were used separately in the original study within the same session. Other factors that had to be considered in making the research design flexible included the timing, ordering and altering of intervention goals in the context of individual characteristics of the children and their family priorities. In the present study, Shaun benefitted from an intervention program with a consistent time schedule. He appeared to have difficulty settling back into the intervention routine following a two-week break for the Christmas holidays. Eddie, on the other hand, benefitted from such breaks. His intervention program was slightly longer than Shaun's and included two, two-week-long breaks. While the design was able to document these differences in scheduling, it is difficult to relate the changes in the intervention plan to progress resulting from intervention. For example, it is difficult to answer the question: "How does a two-week absence affect the measurement of progress in a multiple-baseline design?" Given that clients and clinicians may have unexpected illnesses or vacations scheduled during the period of intervention, this question needs to be addressed in future conversations about effectiveness research. 123. 5.3.2.3 Data Collection Olswang and Bain (1994) discuss the need to keep several kinds of data in order to measure the total value of an intervention program. Further, they suggest that data is only as useful as the information it provides as part of the ongoing clinical decision-making process. Bernhardt (1993b) discussed signs of progress in phonological intervention in addition to the generalization of newly learned targets to conversational speech. These included increases in: (1) the child's enjoyment of direct speech games, (2) the child's enjoyment of talking in general, (3) ease and speed of imitation, (4) the non-imitative use of words in direct speech activities. Further, Bernhardt (1993b) indicated the need to observe small changes in the child's phonological development such as: (1) occasional use of a target word in conversation, (2) changes to speech sounds or syllable structures that are not the focus of intervention, and (3) improved performance on minor probe measures, indicating immediate memory for the production of the target. This data could be collected in a journal kept by the caregiver. Journal entries recording the above data could be made daily, or alternatively, the journal could be housed at the clinic and parents could spend the first five to ten minutes of every session recording recently observed data. 5.3.2.4 Optimization: Beyond Phonology Olswang and Bain's (1994) data collection focused mainly on speech and language data but they also included the child's willingness to attend treatment sessions as a qualitative measure of progress. In the present study, no formal attempt was made to keep data on non-phonological goals. However, parents frequently reported benefits derived from participating in intervention with their child. For the child, this included: (1) improved ability to be understood by members of the extended family and neighbours (e.g., grandparents), (2) increased confidence in speaking situations (e.g., sharing time at preschool), (3) enjoyment of individual time with a parent for practising weekly targets, (4) improved social relationships with family and peers due to reduced 124 frustration related to communication breakdowns. For the parents, non-phonological benefits included: (1) their child's improved behaviour and communication skills, (2) improved interactions between them and their child, (3) the empowerment to deal effectively with communication breakdowns, and (4) increased confidence in participating in their child's intervention program as a result of education sessions and participation in intervention sessions. 5.3.3 Summary of Optimization Questions Intervention based on nonlinear phonological principles was effective for Shaun and Eddie. It appeared that twice weekly intervention was at least as effective as three times weekly intervention. Issues of optimization that need to be considered at the outset of any effectiveness design include a faithful but flexible research design and thoughtfully designed data collection methods that inform the clinician about progress and provide direction within the intervention process. Further, the family system must be factored into data collection so that gains related to but not specific to phonological intervention can be collected and used to plan and provide the most effective intervention. I will now turn to the theoretical questions posed, namely, how does progress inform us about phonological theory and what is the nature of the interaction between phonology and phonetics. 5.4 Informing Phonological Theory The author's main assumptions about phonological theory need to be reviewed. These include: 1. An emphasis on hierarchical representations rather than process and rule-based operations so that figures depicting underlying and surface representations will not demonstrate an intermediate stage of repair rules that delink or relink nodes 125 2. The assertion that in phonological representations, the prosodic and segmental tiers are autonomous 3. That constraints dictate the association between tiers that occur and that these are related to restrictions on output (i.e. Grounded Phonology) 4. That features are specified in a single lexicon and that underspecification is learned in development from a combination of adult input and the child's own output, allowing for child-specific defaults that differ from universal or adult defaults The above topics will not necessarily be dealt with in order but provide a reference point for the reader as the discussion proceeds. 54.1 Underlying Specification We assume that each child comes to the task of learning the phonology of the ambient language with the ability to perceive contrasts and make predictions about the meaning attached to those contrasts in the absence of hearing impairment or cognitive delay. There is some disagreement about how children learn which features are relevant to their language, especially concerning whether the child is programmed with an innate, universal default feature setting that includes all of the unmarked values of every language in the world. If there were, children would change the settings of the marked versus unmarked value according to what they heard in the ambient language and store the feature representations in their phonological lexicon. Another point of view is that the representation of features is learned rather than innate and is a result of a combination of the adult input available to the child and some independent property of the child's own speech. Features are then represented in the child's lexicon according to the 126 frequency of the input, the perceptual abilities of the child (i.e. to distinguish /s/ vs. Ip; /it vs. /&/, etc.) and the conclusions drawn from what was perceived. Menn (1983) suggested that the child may store the wrong representations and not self-correct later because of the belief that it was stored properly the first time. An example of this was seen with Eddie who pronounced "yes sir, yes sir" from "Baa Baa Black Sheep" as [jaekoo jaekoo] but produced other word-medial (Coronal, [+continuant]) segments as either (Coronal, [-anterior]) or Coronal-Dorsal (i.e. palatal). When the correct pronunciation was pointed out to him, he correctly produced this token from then on, suggesting that it was an isolated mis-storing in the lexicon rather than an across-the-board problem with his underlying representation of (Coronal, [+continuant]). For both Shaun and Eddie in this study, it appeared that their own productions (and limitations on productions) affected the development of their underlying representations less that it did their ability to produce contrasts on the surface level. Shaun and Eddie demonstrated different levels of productive evidence of their phonological knowledge. Shaun's productive use of combinations of features was limited by output constraints (discussed in section 5.1.2 below). For Eddie, limited contrasts represented by default values and resulting from his unusual vocal tract shape affected his phonological output. Both children demonstrated evidence of segmental and complexity constraints in their phonologies. 5.4.1.1 Child Default Productions I will start with some examples of adult representations versus child output in the instance of Coronal vs. Dorsal contrasts. Stoel-Gammon and Stemberger (1993) predicted that the order of acquisition of segments progresses such that the simpler (i.e. less specified) segments should be mastered first and that underspecified segments often serve as substitutes for more specified 127 segments. For example, stopping and velar fronting would be explained by not specifying [^ continuant] or [Dorsal] and getting a default stop or alveolar. This was the case for Shaun but not for Eddie as can be seen in the examples found in Table 5-1. Table 5-1 Coronal versus Dorsal productions Shaun Eddie Adult UR 0 'coffee' /khan7 i c [ a ] I • V[+son, 'i+co \ Larval ^ \ Dorsal n>~oi P01* [+cont]l [f] [i] l [+son] T.V.Vthivi/ [th] [i] M [i] M+sonj ^ + c Q n t j ^| + s o n ] [+cons] [+s"g] I I [Jc] Dorsal [ibk] [ilo] Dorsal [iii] / P-cons] Coronal Dorsal Labial [iii] Dolsal [iii] 'coffee' [dati] [f] [a] M [i] ^l+J°n] icons] 'T.V.'/tivi/ [t] l i ] [b] [+vc] y J1 [+cons] I . Doi ¥ I f I v c W+son] ]r>cons]| Afr Coronal rsal -Ibk] [+lo] Coronal rj 0 ] >rsal L [+cons] [+vc] 0 Coronal Dorsal Labial Ddrsal [anl] [+hi] [Jhi] 1 'coffee' /tanV M [a] 1 ! t+cont^ V [+son] [i] T+cons] |t; s o n5 'T.V.'/tivi/ [tl [i] J v [+cons] I \ Laolal , \ I „N . Dors; Coronal Dorsal [Ibk] [-ilo] al [iii] [+son] [b] I c cons] [+vc] T l^ +son] [+cons] Coronal Dorsal Labial Do^al [+hi] [+hi] 128 Table 5-1 Coronal versus Dorsal productions Shaun Eddie T 3 'coffee'/kafi/ [+cons] \ \ Labial ^ \ , D ° r e a l Etoreal ? ° r e a l [J-bk] [+lo] 1+hi] 'T.V.'/t hibi/ [th] [i] [b] [i] I A- J y + s o n ] yKcons] yj^ cons] 1+Sgl [+vc] Coronal Do-sal Labial L\!rsal [Jhi] [+hi] In the case of Shaun's production of 'coffee' as [dati], Dorsal place was not available word-initially; default Coronal place surfaced as a [d] because of an additional restriction on word-initial voiceless segments. The /f/ in C2position was pronounced as a [t], as a result of the Coronal feature spreading (Figure 5-1). 'coffee' [dati] If [a] W [ j ] [+cons] [+vc] real cf [-Ibk] [ilo] Coronal-T orsal D i t+U Figure 5-1 Coronal feature spreading 129 As Shaun's WI restrictions requiring [+voice] in onset position decreased (see T 2), he was able to produce a WI voiceless default (i.e. [tafi]) and [+continuant] emerged in SI position. Finally, [kafi] was produced but without aspiration on WI /Id. The progression from (1) the WI default production [d], to (2) [-voice] Coronal [t], to (3) an unaspirated /Id demonstrates both increasing specification and decreasing onset syllable restrictions on Shaun's phonological output. Eddie's progression from default to matched value is similar except that he produced Coronals with a default that ranged from [-anterior] (e.g. as in [tibi] above) to Dorsal (e.g. 'doll' -> [gAo]), as opposed to Shaun's use of Coronals for Dorsals. Dorsal defaults are less common, but do occur, both in typical (Menn, 1983) and protracted development. Colin (Bernardt, MacNeill & Boleen, 1994) is another example of a child who has a major Dorsal Place stop default. Eddie's Coronal-Dorsal default setting applied to stops, fricatives and affricates. Child-specific defaults can be described within the theory of "Combinatorial Specification" (Archangeli &Pulleyblank, 1994). Given a child's immature oral-motor system and non-mastery of the phonetic content of his speech, deviations from universal default or markedness values would be allowed under the grounding principle. Bernhardt and Stoel-Gammon (in press) also describe language specific predictions and acquisitional factors such as artieulatory immaturity, reduced tongue tip control and a changing vocal mechanism in developmental phonology. Eddie's productions demonstrate a more stable underlying representation in that the output restrictions appeared to apply at the level of phonetic implementation such that he was able to use alternative Place of articulation contrastively to reflect meaning differences. As therapy progressed, increased control of his articulators resulted in more accurate production of the distinctions within the anatomical range between Coronal and Dorsal. 130 5.4.1.2 Universal Specification Predictions Before leaving the discussion of underlying representation and default productions I would like to examine the progression of Shaun's and Eddie's phonological development as seen in their surface production feature geometries from T r T 3 and compare this to the universal default1 predictions made by Bernhardt and Stoel-Gammon (in press). They reported that the following universal defaults are expected in early child phonology: 1. At Root, [-continuant] and [-nasal] are defaults. Early specification starts with [+nasal] then includes [+ continuant] before [+lateral] is specified. 2. At Laryngeal, [-voice] and [-spread glottis] are default features, with gradual specification of [+voice] in onset position prior to [+voice] in Coda position. 3. At Place, Coronal [+anterior] is the default specification, with development of Labial prior to that of Dorsal. Shaun's data comply with Bernhardt and Stoel-Gammon's predictions. Even though [+lateral] was established before the specified value [+continuant], [+continuant] emerged first. Possible explanations for the difference in establishment versus emergence of these features include: 1. Coronal place was the default setting and [+lateral] taught in onset position was supported both by a strong place (i.e. Coronal) and word position; 2. [+continuant] was subject to a feature cooccurrence constraint so development of this feature was restricted. 1. Universal defaults refer to the least marked features cross-linguistically. 2. [+continuant] co-occurs with [+lateral] as a default but with Labial as a non-default. 131 Eddie's development of specified features corresponds with the Bernhardt and Stoel-Gammon (in press) universal predictions at Root and Laryngeal but his Place specification began with Labial and Dorsal. He had not quite established Coronal place by the end of the intervention period. In summary, Shaun and Eddie demonstrated developmentally underspecified phonological representations. Default features surfaced in their output because of constraints on their phonological systems. Shaun produced the universally predicted Coronal place default while Eddie produced a child-specific Dorsal default place. Feature specification followed a generally predictable course with the exception that Dorsal developed prior to Coronal for Eddie. I will now explore the interactions and associations between tiers as they relate to segmental and sequence constraints. 5.4.2 Segmental and Sequence Constraints The discussion about underlying representations and feature specification above was focused on the principles of autonomous, hierarchically organized features. Bernhardt (1995) concluded that the interaction of feature cooccurrence and feature sequence constraints was an intriguing area to pursue in analyses of children's phonological developmental patterns. The next sections deal with the associations between elements and tiers and the constraints that applied for Shaun and Eddie. 5.4.2.1 Segmental Constraints and Feature Cooccurrence Following, a reanalysis of the data, it became apparent that the prevalence of the default feature [+sonorant] in Shaun's data was due to a feature cooccurrence constraint which appeared to allow only binary sets of features. Hence, the most instances of default manner [+sonorant] 132 occurred when there was an additional place or manner specified. For example, Shaun was able to combine the following features: 1. [+sonorant] and Place leading to ([+nasal], Labial) or ([+nasal], Coronal) and accounting for the segments /m,n/ 2. [-consonantal] and Place leading to ([-consonantal], Labial) or ([-consonantal], Coronal), accounting for the segments /w/ and 1)1 which only have V-place 3. ([+sonorant], [+continuant]:), accounting for /h/ which has no C-place. The combination of [+consonantal] and Place (Labial or unspecified) accounted for /p,b,t,d/. If [-(-continuant] was preserved, place was usually delinked leaving Laryngeal [+spread glottis] to fill the C-slot (e.g. sun -> [hAn]). In the case of (Labial, [+continuant]), Place specification was completely delinked leaving [-•-spread glottis] to accommodate the [+continuant] specification (e.g. 'four'-> [hoo]). Glide substitutions were used only for (Labial, [+continuant]) combinations (e.g. 'five' -> jjai], 'feather' -> [wEho]) and will be discussed further in Section 5.4.3. If one were using a rule-based delinking process, the difference between the radically underspecified adult form for 'sun' and Shaun's surface form [hAn] would be represented as in Figure 5-2 where Coronal place is delinked, leaving no place and the default production [hAn]. Alternatively, the production of [h] in onset could be described as the failure of [Coronal] to surface in the output. 133 ' sun' / sAn/ [hAn] W M W i i M [A] [n] • i 1 1 1 t+contl^C V^ +son] C [+cont] fj _> i y jf+son] ^ Vcons] [+nasal] / [+nasal] [+S9J Radically underspecified adult form Shaun's surface production Figure 5-2 Adult form versus Shaun's surface production Similarly, for Shaun's other production of 'sun' [+continuant] would be delinked from Root, resulting in the default production [dAn] shown in Figure 5-3. ' sun' / sAn/ [dAn] [?] [A] M 1 [?] [A] [n] | | | 1 1 1 [+cont]. 9 yt+son] C [+cons] [+nasal] [^^C [ + c o n s ] V[+son] C y [+nasal] [+vc] O Coronal Radically underspecified adult form Shaun's surface production Figure 5-3 [+continuant] delinked from Root In summary, Shaun was able to produce [+continuant] with no specified place (i.e. Ihl) but a feature cooccurrence restriction applied when place was specified in the output (e.g., /f/). When 134 Shaun preserved place of articulation in the output (e.g., [dAn]), [+continuant] could not cooccur with the [+consonantal] feature at Root. Further evidence of Shaun's cooccurrence restrictions was demonstrated in the contrast between his optional glide or stop substitutions for (Coronal, [+continuant]) shown above and his glide only substitutions when the feature was specified underlyingly for place, in his case [Labial]. For example, 1. 'fish' Itif/ -> [wath] (see Figure 5-4) 2. 'five' /farv/ —> [jai] (see Figure 5-6) 3. 'four' /for/ --> [hoo] Given If I which is specified for (Labial, [-(-continuant], [+consonantal]), Shaun was able to produce an output that was specified as: 1. (Labial, [+continuant], [-consonantal]), i.e. [w] 2. (Coronal, [+continuant], [-consonantal]), i.e. [j] 3. ([+continuant], [-consonantal], [-(-spread glottis]), i.e. [h] Therefore, in order to specify [-fcontinuant] with a [+consonantal] (which Ihl becomes in onsets), Shaun was unable to specify place. There were no cases where a [Labial, -fcontinuant, -(-consonantal] segment had a [Labial, -fconsonantal] substitution (i.e. If\nl —> \b\n]) due to feature cooccurrence restrictions described above. 135 'fish' /fijV [f] M I I [+contl 9 V[+son] fx ^ P/+cons] ) m i C [+cont] I [^ cons] Labial Corona] [-anlerior] Radically underspecified adult form 'fish' /wath/ [w] [a] m C[-cons] y [ + s o n ] C Labial \ Dorsal I I [+bk] [+lo] Shaun's output production rt '[+cons] Coronal Figure 5-4 Feature cooccurrence constraints In terms of repair rules resulting from the constraints on Shaun's output for (Labial, [+continuant], [+consonantal], the representation in Figure 5-5 would describe the process: m [w] [w] i t+contl^^ ^cons] Labial c [+contl y [+cons] Labial CT-cons] / Labial 1 C [-cons] Labial (1) Fully specified adult form (2) Delink [+cont] and [+cons] at Root (deletion) (3) Default [-cons] at Root (addition) (4) Shaun's output production. Figure 5-5 Delink-relink repair 136 Word-final /f/ surfaces as [th] to preserve [+consonantal] and can do so (rather than be deleted) because of the [Coronal] specification for [t11]1. The failure of [+continuant] to surface in this example is the same as the /sAn/ -> [dAn] example. If the word-final consonant was specified for Labial or Dorsal, it would be predicted that C -> obeeause Shaun's default glides [h,w,j] would not appear in codas in the same way as stops, nasals or fricatives. That is exactly what happens in the case of 'five' (pronounced as [ J A 1 0 ] ; alternatively transcribed as [JAI1]). In this case, the default is operating at the timing tier rather than at the feature tier. Eddie's data demonstrated a "NotCooccurring" ([^ -continuant], Coronal]) example for the word 'soap' (e.g.,/scop/-> [hoop] Ti, [woop]T2, [Jbop]T3). AtTi , Place was delinked from Root and the [-fcontinuant] feature was preserved. AtT 2 , Labial Place was assimilated from WF/p/,and finally, at T 3 , Labial place and [-fcontinuant] was linked to be produced as [fj. 1. Aspiration on [th] may be an attempt to mark the [-fcontinuant] feature. [+bk] Dorsal [+hi] Figure 5-6 Timing tier constraint 137 5.4.2.2 Complexity Constraints Bernhardt and Stemberger (in prep) propose that when children retain one element of an /st/-cluster, then either "NotComplex" (Onset) or "NotTwice" (C-Place) is operating as the more relevant constraint, depending on what happens to other clusters in the child's phonological system. If other clusters are also reduced, the complexity constraint is the most likely constraint otherwise a particular segment may be prohibited. Structural complexity and related segmental complexity, such as that found in consonant clusters will be discussed in this section with examples of Shaun and Eddie's data to illustrate examples of repairs. Bernhardt and Stemberger (in prep) stress the importance of considering the child's entire phonological system including onset-coda asymmetries and variability of repair strategies. Further, they differentiate constraints from repairs stating that specified elements are more likely to be implicated in sequence constraints and unspecified elements in repairs, citing Stemberger and Stoel-Gammon's (1991) example of assimilation. Smit's (1993) observation that substitution errors in preserved clusters were similar to those produced for consonant singletons and that the errors were unmarked relative to the target (less specified than the target) provide a baseline for discussion. If the elements of the cluster are patterning differently than they would as singletons, then the elements could be analyzed in terms of complexity constraints. As mentioned above, repair strategies are described in terms of "deletion" or "subtraction" repairs, "addition" repairs, and "delink-relink" repairs. These repairs will now be discussed in terms of how they appeared in Shaun and Eddie's cluster data from T1-T3. 5.4.2.3 Deletion Repairs Deletion repairs occur when one or more of the elements violating the sequence constraint fails to surface as a result of delinking one of the levels of representation. Shaun and Eddie used 138 deletion repairs almost exclusively at T l and T2 suggesting that there was a NotComplex (Onset) sequence constraint. Shaun and Eddie's repair patterns at both times were similar and are summarized as follows: 1. C1 (Obstruent), C2 (Approximant) -> Delete C1 (e.g. /pi/ -> [w]); 2. C l (Is/), C2 (Obstruent) -> Delete C2 (e.g. /sp/ -> [b]); 3. IB*/ ->• Fuse to single segment (e.g. IQJI -> [b], [v]) for Shaun; Eddie's production was a "match" for wordshape but not for features (i.e. IBs/ -> [fw]). These results correlated rather closely to those found by Smit (1993). The examples in Table 5-2 show the deletion of the C-slot at the level of the timing tier. Table 5-2 Deletion of C l and C2 in clusters Examples Delete C l Delete C2 Shaun 'sweater' /swct / -> [wedo] 'snake' /sneik/ -> [neit] 'music' /mjuzik/ -> [must] Eddie 'spoon' /spun/ -> [pun] 'black' /btek/ -> [b33ph] These types of constraints and repairs were documented by Olmstead (1971) who showed that for /l/-clusters, C l was usually retained and for /s/-clusters, C2 was usually retained. An interesting pair of clusters to analyze, given these general patterns, are /si/ and /sw/. Smit (1993) states that /si/ generally behaves more like the /l/-clusters and /sw/ like the Is/-clusters. The data in see Table 5-3 and see Table 5-4 demonstrate Shaun's and Eddie's different strategies for dealing with output restrictions on /si/ and /sw/ clusters. 139 Table 5-3 Shaun's /si/ and /sw/ cluster data Cluster T i T 2 T 3 /si/ [w], {[j]} [w] [sw], [sv], [vw], [svw] /sw/ [w] [w] [w], [s°w], [sw] Shaun's singleton III was produced with similar substitutions at Ti and T 2 . At T 3 , IM was established. Both /si/ and /sw/ patterned as /l/-clusters then later, labial variants were produced when a complex onset was licensed. The production of [v] in the /l/-clusters was unusual and was likely due to a feature cooccurrence constraint. Smit (1993) observed that it was relatively uncommon for her subjects to reduce /pi, bl, kl, gl, fl, si/ to an approximant, but Shaun preferred to have [+sonorant], a well-established default in his system. Table 5-4 Eddie's /si/ and /sw/ cluster data Cluster T l T2 T3 /si/ [h] [c] [s1] /sw/ ® [fw], [pw] [fw] The /si/ cluster patterned like an /s/-cluster, while the /sw/ could be interpreted as either. Eddie's singleton productions for IM were variable, including [j] and [w] word-initially and [j], [w], [d] and [b] in initial-word position. Smit (1993) observed that while typical errors comprised a small set, there was a wide range of variability for other errors seen. Neither Shaun nor Eddie used addition of [a] or spreading to resolve any of the constraints on clusters at Ti or T2, although Shaun used epenthetic vowels to learn clusters in Block I. At T 3 , Shaun demonstrated the use of a variety of repair rules, including 140 spreading and fusion (e.g., 'sweep' -> [fip] where HI represents the [-(-continuant] feature from7s/ and the Labial feature from Iwl) of the features, which are discussed in the following sections. 5.4.2.4 Addition Repairs Addition repairs usually imply resyllabification and result in the creation of an unstressed syllable. In terms of child phonological development, a child may insert an epenthetic vowel between the first and second elements of a consonant cluster to assist in the production of a structure that might otherwise have a sequence constraint. Smit (1993) reported that epenthesis was used by up to one-fifth of the some of the children between ages of 2; 6 and 8; 0 in her study for some clusters. The most frequent targets for schwa-insertion were the clusters /9r/, /gl/, /dr/, /br/ and the least frequent targets were /kr/, /sp/, /bl/, /st/ and /kl/). Shaun's use of an addition repair may have been phonetically grounded. He prolonged the Is/ in CI position of /s/-clusters (e.g. [swipirj], [s:noo], [small, [s:pun], [sto:]). Shaun used an epenthetic vowel in a context that would be expected (i.e. /swsd / -> [s0wsdo], further demonstrating the variability that was apparent in his repair strategies at T 3 . The most likely conclusion is that since Shaun was taught these clusters using a mora "edges" approach, some of his earlier strategy of lengthening the extrametrical Is/ prior to adding it to the onset of the syllable had carried over. Eddie did not use addition repairs to resolve sequence constraints in obstruent plus approximant clusters but Shaun began to use them by T 3 , at which time his use of deletion repairs was considerably reduced. Additions included epenthetic [a] and [o] as in the examples below: 1. 'black' /btek/ --> [batek] 2. 'brush' /OJAJV -> [b°WAf] 141 3. 'clouds' /klaodz/ --> [k°laodz] Smit (1993) suggested that the fact that children use schwa insertion more frequently for clusters with approximants than for /s/-clusters indicates that the schwa may be a lengthened transition into an Irl or III rather than a true vowel-insertion. Schwa insertion may also be compatible with motoric difficulty. Although Eddie did not use the lengthening or epenthesis strategies for clusters, his nasal air emission for Isl + nasal clusters may have been marking the [+continuant] feature of Isl. This could be explained from a motoric point of view, in that the coordination needed to raise the velum for [s] and then drop it for [n] are too complex, such that the fricative was produced nasally, followed by the true nasal consonant. 5.4.2.5 Delinking-Relinking Repairs Delinking-relinking repairs involve delinking the disallowed segment and then replacing it with a (a) default value, (b) "double linking" strategy (i.e. assimilation, harmony or spreading) or (c) using a feature movement repair that OT refers to as "flop" (i.e. metathesis). In these cases, we know that the repair is not motivated at the timing tier because all or part (i.e. features) of both segments surface. Therefore, the repair must apply to the feature tier. 5.4.2.6 Default Value Repairs When one of Root, Laryngeal or Place is delinked, the easiest option available to the system is to insert a default value (or allow the default to surface) rather than inserting another specified feature. Default values were common in Shaun and Eddie's data and fulfilled the requirement for the timing tier to be linked to features through Root but could not give the output needed due to constraints on their systems. Some examples of their use of defaults are found in Table 5-5. 142 Table 5-5 Default value repairs Examples Delink-allow default to surface Shaun 'tiger' /taig / -> [daidoo] (delink Dorsal Place) 'candy' /kaendi/ -> [daendi] (delink Dorsal Place) 'clock' /klak/ -> [jot11] ( delink C-slot on timing tier) Eddie 'dress' /dies/ -> [gak] (delink Coronal Place) 'candy' /kaendi/ -> [khAgi] (delink Coronal Place) 'dumptruck' /dAmptiAk/ -> [gAkAk] (delink Coronal Place) When the C-slotis deleted at the timing tier for clusters, the remaining C can be produced using default values. For example, Shaun's constraint against "NotComplex "(Onset) in 'sled' (/skd/ -> fjed]) was resolved by deleting C-Place and its Root which resulted in the [-consonantal] default [j]. Shaun produced [h] frequently for singleton onsets, especially when cooccurrence constraints disallowed specification of Place when [-(-continuant] surfaced in the output but he used [h] infrequently for clusters. Shaun's use of glide defaults (i.e. [w], [j] and [h]) was prominent at Ti and T 2 . By T 3 , he used [h] predominantly and this use was linked to word-position. In word-initial position he used [h] for [-voice] segments (e.g. If, Q,sl) and in. SI position he used [h] for [+voice] segments (e.g. Iv, 61). Bernhardt and Stemberger (in prep) suggest that acoustic similarity may be a factor in the use of [h] for unvoiced segments but find that [h] for voiced fricatives is less explicable. In the case of the following examples, Shaun's use of [h] for voiced fricatives appears to be related to the feature cooccurrence constraint of Place with [+continuant] once again but it is interacting between syllables. Eddie used [h] as a default value in multisyllabic words (e.g., 'together' -> [thug£ho]). Likewise, his production of 'feather' between T l and T3 appears to move from default to Coronal Place WI+glide to Labial Place [-(-continuant] WI with [+continuant] doubly linked to [h] but Place 143 delinked because of limits on feature specification (e.g., 'feather' ->T1: [weho:]; T2: [tcjo]; [lshoo]; T3: [fsho]). The word 'screwdriver' which was inconsistently produced as [srkwusdwAiho], [kuwAiho], [sivuyAiho] and [kwufwAiho] is the only word where [h] was used for Nl and it seems that Eddie 'runs out' of specification by the time he reaches the fourth syllable. 5.4.2.7 Double Linking Repair (Assimilation) Double linking repairs can occur in response to segmental/sequence or complexity constraints. Sometimes, when segmental constraints on singletons have been resolved they continue to operate on clusters (e.g., Not ([+continuant], Place) for Shaun described above). The directionality of double-linking repairs (i.e. assimilation or spreading) is Right to Left in most children (Vihman, 1978). However, Bernhardt and Stemberger (in prep) suggest that directionality is linked to the type of constraint acting on the consonant sequences. Table 5-6provides examples of double linking repairs observed in Shaun's data: 144 Table 5-6 Double linking repairs: Shaun 'gloss' /target/ T i T 2 T 3 Repair 'plum' /p i A m / [WAITI] [plAm] [pUm] Labial from [p] with [+son] from [1] at Root (fusion) 'brush' /bTAf/ fWAf](E) [WAj] [b°WAf] Labial Survived then L->R (d-1) spread to Coronal [-ant] 'truck' /LJAk/ [WAp h] [WAp] [tfOWAp], [VAp], [tfA P ] C l delinked. C2 Labial Survived then L->R spread/d-1 to Dorsal. Still present at T3, Dorsal not established. 'green' /grin/ [wim] [wim] data gap C l delinked. C2 Labial survived then L->R spread (d-1) to Coro-nal [n] 'fish' /hp [wat>], [fif] (E) [f:wc] (E) [ W A S ) ] , [WAp], [we 1;] [fis] Labial Survived (delink [+cont] insert Coronal default 'three' /Oil [vi], [bi] [wi] [vi] [v]: [+cont] and Labial Survived [b]: Labial Surv. with default [-cont] manner. In the case of "brush", "truck" and "green" and for Eddie's production of "black" as rbaeph], spreading occurs in a Left to Right direction as a result of a constraint against Root adjacent consonant sequences1. Figure 5-7 illustrates repairs for the word "green". 1. Root adjacent sequences are those which share similar features at Root, example, Ibl and IdJ are Root adja-cent; they share the feature [+consonantal]. Root adjacent elements can affect each other when they appear in surface contiguous sequences (e.g., clusters). 145 'green c [il I v [vc] [+cons] i\ [+cons] Dorsal LabiaKCoronal Dol [+hi [+son [n] I c +nas] [+cons] taal Coronal Delink repairs [CJ] [-j] [i] [ n ] c v c [+son] , J+nas] [vc] I [+son] , J \ »A f [+cons]^ +V f^cons] t + c t / i l l Dorsal Labialtoronal Doteal Coronal [+cons] [+hi] Double-linking of Labial (L->R) [w] PI En] c v r C ' ' • [+son] I J+nas] [+v, t-son] [+cons] Dorsal 1 [+hi] /> ' Lalbial Figure 5-7 Double linking repair Shaun The examples of "brush", "truck", "green" and "black" support Stemberger and Stoel-Gammon's (1991) conclusion that when the Labial feature is present it dominates other Place specifications. Further, since default (i.e. Coronal for Shaun, Dorsal for Eddie) or unestablished places (i.e. Dorsal for Shaun) were the affected segments, Stemberger and Stoel-Gammon's statement that the constraints usually apply to specified segments and repairs affect unspecified segments was further supported. Smit (1993) discussed labial variants substituting for clusters that did and did not contain labials, excluding /sn, st, ski because they were unlikely to have any labial substitution/default. She states that "most often these labial variants are fricatives [f, <J>, v, fw, vw], but sometimes they are stops or glides [p, b, w, pw]"(Smit, 1993, p. 945). When Labial Place is not a possibility (e.g., in /sn, st, sk/), then some other feature dominates, often the manner feature at Root. Also, according to Stemberger and Stoel-Gammon (1991), when both labial and velar place can be specified for a cluster then Dorsal Place dominates. The use of labial variants for these clusters was less common than for other clusters in Smit's (1993) study. This pattern applied to Eddie (e.g., 'green' -> [gin]), but not for Shaun (e.g., 'green' -> [wim]), for whom Dorsal was an unestablished Place feature. Stoel-Gammon and Stemberger (1993) predicted [Labial] and [Dorsal] features would spread to Coronal Cs (e.g. [fm] for /sm/) in clusters. They did not predict reduction to default l\J from /sp/ (i.e. alveolar stop). Rather, they predicted the fusion of the feature [+continuant] from Is/ with [Labial] from /p/ to produce [f] (Labial, [+continuant]). Eddie produced 'sweater' as [fwcjo] and Shaun said "slipping" as [vwiprji]. In these cases, a complex onset was allowed but not with Is/. Instead, the substitutions of [f] in "sweater" and [v] 147 in "slipping" preserve the [+continuant] from [s] and doubly links Labial Place (from HI and /v/) with the Labial feature of [w]. In general, assimilation (i.e. double linking) in non-contiguous sequences is usually seen in early development but can persist in children with phonological disorders even after six years of age (Bernhardt & Stemberger, in prep). Shaun and Eddie's data from Tj to T 3 showed a development in repair types: (1) deletion and double linking of non-contiguous sequences to (2) addition and double-linking of contiguous sequences of segments. Further, their data followed predictions made by Stemberger and Stoel-Gammon (1991) concerning Labial dominance and repairs applying to default segments. In the following section, one more type of repair will be discussed, namely "flop", which has been referred to previously as "metathesis" in process terminology. 5.4.2.8 Feature Movement Repair (OT: "flop") Feature movement repairs can occur in surface non-contiguous sequences (i.e. long-distance flop) or surface contiguous sequences (i.e. clusters). Smith (1973) talked about "metathesis" in his son Amahl's speech for Cw clusters ending in a consonant and gave examples such as (1) 'quick' /kwik/ -> [kip] and (2) 'queen' /kwin/ -> [ki:m]. Eddie's production of 'dress' /diss/ as [gApi] is an example of Labial flop occurring in at least one direction in a word. The Labial feature from lit moved to the right edge of the wOrd (i.e. [p]). It is also possible that the Dorsal default feature (i.e. in Igl) that would have replaced the Coronal feature (i.e. in Is/) moved to the left. On the other hand, the production of [g] may have been a simple default value for the WI Coronal stop (i.e. Idl). The production of a WF stop for a [-(-continuant] may be related to Eddie's relatively infrequent use of the combination of (Labial, [-(-continuant]) in WF position. 148 Flop in clusters also occurred in Shaun and Eddie's data, during the intervention sessions. Both subjects produced the cluster [ps] in an attempt to produce an /sp/ cluster. Eddie was imitating several /s/-cluster words, using a prolonged [s:] as in [s:wip], [s:mao], [s:naep] and [s:prt]. When he attempted to say 'spoon' non-echoically, he produced [ps:pun]. Likewise, Shaun's spontaneous production of 'spot' in a therapy activity was [ps:pot]. Such flop has been described previously in Chin (1994) and Bernhardt (1995). 5.4.2.9 Summary of Constraints and Repairs Much of the data that was previously inexplicable under generative phonology based on rules and processes can be explained in terms of segmental and complexity constraints. Phonological development consists of the child's attempts to overcome output constraints through the use of repair strategies. Strategies discussed included addition, deletion, assimilation and flop. Some of the data revealed interactions between consonant and vowel features which will now be discussed. 5.4.3 Consonant-Vowel Interactions: Shaun Consonant-vowel interactions were apparent in two different areas of Shaun's phonological development: (1) the interaction between V-Place and default C-Place at T i 1 , and (2) the development of Dorsal Place during Block I of the intervention program. Data related to the interaction between V-Place and default C-Place are found inTable 5-7. 1. Although [h], [w] and [j] are underspecified as [-consonantal], they surface as [-(-consonantal] in onset posi-tion, making a discussion of consonant-vowel interactions possible. 149 Table 5-7 Consonant-vowel interactions 'gloss' /target/ T i 'five' /faiv/ [jail 'four' / fDj/ [hoo] 'feather' Ifcb / [wcho:] 'van' /vsen/ [hasn] 'sun' /SAn/ [dAn], [fiAn] 'side(s)' /said(z)/ Cjai] 'zipper' /zip / [hebo:] 'shoe' Iful [fu](DE) 'shower' If aim I [haowo] 'thing' /eirj/ [bin] Shaun's use of the WI default segments appeared to be related to vowel context. His most frequently produced default segment, [h], appeared in most vowel contexts. However, [j] appeared in the context of the diphthong [ai] indicating double linking of the consonant and vowel. If this context facilitated assimilation, it provides an argument for these vowels being specified as having Coronal Place (Clements, 1985). The use of default [w] for If/ (also in 'fish' -> [wati] and 'finger' -> [wendo]) was inconsistent and indicated that the cooccurrence of Labial and [+continuant] was developing. Thus, the underspecified Coronal glide benefitted from feature cooccurrence while the specified Labial sometimes had a restriction against cooccurring features. Shaun's development of Dorsal Place was vowel-dependent. Vowel contexts of varying facilitative value are summarized in Table 5-8. During the initial therapy session for word-final Dorsal, Shaun was able to produce /k/ in the contexts of the vowel Id (Dorsal, [+back], [+low]) and the diphthong /co/ (Dorsal [+back][-high]; Dorsal [+back] [+high]) on the first prompted echoic response, suggesting that the [+back] feature of the vowels might be the common factor 150 facilitating the production of Jkl. Vowels that did not facilitate the production of Ikl, and in which Shaun actively refused to produce IkJ, included HI ([Dorsal],[+high]) and /u/ ([Dorsal], [+back], [+high]) which triggered Shaun's Coronal place default (i.e. he produced [it] for/ik/). Following repeated attempts, Shaun was able to produce Ikl in the context of /ae/ (Dorsal, [-high]) and the diphthongs laol (Dorsal, [+back], [-high]); (Dorsal, [+back], [+high]) and / A I / (Dorsal, [+back]); (Dorsal, [+high]), all of which keep the [+high] feature at a distance of at least one vowel away from the Ikl phoneme. Table 5-8 Facilitative vowel contexts for Dorsal Place Vowels/ Diphthongs Features Facilitative Value Id /oo/ (Dorsal, [+back], [-high]) (Dorsal, [+back], [-high]);(Dorsal, [+back], [+high]) Positive /ae/ /au/ /Al/ (Dorsal, [+low]) (Dorsal, [+back], [-high]); (Dorsal, [+back], [+high]) (Dorsal, [+back], [-high]); (Dorsal, [+high]) Secondary HI l\xl ([Dorsal],[+high]) ([Dorsal], [+back], [+high]) Negative Another proposal is that the facilitating effect of Id and /oo/ (and later, /ae/) is related to the backness of the vowels in physical relation to the pharyngeal wall. In fact, Shaun may have been producing a phoneme that more closely approximated a pharyngeal stop initially rather than a true velar. MRI slides of an axial section of the vocal tract during Id (and by analogy /ae/) production showed almost complete constriction (Cooper et al.,1994). Shriberg and Kent (1982) stated that /ae/ is so far back that the tongue root, which is just posterior to the dorsum, comes close to the pharyngeal wall. Further, Ohde and Sharf (1992) discuss how the tongue closely approximates the palate for HI and Id but is far away for Id and /ae/. In approximating the palate, these vowels may have triggered Shaun's Coronal default production for Dorsals as in his production of [utfutfudu] 151 for [ukukuku]. My hypothesis is that /u/ was not facilitative for the production of IkJ even though it is a [+back] vowel because of its height. Further, IvJ grouped with I'll in this case which suggested that they would both have a similar specification (i.e. such as Dorsal) rather than grouping central and back vowels separately from front vowels as proposed by Lahiri and Evers (1991). Finally, from a grounded phonology (Archangeli & Pulleyblank, 1994) perspective, the features [+ATR] and [-ATR] which relate to the physical position of the tongue root support the grouping of [+high] vowels with [+ATR] and [+low, +back] vowels with [-ATR]. These data would support the assimilatory effect of [+low, +back] vowels and [-ATR] and the physiological grouping by height rather than by tongue position on a front to back scale. Eddie's development of Coronal Place also appeared to be vowel dependent. The (Coronal, [-continuant]) phonemes were produced more accurately in the context of the vowels III and Ixxl (e.g. 'tea' -> [di] and 'two' -> [to, thu]). These data support the grounding path condition that [+hi] implies [+ATR]. Eddie was using a [Coronal] place for Izl and /j7 (lateralized) and a palatal place for Isl (lateralized) at T x . At T 2 , Izl and /J"/ are both produced with palatal specification and lateralized. A Coronal off-glide provided a phonetic link between the [g] and [u] in the case of 'zoo' (i.e. [g^ u]). The Is/ phoneme in the word 'suitcase' (i.e. [cWeic;]) was also lateralized and was produced as a palatal. By T 3 , Eddie's feature specification and phonetic implementation had balanced so that the Coronal-Dorsal default was not as pervasive. Further, he had acquired the phonological distinction between (Coronal, [-anterior]) and (Coronal, [+anterior]). Interactions between consonants and vowels are difficult to discuss given that phonologists can not agree on the feature representations of either. Further, a variety of interactions may be taking place including a combination of height, place, tenseness and roundness of vowel features. 152 Many consonant-vowel interactions appear to be phonetically motivated which leads to a discussion concerning the interactions between phonology and phonetics. 5.4.4 Interactions Between Phonology and Phonetics As mentioned previously, Laver (1994) described the tendency toward phonetic similarity in the segmental realizations that make up a given stretch of speech as a 'phonetic setting' (see also Bernhardt & Stoel-Gammon, 1994: 'child default features'; Edwards and Shriberg, 1983 and Weiner (1981): 'systematic sound preference'). Laver explains that settings most strongly apply to vocal characteristics in speakers with intact vocal tract physiology. However, for the example of the child with craniofacial anomalies, an artieulatory setting bias might be related to the habitually restricted movements of the articulators in relation to a high palate, restricted tongue mobility or velar-pharyngeal incompetence. The impact of the setting depends On what Laver terms 'segmental susceptibility' which relates to the number and frequency of target articulations affected and whether or not these are key segments in the child's developing phonology. Given that Eddie's default feature place was Coronal-Dorsal and that this default surfaced for Coronals that were unspecified for place in the majority of the cases at the beginning of intervention, it would be predicted that his conversation would be less intelligible than it actually was. Without actual spectographic data it is difficult to prove that Eddie was making a distinction between more front versus more back dorsals; however, several listeners agreed that his intelligibility was not as poor as they would expect given his pervasive use of dorsal default segments which may indicate that he was making a phonological and phonetic distinction between dorsals and alveopalatals. Eddie's Coronal-Dorsal Place default place segments were: 153 1. [-continuant]: [t, cL Ji, c]; 2. [+continuant]: [c]; 3. [j]1 Eddie's Dorsal default segments were: 1. [-continuant]: [k,g] 2. [+continuant]: fo] Eddie used these segments contrastively. For example: 1. [p] was used when the target was a Coronal and [rj] was used when the target was a Dorsal 2. [9] was used when the target was a Coronal and [%\ was used when the target was specified for Dorsal Place. Although Eddie's phonetic setting was generally a [+back] setting, he demonstrated different degrees ofbackness depending on the target phoneme. In general, it seems clear that phonology and phonetics are inter-related and this relationship has an impact on the assessment and treatment of phonological disorders. Stengelhofen (1989) suggests that children with high, narrowly arched hard palates in the absence of cleft palate are at risk for similar articulatory, phonatory and resonatory disturbances as cleft palate children due to the reduced surface area of the palate and the tongue's restricted ability to either reach or make proper contact with the palate. Additionally, if the alveolar ridge is underdeveloped or elevation of the tongue to the ridge is prohibited or imprecise, "the articulation 1. Some argue that /j/ is specified for Coronal-Dorsal Place 154 of a number of fricatives and plosives may be affected" (Stengelhofen, 1989, p. 15). Eddie's phonological development appeared to be affected by a high, narrowly arched palate, an underdeveloped alveolar ridge and limited tongue control. Hewlett (1985) argues that structural abnormalities and the compensatory articulation required to adapt to them may affect a child's speech-related neurological development and, ultimately, representation of phonological information. He suggests that early compensations for structural limitations could lead to patterns resistant to change, even following the alleviation of the original physical limitations. Eddie's difficulty maintaining therapy progress following a break may have been indicative of a well-entrenched pattern of compensation. However, Eddie did make progress on the Coronal Place targets and was able to shift his setting forward such that fewer segments containing the Coronal-Dorsal default place feature were present in his speech at the end of the intervention program. 5.4.5 Summary of Theoretical Questions Shaun's and Eddie's data were analyzed using nonlinear phonological theory to describe feature specification and child default productions. Segmental and sequence constraints were apparent in Shaun and Eddie's phonological development over the period of intervention. Constraints on output resulted in a variety of repair strategies including addition, deletion, delink-relink, assimilation and feature movement (i.e. 'flop'). 5.4.6 Limitations of the Present Study 5.4.6.1 Sample Size Sample size limits the examiner's ability to generalize results to the population of children with phonological disorders. However, the detailed observation of these two subjects compensates 155 for that limitation and adds to the growing data base of case studies using nonlinear phonological theory to guide assessment, goal choice and intervention. 5.4.6.2 Design The multiple baseline design proved to be useful in demonstrating the lack of gains on untreated goals, especially in Block I. However, it is more difficult to tease out the effects of intervention in Block II due to (1) the use of cycles in which goals were repeated, and (2) the targeting of clusters in which it was necessary to include singletons from previous treatment goals (e.g., Ill in /l/-clusters). It appeared that continued development of singleton targets was attributable to their inclusion in cluster goals and that progress might have leveled off or continued at a less rapid rate had these segments not been taught in the cluster goals. Unfortunately, excluding them would have reduced the number of words available for therapy activities substantially and would have served no other purpose than testing the efficiency rather than the effectiveness of intervention. Target selection was based on segmental and prosodic goals and the decision to mix some of the goals may have led to instability of results. For example, Shaun mastered III at the end of four sessions and was expected to produce III in clusters four sessions later. The ordering of goals may affect the stability of the targets such that a recently mastered or developing segment may not be optimal for inclusion in a new wordshape goal. If it is not optimal to include an unstable segment in cluster intervention but unavoidable (as was the case with Shaun), the clinician might expect generalization of the cluster goal to be less rapid than mastery of the singleton. Another potential problem with the design is that the examiner must be aware that a breakthrough moment has occurred and have the time and flexibility to steer from the original course of intervention in order to build immediately on an emerging skill. One way to be 156 intentional about probing for emerging features is to include future target segments or wordshapes in each minor probe word list of untrained targets. Further, the development of a record-keeping system in which breakthrough moments are noted while an intervention session is in progress may alleviate the necessity of tape recording every session in order to detect these moments. While having a single experimenter involved in assessment and intervention provided continuity across the study, it may also have increased the probability of experimenter bias. Areas of bias included the experimenter's enthusiasm and motivation to procure favourable results and the fact that the experimenter, as a graduate student working only two days in the clinic, had more flexibility than a clinician holding the responsibilities of a full caseload. However, aside from time spent on transcription, the experimenter made every attempt to limit time spent on assessment, goal selection and intervention planning. Time that was spent on family contact by telephone, consulting with outside agencies and home visits may have been considered as extraordinary, but with any severe communication disorder, these contacts are essential for optimal service. Subjects will respond to treatment designs individually and that is why flexibility is important. Shaun's only problem seemed to be his reluctance to participate after the two-week Christmas break; He also appeared to benefit from direct teaching of targets and may have mastered IQI, /&/ and /r/-clusters had an extra block of intervention been conducted. Since these are generally considered to be later-developing phonemes, it was predicted that they would emerge without intervention at a later date. Contrastively, it was felt that the design could have been more optimal for Eddie with the following suggested changes: 1. Eddie could have had more prosodic goals on easier targets to initiate progress and increase his experience of success early on in the intervention program. In 157 order to avoid Eddie's use of default place, the thing he did best was intentionally avoided, leading to a very challenging intervention program for him. 2. Dorsals may have been used in clusters since he was good at them. Even so, this could have caused a phonetic coordination problem for him or else reinforced the default Dorsal/Coronal-Dorsal setting. 3. Eddie may have had two eight-week blocks of therapy with an eight-week break in between to give a needed rest and to improve gain generalization and maintenance 4. The timing of second weekly session (4pm Friday afternoon) was not the best but was unavoidable in order to allow both parents to participate. Since this was a high-priority family goal, the session time was not changed. 5.4.7 Directions for Future Research Expanding the number of experimenters involved would reduce experimenter bias and lead to an expanded understanding of optimizing treatment designs. Experimenter rationale for treatment decisions and changes to their original intervention plan for each child could lead to a broadening of how services are provided in the province of British Columbia. While most experienced clinicians follow the child's lead in therapy, the process of laying out a program, then documenting decisions and rationale for changing that program could lead to better research designs. These outcomes will then be more broadly applicable to the general population of children and speech-language pathologists in everyday clinical practise. 158 5.4.7.1 Data Collection In terms of future studies, behavioural measures of child characteristics in terms of needing to alter therapy within sessions by changing activities or techniques and parent characteristics in terms of proportion parent participation (direct and indirect) in activities and parental interruptions of child speech could be charted. Direct participation could be defined as the parent participating by (1) planning a therapy activity, and (2) actively participating (verbally) in a clinic therapy session. Indirect participation could be defined as (1) attending a parent information session, and (2) the physical presence of the parent in therapy room (verbal or non-verbal participation). Further, parental interruptions could be monitored and differentiated from participation in therapy activities on the basis of whether or not comments detracted from or contributed to the focus of the activity. Other optimization questions that need to be addressed include the following: 1. How does the level of severity at the initiation of the program impact the amount of progress achieved by subjects in this type of intervention? 2. What kinds of time frames are optimal for phonological intervention studies (i.e. 16 continuous weeks versus other permutations of time blocks)? 3. How can results obtained from the combination of segmental and prosodic goals be meaningfully interpreted? 4. What level of mastery is optimal for a segment prior to combining it with a new wordshape (or visa versa)? 5. Is the emergence of strategies such as epenthesis, initial consonant lengthening or marking a syllable slot with nasal air emission sufficient to document that a wordshape is developing? Can intervention on that target be terminated with 159 the expectation that sufficient learning has taken place to generalize to a mastery level? 6. Is this kind of intervention as effective once weekly or in groups? How would subjects respond to intervention provided under the following circumstances: a. Individual intervention once weekly b. Individual intervention once weekly plus group intervention once weekly c. Group intervention only (once versus twice versus three times weekly)? Theoretical questions that need to be addressed in future research include: 1. How do developmental data inform phonologists about what the vowel features are and how they are represented within the feature geometry? 2. What is the role of consonant-vowel interactions in phonological development? 3. How can constraint theory be applied to the assessment and intervention of children with phonological disorders? 5.5 Conclusions This study was an optimization of previous intervention research in NLP (Bernhardt, 1990, 1993a, Von Bremen, 1990). Both subjects made phonological gains as a result of intervention based on nonlinear phonological theory and twice weekly intervention was at least as effective as intervention provided three times weekly to matched subjects from the Bernhardt studies. The nature of phonological change during intervention using NLP theory was observed and related to developments in the theory, including Optimality Theory (Prince & Smolensky, 1993; Bernhardt & Stemberger, in prep), Grounded Phonology (Archangeli & Pulleyblank, 1994) and 160 related topics such as complexity constraints and consonant-vowel interactions. The nature of the interaction between phonology and phonetics was child-specific. For Shaun, consonant-vowel feature cooccurrence restrictions appeared to be phonetically motivated. For Eddie, who presented with a Coronal-Dorsal phonetic setting, reduced overall progress in phonological therapy appeared to be related to his unusual oral structures. In conclusion, clinical research must be theory -driven, flexible and applicable to everyday practise. Modifications to nonlinear phonological theories may result from the rigorous testing of these theories in case-by-case studies. Furthermore, carefully planned effectiveness studies can provide useful information to the speech-language pathologist working with any population. 161 References Archangeli, D. (1988). Aspects of underspecification theory. Phonology Yearbook, 5, 183-208. Archangeli, D., & Pulleyblank, D. (1994). Grounded phonology. Cambridge, MA: MIT Press Baum, J. D., & Kinmonth, A. -L. (1985). Care of the chUd with diabetes. Edinburgh: Churchill Livingstone. Bernhardt, B. (1990). Applications of nonlinear phonological theory to intervention with six phonologically disordered children. Unpublished doctoral dissertation, University of British Columbia. Bernhardt, B. (1992a). Developmental implications of nonlinear phonological theory. Clinical Linguistics and Phonetics, 6, 259-281. Bernhardt, B. (1992b). The application of nonlinear phonological theory to intervention with one phonologically disordered child. Clinical Linguistics and Phonetics, 6, 283-316. Bernhardt, B. (1993a). A nonlinear phonological intervention field study: Preliminary outcomes. ASH A Convention Abstracts, November, 1993,127. Bernhardt, B. (1993b). It works! Proving the effectiveness of phonological therapy in schools and clinics. Paper presented at the Conference of Speech and Hearing Association of Alberta, October, 1992, Banff, Alberta. Bernhardt, B. (1994a). Linking theory to therapy: Contributions of nonlinear phonology and metaphon. ASHA Convention Abstracts, November, 1994, 203. Bernhardt, B. (1994b). Phonological intervention techniques for syllable and word structure development. Clinics in Communication Disorders, 4,1, 54-65. Bernhardt, B. (1994c). Word and syllable structure development: Outcomes from a nonlinear phonological intervention field study. Paper presented at the Child Phonology Conference, May, 1994, Sun Valley, Idaho. Bernhardt, B. (1995). Cluster metatheses in phonologically disordered speech: Constraints on sonority sequence, mapping direction, or segment composition? Proceedings of the Canadian Linguistics Society Annual Conference, June, 1994, Calgary, AB (pp. 39-50). University of Toronto Working Papers in Linguistics. Bernhardt, B., & Cam, M. (1994). Speech Analysis. Unpublished phonological computer analysis program. University of British Columbia. Bernhardt, B., & Gilbert, J. (1992). Applying linguistic theory to speech-language pathology: The case for nonlinear phonology. Clinical Linguistics and Phonetics, 6(I&2), 123-145. 162 Bernhardt, B., MacNeill, H. & Bohlen, C. (1994). Nonlinear phonological intervention: Group and individual case study results. Journal of Speech-Language Pathology and Audiology Conference Abstracts, 18, 1,45. Bernhardt, B., Miller, M., Barton, M., MacAulay, K., MacKenzie, M., & Wastie, S. (1992). University and clinic in collaboration: Studying efficacy of phonological intervention. Journal of Speech-Language Pathology and Audiology Conference Abstracts, 17,47. Bernhardt, B., & Stemberger, J. P. (in prep). Multilinear phonology and phonological development. Bernhardt, B., & Stoel-Gammon, C. (1994). Nonlinear phonology: Introduction and clinical application. Journal of Speech and Hearing Research, 15, 811-820. Bernhardt, B., & Stoel-Gammon, C. (in press). Underspecification and markedness in normal and disordered phonological development. To appear in Children's Language, Vol. 7, Lawrence Erlbaum Associates. Bishop, D. (1988). Language development in children with abnormal structure or function of the speech apparatus. In Bishop, D. and Mogford, K. (Eds.). Language development in exceptional circumstances (pp. 220-238). Hillsdale, N.J.: Lawrence Erlbaum Associates. Blachman, B. (1991). Early intervention for children's reading problems: Clinical applications of the research in phonological awareness. Topics in Language Disorders, 12 (1), 51-65. Blockberger, S. (1993). From efficacy to effectiveness in AAC: Service delivery issues for the next decade. Unpublished manuscript, University of British Columbia. de Boysson-Bardies, B., & Vihman, M. (1991). Adaptation to language: Evidence from babbling and first words in four languages. Language, 67, 297-319. Bryant, P. E , Bradley, L., MacLean, M. & Crossland, J. (1989). Nursery rhymes, phonological skills and reading. Journal of Child Language, 16, 407-428. Chapey, R. (1994). Language intervention strategies in adult aphasia. Baltimore, MD: Williams & Wilkins. Chin, S. B. (1994). The role of the sonority hierarchy in delayed phonological systems. Paper presented at the Fourth Annual Congress of the International Clinical Phonetics and Linguistics Association, November 14,1994, New Orleans, USA. Chomsky, N., & Halle, M. (1968). The sound pattern of English. New York: Harper & Row. Clements, G. (1985). The geometry of phonological features. Phonology Yearbook, 2, 225-52. Clements, G., & Keyser, S. (1983). CV-phonology: A Generative Theory of Syllable. Cambridge: MIT Press. Connell, P., & Thompson, C. (1986). Flexibility of single subject designs: Part III: Using flexibility to design or modify experiments. Journal of Speech and Hearing Disorders, 51, 214-225. 163 Cooper, D. S., Ladefoged, P., Gracco, L. C , Stone, M. L., Westbury, J. R., Svirsky, M. A., Milenkovic, P. H., & Schroeter, J. (1994). Developments in vocal tract estimation. ASH A Convention Abstracts, November, 1994, 60. Dean, E., & Howell, J. (1986). Developing linguistic awareness: A theoretically based approach to phonological disorders. British Journal of Disorders of Communication, 21, 223-238. Dean, E., & Howell, J. (1994). Linking theory to therapy: Contributions of nonlinear phonology and metaphon. ASH A Convention Abstracts, November, 1994, 203. Dinnsen, D. A. (in press). Nonsegmental phonologies. In R. D. Kent & M. J. Ball (Eds.), The new phonologies: Developments in clinical linguistics. San Diego, CA: Singular. Durand, J. (1990). Generative and non-linear phonology. London: Longmans. Dworkin, J. P., & Culatta, R. A. (1985). Oral structural and neuromuscular characteristics in children with normal and disordered articulation. Journal of Speech and Hearing Disorders, 50, 150-156. Edwards, M. L., & Bernhardt, B. (1973). Phonological analyses of the speech of four children with language disorders. Unpublished manuscript, Stanford University, Institute for Childhood Aphasia, Stanford, CA. Edwards, M. L., & Shriberg, L. D. (1983). Phonology: Applications in communicative disorders. San Diego: College-Hill. Elbert, M., & Geirut, J. (1986). Handbook of clinical phonology: Approaches to assessment and treatment. San Diego, CA: College Hill. Ferguson, C , & Farwell, C. (1975). Words and sounds in early language acquisition: English initial consonants in the first fifty words. Language, 51, 419-439. Fletcher, R., Fletcher, S. & Wagner, E. (1988). Clinical epidemiology: The essentials. (2nd ed.). Baltimore, MD: Williams & Wilkins. Fraser, L., & Avery, L. (1985). Treating the young stutterer: Manipulating variables in favour of successful outcomes. Miniseminar presented at the Canadian Speech and Hearing Association Conference, May, 1985, Toronto, Ontario. Goldberg, B. (1993). Translating data into practice. ASHA, 35,45-47. Goldsmith, J. A. (1976). Autosegmentalphonology. Ph.D. thesis, Indiana University; Published 1979, New York: Garland Press. Goldsmith, J. A. (1990). Autosegmental & metrical phonology. Oxford: Blackwell. Grunwell, P. (1985). Phonological assessment of child speech. San Diego, CA: College-Hill Press. Halle, M., & Vergnaud, J. (1980). Three dimensional phonology. Journal of Linguistic Research, 1, 83-105. 164 Harris, J., & Cottam, P. (1985). Phonetic features and phonological features in speech assessment. British Journal of Disorders of Communication, 20, 61-74. Hayes, B. (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry, 20, 253-306. Hewlett, N. (1985). Phonological versus phonetic disorders: Some suggested modifications to the current use of the distinction. British Journal of Disorders of Communication, 20, 155-164. Hodson, B., & Paden, E. (1991). Targeting intelligible speech: A phonological approach to remediation. (2nd ed.: first ed.; 1983). Austin: PRO-ED. Hodson, B., & Scudder, R. (1990). The efficacy of intervention by speech-language pathologists: Phonological disorders in children. Seminars in Speech and Language, 11, 192-199. Holocmbe, A., Wolery, ML, & Gast, D. L. (1994). Comparative single-subject research: Description of designs and discussion of problems. TECSE, 14 (1), 119-145. Hyman, L. (1985). A Theory of Phonological Weight. Dordrecht: Foris. Ingram, D. (1981). Procedures for the Phonological Analysis of Children's Language. Baltimore: University Park Press. International Phonetics Association (1989). Recommended phonetic symbols: Extensions to the IPA, Clinical Linguistics and Phonetics, 6 (3), 259-262. Jakobson, R. (1968). Child language, aphasia, and phonological universals (A. R. Keiler, Trans.). The Hague: Mouton. (Original: Kindersprache, Aphasie, und Allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell, 1941.) Johnston, J. M. (1988). Strategic and tactical limits of comparison studies. The Behavior Analyst, 11, 1-9. Kaye, J. (1989). Phonology: A cognitive view. Hillsdale, NJ: Lawrence Erlbaum Associates. Kaye, J., Lowenstamm, J. & Vergnaud, J.-R. (1990). Constituent structure and government in phonology. Phonology, 7, 193-231. Kenstowicz, M. (1994). Phonology in generative grammar. Oxford: Blackwell. Kenstowicz, M., & Kisseberth, C. (1979). Generative phonology: Description and theory. New York: Academic Press. Kent, R. D. (1992). Phonological development as biology and behavior. In R. S. Chapman (Ed.), Processes in Language Acquisition and Disorders, (pp. 67-85). St Louis: Mosby Yearbook. Kent, R. D., & Hodge, M. (1991). The biogenesis of speech: Continuity and process in early speech and language development. In J. F. Miller (Ed.), Research on child language disorders: A decade of progress (pp. 25-53). Austin, TX: Pro-Ed. Kiparsky, P. (1979). Metrical structure assignment is cyclic. Linguistic Inquiry 10:421-441. 165 Kiparsky, P. (1985). Some consequences of Lexical phonology. Phonology Yearbook, 2, 83-138. Ladefoged, P. (1980). What are linguistic sounds made of? Language, 56, 485-502. Lahiri, A., & Evers, V. (1991). Palatalization and coronality. In C. Paradis & J.-F. Prunet (Eds.) Phonetics and phonology: The special status of coronals (77-100). San Diego: Academic Press Inc. Laver, J. (1994). Principles of Phonetics. Cambridge: Cambridge University Press. LeLaurin, K., & Wolery, M. (1992). Research standards in early intervention: Defining, describing, and measuring the independent variable. Journal of Early Intervention, 16, 275-287. Leonard, L. B. (1992). Models of phonological development and children with phonological disorders. In C.A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.) Phonological Development: models, research, implications (pp. 495-507). Timonium, Maryland: York Press. Lleo, C , & Prinz, M. (1994). Consonant clusters in child phonology and the directionality of syllable structure assignment. Unpublished manuscript. Louko, L. J., Edwards, M. L., & Conture, E. G. (1990). Phonological characteristics of young stutterers and their normally fluent peers: Preliminary observations. Journal of Fluency Disorders, 15, 191-210. Macken, M. A. (1978). Permitted complexity in phonological development: One child's acquisition of Spanish consonants. Lingua, 44, 219-253. Macken, M. A. (1987). Representation, rules, and overgeneralization in phonology. In B. MacWhinney (Ed.), Models of language acquisition (pp.367-397). Hillsdale, NJ: Erlbaum. Macken, M. A. (1992). Where is Phonology? In C. A. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.) Phonological development: Models, research, implications (pp. 249-269). Timonium, Maryland: York Press. Macken, M. A., & Ferguson, C.A. (1983). Cognitive aspects of phonological development: Model, evidence, and issues. In K. Nelson (Ed.), Children's language, vol.4, (pp. 255-282). Hillsdale, NJ: Erlbaum. Magnusson, E., & Naucler, K. (1990). Reading and spelling in language-disordered pre-school children-linguistic and metalinguistic prerequisites: Report on a longitudinal study. Clinical Linguistics and Phonetics, 4, 49-61. McCarthy, J. (1988). Feature geometry and dependency: A review. Phonetics, 43, 84-108. McCarthy, J. (1989). Linear order in phonological representation. Linguistic Inquiry, 20(1), 71-99. McCarthy, J., & Prince, A. (1986). Prosodic morphology. Unpublished manuscript, University of Massachusetts and Brandeis University. 166 McCarthy, J., & Prince, A. (1993). Generalized alignment. Yearbook of Morphology, 79-154. McCauley, R. J. (1993). A comprehensive phonological approach to the assessment and treatment of sound system disorders. Seminars in Speech and Language, 14 (2), 153-165. Menn, L. (1978). Phonological units in beginning speech. In A. Bell and J. Bybee Hooper (Eds.), Syllables and Segments (pp. 155-171). Amsterdam: NorthrHolland. Menn, L. (1983). Development of articulatory, phonetic, and phonological capabilities. In B. Butterworth (Ed.), Language production 2: Development, writing and other language processes (pp. 3-50). New York: Academic. Nippold, M. A. (1990). Concomitant speech and language problems in stuttering children. Journal of Speech and Hearing Disorders, 55, 51-60. Odden, D. (1994). Adjacency parameters in phonology. Language, 70, 289-330. Ohde, R. N., & Sharf, D. J. (1992). Phonetic analysis of normal and abnormal speech. Columbus, Ohio: Macmillan Publishing Company. Olmsted, D. L. (1971). Out of the mouth of babes: Earliest stages in language learning. The Hague: Mouton. Olswang, L. B. (1993). Developmental speech and language disorders, ASH A, 35, 42-44. Olswang, L. B., & Bain, B. (1994). Data collection: monitoring chidren's treatment progress. American Journal of Speech-Language Pathology, 3 (3), 55-66. Paradis, C. (1988). On constraints and repair strategies. The Linguistic Review, 6, 71-97. Paradis, C. (1993). Ill-formedness in the dictionary: a source of constraint violation. Canadian Journal of Linguistics, 38, 215-234. Paradis, C , & Prunet, J.-F. (1991). Phonetics and phonology: The special status of coronals. San Diego: Academic Press Inc. Pike, K., & Pike, E. (1947). Immediate constituents of Mazateco syllables. International Journal of American Linguistics, 13, 78-91. Prince, A., & Smolensky, P. (1993). Optimality theory: Constraint interaction in generative grammar. Unpublished manuscript, Rutgers University & University of Colorado. Raymond, M. (1992). The human side of diabetes. Chicago: The Noble Press. Rice, K., & Avery, P. (in press). Variability in a deterministic model of language acquisition: A theory of segmental elaboration. In J. Archibald (Ed.), Phonological acquisition and phonological theory. Hillsdale, NJ: Lawrence Erlbaum Associates. Ringo, C. C. (1985). The nature of change in phonological development: Evidence from the acquisition of Isl + stop and Isl + nasal clusters. Unpublished Ph.D. dissertation, Brown University. 167 Robbins, J., and Klee, T. (1987). Clinical assessment of oropharyngeal motor development in young children. Journal of Speech and Hearing Disorders, 52, 271-277. Sagey, E. (1986). The representation of features and relations in non-linear phonology. Unpublished Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, MA. Shriberg, L. D., & Kent, R. D. (1982). Clinical Phonetics. New York: John Wiley & Sons. Shriberg, L. D., & Kwiatkowski, J. (1980). Natural process analysis procedure for phonological analysis of continuous speech samples. New York: Wiley. Shriberg, L. D., & Kwiatkowski, J. (1994). Developmental phonological disorders I: A clinical profile. Journal of Speech and Hearing Research, 37, 1100-1126. Shriberg, L. D., Kwiatkowski, J. & Gruber, F. A. (1994). Developmental phonological disorders II: Short-term speech-sound normalization. Journal of Speech and Hearing Research, 37, 1127-1150. Shriberg, L. D., Gruber, F.A. & Kwiatkowski, J. (1994). Developmental phonological disorders III: Long-term speech-sound normalization. Journal of Speech and Hearing Research, 37, 1151-1177. Siegel, G. M. (1993) Research: a natural bridge. ASHA, 35,36-37. Smit, A. B. (1993). Phonological error distribution in the Iowa-Nebraska Articulation Norms Project: Word-initial consonant clusters. Journal of Speech and Hearing Research, 36, 931-947. Smit, A. B., Hand, L., Freilinger, J. J., Bernthal, J. E., & Bird, A. (1990). The Iowa Articulation Norms Project and its Nebraska replication. Journal of Speech and Hearing Disorders, 55, 779-798. Smith, N. (1973). The acquisition of phonology: A case study. Cambridge: Cambridge University Press. Spencer, A. (1984). A nonlinear analysis of phonological disability. Journal of Communication Disorders, 17, 325-348. Stampe, D. (1969). The acquisition of phonetic representation. In R. T. Binnick et al. (Eds.), Papers from the Fifth Regional Meeting of the Chicago Linguistic Society, (pp. 433-444). Chicago: Chicago Linguistic Society. Stampe, D. (1973). A dissertation on natural phonology. Unpublished Ph.D. dissertation, University of Chicago. Stengelhofen, J. (1989). Cleft palate: The nature and remediation of communication problems. New York: Churchill Livingstone Inc. 168 Stemberger, J. P., & Stoel-Gammon, C. (1991). The underspecification of coronals: Evidence for language acquisition and performance errors. In C. Paradis & J.-F. Prunet (Eds.), Phonetics and phonology, Volume 2. The special status of coronals: Internal and external evidence (pp. 181-199). San Diego: Academic Press. Steriade, D. (1988). Reduplication and syllable transfer in Sanskrit and elsewhere. Phonology, 5, 73-155. Stoel-Gammon, C., & Cooper, J. (1984). Patterns of early lexical and phonological development. Journal of Child Language, 11, 247-271. Stoel-Gammon, C , & Dunn, C. (1985). Normal and disordered phonology in children. Baltimore, MD: University Park Press. Stoel-Gammon, C , & Stemberger, J. P. (1994). Consonant harmony and phonological underspecification in child speech. In M. Yavas (Ed.), First and second language phonology (pp. 63-80). San Diego: Singular Press. Templin, M. C. (1957). Certain language skills in children: Their development and interrelationships. Institute of Child Welfare Monographs, 26. Minneapolis: University of Minnesota. Van Riper, C. (1972). Speech correction: Principles and methods. Englewood Cliffs, N.J.: Prentice-Hall. Vennemann, T., & Ladefoged, P. (1973). Phonetic features and phonological features. Lingua, 32, 61-74. Vihman, M. M. (1978). Consonant harmony: Its scope and function in child language. In J. H. Greenburg (Ed.), Universals of human language, vol.2: Phonology (pp. 281-334). Stanford, CA: Stanford University Press. Von Bremen, V. (1990). A nonlinear phonological approach to intervention with severely phonologically disordered twins. Unpublished Master's thesis, University of British Columbia. Weiner, F. (1981). Treatment of phonological disability using the method of meaningful contrasts: Two case studies. Journal of Speech and Hearing Disorders, 46, 97-103. Wolery, M., & Ezell, H. K. (1993). Subject descriptions in single-subject research. Journal of Learning Disabilities, 26, 642-647. 169 A Shaun's Data Table A-1 Phonetic inventory3: ShaunTj W I b SI SF WF N A S A L S Labial m m m m Coronal n n n n Palatal Dorsal STOPS Labial (P)b ((P))b ((b*)) Coronal ((th)) (t)d d((t)) (((t)))(Y*# ((CM)) Palatal Dorsal FRICTIVES/ AFFRICATES Labial (((il)) ((v)) (((f:w))) ((v)) ((D) Coronal ((s>)) ((f)) ((d3)) ((z)) ((s>)) (D (((s))) (((s>))) (((z*))) Palatal Dorsal LIQUIDS a s GLIDES j w j (w) GLOTTALS h? h((?)) (?) (?) a. (((marginal)))=l-10%; ((emergent))=ll^ H)%; (developing)=41-75%; established=75-100%; {1 token only}; ita/i'c^ subsritution only; bold=substitution and match b. WI=Word-Initial, SI=Syllable Initial Within Word, SF=Syllable Final Within Word, WF=Word-Final 170 Table A-2 Phonetic inventory2: ShaunT2 W I b SI SF WF N A S A L S Labial m m m m Coronal n n n (n) ((J?J Palatal Dorsal STOPS Labial (P)b b {p}b Coronal th((t))d ((t))d (((t)))d ((tV (((tV) Palatal Dorsal Wig} ((g)) {k${kf{g} FRICTIVES/ AFFRICATES Labial ((v)) {f} ((v)) f(v) Coronal ({8))(S>){Z>} i(s)){z>} ((s)) ((s>)) (3 (S) Palatal Dorsal LIQUIDS 1 (1) GLIDES j w j w GLOTTALS h? h ( ( ( ? ) ) ) a. (((marginal)))=l-10%; ((emergent))=ll-40%; (developing)=41-75%; established=75-100%; {1 token only}; Jte/ic=substitution only; bold=substi tution and match b. WI=Word-Initial, SI=Syllable Initial Within Word, SF=Syllable Final Within Word, WF=Word-Final Table A-3 Phonetic inventory3: ShaunT3 W I b SI SF WF N A S A L S Labial m m m m Coronal n n n n Palatal ((A) Dorsal to) ((g)) STOPS Labial pb{B} pb (p)b(((Pw Coronal th((t))d (t)d d (t)d((tV(((tti) Palatal Dorsal (((k h)))gk k g k(g){k}(((gV) FRICTIVES/ AFFRICATES Labial fy(((f:))) f(v) (f)(v) (0 (v) Coronal sz((s>)) a)tr ds (s) z.tr (fe (a)) cm (A s sz{jf tf ((«)) ((/)) Palatal Dorsal LIQUIDS i (i) ((D) ((D) GLIDES w(j) (w) j GLOTTALS h? h ? a (((marginal)))=l-10%; ((emergent))=ll-40%; (developing)=41-75%; established=75-100%; {1 token only}; jta/ic=substitution only; bold=substitution and match b. WI=Word-Initial, SI=Syllable Initial Wthin Word, SF=Syllable Final Within Word, WF=Word-Final Table A-4 Intervention goal progress summary: Shaun T r T 3 Target TO T l T2 %Gai n T l -T2 T3 %Gai nT2-T3 %Gai n T I -TS Onset Coda Onset Coda Onset Coda #l.[+con s] vs. [+son]a 8/43 (19%) 5/12 (42%) 19/36 (53%) 15/27 (56%) 23% /14% 114/126 (90%) 54/61 (89%) 37% /33% 71% /47% (#1. o.g.)b 0/2 (0/20) 0/1 (0/5) 0 (0) 0/10 (0/23) 0 (0) 0 (0) #2. Liq-uid III WI (WM) 0/8 (0/5) 3/3 (3/5) 100% (60%) 28/28 (9/16) 0 (-4%) 100% 56% Liquid lil WI (WM) (#2 o.g.) 0/12 (0/6) 0/5 (0/3) 0 (0) 0/25 (0/14) 0 (0) 0 (0) #3. Dorsal Ikl 0/16 0/12 10/16 (62%) 3/23 (23%) 62% WI 23% ' WF 39/42 (93%) 26/34 (76%) 31% WI 53,% WF 93% WI 76% WF #3. Dorsal lg/ 0/16 0/2 1/2 (50%) 2/6 (33%) 50% WI 33% WF 15/15 (100 %) 5/10 (50%) 50% WI 17% WF 100% WI 50% W F Dorsal /n/ (#3. o.g.) 0/20 4/29 (14%) 14% 12/67 (18%) 4% 18% #4. CG_+ ( C L J 2/20 (10%) 1/17 (6%) (-4%) 35/55 (64%) 58% 54% sC_ (#4.o.g.) 0/11 0/9 0 31/45 (69%) 0 69% #5. map /f,v/ to WI 3/22 (14%) 0/8 4/19 (21%) 6/8 (75%) 7% WI 54% WF 56/70 (80%) 25/29 (86%) 59% WI 11% WF 66% WI 86% WF 173 Table A-4 Intervention goal progress summary: Shaun T r T 3 Target TO T l T2 %Gai n T l -T2 T3 %Gai nT2-T3 %Gai n T I -TS Onset Coda Onset Coda Onset Coda #6. sC_/_Cs (sC above) 0/7 0/4 0 24/29 (83%) 83% 83% Cr_ (#6. o.g.) 0/19 0/20 0 1/47 (2%) 2% 2% _cc (#6. o.g.) 2/12 (17%) 4/6 (67%) 50% 36/43 (84%) 17% 67% a. Onset refers to WI and SI position while coda refers to SF and WF positions for goals #1 and #5 only. b. o.g. = observation goal Table A-5 Vowel development: Shaun Tj-T3 Vowel/ Diphthong T l T2 T3 % gain T1-T3 x/y % x/y % x/y % HI 50/54 93% 43/44 98% 139/147 95% 2% hi 23/53 43% 21/54 39% 143/171 84% 41% IdSl 31/39 79% 17/25 68% 70/76 92% 13% lul 34/35 97% 17/17 100% 55/59 93% (-4%) hi 3/3 100% 4/6 67% 15/17 88% (-12%) lal 19/35 54% 14/22 64% 58/89 65% 11% /ai/ + IAII 26/30 87% 19/19 100% 35/52 67% (-20%) laol + IAOI 13/22 59% 9/11 82% 34/38 89% 30% /oo/ 33/37 89% 16/17 94% 52/57 91% 2% 174 Table A-6 Wordshape Accuracy T r T 3 : Shaun T i T 2 T 3 o 3 O 5 Wordshape x/y (% correct) Wordshape x/y (% correct) Wordshape x/y (% correct) 2VV 1/1 (100) GV 2/2(100) ?V 1/1 (100) GV 3/4 (75) CV 37/39 (95) ?VV 10/11(91) G V V 1/1 (100) C W 8/8(100) GV 4/4(100) CV 27/31(87) C V V C 14/18(78) G V V 1/1 (100) C W 22/24 (92) C V C V 8/8(100) CV 37/39(95) cvvcv 1/1 (100) cvvcv 3/4 (75) C W 14/15(93) 2VVCVC 1/1 (100) G V C ?VVC C V V C C C V V 11/14(79) 4/5 (80) 38/52(73) 8/11(73) C V C V 25/32(78) cvvcv 9/9(100) ?vvcv 2/2 (100) W G V C V C 3/3 (100) ?v 1/2 (50) G V C 5/10 (50) C V C 94/143 (66) ?VC 5/9 (56) C V C 25/43 (58) C C V C 33/56 (59) G V C 7/13 (54) G V C V C 1/2 (50) C V C C 25/52(48) C V V C 14728 (50) C V C V C 9/22 (41) C V C V V 1/2 (50) .If C V C V 7/14 (50) C V C C V C 2/5 (40) C C C V C 1/2(50) O C V C V C 33/58 (57) ?vcvc 4/6 (67) Q G V V C V C 2/4(50) ?vvc 2/6 (33) C V C C 4/18 (22) 2VC 4/27(15) C V C 29/80 (36) C V V G V C 1/5 (20) C G V 1/3 (33) C G V V 2/6(33) ccv 3/12(25) ?vcvc 2/5 (40) G V C C 3/9 (33) C V C V C 5/19(26) C V V C C 1/6(17) C V V G V C 2/5(40) C C V V C 3/11(27) cvccvc 1/6(17) ccvcc ccvcv C G V C V C 6/18 (33) 1/7(14) 1/9(11) | C V V G V C 3/10 (30) ccvcvc 2/9 (22) w cvccvc 10/25(40) 1 C V C C 1/20(5) none none none none S 175 Table A-1 Cluster development: Shaun T r T 3 Cluster Shape T l T 2 T3 Monosyllabic 2 Element Clusters WI CCVV CCV 0/2 0/4 0/5 0/1 8/11 3/12 73 25 CGV CGVV 2/6 33 - - 1/3 33 CCVC CCVVC 0/20 0/1 0/16 0/2 33/56 3/11 59 27 WI& WF CCVVCC CCVCC 0/2 -0/1 -6/18 33 WF 2VCC GVCC GVVCC CVCC CVVCC 0/1 0/1 1/20 0/4 5 4/18 22 0/5 .3/9 25/52 1/6 33 48 17 Multisyllabic 2 Element Clusters WI CCVCV CCVCVC CGVCVC CCVVCVC 0/2 0/5 0/2 0/4 0/1 0/6 0/2 0/5 1/7 2/9 1/9 14 22 11 WF cvcvcc cvcvcc 0/1 - 0/3 WM cvccv cvccvc 1/1 1/6 100 17 0/1 2/5 40 0/1 10/25 40 Monosyllabic 3 Element Clusters WI cccvc 0/2 0/1 1/2 50 WF cvccc 0/2 - 0/2 176 B Eddie's Data Table B - l Phonetic inventory3: Eddie Tj WI b SI SF WF N A S A L S Labial m m ((m)) m Coronal e Palatal J ((J)) (J) J Dorsal STOPS Labial p b pb (P)fr*b Coronal {*}(t)(d)(((tj)) ((t)) {d} (($){£${&}, (((!))) ((d)) Palatal ((J))c {c} Dorsal &g«(k))) k g k g FRICTIVES/ AFFRICATES Labial f(v) (0 V Coronal is}(((s1))) ((z*)f{s}{z1} Palatal «f)) Dorsal (((*))) LIQUIDS GLIDES j w j w GLOTTALS h,? h a. (((marginal)))=l-10%; ((emergent))=ll-40%; (developing)=41-75%; established=75-100%; {1 token only}; ita/ic=substitution only; bold=substitution and match b. WI=Word-Initial, SI=Syllable Initial Within Word, SF=Syllable Final Within Word, WF=Word-Final c. J=palatal stop d. * refers to a voiceless segment 177 Table B-2 Phonetic inventory 3: Eddie T 2 W I b SI S F W F N A S A L S Labial m m m m Coronal n ((n)) Palatal ((n!)) Dorsal STOPS Labial p h b pb Coronal t h ((d*)) (t) (d) {°d} {tU(d*}(&) Palatal (((J))) {c} Dorsal k*g((k)) k g k g F R I C T I V E S / A F F R I C A T E S Labial f (data gap-v) f((v)) (*)(v) Coronal (((z>*))) ((z*)) Palatal ((d) (((<?))) ((C)) {tC} Dorsal { k > ^ L I Q U I D S ((D) G L I D E S j w j w G L O T T A L S h? ? a. (((marginal)))=l-10%; ((emergent))=ll-40%; (developing)=41-75%; established=75-100%; {1 token only}; ifa/i'c=substiration only; bold=substitution and match b. WI=Word-Initial, SI=Syllable Initial Within Word, SF=Syllable Final Within Word, WF=Word-Final Table B-3 Phonetic inventory3: Eddie T 3 WI b SI SF WF N A S A L S Labial m m {m} m Coronal n (n) {n} n Palatal V V (((n!))) Dorsal STOPS Labial p h b pb p b Coronal t h d (t) (d) {°d} (t)d {t\t\v} Palatal {c} {c} Dorsal &g(((k))) k g k g FRICTIVES/ AFFRICATES Labial f((v)) f ( ( v » (t)(v) Coronal ((s)) ((s>, s1))) (((ts.ts'.t;1))) Palatal (D c Dorsal (O)) LIQUIDS ((D) GLIDES (j) w j (w) GLOTTALS h,? a. (((marginal)))=l-10%; ((emergent))=ll-40%; (developing)=41-75%; established=75-100%; {1 token only}; ifa/ic=substitution only; bold=substitution and match b. WinWord-Initial, SI=Syllable Initial Within Word, SF=Syllable Final Within Word, WF=Word-Final 179 Table B-4 Intervention goal progress summary: Eddie Tj -T 3 Target T3 / 0 b T l T2 %Gain T1-T2 WI (WF) T3 %Gain T2-T3 WI (WF) %Gain T1-T3 WI (WF) Onset Coda Onset Coda Onset Coda #1. Coro-nal IM WI(WF) T 2/22 (4%) 5/15 (33%) 6/15 (40%) 4/15 (27%) 31 (-6) 24/26 (92%) 16/24 (67%) 52 (40) 83 (34) Coronal /d/WI (WF) T 5/11 (45%) 3/7 (43%) 2/6 (33%) 3/4 (75%) -12 (32) 16/17 (94%) 5/6 (83%) 61 (8) 49 (40) Coronal /n/WI (WF) O 0/8 0/17 5/5 3/10 (30%) 100 (30) 20/20 31/34 (91%) 0(61) 100 (91) Coronal /s/WI (WF) O 1/19 (5%) 0/8 2/9 (22%) 0/3 17(0) 9/25 (36%) 2/12 (17%) 14 (17) 31 (17) #2. Liq-uid /1/WI T 0/14 0/4 0 2/16 (13%) 13 13 Liquid WI O 0/7 0/6 0 0/14 0 0 #3 WI sC_ T 0/20 0/13 0 1/13 (8%) 0 8 WI C G _ O 3/16 (19) 3/11 (27%) 8 4/30 (13%) (-14) (-14) #4. WI Cor /s/ link to [+cont]\ V [+hi] T 1/19 (5%) 2/9 (22%) 17 9/25 (36%) 14 31 #4for/z/ T 0/2 0/3 0 0/4 0 0 / /A V [+hi] O 0/1 0/4 0 3/5 (60%) 0 60 180 Table B-4 Intervention goal progress summary: Eddie T1-T3 Target Tl T l T2 %Gain T1-T2 T3 %Gain T2-T3 %Gain T1-T3 0 b Onset Coda Onset Coda WI (WF) Onset Coda WI (WF) WI (WF) #5. WI [+dist] IQI T 0/3 0/2 0 0/3 0 0 #5. Ibl T 0/9 0/2 0 0/6 0 0 /tf/ O 0/6 0/5 0 5/11 0 0 /<fe/ O 0/2 0/4 0 0/4 0 0 a. T=Treatment goal b. 0=Observation goal 181 Table B-5 Wordshape accuracy Tj -T 3 : Eddie Tl Degree of Accuracy Wordshape x/y(% correct) T2 Wordshape x/y(% correct) T3 Wordshape x/y (% correct) Estab-lished C V G V C C V V G V V C C V V C G V C V C ? V G V V ? V C V C G V V C V C 7 V V C V V ? V V C V C 13/16 (81) 2/2(100) 9/11 (82) 1/1 (100) 14/16 (88) 2/2(100) 1/1 (100) 2/2(100) 1/1 (100) 1/1(100) 1/1 (100) GV CV G V C C V V ?VC ?VV ? V V C C V C C V ?vcvc G V V C V C cvvcvc ?vvcvc 2/2(100) 7/9 (78) 1/1 (100) 5/5(100) 2/2(100) 1/1 (100) 1/1 (100) 1/1 (100) 1/1 (100) 1/1 (100) 1/1 (100) 1/1(100) C V C V V ? V V ?vvc G V C V C C V C C V ?vvcc G V V C V C CVGCVVCC 19/24(100) 17/19 (89) 6/6(100) 5/5(100) 3/3(100) 2/2(100) 1/1 (100) 2/2(100) 1/1 (100) Develop-ing C V C ?VC C V C V ? V V C C V V C V C C V V G V C 27/57 (47) 3/5 (60) 7/15 (47) 1/2 (50) 1/2 (50) 1/2 (50) C V C G V V C C V V C C V C V G V C V C C G V C V C 12/25 (48) 1/2(50) 5/10 (50) 4/7 (57) 2/3 (67) 1/2 (50) G V C C V C ?VC C V V C C V C V C V V C V ? V C V C C V V G V C cvvcvc C G V C V C ? V V C V C 7/14 (50) 47/83 (57) 11/26 (42) 23/33 (70) 9/16(56) 4/6(67) 2/3 (67) 2/4 (50) 2/3 (67) 1/2 (50) 1/2(50) Emerging Marginal C V C V C C V V C V C G V C V C C V C C V C C V C C 3/15 (20) 1/3 (33) 1/3 (33) 1/9(11) 1/12 (8) C V C C C V C V C C V C C V C 1/8(12) 3/13 (23) 1/5 (20) GV C V C C C V C V C C V C C V C 2/6(33) 5/24(21) 12/39 (31) 2/11 (18) 182 Table 5-9 Cluster development: Eddie T j - T 3 Cluster Shape T l T2 T3 Monosyllabic 2 Element Clusters WF ? V V C C C V C C 0/3 1/12 8% 0/1 1/8 12% 1/1 5/24 100% 21% Multisyllabic 2 Element Clusters WI C G V C V C 1/3 33% 1/2 50% 1/2 50% W M C V C C V C V C C V C 0/1 1/9 11% 1/1 1/5 100% 20% 2/2 2/11 100% 18% Table B-6 Intervention session data: Eddie Sess # Goal #SLP models S PE DE E Total use % cor-rect Comments 1 /t,d/A Discrim: 4/8 (50%) Coronal + V[+hi] 2 /t,d/A Discrim: 10/20 (50%) 3 WI /t,d/A 176 Discrim: 11/13 (85%) 4 WF/t,d/ A + OM 140 3/26 2/14 1/25 6/65 9% Discrim: 10/14(71%) 5 /1/A Reluctance to speak; Dad's 1st session 6 WI III syllable/ words 14/25 (56%) 2/11 16/36 36% Syllables (16/36) = words (16/36) 7 WI III syll/word 167 6/39 (15%) 25/57 (44%) 3/8 (38%) 3/13 (23%) 37/117 32% Discrim: 3/5 (60%) 8 WI III word 43 2/7 (29%) 10/35 (29%) 12/42 29% Eddie sC_ target-sylls but not words 9 sCA 129 1/1 5/11 (45%) 0/3 0/5. (0%) 6/20 30% Skewed data (+4) nasal emiss. 10 sC 131 1/9 (11%) 5/7 (71%) 4/24 (17%) 7/15 (47%) 17/55 31% Break- "C" [si] with only slight lat. 11 sC 33 31/50 (62%) 6/14 (43%) 37/64 58% 1- 10: 8/28 (29%) 2- 10: 29/36 (81%) 183 Table B-6 Intervention session data: Eddie Sess # Goal #SLP models S PE DE E Total use % cor-rect Comments 12 sC 51 4/26 (15%) 1/1 0/5 0/4 5/36 14% Non-compliant day 13 /s,z/ \_V[+hi] 125 0/6 10/10 2/3 1/1 13/20 65% Many prompts & cues 14 /s,z/ \_V[+hi] 35 7/8 (88%) 9/12 (75%) 16/20 80% 15 /s,z/ \_V[+hi] home visit (notesa: lat./c/; IWok over syll (e.g./pitsa/, /idzi/ for 'easy') 16 /s,z/ \_V[+hi] tape corrupt 17 /s,z/ \V[+hi]_ IW: most accurate position 18 /s,z/ \V[+hil_ WF: lateral release 19 WI/6,6/A 96 0/1 A=4/10; Break-[9noomsen] (x2) 20 WI /6,&/ Is/syll 27 2/7 (29%) 13/13 0/17 1/1 16/38 42% lst-10: 4/9 (44%) 2nd-10: 15/20(75) 21 22 WI /e,&/ Is/words 85 Is: 13/26 w: 1/26 50% 4% Back-2wk break I>W4/26t.o. not blowing 23 24 WI IQ,bl words 27 0/19 8/18 (44%) 5/14 (36%) 13/51 25% [6soa6abiz]' these' using oral prompts 25 WI /e,6/ syll/word 73 w. 0/1 s: 35/45 w:0/16 35/45 0/17 78% 0% Isolation 100%; sylls 78%; words 0 26 WI iBJbl words 132 5/19 (26%) 41/72 (57%) 46/91 51% Inc spon. words self-practise 27 WI 11/ 40 10/16 (63%) 15/33 (45%) 15/22 (68%) 40/71 56% Incr. correct spont. lst-10: 15/33(45%) 2nd: 15/22 (68%) 28 /t,d/ 209 21/22 (95%) 36/39 (92%) 11/11 68/72 94% Good spontaneous sc with false assert. 29 sC 72 0/3 4/10 23/40 (58%) 5/12 (42%) 32/65 49% sn-2/54; s:n-26/54; to.-3/54; %-8/54; n-15/54 30 WI /9,6/ words 68 6/11 (55) 13/25 (52) 3/3 (100) 22/39 56% Held previous gains/co-ord low 184 C Intervention strategies C.1 Oral-Motor Strategy The use of a variety of cues (e.g. visual, verbal, auditory and tactile) in therapy to elicit production of speech sounds has been practised widely for decades (e.g. Van Riper, 1972). This strategy provides opportunities for the child to learn through a preferred sensory system or by combining input from several systems. Further, physical practise using the musculature of speech provides opportunities to establish and coordinate the neurolinguistic pathways associated with the production of speech and language. Sitting in front of a large mirror with a child and looking at the position of each other's tongues before eliciting an IM or rubbing the child's alveolar ridge to facilitate tongue placement are two examples of the oral-motor strategy. C.2 Awareness Strategy Dean and Howell (1986) advocate awareness training of phonemic constrasts as the primary focus of intervention instead of requiring the child to perform production tasks. More recently, Dean and Howell (1994) discussed that a period of therapy that requires the child to produce phonological contrasts appears to be beneficial, especially following a period of awareness training. Awareness activities range from the detection of minimal pair contrasts (e.g. two/Sue) to stories involving character names, settings and events that involve repetitive contrasts over the period of the story, (e.g. Winnie the Pooh and Roo live at the Zoo. They are hungry for Stew and don't know what to Do. They're friend Woo goes around to find out Who has some Stew...). 185 D Syllable structure intervention strategies D.1 Onset-Rime Strategy Using this theory, an intervention strategy for establishing word-initial (e.g. /bl/V consonant clusters would involve identifying the onset (i.e. [bsl]) and the rime (i.e. [oo]) then providing some visual anchor (e.g. a toy "bull" and magnetic letter "O") to represent the parts. The child and speech-language pathologist would then work together to connect the onset to the rime by gradually bringing the objects closer together in space and the sounds closer together in time, eventually making the word "blow". Likewise, an intervention strategy for mapping a word-final consonant (e.g. Ikl) to word-initial position would involve quickly producing a series of V C (e.g. [ak]) alternations. Given the phonetic tendency to optimize the onset, C V and C V C would be produced in a series. The final C can be "cut o f f short, leaving an open syllable, if desired, (e.g. [ok] > [akakak] > [akaka] > [ka]). D.2 Mora 'Edges' Strategy This technique can be used to establish consonant clusters (e.g. /st/) by adding the [s] to the edge of an existing word such as [tu]. An intervention activity would then involve bringing a representation of the [s] such as an "s" drawn with a snake head closer and closer in space and time to a representation of [tu] such as a magnetic or drawn "2" to produce the word "stew". Moraic theory can also be used to map consonants from one syllable position to another, the idea being that the weight unit is established in the rime of a C V C word when a C V C V word is "cut off" short. For example, "bunny" becomes "bun" by clapping out the syllables on the one's knees and stopping short of saying [i] on the last production (e.g. [bA.ni], [bA.ni], [bAn.0]). 186 

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0087117/manifest

Comment

Related Items