Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Temporal aspects of children’s production of consonant clusters Purves, Barbara Anne 1976

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

Item Metadata

Download

Media
831-UBC_1976_A6_7 P87.pdf [ 5.7MB ]
Metadata
JSON: 831-1.0093701.json
JSON-LD: 831-1.0093701-ld.json
RDF/XML (Pretty): 831-1.0093701-rdf.xml
RDF/JSON: 831-1.0093701-rdf.json
Turtle: 831-1.0093701-turtle.txt
N-Triples: 831-1.0093701-rdf-ntriples.txt
Original Record: 831-1.0093701-source.json
Full Text
831-1.0093701-fulltext.txt
Citation
831-1.0093701.ris

Full Text

TEMPORAL ASPECTS OF CHILDREN'S PRODUCTION OF CONSONANT CLUSTERS by BARBARA ANNE PURVES B.A.(Honors) Simon Fraser University, 1972 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE i n the D i v i s i o n of Audiology and Speech Sciences i n the Department of Paediatrics We accept t h i s thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA July,. 1976 Barbara Anne Purves, 1976 In p resent ing t h i s t he s i s in p a r t i a l f u l f i l m e n t o f the requirements f o r an advanced degree at the U n i v e r s i t y of B r i t i s h Columbia, I agree that the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r reference and study. I f u r t h e r agree that permiss ion fo r ex tens i ve copying of t h i s t he s i s f o r s c h o l a r l y purposes may be granted by the Head of my Department or by h i s r ep re sen ta t i ve s . It i s understood that copying or p u b l i c a t i o n of t h i s t he s i s f o r f i n a n c i a l gain s h a l l not be al lowed without my w r i t t e n permis s ion. Department of /^(^.c^'ruA^'aS^ /TL^V?/*a*i osr, J A^csf ^tx^c's The Un i v e r s i t y of B r i t i s h Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 Date pz ABSTRACT Th i s . i n v e s t i g a t i o n was concerned with reduction i n duration of consonants i n c l u s t e r s r e l a t i v e to duration i n unclustered contexts. S p e c i f i c a l l y , the i n v e s t i g a t i o n was designed: (a) to determine whether c e r t a i n consonants' durations were s i g n i f i c a n t l y reduced i n consonant c l u s t e r s i n adult speech, (b) to determine whether differences i n the amount of such reduction existed between age groups, and (c) to deter-mine whether a linear, trend c o r r e l a t i n g age with amount of consonant reduction existed. Five age groups with f i v e subjects i n each group repeated words with i n i t i a l consonants /s/, / f / , / l / , /w/, / s i / , / f l / , and /sw/. The age groups tested included five-year-olds, seven-year-olds, nine-year-olds, and eleven-year-olds, and an adult c o n t r o l group. Subjects' u t t e r ances were recorded and oscillograms of the recordings were measured for consonant duration. Analysis of durations indicated that / s / and /%/ durations were s i g n i f i c a n t l y reduced, i n c l u s t e r s by a l l age groups but no s i g n i f i c a n t d ifferences between groups existed i n amount of reduction of these con-sonants. Duration of /w/ was not s i g n i f i c a n t l y reduced by any age group duration of / l / was s i g n i f i c a n t l y changed by the f i v e - y e a r - o l d group only. This group lengthened duration of / l / i n c l u s t e r s . A trend anal-y s i s of age correlated with reduction of / l / demonstrated that amount of reduction of duration of /!/ increases l i n e a r l y with age between f i v e i i i and eleven years. Consonant reduction was discussed in terms of speech timing pro-grammes. Differences between age groups in consonant duration reduction were discussed in terms of processes observed in children's acquisition of consonant clusters, and possible differences in speech timing pro-grammes used by children and adults. TABLE OF CONTENTS Page LIST OF TABLES v i i i LIST OF FIGURES i x Section 1.0 INTRODUCTION 1 1.1 Introduction 1 1.2 Review of the L i t e r a t u r e 2 1.2.1 Temporal Organization of Adult Speech 2 1.2.2 Temporal Organization of Children's Speech 16 1.2.3 A c q u i s i t i o n of Consonant Clusters 26 1.2.4 Summary 32 1.3 Aims of the Investigation 32 2.0 METHOD 34 2.1 Subjects 34 2.2 Construction of Word L i s t s 35 2.3 Selection of a C a r r i e r Phrase 36 2.4 Data C o l l e c t i o n 39 2.4.1 Recording Environment 39 2.4.2 Recording Equipment 39 2.4.3 Screening Test Administration . 39 2.4.4 Test Word E l i c i t a t i o n 40 2.5 Data Analysis 40 i v V Section Page 2.5.1 Preparation of Tapes f o r Analysis 40 2.5.2 Instrumental Analysis 41 2.5.3 Segmentation of Utterances 41 2.5.4 Segmentation C r i t e r i a 43 2.5.4.1 P r i n c i p l e s of Segmentation 43 2.5.4.2 Vowels and Semi-vowel / r / 4 4 2.5.4.3 Semi-vowel /w/ and Resonant /!/ 4 4 2.5.4.4 F r i c a t i v e s 4 9 2.5.4.5 Voiceless Plosives 4 9 2.5.5 Measurement R e l i a b i l i t y . . . . . 4 9 2.5.6 Test Word Rejection 50 2.5.7 S t a t i s t i c a l Analysis 50 2.5.7.1 Analysis I 50 2.5.7.2 Analysis II 52 2.5.7.3 Analysis I I I 53 3.0 RESULTS 54 3.1 Rate Control Experiment 54 3.2 Duration Measurement Data 55 3.3 Results of Analysis I (a-d) 58 3.3.1 Analysis Ia . 63 3.3.2 Analysis Ib 63 3.3.3 Analysis Ic . 63 3.3.4 Analysis Id 64 v i Section Page 3.4 Results of Analysis II (a-d) 64 3.4.1 Analysis I l a 66 3.4.2 Analysis l i b 66 3.4.3 Analysis l i e 66 3.4.4 Analysis I Id 67 3.5 Results of Analysis III (a-d) 67 3.5.1 Analysis I l i a 69 3.5.2 Analysis I l l b 69 3.5.3 Analysis IIIc . 69 3.5.4 Analysis H i d 69 3.6 Summary of Main Points of Interest i n the S t a t i s t i c a l Analysis 70 3.7 Other Points of Interest i n the Data 70 3.8 Measurement R e l i a b i l i t y 71 4.0 DISCUSSION AND CONCLUSIONS 72 4.1 Discussion of Results 72 4.2 Theoretical Implications 76 4.3 Conclusion 81 SELECTED BIBLIOGRAPHY 83 APPENDICES A. Background Information About Subjects 86 B. L i n g u i s t i c Background of Subjects' Parents 88 C. Word L i s t s • 90 v i i Appendix Page D. Duration of Voiced and Voiceless /!/ and /w/ 92 E. Subject Duration Data 94 F. Summary Tables f o r Analysis I (a-d) 100 G. Homogeneous Subtests Determined by Newman-Keuls (Analysis I a-d) 102 H. Summary Tables f o r Analysis II (a-d) 104 I. Summary Tables f o r Analysis I II (a-d) 107 J. Graphic Representation of Linear Regres-sion of /!/ Duration 109 LIST OF TABLES Table Page 2.1 L i s t of Words used i n Data C o l l e c t i o n 37 2.2 Summary of Variables used i n S t a t i s t i c a l Analyses 51 3.1 Mean Durations f o r a l l Age Groups, / s / and / f / 56 3.2 Mean Durations f o r a l l Age Groups, / l / and /w/ 57 3.3 Analysis I (a-d): Summary of Analyses of Variance 61 3.4 Analysis I (a-d): Summary of Mu l t i p l e Range Tests (Newman-Keuls) 62 3.5 Analysis II (a-d) 65 3.6 Analysis III (a-d): F - r a t i o s f o r Trend C o r r e l a t i n g Age (5-11 years) with Reduction 68 v i i i LIST OF FIGURES Figure Page 2.1 Arrangement of Instrumentation 42 2.2 Segmentation of / s i / on Mingographic Trace 45 2.3 Segmentation of / f l / on Mingographic Trace 46 2.4 Segmentation of /sw/ on Mingographic Trace 47 3.1 Graphic Representation of Durations i n msec. (Adult, 11-year, and 9-year groups) 59 3.2 Graphic Representation of Durations i n msec. (5-year and 7-year groups) 60 ix ACKNOWLEDGEMENTS I wish to express my sincere thanks to Dr. John G i l b e r t , f o r h i s guidance and considered r e v i s i o n of a l l d r a f t s , and to Dr. David Ingram for h i s c r i t i c a l reading of the f i n a l d r a f t . Thanks are also due to Dr. Andre-Pie r r e Benguerel for h i s te c h n i c a l advice and to Dr. John Delack for h i s generous loan of recording equipment and for many h e l p f u l discussions. In addition, I would l i k e to thank Catherine Millway f or her t r a n s c r i p t i o n s of a l l tapes, S a l l y Spicer for scoring a r t i c u l a t i o n t e s t s , and Dr. Brenda Fraser, Ronnie S i z t o , and Adele Morton f o r s t a t i s t i c a l advice. A l l my subjects and t h e i r f a m i l i e s also deserve sincere thanks f o r t h e i r patience and co-operation. Above a l l , I would l i k e to thank my husband, Terry, f or understanding. x CHAPTER 1 INTRODUCTION 1.1 Introduction Investigations into a c q u i s i t i o n of consonant c l u s t e r s i n c h i l -dren's speech, while often noting processes i n d i c a t i v e of reorganization of c l u s t e r features into single rather than multiple segments, have generally not considered these processes i n terms of temporal organiza-t i o n of speech. Several investigations have indicated that timing con-t r o l i n c h i l d r e n , i n p a r t i c u l a r , timing control of speech, i s i n f e r i o r to that of the adult. Such r e s u l t s are suggestive of the s i g n i f i c a n c e of temporal constraints i n c h i l d speech. I t i s apparent that a thorough i n v e s t i g a t i o n of temporal organi-zation of speech i n both c h i l d r e n and adults i s c r u c i a l to an under-standing of children's speech production. Recent studies of such organizations have l a r g e l y focussed on the way i n which adult speech may be co-ordinated into temporal units both smaller and larger than s y l -l a b l e s . Evidence that such units e x i s t has to a c e r t a i n extent been provided by studies of c o - a r t i c u l a t i o n (e.g., Ohman 1967), which showed that a r t i c u l a t o r y movements for a given segment may be i n i t i a t e d during production of preceding segments. Studies of aspects of temporal organization i n children's speech have generally been r e s t r i c t e d to studies of v a r i a b i l i t y of durations and r e l a t i v e onset of a r t i c u l a t o r y gestures. While such studies are of 1 2 considerable value i n demonstrating differences i n timing control which may e x i s t between c h i l d r e n and adults, they do not i n any way elucidate compensatory techniques which c h i l d r e n could use to overcome the e f f e c t of t h i s poor c o n t r o l . Two processes noted frequently i n children's production of con-sonant c l u s t e r s which may be i n d i c a t i v e of such temporal compensatory techniques are s p l i t t i n g and d e l e t i o n . A c l e a r e r understanding of the way•in which such techniques may be used can only be obtained by consider-ing differences i n temporal organization i n c h i l d and adult speech. 1.2 Review of the L i t e r a t u r e 1.2.1 Temporal Organization of ;Adult Speech Several approaches to the problem of determining which factors control segment durations i n adult speech production are evident i n the l i t e r a t u r e . Studies have been conducted to determine whether speech gestures are i n i t i a t e d . i n accordance with a timing programme for a u n i t larger than a phone, or whether each gesture i s triggered by completion of the preceding one. Other studies have been concerned with: (a) the i n v e s t i g a t i o n of various components contributing to v a r i a b i l i t y of dura-t i o n of a given utterance repeated several times, and (b) the e f f e c t s of context on the duration of a speech segment. Some o r i g i n a l experiments conducted i n the timing of speech events are reported by Kozhevnikov and Chistovich (1965). In one experi-ment, t h e i r purpose was to determine whether i n i t i a t i o n of sequential a r t i c u l a t o r y units was triggered by completion of the preceding un i t , or whether t h i s i n i t i a t i o n proceeded according to a higher l e v e l timing 3 programme extending over a unit larger than the s y l l a b l e . They reasoned that i f the f i r s t case was true, then the t o t a l variance of the whole utterance.would be equal to or greater than the sum of the variances of the i n d i v i d u a l units, since the v a r i a t i o n s i n the i n t e r v a l s ought to accumulate. I f , on the other hand, the second case was true, then the sum of the variances of the component i n t e r v a l s should be greater than the variance of the t o t a l i n t e r v a l , t h e i r reasoning being that i n i t i a t i o n of a u n i t would proceed regardless of whether a r t i c u l a t i o n of the pre-ceding unit was completed. In an attempt to elucidate these two hypotheses, Kozhevnikov and Chistovich recorded several hundred tokens of a sentence repeated by two speakers, who attempted to minimize rate change. Analysis of the data showed that the v a r i a t i o n of the t o t a l i n t e r v a l was less than the sum of the v a r i a t i o n s of the component i n t e r v a l s , i . e . , s y l l a b l e s . Thus, Kozhevnikov and Chistovich concluded that speech gestures for successive units are executed according to a higher-order timing programme. Ohala (1970) describes t h i s higher-order timing programme as part of a timing-dominant system, i n which a rigorous time schedule i s main-tained, perhaps at the expense of pr e c i s e a r t i c u l a t i o n of the component segments. Thus, the f i r s t case discussed and rejected by Kozhevnikov and Chistovich i s described by Ohala as an articulation-dominant system i n which precise a r t i c u l a t i o n i s maintained, and no rigorous time schedule i s followed. Ohala uses t h i s terminology i n preference to the terms 'chain' and 'comb' suggested by Kozhevnikov and Chistovich because i m p l i c i t i n t h e i r models i s the assumption that feedback i s used i n the 4 chain model (otherwise equivalent to the articulation-dominant system), whereas feedback i s not used i n the comb model (p a r a l l e l e d i n Ohala's terminology by the timing-dominant system). Ohala points out that t h e i r experiments do not indicate whether feedback i s used, thus making i t impossible to d i f f e r e n t i a t e between the chain and comb models. Ohala presents several arguments to show that Kozhevnikov and Chistovich's r e s u l t s do not necessarily lead to acceptance of an exclu-s i v e l y timing-dominant system. One of the stronger arguments he presents suggests that use of the sum of the variances of component i n t e r v a l s pre-supposes that i t i s possible to determine with c e r t a i n t y what such i n t e r -vals are. Yet miscalculation of the true number of i n t e r v a l s could lead to erroneous acceptance of one system over the other. Kozhevnikov and Chistovich selected the s y l l a b l e as the appropriate i n t e r v a l ; Ohala points out the appropriate i n t e r v a l may i n f a c t be larger than t h i s . Another d i f f i c u l t y experienced i n i n t e r p r e t i n g the r e s u l t s of the Kozhevnikov and Chistovich experiment i s that t h e i r conclusion presupposes that the existence of one system implies the exclusion of the other. Ohala points out that a hybrid system could e x i s t , i . e . , the words of a sentence may be executed according to a timing-dominant system but each successive word may be triggered by completion of the preceding word. While Ohala suggests that existence of such a system may be obscured by miscalculation of the number of component i n t e r v a l s , an even greater d i f -f i c u l t y i s apparent, for i t i s unclear what the r e l a t i o n s h i p between the sum of the variances and the t o t a l variance would be i n the case of a hybrid system. I t seems l i k e l y that, even assuming one could be c e r t a i n 5 of the correct i n t e r v a l s i z e , the r e s u l t s could only exclude the p o s s i -b i l i t y of one of the two 'pure' systems; the dec i s i o n between the remain-ing pure system and the hybrid system could not be made with c e r t a i n t y . Reasoning that one cannot conclusively determine the size of com-ponent i n t e r v a l s , Ohala attempts to determine i n another way whether speech i s an articulation-dominant or timing-dominant system. He states that i f the variance i s constant for d i f f e r e n t - s i z e d i n t e r v a l s , i t follows that a timing-dominant system e x i s t s , whereas i f the variance i s proportional to the mean i n t e r v a l , one can conclude that an a r t i c u l a t i o n -dominant system e x i s t s . Only i n t e r v a l s thought to r e s u l t from neuro-muscular signals can be measured. Further, average speaking rate must be constant, or the variance of larger i n t e r v a l s w i l l tend to be larger, no matter which system holds. To solve t h i s problem, Ohala suggests a normalization procedure i n which only those sentences with the same or nearly the same duration are selected for an a l y s i s . An experiment was conducted i n which twenty repeated sentences with the same t o t a l duration were recorded. The variances of the measured i n t e r v a l s for the sentences were p l o t t e d against mean duration i n t e r v a l . The r e s u l t s indicated that variance increases with i n t e r v a l s i z e , thus favouring the articulation-dominant system. Ohala does not accept t h i s conclusion, however, pointing out that instantaneous rate within a sen-tence could vary, rendering the conclusion i n v a l i d . Another d i f f i c u l t y experienced with experiments of t h i s type i s that there i s no pr o v i s i o n for the existence of a hybrid system. I f the r e s u l t s showed a constant variance for a l l i n t e r v a l s i z e s , the conclusion would point to a 6 timing-dominant system; on the other hand, a l i n e a r increase with i n -creasing i n t e r v a l s i z e would point to an articulation-dominant system. A hybrid system i n which words are subject to a timing-dominant system, and successive words are subject to an articulation-dominant system, would also show greater variance f o r larger i n t e r v a l s , however, and could therefore not be conclusively demonstrated. The d i f f i c u l t i e s encountered i n determining whether speech i s a timing-dominant or articulation-dominant system were further investigated by Ohala i n a l a t e r paper (Ohala 1972). Ohala pointed out that the e f f e c t of measurement error i n segmenting an utterance i s to introduce negative c o r r e l a t i o n s between adjacent segments, reducing the value of the term representing the t o t a l variance of the utterance, and pos s i b l y causing i t to be less than the sum of the variances of component i n t e r v a l s . Further, Ohala demonstrated that to conclude that either an a r t i c u l a t i o n -dominant or timing-dominant system predominates i s c l o s e l y r e l a t e d to the number of sub-intervals, selected by the experimenter, i n a given utterance. A preliminary version of a t e s t of whether sensory feedback i s used to determine when a given gesture has been executed was conducted by Ohala (1972). I f sensory feedback i s necessary, then speech may be regulated i n a chain manner, whereas i f i t i s not necessary, a comb model may be appropriate.. In h i s experiment, Ohala i n t e r f e r e d with two speakers' auditory and t a c t i l e feedback channels; he then studied the e f f e c t s of t h i s interference on the speakers' temporal p r e c i s i o n . The r e s u l t s for the two speakers, however, are contradictory, and as Ohala 7 pointed out, there are too few controls i n the experiment to permit any r e l i a b l e d e c i s i o n to be made about "chain" versus "comb" models. There i s not conclusive evidence at t h i s time from feedback studies or comparative variance studies to support either a chain or a comb model of temporal organization of speech. On the basis of such studies i t i s not p o s s i b l e to claim with any c e r t a i n t y that i n i t i a t i o n of successive speech gestures i s regulated by a timing programme i n which precise a r t i c u l a t i o n i s secondary to.precise timing. Another approach to the problem of factors underlying segment durations has been investigated by A l l e n (1973). A l l e n has developed a model which also e x p l o i t s the v a r i a b i l i t y of i n t e r v a l durations across r e p e t i t i o n s of an utterance. The model postulates that there are three determinants of phonetic duration of any a r b i t r a r i l y defined i n t e r v a l or segment: (1) the speaker's intended, speech rate, which either lengthens or shortens a l l segments of an utterance; (2) the underlying phonological length of the segment, and (3) peripheral e f f e c t s of transmission time, associated with the nerves and muscles used i n production and bio-mechanical constraints, and measurement errors i n determining the end-points of a segment. In examining the v a r i a t i o n s i n segment durations over several r e p e t i t i o n s of the same utterance, A l l e n suggested that i t i s necessary to account for t h i s v a r i a b i l i t y i n terms of the three determinants of phonetic duration. V a r i a b i l i t y of durations must therefore be ascribed to: (.1) changes i n the speaker's o v e r a l l speech rate, (2) segment errors, i . e . , those made by the speech motor co n t r o l programme i n t r a n s l a t i n g 8 phonological, length into phonetic duration, and (3) an error factor asso-ci a t e d with, l o c a t i o n of segment endpoints. (These three factors con-t r i b u t i n g t o . v a r i a b i l i t y are also discussed by Ohala (1972), who described the second term as v a r i a b i l i t y due to neuro-muscular noise or randomness i n transmission of neural impulses.) This d e s c r i p t i o n i n d i -cates that these errors are, i n Allen's terminology, of peripheral rather than c e n t r a l o r i g i n . In f a c t , A l l en's discussion of the second term of v a r i a b i l i t y does not adequately describe the nature of t h i s e r r o r . A l l e n proposed that i t i s of ce n t r a l o r i g i n and varies proportionately with i n t e r v a l s i z e , i n d i c a t i n g that errors i n successive segments are uncor-r e c t e d . On the other hand,.Ohala stated that v a r i a b i l i t y due to neuro-muscular noise i s proportional to mean i n t e r v a l duration. Allen's model i n f a c t does not appear to account f o r v a r i a b i l i t y of t h i s kind. Because i t i s c l e a r l y stated that the f i r s t and second terms of the equation represent v a r i a b i l i t y due to ce n t r a l f a c t o r s , v a r i a b i l i t y due to neuro-muscular noise cannot be included i n these terms. The t h i r d term i n Allen's equation, however, i s a constant representing error i n l o c a t i n g segment endpoints. Ohala pointed out that v a r i a b i l i t y due to neuro-muscular noise varies with mean i n t e r v a l duration, so, i n accordance with t h i s f a c t , i t i s not possible to include t h i s v a r i a b i l i t y i n the t h i r d (constant) term. The only s o l u t i o n to t h i s problem appears to be to consider the second term i n Allen's model as v a r i a b i l i t y due to both c e n t r a l and peripheral e f f e c t s . In order to further examine the r o l e of the three postulated com-ponents i n segment duration v a r i a b i l i t y , A l l e n conducted an experiment 9 i n which nine normal speakers repeated two sentences t h i r t y times on one occasion, and eighty times on another. The speakers a l l attempted to keep speech rate constant throughout. Small rate v a r i a t i o n s are never-theless present and contribute s u b s t a n t i a l l y to o v e r a l l v a r i a b i l i t y . In order to eliminate these rate v a r i a t i o n s , A l l e n employed a variance s t a t i s t i c which considers only differences between successive utterances. The r a t i o n a l e f o r use of t h i s s t a t i s t i c was that an examination of the data revealed that although o v e r a l l f l u c t u a t i o n s i n rate were s u b s t a n t i a l , rate f l u c t u a t i o n s between successive r e p e t i t i o n s were n e g l i g i b l e . Only two sources of error therefore remained to be separated: second term (segment) errors and measurement error . A l l e n pointed out that measure-ment error associated with l o c a t i n g segment endpoints remains constant as the size of the i n t e r v a l to be considered increases. Segment v a r i -a b i l i t y , however, was postulated to increase as i n t e r v a l s i z e increased, decreasing the e f f e c t of measurement error . Consequently, by considering only the i n t e r v a l of the e n t i r e sentence, i t was possible to estimate the variance due to segment v a r i a b i l i t y . A l l e n predicted that i f segment timing errors are e s s e n t i a l l y uncorrelated, the second component of variance should follow a square root law, so that as the i n t e r v a l gets bigger, the second variance com-ponent should converge to a multiple of mean duration. To t e s t t h i s hypothesis, he p l o t t e d the variance of a given i n t e r v a l divided by mean duration against mean duration, f o r each of the r e p e t i t i o n tasks. The postulated convergence i s c l e a r l y v i s i b l e for eight of the eighteen r e p e t i t i o n tasks, and can.be estimated i n four of the eighteen tasks. 10 In the remaining s i x cases, the points do not converge to a constant. A l l e n c a l l e d the constant of p r o p o r t i o n a l i t y c alculated f or each speaker the ' r e l a t i v e variance.' This r e l a t i v e variance appears to be a useful measure of a r t i c u l a t o r y v a r i a b i l i t y present i n a given speaker's timing control mechanism. The term, however, must be interpreted as inclu d i n g both c e n t r a l and-peripheral v a r i a b i l i t y . I t i s impossible to determine the extent to which c e n t r a l timing control i s tested by t h i s measure. While the r e l a t i v e variance term i s useful as a measure of timing c o n t r o l , i t should not be used, as A l l e n suggested i t could be, to deter-mine temporal groupings within a sentence. A l l e n suggested that i f seg-ments are grouped i n the same temporal u n i t , they w i l l be negatively correlated, so that such a grouping could be detected i n a segment f o r which the r e l a t i v e variance i s less than the constant of r e l a t i v e v a r i -ance obtained, f o r that speaker. However, Allen's data c l e a r l y show that f o r short i n t e r v a l s , measurement error i s a substantial factor i n v a r i -ance, causing a wide spread of data points, which would make i t d i f f i c u l t to accurately determine the variance for a short i n t e r v a l . While to a c e r t a i n extent Allen's model c l a r i f i e s the r o l e of various factors i n determining segment durations, i t does not provide any information about the way i n which speech gestures may be integrated into h i g h e r - l e v e l units with a cohesive temporal programme. Co-a r t i c u l a t i o n studies, such as those by Ohman (1967) have provided e v i -dence suggestive of a r t i c u l a t o r y preprogramming. Ohman examined the co-a r t i c u l a t i o n of voiced stops with vowels i n VCV utterances. His r e s u l t s 11 indicated that the a r t i c u l a t o r y system prepares during a speech gesture for the succeeding gesture, suggesting that the gestures over which co-a r t i c u l a t i o n e f f e c t s are observed may be organized i n h i e r a r c h i c a l u n i t s . Evidence for such i n t e g r a t i o n i n the temporal dimension has been sought i n studies of negative c o r r e l a t i o n s between segments. Lehiste (1973) and Gregorski and Shockey (1971), for example, have extensively studied such c o r r e l a t i o n s , and have hypothesized temporal units on the basis of negative c o r r e l a t i o n s found i n the data. The value of such evidence, however, has been c r i t i c a l l y questioned by Ohala (1972) and Gregorski and Shockey (1972). F i r s t , the r e s u l t of measurement error i n l o c a t i n g segment endpoints i s the introduction of negative c o r r e l a t i o n s between segments. Second, i f the data are not normalized for rate d i f -ferences, the e f f e c t of such di f f e r e n c e s cannot be cal c u l a t e d . I f the data are normalized, negative c o r r e l a t i o n s are introduced. Third, nega-t i v e c o r r e l a t i o n s can be found between non-adjacent segments. The s i g -n i f i c a n c e of such c o r r e l a t i o n s i s not c l e a r , and u n t i l t h i s question has been answered, any i n t e r p r e t a t i o n of negative c o r r e l a t i o n s w i l l be inadequate. Although phonological units defined by a common timing programme and c o n s i s t i n g of several adjacent segments may e x i s t , the s i z e and organization of such units are as y e t l a r g e l y undetermined. I t has been suggested (Hawkins 1973) that units of d i f f e r e n t s i z e s may e x i s t at d i f -ferent h i e r a r c h i c a l l e v e l s i n the production and perception processes, and that these units may be organized i n d i f f e r e n t ways. Furthermore, units at the same l e v e l may be organized i n d i f f e r e n t ways. 12 The hypothesis that units at the same h i e r a r c h i c a l l e v e l may not be subject to the same organizational p r i n c i p l e s has been investigated by Haggard (1973a) who examined the durations of consonants i n c l u s t e r s r e l a t i v e to the durations of the same consonant i n CV- or -VC sequence. I n i t i a l and f i n a l two-consonant c l u s t e r s were studied: several hundred tokens of words containing the c l u s t e r s or one of the component con-sonants were recorded and the duration of each consonant was measured. Duration of each consonant i n a CCV- or a -VCC context was then compared to duration of the same consonant i n a CV- or-VC structure. Haggard's r e s u l t s demonstrated that durations of consonants i n c l u s t e r s are shorter than durations of consonants i n non-clustered contexts. On the basis of h i s r e s u l t s , Haggard has postulated a model for the c e n t r a l c o n t r o l of durations of consonants i n c l u s t e r s of d i f f e r e n t types, based only on those r e s u l t s which cannot r e a d i l y be explained with reference to aerodynamic and p h y s i o l o g i c a l f a c t o r s . The value of d i f f e r -e n t i a t i n g c l u s t e r types l i e s i n i t s relevance to a s p e c i f i c a t i o n of the way i n which a r t i c u l a t o r y commands might be temporally organized i n at le a s t one possible phonological unit, i . e . , the consonant c l u s t e r . Such information i s of value i n c l a r i f y i n g the factors underlying the a r t i c u -l a t i o n of co-ordinated sequences, and more s p e c i f i c a l l y , the phenomenon of c o - a r t i c u l a t i o n . (Haggard. 1973a, p. 20). Haggard has categorized c l u s t e r sequences as: (a) homorganic compatible sequences, e.g., / s i / , (b) homorganic incompatible sequences, e.g., /dr/, and (c) heterorganic or nonhomorganic sequences, e.g., /sw/. In a l a t e r study, however, Haggard pointed out that d i f f e r e n t speakers 13 employ d i f f e r e n t t a c t i c s for coping with i n c o m p a t i b i l i t y , so that the concept of homorganic i n c o m p a t i b i l i t y may be d i f f i c u l t to demonstrate as a relevant f a c t o r i n the performance of speakers (Haggard 1973b, p. 116). In addition to these sequences, e f f e c t s of sequence type, i , e . , closed-open versus open-closed, are considered to account for the r e s u l t s noted i n the data. Haggard concluded that i n general there i s more reduction i n duration where the component consonants of a c l u s t e r are nonhomorganic; an exception to t h i s g e n e r a l i z a t i o n i s that more reduction i s observed i n homorganic / s i / than i n nonhomorganic /sw/. I t appears, then, that i n nonhomorganic c l u s t e r s the periods for which target values for sep-arate a r t i c u l a t o r s are held can be overlapped and s t i l l held constant with a net reduction i n duration. This accounts f o r the reduction of the more 'open' consonant, though i n f a c t the 'closed' consonant was also shortened, suggesting the p o s s i b i l i t y of an a l t e r e d time programme i n the production of consonants i n c l u s t e r s . Further i n v e s t i g a t i o n of the model i s necessary before a s a t i s f a c t o r y explanation of t h i s r e s u l t can be proposed. As Haggard pointed out, the r e s u l t s of h i s experiment c l a r i f y the requirements from a model of temporal organization of consonant c l u s t e r s , but do not i n themselves s a t i s f y these requirements. His model has not been s e r i o u s l y pursued, however, so the organizational p r i n c i p l e s of timing in.consonant c l u s t e r s (beyond the suggestion that place of a r t i c u l a t i o n i s a relevant variable) remain unspecified. A serious model for temporal organization i n c l u s t e r s would, i n 14 any case, require a more d e t a i l e d s p e c i f i c a t i o n of c e r t a i n e f f e c t s than that j u s t described. As Hawkins has pointed out, Haggard's suggestion for a model does not adequately consider differences i n manner of a r t i c u -l a t i o n , except insofar as they contribute to aerodynamic and physio-l o g i c a l e f f e c t s on duration. There i s considerable evidence, however, that manner differences i n consonants can influence the duration of at l e a s t preceding vowels (Peterson and Lehiste 1960). I t has been shown also that a r t i c u l a t o r y e f f o r t can a f f e c t vowel length ( S l i s 1970). S l i s has demonstrated by means of electromyographic studies of l i p a c t i v i t y that i n Dutch the commands for tense consonants are more advanced i n time than those for lax consonants, thus reducing duration of the preceding vowel. Lehiste (1972) presented evidence i n d i c a t i n g that timing patterns i n perception and production are associated with major changes i n manner of a r t i c u l a t i o n . Her i n v e s t i g a t i o n of manner of a r t i c u l a t i o n indicated that resonant-vowel and vowel-resonant sequences d i d not constitute such major changes, at l e a s t f o r sequences with long vowels. I f t h i s hypo-thesis proves to be true, then Haggard's method of analyzing consonant cl u s t e r s with resonants independently of the accompanying vowel may not be v a l i d . Even though Lehiste's study was concerned with, units (words) which may be on a d i f f e r e n t h i e r a r c h i c a l l e v e l from c l u s t e r s , i f resonant duration i s more affe c t e d by vowel duration than by consonant e f f e c t s , the consonant c l u s t e r may prove not to be a cohesive u n i t at any l e v e l i n the possible hierarchy. The foregoing discussion of experiments i n timing control and 15 models sp e c i f y i n g relevant factors i n determining timing c o n t r o l reveals several i n t e r e s t i n g points to be considered. First, preprogramming of temporally organized units i s suggested by several f a c t s , such as co-a r t i c u l a t i o n and reduction of consonants i n c l u s t e r s ; existence of such preprogramming, however, i s as yet not conclusively determined because of weaknesses i n d i f f e r e n t methodological approaches. I t seems l i k e l y nonetheless that once the components of a unit of some undetermined s i z e have been selected, they are temporally reorganized so that i n i t i a t i o n of a gesture may occur before the preceding gesture has been completed. Second, there are many factors c o n t r o l l i n g duration of a given segment of the speech s i g n a l . Speech rate, phonological length of the underlying segment, and peripheral e f f e c t s of neural transmission time and biomechanical constraints a l l contribute to segment duration. In addition to these factors,, e f f e c t s of neighbouring segments, inc l u d i n g differences i n place and manner of a r t i c u l a t i o n , must be considered. S p e c i f i c a t i o n of the way i n which a l l these variables a f f e c t duration has been discussed, and i t has been shown that although s i g n i f i c a n t ad-vances i n t h i s area have been made, a model incorporating a l l these elements cannot as yet be postulated. Third, several methodological approaches have been evaluated. C o r r e l a t i o n a l analyses and comparative variance studies have been shown to involve several weaknesses which s e r i o u s l y reduce the impact of t h e i r results.. A useful measure, of timing control for a given speaker has been discussed, but i t has been shown that the measure includes both peripheral and c e n t r a l determinants of t h i s c o n t r o l . 16 While considerable progress has been made i n determining the nature of temporal organization of spoken language, there are s t i l l numerous unresolved questions. The r e s u l t s of the foregoing experiments are extremely useful i n d e l i n e a t i n g the questions to be answered, the most productive aspects to in v e s t i g a t e , and the requirements to be f u l -f i l l e d i n a s a t i s f a c t o r y model. 1.2.2 Temporal Organization, of Children's Speech Hawkins (1973) investigated temporal co-ordination of consonant c l u s t e r s i n seven c h i l d r e n ranging i n age from four to seven years. A l -though she did not endorse Haggard's model of co-ordination of consonant c l u s t e r s (discussed i n section 112.1), Hawkins used i t as a s t a r t i n g point for c l a s s i f y i n g and analyzing c l u s t e r sequences i n children's speech. She also used Haggard's data for adult speakers as a basis of comparison for determining whether the constraints i n children's speech are q u a l i -t a t i v e l y d i f f e r e n t to those i n adult speech, or whether they are quanti-t a t i v e l y d i f f e r e n t with a gradual decrease i n phone duration as a co r r e l a t e of increasing maturity. Hawkins c o l l e c t e d tokens of s p e c i f i e d monosyllabic words and measured the durations of consonants i n clustered and unclustered con-texts on a duplex o s c i l l o g r a p h i c trace. Two types of measurements were used for / f l / and / s i / sequences: unrefined measures, which assigned a l l aperiodic portions of the c l u s t e r to the f r i c a t i v e , and r e f i n e d measures, which assigned that aperiodic i n t e r v a l i n which i t appeared that the sides of the tongue were forming the / l / to the / l / segment. Hawkins then compared the clustered consonants' durations with the unclustered 17 consonants' durations. Several differences i n durational aspects were revealed i n a com-parison of children's and adults' data: (1) i n c l u s t e r s with i n i t i a l f r i c a t i v e s , e s p e c i a l l y those with prevocalic / l / , segments tended to be lengthened or at l e a s t l e s s abbreviated i n children's speech; (2) post-v o c a l i c / l / tended to be s i g n i f i c a n t l y longer before homorganic than before nonhomorganic consonants i n children's speech, whereas the reverse was true for adults; and (3) i n i t i a l c l u s t e r s of stop plus prevocalic / r / are probably lengthened also, e s p e c i a l l y i n homorganic c l u s t e r s . In i n t e r p r e t i n g her r e s u l t s , Hawkins suggested that the d i f f e r -ences i n durational changes observed between the adults' and children's data,•especially i n those c l u s t e r s i n v o l v i n g both pre- and post-vocalic / l / a r t i c u l a t i o n s r e f l e c t the d i f f i c u l t y c h i l d r e n have with / l / i n c l u s t e r contexts. S i m i l a r l y , the observation that more lengthening was seen i n / s i / c l u s t e r s than i n / f l / c l u s t e r s may be explained by r e f e r r i n g to the greater " d i f f i c u l t y " associated with the / s / phoneme; Hawkins claimed that evidence supporting such a notion i s to be found i n the order of a c q u i s i t i o n of these two phonemes, and i n the frequency of mis-a r t i c u l a t i o n of / s / i n children's speech. Hawkins further stated that on the basis of her r e s u l t s , homorganicity may not be as important an i n f l u -ence i n children's speech as i t i s i n the speech of an adult. She specu-lated that a r t i c u l a t o r y d i f f i c u l t y , p a r t i c u l a r l y i n terms of the d i f f i c u l t y encountered.by c h i l d r e n i n co-ordinating temporal regulation of i n i t i a t i o n of a r t i c u l a t o r y gestures, may be a more important influence i n children's speech. She hypothesized that evidence of t h i s d i f f i c u l t y 18 may be found i n the r e l a t i v e duration of the voi c e l e s s / l / i n t e r v a l i n cl u s t e r s i n children's speech, compared with that i n adults' productions of the same c l u s t e r s . Hawkins's study, however, d i d not investigate t h i s p o s s i b i l i t y . Another conclusion reached by Hawkins i s that there i s no convin-cing age trend i n changes i n phone duration which might indi c a t e that a decrease in. phone duration i n c l u s t e r s i s corr e l a t e d with increasing maturity. Hawkins c i t e d the work of Port and Preston (1974), who found no evidence of a refinement i n the p r e c i s i o n of voice onset time i n the production of a p i c a l stops i n the speech of c h i l d r e n ranging i n age from two to four-and-a-half years, as supportive evidence f o r the observed lack of age trend i n her own data, speculating that once acceptable though not nec e s s a r i l y mature forms of sounds have been produced f o r some time, the c h i l d ' s speech does,not noticeably change u n t i l a l a t e r time (Hawkins 1973, p. 205). Although Hawkins's r e s u l t s suggest several i n t e r e s t i n g d i f f e r -ences which might d i f f e r e n t i a t e c h i l d versus adult speech, there are some weaknesses i n the design and i n t e r p r e t a t i o n of her study which preclude u n c r i t i c a l acceptance of her conclusions. F i r s t , the differences between c h i l d and adult speech are based on r e s u l t s from two studies using quite d i f f e r e n t methods of data c o l l e c t i o n . Haggard obtained data f o r analysis i n l i s t form from speakers who read the supplied word l i s t s . Hawkins obtained her data by .means of a question and answer method, combined with some e l i c i t a t i o n by im i t a t i o n . Differences i n recording equipment also e x i s t . Second, Hawkins d i d not adequately control the f i n a l 19 consonant of the words i n which prevocalic consonants were analyzed. Differences i n manner of a r t i c u l a t i o n of f i n a l consonants have been shown to a f f e c t duration of at l e a s t the preceding vowel (Peterson and Lehiste 1960), and could possibly have some e f f e c t on duration of prevocalic con-sonants. Third, the s i z e of the population used i n Hawkins's study may have obscured any age trend which might e x i s t . Only seven subjects were involved: one four-year-old, two four-and-a-half-year-olds, two f i v e -year-olds, one six-year-old, and one seven-year-old. The lack of age trend i n Hawkins's data i s therefore not conclusive evidence that phone duration i n c l u s t e r s does not decrease with increasing maturity. In addition to these weaknesses i n the design of her study, Hawkins's i n t e r p r e t a t i o n of her r e s u l t s i s somewhat questionable. F i r s t , she stated that a r t i c u l a t o r y d i f f i c u l t y may be a relevant v a r i a b l e i n temporal co-ordination of consonant c l u s t e r s i n c h i l d speech. The con-cept of a r t i c u l a t o r y d i f f i c u l t y , however, i s not easy to demonstrate, and Hawkins offered l i t t l e evidence supporting the hypothesis that /!/ i s p a r t i c u l a r l y d i f f i c u l t for c h i l d r e n . In f a c t , i n a study by Locke (1972), i t was shown that, on the basis of subjective judgements by adult speakers, /w/ i s more d i f f i c u l t to produce than / l / . This r e s u l t does not c o r r e l a t e well with the a c q u i s i t i o n schedule found by other experimenters (e.g., Templin 1957). Thus, there appears to be at l e a s t some discrepancy between d e f i n i t i o n s of a r t i c u l a t o r y d i f f i c u l t y based on age of a c q u i s i t i o n , which might be supposed to assume that e a r l i e r a c q u i s i t i o n indicates an easier sound, and d e f i n i t i o n s based on subjec-t i v e judgements of adult, speakers. Even i f such discrepancy d i d not 20 e x i s t , these types of d e f i n i t i o n s are not p a r t i c u l a r l y useful i n that they do not explain why sounds are more ' d i f f i c u l t , ' i . e . , acquired e a r l i e r , or judged d i f f i c u l t by adult speakers. One could argue that acceptable evidence for a r t i c u l a t o r y d i f f i c u l t y can only be provided by c a r e f u l studies of the degree of a c t i v i t y of a r t i c u l a t o r s required for each sound. As yet, such evidence i s not a v a i l a b l e for a l l sounds, even i n adult speech. Even i f one could accept Hawkins's unsupported notion of a r t i c u -l a t o r y d i f f i c u l t y of the component consonants of c l u s t e r s , i t i s probably misleading to consider the d i f f i c u l t y of each consonant as a separate e n t i t y . Instead, one should consider the d i f f i c u l t y encountered i n exe-cuting the gestures required for the e n t i r e consonant complex. Thus the greater lengthening associated with / s i / c l u s t e r s , compared with / f l / c l u s t e r s , may be a r e s u l t of the d i f f i c u l t y encountered i n co-ordinating the gestures for the c l u s t e r , rather than a r e s u l t of the greater d i f -f i c u l t y of / s / . In f a c t , evidence of the greater d i f f i c u l t y of / s / , sup-posedly provided by the observation that i t i s acquired l a t e r , i s not conclusive: Ingram (1975.), i n reviewing several diary studies, demon-strated that / s / and ft/ are acquired i n d i f f e r e n t orders by d i f f e r e n t subjects."'" In any case, on the basis of Hawkins's and Haggard's data, i t i s not at t h i s time necessary to account for differences between children's and adults' speech i n co-ordination of / f l / and / s i / c l u s t e r s , Cf. Nelson's (1973) monograph on t a l k i n g s t y l e s for a more thoroughgoing examination of rates of a c q u i s i t i o n which apply probably at the phonetic l e v e l . 21 as the adult data for / f l / c l u s t e r s are incomplete, and there may not i n fa c t be any d i f f e r e n c e between the two groups. If differences do not e x i s t , one could conclude a r t i c u l a t o r y d i f f i c u l t y i s a relevant v a r i a b l e i n adults' temporal organization, or that the place and manner of a r t i c u -l a t i o n variables a f f e c t i n g adult production also provide the explanation f o r the children's data. Hawkins's study, i n spite of the weaknesses discussed, does indicate that there may be durational r e l a t i o n s h i p s of timing c o n t r o l i n children's speech d i f f e r e n t to those i n adults' speech. Supportive e v i -dence for t h i s hypothesis i s found i n a study by Tingley and A l l e n (1974). Tingley and A l l e n studied the extent to which speech timing control improves with age i n four groups: five-year-olds, seven-year-olds, nine-year-olds, and eleven-year-olds. The measure of timing control used was Allen's r e l a t i v e variance term discussed i n section 2.1 above. In addition to measuring the r e l a t i v e variance of repeated utterances, a measure of r e l a t i v e variance of the i n t e r v a l s between taps tapped out for two minutes by each subject was made. Tingley and A l l e n hoped to demonstrate by comparing speech r e l a t i v e variance scores and tapping r e l a t i v e variance scores that timing c o n t r o l extends over a l l motor s k i l l s and i s not s p e c i f i c to speech. The r e s u l t s of Tingley and Allen's study showed that consistency i n timing control does improve with age, at l e a s t i n the speech task. Further, the r e l a t i v e variance scores f o r the speech task for the nine-year-old and eleven-year-old groups f a l l within the range of scores found f o r adults i n Allen's e a r l i e r study (1973). Evidence for the 22 tapping scores i s l e s s c l e a r , as data from only nine of the twenty sub-jec t s were measurable. A s i g n i f i c a n t d i f f e r e n c e between the f i v e - and eleven-year-old groups was found for the tapping scores, i n the same d i r e c -t i o n as that of the speech scores. However, no obvious r e l a t i o n s h i p was observed between the speech and tapping r e l a t i v e variance scores for each subject. Though the data do not provide conclusive evidence for a common timing-control mechanism, they do not contradict the p o s s i b i l i t y . That timing control improves with age i s not a s u r p r i s i n g con-clusion; as Tingley and A l l e n point out, neurological development and improved motor s k i l l s p r e d i c t t h i s r e s u l t . I t i s unclear, however, what the mechanism underlying t h i s control i s . In Allen's study (1973), i t was implied that the error measured by the r e l a t i v e variance term was caused by errors i n a "clock mechanism" responsible for t r a n s l a t i n g phonological length into phonetic duration. In section 1.2.1 above, however, i t was pointed out that the r e l a t i v e variance term also must measure v a r i a b i l i t y due to randomness i n transmission of neural impulses. The r e s u l t s of Tingley and Allen's study showing that a timing control mechanism i s not necessarily speech-specific, could i n d i c a t e that the variance measured by the r e l a t i v e variance s t a t i s t i c i s more af f e c t e d by randomness i n transmission of neural impulses than by a speech-specific error i n t r a n s l a t i n g phonological length into phonetic duration. Further evidence of development of timing control i n c h i l d r e n can be found i n studies of voice onset time. Port and Preston (1974) investigated developmental aspects of p r e c i s i o n of voice onset time i n the speech of one two-year-old, one three-and-a-half-year-old, and two 23 four-and-a-^ half-year-olds. They measured the i n t e r v a l s between the onset of v o i c i n g and the release of a p i c a l stops i n the spontaneous utterances of the subjects. Their r e s u l t s demonstrated that although the adult value of the voice onset time c h a r a c t e r i z i n g 'voiced' (or short l a g stops), such as /d/ i s acquired very e a r l y (by three years of age; G i l b e r t 1976), the adult value for the voiceless stop was not reached even by age four-and-a-half years. Instead, there was a wide scatter of data points about the point marking the adult value on the temporal dimension of voice onset time. Port and Preston proposed that there are d i s t i n c t p h y s i o l o g i c a l mechanisms underlying the production of voiced and v o i c e l e s s stops which are characterized not only by d i f f e r e n c e i n r e l a t i v e timing between events, but also by d i f f e r e n t events. The production of stops described by Lisker and Abramson (1965) as short v o i c i n g lag stops (such as /d/ i n English) i s easier to produce i n that the timing of the onset of v o i c i n g following the release of the stop i s determined by aerodynamic f a c t o r s , rather than by a timing programme c o n t r o l l e d by the speaker. Conse-quently, to produce the correct short v o i c i n g lag stop, the speaker needs to i n i t i a t e only one a r t i c u l a t o r y gesture, i . e . , release of the stop. The vocal folds can be held i n p o s i t i o n for any unspecified time before the release of the stop, and they w i l l s t i l l only v i b r a t e a f t e r the stop has been released. For production of stops with long v o i c i n g l a g , how-ever, two a r t i c u l a t o r y gestures must be co-ordinated by an accurate timing programme. F i r s t , the stop must be released before the vocal folds are approximated; then, the vocal folds must be approximated at a 24 s p e c i f i e d time a f t e r the release of the stop, i f the adult value of voice onset time i s to be obtained. Young c h i l d r e n are presumably not capable of the precise timing control required to produce stops with the narrowly s p e c i f i e d adult values. Port and Preston, i n reviewing t h e i r r e s u l t s , pointed out that there was no age trend observable i n t h e i r data. They showed that the values obtained for the two-year-old were s i m i l a r to those for the four-and-a-half-year-olds, i n d i c a t i n g that there was possibly not a gradual approximation to the adult values. The l i m i t e d s i z e and age range of the population studied may have obscured any age trend which might never-t h e l e s s e x i s t . G i l b e r t (1976) suggested i n h i s study of voice onset time i n s i x three-year-olds that, since the adult values for / t / at l e a s t have not been, acquired at that age, a c r o s s - s e c t i o n a l analysis of VOTimes for / t / at ages up to nine years (when a r t i c u l a t o r y mastery might be assumed to have been achieved) could give i n s i g h t s into the c h i l d ' s emerging p h y s i o l o g i c a l mechanism for speech production. Such a cr o s s - s e c t i o n a l analysis could also reveal a gradual approximation of the range of voice onset time values for / t / to that observed i n adult speech. Evidence that a gradual approximation to adult values for voice onset times does e x i s t i s provided i n an experiment by Eguchi and Hirsh (1969). They studied voice onset time, measured as the i n t e r v a l between pl o s i v e release and subsequent voiced sounds i n the words .'blue, 1 'pen,' and ' t a l l , ' i n order to determine whether there e x i s t s a developmental trend with respect to t h i s temporal feature. The population used was 25 comprised of eighty-four subjects, ranging i n age from three to t h i r t e e n years, and including an adult control group. Each age group contained f i v e or s i x subjects. The t e s t words were e l i c i t e d i n sentences repeated f i v e times by each subject. The r e s u l t s obtained by Eguchi and Hirsh showed no s i g n i f i c a n t d i f f e r e n c e between age,groups f o r the mean values of each time i n t e r v a l . Intra-subject v a r i a b i l i t y , however, decreased systematically with age, reaching a minimum value at seven to eight years which corresponds to the adult value. This decrease i n int r a - s u b j e c t v a r i a b i l i t y demonstrates that there i s an age trend i n acquiring the adult timing p r e c i s i o n needed i n production of the long v o i c i n g l a g stops, e.g., / t / , described by Lisker and Abramson (1965). Further, the age at which adult v a r i a b i l i t y values are achieved corresponds reasonably well with Tingley and Allen's r e s u l t s . Several i n t e r e s t i n g aspects of development of timing c o n t r o l i n chi l d r e n have emerged i n the preceding discussion. Evidence of increas-ing time co n t r o l with age has been presented, and i t i s apparent that timing .control observed i n chi l d r e n of eight or nine years i s probably comparable to that demonstrated i n adults. From the evidence presented, i t seems probable that development i n timing control i n speech production i s p a r a l l e l e d i n development of timing control of other motor s k i l l s . The study of reduction of consonant c l u s t e r s i n the speech of child r e n also r a i s e s some i n t e r e s t i n g questions f o r consideration. Although differences between children's and adults' co-ordination of con-sonant c l u s t e r s were observed, convincing explanations underlying these 26 differences have yet to be proposed. The p o s s i b i l i t y that children's co-ordination of c l u s t e r s i s a f f e c t e d by t h e i r demonstrated i n a b i l i t y to p r e c i s e l y co-ordinate a r t i c u l a t o r y gestures must be considered. I t i s also possible that p r i n c i p l e s permitting co-ordination of consonant c l u s t e r s into a u n i f i e d complex have not been acquired by young c h i l d r e n , so that, even i f they had the necessary timing c o n t r o l , they would not be capable of i m i t a t i n g the adult's reduced form of the c l u s t e r . I t i s l i k e l y that timing control can account for greater v a r i a b i l i t y i n children's productions but cannot account f o r d i f f e r e n t p r i n c i p l e s of organization which may be found i n children's speech. Corroborating evidence for t h i s hypothesis i s found i n Eguchi and Hirsh's study which shows that c h i l d r e n have approximately the same target value for voice onset time as adults. One would expect to f i n d reduction i n consonant c l u s t e r s , accompanied by more v a r i a b i l i t y , i n children's speech, i f timing control were t o t a l l y responsible for d i f f e r e n c e s . That one also finds less general reduction i s suggestive of d i f f e r e n t organizational p r i n c i p l e s i n children's speech. Further i n v e s t i g a t i o n of t h i s p o s s i -b i l i t y requires a study of theories of a c q u i s i t i o n and underlying repre-sentations of consonant c l u s t e r s . 1.2.3 A c q u i s i t i o n of Consonant Clust e r s : Developmental data on the a c q u i s i t i o n of consonant c l u s t e r s i s not r e a d i l y a v a i l a b l e from most studies of phonological development. One of the most s i g n i f i c a n t reasons for t h i s lack i s that consonant c l u s -ters are acquired over a long period of time; developmental data therefore should be obtained from c h i l d r e n across several age l e v e l s . 27 As Ingram (1975) has pointed out, there are two basic methods of study of phonological a c q u i s i t i o n : diary studies and large sample studies, both of which are generally designed to provide an overview of a l l the sounds a c h i l d uses and are consequently often d e f i c i e n t i n pro-v i d i n g s u f f i c i e n t information on more s p e c i f i c sounds, such as consonant c l u s t e r s . Ingram discussed, the d i f f i c u l t i e s inherent i n each of these approaches. Diary studies usually c o l l e c t data from only one c h i l d , who may or may not be t y p i c a l of other c h i l d r e n i n h i s a c q u i s i t i o n of i n d i -v i d u al sounds. Further, these studies usually extend over a short period of time only, so the sounds may not be observed to t h e i r point of a c q u i s i t i o n . Ingram attempted to p a r t i a l l y compensate for these weaknesses i n his study of the a c q u i s i t i o n of f r i c a t i v e s by comparing several d i a r i e s to ascertain general patterns. While t h i s method i s more r e l i a b l e than analyzing data from only one diary, i t s t i l l cannot compensate for other weaknesses i n diary studies. Methods of data c o l -l e c t i o n i n v o l v i n g frequency of data t r a n s c r i p t i o n , type of t r a n s c r i p t i o n employed, and method of e l i c i t a t i o n may be expected to vary with each study, thereby reducing the value of cross-study comparisons. In any case, a cross-study comparison of a c q u i s i t i o n of consonant c l u s t e r s i s at t h i s point not a v a i l a b l e , and so the data provided by diary studies i s a d d i t i o n a l l y subject to the weaknesses discussed by Ingram. Large sample studies (e.g., Templin 1957; Wellman 1936) are also not e n t i r e l y r e l i a b l e . Ingram analyzed i n considerable d e t a i l the d i f -f i c u l t i e s involved i n such studies (Ingram 1976). Data i s usually c o l -l e c t e d from large groups of c h i l d r e n selected from a cross-section of 28 ages; age norms for the point at which sounds are acquired are then given, usually based on an i m i t a t i o n task and some a r b i t r a r y c r i t e r i o n of acqui-s i t i o n (75 percent of the population i s used i n Templin's study). Ingram pointed out that the norms are questionable since a c h i l d ' s a b i l i t y to produce a sound may vary according to the word used to t e s t that a b i l i t y : Templin's study tested each sound i n only one word. Ingram presented convincing evidence that i m i t a t i o n may y i e l d d i f f e r e n t r e s u l t s to those obtained i n spontaneous production tasks. Further, Templin scored responses as r i g h t or wrong only, obscuring developmental trends i n the gradual a c q u i s i t i o n of i n d i v i d u a l sounds. C l e a r l y , the evidence provided by large sample studies does not adequately r e f l e c t the development of correct production of sounds. Certain trends i n a c q u i s i t i o n of consonant c l u s t e r s can be noted nonetheless, even though they have not yet been adequately tested and documented. Reduction of consonant c l u s t e r s , r e f e r r i n g to the process i n which one consonant of a c l u s t e r i s deleted i s one example. Splitting i s another process used, although very r a r e l y (Olmsted 1971); the c l u s t e r i s s p l i t by i n s e r t i o n of a schwa vowel between the two consonants. The order of the reduction and s p l i t t i n g processes i s apparently f i x e d : i f a c h i l d uses both processes, s p l i t t i n g always appears l a t e r than reduction (Kornfeld 1971). Reduction of consonant c l u s t e r s appears to be a systematic pro-cess by which c e r t a i n consonants .in a c l u s t e r are always deleted. Though few studies have s p e c i f i c a l l y investigated the rules governing t h i s dele-t i o n , a few trends can be stated on the basis of data i n Kornfeld's 29 (1971a,b) and Menyuk and K l a t t ' s (1968) experiments and of personal obser-vation. In a stop-resonant c l u s t e r , i t i s the resonant, / l / or / r / , which i s always deleted, so that / b l u / becomes /bu/ and / t * A k / becomes / t A k / . In a f r i c a t i v e - s t o p c l u s t e r , i t i s the f r i c a t i v e which i s deleted, so that /step/ becomes /tep/. In a fr i c a t i v e - r e s o n a n t c l u s t e r , i t i s the resonant which i s deleted, so that / s l i p / becomes / s i p / . In these examples, i t i s always the ontogenetically l a t e r sound.which i s deleted. I t i s d i f f i c u l t , i f not impossible, to f i n d examples i n which t h i s p r i n c i p l e does not hold. The explanation of why c l u s t e r s are reduced and why c e r t a i n con-sonants are cons i s t e n t l y deleted i s by no means cl e a r . I t i s a l i k e l y p o s s i b i l i t y that ' l a t e r ' sounds are deleted because the c h i l d finds them d i f f i c u l t to co-ordinate into a consonant sequence, i n that they are inherently more " d i f f i c u l t . " Unfortunately, c r i t e r i a determining which sounds are d i f f i c u l t are, as mentioned i n section 1.2.2, not obvious at t h i s time. Locke (1972), f or instance, asked adults to judge which sound (presented i n a p a i r of sounds) was the more d i f f i c u l t i n terms of muscular e f f o r t . His r e s u l t s c o r r e l a t e d quite well with the observed sequence of a c q u i s i t i o n of sounds. I t i s therefore possible that 'ease' of a r t i c u l a t i o n i s measured.in terms of muscular e f f o r t i n production, and that sounds acquired l a t e r i n time require more muscular e f f o r t , so they are correspondingly more d i f f i c u l t . Unfortunately, i t i s not at a l l c e r t a i n that a subjective judgement of muscular e f f o r t i s not affected i n the adult's case by frequency of occurrence of a sound; cer-t a i n a r t i c u l a t o r s are probably more frequently used than others, and 30 judgement may be aff e c t e d by t h i s . Locke's r e s u l t s have not thus far been corroborated by electromyographic data. Port and Preston's (1974) r e s u l t s suggest the p o s s i b i l i t y that more d i f f i c u l t y i s encountered i n those sounds which require more precise r e l a t i v e timing of onset of a r t i c u l a t o r y events occurring i n the speech t r a c t . I t i s possible that the.sounds deleted i n consonant c l u s t e r reduc-t i o n are those re q u i r i n g the highest degree of temporal p r e c i s i o n . Again, empirical.evidence i n support of t h i s hypothesis i s not a v a i l a b l e . Two studies investigated the p o s s i b i l i t y that consonant c l u s t e r s are at some point i n the developmental sequence of a c q u i s i t i o n treated by the c h i l d as a singl e segment rather than as a segment sequence. Menyuk and K l a t t (1968) c o l l e c t e d data from a c h i l d at weekly i n t e r v a l s from age eighteen months to age t h i r t y months. Samples at six-month periods were selected and utterances containing i n i t i a l c l u s t e r s were i s o l a t e d . Two examples produced several times throughout the study were 'Brian' and 'truck'; these were heard as CVC (Bian, tuck, guck) and CWVC (Bwian, twuck). Instances of CVC decreased with age whereas instances of CWVC increased with age. Results of spectrographic analysis of words with i n i t i a l c l u s t e r s and singleton i n i t i a l consonants the same as those in i n i t i a l p o s i t i o n i n a c l u s t e r , plus the same vowel, indicated that i n the case of c l u s t e r s , there was a more gradual t r a n s i t i o n i n t o the vowel than that observed i n the case of singleton consonants. Menyuk and K l a t t suggested that at t h i s point the c h i l d i s producing the c l u s t e r as a sin g l e segment incorporating features from both elements, and the c l u s t e r may be at some l e v e l of production considered by the c h i l d as a si n g l e 31 segment. Rules for f i l l i n g i n the complete set of features of the c l u s t e r and for separating them into two or more elements have at t h i s stage not been acquired. As Menyuk and K l a t t pointed out, however, the data i n t h e i r study cannot be considered conclusive evidence i n support of t h i s hypothesis. Additional data supporting Menyuk and K l a t t ' s hypothesis can be found i n a study by Kornfeld (1971b). Kornfeld c o l l e c t e d data from t h i r t e e n c h i l d r e n aged one-and-a-half to two-and-a-half years and analyzed spectrographically the i n i t i a l consonant c l u s t e r s produced. Her analysis showed that consonant c l u s t e r s may be produced at t h i s age as a singleton consonant with features from both elements: for example, the /b/ i n /bu/ produced for 'blue' i s spectrographically not the same as the /b/ i n 'boo.' Another p o s s i b l e example of t h i s process of com-bining features of two consonants into one segment i s the s u b s t i t u t i o n of a l a b i a l f r i c a t i v e f o r /sw/ (personal observation). One d i f f i c u l t y involved i n i n t e r p r e t i n g data such as those j u s t discussed i s i n making a decision concerning the abstract l e v e l at which a c h i l d i s reorganizing the c l u s t e r into a s i n g l e segment. I t i s at present v i r t u a l l y impossible to conclude: (a) whether the c h i l d perceives adult c l u s t e r s as s i n g l e segments; (b) whether he i s simply not capable, possibly for neurophysiological reasons, of producing the necessary precisely-timed a r t i c u l a t o r y gestures;, and (c) whether, although the c h i l d perceives the c l u s t e r s as successive segments, he has not yet learned to c o r r e c t l y apply the segmentalization rules required to sepa-rate feature 'bundles' into more than one segment. 32 1.2.4 Summary It i s possible to incorporate information obtained i n studies of timing control i n adults and c h i l d r e n and information about consonant c l u s t e r production i n c h i l d r e n into an i n v e s t i g a t i o n of the way i n which a timing-dominant or articulation-dominant system (Ohala 1970) may be r e f l e c t e d i n children's speech production. If a timing-dominant system exi s t s at some point i n the c h i l d ' s production, the implication i s that the c h i l d must execute s p e c i f i e d gestures i n accordance with a l i m i t e d , r i g i d time schedule. The hypothesis that c l u s t e r s are at some point pro-duced as singleton consonants i s consistent with the hypothesis that at some l e v e l of phonological organization a timing-dominant system may be i n e f f e c t . The c h i l d , with h i s demonstrated lack of timing c o n t r o l , can only meet the r i g i d demands of t h i s system by combining features of each component consonant into a singleton consonant. The s p l i t t i n g process l a t e r observed i n some ch i l d r e n may be one of the s t r a t e g i e s employed to overcome the r i g i d timing demands.in an e f f o r t to apply the appropriate segmenting r u l e s . As the c h i l d develops increasing control over timing of onset of a r t i c u l a t o r y gestures, he also learns to apply new reduction p r i n c i p l e s for co-ordination of consonants into a cohesive c l u s t e r u n i t . Obviously, the data supporting such an hypothesis are f a r from conclusive. Further evidence of possible timing systems used by adults, as well as more complete and r e l i a b l e developmental data on the acqui-s i t i o n of consonant c l u s t e r s by c h i l d r e n are required. 1.3 Aims of the Investigation The aim of the present i n v e s t i g a t i o n was not to answer the broader 33 t h e o r e t i c a l questions of temporal organization of speech, but rather to answer three s p e c i f i c questions concerning the development of temporal co-ordination of consonant c l u s t e r s i n children's speech: 1) whether at a given age l e v e l , the duration of c e r t a i n consonants i n c l u s t e r s i s reduced r e l a t i v e to duration of the same conson-ants as singleton segments; 2) whether there are differences between age groups i n terms of the amount or d i r e c t i o n of durational change i n consonants i n cl u s t e r s r e l a t i v e to singleton consonants; and 3) whether an age trend e x i s t s , i n the event that differences be-tween age groups occur. This age trend may be expected to r e f l e c t c l o s e r approximation to adult patterns with increasing maturity. 1 In answering these questions, an attempt was made to determine whether the same variables a f f e c t i n g durational changes e x i s t f o r each group. As two possible relevant v a r i a b l e s , i . e . , homorganicity (Haggard 1973a) and a r t i c u l a t o r y d i f f i c u l t y , associated p a r t i c u l a r l y with / l / (Hawkins 1973), have been postulated, the c l u s t e r s analyzed included a homorganic c l u s t e r with / l / , a nonhomorganic c l u s t e r with / l / , and a non-homorganic c l u s t e r without / l / . Such c l u s t e r s provide the most s e n s i t i v e measure of possible differences between age groups r e s u l t i n g from q u a l i -t a t i v e l y d i f f e r e n t constraints. Accordingly, / s i / , / f l / , and /sw/ were selected for analysis. CHAPTER 2 METHOD 2.1 Subjects Four c h i l d r e n , ranging i n age from s i x years to ten years, were selected as subjects f o r . a . p i l o t study. Five c h i l d r e n i n each of the following age groups served as subjects f o r the experiment proper: (1) f i v e to five-and-a-half years; (2) seven to seven-and-a-half years; (3) nine to nine-and-a-half years; and (4) eleven to eleven-and-a-half years. Five adults between the ages of twenty-three and twenty-eight years served as controls. Appendix A gives the sex, and age at t e s t i n g of each subject. Several c h i l d r e n came from the same f a m i l i e s . Appendix A also shows s i b l i n g r e l a t i o n s h i p s of the c h i l d r e n . A l l subjects were native speakers of English, r a i s e d i n the Greater. Vancouver area. L i n g u i s t i c background of the children's parents i s given i n Appendix B. The ch i l d r e n came from two suburbs i n the Greater Vancouver, area. Parental occupations include: academic, labour, c l e r i c a l , and pr o f e s s i o n a l , i . e . , the range of socio-economic cl a s s grouping was from lower-middle to upper-middle. Adult subjects a l l hold academic or professional jobs. Speech sound a r t i c u l a t i o n scores of a l l subjects were within normal l i m i t s as measured by the Templin-Darley A r t i c u l a t i o n Screening Test. A l l subjects showed hearing within normal l i m i t s . 34 35 2.2 Construction of Word L i s t s In order to obtain tokens of consonants and consonant c l u s t e r s for analysis, t e s t l i s t s were constructed. Use of l i s t s ensured maximum control of the context i n which the consonants were produced; durational differences between tokens of consonants could not therefore be a t t r i b u t e d 2 to e f f e c t s of following vowels or consonants. Test words selected met the following c r i t e r i a : (1) monosyllabic; (2) meaningful, and (3) (C)CVC structure. The corpus included nouns, verbs, and adjectives. A l l tokens (except 'feet') were sing l e morphemes. The following i n i t i a l consonants (or consonant clusters) were used: / s / , / f / , / l / , /w/, / s i / , / f l / , and /sw/. Singleton consonants were included to allow l a t e r comparison with that consonant i n the c l u s -tered contexts. Vowels used included: / i / , / I / , and /p/, i . e . , vowels which would allow construction of meaningful words given the i n i t i a l and f i n a l con-sonant constraints. F i n a l consonants i n the (C)CVC included the unvoiced stops: /p/, / t / , and /k/, except for the unvoiced a f f r i c a t e / t / / , which was used i n the token 'switch.' I t i s recognized that, i d e a l l y , the f i n a l consonant should not vary. A v a i l a b i l i t y of meaningful words, however, necessitated some v a r i a t i o n . I t has been shown (Peterson and Lehiste 1960) that the 2 Although i t can be argued that e l i c i t a t i p n by means of l i s t s i s an a r t i f i c i a l production s i t u a t i o n , i t i s nonetheless j u s t i f i a b l e to ana-lyz e any differences obtained within that s i t u a t i o n . No attempt should be made, however, to extend r e s u l t s based on l i s t e l i c i t a t i o n to continu-ous, spontaneous speech u n t i l that context has i t s e l f been investigated. Such i n v e s t i g a t i o n i s beyond the scope of the present study. 36 four f i n a l consonants selected have very s i m i l a r e f f e c t s on the duration of at l e a s t the preceding s y l l a b i c nucleus. These consonants, therefore, might reasonably be expected to have s i m i l a r e f f e c t s ( i f any) on the dura-t i o n of i n i t i a l consonants and consonant c l u s t e r s . Combining each i n i t i a l consonant and consonant c l u s t e r with each vowel and with a f i n a l consonant generated a l i s t of twenty-one t e s t words. This t e s t l i s t was then doubled either by (1) changing the f i n a l consonant of each word, or (2) i n those instances where only one f i n a l consonant could be used to make a meaningful word, including t h i s word i n the l i s t twice. The t e s t l i s t , therefore, included forty-two tokens, i . e . , two tokens of each i n i t i a l consonant or consonant c l u s t e r with each vowel; these words are given i n Table 2.1. To allow maximum randomization of task across subjects, three further l i s t s of forty-two words each were constructed by randomly order-ing the words i n the o r i g i n a l l i s t . Using such l i s t s i t was possible to obtain s i x tokens of each i n i t i a l consonant or consonant c l u s t e r , i n each vowel context, from each subject. The three t e s t l i s t s are given i n Appendix C. 2.3 Selection of a C a r r i e r Phrase One of the problems encountered i n almost any study of temporal organization of speech i s that of rate v a r i a t i o n . In order to draw con-clusions about differences between i n t r i n s i c durations of phones i n d i f -ferent utterances, i t i s necessary to eliminate the e f f e c t of e x t r i n s i c , i . e . , rate, d i f f e r e n c e s . The assumption i n the present study i s that rate v a r i a t i o n i s random throughout the production of the t e s t l i s t s and, Table 2.1. L i s t of Words Used i n Data C o l l e c t i o n Vowel / s / / f / A / /w/ / s i / / f l / /sw/ / i / seat fe e t leap week sleep f l e e t sweep seep leak wheat sl e e t sweet / I / s i t f i t l i c k whip s l i p f l i p switch sic k l i p wick s l i t f l i c k / P / sock fought lock walk s l o t f l o p swap sop l o t watch slop f l o c k swat 38 because word order i n the l i s t s i s also random, t h i s v a r i a t i o n should not s i g n i f i c a n t l y a f f e c t r e s u l t s . I f po s s i b l e , however, the amount of rate v a r i a t i o n within a session should be measured. A c a r r i e r phrase preceding each word was therefore included. A p i l o t study was designed to s e l e c t an appropriate phrase. The three phrases tested were: " I t ' s a ," "I say ," and "Repeat ." Since a l l words i n the l i s t s were not sing l e nouns, the phrase " I t ' s a " resulted i n several nongrammatical utterances. Testing with p i l o t subject one c l e a r l y demonstrated that for t h i s reason the phrase would be unsuitable. The phrase "I say " was tested with a l l four p i l o t subjects and, i n terms of e l i c i t a t i o n of utterances, proved to be very successful. The phrase was rejected, however, on the basis of preliminary analysis of mingograms made of p i l o t subject four's utterances: the boundaries between f i n a l / e i / i n 'say' and succeeding word i n i t i a l voiced segments were very d i f f i c u l t to determine. The youngest p i l o t subject (four), was then retested using the c a r r i e r phrase "Repeat " to determine s u i t a b i l i t y of the phrase i n e l i c i t a t i o n with young subjects and to obtain mingograms comparable with those made using the c a r r i e r phrase "I say ." R e l i a b i l i t y of segmen-t a t i o n of "Repeat " b e f o r e t e s t words was then compared with r e l i -a b i l i t y of segmentation of "I say " before the same words. The more r e l i a b l e measurements were.obtained with "Repeat " as c a r r i e r phrase. This phrase was consequently selected, even though s l i g h t l y more d i f f i -c u l t y i n e l i c i t a t i o n with younger subjects was anticipated. 39 2.4 Data C o l l e c t i o n 2.4.1 Recording Environment Data were c o l l e c t e d from each subject i n one session only. A l l sessions (with two exceptions) took place i n eit h e r E's home or E's parents' home, where optimal recording conditions were e a s i l y c o n t r o l l e d . Sessions for S's 1.1 and 1.5 were conducted i n t h e i r homes. In these two instances, attempts were made to minimize background noise by turning o f f appliances and recording i n a room with the lowest noise l e v e l pos-s i b l e . Recording conditions for SI.5 were q u a l i t a t i v e l y comparable with those of other subjects. Recording conditions for S l . l were character-ized by considerable background noise. This noise may have affected r e l i a b i l i t y of measurements for S l . l . 2.4.2 Recording Equipment A Nagra IV-D portable tape recorder, AKG D202E d i r e c t i o n a l micro-phone, and Ampex 434 low-noise tape were used for data c o l l e c t i o n . The tape recorder was c a l i b r a t e d to give a f l a t response (± 2 dB) over the freguency range 50 to 10,000 Hz. Tape speed was 7.5 i n . per sec. The microphone was held at a distance between four and ten inches from the subject's mouth. The peak VU record l e v e l was optimally 0 dB SPL; inten-s i t y l e v e l of the younger children's voices, however, varied so ra p i d l y and extensively that a constant record l e v e l was not possi b l e . 2.4.3 Screening Test Administration At the beginning of each session, the Templin-Darley A r t i c u l a t i o n Screening Test was administered to the subject. Responses were recorded 40 and l a t e r scored by two judges. Subjects' scores are reported i n Appen-dix A. 2.4.4 Test Word E l i c i t a t i o n Before e l i c i t i n g tokens of t e s t words, an informal check was con-ducted to determine that subjects were f a m i l i a r with these words; meanings of any unfamiliar words were c a r e f u l l y explained. In order to e l i c i t tokens of t e s t words i n the frame sentence "Repeat ," subjects were asked to repeat each e n t i r e sentence when presented. The examiner then said each t e s t word i n the sentence "Repeat (test word)." Af t e r the model sentence, the subject responded "Repeat (test word)." Every subject received the three t e s t l i s t s presented i n t h i s way. Order of presentation of l i s t s was randomized. Recordings were made of both model utterances and subjects' responses. Test ses-sions ranged i n duration from twenty minutes to one hour. Several sub-jects required frequent r e s t periods. Subjects completed the three t e s t l i s t s i n one session. 2.5 Data Analysis 2.5.1 Preparation of Tapes for Analysis Only frame sentences containing target words with the vowel / i / were used i n the present a n a l y s i s . Recordings of a l l t e s t sessions were therefore edited. O r i g i n a l recordings were reproduced on. an Ampex 440B tape recorder having approximately the same frequency response as the Nagra IV-D recorder. Utterances for l a t e r mingographic analysis were then edited onto an Ampex Micro 50 Cassette recorder. A l l edited 41 utterances were transcribed independently by two phonetically trained judges. When t r a n s c r i p t i o n s d i f f e r e d , a t h i r d judge determined the most accurate representation. 2.5.2 Instrumental. Analysis Mingograms were made of a l l edited items. Figure 2.1 presents a schematic representation of instrumentation used for t h i s purpose. The Ampex 50 Cassette recorder i s designated R 1. A REVOX Model A l l i s des-ignated R 2. A Frc^kjaer-Jensen Transpitch meter provided the Channel 2 tr a c i n g . Channel 3, the speech power c i r c u i t , has been described by Peterson and McKinney (1961). A Siemens Oscillomink, with paper speed of 10 cm. per s e c , was used to record the traci n g s . Three signals were displayed as traces on the mingograms: (a) speech wave s i g n a l ; (b) duplex oscillogram; and (c) log of average speech power. Segmentation of utterances was made on these three traces. The most r e l i a b l e trace was given by the duplex oscillogram i n which frequen-c i e s above 2.5 kHz are converted to a negative going i n t e n s i t y curve and mixed with the p o s i t i v e part of the audio frequency s i g n a l . This trace was e s p e c i a l l y e f f e c t i v e i n determining f r i c a t i v e and stop boundaries. 2.5.3 Segmentation of Utterances 1 To determine acceptable segmentation c r i t e r i a , twenty-one u t t e r -ances of P.4 were selected for comparative spectrographic and mingo-graphic a n a l y s i s . Each i n i t i a l consonant and consonant c l u s t e r was included. Spectrograms were made of these utterances (Kay Sonagraph Type A, frequency range 80 to 80,000 Hz). A scale factor was used to Mingograph Channel 1 Duplex Oscillograph Channel 2 Speech Power C i r c u i t Channel 3 1. Arrangement of Instrumentation 43 c a l c u l a t e durations of acoustic segments measured on the sonagrams. The durations of segments measured on the sonagrams were then compared to those measured on the mingograms. For c a l c u l a t i n g the duration of the c a r r i e r phrase, measurements were made of the distance from onset of the unstressed s y l l a b l e / r a / to the release of the f i n a l consonant / t / . Segmentation c r i t e r i a are described below (2.5.4). Three measures were made on those t e s t items having a CVC struc-ture. They included: (1) the i n i t i a l f r i c a t i v e , semi-vowel, or resonant; (2) the vowel / i / ; and (3) the f i n a l stop consonant. Five measures were made on the t e s t words having a CCVC structure. These included: (1) the i n i t i a l f r i c a t i v e ; (2) the voic e l e s s t r a n s i t i o n segment between the f r i c a t i v e and following voiced consonant; (3) the voiced resonant / l / or semi-vowel /w/; (4) the vowel / i / ; and (5) the f i n a l stop consonant. The voiceless t r a n s i t i o n segment was measured i n order to determine the extent to which i t s duration r e l a t i v e to the following voiced segment may account for any observed differences between children's and adults' pro-ductions. A l l measurements were made i n accordance with the segmentation c r i t e r i a described below. 2.5.4 Segmentation C r i t e r i a  2.5.4.1 P r i n c i p l e s of Segmentation The basic assumption underlying the segmentation c r i t e r i a used i n the present study i s that timing information i n both speech production and perception i s associated with major changes i n manner of a r t i c u l a t i o n (Lehiste 1972). Segment durations can therefore be determined by i 44 examining the manner of a r t i c u l a t i o n cues represented as abrupt changes i n the v i s u a l display of the acoustic waveform. 2.5.4.2 Vowels and Semi-vowel / r / A comparison of a l l three traces y i e l d e d reasonably accurate measures of both vowel onset and vowel o f f s e t . Vowel onset was marked by the onset of a p e r i o d i c wave envelope characterized by an abrupt growth of amplitude, i n the speech wave trace. This point coincided with an abrupt increase i n amplitude of the wave shown i n the speech power trace. Vowel onset a f t e r a semi-vowel was not as r e a d i l y determined as vowel onset a f t e r a v o i c e l e s s f r i c a t i v e or onset of the semi-vowel / r / i n i n i t i a l p o s i t i o n . C r i t e r i a used for vowel onset a f t e r / l / or /w/ i s d i s -cussed below (2.5.4.3). Vowel o f f s e t was marked by the end of p e r i o d i c e x c i t a t i o n i n the waveform shown i n the speech wave trace. As a l l vowels were immediately followed by vo i c e l e s s stop consonants, t h i s point i n the trace was e a s i l y located. 2.5.4.3 Semi-vowel /w/ and Resonant /!/ In w o r d - i n i t i a l unclustered p o s i t i o n , the onset of these segments was measured at the onset of p e r i o d i c e x c i t a t i o n i n the waveforms shown i n the speech wave and.duplex oscillogram traces. Following a f r i c a t i v e , onset of the voiceless portion of the seg-ment was determined with reference to several cues, shown i n Figure 2.2 to Figure 2.4. Thus: (a) i n / s i / c l u s t e r s , the speech wave si g n a l trace for segment onset showed a sharp burst, r e f l e c t i n g movement of the sides 4 5 FIGURE 2.2 SEGMENTATION OF / s i / ON MINGOGRAPHIC TRACE 46 FIGURE 2.3 SEGMENTATION OF / f l / ON MINGOGRAPHIC TRACE FIGURE 2.4 SEGMENTATION OF /sw/ ON MINGOGRAPHIC TRACE 48 of the tongue i n the formation of / l / (Haggard 1973a; Hawkins 1973); (b) i n / f l / c l u s t e r s , onset of v o i c e l e s s / l / was usually s i g n a l l e d by a s h i f t i n the duplex oscillogram trace, marking a change i n amplitude associated with the change to the / l / a r t i c u l a t i o n ; the speech wave si g n a l trace often showed a corresponding change i n wave amplitude (Haggard 1973a; Hawkins 1973); and (c) i n /sw/ c l u s t e r s , d e f l e c t i o n of the duplex o s c i l -logram trace from i t s baseline decreased sharply at the onset of voice-less /w/, c l e a r l y d e f i n i n g t h i s segment boundary (Hawkins 1973). In a l l cases, o f f s e t of.the voiceless /w/ and / l / segments was indicated i n the s i g n a l traces by the onset of the following p e r i o d i c e x c i t a t i o n . This point.also marked the onset of voiced /w/ and / l / seg-ments i n the three types of c l u s t e r s . ;'" Some d i f f i c u l t y was encountered i n determining o f f s e t of voiced / l / and /w/ before a vowel, i n both clustered and unclustered contexts. Haggard suggests that a useful cue i n determining t h i s boundary i s the point of maximum change of the waveform peak amplitude, which often coincides with a point of noticeable change i n the waveshape (Haggard 1973, p. 11). Further segmentation cues for t h i s boundary are described by Hawkins (1973, p. 210): o f f s e t of / l / and /w/ before the vowel / i / i s often marked by an increase i n amplitude represented as a downward s h i f t i n the duplex oscillogram trace, i n d i c a t i n g the presence of high f r e -quency components associated with onset of the vowel / i / . On the basis of these cues, i t was possible to segment the majority of utterances. Spectograms were made i n those cases where the boundary between /!/ or /w/ and the following vowel could not be determined by means of 49 the foregoing c r i t e r i a . The onset and o f f s e t of formants was then used as a segmentation cue. 2.5.4.4 F r i c a t i v e s The onset of a l l f r i c a t i v e s was measured at that point where a downward s h i f t i n the duplex oscillogram trace showed a r i s e i n amplitude. The o f f s e t of f r i c a t i v e s i n unclustered contexts was taken to be that point where p e r i o d i c i t y i n the speech wave trace indicated v o i c i n g i n the following segment. F r i c a t i v e o f f s e t i n clustered contexts was marked by the onset of the voiceless / l / or /w/ described above. 2.5.4.5 Voiceless Plosives The onset of a l l stops was measured at that point where both the speech wave trace and the duplex oscillogram trace marked the cessation of e x c i t a t i o n which characterized preceding voiced segments. The burst at the release of closure s i g n a l l e d stop o f f s e t . The a s p i r a t i o n period of v o i c e l e s s stops, represented as aperiodic e x c i t a t i o n i n the speech wave and duplex oscillogram traces, was not included i n stop measurements. Exclusion of the a s p i r a t i o n period i s consistent with the underlying p r i n c i p l e s of segmentation, as the a r t i c u l a t o r y gestures involved are d i s t i n c t from those involved i n producing the stop (Hawkins 1973, p. 212). 2.5.5 Measurement R e l i a b i l i t y Duration measurements were made to the nearest 10 msec. Maximum possible accuracy was therefore 5 msec. R e l i a b i l i t y of duration measure-ments was tested by measuring one subject's utterances twice on d i f f e r e n t occasions. 50 2.5.6 Test Word Rejection A l l tokens of t e s t words containing vowels other than / i / were rejected, as were tokens containing speech errors. Several subjects, for example, produced as many as f i v e consonant segments before producing the appropriate phones; i t was impossible to determine onset of the appropriate, c l u s t e r i f preceded by one or more inappropriate phones. Some d i f f i c u l t y was encountered, however,.in determining whether pro-longed i n i t i a l f r i c a t i v e s , i n c e r t a i n tokens constituted speech errors which should be excluded. I t was f i n a l l y decided to exclude those tokens i n which the f r i c a t i v e duration was s u b j e c t i v e l y and unequivocably classed as a speech error by.two judges. Durations of such f r i c a t i v e s a l l exceeded 300 msec. A l l other f r i c a t i v e s , i ncluding those described as 'somewhat long' were measured. 2.5.7 S t a t i s t i c a l Analysis To answer the three, questions outlined i n the aims of the inves-t i g a t i o n (1.3), three s t a t i s t i c a l analyses were performed. Each of these analyses was comprised of four separate analyses. Table 2.2 de-scribes the variables used i n each a n a l y s i s . 2.5.7.1 Analysis I In Analysis I, main e f f e c t s and i n t e r a c t i o n e f f e c t s of age (with subjects nested under age), context, and consonant type on the duration of a consonant i n i n i t i a l p o s i t i o n of a CVC word and on duration of both consonants, considered separately, i n a CCVC word, were tested. The dependent va r i a b l e was duration of the s p e c i f i e d consonant i n a p a r t i c u l a r 51 Table 2.2. Summary of Variables used i n S t a t i s t i c a l Analyses Analysis Age (Subject nested) Context Phone Dependent Variable Ia 5 groups (5,7,9,11 Adults) Alone Before 1 s f Duration Ib 5 groups Alone Aft e r / s / 1 w Duration Ic 5 groups Alone Before /w/ Before / l / s Duration Id 5 groups Alone A f t e r / s / Afte r / f / • 1 Duration I l a 5 groups Before / l / f Difference i n msec. l i b 5 groups Aft e r / s / w Difference i n msec. l i e 5 groups Aft e r / s / Afte r / f / 1 Difference i n msec. l i d 5 groups Before / l / Before /w/ s Difference i n msec. I l i a 4 groups (5,7,9,11) Alone/Cluster f Difference i n msec. I l l b 4 groups Alone/Cluster s Difference i n msec. IIIc 4 groups Alone/Cluster 1 Difference i n msec. H i d 4 groups Alone/Cluster w Difference i n msec. 52 context, as produced by a given subject i n a given age group. Each c e l l contained several n's, because each subject produced several tokens of each consonant i n a given context. Analyses Ia and Ib were four-way analyses of variance (age, sub-j e c t , context, and phone). Analyses Ic and Id were three-way analyses of variance (age, subject, and context); phone was s p e c i f i e d as / s / i n Analysis Ic and / l / i n Analysis Id. By performing a Newman-Keuls multiple range t e s t on predicted means for duration of each consonant i n each context at each age l e v e l , i t was possible to determine whether the e f f e c t of context was s i g n i f i -cant for each consonant at each age l e v e l . M u l t i p l e range t e s t s were performed when eit h e r context as a main e f f e c t was s i g n i f i c a n t , or the i n t e r a c t i o n e f f e c t of age x context was s i g n i f i c a n t . 2.5.7.2 Analysis II In Analysis I I , the e f f e c t s of age and context (with subjects nested under age) on the amount of reduction i n duration of a consonant i n a c l u s t e r , r e l a t i v e to duration of that consonant as a singleton were tested. One duration value for each consonant i n each context for each subject was obtained by f i n d i n g the median duration of the several tokens produced by the subject. The dependent v a r i a b l e i n the analyses was the difference i n msec, between (median) duration of a consonant as a singleton, and (median) duration of that consonant as part of a c l u s t e r . Lengthening of a consonant i n a c l u s t e r r e l a t i v e to singleton duration was considered negative reduction. Four analyses of variance of t h i s type were included i n Analysis 53 II . Analyses I l a and l i b were simple analyses of variance, t e s t i n g e f f e c t of age on consonant reduction. Analyses l i e and l i d were two-way analyses of variance, t e s t i n g e f f e c t s of age and context on consonant reduction. A Newman-Keuls t e s t was done for those analyses having s i g n i f i -cant main, e f f e c t s or i n t e r a c t i o n e f f e c t s . 2.5.7.3 Analysis III Analysis I I I tested possible trends c o r r e l a t i n g age with amount of consonant reduction. Only four age groups, f i v e years, seven years, nine years, and eleven years, were included i n t h i s a n a l y s i s . Reduction of each of the four consonants /s/, / l / , / f / , and /w/ was considered, g i v i n g four trend analyses. Analysis II demonstrated that i t was pos-s i b l e to consider each consonant i n c l u s t e r s without s p e c i f y i n g c l u s t e r type; consequently, only one di f f e r e n c e , i . e . , d i f f e r e n c e between dura-t i o n as a singleton and duration i n c l u s t e r s , was calculated f o r each consonant f or each subject. Again, median values for each consonant i n each context for each subject were used i n c a l c u l a t i n g d i f f e r e n c e s . In those cases where the F r a t i o f o r l i n e a r regression did not reach significance,, quadratic and cubic regressions were tested. CHAPTER 3 RESULTS 3.1 Rate Control Experiment An observational analysis of durations of the c a r r i e r p h r a s e — 'Repeat'—in the data for each subject, indicated that v a r i a t i o n s i n these durations were not representative of the rate changes i n the t o t a l data for the subject for several reasons. F i r s t , each subject tended to vary the stress on the f i r s t s y l l a b l e , often reducing i t so greatly that the s y l l a b l e was not c l e a r l y represented on the o s c i l l o g r a p h t r a c -ings.. Second, duration of the stressed s y l l a b l e of the c a r r i e r phrase was often lengthened to such an extent that i t was apparent that the sub-j e c t was h e s i t a t i n g , while u t t e r i n g 'Repeat,' before producing the re-quired t e s t word. Third, i t was often d i f f i c u l t to determine the point on the o s c i l l o g r a p h trace i n d i c a t i n g release of the f i n a l consonant / t / ; i n many cases, phonetic t r a n s c r i p t i o n of the data demonstrated omission of the f i n a l consonant. These f a c t o r s , contributing to v a r i a b i l i t y of duration of the c a r r i e r phrase, were not evident i n the analysis of the t e s t words. F i r s t , because a l l t e s t words were monomorphemic, stress v a r i a b i l i t y was never s u f f i c i e n t to cause a l t e r a t i o n s of vowel q u a l i t y , or omission of any portion of the t e s t word. Second, no subject ever hesitated while u t t e r i n g a t e s t word. In view of the d i f f e r e n t factors contributing to 54 55 duration v a r i a b i l i t y for c a r r i e r and t e s t words, i t was decided that rate v a r i a b i l i t y f o r t e s t words would be more r e l i a b l y represented by a descrip-t i o n of the range of durations f o r each segment for each subject. Though all.segments (as defined i n section 2.5.4) were measured, only the ranges of the four i n i t i a l consonants and of the component consonants of c l u s t e r s are shown i n the following tables. 3.2 Duration Measurement Data Tables 3.1 and 3.2 show the mean durations i n msec. for each con-sonant i n each context. The means shown are f o r a l l durations of a given consonant i n a given context produced by a l l subjects i n each age group. Also included are: standard,deviation, range, and the number of tokens included i n each mean c a l c u l a t i o n . Although sample s i z e f o r each conso-nant i n each context i s optimally t h i r t y tokens (six tokens per subject; f i v e subjects per group) for each age group, several samples have N < 30 since tokens were occasionally necessarily rejected. Notation used f or context i n several of the tables i s to be i n -terpreted as follows: 'Alone' r e f e r s to the consonant as a singleton i n a CVC word; /-IA used only with reference to / s / or / f / as the consonant under consideration, indicates that the durations are those of the f i r s t consonant i n a CCVC word, where the second consonant i s / l / ; /-w/, used only with reference to /s/, indicates that the duration i s for / s / before /w/ i n a CCVC word; /s - / , used with / l / and /w/, indicates that the dura-t i o n shown i s for / l / or /w/ a f t e r / s / i n a CCVC word; and / f - / , used only with / l / , i ndicates that the duration shown i s for / l / a f t e r / f / i n a CCVC word. Table 3.1. Mean Durations f o r a l l Age Groups, / s / and / f / . _ /•?/ / J - / Age Content X S.D. Range N X S.D. Range N Adult Alone 167.0 25.7 110 30 143.3 37.2 160 30 /-!/ 134.7 22.3 100 30 117.7 27.3 100 30 /-w/ 143.0 31.0 130 30 11 yr. Alone 166.1 31.0 120 28 149.0 31.8 140 29 / - l / 132.8 26.5 100 29 116.7 22.4 80 27 /-w/ 143.2 30.6 170 28 9 yr. Alone 168.7 29.1 90 30 151.4 38.5 160 28 / - l / 142.8 36.5 160 29 121.1 30.0 110 28 /-w/ 154.1 28.3 120 27 7 yr. Alone 196.3 37.8 140 30 162.3 33.5 140 29 / - l / 173.2 27.5 110 28 143.3 33.0 130 27 /-w/ 155.3 39.4 180 32 5 yr. Alone 194.5 42.1 160 31 167.1 52.2 230 28 160.0 36.0 120 29 163.1 50.0 190 26 /-w/ 170.0 43.9 180 24 Ul (TV Table 3.2. Mean Durations f o r a l l Age Groups, /!/ and /w/ /w/ Age Content X S.D. Range N X S.D. Range N Adult 11 yr. 9 yr. 7 yr. 5 y r . Alone / s - / / f - / Alone / s - / / f - / Alone / s - / / f - / Alone / s - / / f - / Alone / s - / 84.0 72.3 69.0 69.7 68.3 63.3 77.4 77.2 82.1 87.0 94.0 90.0 94.8 115.8 107.7 19.3 14.1 19.4 24.4 19.8 21.4 22.0 23.0 26.6 21.4 17.6 17.8 41.7 35.1 29.1 80 60 80 110 60 90 100 100 100 70 80 70 150 150 130 30 30 30 30 29 27 27 29 28 27 28 27 29 29 26 69.7 69.0 67.2 72.1 72.0 76.3 85.9 87.1 95.9 92.1 24.4 17.4 20.2 17.4 23.0 19.7 35.7 27.0 43.0 27.4 90 60 70 90 100 90 160 120 200 110 29 30 29 28 30 27 29 32 29 24 Ul 58 Figures 3.1 and 3.2 are graphic representations of mean durations i n msec, for each consonant i n each context; the means shown are the same as those i n Tables 3.1 and 3.2. The bars representing durations of / l / and /w/ are divided into two sections: the shaded portion at the bottom of the bar represents duration of the voiceless t r a n s i t i o n phase immedi-a t e l y following the i n i t i a l f r i c a t i v e , and the unshaded p o r t i o n repre-sents duration of the voiced part of / l / or /w/. Numerical values f o r these segments are given i n Appendix D. Median values f o r duration of each consonant i n each context for each subject are given i n Appendix E. The ranges and number of tokens used i n each c a l c u l a t i o n are also given. 3.3 Results of Analysis I (a-d) The r e s u l t s of Analysis I, which included four analyses of v a r i -ance, are given i n Table .3.3 and Appendix F. For a l l four analyses, r e s u l t s of multiple range tests (Newman-Keuls) on s i g n i f i c a n t main e f f e c t s are given i n Appendix G. Results of Newman-Keuls multiple range t e s t s on means for each, consonant i n each context f o r each age group are summarized i n Table 3.4. Table 3.3 also shows the F values f o r each main e f f e c t i n each an a l y s i s . S i g n i f i c a n t i n t e r a c t i o n e f f e c t s are shown i n the table; none of the other i n t e r a c t i o n e f f e c t s reached s i g n i f i c a n c e . An a d d i t i o n a l l e v e l tested i n the four analyses of variance i s 'Subject,' nested under age. E f f e c t s due to t h i s factor were s i g n i f i c a n t i n every analysis; F - r a t i o s f o r t h i s e f f e c t are not given i n the r e s u l t s . Adults 59 1 2 3 / s / 1 2 / f / M. 1 4 5 1 4 /w/ 11-year group 1 2 3 / s / 1 2 / f / 1 4 5 /!/ 1 4 /w/ 9-year group 150 -| 100 50 H 1 1 2 3 / s / 1 2 / f / 1 4 5 /!/ 1 4 /w/ F i g . 3.1. Graphic representation of durations i n msec. l=Alone; 2=Before / l / ; 3=Before /w/; 4=After /s/; 5=After / f / . 7-year group 60 200 150 100 50 J 1 2 3 / s / 1 2 / f / 1 4 5 /!/ 1 4 /w/ 5-year group 200 A 150 100 50 J 1 2 3 / s / 1 2 / f / 1 4 5 /!/ 1 4 /w/ F i g . 3.2. Graphic representation of durations i n msec. l=Alone; 2=Before / l / ; 3=Before /w/; 4=After /s/; 5=After / f / 61 Table 3.3. Analysis I (a-d): Summary of Analyses of Variance Analysis V a r i a b l e s R e : s u i t s Ia Age (5 levels) F(4,541) 30.93** Phone (/s/, /£/) F(l,541) 57.73** Context (Alone, / - l / ) F(l,541) 103.16** Ib Age (5 levels) F(4,540) 28.39** Phone (/!/, /w/) F(l,540) 7.92** Context (Alone, /s-/) F(l,540) .74 Age x Phone x Context F(4,540) 2.50* Ic Age (5 levels) F(4,399) 20.43** Context (Alone, / - l / , /-w/) F(2,399) 42.96** Phone (/s/) Id Age (5 levels) F(4,391) 40.58** Context (Alone, / s - / , / f - / ) F(2,391) .79 Age x Context F(8,391) 2.59** Phone (/!/) S i g n i f i c a n t at P < .05. * S i g n i f i c a n t at P < .01. 62 Table 3.4. Analysis I (a-d): Summary of Multiple Range Tests (Newman-Keuls) Analysis Age Group / s / alone vs. / s / before / l / / f / alone vs. / f / before /!/ Ia Adults * * 11 years * * 9 years * * 7 years * * 5 years * N.S. Analysis Age Group / l / alone vs. /!/ a f t e r / s / /w/ alone vs. /w/ a f t e r / s / Ib Adults N.S. N.S. 11 years N.S. N.S. 9 years N.S. N.S. 7 years N.S. N.S. 5 years * N.S. Analysis Age Group / s / alone vs. / s / before /w/ / s / before / l / vs / s / before /w/ Ic Adults * N.S. 11 years * N.S. 9 years N.S. N.S. 7 years * N.S. 5 years * N.S. Analysis Age Group / l / alone vs. /!/ a f t e r / f / / I / a f t e r / s / vs. /!•/ a f t e r / f / Id Adults N.S. N.S. 11 years N.S. N.S. 9 years N.S. N.S. 7 years N.S. N.S. 5 years * N.S. * S i g n i f i c a n t at P < .05. 63 3.3.1 Analysis Ia Analysis Ia tested e f f e c t s of age, subject, phone (/s/ or /f/) and context (alone, and / - l / ) on consonant duration. A l l three main e f f e c t s were s i g n i f i c a n t . Results of a multiple range t e s t comparing means of each consonant i n each context for each group, summarized i n Table 3.4, indicated that f o r each age group the means f o r / s / alone and /s/ before / l / were s i g n i f i c a n t l y d i f f e r e n t . The means for / f / alone and / f / before / l / were s i g n i f i c a n t l y d i f f e r e n t f o r each of the four older age groups. For the fi v e - y e a r - o l d group, however, the means f o r / f / alone and / f / before / l / were not s i g n i f i c a n t l y d i f f e r e n t . 3.3.2 Analysis Ib Analysis Ib tested e f f e c t s of age, subject, phone ( / l / or /w/) and context (alone, and /s-/) on consonant durations. Age, subject, and phone were a l l s i g n i f i c a n t , but context was not. However, the i n t e r -action e f f e c t , age x phone x context, was s i g n i f i c a n t . Results of a multiple range t e s t comparing means of each consonant i n each context for each age group are summarized i n Table 3.4 and show that the means for / l / alone and / l / a f t e r / s / were not s i g n i f i c a n t l y d i f f e r e n t for the four older age groups. For the f i v e - y e a r - o l d group, however, these means were s i g n i f i c a n t l y d i f f e r e n t . The means for / s / alone and /w/ a f t e r / s / were not s i g n i f i c a n t l y d i f f e r e n t f o r any age group. 3.3.3 Analysis Ic Analysis Ic tested e f f e c t s of age, subject, and context (alone, /-!/, and /-w/) on duration of / s / . A l l three main e f f e c t s were 64 s i g n i f i c a n t . A multiple range t e s t comparing means of durations i n each context f or each age group, the r e s u l t s of which are summarized i n Table 3.4, indicates that the means of / s / alone and / s / before / l / were s i g -n i f i c a n t l y d i f f e r e n t f o r a l l age groups. Means of duration of / s / alone and / s / before /w/ were s i g n i f i c a n t l y d i f f e r e n t f o r a l l age groups except the nine-year-old group. Means of duration of / s / before / l / and / s / before /w/ were not s i g n i f i c a n t l y d i f f e r e n t f o r any age group. 3.3.4 Analysis Id Analysis Id tested e f f e c t s of age, subject, and context (alone, /s- / , and /f-/) on duration of / l / . Age and subject both gave s i g n i f i -cant main e f f e c t s . Gontext, however, was not s i g n i f i c a n t . The i n t e r -action e f f e c t , age x context, was s i g n i f i c a n t . A multiple range t e s t comparing means of durations i n each context f or each age group, the re-s u l t s of which are summarized i n Table 3.4, indicated that the means of duration of / l / alone and / l / a f t e r / s / were s i g n i f i c a n t l y d i f f e r e n t only f o r the f i v e - y e a r - o l d group. S i m i l a r l y , the means f o r durations of / l / alone and / l / a f t e r / f / were s i g n i f i c a n t l y d i f f e r e n t f o r the f i v e -year-old group only. The means for duration of / l / a f t e r / s / and / l / a f t e r / f / were not s i g n i f i c a n t l y d i f f e r e n t f o r any age group. 3.4 Results of Analysis II (a-d) The res u l t s , of Analysis I I , which includes four analyses of v a r i -ance, are summarized i n Table 3.5. Summary tables for the s t a t i s t i c a l analyses are given i n Appendix H. S i g n i f i c a n t main e f f e c t s were further analyzed by means of a Newman-Keuls multiple range t e s t ; these r e s u l t s Table 3.5. Results of Analysis II (a-d) Analysis E f f e c t s Tested Levels F r a t i o Newman-Keuls I l a Age on duration d i f f e r e n c e between / f / alone and / f / before / l / 5 ages F(4,20)=.38 l i b Age on duration d i f f e r e n c e between /w/ alone and /w/ a f t e r / s / 5 ages F(4,20) = .05 l i e Age on duration d i f f e r e n c e between / l / alone and / l / i n c l u s t e r s 5 ages F(4,20)=3.66* 5-year-olds d i f f e r e n t to adults* Context on duration d i f f e r e n c e between / l / alone and / l / i n c l u s t e r s 2 contexts F(l,20)=.13 l i d Age on duration d i f f e r e n c e between / s / alone and / s / i n c l u s t e r s 5 ages F(4,20)=.37 Context on duration d i f f e r e n c e between / s / alone and / s / i n c l u s t e r s 2 contexts F(l,20)=1.33 * S i g n i f i c a n t at P < .05. 66 are shown i n Table 3.5. 3.4.1 Analysis I l a Analysis I l a tested the e f f e c t of age on the differ e n c e i n msec, between the duration of / f / alone and / f / before / l / . A l l f i v e age groups were included i n the analysis. The r e s u l t s showed that age d i d not s i g -n i f i c a n t l y a f f e c t the differ e n c e between the two durations f o r / f / . 3.4.2 Analysis l i b Analysis l i b tested the e f f e c t of age on the differ e n c e i n msec, between the durations of /w/ alone and /w/ a f t e r / s / . A l l f i v e age groups were included i n the analysis. The r e s u l t s showed that age d i d not s i g -n i f i c a n t l y a f f e c t the differ e n c e between the two durations f o r /w/. 3.4.3 Analysis l i e Analysis l i e tested the e f f e c t of age and the e f f e c t of context on the difference i n msec, between the durations of / l / alone and / l / i n a c l u s t e r . A l l f i v e age groups were included i n the analysis. Two con-texts were considered: / l / a f t e r / f / , and / l / a f t e r / s / . The r e s u l t s showed that context did not s i g n i f i c a n t l y a f f e c t the difference between durations of / l / alone and / l / i n a c l u s t e r . Age, however, d i d s i g n i f i c a n t l y a f f e c t t h i s d i f f e r e n c e . A Newman-Keuls mul-t i p l e range t e s t which pooled contexts and tested differences between ages showed that the only groups which were s i g n i f i c a n t l y d i f f e r e n t were the five-year-olds and the adults. 67 3.4.4 Analysis l i d Analysis l i d tested the e f f e c t of age on differences i n duration between / s / alone and / s / i n a c l u s t e r , and the e f f e c t of context, i . e . , following consonant, on the differences i n duration between / s / alone and / s / i n a c l u s t e r . A l l f i v e age groups were included, and the contexts tested were / - l / and /-w/. Results of the analysis of variance indicated that neither age nor type of following consonant a f f e c t the d i f f e r e n c e i n duration between / s / alone and / s / i n a c l u s t e r . 3.5 Analysis III A trend a n a l y s i s , including four analyses summarized i n Table 3.6 was done to determine whether the di f f e r e n c e i n duration between each consonant alone and that consonant i n clusters increases l i n e a r l y with age between the ages of f i v e and eleven years. In those cases where a l i n e a r regression was not observed,, quadratic and cubic regressions were tested. Summary tables of the analyses are given i n Appendix I. Since Analysis l i e indicated that the type of consonant preceding / l / i n a c l u s t e r does not a f f e c t the d i f f e r e n c e i n duration between / l / alone and / l / i n a c l u s t e r , i t was possible to pool the values for the / s i / and / f l / c l u s t e r s , r e s u l t i n g i n one d i f f e r e n c e only, i . e . , the d i f -ference between / l / alone and/1/ i n a c l u s t e r . A s i m i l a r procedure was followed for / s / alone and / s / i n a c l u s t e r , regardless of the second consonant i n that c l u s t e r . Table 3.6. Analysis III (a-d): F-ratios for Trend Correlating Age (5-11 years) with Reduction I l i a I H b H i e H i d Reduction / f / alone -/ f / before / I / / s / alone -/ s / i n c l u s t e r /!/ alone -/ l / i n c l u s t e r /w/ alone -/w/ i n c l u s t e r Linear F(l,16)= .47 F(l,16)= .46 F(l,16)=7.05* F(l,16)= .11 Quadratic F(l,16)= .11 F(l,16)= .03 - F(l,16)= .02 Cubic F(l,16)= .00 F(l,16)= .49 - F(l,16)= .01 Deviation - - F(2,16)= .62 -S i g n i f i c a n t at P < .05. 69 3.5.1 Analysis I l i a The trend analysis performed on the differ e n c e between durations of / f / alone and / f / before /l/showed that neither a l i n e a r , quadratic, nor cubic regression l i n e appropriately represents the r e l a t i o n s h i p be-tween age and difference i n duration. 3.5.2 Analysis I l l b The trend analysis performed on the differ e n c e between durations of / s / alone and / s / i n a c l u s t e r showed that neither a l i n e a r , quadratic, nor cubic regression l i n e appropriately represents the r e l a t i o n s h i p be-tween age and difference i n duration. 3.5.3 Analysis IIIc The trend analysis performed on the difference between durations of / l / alone and. / I / i n a c l u s t e r showed that a l i n e a r regression repre-sents the r e l a t i o n s h i p between age and di f f e r e n c e i n duration. Further, the deviation term demonstrated that, the l i n e a r regression i s a good f i t to the trend observed i n the.data. This regression i s shown i n Appendix J. 3.5.4 Analysis H i d The trend analysis performed on.the diffe r e n c e between durations of /w/ alone and /w/ i n a c l u s t e r showed that neither a l i n e a r , quadratic, nor cubic regression appropriately represents the r e l a t i o n s h i p between age and differ e n c e i n duration. 70 3.6 Summary of Main Points of Interest i n the S t a t i s t i c a l Analyses  The main points of i n t e r e s t i n the r e s u l t s of the s t a t i s t i c a l analyses can be summarized as follows: 1) / s / was s i g n i f i c a n t l y reduced before / l / by a l l age groups; 2) / s / was s i g n i f i c a n t l y reduced before /w/ by a l l groups except the nine-year-old group; 3) / f / was s i g n i f i c a n t l y reduced before / l / by a l l age groups except the f i v e - y e a r - o l d group; 4) / l / was s i g n i f i c a n t l y lengthened a f t e r / s / and / f / only by the fiv e - y e a r - o l d group; 5) / l / d i d not s i g n i f i c a n t l y , change with context for any other group; 6) /w/ d i d not s i g n i f i c a n t l y change with context for any age group; 7) the fiv e - y e a r - o l d group, d i f f e r e d s i g n i f i c a n t l y from the adult group i n the amount by which the / l / was reduced i n a c l u s t e r , r e l a t i v e to i t s duration as a singleton; no other p a i r s of age groups for any consonants d i f f e r e d s i g n i f i c a n t l y i n the amount by which consonants were reduced; and 8) diff e r e n c e i n duration of / l / alone and /!/ i n c l u s t e r s increased l i n e a r l y with age between f i v e and eleven years; a l i n e a r i n -crease was not seen f o r the differences i n duration of other con-sonants as singletons and i n c l u s t e r s . 3.7 Other Points of Interest i n the Data Several i n t e r e s t i n g observations should be made concerning the 71 raw data, even though these points have not been s p e c i f i c a l l y and s t a t i s -t i c a l l y tested: 1) observation of variances of durations of consonants i n each age group revealed.that greater variance i s associated with younger age groups; 2) i n v e s t i g a t i o n of the multiple range te s t s of age as a main e f f e c t on durations of f r i c a t i v e s and resonants (Appendix F) showed that two age groups can roughly be defined by the durations of these consonants: five-year-olds and seven-year-olds form one group, and nine-year-olds, eleven-year-olds, and adults form the other group; and 3) inspection of the s o - c a l l e d voiceless portion of / l / or /w/ showed that the duration of t h i s p ortion r e l a t i v e to the follow-ing voiced / l / or /w/ was approximately the same for a l l age groups. 3.8 Measurement R e l i a b i l i t y The durations of i n i t i a l consonant and consonant c l u s t e r segments of a l l t e s t words i n S2.5's data were measured twice by the same experi-menter. The second set of measurements was made one month a f t e r the f i r s t ; the same c r i t e r i a were used for both sets. The greatest d i f f e r e n c e between the two measurements f o r any segment was ±10 msec. This d i f f e r -ence was observed, for only nine of the t o t a l number of segments remea-sured, i n d i c a t i n g that segmentation c r i t e r i a were co n s i s t e n t l y applied. CHAPTER IV DISCUSSION AND CONCLUSIONS The purpose of t h i s i n v e s t i g a t i o n was to determine whether reduc-t i o n of duration of consonants i n c l u s t e r s d i f f e r s i n terms of amount and d i r e c t i o n of change with d i f f e r e n t age groups. In addition, the purpose was to determine, i n the event of such di f f e r e n c e s , whether an age trend c o r r e l a t i n g increasing reduction with increase age e x i s t s . 4.1 Discussion of Results Data i n the present i n v e s t i g a t i o n show that i n t e r e s t i n g d i f -ferences e x i s t f or consonant durations both within and across age groups. While the number of c l u s t e r s considered i s not large enough to permit a generalization of p r i n c i p l e s underlying consonant reduction, they are suggestive of (a) the way i n which consonants are co-ordinated into i n i t i a l c l u s t e r s by adults and (b) the development of such co-ordination i n c h i l d r e n . Before elaborating on the general trends demonstrated by the r e s u l t s , two inconsistencies should be noted. F i r s t , / s / was s i g n i f i -cantly reduced before /w/ by a l l but the nine-year-old group.' An inves-t i g a t i o n of the means for each subject i n the group revealed that although / s / was reduced i n a l l cases before /w/, the d i f f e r e n c e was not large enough to suggest s i g n i f i c a n c e i n three cases. No reasonable t h e o r e t i c a l explanation can be found for t h i s observation; the most p l a u s i b l e 72 73 explanation i s therefore that the r e s u l t must be a t t r i b u t e d to the small sample siz e which i n t h i s instance i s probably not representative of the nine-year-old population. The second point to be noted i s that / f / was not s i g n i f i c a n t l y reduced i n c l u s t e r s by the f i v e - y e a r - o l d group, although i t was s i g n i f i -cantly reduced by a l l other age groups. An i n v e s t i g a t i o n of i n d i v i d u a l means within the f i v e - y e a r - o l d group showed that / f / was reduced by only two of the f i v e subjects and was i n f a c t lengthened i n a l l other cases. I f median values are considered, however, i t i s noted that / f / was re-duced by a l l c h i l d r e n i n the group. An examination of the t o t a l measure-ments for t h i s group indicates that medians may be more, representative of the t o t a l sample, since there are some extremely short durations i n the category / f / alone. I t i s suggested therefore that t h e o r e t i c a l s i g n i f i c a n c e should not be attached to the d i f f e r e n c e between age groups i n terms of reduction of / f / . Reduction of consonant durations by adult speakers can be de-scribed by two general statements:. (1) i n i t i a l f r i c a t i v e consonants were always reduced;. (2) the second, consonant i n a c l u s t e r , whether semi-vowel or resonant, was never reduced. These r e s u l t s cannot be explained i n terms of the place of a r t i c u l a t i o n , hypothesis proposed by Haggard (1973) which suggests that nonhomorganic c l u s t e r s generally show more reduction than homorganic c l u s t e r s : i n the present data, duration of / s / d i d not undergo s i g n i f i c a n t l y more reduction before / l / (homorganic) or /w/ (non-homorganic) . S i m i l a r l y , duration of / l / a f t e r homorganic / s / was not s i g n i f i c a n t l y d i f f e r e n t to duration of /!/ a f t e r nonhomorganic / f / . 74 Further, although the data were not s t a t i s t i c a l l y analyzed duration measurements indicated, that / s / before / l / d i d not undergo s i g n i f i c a n t l y more or le s s reduction than /f/..before / l / . Thus the concept of homorgan-i c i t y as a major factor i n consonant duration reduction i s not at a l l substantiated i n the present data. Two explanations of consonant reduction i n i n i t i a l c l u s t e r s can be proposed,. though there i s . n o t : s u f f i c i e n t evidence i n the present i n -v e s t i g a t i o n to j u s t i f y accepting e i t h e r of these i n p a r t i c u l a r . F i r s t , i t i s possible that reduction of .:consonants i s determined l a r g e l y by considerations of manner of a r t i c u l a t i o n , such that, for example, a l l f r i c a t i v e s ,are permitted, to be.; reduced by a l i m i t e d and r e l a t i v e l y con-stant amount, regardless of the accompanying consonant. This amount may be l a r g e l y determined by the minimal duration necessary to ensure pro-duction of the target' a r t i c u l a t i o n . . Semi-vowels and resonants, however, which are considerably shorter.In duration as singleton consonants than are f r i c a t i v e s , may not be reduced simply because the singleton duration i s too short to permit reduction, i . e . , i f t h e i r durations were reduced i n c l u s t e r s there would not be s u f f i c i e n t time to achieve the target a r t i c u l a t i o n required for perception of the consonant. The second possible explanation of the type of consonant reduc-t i o n seen i n the present data i s that only the f i r s t consonant i n the c l u s t e r reduces, regardless of i t s place of a r t i c u l a t i o n r e l a t i v e to that of. the second consonant.. A r t i c u l a t o r y gestures for the second consonant may be moved forward i n time, reducing duration of the f i r s t consonant. The observation.that /!/ a f t e r / f / i s not s i g n i f i c a n t l y shorter than /!/ 75 a f t e r / s / suggests that the amount by which i n i t i a t i o n of gestures f o r second consonant may be temporally advanced i s constant and i s not affected by r e l a t i v e places of a r t i c u l a t i o n of the consonants. Inherent i n both explanations i s the assumption that there may be a maximum durations a l l o t t e d to i n i t i a l segments of higher-order units such as words which require that t o t a l duration of consonants i n c l u s t e r s must be less than the sum of duration of those consonants as singletons. I t seems possible that c o - a r t i c u l a t o r y information provided by neigh-bouring segments compensates for any reduction i n duration of c e r t a i n consonants. For example, the voiceless phases of the resonants and semi-vowels which have i n t h i s study been assigned to the resonants and semi-vowels, provide information required f o r perception of both consonants. Thus, overlapping of gestures s a t i s f i e s the requirements of both tem-poral organization and target a r t i c u l a t i o n s required for perception. The only way i n which i t may be possible to choose between the alternate hypotheses proposed above would be to analyze c l u s t e r s i n which the second consonants (perhaps stops) had. s u f f i c i e n t l y long dura-tions as singletons to allow s i g n i f i c a n t reduction i n the c l u s t e r . Such reduction would be strong evidence i n support of the f i r s t explanation. One could a l s o . i n v e s t i g a t e post-vocalic c l u s t e r s i n which the f r i c a t i v e i s the second consonant; there i s not as yet s a t i s f a c t o r y evidence show-ing, however, that post-vocalic c l u s t e r s are subject to the same con-s t r a i n t s as p r e - v o c a l i c . Further, i n s i n g l e words, f r i c a t i v e lengthening i n f i n a l p o s i t i o n would always occur. Differences between age groups i n terms of temporal organization 76 of c l u s t e r s are, i n the present data, e n t i r e l y r e s t r i c t e d to those c l u s t e r s with /!/. Duration of / l / a f t e r f r i c a t i v e s was increased r e l a -t i v e to duration of /!/ as a singleton consonant by the f i v e - y e a r - o l d group, whereas for a l l other age groups, duration of / l / d i d not change s i g n i f i c a n t l y with context. As shown i n the r e s u l t s , differences between age groups i n terms of the amount by which duration of / l / reduced (with lengthening interpreted as negative reduction) were s i g n i f i c a n t only for the f i v e - y e a r - o l d group.compared with the adult group. Further, the age trend analysis indicated that duration of / l / i n c l u s t e r s r e l a t i v e to singleton / l / decreased l i n e a r l y with increasing age between f i v e and eleven years of age. 4.2 Theoretical Implications The differences between age groups observed i n production of c l u s t e r s with / l / , considered together with other information about con-sonant c l u s t e r s , suggest several i n t e r e s t i n g hypotheses about development of co-ordination of consonant c l u s t e r s i n general. F i r s t , Hawkins (1973) has suggested that lengthening of segments i n children'.s speech occurs generally i n c l u s t e r s with i n i t i a l f r i c a t i v e s , i n d i c a t i n g that lengthen-ing i s also present i n /sw/ c l u s t e r s . Data i n the present i n v e s t i g a t i o n demonstrate that t h i s i s not the case for the age groups studied, and that lengthening i s r e s t r i c t e d to c l u s t e r s with / l / . Differences between at l e a s t these age groups are hence apparently r e l a t e d to the combination of / l / w i t h - f r i c a t i v e s . . Hawkins further suggested that most of the elon-gation i n /!/ c l u s t e r s occurs during the voiceless period of the / l / a r t i c u l a t i o n . This observation, however, has not been corroborated i n 77 the present data, since the r e l a t i v e duration of v o i c e l e s s to voiced / l / i s approximately the same for c h i l d r e n and adults. Thus Hawkins's hypothesis (p. 206) that a lack of co-ordination between v o i c i n g onset and p o s i t i o n i n g of upper a r t i c u l a t o r s i s l a r g e l y responsible for elonga-t i o n of / l / i s not substantiated i n the present i n v e s t i g a t i o n . Hawkins has postulated that lengthening of / I / i n young c h i l -dren's speech may be r e l a t e d to greater d i f f i c u l t y of a r t i c u l a t i o n of that consonant; / s / may also be lengthened more (or reduced less) than other f r i c a t i v e s i n c l u s t e r s because i t may be a r t i c u l a t o r i l y more d i f -f i c u l t . In the present data, no apparent difference between reduction of / s / and reduction of / f / i n c l u s t e r s was noted, i n d i c a t i n g that par-t i c u l a r d i f f i c u l t y which may be associated with / s / i n c l u s t e r s i s not r e f l e c t e d i n duration. Second, the concept of " a r t i c u l a t o r y d i f f i c u l t y " should not i n any case be used to explain elongation of / l / i n c l u s t e r s as the concept i t s e l f i s not well-defined. Third, there i s l i t t l e e v i -dence, apart from data on sequence of a c q u i s i t i o n of consonants (Wellman et a l . 1936; Templin 1957; Olmsted 1971), to support the proposal that / l / i s more d i f f i c u l t to produce i n a l l contexts, or that i t i s only more d i f f i c u l t to co-ordinate i n c l u s t e r s . Evidence of a c q u i s i t i o n of c l u s t e r s i n very young ch i l d r e n i s of considerable s i g n i f i c a n c e i n i n t e r p r e t i n g elongation of / l / i n c l u s t e r s by five-year-olds. I t has been shown (Menyuk and K l a t t 1968; Kornfeld 1971a,b) that c h i l d r e n often delete one consonant of a c l u s t e r ; as the c h i l d grows older, features of the deleted consonants are included with the remaining consonant u n t i l eventually the c h i l d develops the capacity 78 (at e i t h e r a phonetic or phonological level) to segment the c l u s t e r of features into two or more consonants. I t i s possible that i n the c h i l d ' s speech, temporal constraints are i n f l u e n t i a l i n consonant d e l e t i o n i n c l u s t e r s . A timing-dominant system, s i m i l a r to. that described by Ohala (1970) i n which a r e l a t i v e l y , f i xed time i s a l l o t t e d . f o r s a t i s f a c t o r y production of a consonant c l u s t e r may necessitate the d e l e t i o n of a consonant. This hypothesis pre-supposes that the c h i l d ' s c o n t r o l of timing of onset and completion of a r t i c u l a t o r y gestures i s not s u f f i c i e n t l y developed to s a t i s f y the requirements of a r i g i d timing system while s t i l l achieving the necessary a r t i c u l a t o r y gestures within the utterance. Support for t h i s assumption i s to some extent provided by Eguchi and Hirsh's (1969) and Tingley and Allen's (1974) findings that timing control i n c h i l d r e n i s generally sub-j e c t to greater v a r i a b i l i t y than i n adults, and i n Port and Preston's observation (1974) that voice onset time values for / t / , which may require considerable p r e c i s i o n of r e l a t i v e onset of gestures, show greater v a r i a b i l i t y i n young c h i l d r e n . In addition, measurements i n the present study indicate greater, o v e r - a l l variance of consonant duration i n the younger age groups. The s p l i t t i n g process which i s observed, though r a r e l y , i n con-sonant c l u s t e r production, and which i s characterized by i n s e r t i o n of a schwa vowel between two consonants (Kornfeld 1971b) can be considered i n t h i s i n t e r p r e t a t i o n as one attempt to overcome the demands of a timing-dominant system;, the c h i l d ' s timing,control i s not developed enough to enable him to produce the required segments within the l i m i t e d time 79 allowed. Consequently, the segmentation of clustered features i s exagger-ated i n the s p l i t c l u s t e r s , allowing target a r t i c u l a t i o n s of consonants to be achieved. At t h i s point, an articulation-dominant system (Ohala 1970) best describes the c h i l d ' s production. With neurological development, i t i s presumed that the c h i l d acquires the a b i l i t y to co-ordinate the consonants of the c l u s t e r into a more cohesive unit characterized by reduction of o v e r a l l durations r e l a -t i v e to the sum of the duration of consonants as singletons, which again s a t i s f i e s the requirements of a timing-dominant system. Several observations can be c i t e d i n support of the above i n t e r -p r e t a t i o n of development of consonant c l u s t e r s . In production of / l / c l u s t e r s , . / I / i s frequently omitted.by very young c h i l d r e n such that /slip/becomes / s i p / and / b l u / becomes /bu/. Later, the c h i l d may s p l i t the c l u s t e r so that / b l u / becomes /belu/. Results of the present study indicate another stage i n a c q u i s i t i o n of c l u s t e r s : elongation of / l / by five-year-olds may be interpreted as a further stage of the s p l i t t i n g process, i . e . , an attempt by the c h i l d to achieve target a r t i c u l a t i o n of / l / by r e l a x i n g the demands of the timing programme. The l i n e a r age trend noted i n the data provides further support for the hypothesis. As the c h i l d ' s o v e r a l l timing control improves with age, as Tingley and Allen's (1974.) research i n d i c a t e s , he i s able to achieve target a r t i c u l a t i o n of. / I / within an i n c r e a s i n g l y r i g i d time schedule.. The l i n e a r age trend would not be expected to extend beyond eleven years, as the c h i l d at t h i s age has approximately the same timing control as adults. A d d i t i o n a l data are required, however, to t e s t t h i s conclusion. 80 Another point of i n t e r e s t i n the a c q u i s i t i o n of consonant c l u s t e r s i s that i n c l u s t e r d e l e t i o n i t i s generally the ontogenetically l a t e r con-sonant which i s deleted. Evidence, such as EMG data, which may reveal why such consonants are deleted i s at present not a v a i l a b l e ; s i m i l a r l y , evidence explaining the o v e r a l l sequence of a c q u i s i t i o n of consonants has not yet been produced. Although the reason for d e l e t i o n of ontogenetically l a t e r conson-ants i s at present not c l e a r , the observation may be of use i n t e s t i n g the hypothesis that elongation i s not s p e c i f i c to c e r t a i n consonants, regardless of context, but i s a phase i n the i n t e g r a t i o n of consonants into a c l u s t e r which represents an attempt by the c h i l d to r e a l i z e the previously omitted, segment. I f t h i s i s the case, one would expect elonga-t i o n of / s / i n a c l u s t e r such as /sp/ or / s t / , i n which the / s / i s often deleted at an e a r l i e r age. I f elongation of / s / i n / s / + stop c l u s t e r s , and.reduction of / s / i n / s i / c l u s t e r s , occurred together i n the speech of a given subject, i t would be strong support for the hypothesis that elon-gation may be the r e s u l t . o f the c h i l d ' s attempt to achieve the a r t i c u l a -t i o n e a r l i e r precluded, by a more r i g i d time schedule. The implication.of the proposed sequence of a c q u i s i t i o n of con-sonant, c l u s t e r s i s that i n the adult a f a i r l y r i g i d timing-dominant system controls duration of speech segments; t h i s system allows s u f f i c i e n t time for each segment to be completed. For the c h i l d , however, the time allowed i n the adult model i s not s u f f i c i e n t for completion of a l l gestures. Consequently, to comply with.the.adult temporal model, he f i r s t omits c e r t a i n features; eventually he learns to e s t a b l i s h h i s own temporal system 81 which allows enough time to complete a l l necessary segments. The obser-vation that five-year- and seven-year-old groups can be roughly separated from the older age groups on the basis of absolute duration of a l l con-sonants measured.is further evidence that the timing-programme used by young c h i l d r e n i s d i f f e r e n t from that of adults. Gradually, as h i s neuro-muscular controls develop, the time programme for the c h i l d ' s speech i s a l t e r e d to approximate that of the adult. 4.3 Conclusion Although the above hypothesis i s l a r g e l y speculative, the e v i -dence presented i s c e r t a i n l y suggestive of the importance of temporal constraints i n a c q u i s i t i o n of consonant c l u s t e r s . Further i n v e s t i g a t i o n of such constraints i s nonetheless e s s e n t i a l to an understanding of the p r i n c i p l e s governing consonant d e l e t i o n i n children's speech and conson-ant reduction i n adults'. Further i n v e s t i g a t i o n i s also necessary to more c l e a r l y reveal the manner i n which a l l segment durations are deter-mined i n a d u l t s 1 speech, i . e . , the r e l a t i v e importance of timing-dominant and articulation-dominant . systems. Although the hypothesis pro-posed above i s based.on the assumption that segment durations i n adult speech are l a r g e l y determined, by a timing-dominant system, the evidence supporting t h i s assumption i s by no means conclusive. One l a s t point which requires c l a r i f i c a t i o n i n studies of temporal constraints a r i s e s from Lehiste's hypothesis (197.2) that resonants and vowels are fused into a s i n g l e temporal unit. I f t h i s hypothesis i s substantiated by further research, i t may be more p r o f i t a b l e to investigate duration i n terms of resonant + vowel sequences. 82 The r e s u l t s of the present study suggest s e v e r a l i n t e r e s t i n g aspects of development o f consonant c l u s t e r production i n c h i l d r e n ' s speech, and even more g e n e r a l l y , of development of temporal o r g a n i z a t i o n i n c h i l d r e n ' s speech. Although these r e s u l t s cannot i n themselves pro-vide the answers to many of the questions r e l a t e d to t h i s development, they do suggest r e l e v a n t aspects of the problem f o r f u r t h e r research. SELECTED BIBLIOGRAPHY A l l e n , G. D. 1973. Segmental timing c o n t r o l i n speech production. Journal of Phonetics 1. 207-225. Eguchi, S. and I. J . Hirsh. 1969. Development of speech sounds i n c h i l d r e n . Acta Oto-Laryngologica Suppl. 257. 1-51. Gi l b e r t , J . H. 1976. A voice onset time analysis of a p i c a l stop production i n 3-year olds. To appear i n Journal of Child Language. Gregorski, R. and L. Shockey. 1971. Word-unit temporal compensation. O.S.U. Working Papers in Linguistics 9. 145-165. Gregorski, R. and L. Shockey. 1972. A note on temporal compensation. O.S.U. Working Papers in Linguistics 12. 87-88. Haggard, M. P. 1973a. Abbreviation of consonants i n English pre- and post-v o c a l i c c l u s t e r s . Journal of Phonetics 1. 9-24. Haggard, M. P. 1973b. Correlations between successive segment durations: values i n c l u s t e r s . Journal of Phonetics 1. 111-116. Hawkins, S. 1973. Temporal coordination of consonants i n the speech of ch i l d r e n : preliminary data. Journal of Phonetics 1. 181-218. Ingram, D. 1975. The a c q u i s i t i o n of f r i c a t i v e s and a f f r i c a t e s i n normal and l i n g u i s t i c a l l y deviant c h i l d r e n . In A. Caramazza and E. Z u r i f f (eds.) The Acquisition and Breakdown of Language. Baltimore: The John Hopkins University Press. Ingram, D. 1976. Phonological Disability in Children. To appear: London: Edward Arnold Ltd. Kornfeld, J . R. 1971a. Theoretical issues i n c h i l d phonology. Papers, 7th Regional Meeting, Chicago Linguistics Society. 454-468. Kornfeld, J . R. 1971b. What i n i t i a l c l u s t e r s t e l l us about a c h i l d ' s speech code. Q.P.R., Research Laboratory of Electronics, M.I.T. 101. 218-221. Kozhevnikov, V. A. and L. A. Chistovich. 1965. Speech: Articulation and Perception. Washington, D.C: J o i n t Publications Research Service. Lehiste, I. 1970. Temporal organization of spoken language. O.S.U. Work-ing Papers in Linguistics 4. 96-114. 83 84 Lehiste, I. 1972. Manner of a r t i c u l a t i o n , p a r a l l e l processing, and the perception of duration. O.S.U. Working Papers in Linguistics 12. 33-52. Lisker, L. and A. Abramson. 1965. Voice onset time i n stop consonants: acoustic analysis and synthesis. In D. E. Commins (ed.) Proc. Fifth Int. Cong. Acoust. Liege: Imp. G. Thone. Locke, J . L. 1972. Ease of a r t i c u l a t i o n . Journal of Speech and Hearing Research 15. 194-200. Menyuk, P. and D. H. K l a t t . 1968. The c h i l d ' s production of i n i t i a l con-sonant c l u s t e r s . Q.P.R., Research Laboratory of Electronics, M.I.T. 91. 205-213. Nelson, K. 1973. Structure and strategy i n learning to t a l k . Monographs of the Society for Research in Child Development 38. Ohala, J . 1970. Aspects of the c o n t r o l and production of speech. U.C.L.A. Working Papers in Phonetics 15. Ohala, J . 1972. The regulation of timing i n speech. Paper presented to the 1972 Conference on Speech Communication and Processing, Newton, Mass. Ohman, S. E. G. 1966. C o a r t i c u l a t i o n i n VCV utterances: spectrographic measurements. Journal of the Acoustical Society of America 39. 151-168. Olmsted, D. 1971. Out of the Mouth of Babes. The Hague: Mouton. Peterson, G. E. and I. Lehiste. 1969. Duration of s y l l a b l e n u c l e i i n English. Journal of the Acoustical Society of America 32. 693-703. Peterson, G. E. and N. P. McKinney. 1961. The Measurement of Speech Power. Phonetica 7. 65. Port, D. K. and M. S. Preston. 1974. Early s p i c a l stop production: a voice onset time a n a l y s i s . Journal of Phonetics 2. 195-210. S l i s , I. H. 1971. A r t i c u l a t o r y e f f o r t and i t s durational and EMG cor-r e l a t e s . Phonetica 23. 171-188. Templin, M. C. 1957. Certain Language Skills in Children. Minneapolis: University of Minnesota Press. Tingley, B. M. and G. D. A l l e n . 1975. Development of speech timing con-t r o l i n c h i l d r e n . Child Development 46. 186-194. 8 5 Wellman, B. L., I. M. Case, I. G. Mengert, and D. E. Bradbury. 1936. Speech sounds of young c h i l d r e n . University of Iowa Studies in Child Welfare 2. Winer, B. J . 1962. Statistical Principles in Experimental Design. New York: McGraw-Hill. APPENDIX A BACKGROUND INFORMATION ABOUT SUBJECTS 86 87 Subject Age at Testing Sex Family Templin-Darley Score P . l 10 yr . 3 mo. M A P.2 9 yr. 7 mo. F B P.3 6 yr. 8 mo. F A P.4 6 yr . 1 mo. F C 1.1 5 yr . 3 mo. M D 41 1.2 5 yr . 1 mo. M E 34 1.3 5 yr . 2 mo. F F 47 1.4 5 yr. 5 mo. F G 40 1.5 5 yr. 1 mo. F H 34 2.1 7 yr. 4 mo. F I 45 2.2 7 yr. 1 mo. F . J 48 2.3 7 yr. 5 mo. M G 47 2.4 7 yr. 2 mo. F C 48 2.5 7 yr. 2 mo. F K 47 3.1 9 yr. 4 mo. F I 49 3.2 9 yr. 2 mo. M L 50 3.3 9 yr. 5 mo. F M 50 • 3.4 9 yr. 4 mo. M N 47 3.5 9 yr. 3 mo. M C 50 4.1 11 yr. 5 mo. F I 50 4.2 11 yr. 5 mo. F M 50 4.3 11 yr . 3 mo. F 0 50 4.4 11 yr. 6 mo. F P 49 4.5 11 yr . 6 mo. F Q 50 5.1 24 yr. 5 mo. F R 50 5.2 26 yr. 8 mo. F S 50 5.3 28 yr. 7 mo. M T 50 5.4 23 yr. 2 mo. M U 50 5.5 23 yr . 7 mo. F V 50 APPENDIX B LINGUISTIC BACKGROUND OF SUBJECTS' PARENTS 88 89 Subject Mother's Native Language Father's Native Language p . l Canadian English Canadian English P.2 Canadian English Canadian English P.3 Canadian English Canadian English P.4 Canadian English Canadian Engl i s h 1.1 German German 1.2 Canadian English Canadian English 1.3 Canadian Engl i s h Canadian English 1.4 Canadian English Canadian Engl i s h 1.5 Canadian English Canadian English 2.1 Canadian English Canadian Engl i s h 2.2 Canadian English Canadian English 2.3 Canadian English Canadian English 2.4 Canadian English Canadian English 2.5 Canadian English Canadian English 3.1 Canadian English Canadian English 3.2 German B r i t i s h English 3.3 Canadian English Canadian English 3.4 B r i t i s h English B r i t i s h E n g l i s h 3.5 Canadian English Canadian English 4.1 Canadian English Canadian English 4.2 Canadian English Canadian English 4.3 B r i t i s h English B r i t i s h English •4.4 Canadian English Canadian English 4.5 Canadian Eng l i s h I t a l i a n APPENDIX C WORD LISTS 90 91 L i s t l L i s t 2 L i s t 3 fought l i p feet fought fought slop f l e e t seep switch slop l i c k feet sweep f l i p wick lock watch f l e e t sweet s l o t sick leak f l o p s l i p whip feet s l o t swap swat sweep l o t f i t s l i t l o t leap lock swap sock sweep sop feet week sick s l i p f l i t f l o p whip fought f l e e t fought swat sock s l o t sleep leap s i t f l e e t walk leak sock switch switch switch s i t sweet swap sleep l i c k wheat f i t seep f l i c k week s l e e t s l e e t seep wick f l i p swat sop l i p f l i c k f l e e t f i t seat wick seat sweet switch s l i t slop walk s l i p whip walk l o t f l o p f l o c k fought sleep seat f i t s l i t leak leap watch switch s l e e t lock watch feet f i t l i p s i t f i t wheat f l e e t week fl o c k s i c k feet f l i c k l i c k f l o c k wheat sop In Vancouver English, spoken by a l l subjects, the low back vowel i n a l l words i n the above l i s t s i s the same vowel /o/. APPENDIX D DURATION OF VOICED AND VOICELESS /!/ and /w/ 92 93 Age Context Voiceless / l / Voiced / l / N X S.D. Range X S.D. Range Adult / s - / 30.7 13.6 50 41.7 10.7 40 30 / f - / 26.0 18.9 80 43.0 19.4 80 30 11.yr. / s - / 37.2 18.7 70 31.0 15.4 60 29 / f - / 26.3 11.6 50 37.0 18.8 80 27 9 yr. 35.5 19.8 80 41.7 23.6 110 29 / f - / 37.1 19.6 80 45.0 20.8 80 28 7 y r . 36.1 22.4 120 57.9 24.8 110 28 / f - / 36.3 17.9 100 53.7 17.2 80 27 5 yr. 48.6 24.2 120 67.2 31.7 150 29 / f - / 40.0 22.0 120 67.7 30.5 120 26 Age Context Voiceless /w/ Voiced /w/ N X S.D. Range X S.D. Range Adult / s - / 37. 0 23. 3 70 32.0 13.5 60 30 11 y r . / s - / 43. 6 18.1 70 28.6 16.2 80 28 9 yr. / s - / 39. 6 26.6 120 35.7 20.2 70 27 7 yr. / s - / 43. 8 26.7 110 43.4 22.3 120 32 5 yr. / s - / 39. 6 30.8 140 52.5 31.4 130 24 APPENDIX E SUBJECT DURATION DATA 94 95 Measurements i n msec. for 5-year--old group (/s/. and / f / ; /!/ and /w/) / s / / f / Subject Context Mdn. Range N Mdn. Range N 1.1 Alone 200 150 6 170 130 6 / - l / 190 120 5 140 60 7 /-w/ 190 80 5 1.2 Alone 230 110 6 180 70 6 / - ! / 180 60 6 165 90 4 /-w/ 185 30 6 1.3 Alone 195 70 6 180 80 6 / - I / 125 50 6 220 150 5 /-w/ 160 80 5 1.4 Alone 210 80 7 240 150 4 / - l / 155 100 6 160 130 4 /-w/ 220 180 3 1.5 Alone 140 40 6 140 80 6 / - ! / 135 110 6 140 120 6 V-w/ 120 80 5 /V /w/ Subject Context Mdn. Range N Mdn. Range N 1.1 Alone 100 60 5 95 50 7 / s - / 100 40 5 100 40 5 / f - / 100 40 7 1.2 Alone 130 70 5 110 140 6 / s - / 155 140 6 105 90 6 / f - / 145 60 4 1.3 Alone 75 60 6 65 70 6 / s - / 110 90 6 80 70 5 / f - / 90 40 5 1.4 Alone 85 120 8 110 130 4 /s-/ 90 40 6 70 20 3 / f - / 75 90 4 1.5 Alone 70 40 5 55 40 6 /s-/ 115 100 6 90 100 5 / f - / 110 50 6 96 Measurements in msec. fo r 7-year--old group ( A V and / f / ; /!/ and /w/) / s / / f / Subject Context Mdn. Range N Mdn. Range N 2.1 Alone 195 140 6 190 40 6 / - ! / 180 40 5 175 130 6 /-w/ 190 130 6 2.2 Alone 180 80 6 130 70 6 / - ! / 155 50 6 120 40 6 /-w/ 125 80 6 2.3 Alone 240 100 6 180 140 5 / - ! / 190 80 5 150 110 5 /-w/ 150 90 6 2.4 Alone 190 90 6 180 100 7 / - ! / 150 70 5 160 50 6 /-w/ 140 110 8 2.5 Alone 165 .40 6 140 30 5 / - l / 180 80 7 135 50 4 140 40 6 Subject Context / ! / /w/ Mdn. Range N Mdn. Range N 2.1 Alone 85 50 4 60 20 5 / s - / 80 60 5 70 40 6 / f - / 105 20 6 2.2 Alone 70 50 5 105 80 6 / s - / 85 30 6 100 40 6 / f - / 90 30 6 2.3 Alone 85 70 6 70 160 6 /s - / 110 60 5 90 70 6 / f - / 90 60 5 2.4 Alone 100 50 6 80 60 6 / s - / 90 30 5 85 90 8 / f - / 80 30 6 2.5 Alone 80 50 6 80 40 6 / s - / 90 40 7 65 60 6 / f - / 90 50 4 97 Measurements i n msec. for 9-year-•old group and / - f / ; /!/ and /w/) / s / / f / Subject Context Mdn. Range N Mdn. Range N 3.1 Alone 190 60 6 150 60 6 / - ! / 175 50 6 150 60 5 /-w/ 180 80 5 3.2 Alone 140 90 6 95 70 6 / - l / 110 40 6 105 90 6 /-w/ 150 50 5 3.3 Alone 170 90 7 185 40 6 /-V 170 60 5 130 50 6 /-w/ 165 60 6 3.4 Alone 160 100 5 125 30 4 / - ! / 115 50 6 100 50 5 /-w/ 150 50 5 3.5 Alone 180 50 6 180 110 6 /-V 130 140 6 145 70 6 /-w/ 130 90 6 / V /w/ Subject Context Mdn. Range N Mdn. Range N 3.1 Alone 95 40 6 75 50 6 / s - / 75 40 6 90 20 5 / f - / 100 50 5 3.2 Alone 70 60 6 70 50 4 / s - / 75 30 6 70 30 5 / f - / 85 30 6 3.3 Alone 95 70 6 65 60 6 / s - / 50 70 5 65 40 6 / f - / 95 60 6 3.4 Alone 90 40 3 65 80 8 / s - / 75 80 6 70 90 5 / f - / 50 30 5 3.5 Alone 65 50 6 85 50 6 / s - / 90 50 6 80 30 6 / f - / 60 40 6 98 Measurements i n msec. for 11-year--old group (/s/ and / f / ; /!/ and /w/) / s / / f / Subject Context Mdn. Range N Mdn. Range N 4.1 Alone 160 90 6 180 70 6 / - l / 180 60 6 135 60 6 /-w/ 165 110 6 4.2 Alone 150 40 4 120 60 6 / - l / 120 40 5 100 40 4 /-w/ 135 50 6 4.3 Alone 180 40 6 140 50 5 /-!/ 120 40 6 125 20 6 145 60 6 4.4 Alone 155 110 6 135 80 6 /-V 105 60 6 120 80 5 /-w/ 115 40 4 4.5 Alone 150 60 6 150 80 6 / - l / 135 40 6 100 50 6 /-w/ 140 40 6 / ! / /w/ Subject Context Mdn. Range N Mdn. Range N 4.1 Alone 80 40 6 75 30 6 / s - / 80 40 6 85 50 6 / f - / 75 80 6 4.2 Alone 65 50 6 60 60 6 / s - / 60 30 5 55 40 6 / f - / 40 40 4 4.3 Alone 65 50 6 55 40 6 / s - / 60 10 6 70 30 6 / f - / 60 50 6 4.4 Alone 60 30 6 60 50 5 / s - / 45 60 6 70 20 4 / f - / 50 50 5 4.5 Alone 75 100 6 80 70 6 / s - / 85 20 6 80 60 6 / f - / 80 20 6 99 Measurements i n msec. for adult group (/s/ and / f / ; /!/ and /w/) / s / / f / Subject Context Mdn. Range . N Mdn. Range N 5.1 Alone 175 40 6. 175 70 6 /-V 145 40 6 140 50 6 /-w/ 165 30 6 5.2 Alone 195 30 6 175 20 6 / - ! / 150 30 6 140 50 6 /-w/ 140 80 6 5.3 Alone 175 40 6 150 130 6 /-V 145 - 20 6 95 70 6 /-w/ 165 30 6 5.4 Alone 130 80 6 130 70 6 /-V 100 60 6 105 70 6 /-w/ 95 60 6 5.5 Alone 165 40 6 140 30 6 / - l / 125 40 6 120 20 6 /-w/ 130 60 6 / l / /w/ Subject Context Mdn. Range N Mdn. Range N 5.1 Alone 90 60 6 50 60 6 / s - / 60 30 6 45 20 6 / f - / 75 40 6 5.2 Alone 80 40 6 70 50 5 / s - / 80 40 6 80 30 6 / f - / 75 50 6 5.3 Alone 95 70 6 80 80 6 / s - / 80 30 6 80 50 6 / f - / 85 70 6 5.4 Alone 75 50 6 65 70 6 / s - / 70 20 6 70 30 6 / f - / 60 30 6 5.5 Alone 80 40 6 70 70 6 / s - / 70 20 6 70 40 6 / f - / 60 .10 6 APPENDIX F SUMMARY TABLES FOR ANALYSIS I 1 0 0 101 Source SS df MS F Analysis Ia: E f f e c t s on absolute durations of consonants (/s/, /f/) = Age 1,162.82 4 290.70 30.93** Subject 1,745.21 20 87.26 9.28** Phone 542.56 1 542.56 57.73** Context 969.64 1 969.64 103.16** Age x Phone 65.31 4 16.32 1.74 Age x Phone x Context 61.67 4 15.41 1.64 Error 5,084.85 541 9.40 Analysis Ib: E f f e c t s on absolute durations of consonants (/!/, /w/): Age 692.62 4 173.15 28.39** Subject 641.26 20 32.06 5.26** Phone 48.30 1 48.30 7.92** Context 4.53 1 4.53 .74 Age x Phone 30.16 4 7.54 1.23 Age x Phone x Context 61.12 4 15.28 2.50* Error 3,294.10 540 6.10 Analysis Ic: E f f e c t s on absolute durations of consonants (/s/): Age 693. .91 4 173. ,48 20. .43** Subject 1,362. .25 20 68. .11 8. .02** Context 729. .18 2 364, .59 42. .96** Age x Context 110. .66 8 13, .83 1, .63 Error 3,386, .41 399 8, .49 Analysis Id: E f f e c t s on absolute durations of consonants (/!/) : Age 839, .69 4 . 209, .92 40, .58** Subject 532, .34 20 26. ,62 5, .14** Context 8, .19 2 4, .10 .79 Age x Context 107, .18 8 13. ,40 2, .59** Error 2,022, .92 391 5. .17 * S i g n i f i c a n t at P < .05. ** S i g n i f i c a n t at P < .01. APPENDIX G HOMOGENEOUS SUBSETS DETERMINED BY NEWMAN-KEULS (ANALYSIS I a-d) 102 103 Analysis Ia: Age: (11-year-olds, Adults, 9-year-olds); (7-year-olds, 5ryear-olds)* Phone: ( f ) ; (s) Context: (Before /I/); (Alone) Analysis Ib: Age: (11-year-olds, Adults, 9-year-olds); (7-year-olds); (5-year-olds) Phone: ( / l / ) ; (/w/) Context: (After / s/, Alone) Analysis Ic: Age: (11-year-olds, Adults, 9-year-olds); (5-year-olds, 7-year-olds) Context: (Before / l / , Before /w/); (Alone) Phone i n Analysis Ic i s / s / Analysis Id Age: (Adults, 9-year-olds); (11-year-olds); (7-year-olds); (5-year-olds) Context: (After / f / , Alone, A f t e r /s/) Phone i n Analysis Id i s /!/ * Variables within parentheses are those which are not s i g n i f i -cantly d i f f e r e n t at P < .05, i . e . , parentheses i n d i c a t e homogeneous sub-sets. Subsets are s i g n i f i c a n t l y d i f f e r e n t at P < .05. APPENDIX H SUMMARY TABLES FOR ANALYSIS II 104 105 Source SS df MS F Analysis I l a : Difference i n msec, between / f / alone and / f / before /!/: Between groups 979.00 4 244. 75 .38 Within groups 12,957.00 20 647. 85 Total 13,936.00 24 Analysis l i b : Difference i n msec. between /w/ alone and /w/ a f t e r / s / : Between groups 54.00 4 13. 50 .05 Within groups 5,340.00 20 269. 70 Total 5,394.00 24 Analysis l i e : Difference i n msec. between /!/ alone and / l / i n c l u s t e r s : Between subjects 13,308.00 24 Age 5,623.00 4 1,405. 75 3 .66* Subjects within groups 7,685.00 20 384. 25 Within subjects 3,862.50 25 Context 24.50 1 24. 50 .13 Age x Context 273.00 4 68. 25 .38 Context x subjects within groups 3,565.00 20 178. 25 Analysis l i d : Difference i n msec. between / s / alone and / s / i n c l u s t e r s : Between subjects 19,248.00 24 Age 1,338.00 4 334. 50 .37 Subjects within groups 17,910.00 20 895. 50 Within subjects 7,512.50 25 Context 312.50 1 312. 50 1 .33 Age x context 2,520.00 4 630. 00 2 .69 Context x subjects within groups 4,680.00 20 234. ,00 * S i g n i f i c a n t at P < .05. 106 Age Adult 9-yr. 11-yr. 7-yr. 5-yr. Analysis l i e : Newman-Keuls t e s t f o r s i g n i f i c a n c e of differences i n age for amount of reduction of / l / i n c l u s t e r s : Adult 5.0 7.0 20.5 38.5* 9-yr. .2.0 15.5 23.5 11-yr. 13.5 13.5 21.5 7-yr. Truncated r 8.0 .95 (r,20) /n x MS error 25.87 31.40 34.72 37.09 S i g n i f i c a n t at P < .05. APPENDIX I SUMMARY TABLES FOR ANALYSIS I I I (a-d) 107 108 Source SS df MS F Analysis I l i a : Age of , trend analysis / f / i n c l u s t e r s : f o r reduction Linear 420.25 1 420. 25 .47 Quadratic 101.25 1 101. 25 .11 Cubic 2.25 1 2. 25 .00 Deviation 0.00 0 Within groups 14,340.00 16 896. 25 Total 14,863.75 19 Analysis I l l b : Age of trend analysis / s / i n c l u s t e r s : f o r reduction Linear 232.56 1 232. 56 .46 Quadratic 15.31 1 15. 31 .03 Cubic 248.06 1 248. 06 .49 Deviation 0.00 0 Within groups 8,102.50 16 506. 40 Total 8,598.44 19 Analysis I I I c : Age of trend analysis /!/ i n c l u s t e r s : f o r reduction Linear 1,600.00 1 1,600. 00 7.05' Deviation 281.25 2 140. 63 .62 Within groups 3,630.00 16 226. 88 Total 5,511.25 19 Analysis H i d : Age trend analysis of /w/ i n c l u s t e r s : for reduction Linear 36.00 1 36. 00 .11 Quadratic 5.00 1 5. 00 .02 Cubic 4.00 1 4. 00 .01 Deviation 0.00 0 Within groups 5,110.00 16 319. 38 Total 5,155.00 19 * S i g n i f i c a n t at P < .05. APPENDIX J GRAPHIC REPRESENTATION OF LINEAR REGRESSION OF /!/ DURATION 109 110 J N — T 5-year 7-year 9-year 11-year Mean (•) and range (I) of amount of re d u c t i o n i n msec, of / l / f o r the f i v e subjects i n each of four age groups. Dashed l i n e i s l i n e a r r e g r e s s i o n demonstrated by tren d a n a l y s i s . 

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.831.1-0093701/manifest

Comment

Related Items