Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Protracted phonological development in Granada Spanish : a case study Klassen, Josiah 2017

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

Item Metadata

Download

Media
24-ubc_2017_november_klassen_josiah.pdf [ 1.9MB ]
Metadata
JSON: 24-1.0344008.json
JSON-LD: 24-1.0344008-ld.json
RDF/XML (Pretty): 24-1.0344008-rdf.xml
RDF/JSON: 24-1.0344008-rdf.json
Turtle: 24-1.0344008-turtle.txt
N-Triples: 24-1.0344008-rdf-ntriples.txt
Original Record: 24-1.0344008-source.json
Full Text
24-1.0344008-fulltext.txt
Citation
24-1.0344008.ris

Full Text

PROTRACTED PHONOLOGICAL DEVELOPMENT IN GRANADA SPANISH: A CASE STUDY by  Josiah Klassen  B.A., McGill University, 2013  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Audiology and Speech Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  April 2017  © Josiah Klassen, 2017   ii Abstract  This case study examines single-word productions of “Emilio”, a Spanish-speaking boy from Granada with protracted phonological development (PPD), at ages 4;8 and 7;6. Emilio was the only child in a cohort of 15 to exhibit persisting difficulties at the second time point. At time 2, he showed differences in word-initial /r/, syllable-final /ɾ/ and one polysyllabic word, hipopótamo. In an acoustic comparison of /r/ data from children aged 4 to 9 years (Carballo & Mendoza, 2000), Emilio’s /r/ attempts had higher formant frequencies (F1 and F2), less distinct occlusions and apertures, and longer duration. The current study examines Emilio’s first dataset to determine phonological factors that may have predicted persisting PPD. The analysis adopts a nonlinear phonological framework (Bernhardt & Stemberger, 1998), and evaluates his data with respect to 16 expectations about Spanish acquisition derived from research literature. His 4-year-old phonological skills are compared with those of two 4-year-old peers who had similar whole word match (WWM) scores at the first data collection (WWM is the percent of words produced that match their targets; Bernhardt et al., 2015a). For Emilio, global measures including WWM were within the expected range for 4-year-olds with PPD. Three word structure measures, however, differentiated his phonology at age 4 from his two peers: overall word shape match (WSM), WSM of words with clusters, and word-initial cluster match (accuracy) in stressed syllables. (WSM is the percent of words produced with consonant-vowel [CV] sequences matching the adult target sequences; Bernhardt et al., 2015a.) Emilio also showed more difficulty with dorsal consonants, and words with alternating coronal-labial cross-vowel sequences than Sofía and Tomás. Those particular variables may be good candidates for predicting long-term   iii difficulty. Other measures, including liquid and fricative matches and phonological processes, were not differentiators for these 4-year-olds. These results highlight the relevance of word structure for assessment and treatment of PPD, in alignment with Bernhardt et al. (2015a) and Schretlen (2013). The relevance of two ways of measuring WSM, with or without glides in the consonant category, are discussed. Further research is needed to confirm or disprove the usefulness of these predictors.    iv Preface  This thesis is original and independent work by the author, Josiah Klassen. The author was not involved in the original data collection, but he conducted the literature review, all analyses and the writing of this thesis in its entirety. This thesis was written under supervisor Dr. B. May Bernhardt and thesis committee members Dr. Joe Stemberger and Dr. Laura Colantoni.     v Table of Contents  Abstract .......................................................................................................................................... ii	  Preface ........................................................................................................................................... iv	  Table of Contents ...........................................................................................................................v	  List of Tables ..................................................................................................................................x	  List of Figures ............................................................................................................................... xi	  List of Symbols ............................................................................................................................ xii	  List of Abbreviations ................................................................................................................. xiv	  Acknowledgements ......................................................................................................................xv	  Dedication ................................................................................................................................... xvi	  Chapter 1: Introduction ................................................................................................................1	  1.1	   Motivation for this project .............................................................................................. 1	  1.1.1	   Why PPD? ............................................................................................................... 1	  1.1.2	   Why Granada Spanish? ........................................................................................... 5	  1.1.3	   Why a case study? ................................................................................................... 5	  1.1.4	   Why Emilio? ........................................................................................................... 7	  1.2	   Theoretical background: constraint-based nonlinear phonology .................................... 7	  1.3	   The phonology of Granada Spanish .............................................................................. 12	  1.3.1	   Variation in Granada Spanish ............................................................................... 15	  1.4	   Acquisition of Spanish phonology: typical and protracted ........................................... 16	  1.4.1	   Global measures .................................................................................................... 17	  1.4.1.1	   Previous data for global measures .................................................................... 18	    vi 1.4.2	   Acquisition of word structure ............................................................................... 20	  1.4.2.1	   Consonant clusters: timing unit match .............................................................. 22	  1.4.2.2	   Consonant clusters: full segmental match ......................................................... 24	  1.4.2.3	   Word-final consonants ...................................................................................... 26	  1.4.2.4	   Diphthongs ........................................................................................................ 27	  1.4.3	   Acquisition of segments ........................................................................................ 28	  1.4.3.1	   Singleton vowels ............................................................................................... 29	  1.4.3.2	   Singleton consonants ........................................................................................ 29	  1.4.3.3	   Singleton consonants by word position ............................................................ 31	  1.4.4	   Acquisition of features and feature combinations ................................................ 31	  1.4.4.1	   [Dorsal, +consonantal] (dorsal consonants) ...................................................... 32	  1.4.4.2	   [-sonorant, +continuant] (fricatives) ................................................................. 33	  1.4.4.3	   [+sonorant, +consonantal] and [+lateral] or [+vibrant] (liquids) ...................... 34	  1.4.5	   Mismatch patterns during acquisition ................................................................... 36	  1.4.6	   Expected variables of relevance at age 4;8 ........................................................... 39	  Chapter 2: Method .......................................................................................................................41	  2.1	   Participants .................................................................................................................... 41	  2.2	   Procedures ..................................................................................................................... 42	  2.3	   Analysis......................................................................................................................... 43	  Chapter 3: Results ........................................................................................................................45	  3.1	   Expectations and results ................................................................................................ 45	  3.2	   Descriptive findings: age 4;8 ........................................................................................ 46	  3.2.1	   Global measures .................................................................................................... 46	    vii 3.2.2	   Word structure ...................................................................................................... 47	  3.2.2.1	   Consonant clusters: timing unit match .............................................................. 49	  3.2.2.2	   Consonant clusters: full segmental match ......................................................... 52	  3.2.2.3	   Codas................................................................................................................. 52	  3.2.2.4	   Vowel sequences ............................................................................................... 52	  3.2.3	   Segments ............................................................................................................... 54	  3.2.4	   Features and feature combinations ........................................................................ 58	  3.2.4.1	   [Dorsal, +consonantal] (dorsal consonants) ...................................................... 58	  3.2.4.2	   [+continuant, -sonorant] (fricatives) ................................................................. 58	  3.2.4.3	   [+sonorant, +consonantal] and [+lateral] or [+vibrant] (liquids) ...................... 60	  3.2.5	   Phonological mismatch patterns ........................................................................... 61	  3.2.5.1	   Singleton substitutions ...................................................................................... 61	  3.2.5.2	   Phonetic differences .......................................................................................... 64	  3.2.5.3	   Cross-vowel consonant harmony ...................................................................... 64	  3.2.5.4	   Migration ........................................................................................................... 66	  3.2.5.5	   Coalescence? ..................................................................................................... 67	  3.2.5.6	   Epenthesis ......................................................................................................... 67	  3.2.6	   Description of constraints ..................................................................................... 68	  3.2.7	   Variability in productions ..................................................................................... 69	  3.3	   Descriptive findings: age 7;6 ........................................................................................ 70	  3.3.1	   Global measures .................................................................................................... 70	  3.3.2	   Word structure ...................................................................................................... 71	  3.3.2.1	   Consonant clusters ............................................................................................ 71	    viii 3.3.2.2	   Vowel sequences ............................................................................................... 71	  3.3.3	   Segments ............................................................................................................... 71	  3.3.3.1	   Phonetic description of “irregular” /r/ .............................................................. 72	  Chapter 4: Discussion ..................................................................................................................74	  4.1	   Review of descriptive results: age 4;8 .......................................................................... 74	  4.2	   Review of descriptive results: age 7;6 .......................................................................... 77	  4.2.1	   Comparison of difficult words across data collections ......................................... 78	  4.2.2	   Discussion of challenges at age 7;6 ...................................................................... 79	  4.3	   Comparisons with expectations (Time 1) ..................................................................... 81	  4.3.1	   Variables meeting expectations ............................................................................ 81	  4.3.2	   Variables that did not meet expectations .............................................................. 84	  4.3.3	   WSM values from Bernhardt et al. (2015a) .......................................................... 85	  4.4	   Results relative to expectations ..................................................................................... 87	  4.5	   Relevance of other measures ........................................................................................ 88	  4.6	   Predicting persistent difficulty in phonological development ...................................... 89	  4.7	   Limitations of this study ............................................................................................... 90	  4.8	   Clinical implications ..................................................................................................... 91	  4.9	   Future directions ........................................................................................................... 92	  Chapter 5: Conclusion .................................................................................................................93	  References .....................................................................................................................................94	  Appendices ..................................................................................................................................112	  Appendix A. Wordlists ........................................................................................................... 112	    ix Appendix B. Scan analysis at time 1, age 4;8 (adapted from Bernhardt et al., 2016) ............ 117	  Appendix C. Spectrogram of word with “irregular” /r/ at age 7;6 ......................................... 122	     x List of Tables  Table 1.1 Consonant and glide inventory of Granada Spanish ..................................................... 12	  Table 1.2 Global measures for preschoolers ................................................................................. 19	  Table 1.3 Tendencies in phonological mismatch patterns ............................................................ 37	  Table 1.4 Expected performance for 4-year-old Granada Spanish speakers with PPD ................ 39	  Table 3.1 Expectations versus results at age 4 .............................................................................. 45	  Table 3.2 Falling diphthong matches and mismatches ................................................................. 53	  Table 3.3 Emilio's diphthongs compared with Ponce (2014) ....................................................... 54	  Table 3.4 Overall and singleton matches by consonant ................................................................ 55	  Table 3.5 Singleton matches by consonant ................................................................................... 56	  Table 3.6 Singleton matches by word position ............................................................................. 57	  Table 3.7 Singleton substitutions by word position ...................................................................... 62	  Table 3.8 Phonetic measures of /r/ at time 2 ................................................................................. 73	  Table 4.1 WWM and WSM of 3- and 4-year-olds in Emilio's cohort (Bernhardt et al., 2015a) . 86	     xi List of Figures  Figure 1 The nonlinear prosodic word: peine, ‘comb’ ................................................................... 8	  Figure 2 The timing unit tier: pan, /ˈpan/, ‘bread’ .......................................................................... 8	  Figure 3 Feature geometry (adapted from Bernhardt & Stemberger, 2000, with permission) ....... 9	     xii List of Symbols  All of the data in this case study are coded in the International Phonetic Alphabet (IPA, International Phonetic Association, 2015). IPA in this document is recognizable by being enclosed in either slashes, /r/, or square brackets, [r], which represent target/adult pronunciations and actual/child pronunciations respectively or phonemic transcriptions and phonetic transcriptions (allophones), respectively. Fortunately many letter-symbols in Spanish correspond to their equivalents in IPA, and so many pronunciations can be deciphered by a careful reader. Here are some of the non-English IPA symbols: Symbol Description in words β voiced bilabial fricative/approximant r apical alveolar trill (vibrant) ʝ voiced palatal ungrooved fricative n ̪ dental nasal (diacritic: dental consonant) ɲ palatal nasal x voiceless velar fricative ɣ voiced velar fricative/approximant ˠ voiced velar fricative/approximant (superscript: negligible presence) a,̞ r ̞ diacritic: lowered segment ɬ voiceless lateral coronal fricative    xiii Ages in this document are represented by two numbers separated by a semicolon, which represent years and months respectively. “4;8” is therefore 4 years and 8 months; “7;6” is 7 years 6 months.   xiv List of Abbreviations  Abbreviations in this document are usually introduced at least once with their full form, but some common ones are listed here:  C: consonant IPA: International Phonetic Alphabet G: glide MSW: multi-syllabic word PCC: percent consonants correct PMLU: phonological mean length of utterance PPD: protracted phonological development PVC: percent vowels correct PWP: proportion of whole word proximity SLP: speech-language pathology/pathologist TD: typically developing / typical development V: vowel WSM: word shape match WWM: whole word match    xv Acknowledgements  A heartfelt thank you to everyone who made my Master’s program and this thesis possible. Most importantly, the boy from Granada whom I never met, whom I call Emilio: your participation in the Granada cohort made it all possible. Second, “Sofía” and “Tomás,” two more children from Granada whose participation gave me the ability to write this thesis. My thesis committee: especially my tireless supervisor Barbara May Bernhardt, but also Joe Stemberger and Laura Colantoni. Those who proofread my proposal and drafts: Karyn Santiago, Meagan Klassen, Gillian O’Toole. Nicholas Aultman, who provided technical consultation in Microsoft Word. Gaby Raymond, who provided moral support and also a sound-file of her pronouncing rojo with a typical Spanish /r/ (Appendix 3). My classmates: UBC SASS grad classes of 2016-2018, especially Jess Spoor, who paved the way for me with her case study in Québec French. Those who babysat and otherwise helped out with the baby, especially my immediate family and family-in-law. And said baby: Kalia Jane Emilia Klassen.   xvi Dedication  To Meagan.   1 Chapter 1:  Introduction This thesis examines phonological data from a boy from Granada, Spain, "Emilio", who participated in a phonological acquisition study at two time points: when he was four years, eight months of age (4;8), and again when he was seven years and six months (7;6). Emilio was the only child in a longitudinal cohort of 15 to exhibit difficulties at a second time point, and he was receiving speech therapy focusing on trilled /r/. The introduction describes the motivation for this study and provides background regarding the term protracted phonological development (PPD), the phonology of Granada Spanish, theoretical frameworks for the analysis and previous reports on Spanish phonological acquisition.  1.1   Motivation for this project This project continues the research program headed by Barbara May Bernhardt into PPD across languages. The following sections outline the reasons such research is warranted and in particular this study, justifying the choice of topic, language, research method and participant.  1.1.1   Why PPD? Protracted phonological development (PPD) is a term used to describe the acquisition of the forms of speech which takes place over a longer period of time than average. The term “PPD” is used in place of others such as “speech sound disorder”, “phonological impairment”, “speech delay”, or others for its comparatively neutral/positive valence and its greater accuracy than “delay” (Bernhardt & Stemberger, 1998; Chávez-Peón et al., 2012; Dubasik & Ingram, 2013). PPD is also more specific to phonological description than the terms “disorder” and   2 “impairment” which also imply the functional or life consequences of the longer period of acquisition. PPD can occur in the absence or presence of other anatomical (e.g. cleft palate), sensory (e.g. hearing loss), motor (e.g. cerebral palsy) or genetic (e.g. Down Syndrome) differences. PPD is more likely in the case of male sex, low maternal education and positive family history (Campbell et al., 2003).  Some researchers distinguish “articulation disorders” (inability to pronounce certain phones, such as /s/ or /r/ on target) from “phonological delays” (in which typical phonological errors are seen, but of a younger age), or further differentiate the observed difficulties into “consistent phonological disorders” (in which atypical errors are consistently produced) and “inconsistent phonological disorders” (in which atypical errors are inconsistently produced; Dodd, Holm, Crosbie & McIntosh, 2006). PPD encompasses all of the categories; typicality and consistency of errors are further distinctions that can be made within PPD.  Estimates of the prevalence of PPD vary widely, ranging from 2.3% to 24.6% (Harasty & Reed, 1994; Kirkpatrick & Ward, 1984; Law, Boyle, Harris, Harkness & Nye, 2000; Stewart, Hester & Taylor, 1986; Warr-Leeper, McShea & Leeper, 1979). In any case, speech assessment and treatment form a large part of pediatric SLP caseloads (Broomfield & Dodd, 2004). Nevertheless, cross-linguistic research in PPD is only recently gaining traction. With almost 35% of Canadians speaking more than one language, and around 55 million Americans, research into other languages is more necessary than ever (Grosjean & Li, 2013). Even for serving English-  3 speaking clientele, cross-linguistic research can only strengthen the phonological theories which motivate assessment and treatment goals.  Preschoolers with PPD are also quite likely (50-75% chance) to have expressive language impairment (Shriberg & Kwiatkowski, 1994a). Furthermore, children with speech and language difficulties can be at risk for continued communication difficulties even many years later, although those with speech impairments alone are less likely to have continued communication difficulties (Johnson et al., 1999), reading difficulties (Bernhardt & Major, 2005; Skebo et al., 2013) and psychological or psychosocial conditions (Beitchman et al., 2001; Lewis et al., 2016a). Other related aspects of phonology, such as phonological awareness and phonological working memory, are often correlated with later literacy skills and/or later spoken language skills (Bernhardt & Major, 2005; Lewis et al., 2011). Adults with a history of moderate-to-severe PPD are more likely work in occupations viewed as semi-skilled or unskilled than age- and education-matched peers and siblings (Felsenfeld & Broen, 1994), but are not more likely to have atypical scores on personality measures such as neuroticism or extroversion (Felsenfeld & Broen, 1992).  Persistence of phonological difficulties over time is a less-researched area than PPD overall, but some researchers have examined factors that may predict or accompany “persistent developmental speech disorder”, or persistent PPD, in English-speaking children and adults (Flipsen, 2015). Persistent PPD has been indicated through single-word samples (percentage of deletions and substitutions), connected speech samples and non-word repetition tasks (Wren, McLeod, White, Miller & Roulstone, 2013). Wren, Miller, Peters, Emond and Roulstone (2016) estimated the prevalence of persistent PPD at 3.6%, and identified motor, cognitive and linguistic   4 factors that put children at risk. Factors that co-occur with persistent PPD include speech perception and short-term memory deficits (Kenney, Barac-Cikoja, Finnegan, Jeffries & Ludlow, 2005). Predictive factors identified include number of deletions (Shriberg & Kwiatkowski, 1994a; Shriberg, Kwiatkowsi & Gruber, 1994b), articulation and phonological processing measures (Lewis et al., 2006), co-occurrence of phonological and phonetic errors, male gender, lower IQ, and lower socio-economic status (Wren, Roulstone & Miller, 2011). Children with persistent PPD may also have lower reading skills than those with “transient” PPD, possibly as a result of weaker underlying phonological representations (Lewis et al., 2006; Raitano, Pennington, Tunick, Boada & Shriberg, 2004; Stothard, Snowling, Bishop, Chipcase & Kaplan, 1998).  Some speech sound differences (or PPD) persist into adulthood, while others resolve either spontaneously or in conjunction with treatment (Flipsen, 2015). Flipsen categorizes persistent PPD into “residual” (which resisted treatment) and “persistent” (untreated for various reasons), and argues that the two types of speech sound differences have different acoustic properties and optimal treatment approaches (Flipsen, 2015). (For this thesis the term "persistent" is used, in keeping with most previous papers.) Treatment approaches vary, but there is some evidence that instruments including electropalatography (Gibbon, Stewart, Hardcastle & Crampin, 1999) and ultrasound (Adler-Bock, Bernhardt, Gick & Bacsfalvi, 2007; Cleland, Scobbie & Wrench, 2015; Preston et al., 2014) can be effective for treating persistent PPD, as well as specific treatment approaches such as those based on nonlinear phonology (Bernhardt, 1992a; Bernhardt & Major, 1995) or stimulability (Powell, 1996).    5 This study will build on this body of research by piloting exploration into persistent PPD in Granada Spanish, focusing on phonological measures in single words.  1.1.2   Why Granada Spanish? Spanish is spoken by over 500 million speakers around the world in at least 30 countries, in North and South America, Europe, Africa and Asia (Lewis et al., 2016b). As a major world language, it was selected for the cross-linguistic study of 13 other languages, led by Bernhardt and Stemberger (in press). The phonology of Granada Spanish (Andalusian) is similar to other varieties of Spanish spoken in Andalusia and in Latin America, especially in the optional deletion of codas (Bernhardt et al., 2015a). The current study thus will deepen our understanding of PPD in Granada Spanish, but also in other places with similar dialects..  1.1.3   Why a case study? A case study, as defined by Damico and Simmons-Mackie (2003), is characterized by four attributes: 1) it has a specified object of inquiry; 2) the object of inquiry can be studied separately; 3) the study has a relatively short timeframe (as opposed to a biographical study); and 4) something of value can be learned from this individual case. Emilio’s case satisfies these four criteria: 1) his protracted phonological development is the object of inquiry; 2) his datasets enable study of that object alone; 3) the longitudinal timespan is 3 years (certainly shorter than a biography); and 4) there is potential to glean predictive factors from his case because he was unique in his cohort, that is, only his difficulties persisted into the time of the second data collection. According to Damico and Simmons-Mackie’s (2003) further distinction, this case is both “instrumental” because it may shed light on a greater phenomenon, i.e. persistent PPD in   6 Spanish, and “collective” because its information can be combined with the others in the cross-linguistic series to shed light on PPD generally.  Case studies can play a pioneering role in developing fields and open up new avenues of research, which would not be feasible nor practical without the suggestive evidence that they can provide. Case studies are also useful for providing qualitative and quantitative information that would not be possible to explore with a larger number of participants.   To date, PPD in Spanish has not been described in great detail. The question of persistence over time and predictive factors earlier on is even less well understood. As part of a series of case studies, Emilio’s case will be comparable to its counterparts in other languages, which will make each of them more accessible to researchers more familiar with any one of the other languages. As Chávez-Peón et al. (2012) note, such a case study can shed light on which constraints on Spanish acquisition may be language- or dialect-specific, which may be universal, and which may be child-specific. Emilio's specific case study may also shed further light on the generally later acquisition of rhotic segments. Additionally, it may provide evidence on aspects of markedness in Spanish phonology, which could both develop our understanding of Spanish phonology and clarify choices in complexity-based intervention (Barlow, 2001b; Gierut, 2001). Relative to intervention, a case study can provide clinicians with a detailed picture of one actual client, rather than an amalgam of generalizations from groups of clients. While simple in some respects (the boy is monolingual and has no known concerns other than PPD), thus allowing for a clear focus on the topic, this study also explores the complexity and richness of one child’s phonological development.   7 1.1.4   Why Emilio? Emilio was selected from the cohort described in Bernhardt et al. (2015a, 2015c) and Schretlen (2013), comprising fifty-nine 3-, 4- and 5-year-olds with typical development (TD, 30 participants) and PPD (29). Fifteen of the children with PPD were assessed at two time-points, and on the second all but one participant (Emilio) had resolved their phonological difficulties, most following speech therapy. Emilio had also received therapy, including for trilled /r/, but his difficulties persisted through the second assessment point. Examining his data from the first time-point relative to patterns described in the literature about PPD in Spanish could shed light on patterns most predictive of long-term phonological difficulties. These findings may stimulate future research into PPD and treatment, and could also inform clinicians as they make predictions and treatment decisions. Analysis and research by nature depend on theory; the following section outlines the theories relied upon in this research.  1.2   Theoretical background: constraint-based nonlinear phonology The descriptive analysis largely draws from nonlinear phonology and optimality theory. Nonlinear phonology focuses on hierarchies of tiers of phonological units (Bernhardt & Stoel-Gammon, 1994). These tiers are autonomous and interactive and include the prosodic word tier, foot tier, syllable tier, timing unit tier, segment tier and/or feature tiers. In the description below, Spanish examples are used. The following section provides additional information on Spanish.  Starting at the top of the hierarchy, Figure 1 below describes the Prosodic Word and Foot levels; “σ” refers to the syllable. Peine is single-footed, as most Spanish words are, which means that it only has one stressed syllable (in this case, the first).   8  Figure 1 The nonlinear prosodic word: peine, ‘comb’  Pan (Figure 2 below) is another single-footed word, which also has only one syllable. This figure shows the level of form below the syllable, i.e. the timing unit tier, where each segment of the word has at least one timing unit. In words with geminate consonants or long vowels, such as Spanish /penne/ (variant of peine, ‘comb’), one phone can be linked to two Cs or two Vs. The word shape for pan is CVC (consonant-vowel-consonant). In terms of syllable structure, its form is onset-nucleus-coda, the nucleus and coda together forming the rhyme/rime.   Figure 2 The timing unit tier: pan, /ˈpan/, ‘bread’ The features of the various consonants and vowels are linked below the timing units. Features, too, are related to one another in a “feature geometry” of interconnected nodes (Figure 3).   9  Figure 3 Feature geometry (adapted from Bernhardt & Stemberger, 2000, with permission)  The nodes in Figure 3 correspond to the phonological features that distinguish Spanish segments from one another. They are all related to one another in a hierarchy. The bolded nodes dominate others: [Root], which is the feature under which all others are organized. Manner of articulation features fall directly under this node, including [+/-nasal], [+/-sonorant], and the only feature on this tree that is not relevant to English phonology: [+/-vibrant]. The feature [Laryngeal] describes the glottal features of a segment, which include [+/-voiced] and [+/-spread glottis] for adult Spanish phonology. [+/-constricted glottis] is relevant for glottal stops, which are not phonemic in Spanish, but which do appear once in Emilio’s productions. Finally, [Place] designates the supralaryngeal articulators, namely the lips (Labial) and tongue (Coronal, or tip and blade, and Dorsal, tongue body). These three place features differentiate consonants by major place, but each major place feature dominates other specific articulations, e.g. [+/-anterior], [+/-distributed]   10 and/or [+/-grooved] for coronals. Relevant features for Emilio's analyses are discussed further in those sections outlining his results.   Regarding vowels, Spanish does not require the [+/-tense] distinction that English needs to distinguish some vowels. The five vowels of Spanish can be distinguished entirely on three binary dimensions: [+/-high], [+/-low] and either [+/-back] or [+/-rounded], which are redundant with one another.  The framework above provides a basis for description of phonological mismatch patterns between targets and produced (actual) forms in developmental phonology. Mismatches can occur at any level of the phonological hierarchy with e.g. syllable deletion, segment substitution, stress shift or feature metathesis. A key distinction between nonlinear phonology and traditional generative phonology is the multi-tiered rather than solely sequential description of phonological form (Bernhardt, 1992b; Bernhardt & Gilbert, 1992). In this way nonlinear phonology is able to account for complex and interacting changes between distant elements in a word without needing to force them into sequences.  Another theoretical framework also provides a way to describe phonological form during development. In optimality theory, markedness and faithfulness constraints describe the tension and interplay between complex target forms and the requirement to match such forms (Barlow, 2001a; McCarthy, 2007) These constraints are ranked in relative importance within a word and across a phonological system and their ranking determines the output forms chosen to best (not necessarily most closely) match target forms, the optimal (and actual) outputs (Barlow & Gierut,   11 1999). Faithfulness constraints describe the various ways that forms can be similar or identical to target forms (such as MAX - no deletion), while markedness constraints describe the avoidance of complex or difficult forms (such as *CODA - no codas).   Phonological development can therefore be described as a series of re-ranked constraints. For example, children begin with very strong markedness constraints, ranking, e.g. *CODA (no coda) and *DORSAL (no dorsal consonants) very high, with typical consequences of coda deletion (with or without compensatory lengthening) and potentially dorsal fronting to coronals. In English, /l/ is a later-developing phoneme than in Spanish (Cataño, Barlow & Moyna, 2009). This may be in part due to the [Dorsal] feature /l/ has in English coda position (“dark l”). As a result, *LATERAL can be ranked quite high for English-speaking children, but not for Spanish-speaking ones. As the child becomes more capable of producing codas and dorsal consonants, faithfulness to these targets increases and the negative markedness constraints become lower-ranked than those faithfulness constraints. This can happen gradually (optional re-rankings) or more suddenly. For a comprehensive account of constraint-based nonlinear phonology as it applies to phonological acquisition, please see Bernhardt and Stemberger (1998).  In this next section, Granada Spanish is described according to the levels of nonlinear phonology; each tier (segment, syllable, foot, word) is considered both independently from and in interaction with the others.    12 1.3   The phonology of Granada Spanish Table 1.1 lays out the consonants of Granada Spanish. Granada Spanish has five primary vowels, and they are described after the consonants.   [Labial] [Coronal, +anterior] [Coronal,  -anterior] [Dorsal] [Laryngeal] Plosives [-continuant] p ba t db  k gc  Approximants [+sonorant] βa ðb ʝd ʝd    ɣc  Fricatives [-sonorant] [+continuant] f s/θ (ʃ)e xf  Affricates [-continuant], [+continuant]   t ͡ʃe dʒ͡d   Nasals [+nasal] m ng, (n)̪g ɲ ɲ  (ŋ)g  Lateral [+lateral]  l    Rhotics [+vibrant] (Trill/tap)  ɾ, r    Glides [-consonantal] w  j j   w hf Table 1.1 Consonant and glide inventory of Granada Spanish (a-g Allophonic variants are labelled with matching letters. The allophone [h] can also substitute for /s/ in coda position. Glides /w/ and /j/ are the first elements of rising diphthongs, not true consonants. The palatals have both [Corona] and [Dorsal] place features. The parentheses indicate less pervasive variants.)  The segments in parentheses are less frequent allophonic variants. The voiced fricatives/approximants [β], [ð] and [ɣ] and the voiced oral stops [b], [d], and [g] are three pairs of allophones, with the voiced fricatives/approximants occurring intervocalically (Barlow, 2003b; Eddington, 2011). Intervocalic [d] ([ð]) can also be deleted, as was found in 23% of university-educated Granada adults in Gómez-Molina and Gómez-Devís (2008; cited in Samper, 2011). “Yeísmo” refers to the pronunciation of “ll” as [ʝ] (Real Academia Española y Asociación de Academias de la Lengua Española [RAE], 2011). The affricate [dʒ͡] is the word-initial   13 allophone of /ʝ/. The fricative [ʃ] is a variant of /t ͡ʃ/, especially word-medially (e.g. flecha, [ˈfle{tʃ͡/ʃ}a]; RAE, 2011), although Samper (2011) notes that the fricative realization of /t ͡ʃ/ is in regression in Granada.   The nasals [n]̪ and [ŋ] are allophones of /n/ before dental and velar obstruents, and sometimes [ŋ] occurs word-finally where other dialects have [n] (although the pronunciation of most [+anterior] coronals is more dental than alveolar). The glottal [h] replaces /s/ in coda position (or the /s/ simply does not appear, with vowel shifts instead as noted below); [h] is also an alternative pronunciation of /x/.  Samper and Hernández-Cabrera (1995) found that coda /s/ is realized most commonly as [h] word-internally and /Ø/ word-finally (cited in Samper, 2011). /θ/ and /s/ can be distinct or replace one another variably, with [θ] use termed ‘ceceo’ and [s], ‘seseo’ (RAE, 2011). Samper (2011) notes that distinction between /s/ and /θ/ is now the norm, but merging is more common in some, more traditional areas of Granada.  The vibrants /r/ and /ɾ/ are only distinguished in intervocalic contexts and may be considered allophones in free variation in some Spanish varieties (Gómez, 2004). The trill /r/ occurs as a   14 singleton in word-initial and word-medial (intervocalic) contexts, and word-medially after /n/, /l/ and /s/. The other vibrant, tap /ɾ/, can occur as the second element in onset clusters, in coda position (word-medially or word-finally), and as a singleton word-medially. The tap in coda position alternates with /l/ (RAE, 2011) as is discussed further below.  Vowels in Granada Spanish include the five short vowels of other Spanish varieties: /a/, /e/, /i/, /o/, and /u/. They combine to form diphthongs, called “falling” and “rising” (in sonority). The falling diphthongs can be denoted as two vowels: /ai/, /ao/, /au/, /eu/, /ei/, /oi/, /ou/, whereas the rising diphthongs are transcribed here with a glide (/w/ or /j/) followed by a vowel: /wa/, /we/, /wo/, /wi/, /ja/, /je/, /jo/, /ju/. They can also be found in the literature, often with special diacritics, as /ua/, /ue/, /uo/, /ui/, /ia/, /ie/, /io/, /iu/, a transcription which has the feature of matching Spanish orthography. Some rising diphthongs may be realized with an initial consonant as in [ˈʤe.lo] for hielo and [ˈɡwe.so] for hueso (Bernhardt et al., 2015a). When word-final codas delete, vowels lower, e.g., the mid vowels /e/ and /o/ are realized as [ɛ] and [ɔ].  The word structure of Granada Spanish phonology is somewhat simplified with respect to other Spanish varieties. As in the other varieties, syllables may maximally have an onset of a consonant followed by a liquid, a diphthong (rising or falling) or simple nucleus, and a coda. The minimal syllable is a single-vowel nucleus (Quilis, 1997). Compared with other varieties, non-  15 sonorant codas can be deleted or replaced with [h], and the resulting open vowels are sometimes lowered, especially /o/. As noted above, sonorant codas may be deleted, such as /ɾ/ in /floɾ/ and, less commonly, nasals. The liquids /ɾ/ and /l/ in word-medial coda position are also sometimes interchanged or the consonant sequence is geminated, e.g. tortuga, /torˈtuga/ > [totˈtuɣa].  The typical prosodic word is maximally one foot, having one stressed syllable (exceptions being adverbs ending in -miente and some compound words). Disyllabic, trochaic words are the most common (Quilis, 1983) and generally are acquired first (Lleó, 2006). Quilis reports that 41.9% of words have two syllables, 27.7% one, 20.3% three and around 10%, four or more. Of the disyllabic words, 79.5% have stress on the first syllable (left prominent/trochaic). In longer words, right- or centre-prominence is more common, where the stress is either rightmost or on a central syllable.  1.3.1   Variation in Granada Spanish In Granada, like in every city, children can be exposed to more than one language variety. In Emilio’s case, his pronunciations were closer to traditional Granada Spanish in the first sample, but closer to more standard Northern Central Spanish (Castilian) in the second. Likely this reflects exposure to a more standard form of the language in school, which, for example, includes codas and both /s/ and /θ/. Sociocultural level is an important factor across all Spanish varieties according to Samper (2011), an example being that the deletion of intervocalic ‘d’ declines as one becomes more affluent or educated.   16 Bidialectalism, or even “bivarietism”, and its effects on phonological acquisition are still poorly understood and defined, but it might be assumed that it has similar effects to bilingualism: acceleration of some learning and deceleration of other. In any case, dialect must be taken into account: scoring dialect features as errors can easily result in a diagnosis of PPD or a speech sound disorder where there is none, or increase the severity rating of a disorder from mild to moderate or severe (Goldstein & Iglesias, 2001). In this case, exposure to multiple varieties of Spanish requires flexibility in targets: accordingly, I have posited mostly Granada Spanish targets at age 4;8 and more standard ones at 7;6. These forms are closer to targets one might expect from Castilian - the rise of which to a standard form of Spanish is described by Hernández-Campoy (2011). As Goldstein (2007) argues, when comparing the relatively conservative dialect of Mexican Spanish with the more “radical” Puerto Rican, dialect differences should not result in significant differences in measures of phonological development, provided the differences are purposefully taken into account and not misinterpreted as errors in the one dialect.   1.4   Acquisition of Spanish phonology: typical and protracted Given the background above, previous literature on the acquisition of Spanish phonology is discussed below, organized by type of measure from global to word shape, segment and feature level. Broad trends are described, focusing in particular on 3- and 4-year-olds, both TD and with PPD. While some studies focus on Emilio’s demographic, i.e. monolingual Spanish-speaking children from Granada (Bernhardt et al., 2015c; Pérez, 2013; Schretlen, 2013), more studies focus either on monolingual Spanish-speaking children from elsewhere (Bosch, 2004; Burrows & Goldstein, 2010; Ponce, 2014), or on bilingual children in North America (Burrows &   17 Goldstein, 2010; Chávez-Peón et al., 2012; Goldstein & Pollock, 2000, 2004). Word lists used to elicit datasets vary across studies, although several used one nearly identical to that of this case study (see Appendix 1), first published in Chávez-Peón et al. (2012), and including Bernhardt et al. (2015a, 2015c), Pérez (2013), Ponce (2014) and Schretlen (2013), some of which have examined data from the same Granada cohort.  1.4.1   Global measures Global measures calculate properties of entire words or sets of words, rather than more specific aspects of those words, in order to give an overall sense of phonological development or faithfulness to the adult targets. Whole word match (WWM) is a simple measure of how many produced words are exactly identical to their adult targets, including the stress pattern and every segment. Small phonetic differences, nevertheless, such as exact voicing or tongue placement, may be ignored (Bernhardt et al., 2015a).  Percent consonants correct (PCC) measures how many consonant segments match the adult target, without requiring perfect accuracy of whole words. PCC has been proposed as a reflection of severity of speech sound disorder (Shriberg & Kwiatkowski, 1982). Similarly, percent vowels correct (PVC) measures how many vowels match the adult targets.  Phonological mean length of utterance (PMLU) is a measure of whole-word complexity, for both child and adult utterances (Ingram, 2002). It has been suggested to be more indicative of phonological development and impairment than PCC (Arias & Lleó, 2014). PMLU is calculated by assigning points to each word in a dataset (1 point for each segment plus 1 point for each   18 correct consonant) and dividing the sum by the total number of words. Percentage of whole-word proximity (PWP) is calculated by dividing the PMLU of the produced utterances by the PMLU of the corresponding target words.  In order to capture information at the segment, syllable and prosodic word level, Schretlen (2013) developed a multisyllabic word (MSW) rubric, following previous work by Mason, Bernhardt and Masterson (2011; developed further in Mason, 2015). This metric evaluates word structure (stress placement, the number of syllables and syllable shape) and segment-structure interaction mismatches (syllable assimilation or reduplication, for example). Descriptive labels were assigned to various types of mismatch with a score indicating distance from the target (e.g., Syllable Deletion 1.0 and Singleton Consonant Deletion 1.0). Mismatches can compound, as they always do with syllable deletions, for example: a nucleus segment is always deleted when its syllable is deleted.  1.4.1.1   Previous data for global measures Table 1.2 on the next page summarizes global measures from previous research. First looking at WWM, Bernhardt et al. (2015a) studied 30 TD children and 29 children with PPD, all monolingual Spanish speakers from Granada (ages 3-5), and found WWMs of TD 3- and 4-year-olds of 58.0% and 85.4% respectively. The 3- and 4-year-olds with PPD, by contrast, had WWMs of 21.4% and 37.5% respectively. The 3- and 4-year-olds with PPD, by contrast, had WWMs of 21.4% and 37.5% respectively. Pérez (2013) compared monolingual, middle-class, Spanish-speaking children in Granada with Down Syndrome (28 participants), PPD (27), language delay (18) and TD (30). She found that between 3;1 and 5;11, children with PPD had a   19 WWM of 41.7%, compared with TD 79.8%. Ponce (2014), studying 30 TD Chilean 3-, 4- and 5-year-olds (divided evenly among the three age groups) found a WWM for TD 3-year-olds of 60.2% and 4-year-olds of 56.2%. Chávez-Peón et al. (2012) found a WWM of 17% and 9.5% for two speakers of Mexican and Argentinian Spanish with PPD at age 4;9. All of these studies used wordlists approximately identical to the ones used in this study. Measure TD PPD Sources Ages WWM 58.0% 85.4% 79.8% 60.2% 56.2% 21.4% 37.5% 41.7% - - Bernhardt et al. (2015a) Bernhardt et al. (2015a) Pérez (2013) Ponce (2014) Ponce (2014) 3;0-3;11 4;0-4;11 3;1-5;11 3;0-3;11 4;0-4;11 PCC - 64.8% Burrows & Goldstein (2010) 3;2-5;3 PMLU - 6.63 Burrows & Goldstein (2010) 3;2-5;3 PWP - 0.80 Burrows & Goldstein (2010) 3;2-5;3 Table 1.2 Global measures for preschoolers  Regarding PCC, Chávez-Peón et al. (2012) found a percent consonant match (PCM) for singletons of 57.6% and 51.9% for their two participants with PPD aged 4;9. Burrows and Goldstein (2010), studying 8 monolingual English speakers, 8 monolingual Spanish speakers and 8 Spanish-English bilinguals, found monolingual Spanish speakers (ages 3;2 to 5;3) with speech sound disorders (SSD) to have a PCC of 64.8% and their bilingual (Spanish-English) peers, also with SSD, 53.0%, using the phonology subsection (28 items in Spanish) of the Bilingual English Spanish Assessment (BESA; Burrows & Goldstein, 2010).  As for PMLU, Burrows and Goldstein (2010) found monolinguals with SSD (ages 3;2-5;3) had a PMLU of 6.63 and PWP of .80. Bilingual Spanish scores were PMLU: 6.16 and PWP: .74. These are comparable to the numbers in Loatman (2001) for six TD children (American children with   20 over 90% exposure to Mexican Spanish) ages 1;8 to 2;11: PMLU: 6.3 and PWP: .82. Dubasik and Ingram (2013) found higher scores for only 4-year-olds with PPD: PMLU of 7.03, 6.87 and 7.00 (cf. TD 7.94, 7.65 and 7.94) and PWP scores of .85, .83 and .84 (cf. .96, .92 and .96).  Regarding overall trends with a MSW metric, Schretlen (2013) analyzed the cohort of monolingual Spanish speakers from Granada (30 TD & 29 with PPD; ages 3-5) and found increased mismatch scores with increased word length, and age and group effects in polysyllabic words compared with the more comparable disyllabic scores.  1.4.2   Acquisition of word structure Acquisition of word structure can be measured through analyses of word length (in syllables or segments), stress, word shape (consonant-vowel sequence) and further specified consonant-vowel combinations, e.g. accuracy of consonant clusters (CC), or diphthongs (VV). The 3- and 4-year-olds in Emilio’s cohort with PPD had median exact structure matches (stress and consonant-vowel shape) of 66.7% and 75% respectively (Schretlen 2013).  Bernhardt et al. (2015a) found that word shape match (WSM) increased with age in both TD and PPD groups, from 64% to 92% to 95% in TD 3-, 4- and 5-year-olds, and from 48% to 63% to 75% among those with PPD. With clusters in the word, WSM did not exceed 37% in any category for those with PPD. Pérez (2013) notes that word shape match (WSM) was 62.9% for 3- to 5-year-olds with PPD, compared with TD 89.2%. Ponce (2014) found that TD 3- to 5-year-olds’ WSM was 80.1%.    21 In terms of order of syllable structure acquisition, Serra, Serrat, Solé, Bel and Aparici (2000) found that CV and V syllable structures are acquired (with 90% accuracy) at age 1;6, VV and VVC at 2;8, CVVV and VVCC at 3;6, CCVV, VC, CVVC, and CCVC at 3;9, and CVV, VCC, and CVC at 4;0. Lleó & Prinz (1996) posit the acquisition sequence CV>CVC>CVCC>CCVCC (codas before onset clusters). Bosch (2004) considered syllable deletion a sign of risk for age 3.   Relative to word length, Lleó (2002) found that Spanish monolinguals tended to move from single-footed words to words with what she calls an additional unfooted syllable before the trochee. Astruc, Payne, Post, Prieto and Vanrell (2010) studied the word prosody of 12 Spanish-speaking children and found that Spanish 2-year-olds had acquired all elicited patterns (S, Sw, wS, wSw, Sww, wwS, SwSw and SwSww), unlike their Catalan and English counterparts. The children nevertheless still truncated 13% of polysyllabic words, while 4-year-olds truncated around 1% and English 2-year-olds around 26%.   Relative to stress, Spanish-speaking children often master “regular” stress patterns by 3, but may find “irregular” words difficult up to at least age 5 (Hochberg, 1988). Ponce (2014) found a TD stress match of 96.9%, with the lowest category being 5-syllable words at 80%.  In Pérez (2013), patterns wSww and wwSw had stress matches of 75% and 77.1% respectively for children with PPD, compared with a TD score of over 90%. In her study, however, disyllabic and monosyllabic stress patterns were generally well-matched. She found deletion of unstressed syllables and reduplication to be indicators of risk. Bernhardt et al. (2015a) similarly found that 4-year-olds with PPD were more likely than TD peers to have low accuracy in multisyllabic words, with syllable deletion occurring in 3-5 year olds with PPD in 42% of 4-5 syllable words.   22 Epenthesis and stress shift were both rare. Syllable deletions generally preserved or created left-prominent words. Chávez-Peón et al. (2012) found that stress was “relatively well-preserved” in their sample of two 4-year-old boys with PPD.  Concerning possible stress and word length mismatch types, Schretlen (2013) found syllable deletion to be the most common. It did not occur in the TD group except in the 3-year-olds’ 4-5 syllable words. Syllable deletion decreased with age, and no child (with PPD or TD) used syllable deletion by age 5.  1.4.2.1   Consonant clusters: timing unit match Timing unit match is analogous to word shape match: it is the match of produced CV shape to the target. Consonant cluster timing unit match is therefore the match of CC or CCC. It is an important measure because it gauges how well a child matches the structure of clusters, whether or not they match the exact consonants. As such, it is always equal to or greater than full segmental match.  What constitutes a consonant cluster is sometimes unclear in Spanish phonological acquisition. Commonly, glide-vowel (/j/ or /w/ plus a vowel) sequences have been transcribed as their Spanish orthography suggests: as vowels, and therefore as a rising diphthong. Kehoe, Hilaire-Debove, Demuth and Lleó’s (2008) study examining French- and Spanish-speaking 1- to 3-year-olds, however, suggests that these sequences act more like branching onsets in early acquisition. In adult Spanish, phonotactics suggest that GV sequences act like diphthongs, because glides are   23 able to follow nasals, homorganic consonants, consonant-liquid combinations, and /s/ (Kehoe et al., 2008). The transcriptions in this thesis classified /j/ and /w/ as glides (rather than as vowels /i/ and /u/). The analysis both included and excluded /j/ and /w/ from the consonant class, because they are indeed different from clusters with liquids, especially, as will be seen in Emilio’s case.  In a similar vein, following the initial Spanish phoneticians’ practice, geminates are transcribed as CC rather than C: (e.g., pierna is transcribed /ˈpjenna/ rather than /ˈpjen:a/). This would technically place such sequences into the CC category of what have been called “consonant clusters”, but I have noted them where they occur in the analysis so as not to obscure the “true” consonant cluster patterns. The only other word with a target geminate sequence is peine (/ˈpenne/), which occurs twice in Emilio’s Time 1 data.  In Emilio’s cohort, comprising TD and PPD, the 3- and 4-year-olds had word-initial CC shape accuracies of 40.9% and 89.2% (TD), and 15.2% and 30.1% (PPD), with much higher accuracy in left-prominent words (Bernhardt et al., 2015a). Pérez (2013) found that word-initial cluster timing unit matches were 27.7% for children with PPD compared with 88.9% in TD children.  Word-medial cluster timing unit match scores were 12.8% for children with PPD compared with 66.7% in TD children, but these included codas that were deleted from targets (Pérez, 2013). Bernhardt et al. (2015a) found a difference between heterosyllabic clusters and tautosyllabic   24 onset clusters. They found that TD 3- and 4-year-olds had 77.8% and 100% accuracy in heterosyllabic clusters, compared with 61.9% and 81% for their peers with PPD, and 30.6% and 77.5% (TD) in tautosyllabic onsets, compared with 10.7% and 23.2% (PPD).  1.4.2.2   Consonant clusters: full segmental match Full segmental match in clusters is also a key measure of acquisition, analogous to Whole Word Match (WWM). It is less sensitive to structural development, but more sensitive to later stages of acquisition, when timing unit match may reach 100% but exact consonants do not match their targets.  Match levels for consonant-liquid clusters have had various results. Bosch (2004) found an 11.3% error rate (88.7% match) in TD 4-year-olds, whereas Pérez (2013) found that word-initial consonant cluster full segmental match was 85.6% for the whole TD sample (aged 3 to 5 years) and 19.7% for the PPD group. Also in Pérez (2013), word-medial consonant cluster match was 4.2% for PPD compared with 66.7% for TD. In word-initial position, children with PPD were more likely to reduce clusters to one of the consonants (46.1%, cf. 7.8% for TD), replace both with a third consonant (22.2% cf. 1.1%), or replace one with another consonant (11.1% cf. 3.3%), but similarly unlikely to epenthesize a vowel to remain faithful to both consonants (1.1% cf. 2.2%; Pérez, 2013). Ponce (2014) found TD 4-year-olds had 82.9% accuracy for /Cl/ sequences. The /Cl/ sequence with the lowest accuracy was /gl/ at 60%. In contrast, Ponce found /Cɾ/ to have an average accuracy of 46.3%; lowest /Cɾ/ accuracy was for /gɾ/ and /pɾ/ at 35%, /dɾ/ at 36.7% and /tɾ/ at 43.3%.   25  Diez-Itza and Martínez (2004) studied phonological processes affecting consonant clusters in 240 Spanish-speaking children (ages 3;0-5;11). They found that all processes declined with age, that heterosyllabic clusters encountered more reduction than tautosyllabic ones until 5;0 and that deletion of one consonant was the most common process. Substitution of one consonant was another (less) common process, and whole-cluster deletion, coalescence and epenthesis were present but rare.  In examining mismatch patterns, Barlow (2003a) found that clusters were reduced to the least sonorous segment (in this case, the first) with occasional well-defined exceptions, depending on the child (sC for headedness reasons in English and dorsal-liquid, she claims, for markedness reasons). Stop-liquid or stop-glide clusters, therefore, should be reduced, if necessary, to the stop for labial and coronal stops, but to the liquid or glide for dorsal stops (see also Pater & Barlow, 2003). With /l/ being an earlier-acquired sound in Spanish (and therefore less marked), one might expect to see this more frequently in Spanish with dorsal-lateral onset clusters. However, the approximant [ɣ], which may replace /g/ even word-initially, may be more sonorous than /ɾ/ and /l/ in any case, and therefore reduction to the liquid would not be an exception to Barlow’s (2003a) general rule. Supporting this analysis, Gómez Fernández (1998), in a study of 104 Spanish-speaking children from Sevilla (ages 1 to 6), found that the first segment generally survived more often than the second. He also found that children tended to begin by producing neither consonant (cluster deletion), then one , then two with at least one substitution and finally, the on-target cluster (Gómez Fernández, 1998). Vivar (2009) studied complex onsets in 72 TD Chilean children (ages 3-5) and found that omission of the second element was the most   26 common pattern, occurring more frequently for /ɾ/ than /l/. Similarly, Bosch (2004) found that /Cɾ/ clusters were typically acquired with 90% accuracy only by age 6, while /Cl/ clusters reached 90% accuracy by age 4.  1.4.2.3   Word-final consonants In Granada Spanish, codas are often optional, a difference from dialects such as Castilian Spanish. Nevertheless, some codas remain, especially nasals. In this dataset, I have retained only the targets which Emilio produces word-finally, /n/ and /l/. While this restricted set of codas is less comparable to the literature than the full complement of codas found in other dialects, it is worth briefly reviewing their findings.  Bosch (2004) found total deletion of codas in 4-year-old Northern Central Spanish speakers to be indicative of delay. Chávez-Peón et al. (2012) observed that codas were better preserved than onsets in their two participants with PPD (Argentinian and Mexican Spanish), onsets being especially vulnerable in unstressed word-initial position.  Relative to word length and codas, Borràs-Comes and Prieto (2014) compared Catalan and Castilian-speaking children (8 Spanish-speaking 2-year-olds); no word length effect was found for either group between stressed syllables in monosyllables and iambs. Spanish speakers did have a positional effect, with greater accuracy in word-final position (stressed and unstressed) than word-medial position. Lleó (2003), studying two monolingual Spanish children, posited the   27 same position effect, but found rather that word-medial codas were more accurate, which she attributed to stress in the common trochee pattern.  1.4.2.4   Diphthongs Diphthongs are the vocalic equivalent of consonant clusters, and are divided into rising and falling, relative to sonority. For example, /ai/ is a falling diphthong as /a/ is more sonorous than /i/, whereas /ja/ is a rising diphthong as /a/ is more sonorous than /j/. In Emilio’s cohort (Bernhardt et al., 2015c), six 4-year-olds with PPD had an average VV timing unit match of 59% and a full segmental match of 51.2% (Bernhardt & Stemberger, in press). Pérez (2013), for the same cohort, found rising diphthongs to be largely accurate for both TD (98.3%) children and those with PPD (88.1%), but falling diphthongs to have much lower accuracy for children with PPD: 54.6% accuracy compared with 92.2% for their TD peers. Falling diphthongs also had lower accuracy (74.5%) in the Chilean TD cohort of Ponce (2014), especially falling diphthongs with front vowels (/ei/ at 70% and /eu/ at 30%), with /ai/ at 75%. In Pérez (2013), the most common pattern relative to falling diphthongs was deletion of the second vowel: 30.8% in PPD; 3.4% in TD. Bosch (2004) found the opposite pattern relative to rising and falling diphthongs: 90% accuracy in falling diphthongs by age 3, but the same for rising diphthongs only by age 4. This difference is unexplained in the literature, but a key difference between the two studies may be dialect, with Pérez’s participants being from Granada and those of Bosch from central Spain. While both dialects include falling and rising diphthongs, they differ on coda deletion, which may contribute to differences in acquisition of rime structure. In any case, Bosch concluded that   28 simplification of both rising and falling diphthongs was an indicator of phonological difficulty at age 4.  Relative to vowel sequences, however, it is important to clarify that reduction can occur even in adult speech (RAE, 2011) as the examples below show. Thus, some of the children's patterns may be reflecting adult variation.  (1) ei/ > /e/ or /i/ in numbers, e.g. veintidós /bein.̪ti.ˈðos/ → {[ben.̪ti.ˈðos]/[βin.̪ti.ˈðos]}  (2) /au/ → [a], [o] or [u], e.g. precaución /pɾe.kau.ˈθio̯n/ → [pɾe.ka.ˈθio̯n]) 'precaution'  (3) /eu/ → [u], e.g. Europa /eu.ˈɾo.pa/ → [u.ˈɾo.pa] 'Europe'  (4) /ue/ → [e] or [o], e.g. luego /ˈlue.go/ → [ˈlo.go] 'then'  (5) /a.o/ → [o] (e.g. zanahoria /{θ/s}a.ˈna.o.ɾia̪/ → [{θ/s}a.ˈno.ɾia̪] 'carrot'  (6) /a.e/ → [e] (e.g. maestro /ˈma.es.tro/ → [ˈmes.tro] 'teacher'.  (7) Dissimilation with deletion: ciencia /θi.ˈen.θia/ → [ˈθen.θia].  1.4.3   Acquisition of segments By 4 years of age, most Spanish segments are expected to be “acquired”, a word defined variously in the literature (Acevedo, 1993; Bosch, 2004; González, 1981; Jímenez, 1987; Raymond, 2017). Vowels, especially, are expected to have high match rates.    29 1.4.3.1   Singleton vowels Goldstein and Pollock (2000) studied twenty-three 3- and 4-year-old Puerto-Rican children with speech sound disorders (SSDs) and found fewer than 1% mismatches. Most occurred for /o/, and some for /e/, with vowel lowering being most common. The authors suggested that vowel harmony may have played a larger role in those mismatches than typically occurs in English children with PPD.   1.4.3.2   Singleton consonants González (1981) studied consonant production in 150 TD Puerto Rican children (ages 2;6-3;0, 3;6-4;0 and 4;6-5;0), concluding that all phones were “customarily used” by 4, but only the following were “mastered” (90% correct or greater in all word positions tested by 75% of the age group): /ɲ/, /ʝ/, /k/, /g/, /t/, /n/, /m/, /f/.  Further to 4-year-olds, Jímenez (1987) studied acquisition of phones by 120 Mexican Spanish-speaking children: all phones other than fricatives /s/, /x/, nasal /ɲ/ and trilled /r/ were acquired by age 4;7. Acevedo (1993), in a study of 120 Mexican-American children ages 3;0-5;11, found that by 4;6 all phones including vibrants had an accuracy rate of 90% or greater except fricatives [ɣ] (70% accuracy), /x/ and glide /j/ (85% word initially each). “Mastery” level, however, was defined as achieving 90% or greater in two consecutive age groups, and two phones did not achieve that by 5;6: trilled /r/ and voiced velar 'stop' /g/ (including [ɣ]).     30 For Castilian Spanish, Bosch (2004), using a 90% accuracy criterion, observed a general order of acquisition of consonants as: by 3 years, nasals /m/, /n/, /ŋ/, stops /p/, /t/, /k/, /b/, fricative /x/, and liquid /l/; by 4 years, stops /d/, /g/, fricative/affricate /f/, /t ͡ʃ/, and liquid /ɾ/; by 6 years, coronal fricatives /s/, /θ/; and by 7 years, trilled /r/.  Turning to the Granada Spanish cohort for Emilio, Raymond (2017) divided the consonants by three accuracy levels: “still developing” (69% or less), “near mastery” (70-85%) and “mastered” (86-100%), for each of three TD age groups (3-, 4-, and 5-year-olds). By age 4, all phones were either “mastered” or “near-mastery”, similar to the data above. Word initially, consonants at "near mastery" were  trilled /r/ and coronal fricative /s/ (in left-prominent words and for /r/, in right-prominent words), and stops /b/ and /g/ (in centre-prominent words). Word medially, “near-mastery” consonants included trilled /r/, and fricatives /θ/, /s/ and [ɣ] in left-prominent words, the latter two also in centre-prominent words and [ɣ] in right-prominent words.  Clearly some variability is to be expected between dialects and/or individuals, but /r/ tends to develop later than other consonants, and is the only one consistently identified as not yet acquired or mastered by age 4.    31 1.4.3.3   Singleton consonants by word position The literature is somewhat contradictory regarding consonant singleton development by word position. González (1981) reports that TD Puerto Rican children (ages 2;6-5;0) had greater accuracy in word-medial consonants than in word-initial. For Mexican-American children, Avecedo (1993) reports that most children aged 4;6 showed consonants within 5% variance in accuracy across word positions, although word-medial [ɣ] was less accurate than word-initial /g/, and word-initial /x/ and /j/ were less accurate than their word-medial counterparts. For the Granada Spanish-speaking children with PPD, Pérez (2013) reported that word-initial liquids had higher accuracy than word-medial and word-final counterparts. For their two children (Argentinian, Mexican) with PPD, Chávez-Peón et al. (2012) found that codas were better preserved than onsets, and that onsets were especially vulnerable in unstressed word-initial position.  1.4.4   Acquisition of features and feature combinations Order of acquisition is not limited to word structure or segments: features can be described similarly. Dinnsen, Chin, Elbert and Powell (1990) posited a typology based on their analysis of 40 English-speaking “functional misarticulators” (ages 3;6-6;10), noting that their proposed pattern generally aligns with that of typical development, providing evidence for “delayed” acquisition, rather than fundamentally different development (see below). Their stages referred to manner and laryngeal features only and referenced Chomsky and Halle (1968). Cataño et al. (2009) took 39 cross-sectional and longitudinal phonetic inventories from 16 TD Spanish-speaking children from a variety of sources and found that the universal typologies for   32 phonological acquisition proposed by Dinnsen et al. (1990), and earlier by Jakobson (1968), were generally upheld in Spanish. The one major exception was that /l/ of stage D was acquired much earlier in the typology, perhaps due to it lacking the [Dorsal] feature of the English /l/. This occurred then in stage A, requiring a [+nasal] or [+lateral] distinction early on. Modified for Spanish acquisition, the following could be posited for the typology proposed by Dinnsen et al.(1990) for English. A [+/-consonantal], [+/-sonorant], [+/-coronal] & D: [+/-nasal] or [+/-lateral] B [+/-voice], [+/-anterior] C [+/-continuant] and/or [+/-delayed release] E [+/-strident] and/or [+/-lateral]   The following sections focus on three of the more marked combinations of features: dorsal consonants, fricatives, and liquids. Fricatives and liquids tend to be acquired later than other manners of articulation (Bernhardt, Másdóttir, Stemberger, Leonhardt & Hansson, 2015b; Cataño et al., 2009). Dorsals also tend to be acquired later than other consonants cross-linguistically, and some of Emilio’s cohort (including himself) were noted to have more difficulty with dorsals than coronals or labials.  1.4.4.1   [Dorsal, +consonantal] (dorsal consonants) Dorsal consonants have the features [Dorsal] and [+consonantal]. Spanish has five dorsals: /k/, /g/, /x/, [ɣ] and the nasal /ŋ/. (Palatals /ɲ/ and /ʝ/ also have a Dorsal component.) The dorsals can be distinguished from each other with the features [+/-voiced], [+/-continuant] and [+/-nasal].    33 Jímenez (1987) found her Mexican Spanish group to have 90% accuracy for /k/ by 3;7, for /g/ by 4;7 and /x/ by 4;11. Bosch (2004) considered the use of coronals for dorsals a possible indicator of delay in Northern Central Spanish speakers at 4 years of age. Of the six phones still unmastered at age 4 in Emilio’s TD cohort, two were voiced dorsals: [g] (word-initial, unstressed) and [ɣ] (word-medial; Raymond, 2017). As these were only “near-mastery” (i.e., 70-85%) by the TD children in the Granada cohort, one would not expect them to be mastered by 4-year-olds with PPD such as Emilio.  1.4.4.2   [-sonorant, +continuant] (fricatives) Fricatives are uniquely [-sonorant] and [+continuant]. Granada Spanish has three (alternately, four) voiceless fricatives: /f/, /x/ and the coronals /s/ and /θ/. Spanish also has three continuant allophones of the voiced stops, which are sometimes described as fricatives and sometimes as approximants: [β], [ð], and [ɣ] (cf. Eblen, 1982). This study focuses on the voiceless fricatives, with brief comparison to their voiced counterparts. When referring to fricatives, voiceless ones are indicated unless otherwise specified.  Research on first language acquisition of Spanish fricatives has focused on 2- to 5-year-old children. Fricatives appear to be developmentally later than oral and nasal stops (Lleó, 2008). Vivar and León (2009) found /s/ and /x/ to be particularly difficult for Chilean 3-5 year olds, only surpassed by /r/. In Emilio’s cohort, of the six phones left unmastered at age 4 years, three   34 were fricatives (two true fricatives): /s/, /θ/ and [ɣ] (Raymond, 2017). Eblen (1982) notes that substitution patterns for fricatives in Mexican Spanish, used by six 3-year-olds, included those common (and acceptable) to other dialects in the region (e.g., /x/ > [h]) plus idiosyncratic ones (e.g., /s/ > [t ͡ʃ]).  The affricate /t ͡ʃ/ is typically acquired by age 4 (Bosch, 2004; Raymond, 2017). While affricates are not true fricatives in that they are both [+continuant] and [-continuant], they do end with frication. Both /t ͡ʃ/ and [dʒ͡] were analyzed for Emilio along with all the other phones, but were not considered with the fricatives in this section, nor compared with Sofía and Tomás’ productions, because they were not expected to be as difficult as the other segments described here (fricative, liquids and dorsal consonants).  1.4.4.3   [+sonorant, +consonantal] and [+lateral] or [+vibrant] (liquids) In Spanish, liquids include the vibrant and lateral sonorants, with features [+sonorant], [+consonantal] and either [+lateral] or [+vibrant]. Spanish has three liquids: /l/, /ɾ/ and /r/. The lateral liquid is /l/, while the vibrants are tap and trill (/ɾ/ and /r/). Vibrant refers to the complete occlusion of the airstream which is brief enough to allow the vocal folds to continue vibrating unimpeded by excess supraglottal air pressure. Tap refers to a single occlusion; trill refers to multiple, usually around three in word-initial position. The periods between occlusions in a trill are referred to as apertures.   35 The liquids tend to be typically acquired sequentially: /l/ first, between ages 2 and 4 years, the tap second between 3 and 6 years, and the trill between 5 and 7 years (Bosch, 1984; Cataño et al., 2009; González, 1989; Miras, 1992). Rhotics develop late cross-linguistically, possibly due to their motoric complexity, including increased demands on timing abilities (cf. Boyce, Hamilton & Rivera-Campos, 2016; Carballo & Mendoza, 2000).  Further comparing the lateral and rhotics, Pérez (2013) found singleton /l/ to be mostly accurate in TD (>90%) across word positions, but for children with PPD, 81.3%, 63.3% and 67.7% word- initially, -medially and -finally. The vibrant /ɾ/ was more difficult for those with PPD, with 47.2% accuracy compared with TD 96.2%. The trill /r/ was most difficult for both groups, with 39.6% and 30.3% word initially and medially for children with PPD and 79.2% and 80.6%, respectively, for their TD counterparts. Vivar and León (2009) also found /r/ to be the latest-acquired phoneme, followed by fricatives /s/ and /x/. Substitution was the most common pattern affecting liquids. Pérez (2013) found that some children with PPD velarized (backed) the trill /r/, and substituted [l] for both vibrants. Ponce (2014) found that TD 4-year-olds had match proportions for /l/ of 84.5%, for /ɾ/ of 48.1% and for /r/ of 25%, with substitutions the most common mismatch patterns.  Further to the trill and the Granada cohort, Carballo and Mendoza (2000) examined trill productions of 45 other Spanish-speaking children from Granada. They compared three groups   36 of nine children between the ages of 3;0 and 6;6 (high, middle and low intelligibility) and two groups of nine children between the ages of 7;0 and 9;6 (incorrect pronunciation and control group). They compared various measurements of the production of /r/ in several words beginning with /ra/; the groups differed significantly as follows:  (1) higher formant 1 (F1) frequency (over 600Hz) in the younger groups than in the older groups (~518Hz). This is unsurprising considering that vocal tracts tend to increase in size with age during childhood.  (2) higher formant 2 (F2) frequency (2144.44Hz) in the low intelligibility group than the control group (1957.03Hz).  (3) lower amplitude of F1 in the high and low intelligibility groups (~36dB) than the control group (41.22dB).  (4) lower C-V ratio (difference in intensity between consonant and vowel) in the younger groups (around 4.4dB) than in the control group (8.39).  (5) greater consonant duration in the incorrect pronunciation group (163.50ms) than in the low intelligibility (100.33ms) and control groups (115.67).  (6) lower number of apertures and occlusions in the low intelligibility and incorrect pronunciations groups (~1.2 apertures and ~0.3 occlusions) than in the high intelligibility and control groups (~3.4 apertures and ~2.4 occlusions).  1.4.5   Mismatch patterns during acquisition Table 1.3 (following page) shows mismatch patterns observed in several studies. In preschoolers, Goldstein and Iglesias (1998) noted that common phonological processes (>10% of the time) are   37 cluster reduction, liquid simplification and stopping. For those with PPD, they also found initial consonant deletion, weak syllable deletion and dorsal fronting. Less common processes exhibited by both groups were palatal fronting (/j/>[n]), assimilation and final consonant deletion. For TD Spanish-speaking children, Goldstein (2005) found that [ð], [β], /r/, /ɾ/, [ɣ], /t/, /t ͡ʃ/, and /s/ were the most common singletons showing substitutions (at over 5% each), and the most common substitutes were [ɾ], [b], [l], [g], [k], and [ʃ]. Some rare patterns in both TD and PPD groups were de-affrication, backing, spirantization and denasalization. Some processes only seen in TD children were addition and palatization. Those only found in PPD were lisping, nasalization and spirant deletion. Pérez (2013) found the following to be associated with PPD at 4 years: substitution of a liquid for [ð], fronting of dorsal consonants, and the stopping of fricatives. Process Commonly observed Sources Deletion Codas, clusters, liquids, fricatives Heterosyllabic clusters Clusters Goldstein & Cintrón (2001) Diez-Itza and Martínez (2004) Goldstein & Iglesias (1998) Epenthesis Consonant clusters González (1981) Substitution One consonant for a cluster (less common than deletion) Liquids, fricatives Diez-Itza & Martínez (2004)  Goldstein & Iglesias (1998) Assimilation Regressive, non-contiguous, interconsonantal Coronal assimilation Martínez & Diez-Itza (2012) Ponce (2014) Substituted singletons [ð], [β], [r], [ɾ], [ɣ], [t], [t ͡ʃ], [s] Goldstein (2005) Substitutions [ɾ], [b], [l], [g], [k], [ʃ] Goldstein (2005) Table 1.3 Tendencies in phonological mismatch patterns  Relative to assimilation, Ponce (2014) noted that TD 4-year-olds showed some instances of assimilation: 6.4% of the targets showed coronal place assimilation, 2.5% total assimilation and 1.9% dorsal assimilation. Coda aspiration was present 8.2% of the time, substitution of a liquid   38 by a non-liquid 5.6%, other substitution 4.8%, gliding 3.6%, and less than 2% of fronting, labializing, stopping, spirantizing and others. Martínez and Diez-Itza (2012) compared frequency of assimilation processes in a naturalistic corpus study of 240 children (ages 3-6) and found that progressive, contiguous and intervocalic assimilations were less common and regressive, non-contiguous and interconsonantal assimilations were more common and persistent. Examples of metathesis have been noted (but not quantified) in some studies with children both with PPD and TD (Bernhardt et al., 2015c; Chávez-Peon, 2012; Pérez, 2013; Ponce, 2014; Schretlen, 2013).  Spanish-learning children seem to exhibit some similar and some different processes to children learning other languages. Goldstein & Cintrón (2001) compared three 2-year-olds learning Puerto Rican Spanish to others learning English, Cantonese and Igbo. Cross-linguistically, children tended towards CV-type syllables, produced relatively few syllable codas and onset clusters, and deleted liquids, fricatives and clusters. Unique to Spanish were a greater inventory of onset consonants (perhaps due to a relatively small number of codas in the dialect - a commonality with Granada Spanish), and a greater number of bisyllabic and polysyllabic words. Most striking was a lack of substitution for /l/ compared with other languages: Spanish children acquired it earlier. González (1981) also found that most patterns in Puerto Rican Spanish acquisition were similar to those of English acquisition, but noted that epenthesis in consonant clusters was more common in her Spanish sample.    39 1.4.6   Expected variables of relevance at age 4;8 Table 1.4 outlines sixteen general expectations for 4-year-old Spanish-speaking children with PPD, based on the literature reviewed above. Measure Predicted Outcome Sources 1. Whole Word Match (WWM) 10-40% Chávez-Peón et al. (2012), Pérez (2013) 2. Percent Consonants Correct (PCC) 50-65% Burrows & Goldstein (2010), Chávez-Peón et al. (2012) 3. Phonological Mean Length of Utterance (PMLU) 6.1-7.0 Burrows & Goldstein (2010), Dubasik & Ingram (2013) 4. Proportion of Whole-Word Proximity (PWP) .7-.85 Burrows & Goldstein (2010), Dubasik & Ingram (2013) 5. Overall Stress Match 90-100% Bernhardt et al. (2015a), Chávez-Peón et al. (2012), Pérez (2013) 6. Stress Match (3+ syllables) 70-90% Pérez (2013), Schretlen (2013) 7. Word Shape Match (WSM) 55-65% Bernhardt et al. (2015a), Pérez (2013) 8. WSM: words with CC 30-40% Bernhardt et al. (2015a), Pérez (2013) 9. CC Timing Unit Match: Word-Initial, Left-Prominent 35-45% Bernhardt et al. (2015a), Pérez (2013) 10. CC Timing Unit Match: Word-Initial, Centre-/Right-Prominent 10-20% Bernhardt et al. (2015a), Pérez (2013) 11. CC Full Segmental Match 15-35% Pérez (2013) 12. Rising diphthong: Full Segmental Match 75-85% Pérez (2013) 13. Falling diphthong: Full Segmental Match 45-55% Pérez (2013) 14. Coda: Full Segmental Match 65-80% Chávez-Peón et al. (2012), Pérez (2013) 15. Phone with lowest accuracy /r/ Boyce et al. (2016), Vivar & León (2009), Pérez (2013) 16. Singleton vowels: Full Segmental Match 90-100% Goldstein & Pollock (2000, 2004) Table 1.4 Expected performance for 4-year-old Granada Spanish speakers with PPD  The expectations in Table 1.4 (previous page) focus on time 1, at age 4;8. Among the children with PPD in Emilio’s cohort, he was the only one with continuing difficulties at age 7;6, and therefore the expected outcomes for time 2 are a WWM of 100% (and therefore all other   40 measures at 100% too). They are based on three features of his profile: age, dialect and being designated with PPD. Some research suggests that gender has no significant impact on typical Spanish phonological development (Mercedes, Maggiolo, Penaloza & Julia, 2009), although it has been found (in English) to make the incidence of PPD more likely (Campbell et al., 2003). Mercedes et al. (2009) did find that socio-economic status (SES) has an impact, but Emilio’s SES is unknown.  For the current study, Emilio’s values on these measures are compared with those for two children from his cohort, Sofía and Tomás, who at time 1 had similar ages and WWM scores. For reference to other papers that analyze the Granada cohort’s data, Emilio’s original identification number was 316; Sofía and Tomás were 308 and 314, respectively.   41 Chapter 2:  Method This section outlines the participants, procedures and analysis for the study. Background information from the larger study of 59 children in Granada is included.  2.1   Participants Emilio is a boy from Granada, Spain with protracted phonological development (PPD). He participated in Bernhardt et al.’s (2015a) study comparing 30 TD children with 29 children with PPD (ages 3-5 years). Fifteen children with PPD were re-assessed three years after the initial assessment. The chosen participant, Emilio, was the only one with lingering phonological difficulties at the second time period, and so was chosen to be looked into further. He had been receiving speech therapy focusing on trilled /r/. The two samples were taken at ages 4;8 and 7;6. Other than PPD, he had no identified health, hearing or developmental concerns. In the original study, he was screened for hearing at 25 dB from 250 to 4000 Hz, had an oral mechanism screening and language comprehension and production assessment using the Test de Comprensión de Estructuras Gramaticales de 2 a 4 años (Calet, Mendoza, Carballo, Fresneda & Muñoz, 2010), the Peabody Picture Vocabulary Test - Español (Dunn, Dunn & Arribas, 2006) and the Test Breve de Inteligencia de Kaufman (Kaufman & Kaufman, 2009: Spanish adaptation; Cordero & Calonge, 2000). Emilio was assigned to the PPD group based on (1) the phonology subsection of the PLON-R and (2) a conversation sample. PLON-R (La Prueba de Lenguaje Oral Navarra - Revisada, Aguinaga, Armentia, Fraile, Olangua & Uriz, 2004) contains 23 words for 4-year-olds and focuses on segmental acquisition by age. Patterns in the conversation sample were compared with those reported in Bosch (2004) and identified as typical or atypical   42 accordingly. His assignment was then confirmed using the global measure whole word match (WWM) on a 103-word sample designed for the study (see Appendix A for the word list).  Sofía and Tomás were recruited with the same selection process and also had no identified developmental or health concerns other than PPD. At time 1 they were aged 4;3 and 4;10, respectively; the second data collection took place three years later. Their data were also transcribed and double-checked like Emilio’s, as is described in the following section.  2.2   Procedures As in Bernhardt et al. (2015a), single words were digitally recorded by a native Spanish speaker using a 103-word list (Appendix A). In accordance with Spanish word structure proportions, the list has a high frequency of left-prominent disyllabic words and fewer monosyllables and 4-5 syllable words. The words were recorded with a Microtrack II recorder and microphone. Words were elicited using pictures (¿Que es esto?), with spontaneous productions being the aim, but imitation accepted when words were not produced spontaneously.1 At time 1, Emilio's data included 26 imitated words (about one quarter of the words); no imitations were noted in time 2). The words were then transcribed once by native Spanish speakers, with confirmation by a second (North American) transcriber team and acoustic analysis.                                                  1 Goldstein, Fabiano and Iglesias (2004) observed that spontaneous and imitated productions in Spanish-speaking children with phonological disorders were identical 62% of the time, spontaneous more adult-like 25% of the time, and imitated 13% of the time. They recommend that imitated samples be included for additional diagnostic and prognostic value.   43 The two transcription teams (North America and Granada) met to develop conventions for transcription, and they agreed upon a level of narrowness to match the larger study comparing 12 languages (Bernhardt & Stemberger, 2012). The two teams transcribed the first 12 PPD data sets, discussed and agreed upon consensus transcription conventions. Native Granada Spanish speakers transcribed the remainder, and the North American team confirmed the transcriptions perceptually and using waveform and spectrogram analysis where ambiguity occurred. Emilio’s transcriptions were highly reliable, with 80% of words (88 of 110) having no differences in transcription, and 94.9% of segments with no differences (including consonants and vowels). Any disagreements were resolved in a final meeting where perceptual or conventional differences were agreed to be resolved by closer analysis of the acoustic signal. These differences primarily entailed different opinions regarding /ɾ/ versus /l/ and frication versus stop for the voiced bilabial, plus some differences in perception of stop place of articulation.  2.3   Analysis The program Phon 2.2 (Rose et al., 2006; Rose & MacWhinney, 2014) and spreadsheets provided quantitative data for each of the two samples. For all analyses, adult targets were chosen that matched dialectal variations used by the child in each individual word. For example, flor can be pronounced with or without the final consonant, as /ˈfloɾ/ or /ˈflɔ/. Emilio produced this word twice, as [ˈfɔ] and [ˈfo], both of which show deletion of /l/ but only one of which would be considered to have deleted /ɾ/ ([ˈfo]), because the vowel did not undergo laxing and lowering. A variety of measures (PMLU, WWM, PCC, PVC, structural and segmental match and   44 mismatch patterns) compare his data to what is found in the literature. Additional analyses of cross-vowel sequences, imitation and other effects describe the two datasets in terms of variability and specificity. For example, for time 1, a Spanish scan analysis form (Bernhardt et al., 2016, pages 3 to 7) summarizes and analyzes patterns in his phonological output (Appendix B). For time 2, both vibrants and polysyllabic word production were still developing, and are discussed and compared with the phonetic findings in Carballo and Mendoza (2000). Spectrograms of the three words with “irregular” /r/ are provided in Appendix C, and were collected using Praat version 6.0.04 (Boersma, 2001).   45 Chapter 3:  Results Results describe Emilio's datasets at ages 4;8 and 7;6 with relevant comparisons of his data at time 1 with those of Sofía (age 7;2, originally 4;3) and Tomás (age 7;7, originally, 4;10).  3.1   Expectations and results Table 3.1 lays out the findings compared with expectations described in the introduction.  Measure Predicted Outcome Emilio (4;8) Sofía (4;3) Tomás (4;10) WWM 10-40% 26.6% 33.3% 43.1% PCC 50-65% 60.7% 67.8% 70.3% PMLU 6.1-7.0 6.4 7.27 6.86 PWP .7-.85 0.80 0.88 0.85 Overall Stress Match 90-100% 95.4% 94.4% 95.4% Stress Match  (3+ syllables) 70-90% 88.8% - - WSM 55-65% 53.2% 82.4% 73.4% WSM: words with CC 30-40% 28.0% 72.5% 44.9% CC Timing Unit Match: Word-Initial, Left-Prominent 35-45% 26.9% 80.8% 84.0% CC Timing Unit Match: Word-Initial, Centre-/Right-Prominent 10-20% 0.0% 0.0% 0.0% CC Full Segmental Match 15-35% 26.2% - - Rising diphthong: Full Segmental Match 75-85% 93.8% - - Falling diphthong: Full Segmental Match 45-55% 45.5% 90.9% 83.3% Coda: Full Segmental Match 65-80% 55.2% - - Phone with lowest accuracy /r/ /r/ /r/ /r/ Singleton vowels: Full Segmental Match 90-100% 95.5% - - Table 3.1 Expectations versus results at age 4 (blue = within predicted range, green = above same, yellow = below same.   46 3.2   Descriptive findings: age 4;8 The following sections lay out Emilio’s results at age 4;8, compared when relevant/possible with data from Sofía and Tomás. The data follow the same general outline as the description of phonological acquisition of Spanish in the introduction.  3.2.1   Global measures At age 4;8, Emilio’s dataset had a whole-word match (WWM) of 26.6% (29/109 words exactly matching their target). His was the lowest WWM score at time 1 of the eight children in the longitudinal sample who were 4-year-olds at time 1 (average WWM at time 1 of 45%, SD 97%). Comparison participants Sofía and Tomás had WWMs of 33.3% and 43.1% respectively, each slightly higher than that of Emilio, but closest to his scores in the cohort of 4-year-olds in the longitudinal sample of eight children. Both Emilio and Sofía had WWM scores slightly lower than the average of 37.5% (SD 13.5%) of the whole Granada cohort of 4-year-olds with PPD at time 1, but within one standard deviation of that group (Bernhardt et al., 2015a).  Emilio’s consonants correct (PCC) was 60.7% (185/305). Comparison participants Sofía and Tomás each had PCCs of 67.8% and 70.3%, respectively, again higher than Emilio’s. Emilio’s PCC varied with word length, but was higher in shorter words: 1-syllable words had a PCC of 66.6%, 2-syllable words 62.8%, 3-syllable words 54.3% and 4- to 5-syllable words 59.5%. Emilio’s percent vowels correct (PVC) was much higher than consonants at 93.2% (246/264). Sofía and Tomás scored 95.9% and 97.4%, respectively.    47 At age 4;8, target words in Emilio's sample had an average phonological mean length of utterance (PMLU) of 8.0; Emilio's own productions had an average PMLU of 6.4. A comparison of those two numbers yields a proportion of whole-word proximity (PWP) of .80. Sofía and Tomás’ target PMLUs were 8.27 and 8.07 respectively, with actual PMLUs 7.27 and 6.86.   As for Schretlen’s (2013) MSW (multisyllabic word) rubric, Emilio had a total of 2, 16, and 11 structure mismatch points for 2-, 3- and 4-syllable words. These are comparable to the 4-year-old group’s average of 1.75, 16.6 and 10.7. His word-structure/segment interaction points are also comparable to the group’s averages, although slightly higher than his structure points: 2, 10 and 6.5 points compared with the averages of .79, 7.07 and 3.46. Sofía and Tomás, by comparison, had structure points of 2, 6 and 9 (Sofía) and 1, 19 and 4 (Tomás). They had interaction points of 0, 6.5 and 1 (Sofía) and 0.5, 6.5 and 1 (Tomás). Emilio’s mismatch points were therefore not substantially higher than his cohort’s averages, nor Sofía’s or Tomás’ scores.  3.2.2   Word structure Word structure incorporates word length, stress match, word shape match and CV sequences including consonant clusters (CC) and diphthongs (VV). Word length was an overall strength for Emilio, with a general match for all but one word length: 5 syllables (wwSww).  Stress match was also high, at 95.4% (104/109 words having the correct sequence of stressed/unstressed syllables). Stress mismatches therefore included instances of syllable addition or deletion even if the stressed syllable was otherwise correct. Sofía and Tomás had stress matches of 94.4% and 95.4%, respectively. Stress match in polysyllabic words (three or more   48 syllables) was 88.8% (32/36 words). The five stress mismatches (two syllable additions, two syllable deletions and one stress shift) were: (1) luz  /ˈlu/  > [aˈtʃ͡u]  ‘light’  (imitated)  Europa /euˈɾopa/ > [ˈeuopa]  ‘Europe’ (imitated)  triángulo /ˈtɾjaŋɣulo/  > [ˈtiː.ʔanuɾo]  ‘triangle’ (imitated)  elefante /eleˈfant̪e/  > [eˈfan]  ‘elephant’  hipopótamo /ipoˈpotamo/ > [poˈpotamo] ‘hippo’  Imitation may have played a role in these stress errors; the first syllable in [aˈtʃ͡u] may have been the elicitor’s article: la luz. Elefante and hipopótamo, on the other hand, may have suffered syllable deletion as a result of their length (and centre-prominence, cf. Europa). While the syllable creation in triángulo is possible in adult speech as well, its stress placement is highly irregular, not falling on one of the three final syllables as one would expect.  Word shape match (WSM) for Emilio was 53.2% (58/109 words matching their target in CV shape). Sofía and Tomás had higher WSMs of 82.4% and 73.4%, respectively. Emilio’s WSM decreased with the presence of clusters in the word, reaching 28.0% (14/50 words with clusters matching their target shape). Sofía and Tomás’ WSMs also decreased with the presence of clusters, but were still higher than Emilio’s: 72.5% and 44.9% respectively.  WSM measures accuracy of sequences of C (consonant) and V (vowel). The “glides” or “semivowels” /j/ and /w/, however, are somewhere in between consonants and glides. WSM was therefore also calculated in a second way, using a third category: G for glide. Emilio’s overall WSM, however, did not change (it remained 50.5%). Likewise, Tomás’ WSM remained the   49 same at 73.4%. For Sofía, however, calculating WSM with G as different from C, resulted in a drop in WSM from 82.4% to 61.1%, due to the large number of glides she substituted for liquids (e.g., blanco /ˈblaŋko/ > [ˈbjaŋko] ‘white’).  Certain syllable structures posed particular challenges for Emilio. Not counting the rising diphthong onsets /j/ and /w/, he had no onset clusters (i.e., with /l/ or tap). He had some codas (nasals, one [h] and one [l]). He never had two clusters in one word, nor did he ever have a cluster and a word-final consonant in a single word. Cumulative phonological complexity, therefore, may present unique challenges for Emilio’s productions.  3.2.2.1   Consonant clusters: timing unit match True consonant clusters, that is, two different [+consonantal] segments in sequence had an overall timing unit match of 28.9% (13/45). Sequences where one “consonant” was a glide had a timing unit match of 82.4% (14/17). There were also three instances of geminate targets, pierna, /ˈpjenna/ which Emilio produced as [ˈpjena] and two counts of peine, /ˈpenne/, which Emilio matched.  Word-initial clusters had a timing unit match of 26.9% (stressed) and 0% (unstressed), although the unstressed word-initial clusters were all /Cɾ/ targets, a combination never produced by Emilio. Word-initial “true” clusters had a timing unit match of 0% in both stressed and unstressed syllables. Sofía and Tomás had stressed word-initial cluster timing unit matches of   50 80.8% and 84.0%, respectively. Like Emilio, however, word-initial unstressed cluster timing unit match was 0%.  An alternative measure would be cluster reduction or the presence of deletion, which would allow for substitutions of G or V. In stressed, word-initial clusters, Emilio had a reduction rate of 94.1%, only not reducing triángulo (/ˈtɾjaŋɣulo/ > [ˈtiː.ʔanuɾo]), which had compensatory lengthening of the vowel [i]. By contrast, Sofía and Tomás had reduction rates in stressed, word-initial clusters of 11.8% and 17.6%.  Emilio’s onset clusters with a liquid for a second consonant (true clusters) had a timing unit match of 0% (0/23). Including glides in the consonant category, his had an overall timing unit match of 34.2% (13/38). In true clusters, Emilio’s repair strategy was usually to delete the second C unit; the two exceptions, where he deletes the first unit or substitute another phone entirely, are listed here: (2) cocodrilo /kokoˈðɾilo/  > [toloˈliɾo] ‘crocodile’  globos  /ˈgloβo/  > [ˈloβo]  ‘balloons’  The remaining onset clusters had a glide for a second consonant (or were alternatively singleton onsets with rising diphthongs where the onset deleted) and had a timing unit match of 86.7% (13/15). The only two mismatches with glides nevertheless had compensatory lengthening, which was not observed in the onset clusters with liquids: (3) suave  /ˈθwaβe/  >  [ˈwːaβe] ‘soft’  (imitated)  cuadro  /ˈkwaðɾo/  >  [ˈwːaðo] ‘painting’   51 As for clusters spanning two syllables (heterosyllabic clusters), there was again a distinction in performance between those with liquids and those with other sonorants (in the coda position). The overall timing unit match for heterosyllabic clusters was 47.8% (11/23), and the full segmental match was 34.8% (8/23). Clusters with a liquid coda had 14.3% (1/7), and with other sonorants (nasals and [h]) 68.8% (11/16). The one timing unit match with a liquid coda (target) and the five shape mismatches with other sonorant codas are listed here: (4) hermano /eɾˈmãno/  >  [enˈmãno] ‘brother’ (imitated)  elefante /eleˈfant̪e/  > [eˈfan]  ‘elephant’  triangulo /ˈtɾjaŋɣulo/  >  [ˈtiː.ʔanuɾo]  ‘triangle’ (imitated)  pierna  /ˈpjenna/  >  [ˈpjena] ‘leg’  hirviendo /iɾˈβjend̪o/  > [iˈβjen ̪ː o] ‘boiling’ (imitated)  pantalón /pant̪aˈlon/  >  [tataˈlon] ‘pants’  There was also one instance of compensatory lengthening with the deletion of a liquid coda: barco, /ˈbaɾko/ > [ˈbatto]. Elefante may only be difficult due to length, with one consonant cluster. The final four also have either two clusters each or a cluster and word-final consonant - combinations Emilio never produces in the same word.  The dataset included two words containing a cluster with three consonants (or rather two consonants and a glide, CCG). These were never produced, with timing unit match 0% (0/2), although one was realized as CG. The two words with target CCG shapes are hirviendo (/iɾˈβjend̪o/ > [iˈβjen ̪ː o], ‘boiling’) and triángulo (/ˈtɾjaŋɣulo/ > [ˈtiː.ʔanuɾo], ‘triangle’). In triángulo, it is important to note that a quadrisyllabic pronunciation is possible in adult speech, but with the tap and without the glottal stop, unlike Emilio’s production.   52 3.2.2.2   Consonant clusters: full segmental match True consonant clusters, that is, two [+consonantal] segments in sequence, had an overall full segmental match rate of 22.2% (10/45). Onset clusters had an overall full segmental match of 21.1% (8/38 clusters); again 0% excluding /j/ and /w/. Heterosyllabic clusters excluding /j/ and /w/, while above 0% in timing unit match, again fell to 0% for full segmental match. As CCG timing unit match was 0%, so was full segmental match.  3.2.2.3   Codas Emilio had a coda timing unit match of 72.4% (21/29 codas) and a full segmental match of 55.2% (16/29 codas). His word-medial codas had a timing unit match of 66.7% (16/24) and a full segmental match of 45.8% (11/24). His word-final codas had a timing unit match and a full segmental match of 100% (5/5). Because codas are largely optional in Granada Spanish, further analysis was not done.  3.2.2.4   Vowel sequences There were three types of vowel/glide sequences elicited in the dataset: falling diphthongs (nucleus, then high vowel), rising diphthongs (glide/high vowel, then nucleus) and vowel hiatus (two syllables’ nuclei with no consonant in between; V.V). Of these categories, Emilio had the most difficulty with falling diphthongs: he produced 45.5% (5/11) on target although the falling diphthongs /eu/ and /ei/ had 100% accuracy (2/2 and 1/1, respectively). Sofía and Tomás, however, had high falling diphthong matches, at 90.9% and 83.3% respectively. Emilio’s rising diphthongs were more accurate with 93.8% (15/16) on target, and vowel with hiatus were the   53 most accurate at 100% (3/3). The one rising diphthong mismatch was triángulo (/ˈtɾjaŋɣulo/ > [ˈtiː.ʔanuɾo], ‘triangle’), which begins with two consonants before the diphthong - a more challenging syllable structure than the other diphthong contexts. Interestingly, the features of both diphthong elements are preserved - simply divided into two syllables (as some adults do with hiatus rather than a glottal stop). His falling diphthongs productions and mismatch patterns are shown in Table 3.2.  Adult IPA Emilio IPA Phonological mismatch pattern /ˈpaula/ [ˈpaːɾa] V2 deletion, V1 lengthening /ˈaiɾe/ [ˈaðe] V2 deletion /ˈxaula/ [ˈxwala] V1-V2 metathesis /ˈpaula/ [ˈwaða] V1-V2 metathesis or V2 deletion and C1-C2 feature metathesis /baiˈland̪o/ [baˈland̪o] V2 deletion /ˈoiɣo/ [ˈoðo] V2 deletion or V2+C coalescence /dinoˈsauɾjo/ [dinuˈθauɾjo] match /ˈoi/ [ˈoi] match /ˈbeint̪e/ [ˈbeint̪e] match /euˈɾopa/ [ˈeuopa] match /ˈeuɾo/ [ˈeuˠo] match Table 3.2 Falling diphthong matches and mismatches  Further data for Emilio's diphthongs are presented in Table 3.3 (following page). with a comparison from Ponce (2014) for Chilean TD children (ages 3;0-5;11).      54 Participants Rising/Falling Front + /j/ Front + /w/ Back + /j/ Back + /w/ Emilio Rising /je/: 100% (2 of 2) /wi/: not elicited /we/: 100% (3 of 3) /ja/: 67% (2 of 3) /jo/: 100% (1 of 1) /wa/: 100% (5 of 5) Emilio Falling /ei/: 100% (1 of 1) /eu/: 100% (2 of 2) /ai/: 0% (0 of 2) /oi/: 50% (1 of 2) /au/: 25% (1 of 4) Ponce (2014) TD (ages 3;0-5;11) Rising /je/: 100% /wi/: 90% /we/: 100% /ja/: 90% /jo/: 100% /wa/: 92% Ponce (2014) TD (ages 3;0-5;11) Falling /ei/: 70% /eu/: 30% /ai/: 75% /oi/: 100% /au/: 90% Table 3.3 Emilio's diphthongs compared with Ponce (2014)  3.2.3   Segments At 4;8, Emilio used all Granada Spanish phones at least once except for /ŋ/, albeit the one usage of /r/ was a substitute for /ɾ/ and fronted. (See Table 3.4 on the following page, which lists segments by place of articulation in manner order of stop/fricative/affricate/sonorants.)   Overall PCC (percent consonants correct) was 60.7%, while singleton match (percent singleton consonants correct) was 63.1% (113/179 on target). By word position, these singletons had match rates of 60.1% (39/64) word-initially, 63.0% (68/108) word-medially and 85.7% (6/7) word-finally. Codas, whether singleton or in a cluster, had a full segmental match rate of 51.6% (16/31). Singleton vowels were largely correct, with a match rate of 95.5% (211/221 on target).      55  Place of Articulation Target Segment Overall % Match # Match Singleton % Singleton # Labial p 84.0% 21/25 84.2% 16/19  b 81.8% 9/11 71.4% 5/7  f 90.9% 10/11 80.0% 4/5  β 90.0% 9/10 87.5% 7/8  m 80.0% 8/10 75.0% 6/8  w 100% 9/9 100% 1/1 Coronal t 92.3% 24/26 93.3% 14/15  d 90.0% 9/10 100% 3/3  θ 70.6% 12/17 78.6% 11/14  ð 33.3% 1/3 0.0% 0/1  t ͡ʃ 85.7% 6/7 85.7% 6/7  d ͡ʒ 66.7% 2/3 100% 2/2  n 90.3% 28/31 92.3% 12/13  l 48.3% 14/29 61.9% 13/21  ɾ 12.1% 4/33 30.0% 3/10  r 0.0% 0/9 0.0% 0/8 Coronal-Dorsal ʝ 100% 5/5 100% 5/5  ɲ 100% 3/3 100% 3/3  j 87.5% 7/8 none none Dorsal k 11.1% 2/18 0.0% 0/11  g 20.0% 1/5 0.0% 0/2  x 12.5% 1/8 12.5% 1/8  ɣ 30.0% 3/10 14.3% 1/7  ŋ 0.0% 0/3 none none Laryngeal h 100% 1/1 none none Table 3.4 Overall and singleton matches by consonant    56 To roughly judge mastery using the threshold defined by Avecedo (1993) and Bosch (2004) of 90% accuracy, certain cells are highlighted in blue in Table 3.5 below.  Labial Coronal  Coronal-Dorsal Dorsal Laryngeal  Labial Labiodental [+anterior] [-anterior] Palatal Velar Glottal Plosives p b  t d   k g  Fricatives  β f  θ ð  ʝ  x ɣ (h) Affricates     t ͡ʃ d ͡ʒ   Nasals  m   n  ɲ  ŋ  Liquids    l ɾ r   Glides  w     j  w (h) Table 3.5 Singleton matches by consonant (green = 90-100%, blue = 75-89%, yellow = 50-74%, orange = 1-49%, red = 0%;  /j/, /ŋ/ and [h] were not elicited as singletons)  While singleton consonant match is higher than overall percent consonants correct, that pattern is unexpectedly reversed in the case of several segments: /b/, /m/, /f/, [ð], /k/, [g] and [ɣ] show higher accuracy in clusters than as singletons. In the cases of [ð], /k/ and [g], they go from 0% accuracy in singletons to over 10% accuracy overall, with one or two correct in cluster targets. Interestingly, /x/ is also produced with a glide, despite the target being a singleton, in jaula. (5) cruz  /ˈkɾu/   >  [ˈku]  ‘cross’  (imitated)  cuatro  /ˈkwatɾo/  > [ˈkwato] ‘four’  guante  /ˈgwant̪e/  >  [ˈgwant̪e] ‘glove’  jaula  /ˈxaula/ > [ˈxwala] ‘cage’  agua  /ˈaɣwa/  > [ˈaɣwa] ‘water’  cuadro  /ˈkwaðɾo/  > [ˈwːaðo] ‘painting’    57 In the case of jaula, the repair of its falling diphthong to a rising one may reflect both his preference for dorsal consonants in cluster position and his preference for rising over falling diphthong. (See Table 3.6 below.)  Place of Articulation Segment Word-Initial Match Word-Medial Match Word-Final Match Labial p 75% (9/12) 100% (7/7) -  b 66.7% (4/6) 100% (1/1) -  f 100% (2/2) 66.7% (2/3) -  β - 87.5% (7/8) -  m 66.7% (2/3) 80% (4/5) -  w 100% (1/1) - - Coronal t 100% (4/4) 90.9% (10/11) -  d 100% (3/3) - -  θ 50% (3/6) 100% (8/8) -  ð - 0% (0/1) -  t ͡ʃ 100% (3/3) 75% (3/4) -  d ͡ʒ 100% (2/2) - -  n 100% (2/2) 83.3% (5/6) 100% (5/5)  l 75% (3/4) 56.3% (9/16) 100% (1/1)  ɾ - 30% (3/10) -  r 0% (0/5) 0% (0/3) - Coronal-Dorsal ʝ - 100% (5/5) -  ɲ - 100% (3/3) - Dorsal k 0% (0/4) 0% (0/7) -  g 0% (0/3) - -  x 25% (1/4) 0% (0/4) -  ɣ - 16.7% (1/6) - Table 3.6 Singleton matches by word position   58 3.2.4   Features and feature combinations The following sections examine three difficult (marked) combinations of features: dorsal consonants ([Dorsal, +consonantal]), fricatives ([+continuant, -sonorant]), and liquids ([+sonorant, +consonantal, -nasal]).  3.2.4.1   [Dorsal, +consonantal] (dorsal consonants) Dorsal consonants have the features [Dorsal] and [+consonantal], and include /k/, /g/, /x/, [ɣ] and /ŋ/. As the overall and singleton match rates show above, dorsal consonants had low accuracy in Emilio’s productions at time 1: 15.9% (7/44). Singleton match rate was 7.1% (2/28), whereas dorsal consonants in clusters were comparatively more accurate at 31.3% (5/16; p=.08). Dorsal feature match was exactly the same as full segmental match at 15.9% (7/44), which follows from the observation that every non-match was fronted to either coronal or labial place.  In comparison, Sofía and Tomás had exact overall matches of 86.4% and 77.8% for dorsal consonants, much higher than Emilio’s 14.6%. For [Dorsal, +consonantal] feature matches, Sofía and Tomás’ scores rise to 97.7% and 95.6%, also notably higher than Emilio’s 17.1%. Emilio’s dorsal consonants, therefore, distinguish him both from typically developing 4-year-olds and from his peers, Sofía and Tomás.  3.2.4.2   [+continuant, -sonorant] (fricatives) Only voiceless fricatives were considered to be true fricatives, with features [+continuant] and [-sonorant] (cf. Eblen, 1982). Voiced approximants or fricatives were considered to be   59 [+sonorant], with unimpeded vocal fold vibration, as was the allophone [h], which has no supraglottal obstruction of airflow. The phones under consideration, therefore, were /f/, /θ/ and /x/, because Emilio never produced /s/ at time 1 (he used “ceceo”).  Emilio’s true fricatives had a full segmental match of 64.7% (22/34) and a manner feature match ([+continuant, -sonorant]) of 79.4% (27/34), so five of twelve mismatches maintained manner of articulation. As a sound class, therefore, fricatives were more accurate than dorsal consonants. Most of these segmental mismatches which retained manner of articulation were instances of dorsal fronting, as seen below in juguete. The only exception was teléfono, which showed coronal assimilation in place, likely due to a sequence constraint prohibiting a Coronal-Labial sequence in a long word. (6) juguete /xuˈɣete/ > [θuˈðete] ‘toy’  (imitated)  teléfono /teˈlefono/ > [teˈleθeno] ‘phone’  For the sake of comparison, Emilio’s voiced approximants [β], [ð], [ʝ] and [ɣ] had a full segmental match of 62.5% (15/24), and a manner feature match ([+voiced, +continuant, -lateral, -vibrant]) of 83.3%. The affricates /t ͡ʃ/ and /d ͡ʒ/, to make a further comparison, had an overall match of 80.0% (8/10).  Sofía and Tomás, on the other hand, differed notably between each other. Sofía’s true fricative full segmental match was 90.5% (38/42), and her manner feature match was 100% (42/42). Tomás, however, had a full segmental match and a manner feature match of 35.1% (13/37), with   60 stops for those fricatives with mismatches. Therefore Emilio’s fricative match, full segmental or manner feature, did not distinguish him from his two peers. All three differed in this regard (although he was more advanced than Tomás at time 1 in manner feature match).   3.2.4.3    [+sonorant, +consonantal] and [+lateral] or [+vibrant] (liquids) Liquids, like dorsal consonants, had a lower accuracy rate overall, with an overall full segmental match of 23.6% (17/72). Singleton match rate was higher than the overall rate, at 40% (16/40). Overall feature match was higher than full segmental match, at 38.9% (28/72), reflecting a tendency to substitute liquids for other (usually singleton) liquids, retaining general manner features.  The lateral /l/, however, also needs to be considered separately from the vibrants /r/ and /ɾ/ because it is typically acquired earlier in Spanish (Bosch, 2004). Indeed, overall match and singleton match rates were much higher for /l/ than /r/ and /ɾ/: the overall lateral match was 48.3% (14/29) and the singleton lateral match was 61.9% (13/21). This compared with a lower match rate for the vibrants overall of 7.0% (3/43) and for singletons of 15.8% (3/19), with only tap showing any matches. Overall manner feature match was slightly higher (at 11.9%) than overall full segmental match (7.0%).  Sofía and Tomás had lateral (/l/) full segmental match rates of 18.5% and 57.1%, respectively, compared with Emilio’s 48.3%, i.e., the three children differed overall in this regard. They had vibrant full segmental match rates of 11.1% and 37.8% respectively, compared with Emilio’s   61 7.0% (who was more like Sofía). Manner feature matches raised their scores to 20.0% and 48.9%, respectively, while Emilio’s remained 11.9%. Lateral match, therefore, did not distinguish Emilio from his peers, nor did vibrant full segmental match. Vibrant manner feature match, however, was more distinctive, with Emilio’s 11.9% being modestly removed from Sofía’s 20.0% and Tomás’ 48.9%.  3.2.5   Phonological mismatch patterns The following sections examine phonological mismatch patterns for Emilio’s consonants and compare their incidence with those of Sofía and Tomás. These include substitutions, phonetic differences, harmony, metathesis, migration, coalescence and epenthesis.  3.2.5.1   Singleton substitutions Singleton mismatches show various repairs, including substitutions which are outlined in Table 3.7 (following page). There is no column for word-final substitutions because there were only matches and deletions word-finally.  As Table 3.7 shows, word-initially and –medially, coronal is the most commonly produced place in substitutions. Word-initially, the voiceless oral stop [t] was the most common of these, replacing /k/ twice, /x/ twice and /g/ once (gitarra, /giˈtara/ > [tiˈtaŭla]). It also replaced labial segments /p/ and /m/, but only in assimilative contexts in pantalón (/pant̪aˈlon/ > [tataˈlon]) and martillo (/maɾˈtiʝo/ > [taˈtiʝo]).   62  Place of Articulation Target Segment Word-Initial Word-Medial Total Labial p w, f, t  15.8% (3/19 targets)  b m  14.3% (1/7)  β  ð 12.5% (1/8)  f  θ 20.0% (1/5)  m t ð 25.0% (2/8)  w   0.0% (0/1) Coronal t  p 6.7% (1/15)  d   0.0% (0/3)  θ p  7.1% (1/14)  ð  ɾ 100% (1/1)  t ͡ʃ  tj 20.0% (1/5)  d ͡ʒ   0.0% (0/2)  n  ɾ 7.1% (1/14)  l t ͡ʃ ɾ (4), ð (2) 33.3% (7/21)  ɾ  l (2), ð, r,̞ n 41.7% (5/12)  r d (3), n, ɾ lː, l (2) 100% (8/8) Coronal-Dorsal ʝ   0.0% (0/4) Dorsal k n, t (2) l (2), t (4), pː 90.9% (10/11)  g ʝ, b, t  100% (3/3)  x t (2), θ l, θ (2), f 87.5% (7/8)  ɣ  β, ɾ, ð (3) 83.3% (5/6) Table 3.7 Singleton substitutions by word position  The voiced stop [d] only replaced /r/, which it did three times. Interestingly, [n] replaced oral segments on two occasions, mysteriously in casa (/ˈkaθa/ > [ˈnaθa]) and also in ratón (/raˈton/   63 > [naˈton]), perhaps under the influence of the final consonant. Nasal assimilation also occurred in baño (/ˈbaɲo/ > [ˈmaɲo]). Fronting to labial segments occurred assimilatively in zapato (/θaˈpato/ > [paˈpapo]) and less explicably in gorra (/ˈgora/ > [ˈbola]).  The substitution of [w] for /p/ in Paula may rather be analyzed as deletion with diphthong metathesis: /ˈpaula/ > [ˈwaða]. Overall, [-continuant] seems to be a default word initially, although there are two examples of the reverse trend: pájaro (/ˈpaxaɾo/ > [ˈfalalo]), likely under the influence of the following consonant target /x/, and gato (/ˈgato/ > [ˈʝato]). The lateral /l/ is substituted for only once word initially (and it may be more properly phrase-medial with the article) in (la) luz (/ˈlu/ > [aˈt ͡ʃu]).  Word-medially, the segments [ð], /ɾ/ and /l/ saw several exchanges, which is some evidence in favour of [ð] acting more like a sonorant (approximant) segment than an obstruent. The trill is consistently replaced word-medially by /l/. Dorsal consonants are consistently replaced by coronal segments, and labial segments only in cases of assimilation, e.g. bruja (/ˈbɾuxa/ > [ˈbufa]) and fuego (/ˈfweɣo/ > [ˈfweβo]). Coronals also replaced labial segments on three occasions, in teléfono (/teˈlefono/ > [teˈleθeno]), chimenea (/t ͡ʃimẽˈne.a/ > [t ͡ʃiðĩˈɾe.a]) and primavera (/pɾimaˈβeɾa/ > [pimaˈðeɾa]), all polysyllabic words with possible sequence issues.   64 The reverse occurs once, in zapato (/θaˈpato/ > [paˈpapo]). The voiceless affricate is split into two segments once, in flecha (/ˈflet ͡ʃa/ > [ˈfetja]).  3.2.5.2   Phonetic differences There were three cases of consonants with small phonetic differences (or “distortions”): (7) jirafa  /xiˈɾafa/  >  [tiˈra̞fa] ‘giraffe’  princesa /pɾin ̪ˈ θeθa/  > [tin ̟̪ˈ θeθa] ‘princess’  euro  /ˈeuɾo/   >  [ˈeuˠo]  ‘Euro’  (imitated)   The /n/ in princesa likely reflects the phonetic environment, by being an assimilation in place to the /θ/ which is interdental. The two others, while less clearly motivated by environment, seem to fit Emilio’s later difficulties with the vibrants and may reflect specific difficulties with articulation of rhotics rather than phonological development.  Sofía and Tomás each have more instances of phonetic differences in consonants in their datasets (11 and 5, respectively). Tomás lateralized several taps, and Sofía produced coronals in a more anterior position, reduced the length of consonants, and lateralized taps.  3.2.5.3   Cross-vowel consonant harmony There were 36 instances of cross-vowel consonant harmony in total and 15 involving singletons. (Use of coronals for dorsals, even when in the presence of other [Coronal] consonants in the   65 word, was not counted because of ambiguity concerning coronal default insertion versus harmony.) Examples include: (8) zapato  /θaˈpato/  > [paˈpapo] ‘shoe’   Labial  baño  /ˈbaɲo/  >  [ˈmaɲo] ‘bath’ (imitated) Nasal  teléfono /teˈlefono/  > [teˈleθeno] ‘phone’  Coronal  fuego  /ˈfweɣo/  >  [ˈfweβo] ‘fire’   Labial  chocolate /t͡ ʃokoˈlate/  >  [t͡ ʃoloˈlate] ‘chocolate’  Lateral  Example 8 illustrates the variety of possible sequence constraints seen in Emilio’s productions at Time 1. The most striking examples involved labials and coronals, with possible sequence constraints against their alternation in polysyllabic words, such as in zapato and teléfono in Example 8 and chimenea (/t ͡ʃimẽˈne.a/ > [t ͡ʃiðĩˈɾe.a]). (Incidentally, this sequence constraint seemed to survive until time 2, with hipopótamo showing the only non-vibrant substitution, of /m/ > [n].) Zapato might be better described as partial reduplication rather than assimilation – also true of cocodrilo. The nasal assimilation seen in baño was rare in Emilio’s productions, with the other occurrence of baño being produced on target and only one other occurrence in the dataset: ratón (/raˈton/ > [naˈton]). Lateral spreading, however, was more common, as seen in chocolate in Example 8, cocodrilo (/kokoˈðɾilo/ > [toloˈliɾo]), and pájaro (despite having no target /l/; /ˈpaxaɾo/ > [ˈfalalo]). There did not seem to be any constraints against repeated features (cross-vowel dissimilation).  Sequences involving dorsal consonants seemed assimilatory or default-based in nature, with dorsals having a general constraint against their production, rather than in particular   66 environments. Dorsal consonants became coronals, presumably by default (e.g. jamón, /xaˈmõn/ > [taˈmõn]) or they became labials by assimilation (e.g. bruja, /ˈbɾuxa/ > [ˈbufa]).  Sofía and Tomás showed a comparably high number of instances of consonant harmony, with 25 each. These, however, also included instances of patterns resolving marked segments to defaults, i.e. Sofía had many instances of dorsals fronting to coronals, while Tomás frequently had stopping of fricatives. Thus, their sequence constraints may have been notably less than those of Emilio. This is particularly true in the case of alternating coronal-labial sequences, which had an accuracy of 28.6% for Emilio, compared with 85.7% for Sofía and 100% for Tomás.  3.2.5.4   Migration There were four instances of migration, below, including complete segment metathesis (cocodrilo with [...liɾo] but also including lateral harmony), migration of a feature (Paula, ruido, [+/-sonorant]), and migration of the [Labial] feature and its segmental timing unit (b>p:). Ruido could also be analyzed as labial harmony with the stopping of /r/, and simple substitution of [ɾ] for [ð]. (9) cocodrilo /kokoˈðɾilo/ > [toloˈliɾo] ‘crocodile’  Paula  /ˈpaula/  >  [ˈwaða] ‘Paula’  (doll’s name)  ruido  /ˈrwiðo/  > [ˈbwiɾo] ‘noise’  boca  /ˈboka/  >  [ˈopːa]  ‘mouth’ (imitated)    67 There was also one instance of possible segmental unidirectional metathesis but this may have been a dorsal fronting or coronal harmony pattern, with maintenance of the [+sonorant] of the dorsal ‘fricative/approximant’ in the tap: (10) tortuga /toɾˈtuɣa/  > [toˈtuɾa] ‘turtle’  Sofía and Tomás had a comparable number of instances of metathesis and migration, with two and three instances each. The tap /ɾ/ was the most frequently affected segment.  3.2.5.5   Coalescence? There was one instance of possible coalescence in Emilio’s dataset at age 4;8, involving a word-initial consonant cluster. In princesa (/pɾin ̪ˈ θeθa/ > [tin ̟̪ˈ θeθa]), the initial cluster was simplified to [t], which takes most features from the first consonant but [Coronal] place from the second. Sofía also had two instances of possible coalescence (including the same example in princesa), while Tomás had no instances of coalescence.  3.2.5.6   Epenthesis There were no instances of consonantal epenthesis in Emilio’s dataset at age 4;8, nor vocalic epenthesis between consonant clusters. There was only one possible instance of epenthesis, but this may have simply been the article la with the /l/ deleted, in (la) luz. (11) luz  /ˈlu/  > [aˈtʃ͡u]  ‘light’  (imitated)    68 Sofía had no instances of epenthesis, but Tomás had one, shown in Example 12. (12) bruja  /ˈbɾuxa/ > [bəˈɾuka] ‘witch’  3.2.6   Description of constraints Emilio’s productions can be summarized using constraints, both positive and negative, i.e. “faithfulness” and “markedness” constraints. Beginning with word length, Emilio could produce words of one to four syllables in length but not 5 syllables (0/1). In terms of constraints, *5Syllables > MAX > *4Syllables. That is, the constraint against five syllables outranks the constraint against (syllable) deletion, but any constraint against four syllables does not, and Emilio was able to pronounce 4-syllable words.  As for syllable structure, Emilio could produce singleton onsets, word-final codas and word-medial nasal and [h] codas. Formally in terms of constraint rankings, Onset, WFCoda, NasalCoda & hCoda > *Complex. *Complex, when applied to syllable structure, favours simple CV syllables, but Emilio’s syllable structure had partially progressed past that phase. He could produce word-initial sequences when the second element was /j/ or /w/ (as part of rising diphthongs rather than as part of the consonantal onset), so CGV > *Complex. He could not produce word-medial and word-final codas in the same word (*2Codas > MAX), a coda and a consonant cluster (even CG; *Coda+CC > MAX), or onset clusters when the second element was a liquid (*LiquidC2 > MAX). Some examples are as follows. (13) pantalón /pant̪aˈlon/  >  [tataˈlon] ‘pants’ (*2Coda > NasalCoda)  pluma  /ˈpluma/  >  [ˈpuma] ‘feather’ (*LiquidC2 > MAX)  jamón  /xaˈmõn/  >  [taˈmõn] ‘ham’ (*Dorsal > Place)   69 3.2.7   Variability in productions There were seven words that Emilio produced twice each. Two were produced on target both times, one was produced off-target but the same way both times, and the others were produced differently each time. This reflects some variability that is also reflected in the variable treatment of each phone observed above. (14) pez  /ˈpɛ/   >  [ˈpɛ], [ˈpɛ]  ‘fish’  peine  /ˈpenne/  >  [ˈpenne], [ˈpenne] ‘comb’  baño  /ˈbaɲo/  >  [ˈbaɲo], [ˈmaɲo] ‘bath’  (imitated)  zapato  /θaˈpato/  >  [paˈpapo], [aˈpato] ‘shoe’  flor  /ˈfl{oɾ,ɔ}/ >  [ˈfo], [ˈfɔ]  ‘flower’  reloj  /reˈl{ox,ɔ}/  >  [deˈlɔ], [deˈðo]  ‘clock’  Paula  /ˈpaula/  >  [ˈpaːɾa], [ˈwaða] ‘Paula’  (doll’s name)  Bernhardt (personal correspondence, February 27, 2017) noted that some children with PPD in Granada had difficulties pronouncing the doll’s name (Paula), possibly because it was an unfamiliar name to them, and possibly because the /au/ sequence is fairly uncommon. Baño shows cross-vowel nasal assimilation. Zapato may reflect a coronal/labial sequence constrain, also seen in other polysyllabic words in Emilio’s productions at time 1, such as teléfono and chimenea. Reloj shows the variable intersubstitution between /l/ and [ð] (and Paula between those and /ɾ/). Flor and reloj reflect the laxing and lowering of /o/ when word-final codas are deleted.    70 3.3   Descriptive findings: age 7;6 Emilio’s dataset at 7;6 was characterized by very few mismatches between actual and target pronunciations. Of 102 words, the following seven are the only ones with differences from adult pronunciations: (15) hirviendo /iɾˈβjend̪o/  >  [ixˈβjend̪o] ‘boiling’ (imitated)  flor  /ˈfloɾ/   > [ˈflox]  ‘flower’  hipopótamo /ipoˈpotamo/ > [hipopótano] ‘hippo’  hermano /eɾˈmãno/  >  [eʴˈmãno] ‘brother’ (imitated)  (le) regalo /reˈɣalo/  >  [t ͡ɬ:iɾeˈɣalo] ‘gift’  rojo  /ˈroxo/   > [ˈkɬ͡:oxo] ‘red’  (imitated)  ruido  /ˈrwiðo/  >  [ˈɬ:wiðo] ‘noise’   The final three words were originally transcribed with a question mark and a note saying the /r/ was “irregular”. The author therefore transcribed it as best he could, and the transcriptions were confirmed by a committee member and phonetician. Printouts of the spectrograms for those three words (regalo, rojo and ruido) are attached in Appendix 3. Acoustic measurements are provided in section 3.3.3.1 below.  3.3.1   Global measures Whole-word match at 7;6 was 93.1% (95/102). Of the words that contain clusters, 92.6% had a full segmental match (50/54). Overall percent consonants correct (PCC) was 97.8% (308/315). Percent vowels correct (PVC) was 100% (251/251). The target phonological mean length of utterance (PMLU) was 8.6. Emilio’s actual PMLU was 8.4, giving a proportion of whole-word proximity (PWP) of 0.98. Stress match was 100% (102/102), and therefore words with three or more syllables also had a stress match of 100%.   71 3.3.2   Word structure Word shape match was 99% (101/102 words) and in words with clusters only, 98% (52/53). The mismatched word was hermano (/eɾˈmãno/ > [eʴˈmãno]), in which the word-medial coda /ɾ/ was underarticulated rather than deleted completely.  3.3.2.1   Consonant clusters Cluster timing unit match was 98.5% (66/67), with the only non-match being /eɾˈmãno/ as [eʴˈmãno] as noted above. Word initial cluster timing unit match, therefore, was 100%, in both left-prominent and right-prominent words. Consonant cluster full segmental match, however, was 95.5% (64/67), with the non-matches being vibrants in sequences with glides or other consonants (/ˈrwiðo/ > [ˈɬx:wiðo], /eɾˈmãno/ > [eʴˈmãno] and /iɾˈβjend̪o/ > [ixˈβjend̪o]).  3.3.2.2   Vowel sequences Exact vowel match was 100%, and therefore vowel sequences, i.e. diphthongs and hiati also had 100% match.  3.3.3   Segments As noted, vowels had 100% match. Singleton consonant match was 97.8% (175/179), the only non-matches being /r/, /ɾ/ and /m/ in hipopótamo (/ipoˈpotamo/ > [ipoˈpotano]). These were evenly distributed between word-initial, -medial and -final positions, giving match percentages   72 of 96.6% (word-initial), 99% (word-medial) and 95% (word-final). Overall, codas had a full segmental match of 93.5% (43/46), with /ɾ/ being the only vulnerable segment.  As noted, /r/, /ɾ/ and /m/ each had at least one non-match. The /m/ was only missing once, substituted by [n] in hipopótamo (/ipoˈpotamo/ > [ipoˈpotano]), and had 90.9% accuracy overall (10/11). Likely the [n] arose from harmony with the [Coronal] feature of /t/, giving a [Labial][Labial][Coronal][Coronal] consonant place sequence rather than [Labial][Labial][Coronal][Labial], which may have been too complicated to produce in a 5-syllable word. Accuracy of the tap was also high at 90.9% (30/33); the trill was lowest at 62.5% (5/8).  Singleton matches for these phones showed differentiated results by word position. All three word-initial /m/ were correct, compared with four of five word-medial. Word-medial singleton vibrants, however, had 100% accuracy: all eight /ɾ/ and all three /r/. Word-initial singleton /r/ was low at 50% (2/4) and word-final /ɾ/ had only one target, which was mismatched (0/1).  3.3.3.1   Phonetic description of “irregular” /r/ Because Emilio’s /r/ productions at time 2 were not simple substitutions, but showed notable perceptible differences from typical /r/, acoustic phonetic measurements were taken to compare   73 with Carballo and Mendoza’s (2000) findings. Table 3.8 shows Emilio’s /r/ measures for regalo, ruido, and rojo, collected with Praat, version 6.0.04 (Boersma, 2001).  Measure regalo ruido rojo F1 frequency (Hz) 821.6 931.6 726.1 F2 frequency (Hz) 2309.5 2168.5 2069.7 C-V ratio (dB) 2.31 N/A 11.62 Consonant duration (ms) 386.73 355.14 355.63 Table 3.8 Phonetic measures of /r/ at time 2    74 Chapter 4:  Discussion This study looked at 4-year-old Emilio’s case of PPD in order to tease out possible factors that could have predicted his persistent (relatively mild) phonological difficulties. Global, structural, segmental and featural descriptions of his dataset were described at two time points, age 4;8 and 7;6. In order to determine how/whether Emilio differed from his peers with PPD at age 4;8, 16 phonological measurements were extracted from the literature on 4-year-old children with PPD learning Spanish. Emilio's phonological skills at age 4;8 were compared with these scores. Five of Emilio's scores were lower than the expected ranges from the literature: word shape match (WSM), WSM in words with clusters, word-initial consonant cluster (CC) shapes (both left- and center-/right-prominent) and full segmental coda match. When the results (excluding exact coda match) were compared to data from two age and WWM peers, Sofía and Tomás, the peers’ results exceeded the expected range for all but word-initial clusters in center-/right-prominent words. This indicates that word structure was highly relevant for distinguishing Emilio from his peers, in WSM, WSM in words with clusters and word-initial cluster timing unit match in left-prominent words. A review of his descriptive results follows in the next two sections. The Discussion then compares time 1 data with the expectations derived from the literature on 4-year-old Spanish speakers with PPD, discusses the relevance of the measures not included in the expectations and discusses predicting persistent difficulties in PPD. The limitations of this study are discussed, and implications for clinical practice and research are suggested.   4.1   Review of descriptive results: age 4;8 Emilio’s global measures, including WWM, PCC and PWP, were slightly lower than his comparison peers (Sofía and Tomás) and the averages for PPD described in the literature   75 (Dubasik & Ingram, 2013; Pérez, 2013). His polysyllabic mismatch scores (as in Schretlen, 2013), however, were not substantially different from his cohort’s averages or Sofía and Tomás’ scores. Stress match was high, at 95.4%. Word-final codas had a 100% match rate (/n/ and /l/ being the only targets, and optional ones at that). Rising diphthongs had high accuracy (93.8%), but falling diphthongs had low accuracy (45.5%), especially compared with TD peers, e.g., Ponce (2014). Consonant preceding rising diphthongs (CG) always had a timing unit match except for suave (/ˈθwaβe/) and cuadro (/ˈkwaðɾo/), in which C1 was deleted and the glide (/w/) lengthened. It is unclear why those two would be deleted, especially when /θ/ was a relatively accurate phoneme and a word like gwante (/ˈgwant̪e/), was produced on target. Emilio showed some mismatch patterns including substitution, often to /t/, /d/ and [ð], three instances of phonetic variation, cross-vowel harmony with sequence constraints against alternating coronals and labials, lateral spreading and a few (4-5) instances of migration. Other than the sequence constraints, Emilio’s mismatch patterns seemed comparable to those of Sofía and Tomás, and in keeping with Martínez and Diez-Itza’s (2012) finding that (often regressive), non-contiguous and interconsonantal assimilations were common and persistent among 3- to 6-year-old Spanish-speaking children. Emilio’s fricatives were more accurate (80% feature match) than those of Tomás, and liquids had relatively low accuracy for all three of the 4-year-olds. Of the liquids, Emilio’s laterals were stronger, with an overall match of 48.3%, while vibrants had an overall match of 7%. All three children had a trill full segmental match of 0%. Several of Emilio’s segments were never produced accurately as singletons, i.e. /r/, [ð], /k/ and /g/, and others were   76 accurate less than half the time: /x/, [ɣ] and /ɾ/. Some variability in productions was observed, with five of seven words recorded twice being produced differently each time.  Overall, WSM (both including and excluding /j/ and /w/ in the C category) was notably lower than that of his comparison peers. One of the most striking observations was that Emilio had no matches, timing unit or exact, for true onset clusters (excluding /j/ and /w/ as possible onset glides). This (true) onset cluster match of 0% is the number comparable to Bernhardt et al. (2015a): 40.9% for 3-year-olds and 89.2% for 4-year-olds with PPD. In all but two true onset clusters, C2 is deleted. The two exceptions (cocodrilo and globos) show deletion of C1 instead, and are somewhat explicable through the literature: [ð] tends to delete word-medially (albeit intervocalically) in adult Granada Spanish (Gómez-Molina & Gómez-Devís, 2008, cited in Samper, 2011), and /g/ before /l/ in onset clusters in child phonology when deletion occurs (Barlow 2003a). Heterosyllabic clusters with a liquid coda had a timing unit match of 14.3% (1/7), and with other sonorants (nasals and [h]) 68.8% (11/16); both lower than his cohort’s average for 4-year-olds with PPD of 81% (Bernhardt et al., 2015a). Dorsal consonants were also notably low, with [Dorsal, +consonantal] match being 17.1%, although dorsals were noted to be still developing in his TD peers by Raymond (2017). While vibrants had low accuracy for all three children, vibrant feature match was more distinctive (substitutions retaining the [+vibrant] feature): Emilio’s 11.9% was modestly removed from Sofía’s 20.0% and Tomás 48.9%. Finally, as mentioned above, alternating coronal-labial sequences had lower accuracy for Emilio than Sofía and Tomás, with 28.6% accuracy compared with 85.7% and 100%.   77 4.2   Review of descriptive results: age 7;6 As noted earlier, Emilio was the only child in a cohort of 15 followed longitudinally in Granada, to have persistent speech difficulties at age 7;8 (and was the only one still receiving speech therapy, for trilled /r/). As in time 1, the Granada data did not include measurements of intelligibility or other functional or participation measures, due to insufficient grant funding. At age 7;6, Emilio’s dataset was characterized by a much higher WWM: 93.1%. PCC, likewise, was very high, at 97.8%. Stress match was 100%. Only seven words differed from their adult targets, and each in only one consonant, always replaced by another. Only hipopótamo (/ipoˈpotamo/ > [ipoˈpotano]; a 5-syllable word) differed on a non-vibrant segment. The other non-matches were all velarization of /ɾ/ in coda position to [x] and velarization, lateralization and spirantization of word-initial /r/.   While there are no phonetic norms for /r/ productions in the contexts in which Emilio had difficulties, an informal comparison with Carballo and Mendoza (2000) gives the impression of high F1 and F2 frequencies and substantially longer duration, on word-initial /r/ in regalo, ruido and rojo. While Carballo and Mendoza’s (2000) results were all /r/ followed by /a/, it seems fair to comment that the F1 and F2 frequencies were all higher than the highest averages for the younger groups (F1) and the low intelligibility group (F2). C-V is less comparable still due to the differences in vowel stress (regalo) and shape (ruido). Duration, however, is over double the longest group’s (the incorrect pronunciation group). Occlusions and apertures were impractical to count, because the spectrograms seemed to show frication more than trill (see Appendix 3).   78 Incidentally, this rapid pulse characteristic is characterized by Widdison (1997) as the central feature of the alveolar trill, suggesting reduced clarity when it is missing.  4.2.1   Comparison of difficult words across data collections The seven words Emilio produced differently from the adult targets at time 2 were six words involving vibrants and one polysyllabic word. Below are the seven words produced at age 7;6 (example 16) with the same seven words produced at age 4;8 underneath them (example 17). (16) hirviendo /iɾˈβjend̪o/  >  [ixˈβjend̪o] ‘boiling’ (imitated)  flor  /ˈfloɾ/   > [ˈflox]  ‘flower’  hipopótamo /ipoˈpotamo/ > [hipopótano] ‘hippo’  hermano /eɾˈmãno/  >  [eʴˈmãno] ‘brother’ (imitated)  (le) regalo /reˈɣalo/  >  [t ͡ɬ:iɾeˈɣalo] ‘gift’  rojo  /ˈroxo/   > [ˈt ͡ɬ:oxo] ‘red’  (imitated)  ruido  /ˈrwiðo/  >  [ˈɬ:wiðo] ‘noise’   (17) hirviendo /iɾˈβjend̪o/  >  [iˈβjenn̪o̪] ‘boiling’ (imitated)  flor  /ˈfloɾ/   > [ˈfɔ], [ˈfo] ‘flower’  hipopótamo /ipoˈpotamo/ > [poˈpotamo] ‘hippo’  hermano /eɾˈmãno/  >  [enˈmãno] ‘brother’ (imitated)  regalo  /reˈɣalo/  >  [deˈɣaɾo] ‘gift’  rojo  /ˈroxo/   > [ˈɾoθo]  ‘red’  (imitated)  ruido  /ˈrwiðo/  >  [ˈbwiɾo] ‘noise’   The data show a clear progression in accuracy from time 1 (Example 17) to time 2 (Example 16). CCG was produced in hirviendo, albeit with a dorsal fricative in place of /ɾ/, all five syllables are produced in hipopótamo, albeit with a substitution of place. CC was produced in flor, and in each word only one segment was produced differently from the adult target.   79 4.2.2   Discussion of challenges at age 7;6 It is unclear to what extent Emilio’s difficulties at time 2 reflected phonological difficulties (with codas and polysyllabic words) versus articulatory challenges of vibrant consonants. It seems most likely that both phonological representation/processing and actual articulation were at play, however. His difficulties with tap and /r/ were sensitive to word structure, occurring in coda and word-initial position, respectively, and hipopótamo is polysyllabic. Tap and /m/ were substituted with other segments rather than simply showing small phonetic deviations,  suggesting phonological difficulty; however, /r/ was attempted but unsuccessfully produced some of the time ("irregular" was the term used by the Granada transcriber for several of his /r/ attempts). This follows the pattern noted by Carballo and Mendoza (2010), in which younger children tend to substitute other segments for /r/, whereas older ones who produce /r/ incorrectly do so without substitution (showing phonetic differences alone). Pérez (2013) also found that some children with PPD velarized trill /r/ and substituted /l/ for both vibrants; both lateralization and velarization are present in time 2.  Comparing Emilio’s /r/ with Carballo and Mendoza (2000), Emilio patterned with the younger groups in high F1 frequency, the low intelligibility group in high F2 frequency, and the incorrect pronunciation group in lengthened consonant duration. Indeed, he was more extreme than each of those groups in their respective differentiating measures. The higher F1 and F2, following Carballo and Mendoza (2000), may be related to a more fronted tongue body (F2) and whether or not the tongue blade and apex are raised or not (F1). Overall, these formants relate to whether   80 the tongue shape is suited for the aerodynamic requirements for a successful trill production. Ladefoged (1993) hypothesizes that a stiff tongue blade may make it difficult to form the necessary tongue configurations and aerodynamics for a trill. As for consonant duration, Emilio patterned with the older children in Carballo and Mendoza (2000) that produced the trill incorrectly rather than substituting a tap or other consonant for it, as many younger children did (and as Emilio did at time 1). At time 2, Emilio was clearly attempting the trill, but lacked the motor control to successfully execute it, as Carballo and Mendoza (2000) suggest of the incorrect pronunciation group. Widdison (1997) surveys descriptions of the aerodynamic and myoelastic requirements for successful trill production, and notes that /ʀ/, the uvular trill, has less stringent airflow requirements. Perceptually, it does seem that Emilio’s trill attempts had dorsal features – perhaps they were compensatory articulations that approximated the acoustic properties of the apical trill by producing a signal that was pulse-like.  The tap /ɾ/, however, may also have an unforeseen allophonic factor at play, with trill and tap varying with sibilants/fricatives ([r]̆) in a number of dialects of Spanish including that of the Andean highlands (Rafat, 2015). In Mexican Spanish, it is a marker of gender, with females producing the variant and males being highly resistant to it (Rissel, 1989). This variant can take place, among other environments, in coda position both word-medially and word-finally (Bradley, 1999), as we see in Emilio’s data. Indeed, the articulatory-acoustic properties of the assibilated rhotic have been compared to palato-alveolar sibilants (Colantoni, 2006), which would imply that the transcripted [x] is quite a bit closer acoustically and articulatorily to an   81 established variant than it originally appeared. Rhotics also vary, at least in Argentinian Spanish, in both voicing and degree of constriction, from vibrant or fricative to approximant (Colantoni, 2008), possibly shedding light on Emilio’s production of hermano ([eʴˈmãno]). In any case, Emilio’s therapy focused on trilled /r/, and perhaps his speech-language pathologist judged his ‘tap’ productions to be within functional limits. In choosing alternative transcriptions for Emilio’s taps, however, it seems the Granada team judged them as differing from their targets. Without intelligibility ratings for these words, however, it is difficult to judge whether or not Emilio’s productions differed in a meaningful way from his peers.   4.3   Comparisons with expectations (Time 1) Of the 16 expectations derived for a 4-year-old with PPD (end of introduction), certain variables met expectations (10/16) and others did not (6/16). Of those that did not, five of the expected levels were higher than Emilio’s scores and one was lower. (See Table 3.1.)  4.3.1   Variables meeting expectations  Ten of Emilio's scores on the 16 measures fell into the expected ranges for 4-year-olds with PPD; thus, these measures were not able to distinguish Emilio’s productions at time 1. These included global measures WWM, PCC, PMLU, PWP, plus more specific measures: overall and polysyllabic stress match, CC full segmental match, falling diphthong match, singleton consonant with the lowest accuracy, and singleton vowel match.    82 Across global measures, Emilio's scores were within expected ranges for PPD at age 4. His Whole word match (WWM) was 26.6%, in the expected range of 10 to 40% (Chávez-Peón et al., 2012; Pérez, 2013). His PCC of 60.7% similarly fell in the expected range of 50 to 65% (Burrows & Goldstein, 2010; Chávez-Peón et al., 2012). Phonological mean length of utterance (PMLU) was expected to be between 6.1 and 7.0 (Burrows & Goldstein, 2010; Dubasik & Ingram, 2013), and Emilio’s PMLU was 6.4. Similarly his proportion of whole-word proximity (PWP) at .80 was in the expected range of .7 to .85 (Burrows & Goldstein, 2010; Dubasik & Ingram, 2013).   Looking at more specific global measures, his scores matched expected ranges. Overall stress match was expected to be within 90 and 100% (Bernhardt et al., 2015a; Chávez-Peón et al., 2012; Pérez, 2013), and Emilio’s stress match was 95.4%. Emilio’s polysyllabic word stress match was 88.8%, in the expected range between 70 and 90% (Pérez, 2013; Schretlen, 2013). Consonant cluster (CC) full segmental match was expected to be between 15 and 35% (Chávez-Peón et al., 2012; Pérez, 2013), and his score was 26.2%. His falling diphthong (full segmental) match was just within the expected range at 45.5% (45 to 55%: Pérez, 2013).   Considering individual segments, trilled /r/ was expected to have the lowest accuracy (Boyce et al., 2016; Pérez, 2013), and /r/ did indeed have 0% accuracy, along with /k/, /g/ and [ð]. Singleton vowel match was expected to be within 90 and 100% (Goldstein & Pollock, 2000, 2004), and it fell within that range at 95.5%. Rising and falling diphthongs and vowel hiati (V.V   83 sequences across syllable boundaries) were not considered singleton vowels and were not included in this calculation (incidentally, he produced all three instances of hiatus on target).  Of the ten variables that met expectations, seven were compared with scores of Sofía and Tomás; they either scored above or within the expected ranges, and for four of those measurements, had slightly or moderately (falling diphthongs) higher than Emilio's. These seven were chosen for reasons of both availability and interest: the global measures were easily retrieved through Phon, while the least accurate phone was expected to be /r/, a phoneme of interest in this case because of Emilio’s difficulties at time 2. Falling diphthong match was compared because in an original calculation Emilio’s score fell beneath the expected range.  For WWM, Sofía and Tomás scored 33.3% and 43.1% respectively (within or above the expected range of 10-40%, slightly higher than Emilio). Their PCC scores were 67.8% and 70.3% respectively (above the expected range of 50-65%). For PMLU, Sofía’s score was above the expected range of 6.1-7.0, with a PMLU of 7.27. Tomás’ PMLU, on the other hand, was within the range at 6.86, similar to Emilio's. Related to PMLU is PWP, on which Sofía and Tomás scored .88 and .85 respectively, above and just within the expected range of .7 to .85, slightly higher than Emilio's. Overall stress match was expected to be between 90 and 100%, and Sofía and Tomás scored 94.4% and 95.4%, respectively, like Emilio's. Falling diphthong matches were higher than the expected range of 45-55%, at 90.9% and 83.3% and much higher than Emilio's. As for the phone with the lowest accuracy, Sofía and Tomás also confirmed that expectation by having the lowest accuracies with /r/ at 0% (although for Sofía and Tomás, this   84 was the only singleton with 0% accuracy). The slightly higher scores of Sofía and Tomás may have been relevant in distinguishing their performance from Emilio's relative to persisting speech difficulty, but a larger study would be needed to confirm this trend.  4.3.2   Variables that did not meet expectations Six of Emilio's measures did not meet expectations at age 4 for PPD cohorts. Emilio had lower scores than the expected range on five variables and higher scores on one variable. The one higher score was for rising diphthongs (full segmental) match, which was expected to be within 75 and 85%, but Emilio scored 93.8%. Because higher performance was not expected to be a predictor of persistent difficulties, that score was not compared with those of Sofía and Tomás.  The five expectations that overestimated Emilio’s scores were WSM, WSM in words with clusters, word-initial CC timing unit match (stressed and unstressed), and coda full segmental match. Word shape match (WSM) was expected to be between 55 and 65% (Bernhardt et al., 2015a; Pérez, 2013), and Emilio’s WSM was slightly lower at 53.2% (58/109 word shapes). WSM in words with clusters was expected to be within 30 and 40% (Bernhardt et al., 2015a; Pérez, 2013), but Emilio’s score again fell just short at 28.0%. Word-initial stressed CC timing unit match was expected to be between 35 and 45% (Bernhardt et al., 2015a; Pérez, 2013), but Emilio scored 26.9%. Word-initial unstressed CC timing unit match was expected to be within 10 and 20% (Bernhardt et al., 2015a; Pérez, 2013), but Emilio scored 0%. Finally, coda full segmental match was expected to be within 65 and 80% (Chávez-Peón et al., 2012; Pérez, 2013), but Emilio scored 55.2%.    85 Of the five variables on which Emilio had lower scores than expected, three of those were also notably lower than scores of Sofía and Tomás. These measures thus distinguished Emilio quite clearly from his two comparison peers. The three such measures were WSM, WSM in words with clusters, and word-initial stressed CC timing unit match. WSM scores were expected to be between 55 and 65%, whereas Sofía and Tomás scored 82.4% and 73.4%, respectively. WSM in clusters was expected to be within 30 and 40%, whereas Sofía and Tomás scored 72.5% and 44.9%, respectively. Finally, word-initial stressed CC timing unit match was expected to be 35-45%, whereas Sofía and Tomás scored 80.8% and 84.0%, respectively.  The two other variables with higher expected scores than Emilio's, word-initial unstressed CC timing unit match and coda full segmental match, did not distinguish Emilio from Sofía and Tomás. The first did not because Sofía and Tomás each also scored 0%, whereas the second was not compared due to the variability of codas in Granada Spanish. The first measure’s expectation range was subject to sampling bias, however, because the elicitation included only three words with CC in word-initial unstressed syllables, making it impossible to score between 0% (0/3) and 33.3% (1/3). Further research is needed concerning this variable..  4.3.3   WSM values from Bernhardt et al. (2015a) Because WSM appeared to be a potentially useful measure for predicting persistent phonological difficulty, Emilio's WSM values were compared with those of all 3- and 4-year-olds in Emilio’s longitudinal PPD cohort. The resulting values, including WWM and WSM at the first time point, are shown in Table 4.1.     86 Anonymized Code  or Fictional Name Age WWM WSM 325 3;3 15.9% 38.7% 306 3;5 12.8% 45.0% Sofía 4;3 33.3% 58.3% 327 4;3 51.9% 73.5% 322 4;5 55.6% 65.4% Emilio 4;8 27.5% 57.0% 309 4;9 50.0% 75.5% 315 4;9 50.0% 70.4% 323 4;9 48.6% 74.5% Tomás 4;10 43.1% 77.3% Averages 54.1 months = 4;6 38.87% 63.56% Table 4.1 WWM and WSM of 3- and 4-year-olds in Emilio's cohort (Bernhardt et al., 2015a)  Emilio was indeed the 4-year-old participant with the lowest WSM (although Sofía was a close second). WSM in Bernhardt et al. (2015a), however, was calculated with glides /j/ and /w/ separated from C and V. This enables a greater distinction between those who actually substitute liquids for other liquids in clusters from those who substitute glides. As a result, the measure has potentially greater sensitivity to structural difficulties. Serendipitously, I first measured WSM with the simpler C/V distinction, i.e. including all glides in C rather than G. This enabled the measure to separate Emilio from Sofía, who did in fact substitute glides for liquids. With glides considered a match for consonants, her WSM went from 61.1% (my measurement) to 82.4% - while Emilio’s remained in the 50s either way because he did not substitute glides for liquids (he generally deleted the liquids). Other differences in calculation, from method of target choice and calculation as well as human error may also be present - making it difficult to compare too closely across the studies.   87 4.4   Results relative to expectations The measures that distinguish Emilio from the study predictions were largely structural. WWM, despite being the lowest of his longitudinal cohort at time 2, was within the predicted range at age 4;8. WSM, however, was below expectations at time 1, particularly in comparison with his two age and WWM peers at age 4, Sofía and Tomás. This indicates that while Emilio’s percentage of words exactly on target was within expected range for a child with PPD of his age (although still low), he was not approaching the shape of the target words as much as his peers. Other global measures (PCC, PVC, PMLU, PWP) also fell short of distinguishing Emilio from his peers, even when his scores were lower. Interestingly, CC timing unit measures distinguished Emilio from the predicted ranges, but not CC full segmental matches. This again indicates that his peers were not achieving target clusters, but they were achieving or approximating their shape more than Emilio, who simply deleted one consonant in most clusters with liquids. A less-explicable but possibly just as useful contrast involved the falling versus rising diphthongs: Emilio performed above expected range on rising diphthongs but just on the lower edge of the expected range on falling diphthongs – a clear example of the atypical pattern found in children with PPD in Pérez (2013). Coda full segmental match was the final distinctive measure, which again highlights structure, but in this case interacting with segments. Minimal attention was given to this measure, however, due to Granada Spanish’s frequent deletion of codas.  Of these measures, WSM is arguably the easiest measure to determine and the most global. It is also the only one confirmed both by comparison to Sofía and Tomás and again, although with a different measurement method, to the rest of his longitudinal cohort. As discussed briefly in section 4.3.3, the difference in method clarified a distinction between Sofía and Emilio. One   88 possible implication is that WSM is optimally calculated two different ways for different purposes: distinguishing glides from consonants in order to diagnose PPD, and including glides in consonants in order to separate moderate PPD from severe (and possibly persistent) PPD.  4.5   Relevance of other measures While the literature provided sufficient information to make specific predictions for 16 measures, other measures were taken and evaluated informally. Emilio’s results on Schretlen’s (2013) MSW (multisyllabic word) rubric were compared with his cohort’s averages and Sofía and Tomás’ scores, and were not substantially different, suggesting that the metric was less sensitive to his difficulties than WSM alone. His strengths of syllable and stress match may have contributed to this lack of differentiation by the MSW metric. Polysyllabic words are also an important area to consider, however, and were largely the words showing difficulties at the later time point. Of the feature combinations considered (fricatives, dorsal consonants and liquids), dorsal consonants emerged as the most likely candidate for predictive utility, because Emilio’s accuracy was notably lower than Sofía’s and Tomás’, both in full segmental match and feature match. Emilio did comparatively well on fricative match, and all three children had low accuracy for vibrants. Of phonological mismatch patterns considered, coronal-labial alternating cross-vowel sequences presented considerably more difficulty for Emilio than for Sofía and Tomás, with match rates of 28.6% compared with 85.7% and 100%. Especially considering the usefulness of WSM in identifying PPD as noted in Bernhardt et al. (2015a), however, any segmental or feature measure should best be considered within the context of overall WSM. It seems unlikely that a 4-year-old with difficulty with dorsal consonants alone, for example, would be as likely to have persistent difficulties as others if WSM is high.   89 4.6   Predicting persistent difficulty in phonological development While other studies examining PPD in Spanish focus on identifying PPD (e.g., Bernhardt et al., 2015a), this case study focused on identifying potential for long-term difficulties, as has been done in English by Shriberg et al. (1994a, b), Flipsen (2015) and Wren et al. (2011, 2013, 2016), among others. Among other types of predictors, such as socio-economic and other linguistic factors, a phonological predictor identified in English has been number of deletions (Shriberg et al., 1994a, b; Wren et al., 2013).  Comparing Emilio with this study’s expectations, word structure was the most salient area of difficulty. Word shape match (WSM), WSM in words with clusters, and especially word-initial stressed cluster timing unit match differentiated Emilio from Sofía and Tomás. Of the measures not included in the 16 expectations, coronal-labial alternating cross-vowel consonant sequences and dorsal feature and full segmental matches also differentiated Emilio from Sofía and Tomás (although the TD members of his cohort also had some difficulty with dorsals; Raymond, 2017).   Interestingly, however, word structure was less obviously a challenge at age 7;6. There may be interactions, of course: his only non-vibrant mismatch occurred in a 5-syllable word, for example, and his tap substitutions took place only in the less salient, more restricted coda position (although trill substitutions took place word-initially). It was not simply a matter of difficulty articulating vibrants: there was an interaction between the challenging segments or features and challenging aspects of word structure. WSM is then a candidate for the answer to the question: what could have predicted Emilio’s continued difficulties with phonology? WSM has also already been noted to be an identifier for PPD, and so as a measure it can both be a   90 screener for PPD and a measurement of its possible persistence; as noted above, for Spanish, the method should perhaps change to include or exclude /j/ and /w/ from the C category to be more or less sensitive to severity of PPD.  4.7   Limitations of this study One limitation of this study is that no functional measures were taken, such as intelligibility or literacy measures, resulting from research cuts necessitated by funding limitations. While this isolated focus on phonology enables theoretical analyses, it does distance the results from Emilio’s life participation and functioning. This is especially relevant for the later dataset, with very few off-target productions. His /r/ is “irregular”, his /ɾ/ is sometimes realized as a dorsal fricative and one 5-syllable word is pronounced with a minor substitution (/m/ > [n]). It would be helpful to know if his speech is significantly less intelligible than his peers’ as a result. Lousada, Jesus, Hall and Joffe (2014) and Miller (2013), for example, discuss using intelligibility as a robust and functional measure for clinical assessment and outcomes in the context of studies with Portuguese- and English-speaking participants, respectively. Klein and Flint (2006) show how intelligibility can be used as a justification for targeting certain processes over others, in English-speaking children, when research has already been conducted. However, the fact that he was still receiving speech therapy for the trill suggested it was considered to be somewhat impactful for his life.  Another functional measure would be self-perception and others’ perception of difference and any distress caused by the same. Incidentally, this could also save researchers from   91 misdiagnosing lesser-known regional variability as phonological impairment. Purely “objective” measures are only as good as the relevance of the standards to which they are compared.  A further limitation is that the phonological samples were one-word utterances, which is itself distanced from connected speech. In English, some speakers show phonological difficulties in connected speech, but not one-word utterances (Howard, Wells & Local, 2008). These difficulties are missed entirely in single-word elicitations, such as this one.  Another area that would have been interesting to include would be phonological processing, including phonological working memory tasks and phonological awareness tasks. These have been shown to correlate with later literacy skills in English-speaking children (Bernhardt & Major, 2005; Lewis et al., 2011), and would therefore be variables of interest on their own and to compare with persistent PPD.  A central limitation of this study is its suggestive, rather than conclusive, nature. This is directly related to its nature as a case study: predictive value cannot be inferred from one retrospective study alone. As WSM becomes a more widely used measure, we will hopefully see its value proven in more studies with more participants.  4.8   Clinical implications This case highlights the importance of word structure as an independent focus in clinical evaluation. While some clinicians may find it more straightforward to focus on individual speech sounds, word structure may be a more important factor to consider in creating caseload priorities.   92 Emilio's outcomes suggest that word structure needs to be considered as a key target for therapy, independently of individual segments or features. Such intervention, inspired by nonlinear phonology, has been shown to be effective at increasing WSM (Bernhardt, 1992a; Edwards, 1995).  4.9   Future directions As all case studies do, Emilio’s raised more questions than it answered. The central question that has emerged is whether or not the identified variables of low WSM (overall and in words with clusters), timing unit match of word-initial stressed clusters, dorsal feature and full segmental match, and coronal-labial alternating sequence match are reliably correlated with persistence in phonological difficulties. Future longitudinal studies with more participants could answer that question. Then further, the question can be asked: if any of these is a useful predictor for persisting PPD in Spanish-speaking children, will that also be true for children learning other languages?  Once identification of persistent PPD in Spanish is better understood, the question of treatment will be all the more salient. Does intensive intervention at a younger age improve outcomes for those identified as being at risk for persistent PPD? Rhotics in particular are often targets for treatment in persistent PPD in English, because they are a later-developing speech sound and a persistent source of difficulty (Adler-Bock et al., 2007; Boyce, 2015).   93 Chapter 5:  Conclusion Emilio, a 4-year-old boy with PPD from Granada, Spain, was chosen to be a participant in this case study due to the persistence of phonological difficulties, unique in a cohort from Bernhardt et al. (2015a) followed longitudinally. At time 2, Emilio continued to have difficulty with word-initial trilled /r/, possibly tap in coda position and one polysyllabic word, hipopótamo. Two age peers from the cohort, Sofía and Tomás, were selected as comparisons, because they had comparable WWM at time 1 but no persistent difficulties at the second data collection. Sixteen predictions were extrapolated from the literature on the phonological acquisition of Spanish. Of the sixteen, three measures emerged as particularly viable candidates to predict persistence of PPD: word shape match (WSM), WSM in words with clusters and word-initial stressed cluster timing unit match. Two other factors that showed promise in comparison with Sofía and Tomás were difficulty with dorsal consonants (exact or feature match) and difficulty with alternating coronal-labial cross-vowel sequences. This case adds suggestive evidence for the usefulness of structural measures in assessment of PPD and proposes the usefulness of two ways of WSM calculation for Spanish: distinguishing /j/ and /w/ as glides (neither consonants nor vowels) for greater sensitivity to PPD and including them in the consonant category to separate moderate PPD from severe and possibly persistent PPD. Further research is needed to investigate the predictive abilities of these measures in Spanish and other languages.    94 References Acevedo, M. A. (1993). Development of Spanish consonants in preschool children. Journal of Childhood Communication Disorders, 15(2), 9-15. Adler-Bock, M., Bernhardt, B. M., Gick, B., & Bacsfalvi, P. (2007). The use of ultrasound in remediation of North American English /r/ in 2 adolescents. American Journal of Speech-Language Pathology, 16, 128-139. Aguinaga, G., Armentia, M. L., Fraile, A., Olangua, P. & Uriz, N. (2004). Prueba de Lenguaje Oral, Navarra Revisada (PLON-R) [Test of Oral Language, Navarra, Revised]. Madrid: TEA. Arias, J., & Lleó, C. (2014). Rethinking assessment measures of phonological development and their application in bilingual acquisition. Clinical Linguistics and Phonetics, 28(3), 153-175. Astruc, L., Payne, E., Post, B., Prieto, P. & Vanrell, M. del M. (2010). Word prosody in early child Catalan, Spanish and English. Speech Prosody 2010, 11-14. Chicago. Barlow, J. A. (2001a). Case study: optimality theory and the assessment and treatment of phonological disorders. Language, Speech and Hearing Services in Schools, 32, 242-256. Barlow, J. A. (2001b). Prologue: recent advances in phonological theory and treatment. Language, Speech, and Hearing Services in Schools, 32, 225-228. Barlow, J. (2003a). Asymmetries in the acquisition of consonant clusters in Spanish. Canadian Journal of Linguistics-Revue Canadienne De Linguistique, 48(3-4), 179-179.    95 Barlow, J. (2003b). The stop-spirant alternation in Spanish: Converging evidence for a fortition account. Southwest Journal of Linguistics, 22, 51-86. Barlow, J. A. & Gierut J. A. (1999). Optimality theory in phonological acquisition. Journal of Speech, Language, and Hearing Research, 42, 1482-1498. Beitchman, J.H., Wilson, B., Johnson, C.J., Atkinson, L., Young A., Adlaf, E., Escobar, M. & Douglas, L. (2001). Fourteen-year follow-up of speech/language-impaired and control children: psychiatric outcome. Journal of American Academic Child Adolescent Psychiatry, 40(1), 75-82. Bernhardt, B. (1992a). The application of nonlinear phonology to intervention with one phonologically disordered child. Clinical Linguistics and Phonetics, 6(4), 283-316. doi: 10.3109/02699209208985537 Bernhardt, B. (1992b). Developmental implications of nonlinear phonological theory. Clinical Linguistics & Phonetics, 6(4), 259-281. doi: 10.3109/02699209208985536 Bernhardt, B. & Gilbert J. (1992). Applying linguistic theory to speech-language pathology: the case for nonlinear phonology. Clinical Linguistics and Phonetics, 6(1), 123-145. doi: 10.3109/02699209208985523 Bernhardt, B. M., Hanson, R., Pérez, D., Ávila, C., Lleó, C., Stemberger, J.P., Carballo, G., Mendoza, E., Fresneda, D. & Chávez-Peón, M. (2015a). Word structures of Granada Spanish-speaking preschoolers with typical versus protracted phonological development. International Journal of Language and Communication Disorders, 50(3), 298-311. doi:10.1111/1460-6984.12133.   96 Bernhardt, B. & Major, E. (2005). Speech, language and literacy skills 3 years later: a follow-up study of early phonological and metaphonological intervention. International Journal of Language and  Communication Disorders, 40(1), 1-27. Bernhardt, B. M., Másdóttir, T., Stemberger, J.P., Leonhardt L., & Hansson, G.O. (2015b). Fricative acquisition in English- and Icelandic-speaking preschoolers with protracted phonological development. Clinical Linguistics & Phonetics, 29(8-10), 642-655. Bernhardt, B. M. & Stemberger, J. P. (1998). Handbook of phonological development: from the perspective of constraint-based nonlinear phonology. San Diego, CA: Academic Press. Bernhardt, B. M. & Stemberger, J. P. (2000). Workbook in nonlinear phonology for clinical application. Austin, TX: Pro-Ed (copyright reverted to authors, available at phonodevelopment.sites.olt.ubc.ca). Bernhardt, B. M. & Stemberger, J. P. (2012). Translation to practice: transcription of the speech of multilingual children. In S. McLeod & B. Goldstein (Eds.), Multilingual Aspects of Speech Sound Disorders in Children (pp. 170-181). Bristol: Multilingual Matters. Bernhardt, B. M. & Stemberger, J. P. (in press). Investigating typical and protracted phonological development across languages. In E. Babatsouli, D. Ingram & N. Müller (Eds.), Where typical and atypical language acquisition meet cross-linguistically. Bristol, UK: Multilingual Matters. Bernhardt, B.M., Stemberger, J.P., Chávez-Peón, M., Adler-Bock, M., & Raymond, G. (2016). Análisis Fonológico No Lineal: Español_2016. Retrieved from http://phonodevelopment.sites.olt.ubc.ca/afonolito/.   97 Bernhardt, B. M., Stemberger, J.P., Liu, C., Ávila, C., Carballo, G., Fresneda, D., Lléo, C., Muñoz, J., Pérez, D. & Zhao, J. (2015c). Acquisition of diphthongs in Mandarin and Granada Spanish. Paper presented at the International Symposium on Monolingual and Bilingual Speech 2015 (ISMBS 2015), Chania, Crete, Greece, September 7-10, 2015. Bernhardt, B. & Stoel-Gammon, C. (1994). Nonlinear phonology: introduction and clinical application. Journal of Speech and Hearing Research, 37(1), 123-144. Boersma, P. (2001). Praat, a system for doing phonetics by computer. Glot International, 5(9-10), 341-345. Borràs-Comes, J. & Prieto, P. (2014). The acquisition of coda consonants by Catalan and Spanish children: effects of prominence and frequency of exposure. Probus, 26(1), 59-82. Bosch, L. (1984). El desarrollo fonológico infantil: una prueba para su evaluación. [Child phonological development: a test for its evaluation]. In M. Siguan (Ed.), Estudios de psicología infantil (pp. 33-58). Madrid, Spain: Pirámide. Bosch, L., (2004). Evaluación Fonológica del Habla Infantil [Phonological Evaluation of Child Speech]. Barcelona: Elsevier Masson. Boyce, S. E. (2015). The articulatory phonetics of /r/ for residual speech errors. Seminars in Speech and Language, 36(4), 257-270. Boyce, S. E., Hamilton, S. M., & Rivera-Campos, A. (2016). Acquiring rhoticity across languages: an ultrasound study of differentiating tongue movements. Clinical Linguistics and Phonetics, 30(3-5), 174-201.   98 Bradley, T. G. (1999). Assibilation in Ecuadorian Spanish. In J. Authier, B. Bullock, & L. Reed (Eds.), Formal perspectives on Romance linguistics: selected papers from the 28th linguistics symposium on Romance languages (pp. 57-71). Amsterdam: John Benjamins. Broomfield, J. & Dodd, B. (2004). Children with speech and language disability: caseload characteristics. International Journal of Language and Communication Disorders, 39, 303-324. doi: 10.1080/13682820310001625589 Burrows, L. & Goldstein, B.A. (2010). Whole word measures in bilingual children with speech sound disorders. Clinical Linguistics and Phonetics, 24(4-5), 357-368. Calet, N., Mendoza, E., Carballo, G., Fresneda, M. & Muñoz, J. (2010). Test de Comprensión de Estructuras Gramaticales de 2 a 4 años [Test of the Comprehension of Grammatical Structures from 2 to 4 years of age]. Revista de Logopedia, Foniatría y Audiología, 30(2), 62–72. Campbell, T. F., Dollaghan, C. A., Rockette, H. E., Paradise, J. L., Feldman, H. M., Shriberg, L. D., Sabo, D. L. & Kurs-Lasky, M. (2003). Risk factors for speech delay of unknown origin in 3-year-old children. Child Development, 72(2), 346-357. Carballo, G. & Mendoza, E. (2000). Acoustic characteristics of trill productions by groups of Spanish children. Clinical Linguistics and Phonetics, 14(8), 587-601. Cataño, L., Barlow, J. A., & Moyna, M. I. (2009). A retrospective study of phonetic inventory complexity in acquisition of Spanish: implications for phonological universals. Clinical Linguistics and Phonetics, 23(6), 446-472. Chávez-Peón, M. E., Bernhardt, B. M., Adler-Bock, M., Ávila, C., Carballo, G., Fresneda, D., Lleó, C., Mendoza, E., Pérez, D., & Stemberger, J. P. (2012). A Spanish pilot   99 investigation for a cross-linguistic study in protracted phonological development. Clinical Linguistics & Phonetics, 26(3), 255-272. doi:10.3109/02699206.2011.608463. Chomsky, N. & Halle, M. (1968). The sound pattern of English. New York: Harper & Row. Cleland, J., Scobbie, J. M., & Wrench, A. A. (2015). Using ultrasound visual biofeedback to treat persistent primary speech sound disorders. Clinical Linguistics and Phonetics, 29(8-10), 575-597. doi: 10.3109/02699206.2015.1016188 Colantoni, L. (2006). Macro and micro sound variation and change in Argentine Spanish. In J. Toribio & N. Sarra (Eds.), Selected proceedings of the 9th Hispanic Linguistic Symposium (pp. 91-102). Somerville, MA: Cascadilla Press. Colantoni, L. (2008). Variación micro y macro fonética en español [Micro and macro phonetic variation in Spanish]. Estudios de fonética experimental, 17, 65-104 Damico, J. S. & Simmons-Mackie N. N. (2003). Qualitative research and speech-language pathology: a tutorial for the clinical realm. American Journal of Speech-Language Pathology, 12, 131-143. Diez-Itza, E. & Martínez, V. (2004). Las etapas tardías de la adquisición fonológica: procesos de reducción de grupos consonánticos [Late stages in phonological acquisition: processes of consonant cluster reduction]. Annuario de Psicología, 35(2), 177-202. Dinnsen, D. A., Chin, S. B., Elbert, M., & Powell, T. W. (1990). Some constraints on functionally disordered phonologies: phonetic inventories and phonotactics. Journal of Speech and Hearing Research, 33, 28–37.   100 Dodd, B., Holm, A., Crosbie, S. & McIntosh, B. (2006). A core vocabulary approach for management of inconsistent speech disorder. Advances in Speech-Language Pathology 8(3), 220-230. Dubasik, V.L. & Ingram, D. (2013). Comparing phonology of dyads of children with typical development and protracted development. Clinical Linguistics & Phonetics, 27(9), 705-719.  Dunn, L., Dunn, L. & Arribas, D. (2006). PPVT-III Peabody: Test de Vocabulario en Imágenes [Peabody Picture Vocabulary Test]. Madrid: TEA.  Eblen, R.E. (1982). A study of the acquisition of fricatives by three-year-old children learning Mexican Spanish. Language and Speech, 25(3), 201-220. Eddington, D. (2011). What are the contextual variants of /β ð ɣ/ in colloquial Spanish. Probus, 23, 1-19. Edwards, S. M. (1995). Optimal outcomes of nonlinear phonological intervention. Unpublished Master’s Thesis. Vancouver, Canada: University of British Columbia. Felsenfeld, S. & Broen, P. A. (1992). A 28-year follow-up of adults with a history of moderate phonological disorder: linguistic and personality results. Journal of Speech and Hearing Research, 35(5), 1114-1125. Felsenfeld, S. & Broen, P. A. (1994). A 28-year follow-up of adults with a history of moderate phonological disorder: educational and occupational results. Journal of Speech and  Hearing Research, 37(6), 1341-1353.   101 Flipsen, P. Jr. (2015). Emergence and prevalence of persistent and residual speech errors. Seminars in Speech and Language, 36(4), 217-223. doi: 10.1055/s-0035-1562905 Gibbon, F., Stewart, F., Hardcastle, W. J., & Crampin, L. (1999). Widening access to electropalatography for children with persistent sound system disorders. American Journal of Speech-Language Pathology, 8(4), 319-334. Gierut, J. A. (2001). Complexity in phonological treatment: clinical factors. Language, Speech, and Hearing Services in Schools, 32, 229-241. Goldstein, B. A. (2005). Substitution patterns in the phonology of Spanish-speaking children, Journal of Multilingual Communication Disorders, 3(3), 153-168. Goldstein, B. A. (2007). Phonological skills in Puerto Rican and Mexican Spanish-­‐speaking children with phonological disorders. Clinical Linguistics and Phonetics, 21(2), 93-109. Goldstein, B. A. & Cintrón, P. (2001). An investigation of phonological skills in Puerto-Rican Spanish-speaking 2-year-olds. Clinical Linguistics and Phonetics, 15, 343-361. Goldstein, B. A., Fabiano, L., & Iglesias, A. (2004). Spontaneous and imitated productions in Spanish-speaking children with phonological disorders. Language, Speech, and Hearing Services in Schools, 35(1), 5-15. Goldstein, B. A., & Iglesias, A. (1998). Phonological production in Spanish-speaking preschoolers. ERIC Institution of Education Sciences. ERIC number: ED424935. Goldstein, B. A., & Iglesias, A. (2001). The effect of dialect on phonological analysis: evidence from Spanish-speaking children. American Journal of Speech-Language Pathology, 10(4), 394-406.   102 Goldstein, B. A. & Pollock, K. E. (2000). Vowel errors in Spanish-speaking children with phonological disorders: a retrospective, comparative study. Clinical Linguistics and  Phonetics, 14(3), 217-234.  Goldstein, B. A. & Pollock, K. E. (2004). Vowel production in Spanish-speaking children with phonological disorders: dialect and sampling issues. Journal of Multilingual Communication Disorders, 2(2), 147-160. Gómez-Molina, J. R. & Gómez-Devís, B. (2008). Estudio sociolingüistico de la /d/ intervocálica en el español hablado de Valencia [Sociolinguistic study of intervocalic /d/ in Spanish spoken in Valencia]. Actas del XV Congreso Internacional de la Asociación de Lingüistica y Filología de la América Latina (ALFAL). Montevideo: CD. Gómez Fernández, D. (1998). El proceso de adquisición de los grupos consonánticos en los niños de la provincia de Sevilla [The process of consonant cluster acquisition in children in the province of Seville]. Cauce, 20-21, 623-702. Gómez Fernández, D. (2004). La adquisición de las líquidas en niños de Sevilla y su provincia (I): desde los doce a los treinta y seis meses [Acquisition of liquids in children of Seville and surrounding province (I): from 12 to 36 months]. Cauce, 27, 125-185. González, A. (1981). A descriptive study of phonological development in normal speaking Puerto Rican preschoolers. Unpublished doctoral dissertation. State College, PA: Pennsylvania State University. Grosjean, F. & Li, P. (2013). The psycholinguistics of bilingualism. Malden, MA: Wiley-Blackwell.   103 Harasty, J. & Reed, V. A. (1994). The prevalence of speech and language impairment in two Sydney metropolitan schools. Australian Journal of Human Communication Disorders, 22, 1-23. Hernández-Campoy, J. M. (2011). Variation and identity in Spain. In M. Díaz-Campos. (Ed.), The handbook of Hispanic sociolinguistics (pp. 704-727). West Sussex: Blackwell Publishing Ltd. Hochberg, J. G. (1988). Learning Spanish stress: Developmental and theoretical perspectives. Language, 64, 683-706. Howard, S., Wells, B., & Local, J. (2008). Connected speech. In M. Ball, M. Perkins, & S. Howard. (Eds.), The handbook of clinical linguistics (pp. 583-602). Malden, MA: Blackwell Publishing Ltd. Ingram, D. (2002). The measurement of whole-word productions. Journal of Child Language, 29(4), 713-733. doi: 10.1017/S0305000902005275 International Phonetic Association. (2015). International phonetic alphabet (revised to 2015). Retrieved from https://www.internationalphoneticassociation.org/sites/default/files/IPA_K  iel_2015.pdf  Jakobson, R. (1968). Child language, aphasia and phonological universals. The Hague, Netherlands: Mouton.  Jímenez, B. C. (1987). Acquisition of Spanish consonants in children aged 3-5 years, 7 months. Language, Speech, and Hearing Services in Schools, 18(4), 357-363.   104 Johnson, C. J., Beitchman, J. H., Young, A., Escobar, M., Atkinson, L., Wilson, B., Brownlie, E. B., Douglas, L., Taback, N., Lam, I. & Wang, M. (1999). Fourteen-year follow-up of children with and without speech/language impairments: speech/language stability and outcomes. Journal of Speech, Language and Hearing Research, 42, 744-760. Kaufman, A. & Kaufman, N. (2009). K-BIT: Test Breve de Inteligencia de Kaufman; Adaptación Española [K-BIT: Kaufman Brief Intelligence Test; Cordero, A. and Calonge, I., Spanish Adaptation]. Madrid: TEA.  Kehoe, M., Hilaire-Debove, G., Demuth, K. & Lleó, C. (2008). The structure of branching onsets and rising diphthongs: evidence from the acquisition of French and Spanish. Language Acquisition, 15(1), 5-57. doi: 10.1080/10489220701774229 Kenney, M. K., Barac-Cikoja, D., Finnegan, K., Jeffries, N. & Ludlow, C. L. (2005). Speech perception and short-term memory deficits in persistent developmental speech disorder. Brain and Language, 96, 178-190. doi: 10.1016/j.bandl.2005.04.002 Kirkpatrick, E. & Ward, J. (1984). Prevalence of articulation errors in New South Wales primary school pupils. Australian Journal of Human Communication Disorders, 12, 55-62. Klein, E. S. & Flint, C. B. (2006). Measurement of intelligibility in disordered speech. Language, Speech and  Hearing Services in Schools, 37, 191-199. Ladefoged, P. (1993). A course in phonetics. New York, NY: Harcourt Brace. Law, J., Boyle, J., Harris, F., Harkness A. & Nye, C. (2000). Prevalence and natural history of primary speech and language delay: findings from a systematic review of the literature. International Journal of Language and  Communication Disorders, 35(2), 165-188.   105 Lewis, B. A., Avrich, A. A., Freebairn, L. A., Hansen, A. J., Sucheston, L. E., Kuo, I., Taylor, H. G., Iyengar, S. K. & Stein, C. M. (2011). Literacy outcomes of children with early childhood speech sound disorders: impact of endophenotypes. Journal of Speech, Language and Hearing Research, 54, 1628-1643. Lewis, B. A., Freebairn, L. A., Hansen, A. J., Stein, C. M., Shriberg, L. D., Iyengar, S. K., & Taylor, H. G. (2006). Dimensions of early speech sound disorders: a factor analytic study. Journal of Communication Disorders, 39, 139-157. doi: 10.1016/j.jcomdis.2005.11.003 Lewis, B. A., Patton, E., Freebairn, L., Tag, J., Iyengar, S. K., Stein C. M., & Taylor, H. G. (2016a). Psychosocial co-morbidities in adolescents and adults with histories of communication disorders. Journal of Communication Disorders, 61, 60-70. Lewis, M. P., Simons, G. F., & Fennig, C. D. (2016b). Ethnologue: Languages of the World. Dallas, Texas: SIL International. Retrieved from http://www.ethnologue.com Lleó, C. (2002). The role of markedness in the acquisition of complex prosodic structures by German-Spanish bilinguals. International Journal of Bilingualism, 6(3), 291-313. Lleó, C. (2003). Prosodic licensing of codas in the acquisition of Spanish. Probus, 15, 257-281. Lleó, C. (2006). The acquisition of prosodic word structures in Spanish by monolingual and Spanish-German bilingual children. Language and Speech, 49(2), 205-229. Lleó, C. (2008). Research on first language acquisition of Spanish phonology. Studies in Hispanic and Lusophone Linguistics, 1(2), 349-382. Lleó, C., & Prinz, M. (1996). Consonant clusters in child phonology and the directionality of syllable structure assignment. Journal of Child Language, 23(1), 31-56. Loatman, C. (2001). Phonological patterns of two-year-old monolingual Mexican-American Spanish speakers. Unpublished Master’s thesis. Arizona State University.   106 Lousada, M., Jesus, L. M. T., Hall, A. & Joffe, V. (2014). Intelligibility as a clinical outcome measure following intervention with children with phonologically based speech-sound disorders. International Journal of Language and  Communication Disorders, 49(5), 584-601. Martínez, V. & Diez-Itza, E. (2012). Procesos de asimilación en las etapas tardías del desarrollo fonológico [Assimilation processes in the late stages of phonological acquisition]. Psicothema, 24(2), 193-198. Mason, G. (2015). Multisyllabic word production of school-aged children with and without protracted phonological development. Unpublished doctoral dissertation. Vancouver, Canada: University of British Columbia. Mason, G., Bernhardt, B. M., & Masterson, J. J. (2011). Variability in the production of multisyllabic words by 5-year-olds: the British Columbia, Canada studies. Paper presented at the International Child Phonology Conference, York; UK, June, 2011. McCarthy, J. J. (2007). What is Optimality Theory? Language and Linguistics Compass, 1(4), 260-291. Mercedes Pavez, M., Maggiolo, M., Penaloza, C., & Julia Coloma, C. (2009). Desarrollo fonológico en niños de 3 a 6 años: Incidencia de la edad, el genero y el nivel socioeconomico [Phonological development in children from 3 to 6 years old: By age, gender and socioeconomic status]. RLA: Revista De Linguistica Teorica y Aplicada, 47(2), 89-109. Miller, N. (2013). Review: measuring up to speech intelligibility. International Journal of Language and  Communication Disorders, 48(6), 601-612.   107 Miras Martínez, F. (1992). El desarrollo fonoarticulatorio del habla infantil [Phono-articulatory development in child speech]. Granada, Spain: Instituto de Estudios Almerienses. Oxford Spanish-English Dictionary. (2017). Oxford University Press. Retrieved from   https://es.oxforddictionaries.com. Pater, J. & Barlow, J. (2003). Constraint conflict in cluster reduction. Journal of Child Language, 30(3), 487-526. Pérez, D. (2013). Descripción de la fonología en niños con desarrollo fonológico prolongado, trastorno del desarrollo del lenguaje y síndrome de down [Description of the phonology of children with protracted phonological development, developmental language delay and Down syndrome]. Unpublished doctoral dissertation. Granada, Spain: Universidad de Granada. Ponce, P. V. (2014). Patrones fonológicos en niños de entre 3.0 y 5.11 años con desarrollo típico del lenguaje según el enfoque de la fonología no lineal y fonología natural [Phonological processes in children between ages 3;0 and 5;11 with typical language development following nonlinear and natural phonology approaches]. Master’s Thesis, Chile: Universidad de Valparaiso. Powell, T. W. (1996). Stimulability considerations in the phonological treatment of a child with a persistent disorder of speech-sound production. Journal of Communication Disorders, 29, 315-333. Preston, J. L., McCabe, P., Rivera-Campos, A., Whittle, J. L., Landry, E. & Maas, E. (2014). Ultrasound visual feedback treatment and practice variability for residual speech sound errors. Journal of Speech, Language, and Hearing Research, 57, 2102-2115. doi: 10.1044/2014_JSLHR-S-14-0031   108 Quilis, A. (1983). Frecuencia de los esquemas acentuales en español [Frequency of stress patterns in Spanish]. Estudios Ofrecidos a Emilio Alarcos Llorach, 5, 113–126. Quilis, A. (1997). Principios de fonología y fonética españolas [Principles of Spanish phonology and phonetics]. Madrid, Spain: Arco libros. Rafat, Y. (2015). The interaction of acoustic and orthographic input in the acquisition of Spanish assibilated/fricative rhotics. Applied Psycholinguistics, 36(1), 43-66. Raitano, N., Pennington, B., Tunick, R., Boada, R., & Shriberg, L. (2004). Pre-literacy skills of subgroups of children with speech sound disorders. Journal of Child Psychology and Psychiatry, 45, 821-835. Raymond, G. (2017). Consonant acquisition influenced by word stress and word position in preschool speakers of Granada Spanish. Unpublished graduating paper. Vancouver, Canada: University of British Columbia. Real Academia Española y Asociación de Academias de la Lengua Española. (2011). Nueva gramática de la lengua española. Fonética y fonología [New grammar of the Spanish language: phonetics and phonology]. Madrid, Spain: Espasa. Rissel, D. A. (1989). Sex, attitudes, and the assibilation of /r/ among young people in San Luis Potosi, Mexico. Language Variation and Change, 1(3), 269-283. Rose, Y. & MacWhinney, B. (2014). The PhonBank Project: Data and software-assisted methods for the study of phonology and phonological development. doi:10.1.1.725.8916. Rose, Y., MacWhinney, B., Byrne, R., Hedlund, G., Maddocks, K., O'Brien, P., & Wareham, T. (2006). Introducing Phon: a software solution for the study of phonological acquisition. Proceedings of the Annual Boston University Conference on Language Development, 489. NIH Public Access.   109 Samper Padilla, J. A. (2011). Sociophonological variation and change in Spain. In M. Díaz-Campo (Ed.), The handbook of Hispanic sociolinguistics (pp. 98-120). West Sussex: Blackwell Publishing Ltd. Samper Padilla, J. A. & Hernández-Cabrera, C. (1995). La variación de -/s/ en el español culto de Las Palmas de Gran Canaria [Variation in /s/ in Spanish of Las Palmas, Grand Canary Island]. Philologica Canariensia 2, 391-408. Schretlen, C. E. (2013). Prosodic structure patterns in multisyllabic word productions of Granada Spanish-speaking children with typical versus protracted phonological development. Unpublished Master’s Thesis. Vancouver, Canada: University of British Columbia.  Serra, M., Serrat, E., Solé, R., Bel, A. & Aparici, M. (2000). La adquisición del lenguaje [Language acquisition]. Editorial Ariel: España. Shriberg, L. D. & Kwiatkowski, J. (1982). Phonological disorders III: a procedure for assessing severity of involvement. Journal of Speech and Hearing Disorders, 47, 256-270. Shriberg, L. D. & Kwiatkowski, J. (1994a). Developmental phonological disorders I: a clinical profile. Journal of Speech and Hearing Research, 37, 1100-1126. Shriberg, L. D., Kwiatkowski, J. & Gruber, F. A. (1994b). Developmental phonological disorders II: Short-term speech-sound normalization. Journal of Speech and Hearing Research, 37, 1127-1150.   110 Skebo, C. M., Lewis, B. A., Freebairn, L. A., Tag, J., Avrich Ciesla, A., & Stein, C. M. (2013). Reading skills of students with speech sound disorders at three stages of literacy development. Language, Speech and Hearing Services in Schools, 44, 360-373. Spoor, J. (2016). A case study of a Canadian French-speaking child with protracted phonological development: Nonlinear and constraints-based phonological perspectives. Unpublished Master’s thesis. Vancouver, Canada: University of British Columbia. Stewart, J. M., Hester, E. J. & Taylor, D. L. (1986). Prevalence of language, speech and hearing disorders in an urban preschool black population. Journal of Childhood Communication Disorders, 9, 107-123. Stothard, S., Snowling, M., Bishop, D., Chipchase, B., & Kaplan, C. (1998). Language impaired preschoolers: a follow-up into adolescence. Journal of Speech, Language, and Hearing Research, 41, 407-418. Vivar Vivar, P. A. (2009) Evaluación de grupos consonánticos de ataque complejo en un grupo de niños de la ciudad de Concepción con prueba articulatoria CEFI [Evaluation of complex onset consonant clusters in a group of children of Concepción city with the CEFI test of articulation]. Onomázein, 20(2), 33-44.  Vivar Vivar, P. A., & León, H. (2009). Desarrollo Fonológico-Fonético en un grupo de niños entre 3 y 5,11 años [Phonological-phonetic development in a group of children between ages 3 and 5;11]. Revista CEFAC, 11(2). Widdison, K. A. (1997). Variability in lingual vibrants: Changes in the story of /r/. Language & Communication, 17(3), 187-193.   111 Warr-Leeper, G. A., McShea, R. S. & Leeper, H. A. Jr. (1979). The incidence of voice and speech deviations in a middle school population. Language, Speech and Hearing Services in Schools, 10, 14-20. Wren, Y. E., McLeod, S., White, P., Miller, L. L., Roulstone, S. E. (2013). Speech characteristics of 8-year-old children: findings from a prospective population study. Journal of Communication Disorders, 46, 53-69. doi: 10.1016/j.jcomdis.2012.08.008 Wren, Y. E., Miller, L. L., Peters, T. J., Emond, A. & Roulstone, S. E. (2016). Prevalence and predictors of persistent speech sound disorder at eight years old: findings from a population cohort study. Journal of Speech, Language, and Hearing Research, 59, 647-673. doi: 10.1044/2015_JSLHR-S-14-0282 Wren, Y. E., Roulstone, S. E. & Miller, L. L. (2011). Distinguishing groups of children with persistent speech disorder: findings from a prospective population study. Logopedics Phoniatrics Vocology, 37(1), 1-10. doi: 10.3109/14015439.2011.625973   112 Appendices  The following appendices include the wordlists and pronunciations at times 1 and 2 (Appendix 1), the scan form used to analyze time 1 (Appendix 2) and the spectrograms of the words including Emilio’s “irregular” /r/ at time 2 (Appendix 3).  Appendix A. Wordlists Orthography Target (age 4;8) Actual (age 4;8) Actual (age 7;6) English gloss perro ˈpero  ˈpelːo ˈpero dog hueso ˈ{g/}we{s/θ}o  ˈweθo ˈɡweso bone gato ˈgato  ˈʝato ˈɡato cat blanco ˈ{b/β}laŋko  ˈβant̪o ˈblaŋko white casa ˈka{s/θ}a  ˈnaθa ˈkasa house llave ˈd ͡ʒaβe  ˈd ͡ʒaβe ˈʤaβe key techo ˈtetʃ͡o  ˈtetʃ͡o ˈteʧo roof chimenea t ͡ʃimẽˈne.a  t ͡ʃiðĩˈɾe.a ʧimẽˈne.a chimney escalera e{s/h}kaˈleɾa  ehtaˈleɾa ̞ eskaˈleɾa staircase mesa ˈme{s/θ}a  ˈmeθa ˈmesa table silla ˈ{s/θ}iʝa  ˈθiʝa ˈsiʝa chair teléfono teˈlefono  teˈleθeno teˈlefono phone lámpara ˈlampaɾa  ˈlampala ˈlampaɾa lamp luz ˈlu{θ/}  aˈt ͡ʃu ˈluθ light cuadro ˈkwaðɾo  ˈwːaðo ˈkwaðɾo painting fuego ˈfweɣo  ˈfweβo ˈfweɣo fire abierta aˈ{β/b}jeɾta  aˈbjeta aˈβjeɾta open aire ˈaiɾe  ˈaðe ˈaiɾ̯e air   113 Orthography Target (age 4;8) Actual (age 4;8) Actual (age 7;6) English gloss azul aˈθu{l/}  aˈθu aˈθul blue jamón xaˈmõn  taˈmõn xaˈmõn ham uvas ˈuβa{̞s/}  ˈuβa ̞ ˈuβas grapes fruta ˈfɾuta{s/}  ˈfuta ˈfɾutas fruit fresa ˈfɾe{s/θ}a  ˈfeθa ˈfɾesa strawberry pan ˈpan  ˈpan ˈpan bread sed ˈ{s/θ}ɛ{ð/}  ˈθɛ ˈseð thirst leche ˈletʃ͡e  ˈletʃ͡e ˈleʧe milk chocolate t ͡ʃokoˈlate  t ͡ʃoloˈlate ʧokoˈlate chocolate hirviendo iɾˈβjend̪o  iˈβjen ̪ː o ixˈβjend̪o boiling llueve ˈd ͡ʒweβe  ˈʝweβe ˈʤweβe raining nieve ˈnjeβe  ˈnjeβe ˈnjeβe snow primavera pɾimaˈβeɾa  pimaˈðeɾa pɾimaˈβeɾa spring elefante eleˈfant̪e  eˈfan eleˈfant̪e elephant cocodrilo kokoˈðɾilo  toloˈliɾo kokoˈðɾilo crocodile jirafa xiˈɾafa  tiˈra̞fa xiˈɾafa giraffe pájaro ˈpaxaɾo  ˈfalalo ˈpaxaɾo bird pluma ˈpluma  ˈpuma ˈpluma feather jaula ˈxaula  ˈxwala ˈxau ̯la cage hipopótamo ipoˈpotamo  poˈpotamo ipoˈpotano hippo grande ˈ{g/ɣ}ɾand̪e  ˈɣand̪e ˈɡɾand̪e big dinosaurio dinoˈ{s/θ}auɾjo  dinuˈθauɾjo dinoˈsau ̯ɾjo dinosaur conejo koˈnexo  oˈneθo koˈnexo rabbit ratón raˈton  naˈton raˈton mouse zanahorias θanaˈoɾja{s/}  anaˈodja θanaˈoɾjas carrots caballo kaˈβaʝo  taˈβaʝo kaˈβaʝo horse estanque e{s/}ˈtaŋke  eˈtant̪i esˈtaŋke pond   114 Orthography Target (age 4;8) Actual (age 4;8) Actual (age 7;6) English gloss tortuga toɾˈtuɣa  toˈtuɾa toɾˈtuɣa turtle papá paˈpa  paˈpa paˈpa dad hermano eɾˈmãno  enˈmãno eʴˈmãno brother muñeca mũˈɲeka  m̩̩ːˈɲeta mũˈɲeka doll boca ˈboka  ˈopːa ˈboka mouth nariz naˈɾi{θ/}  naˈni naˈɾiθ nose brazo ˈbɾaθo  ˈbaθo ˈbɾaθo arm pierna ˈpjenna  ˈpjena ˈpjenna leg bailando baiˈland̪o  baˈland̪o bai ̯ˈ land̪o dancing oigo ˈoiɣo  ˈoðo ˈoiɣ̯o I hear pelo ˈpelo  ˈpelo ˈpelo hair gorra ˈgora  ˈbola ˈɡora cap pantalón pant̪aˈlon  tataˈlon pant̪aˈlon pants bloques ˈblokɛ{s/}  ˈbotɛ ˈblokes blocks guante ˈgwant̪e  ˈɡwant̪e ˈɡwant̪e glove juguete xuˈɣete  θuˈðete xuˈɣete toy suave ˈ{s/θ}waβe  ˈwːaβe ˈswaβe soft globos ˈgloβo{s/}  ˈloβo ˈɡloβos balloons guitarra giˈtara  tiˈtaŭla ɡiˈtara guitar toca ˈtoka  ˈtota ˈtoka plays foto ˈfoto  ˈfoto ˈfoto picture (le) regalo reˈɣalo  deˈɣaɾo t ͡ɬ:iɾeˈɣalo gift gracias ˈgɾaθja{s/}  ˈnat ͡ʃja ̞ ˈɡɾaθjas thank you fútbol ˈfubol  ˈfubol ˈfutβol soccer saltando θalˈtand̪o  θaˈtand̪o salˈtand̪o jumping martillo maɾˈtiʝo  taˈtiʝo maɾˈtiʝo hammer tres ˈtɾɛ{s/}  ˈtɛ ˈtɾes three   115 Orthography Target (age 4;8)  Actual (age 4;8) Actual (age 7;6) English gloss lápiz ˈlapi{θ/}  ˈlapi ˈlapiθ pencil dragón dɾaˈɣon  daˈðon dɾaˈɣon dragon dos ˈdɔ{s/}  ˈdɔ ˈdos two princesa pɾin ̪ˈ θe{s/θ}a  tin ̟̪ˈ θeθa pɾin ̪ˈ θesa princess cuatro ˈkwatɾo  ˈkwato ˈkwatɾo four bruja ˈbɾuxa  ˈbufa ˈbɾuxa witch triángulo ˈtɾjaŋ{ɣ/g}ulo  ˈtiː.ʔanuɾo ˈtɾjaŋɡulo triangle rojo ˈroxo  ˈɾoθo ˈt ͡ɬ:oxo red cruz ˈkɾu  ˈku ˈkɾuθ cross flecha ˈflet ͡ʃa  ˈfetja ˈfleʧa arrow chocando t ͡ʃoˈkand̪o  t ͡ʃoˈtand̪o ʧoˈkand̪o crashing ruido ˈrwiðo  ˈbwiɾo ˈɬ:wiðo noise llorando d ͡ʒoˈɾand̪o  d ͡ʒoˈɾand̪o ʤoˈɾand̪o crying hoy ˈoi  ˈoi ˈoi ̯ today playa ˈplaʝa  ˈpaʝa ˈplaʝa beach noche ˈnot ͡ʃe  ˈnot ͡ʃe ˈnoʧe night día ˈdi.a  ˈdi.a ˈdi.a day agua ˈaɣwa  ˈaɣwa ˈaɣwa water barco ˈbaɾko  ˈbatto ˈbaɾko boat hoyo ˈoʝo  ˈoʝo ˈoʝo hole veinte ˈ{b/β}eint̪e  ˈbeint̪e ˈβein̯t̪e twenty Paula ˈpaula  ˈpaːɾa, ˈwaða ˈpau ̯la (doll’s name) baño ˈbaɲo  ˈmaɲo, ˈbaɲo ˈbaɲo bath peine ˈp{ei/en}ne  ˈpenne, ˈpenne ˈpein̯e comb zapato θaˈpato  paˈpapo, aˈpato θaˈpato shoe reloj reˈl{ox/ɔ}  deˈlɔ, deˈðo reˈlox clock flor ˈfl{oɾ/ɔ}  ˈfɔ, ˈfo ˈflox flower   116    Orthography Target (age 4;8) Actual (age 4;8) Actual (age 7;6) English gloss pescado / pez {ˈpɛ/pesˈkaðo}  ˈpɛ, ˈpɛ pesˈkaðo fish clavo ˈklaβo  ˈklaβo key sombrero somˈbɾeɾo  somˈbɾeɾo hat Europa euˈɾopa  ˈeuopa  Europe Euro ˈeuɾo  ˈeuˠo  Euro   117  Appendix B. Scan analysis at time 1, age 4;8 (adapted from Bernhardt et al., 2016) Description of Actual Productions Forms used frequently by client Summary by word length 1 syllable (C)V(V), CVC, CCV(V)(C) Single C: WI, WF CC: WI Strength (Y)	  Need (Y) Needs WI CC (tres, flor) Most complex shape: CVC (pan) Not elicited: CVV(C), CG 2 syllables (C)VCV(V), CV.V, CVCVC, CVCCV, CVVC(C)V, VVCV Stress: Su, uS Single C: WI, WM, WF CC: WI, WM 2 CCs/word, CCs & WF C in word Strength (Y) Need (Y) Needs CC, 2 CCs, CC + WF C 3+ syllables uSu, uuS, uuSu, uSuu, Suu e.g. CVCVCCV, CVCVCVCV, CVCVCVVCVV Single C: WI, WM, WF CC: WI, WM CCG: WM 2-3 CCs/word, CCs & WF C in word Strength (Y) Need (Y) Needs CC, 2-3 CCs, CC + WF C Not elicited: CCC Comparison with Adult Target Pattern: Length, stress 1 syllable 2 syllables 3+ syllables Syllable deletion   Yes Syllable addition Yes   Stress shift   Yes More patterns, non-final weak syllables?  Yes Yes Pattern: CV Shape WI WM WF Cs added    Deletion, Single Cs (seen more in longer words) Yes Yes Yes Deletion in CC(C) Yes, often Yes, often  Overused Cs or Vs  Yes, often (t l ɾ ð)  Scan form page 3: word structure (red: absent or marginal, yellow: inconsistent)     118  Word-initial Word-medial Word-final Mostly match: Strength (80% or greater match) t d n f t ͡ʃ d ͡ʒ p b m t n ɲ β θ ʝ n Inconsistent: Partial strength (33% to 79% match) p b m θ l f t ͡ʃ l l Present but not for adult targets (includes segments also in matches) d b p m t n f β θ ʝ ɣ ɾ w(:) p t n f β θ ð tʃ͡ l r ̞ɾ ʔ ˠ  Tested but absent or very marginal (0% to 32%) k g x r k ð x ɣ r ɾ  Non-Spanish speech sounds w: ʔ r ̞ˠ  Frequent substitutions (potential default segments) t t l ɾ ð  Consonants inconsistent in one established word position p b m θ l x f t ͡ʃ l  Scan form page 4: singleton consonants     119 Word-initial cluster inventory: pl pɾ βl bɾ tɾ dɾ kl kɾ gl gɾ ɣɾ fl fɾ Word-medial cluster inventory: across syllables: Nasal.C, /l/.C(C), /ɾ/.C(C), [h].C(C) Word-medial, syllable initial: tɾ, ðɾ Pattern WI WM: across syllables WM: syllable-initial C1 deletion gl nt̪, ɾt, ɾβ, lt  C2 deletion fl, fɾ, pl, pɾ, bl, bɾ, βl, tɾ, dɾ, kɾ, gɾ, ɣɾ  tɾ, ðɾ C1 substitution    C2 substitution  hk>ht  C1Del/C2Sub pɾ>t ɾk>t(:) kokoˈðɾilo>toloˈliɾo? C2Del/C1Sub   kokoˈðɾilo>toloˈliɾo? Migration   kokoˈðɾilo>toloˈliɾo? Assimilation  nd̪>nn̪,̪ ɾm>nm kokoˈðɾilo>toloˈliɾo? Epenthesis    Other (coalescence, C1C2 substitution) pɾ>t, gɾ>n ŋk>nt, ŋɣ>n, nd̪>nn̪ ̪  Scan form page 5: consonant sequences Other information from scan page 5: Assimilation: teˈlefono>teˈleθeno, ˈfweɣo>ˈfweβo pant̪aˈlon>tataˈlon, ˈbɾuxa>ˈbufa pɾin ̪ˈ θeθa>tin ̪ˈ θeθa Metathesis: kokoˈðɾilo>toloˈliɾo xiˈɾafa>tiˈra̞fa ˈpaula>ˈwaða Dissimilation: ˈkaθa>ˈnaθa Coalescence: gɾ>n pɾ>t ðɾ>l  (L=Labial) (C=Coronal) (D=Dorsal) L-L C-C D-D L-C θaˈpato>paˈpapo C-L t ͡ʃimẽˈne.a>tʃ͡iðĩˈɾe.a D-L (probably segmental) L-D ˈboka>ˈopːa C-D koˈnexo>oˈneθo D-C (probably segmental) Other: [+nas]: ˈbaɲo>ˈmaɲo, raˈton>naˈton, naˈɾi>naˈni [+/-cont]: ˈpaxaɾo>ˈfalalo [+lat]: t ͡ʃokoˈlate>t ͡ʃoloˈlate, kokoˈðɾilo>toloˈliɾo [-vc]: maɾˈtiʝo>taˈtiʝo Sequences of consonants related to above   120 Adult feature Adult consonant Word-initial Word-medial Manner: Glides [-consonantal] j (h w)     [+vibrant] ([+rhotic]) (trill or tap) ɾ r  r>d,n r>l, ɾ>l,n,ð,ˠ [+lateral] l  l>ɾ,ð,tʃ͡ Nasals [+nasal] ([-continuant]) m n ɲ[ŋ]   m>t m>ð, n>ɾ Stops: [-continuant]([-nasal]) p b t d k g   b>m, g>ʝ,n, gɾ>n, k>n, p>w,f k>l Fricatives: [+continuant] (& [-son]) f [β] (θ) s [ð] (ʃ ʒ) x [ɣ]   x>t, θ>p (assimilation) ð>ɾ, ðɾ>l,  ɣ>ɾ, x>l Affricates: [-continuant, +continuant] tʃ ͡(dʒ ͡)    t ͡ʃ>tj Place: Labial p b m f (w)[β]   m>t (assimilation), (p>f), p>t, pɾ>t β>ð, f>θ, m>ð [+labiodental] f  f>θ Coronal [+anterior] t d n (θ) s [ð] l ɾ r  θ>p t>p [-anterior] ɲ j tʃ ͡(ʃ ʒ dʒ ͡)  l>t ͡ʃ, ɾ>ˠ [+grooved] s tʃ ͡(ʃ ʒ dʒ ͡)   t ͡ʃ>tj [-grooved] (θ) [ð] (t d l n)   l>t ͡ʃ Dorsal k g ɲ[ŋ] x [ɣ](w) j  g>ʝ,t,b, gɾ>n, k>n,t,l, x>θ,t ɣ>ɾ,ð,β, ŋɣ>n, k>l,t,p: x>l,θ,f Laryngeal: [+voiced] b d g β ð ɣ ( ʒ d ʒ ͡)   g>t  [-voiced] p t k f (θ) s tʃ ͡x (ʃ h)  p>w, k>n k>l,n, x>l [-spread glottis] p t k b d g β ð ɣ ( ʒ d ʒ ͡)  p>f  [+spread glottis] f s (θ) tʃ ͡ x (ʃ h)  θ>p, x>t  Singleton substitutions, scan page 6 Place defaults? [Coronal][+anterior]     121 Singleton Vowel Inventory /a/, /e/, /i/, /o/, /u/ Vowel Patterns of Difference from Adult Target e>o (1), e>i (2), o>u (1)  Phonological Processes vowel harmony  VV/V.V Accuracy ai (0/2), oi (1/1), ei (1/1), au (1/4), eu (2/2), ui (1/1), a.o (1/1) ia (2/3), ie (3/3), io (1/1), ue (3/3), ua (5/5), e.a (1/1), i.a (1/1) VV/V.V Patterns ai>a (2), au>a: (1), au>wa (2), ia>i:ʔa (1) deletion, substitution of V2; metathesis (e.g. jaula, au>wa) Vowels, scan page 7      122 Appendix C. Spectrogram of word with “irregular” /r/ at age 7;6                      t ͡ɬ:       i      ɾ       e       ˈɣ        a            l    o Spectrogram of Emilio’s production of (le) regalo [t ͡ɬ:iɾeˈɣalo]    123                ˈt ͡ɬ:          o      x     o Spectrogram of Emilio’s production of rojo [ˈt ͡ɬ:oxo]  Spectrogram of typical /r/ in female adult production of rojo [ˈroho]    124     ˈɬ:            w                i                    ð       o Spectrogram of Emilio’s production of ruido [ˈɬ:wiðo]   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0344008/manifest

Comment

Related Items