An Analysis of the Variations from Standard English Pronunciation In

DOCUMENT RESUME

ED 064 751 24 CB 500 005 AUTHOR Williams, Frederick, Ed.; And Others TITLE An Analysis of the Variations fromStandard English Pronunciation in the Phonetic Performance of Two Groups of Nonstandard-English-SpeakingChildren. Final Report. INSTITUTION Texas Univ., Austin. Center forCommunication Research. SPONS, AGENCY Office of Education (DHEW), Washington, D.C. Bureau of Research. BUREAU NO BR-0-0336 PUB DATE Jul 71 GRANT OEG -32615-0050-5010(607) NOTE 206p. EDRS PRICE NF-$0.65 BC-$9.87 DESCRIPTORS *American English; *Articulation Ppemict4;*Dialect Studies; *Ninwity Group Children; Negro Dialects; *Nonstandard Dialects; Phonology; Sociolinguistics; Spanish Speaking; Speech Habits

ABSTRACT In this second of two studiesconducted with portions of the National Speech and Hearing Survey data,the investigators analyzed the phonetic variants from standardAmerican English in the speech of two groups of nonstandard Englishspeaking children. The study used samples of free speech andperformance on the Gold-Fristoe Test of Articulation from a group of192 Black children Wades 1-6 from Niagara Falls, New York, and a group of192 Mexican-American children (grades 1-6) from San Antonio, Texas.The study reports the frequencies of omission, substitution, anddistortion. The present data is compared with the previously reportedresults obtained from standard American English speakers from Marshall, Iowa.The major conclusion is that all phonetic variationsfrom standard American English can be attributed to one of thefollowing sources of variation: (1) reduction in the complexity of segments,which usually decreases with age; (2) differing phonologicalrules between standard English and a dialect of English; and (3)phonetic interference between a foreign primary language and English. (Seealso ED 042 756, ED 042 757, ED 042 758, ED 046 938, ED052 213) MI 8kc0004 U.S. DEPARTMENT OPHEALTH. EDUCATION WELFARE OPPICE OP EDUCATION THIS DOCUMENTHAS SUN REPRO. DUCED EXACTLY ASRECEIVED PROM THE PERSON ON ORGANIZATIONORM. MATING IT. POINTS OP VIEWOR OPIN. IONS STATED DO NOTNECESSARILY REPRESENT OFFICIAL OFFICEOF IOW Final Report CATION POSITION OR POLICY. Grant No. 32-15-0050-5010(607)

AN ANALYSIS OF THE VARIATIONS FROMSTANDARD ENGLISH PRONUNCIATION IN THE PHONETIC PERFORMANCEOF TWO GROUPS OF NONSTANDARD-ENGLISH-SPEAKING CHILDREN

Frederick Williams, Editor Helen S. Cairns Charles E. Cairns

Center for Communication Research School of Communication The University of Texas at Austin Austin, Texas July, 1971

This research was conducted by the Center forCommunication Research, The University of Texas at Austin, under sub-contractwith thz National Speech and Hearing Survey, Colorado StateUniversity, which in turn was supported by contract with the U.S. Office of EducationGrant #32-15- 0050-5010(607). The research reported hereinwas performed pursuant to a grant with the Office of Education, U.S. Department of Health,Education, and Welfare. Contractors undertaking such projects under Governmentsponsorship are encouraged to express freely their professionaljudgment in the conduct of the project.Points of view or opinions stated do not, therefore,necessarily represent official Office of Educationposition or policy. U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Office of Education Bureau of Research TABLE OF CONTENTS

Page FOREWORD iv ABSTRACT. Section

1. INTRODUCTION 1

1. 1.Overall Research Plan 1 1. 2 Phase I. 1 1. 3 Phase IIOOOOO . 2 2. PREDICTED VARIATION IN SPEAKERS OF BLACK AND MEXICAN-AMERICAN ENGLISH 5

2. 1.Theoretical Considerations. 6 2. 2.Results. from Marshalltown Study . . 6 2. 3.Published Work on Black and Mexican-American Dialect 9 2. 4.Published Work on Mexican-American DialectOOOOO 11 3. DATA COLLECTION AND INITIALTABULATION. 15 3. 1.National Speech and Hearing Survey Data Collection 15

3. 2.Procedure for the present StudyOOOOOOO. 15

3. 3.Two Excluded Phonemes.... 16 iii

Section Page

4. RESULTS 19 4. 1.Niagara Falls: The Black Sample...... 20

4. 2.San Antonio: The Mexican-AmericanSample. . 95 5. DISC USSION 177 5. 1.Evaluations of Empirical Predictions 177

5. 2.Sociolinguistic Implications...... 180

5. 3.Implications for Linguistic Theory.. OOOOOOOOO 182

REFERENCES.. O . 185

APPENDIX A 187 APPENDIXB...... 205 FOREWORD

This monograph reports the second and final phase of a research project based on speech data collected from United States schoolchildren by the National Speech and Hearing Survey.The project-- a phonological analysis of articulatory variants on Standard American English-- was initiated in June, 1970, with a study of standard English-speaking schoolchildren in Marshalltown, Iowa. The results of that first phase were published January, 1971, in a monograph entitled Aualysis of Production Errors in the Phonetic Performance of School-Age Standard- English-Speaking Children.This second-- an analysis of the speech of two groups of nonstandard-English-speaking children-- was initiated January, 1971. We are indebted to Dr. Forrest Hull, Director of the National Survey, for his cooperation in carrying out this research.Consultants Dr. Roger Shuy, Dr. Walter Stoltz, and Dr. Diana Natalicio aided in the planning stages.Dr. Helen Cairris undertook the day-to-day man- agement of the project.Dr. Charles Cairns served as the main linguistic consultant.Graduate assistants serving as typists and phonetic coders included Roslyn Freeman, Karen Hodges, Leslie Miller, Linda Sobin, and Kathleen Scirneca. Frederick Williams Project Director July, 1971

iv ABSTRACT

In this study the phonetic variantsof Standard American English in the speech of two groupsof nonstandard-English-speakingchildren were analyzed.The study included a group of192 Black children (grades 1-6) from Niagara Falls, New York,and a group of 192 Mexican-Ameri- can children(grades 1-6) from San Aitonio, TexasSpeech samples were obtained by the National Speechand Hearing Survey and weresamples of both Free Speech and ofperformance on the Goldman-FristoeTest of Articulation.This study is the second in aseries of two studies conducted on portions of the NationalSpeech and Hearing Survey data.The first investigated the speech of asample of standard-English-speaking children from Marshalltown,Iowa. This monograph reportsfrequencies of omissions, substitutions, and distortions acrossgrade levels for both groups.Greatest emphasis, however, is placed on the analysisof substitutions according to target phoneme, substituted phoneme,and phonetic features involved ineach substitution. A cluster of substitutions wasfound to be common to all three groups of speakers(considering both studies together);another cluster was found to be unique tothe Black and Mexican-American speakers; one substitution wascharacteristic only of the Mexican- American group. A major conclusion of tllistwo-part study is that allphonetic variations from StandardAmerican English can be explainedby appeal to one of the followingthree sources of variation: (1) Variation attributable to areduction in the complexityof segments, which usuallydecreases with age. Almostall variations of this type can be explainedby a phonological theory incorporatingthe ideas of markedness theory. (2) Variation attributable tothe differing phonological rules between Standard English and adialect of English, such as Black English. (3) Variation attributable tophonic interference between a foreign primary language andEnglish.This interference can be either purely articulatory or can bederived from perceptualinconsistencies resulting from competing perceptual cuesin the two languages. SECTION 1 INTRODUCTION

The research reported here is the second phase of a research project designed to analyze phonetic variations from Standard American English produced by selected groups of school children in the United States.In the total project three dialect areas have been investigated. The first phase (Williams(Ed.), Cairns, Cairns, & Blosser, 1971) dealt with children from Marshalltown, Iowa, grades 1-12 judged to be a Standard English dialect area.This second phase has investigated a group of Black children in NiagaraFalls, New York, (grades 1-6) and a group of Mexican-Americanchildren in San Antonio, Texas (grades 1-6) whose native language is Spanish. 1.1.Overall Research Plan The total research project has had a number of convergent goals, both theoretical and empirical.First, it was important to ascertain the relative incidence of the various types of phonetic variants (i.e. , omissions, distortions, and substitutions) in each of the three groups. The orientation of the research, however, has been toward an analysis of the instances in which one phoneme is substituted for another. The strategy of analysis has been to describe each phoneme as a bundle of phonetic features (Chomsky and Halle, 1968). The substitution is then characterized as a change in the value of certain of the features. The stability of individual features and the inter-relationships among them are being studied. 1.2. Phase I An important theoretical contribution of the first phase of this research was to evaluate a linguistic theory of phonemic inventories incorporating the concept of markedness. Such a theory assigns complexity values to phonetic features and predicts the relative complexity of various phonemes. The marking conventions specified by linguistic theory were supported by the data from the Marshalltown children.In almost every instance the substituted phoneme was less complex than the target phoneme which it replaced.It is important to note that the "complexity" referred to by markedness theory is not simply the opposite of "ease of articulation". Relative complexity in markedness theory derives from a number of sources, including perceptual distinc- tiveness, the frequency of occurrence of certain phonemes among

1 2 languages of the world, and physiological considerations such as deviation of the vocal apparatus from phonetic neutral position. For a complete discussion of the Marshalltown data and its explanation within the markedness framework, see Williams, et al. (1971).Appendix A of the current report presents the theoretical framework on which those analyses were based. 1. 3. Phase II Substitution data collected in this second phase of the research project provided a further test of the theoretical framework supported by the Marshalltown data.In addition, the generality of the phase one state- ments regarding the stability and inter-relationships among the phonetic features were evaluated by the data from the two new dialect areas. The second phase, however, had a more pronounced empirical orientation than did the Marshalltown study.In the present study we were dealing with two quite different types of nonstandard English speech. One of these referred to as Black English (hereafter BE), is commonly regarded as a dialect of English by sociolinguists. We accept all the assumptions which researchers in BE traditionally adopt, i. e. that BE is a true dialect in the sociolinguistic sense. Furthermore, we anticipated that the speech of the children of Niagara Falls would show certain aspects characteristic of BE. The speech of the Mexican-American sample, however, probably cannot be regarded as a dialect of Standard English in the traditional sociolinguistic sense.There is a large population of Mexican-Americans in the Southwest who are either fully bilingual Spanish-English in the home or who have Spanish as the primary language in the home and adopt English as the language of the community. The English of this population shows certain characteristics (which will be discussed in detail in the next section) resulting from the fact that English is either learned simultaneously with or is superimposed upon their native Spanish.It is this characteristic speech which we address in the present study and which we shall refer to as Mexican-American English (hereafter MAE). A number of publications dealing with Black and Mexican-American dialects allow predictions about what variations will be revealed in the study. (Section 2 of this report is devoted to a presentation of the relevant literature and predictions. ) There have been no large-scale quantitative studies, however, to ascertain exactly what kinds of variation, and in particular what kinds of substitutions, (and the incidence of both, ) are associated with these dialects of English. When both phases of the total project are considered together, it is reasonable to expect that a small core of variations common to the three dialect gronps will emerge. These variations might decrease with age, but should certainly be expected to remain at some constant frequency even I

3 into the adult population.In addition, it seems reasonable to anticipate that each dialect group will have a few types of variations which do not occur in the other two groups. The subjects for this project were subsamples of school children interviewed by the National Speech and Hearing Survey, conducted in 1968. Relevant details of the conduct of that survey as well asan explication of the methods involved in the present researchare presented in the third section of this report. Sections 4 and 5 will consist of results andcon- clusions, respectively. SECTION 2 PREDICTED VARIATION IN SPEAKERS OF BLACK AND MEXICAN-AMERICANENGLISH'

It is possible to make predictions about what substitutions will occur in both the Black and theMexican...American dialect populations. There are, essentially, three kinds of sources for these expectations. (I) Theoretical considerations presented in Appendix A; (2) Substitu- tions reported from the Marshalltown study.These substitutions presumably represent performance errors attributable for the most part to an immature vocal apparatus; (3) Published work describingchar- acteristics of Black and Mexican-American dialects.This latter source demands a bit of additional discussion. In the case of the native speakers of Spanish, substitutions can be predicted on the basis of phonetic interference (Weinreich, 1957) in a fairly straight-forward manner.Variants produced by Mexican-American children presumably result from the fact that Spanish and English have different phonemic inventories and the areas of non-overlap produce phonetic variation and perhaps misperceptions as well,This subject will be treated in some detail below when variant predictions for the Mexican-American group are discussed more fully. Literature dealing with Black dialect is, clearly, not so straight- forward. The "difference vs. deficit" controversy rages so intensely that it is questionable whether one is justified in referring to "substitution errors" (Williams, 1970). Many sociolinguists would much prefer the characterization that the black dialect specifies or allows different realizations of certain phonemes in certain positions than does Standard American English (SAE).Henceforth, this report will refer to "variants of SAE," rather than to "errors" or "misarticulations." The researchof Labov and his colleagues has shown that the variations from Standard English are by no means an all-or-none phenomenon. His discoveryof the social stratification of a number of phonological variables (Labov, 1966) shows that in many cases dialect differences are in actuality standard variations which vary quantitatively (rather than qualitatively) according to the race, social class and performance modality (1. e. whether reading aloud or talking conversationally) of the speaker.

1Theauthors would like to thank Linda Sobin for assistance in researching and reporting the literature discussed in this section.

5 6 Literature dealing with Black English (BE),then, is of an entirely different character than the literature dealingwith that dialect of English spoken by native speakers of Spanish (I,. e. ,Mexican-American English, MAE). Mexican-American English is predicted by thecomparison of two well-researched languages.Black English on the other hand, is anunwritten language, long rejected as merely an inferiorversion of SAE. Only recently has the structure and order of BlaukEnglish been recognised and described. A great deal of work remains to bedone before Black English and its relationship to SAE is fully understood.However it is clear that the relationship is a complex combination ofsociological, cultural and linguistic factors. The goal of the present research projectis to compare the variants of Standard English produced bychildren who speak MAE with those of children who speak BE. The making of such acomparison necessitates that consideration of the differential sourcesof Black and Mexican-Americqn dialects be held in abeyance. The remainder of this section will bedevoted to exploring the three available sources of prediction of variationfrom SAE in the present study.

2. 1.Theoretical Considerations Appendix A presents a linguistic theory ofphonemic inventories incorporating the notions of markedness.Such a theory predicts that in the case of articulatorysubstitutions, less complex speech soundswill be substituted for more complex ones.This hypothesis was confirmed by the Marshalltown data.The Marshalltown children, however,made few errors overall. A larger error corpus would represent amuch more stringent test of the complexity hypothesis, as manypredictions derivable from markedness theory were neither falsified norconfirmed by the Marshalltown data. For instance, the theory predicts that/k/ would be substituted for /g/.In the Marshalltown data, however,/g/ was omitted twice in the Goldman-Fristoe data, but it was never the victimof a substitution error. Only in a corpus with more phonetic variationcould some predictions be confirmed or falsified.

2. 2.Results from Marshalltown Study i

1 Table 1 shows all variants which occurred morethan twice in the 1 Marshalltown Goldman-Fristoe data.It was expected that most, if not I all, these variants would occur in theBE and MAE groups.Underlyi-ig 1 the use of the Marshalltown data as aprediction device for the present 1 study is the assumption that those errors areattributable to the immature ! articulatory capabilities of the subjects,rather than to dialect deviance i front SAE in the speech community. 7

Table 1 Variants Occurring more than Twice in the Goldman-Fristoe Data of Speakers of SAE in Grades 1-12 (from Williams, et al.,1971)*

IMP `aft

Target f +1- Feature MI U Feature PhonemeVariant Change Change

f b 6 -cnt; +voi M ant; U cnt M voi; U str

v b 29 -cnt M cor; U cnt; U str

d 18 -cnt U cor; b U cnt; U str

f 21 -cor U cor

t 5 -cnt U cor U cut; U str 8 s 4 +str U cor; U str

omit 5

z s 42 -voi U voi

distort 8 s omit 3

c 8 -cnt U cnt ii s 3 +ant U ant; U cor

t 6 +ant, -str U ant U cor; U str

Ca' g 5 +cnt M cnt Table 1 (contld)

+/- Feature M/ U Feature Target Variant Phoneme Change Change

011111.....gele, 6 -voi U voi g i i d 3 +ant; -str U ant; U car i i ,i w 31 -cns; -ant i 1 -voc; -cor none ( !

I omit 10 i 1 i

w 6 -cns; -voc I -ant; -cor i

A -lat U lat 1 omit 8 1 +voi M voi

In the columns indicatingfeature changes the feature valuesgiven are those of the substitutedphoneme which differ from thoseof the target phoneme. Notice that a feature maychange in marking value without changing in +/- value (see,for instance, the /b/ for/f/substitution). In the Marshalltown study thedata were not organized according to position of the target phoneme, asthey are in the present study. 1 'The target /r/ includes boththe true consonantal/r/ and the vocalic /.6 /.These two phonemes are treatedseparately in the present report. For an explanation of thephonetic features and theirabbreviations, see Table 1, Appendix A. 9

This is a good place to note, also, that all variants discussed in this research project are assumed to be attributable to linguistic per-. formance, not necessarily reflecting the basic linguistic competence of the individuals involved in the study.It is not advisable to make strong claims about an individual's linguistic competence based on information about his linguistic output alone.In the case of the Black and Mexican- American subjects in this study, however, many of the variations from SAE which they exhibit may well be the reflection of different internalized grammars. Labov (1970), Wolfram (1970), and others have been concerned with the development of appropriate formal mechanisms to account for the competence of speakers of dialects and idiolects which differ from SAE. However, data such as these collected by the present study cannot be construed as directly relevant to such theoretical issues. 2.3.Published Work on Black and Mexican-American Dialect 2.3.1.Studies of Black dialect.It is possible to abstract from recent studies of Black dialect specific predictions concerning deviations from SAE in the present study. As Faso ld and Wolfram (1970) point out, almost all the phonetic deviations from SAE associated with Negro dialect alternate with standard English forms in actual speech.Labov's studies indicate that the incidence of Black dialect features varies in the speech of individuals. Stewart (1964) refers to this quantitative variation as alternations between the acrolect (the version of the dialect closest to SAE) and the basilect (the version of the dialect furtherest from SAE). Considerable variation is evidenced within a language community as well as within individual members of the community. Therefore, the variations from SAE which are predicted on the basis of previous work in Black dialect are not expected to occur 100% of the time, but are predicted to account for a significant number of variants in the present study. According to Stewart (discussed in Faso ld, 1969) variations between basilect and acrolect are sensitive to developmental variables. Small children usually use the basilect, while youths and adults are more likely to use the acrolect. The remainder of this section will deal with specific predictions of deviations from SAE. 2.3.1.1. Omission and devoicing of final consonants. A number of writers (see, for instance, Labov, 1970, and Faso ld and Wolfram, 1970) have mentioned the tendency of Black speakers to delete the final consonant from certain types of final consonant clusters.The phonemes of the final cluster must share the same value for the feature of [voice] for this deletion to take place (Wolfram, 1970).Thus, /d/ is deleted 10 from where the final cluster is voiced, but/t/is not deleted from 'count' where the two members ofthe cluster do not match in voicing(i. e., /n/is voiced and /t/is voiceless). Seemingly independent of the tendency tosimplify final clusters is the tendency to devoice final voicedstops.This produces substitutions of /t/ for /d/, /k/ for /g/, and/p/ for /b/ in word final postion. Some writers report the frequentomission of final /d/ in words such as 'good' (i.e.,where the /d/ is not part of avoiced cluster).This sort of /d/ omission is different from syntacticvariation of Black English which realizes the past tense of verbs asthe infinitive form.Thus, the past tense of the verb 'play' is oftenrealized as 'play' rather than 'played. This deletion of final /d/is attributable to syntactic variaticnrather than to articulatory factors, however,and is qualitatively different from the omission of the final /d/ in 'good' or'hand. ' According to Faso ld and Wolfram(1970), final obstruents other than stops are also devoiced, but lessfrequently.This information leads to, the prediction of word-final/s/ for / z/,/f/ for /v/, and /a/ for /j/substitutions as well. 2.3.1.2. Omission of /r/ and121.The omission of /r/ and /A/ from clusters as well as frompost-vocalic positions is considered by most authorities to be characteristicof Black English.Labov's (1966) discovery that this variable is greatlystratified has apparently resulted in many experts' beingcautious in identifying these omissions ascharacteris- tic of Black dialect.In a recent study by Natalicio andWilliams (1971) a panel of 29 language specialists listened totapes of Black and Mexican- American children performing a standardlanguage test.The experts re- sponded with scaled evaluations of eachchild's deviation from SAE and a report of the criteria which hehad used in his evaluation.In that study, while the omission of /r/ andIll were noted by the experts as characteristic of the samples, they emphasizedthe possibility that the omissionscould be explained on the basis of regionaldialect variation or immaturity. 2.3. I. 3.Deviations associated with // and /W. The substitution of /d/ for rt. /in initial position has been shown byLabov (1966) to be included in the speech of probably everyspeaker of every dialect to English. Notice that /d/ for /is a frequent substitution in the Marshalltown study (see Table 1).That study also showed that the incidence of /d/ for /S /substitutions decreased with age. Thefrequency of this substitution is considerably greater inBlack English (Wolfram, 1969; Labov, 1970), however, than in most dialects of SAE.Fasold and Wolfram (1970) also report the substitution of /v/ for/ g/ in word-internal position in some dialects ofblack English.Labov (1970) also reports /v/for fiS / substitutions in word-final position. Wordfinal //is very rare in English 11 however, and is not included in the Goldman Fristoe test of articulation. It seems to be well accepted that /f/substitutes for /V in Black English, especially in medial and final positions.Wolfram (1970) claims, in fact, that this substitution represents the only invar- iant phonetic difference between Black English and Standard English. Other phonetic variations, such as /d/ for /t /, are found in many or all dialects of Standard English as well as those of Black English. He claims, however, that this is not the case for the /f/ for /0/ substitution.He notes the categorical absence of /f/ for /0/ substitutions in middle class speech.Table 1, however, reveals that the /f/for /0/ substitution was quite prevalent in the Marshalltown group. It is interesting to observe that variations in /1 / and RI/ were the phonological characteristics most frequently noted by the panel of experts in the Natalicio and Williams (1971) study.At the same time, the experts insisted that, given the variability of these phonemes in the overall population, their deviance was not regarded as " particularly important to overall linguistic performance." While substitutions per se have not been addressed in any one study of Black dialect, it is possible to abstract out the above predictions. For the most part, the predicted errors are not qualitatively different from those anticipated in any dialect of Standard or non-Standard English. Data gathered in the present study can be expected to test the generality of the above predictions and also to reveal any misarticulations unique to Black dialect if such qualitative deviations do in fact exist. 2.4.Published Work on Mexican-American Dialect In 1957 Weinreich described phonic interference between a person's native language and a second language.In that paper he presented a formula for predicting phonic interference. According to his system, phonic interference occurs when the second language has phonemes in its inventory which do not occur in the native language and also when the native language has phonemes which are not present in the second language.Spanish has a much simpler phonemic system than English (Stock- well and Bowen,1965), and there are many phonemes and phonemic dis- tinctions which exist in English which do not exist in Spanish.Thus, it is possible to predict which English will be misarticulated by native speakers of Spanish simply by looking at the phonemic inventories of the two languages.Reed, Lado, and Shen (1948) report many of the substitution errors typically observed in Spanish-speaking informants. Some of the errors reported in that paper have been substantiated recently by Natalicio and Williams (1971).Considered next are expectations of 12 substitutions for speakers of English who are native speakers of Mexican Spanish. There is neither a /0/ nor an /et /in Spanish, and /s/ and /d/ are usually substitutedfor each, respectively.The only mellow fricatives of Spanish are /f/ and /13/.The latter is a voiced bilabial fricative which does not occur in English. Spanish does not have adistinction between /b/ and /v/ ,so Spanish speakers frequently confuse/b/ and /v/ in English, using /8/ for both phonemes. In certaindialects of Spanish, notably the Texas dialect around the San Antonio area, many speakers have adopted the use of /v/ ,usually having it in free variation with /0/. (Natalicio, personal communication). 2.4.1. Variants involving fricatives and affricates. The /s/ phoneme is the only strident fricative in Spanish; / z/,///, and /6/ are absent. /6/is the only affricate, as Spanish does not have/3/.The foregoing results in /s/ being substituted frequently for /z/. The behavior of/6/ is a bit more complicated./6/is frequently substituted for /i/, but in addition,/6/is frequently misarticulated as /i/.This indicates that the Mexican-American speaker perceives the English distinction between/a/ and /6/ but is unable to use it correctly.The phonemes /i/ and /3/ (both non-occurring in Spanish) and /y/ (which does occur inSpanish) are frequently confused in articulation. Reed, Lado and Shen (1948) suggest that /I/ may be substituted for medial/3/, and /y/ for initial /3/, but there is apparently some question about the generality of thisobservation. As was noted above, /b/ and /v/ are often confused with each other.They may substitute for each other or they may each be replaced by /8/. 2.4.2. Variants involving stops. Word initial voiceless stops are aspirated in English, but not in Spanish.The failure of native Spanish speakers to aspirate initial stops frequently results in the perception of the stop as voiced by native speakers of English.Thus, according to Natalicio and Williams (1971), " 'coat, which in SAE is[khot],may be produced by the Mexican-American as Ekot] but perceived by SAE speakers as/got/, 'goat. ' " 2.4.3.Variants involving glides.The phoneme /h/ does not occur in S?anish, but the voiceless velar fricative /x/ does occur. TheSpanish speaker sometimes omits the English /h/ and sometimes replazes it with an /x/. 2.4.4.Variants involving /r/ and /i/.The liquids of Spanish are articulated differently from those of English.The MAE speaker frequently produces a trilled or flapped /r/ for the SAE/r/.In the present study, 13 however, variations in types of liquids were not recorded and described. Omissions of liquids and substitutions by other phonemes (such as /w/) were, however, noted. There are, then, a number of widely varying sources for error predictions in the present study.Considering the differential sources of articulatory errors for the Black and Mexican-American populations, a recently reported findingof Friedlander (1965) is highly counter- intuitive.He reports the results of articulation tests performed with speakers of Black dialect, SAE, and Mexican-American dialect.The results include only the target phonemes, however -- not the realizations in the case of substitution errors.Friedlander claims that all the phonemes for which variants were produced by the SAE speakers were varied by the Black andMexican-American speakers. In turn, all those for which variants were produced by the Black speakers were also varied by the Mexican-American children.(He excepts medial /t/ from this generalization, meaning, presumably that the Black speakers produced variants of medial /t/ but the Mexican-Americans did not. ) Replication of this finding would mean that there are no variants unique to Black dialect.It was hoped, however, that the present study would reveal a common pool of errors and unique groups of errors attributable to Black and Mexican-American speakers, respectively. SECTION 3 DATA COLLECTION AND INITIAL TABULATION

3. I.National Speech and Hearing Survey Data Collection During the school year 1968-1969 the National Speech and Hear- ing survey (NSHS) obtained speech samples from 38,802 United States school children. NSHS is based at Colorado State University, under the direction of Dr. Forrest Hull.The sample of 38,802 was drawn from the 41, 088, 138 member population of United States school children (1968-69 census). Samples were obtained from 100 school districts in 9 census divisions. A minimum sample of 384 children was tested from each district, evenly divided among grades 1-12, with the sample from each grade equally divided by sex. The speech sample obtained from each child consisted of the Goldman-Fristoe Test of Articulation, a few minutes of connected discourse, selected speech sounds elicited in a vowel context,and the repetition of a set of four sentences. The present study utilized the first two of these speech samples. Segments of connected discourse (Free Speech) were elicited in various ways, depending on the age level of the subject.Children in grades 1-3 were given the Goldman-Fristoe Sounds in Sentences Test; children in grades 4-9 were asked to make up stories in response to pictures; children in the upper grades were asked standar- dized questions to stimulate free speech. The testing of subjects was done in specially equipped soundproof vans by trained NSHS staff members. All speech samples were tape recorded.

3. 2.Procedure for the present Study

3. 2. 1.The samples. For the present study the researchers obtained speech samples from Black and Mexican-American school children.The Black sample was comprised of all speech samples obtained from grades 1-6 in Niagara Falls, New York. The Mexican- American sample consisted of all samples from an elementary school in San Antonio, Texas. Each grade sample contained 32 children. 3. 2.2. Coding and tabulation of variants.The Free Speech sample for each subject was transcribed into English orthography by typists.Phonetic coders then listened to the Goldman-Fristoe test and to each subject's Free Speech sample.Variations from SAE articulation

15 16 were noted for eachsubject (on a "coding sheet") as either asubstitution, an omission, or adistortion.In the case of substitutions, the phonetic realization of the target phoneme wasnoted. In the Free Speech samplesthe number of occurrences of each target phoneme was recorded, aswell as the number of variants produced for each.This was necessary in order to obtain arelative frequency figure for variants occurring inFree Speech.If the realization of a target phoneme was a variant of SAE on theGoldman-Fristoe test, but the target did not occur in the Free Speechsample, a notation of N/O (no occurrence) was used.If the target phoneme occurred infree speech in Standard English form, the notation N/V(no variation) was coded in the free speech column of the coding sheet.Table 2 is a sample coding sheet.The subject reported on Table 2, for instance,produced a /b/ for /p/ substitution in initial position on the Goldman-Fristoetest, but there was subsequently no occurrence of an initial/p/in the Free Speech of Subject #153.Another variant in this subject's Goldrnan-Fristoesample was the substitution of /f/ for /0/ in medial position.All occurrences of medial /8/ in her Free Speech, however, wererealized as /8/, thus the N/V notation appears in the Free Speech columnfor medial /8/.

3. 3.Two Excluded Phonemes With two exceptions all the phonemesand clusters tested by the Goldman-Fristoe Test were evaluated inthe error analysis. The excluded phonemes were /wh/ and 10 I.The substitution of /w/ for/wh/ is without exception the variant associatedwith the /wh/ phoneme. This substitution is not correlated with articulatorydisorders, and is extremely common in English.Many linguists feel that the distinctionbetween /w/ and /wh/ is dying in English, since itcarries a very low functional load. That is, there are very few minimal pairsutilizing this distinction (e. g. IwailYwhale') and those usually involve words which aredifferent parts of speech.Therefore, the distinction almost neveraffects the meaning of a word embedded in an English sentence.It was decided that the inclusion of /w/ for /wh/ substitutions woulddistort the data and magnify the number of variants inappropriately. The /n/ for /9 /substitution was left out of consideration because its distributionalcharacteristics differ from any other phonemein the language. /0 / patterns more like acluster than like an individual phoneme, e.g. it usually occurs only inword-final position. When it is not in word-final position itis in stem-final position at a morphemebound- ary (e. g.'singPsinger', where there is obviously amorpheme boundary after the /b /in 'singer%) In 'finger', wherethere is no evidence for a morpheme boundary following the/ng/ cluster, the intervocalic /0/is disallowed by the rules of English.Of course,1,9 / frequently occurs in

0 4.) Table 2 Subject # 153 Sex Sample Coding Sheet Grade 4 City Niagara Falls Position G. F. Variant f Position occurrences of target# of ' variants # of F. S. Variants of each variantoccurrences # of f 0* 2 , f 11 7 0 7 f d t 1 1 f 4 3 d 4010012! 0t i e b 1 N/0 813 *An omission is indicated fm f by the null 1sign, '0'. N/V f 2 1 t 1 18 clusters before a /k/ or a/g/, usually as a positional variant of /n/ (e.g. 'congressTcongressionalt, where stressis one of the determining characteristics of the assimilation).The distribution of /i9 /is similar to the distribution of theclusters /mb/ and /nd/ in that none of these three may occur in word-initialposition.Thus, to have treated In Ilike any other phoneme,especially like any other nasal phoneme,woula have been highly misleading at best.Therefore, if it had been analyzed, it should have been treated as a cluster.However, the only clusters treated were those occurringin word-initial position.Hence,1.9 / could not have been treated on that basis.Adding to the difficulty with /D /is the fact that the vast majority of /0/ errors occur as an /n/ for /0 /substitution associated with the 'ing' morphemein progressive and gerundive con- structions.The /9 / occurring in the 'ing' morphemeand the/./discussed above (whicli behaves as astem-final cluster) are phonologically very different.To have collapsed the two types of /.0 / as onephoneme for analysis would have simply added to theconfusion. SECTION 4 RESULTS

In this section the quantitative results of the present study will be reported; the discussion of the linguistic and sociolinguistic implications of the data will be reserved for Section 5.These quantitative results consist of tabulations -- for each subject of the articulatory variants which were produced under two speech conditions, responses to the Goldman.. Fristoe Test of Articulation and brief segments of connected discourse (Free Speech).The results are reported in this section in the following order: (1) First reported are the frequency and type of variants for each grade and for each target phoneme. The type of variant (omission, dis- tortion or substitution) is noted, and, in the case of substitutions, the substituted phonemes are noted. (2) In order to investigate trends across grades, several statistical tests are reported and the trends graphically represented. These trends are interpreted relative to a primary goal of discovering which variants from SAE are attributable to dialect differences and which are attributable to developmental factors.It is expected that dialect variants will remain at a fairly stable level across grades, while those variants which disappear at the higher grade levels will be attributable to developmental fa..irs.All reports of trends across grades and all statistical tests were carried out on Goldman-Fristoe data only, because the Goldman- Fristoe test provides a speech corpus which is identical for each subject. It is well known that free speech samples typically suffer from the problem of differential frequency of occurrence of various phonemes within and between subjects.Thus all statistical analyses were carried out on Goldman- Fristoe data.The role of the Free Speech data is discussed below. (3) The stability of variants across the two speech modes, Goldman- Fristoe test and Free Speech, were assessed by a direct comparison of the subjects' performance across the two samples. Only the most frequent variants are subjected to this comparison. (4) Adding across grades, the most frequent variants found in the Goldman-Fristoc data are presented separately, together with the linguistic feature data for each.This presentation, together with the data showing trends across grades, provides the basis for a discussion (in Section 5) of the linguistic and sociolinguistic significance of the present results.

19 20 All the above presentations are given separately for each dirlect group, first for the Niagara Falls group, then for the San Antonio. No comparisons of the two groups are made in this section, but Section 5 incorporates adis- cussion of the implications of similarities anddifferences among the two groups.

4. 1.Niagara Falls: The Black Sample

4. 1. 1.Raw data presentation.Tables 3 and 4 report the raw data for the Niagara Falls children.Specifically Table 3 presents all phonetic variation recorded on the Goldman-Fristoe Tert ofArticulation, while Table 4 reports similar data from the FreeSpeech samples of the children. The rows of both these tables are target phonemes (andblends); the columns are grades (1-6).Each cell reports the number of variants of the row phoneme recorded for the row grade, the number of children who produced a variant, andthe relative frequency of error for that target phoneme. The relativefrequency figure reflects the frequency of variation of a target phoneme visa1vis the frequency of occurrence of that target phoneme among the subjects producing variations. 1 Each cell of Tables 3 and 4 also contains an enumerationof the types of variations observed and the frequency of occurrenceof each.In the case of substitutions, thesubstituted phoneme is noted; distortions are represented as 'dist% and omissions are indicated by the null sign, 0.

10nboth tables the number of occurrences of the target phonemes are indicated.The relative frequency of error is computed by dividing the number of variants by the number of occurrences of the target.In the case of the Free Speech table, the number of occurrences was determinedby counting the tokens of each phoneme on which an error was made.In the case of the Goldrnan-Fristoe test, occurrencesfor each subject was constant, since the language sample was the same for all subjects.Thus, the occurrence value for each target wasobtained by multiplying the number of occurrences of that target in theGoldman-Fristoe by the number of children who produced a variant of that phoneme on the Goldman-Fristoe test.The number of occurrences of each target phoneme on the Goldman-Fristoeis noted in parentheses to the left of the position indicator for targetsymbol on Table 3.Appendix B lists the words in which each target appears. To illustrate this procedure, note for instance /CV in initial position in first grade on Table 3.There are two occurrences of initial g/ on the Goldman-Fristoe test (in chicken and church -- cf.Appendix B).Five subjects produced variants of initial /Z/ in the first grade onNiagara Falls, therefore ten occurrences of initial /Z/ are represented in this cell.There are five variants noted in thecell.The division of the number of variants (five) by the number of occurrences of the target(ten) produces a relative frequency of.50.The analagous cell of Table 4 (initial /14/, first grade, NiagaraFalls) reveals that two subjects produced two occurrences of the target in Free Speech and produced variants on both occasions.Therefore, the relative frequency noted in that cell in 1.0. Variants Produced by Speakers of BE on theTable Goldman-Fristoe 3 Test of Articulation* ,J tu E-4 1bo 1 2 3 Grade 4 5 6 Variants Total n9 f12 .33 rf n10 f15 . 38rf n5 f6 . 30rf n1 f1 . 25rf 4n 4f . 25rf 3n. 4f . 33rf 32n 42f . 33rf (4)m nvariants f f0 12 variantsn f f0 15 nvariants f 0 6 n.variants f 0 1 variantsn f 0 4 nvariants f 0 4 nvariants 0 42f (5)f variants10 f 15 . 30rf15 variants16 f 27 .34 27rf variants17 f f35 .41 35rf variants16 f 34f .40 34rf variants22 f f50 .45 50rf variants22 f 48f .44 48rf 103 variants 209 f 209.41 frf 0 01111111101M MOM SIMS IIIIIM 0 UMW 0 IMMO 1111110 0 0 41111. n 0f rf n1 f rf n Wag0f rf n f rf NOMn f rf 111n OMNI f0 rf n1 0 111f1 rf variants f t1.33 1 0 0 0 variants t . f133 Table 3 (contld) 1 2 3 Grade 4 5 6 Variants Total 06.n6 6f IIIMMINOrf1 n14 f14 rf1 n11 f11 rf1 n11 f11 rf1 n 19 f19 rf1 n17 f17 rf1 n78 f78 frf1 b (1)f variants f 0Pg 231 variants f P0 7 variants f P0 65 variants f p0 . 47 variants f p0 17 2 variants f p0 15 2 variants 0Pg 2651 _ 1 n f rf n f1 rf n f rf OM.n IMOf rf MOND n f rf COMn1 f1 rf n f rf variants5 f 5 . 350 variants1 f . 501 variants4 f 5 . 633 0 variants3 f 3 . 350 variants f . 501 variants14 15 .10 f54 0tI 1 t s ti' 1 t i t0ss 31 Table 3 (cont'd) Grade 4110111111111.1011111/1M 1 2 3 4 5 6 Variants Total 4 4 1 1 1 1 2 2 1 0 2 2 1 0 9 9 1 (1)m ...... ---variants f 0 21...... variants f jV 1MVO variants f t 1 variants f 2 variants 4f tz 1 s, t 0tiz 121 614/00 j 4n 4f 1rf .....n1 f1 OVUM. 1rf NOMn3 f3 rf1 ON/M1n OSIND0f ONINIMINOrf OVID4n 4f 41111.11110rf1 4111110n =MOD 0f rf n12 f12 rf1 (1)f variants s; 1 f variants f 0 1 variants f t9 21 variants f t 21 variants t f35 0t 1 dist 0 1 dist s0.,9 11 Table 3 (cont'd) at, 1 2 3 Grade 4 5 6 Variants Total 1 f1 rf1 n1 f1 rf1 n 0f rf n 0f rf n 0f rf n INM.f0 =MUMSrf MOND2n 2 rf1 d (1)m variants f O 1 variants f 0 1 variants 0 2 4..3I: (1)f variantsn11 f f11 rf1 nvariants17 f f17 rf1 variantsn12 f f12 rf1 nvariants14 f f14 rf1 nvariants19 f f19 rf1 variants22n f 22f rf.-1 variantsn95 f95 frf1. 0 tO 65 0ts 14 21 0t 57 0t 13 1 0t? 15 31 0t ? 20 1 0st? 7515 41 (3)i 2 2f . rf33 n 0f rf n 0f rf n 0f rf n 0f rf n f0 rf variantsn2 2f f.rf 33 variants f 1 . t 1 44 Table 3 (cont'd) E-4 mttr.... 1 2 3 Grade 4 5 6 Variants Total 1 f 1 1100101.0rf1 2 2f rf1 OM. 2 f2 rf1 MOW 3 =NM 3 0.11INIINDrf1 WOO 1 41111 1 1rf WIMP OlISM 0 rf CMS.9 AMMO 9 rf1 variants f 0 1 variants f $ 1 variants f 0 2 variants f 0s 21 variants f $ 1 variants 0s 11111101 63 011110 NNW OMB 0 IMMO MOM *IMO r.'` C% ) n1 f1 rf1 n SIM f rf OM.n f 111100rf n6~ f rf n1 f1 rf1 n f rf n2 f rf1 (1)1 variants f d 1 0 0 0 variants f d 1 0 variants d 2 f2 (1)m MOMvariants2 f 2 2rf1 0 rf 411 MOO 0 rf NOM MOW rf OMNI 0 0 variants2 2 2f1 t71 .1,Worl... 7.. Cr% 0111.011.4110011111.01=111~01/ Table 3 (cont'd) E-4 1 2 3 Grade 4 5 6 Variants Total prowwwww.101111 7n f7 rf1 n10 f10 1rf n7 f7 rf1 n5 f5 rf1 n13 f13 rf1 n10 f10 rf1 n52 52f rf1 2 (1)f 411~1011111111110variants f k0 25NNW variants f k0 46 variants f k0 61 variants f k0 41 011111~~IIIMMIlaill~Mvariants f 0k 9 823IMO variantsNOINNIMINIVIINI V MD f k0 9100 MN variants k0? 1535 2f 1 f1 rf1 2n f2 rf1 0 rf n f0 rf n1 f1 rf1 n f0 rf n4 f4 rf1 (1)1 011101111111variants f s 1 variants f d 2 variants f d 1 MOSIONNI11111010./MOMMvariants d %WM f31 n2 f2 rf1 n1 f 1 rf1 NMIn1 SIMf1 rf1 SINOPn f0 rf MONn1 VIM*f1 rf1 INNWPn f0 rf n5 f5 rf1 (1)m variants f Y 1 variants f g 1 variants f E 1 variants f d 1 variants Y f11 0 1 Cg0d 11 ea* (6.4 wint. . .11 y ..a Table 3 (cont'd) E"4 1 2 3 Grade 4 5 6 Variants Total (1)f variantsn10 f f10 rf1 variantsNNW 8n f . 8f rf1 variantsn10 f f10 rf1 NINO n MO.f0 rf n10 f10 rf1 WOWn6 6f rf1 variants44n 44f frf1 0Z 37 0Z 71 0Z$ 27I variants f dZ 9l variants f c. 6 d0sc. 36 I16 f rf n1 f rf n f rf n f rf n f 111111111rf WINOn WMf rf SONOn f rf k (3)i variants3 f Pt5.56 4I variants f t1.33 I 0 0 0 0 variants4 Pt 6 . f550I Table 3 (cont'd) 00 E-4 as 1 2 3 Grade 4 5 6 Variants Total rf rf rf rf rf IMMOn INf rf n f rf (2)m nfvariants3 f 3 . 150 nfvariants1 f 1 . 501 nfvariants7 f 7 . 50 variantsnf3 f 3 . 50 nfvariants1 f 1 . 501 0 variants15 15 . f50 Oh$ 1 0 0s 21 0i 21 s 0ts 4611 tgr 1 hgtr 1 (1)f variants2 f 2 1 variants1 f 1 1 0 variants3 f 3 1 variants1 f 1 1 0 variants7 7 f1 Ot 11 A 1 0? 21 ? 1 pt0? 231 44 Table 3 (cont'd) H kRSbao 1 2 3 Grade 4 5 6 Variants Total 2n 2f rfI OM.n OMNI f0 0 0 2 2 1 0 4 4 1 (1)i variants f wY 1 variants f w 2 variants Yw ..... f31 i (2)rn 0100variantsn6 f 6.50f rf *NMvariantsn1 f f1.50 rf variantsn3 f f3.50 rf IMOvariantsn1 f f1.50 rf WOWn MOW 0f rf ein MS =MDf0 rf variantsn11 f11 . frf50 n fd0 rf5I n 0f rf1 n fr0 2rf1 0 rf1 rf rf n d0r f 9rf1 (3)f variants11 f 024 . 2473 variants13 f 023 . 2359 variants12 f 019 . 5319 variantsnf4 f w09 . 2775 variantsnf18 f 041 . 4176 nfvariants11 f 026 .79 26 variants69 w0 142 .140 f69 2 Table 3 (cont'd) 01.0 1 2 3 Grade 4 5 6 Variants Total 1 1 1 0 0 0 0 0 1 1 1 (1)1 variants f r 1 variants f1 f : n1 f1 . 50rf n f0 rf 2n 2f . 50rf MINN 0 rf n I0 rf n f0 rf 3n r f3 . 50rf rn (2)m variants f 0 1 variants f 0 9 1 UI variants 0 2f1 WM* 5n f rf dbl.n5 f rf ONM n3 f rf OF.n1 f rf n1 f rf 411n 1 f rf n 16 f19 . 40rf (3)1 variants f 8.53P0 71 variants f p05.33 NM= 41 variants f 03.33 3 variants f n1.33 1 variants f 01.33 1 variants f 01.33 1 variants nP0 16 f21 Table 3 (conttd) ...0111.11111111NOMMINIMINIMI 1 2 3 Grade 4 5 6 Variants Total 01 Mow emslimerf rf MION 1 WNW f1 fimirf1 NOM f rf OM. awef 411001111rf OMB f rf el/irm 1 OW. 1 rf1 0 variants f 1 e0 0 0 variants A f 1 / OMNI IMO OW. C . 8n f11 . 28rf n6 f rf n6 f rf n f0 rf 4n f rf 4n 4f . 20rf n28 f32 . 23rf 4..."- (5)m variants f 91 variants f 7.23 7 variants f A6.20 51 variants f 4.20 31 variants f 31 variants 27 f n k0? 1 0 0 t0 t0 kt01? 21 4.4 Table 3 (cont'd) E-f 1 2 3 Grade 4 5 6 Variants Total n11 f25 . 45rf n17 f29 . 34rf n14 f20 . 29rf n2 4f . 40rf n13 f25 . 38rf n12 f17 . 28rf n69 120 f . f34rf t( : i n (5)f variants f 0? 24 1 variants f rn0 28 1 variants f 0 9 18 2 variants f 0 4 variants f 0? 17 8 variants f 0? 14 3 variants In0? 105 14 1 n5 f5 rf1 7n f7 rf1 111111/0 7n ewef7 ammorf1 n10 f10 rf1 OM=8n IMO8f rf1 M4n 4f rf1 41n 41f rf1 (1)1 variants f b 4 variants f b 7 variants f b 7 variants f b 10 variants f b 8 variants f b 4 variants b 40 f1 0 1 OM" 0 (3)rn 1 f1 . 33rf 1 f1 . 33rf 1 f1 .33 rf n1 f1 0 33rf n1 f1 . 33rf n 0f rf variantsn5 5f .33 frf variants f f 1 variants f 0 1 variants f 9 1 variants f 0 1 variants f b 1 0 21 Table 3 (cont'd) Cdto 1 2 3 Grade 4 5 6 Variants Total n5 f6.60 rf 7n f7.50 rf MOM 7n f rf 1.00 4n 4f . 50rf IMONIIn3 f rf 2n f rf n28 29f . 52rf P (2)f variants f 0 6 variants f 0? 61 variants f f07.50? 51 variants f f0 31 vari amid f 03.50 3 variants f 02.50 2 variants f0? 25 2f2 C7)C .- 4 f5 . 63rf 5 9f . 90rf 4 4f . 50rf ONO aNNHf0 INWIDNOrf 01IND 6 6f . 50rf ONINV4 f5 . 63rf 23n f29 . 63rf (2)1 variants f 41 variants f w 9 variants f w 4 variants f w 6 variants f w 5 variants w 28 f1 (2)m .variants8n f 9f . rf56 variants9n f f10 . rf55 .variantsn7 f 7f . 50rf OS.n OM*0f rf variants41.08n f 9f . 56rf ORMvariants4n f 4f . 50rf variantsn36 f39 . f54rf 0w? I35 w 10 b0w 51 0w 27 w 4 b0w? 31 161 Table 3 (cont'd) bit4.r.: 1 2 AMMINIIIII.II 3 Grade 4 5 6 Variants Total OM.n 41011.f0 rf 0.*n2 f rf 410.1111n OMf0 rf 4111M10n MOMf0 SONININOrf *IOWn1 f rf nMOW f.10.0 41111111111110rf .....n 3 _f3 . rf50 (2)i variants f dist 2.50i 1 variants f 91.50 1 0.01111111111111111011..variants dist 9.s f1 -( 3 _n1 f rf n f0 ...... rf ....n1 f rf 111110 n 01f0 OIIIIIIIIIMrf n1 f rf AMIN.n f0 rf n3 f3 . 50rf s (2)rn variants f O1.50 1 variants f dist 1.50 1 variants f 01.50 1 variants 0 f21 n4 f5 . 63rf n4 f4 . 50rf n9 f9 . 50rf 4n f5 . rf63 n2 f2 . 50rf n1 f1 rf n24 dist f26 . rf54 (2)f variants f O 5 variants f 0 4 variants f 90 81 variants f 0 5 variants f 0 2 variants f 0 .50 1 variants 09 25 f1 Table 3 (cont'd) to.z) 1 2 3 Grade 4 5 6 Variants Total 2n f2 rf1 n3 f3 rf1 2n 2f rf1 n 1 f1 rf1 n2 f2 rf1 n1 f1 rf1 n11 f11 rf1 (1)1 variants f Cs 1 variants f dist E 21 variants f dist s 1 variants f $ 1 variants f sCs 1 variants f C 1 variants Zs 425f f rf1 n f rf n f rf1 n f rf n f rf1 n f rf n dist f rf1 variants3 3 1 f 0 variants2 f 2 2 0 variants3 f E3 2 0 variants8 8 3f 0Zs 1 s dist 1 dist 0sa 131 4.) Table 3 (cont'd) E'4 ostos44) 1 2 3 Grade 4 5 6 Variants Total n3 3f rf1 f um*n Mill0f rf n2 f2 rf1 100106n f0 011111/21111110rf .11n3 UMWf3 rf1 ...... n .....f0 41101.1111Mrf aonmenvariants8 _8f frf1 g (1)1 variants ZO$ 11 variants f dist $ 1 variants f dist Z 21 dist sZ 2231 41111.111MO 0 1 f1 rf MOM IMMO rf 0/11111 111111 rf 3 f3 rf IRMO MUM 0 rf MOM =MD 0 . 50 (2)m variants f O . 501 0 0 variants f . 503 variants4 4 4f 4 4f . rf50 5 f5 . rf50 n4 4f . 50rf n 1 f01 . rf50 4n f5 . 63rf 4n 14 . 50rf n22 0 f23 . rf52 (2)1 variants f O 31 variants f 0? 32 variants f 0? 22 variants f 0 1 variants 0 9 41 f variants 0 9 31 f variants 0? 16 f 7 Table 3 (cont'd) 1 2 3 Grade 4 5 6 Variants Total (1)i 9n 9f rf1 9n 9f 1rf a9n 9f rf1 5n 6f rf1 n2 2f rf1 2n 2f rf1 37n 37f 1rf 0 variants f fdt 243 variants f dtf 351 variants f tf 36 variant:. df 15 f variants f t 2 variants f t 2 variants dtf 1418 f51 -r" C-) (1)m variantsn12 f f12 1rf variantsn18 f f18 01rf 28n 28f rf1 27n 27f 1rf 22n 22f 1rf n19 f19 rf1 126 n 126 f rf1 0f 11 1 f0$ 2691 variants f 0f ? 21 16 variants f 50 21 15 variants f 0f 20 2 variants f 0t 16 21 variants 0f 2198 3f t f f itst? 21 Table 3 (contld) CbIA) 4.) eu E-4 s4asbO 1 2 3 Grade 4 5 6 Variants Total n17 f17 rf1 n18 f18 rf1 25n 25f rf1 n22 f22 rf1 n18 f18 rf1 n10 f10 rf1 110 n 110 f rf1 8 (1)1 variants 0f 15 f 2 variants 01 16 f 2 variants 0f 20 f 5 variants 0tf 19 f 2.i variants ft 17f 1 variants tf 91 f variants t0f 96f11 3 n15 f15 rf1 n14 f14 rf1 20n f20 rf1 n17 f17 rf1 n23 23f rf1 n18 f18 rf1 107 n 107 f rf1 (1)1 variants f d 15 variants f d 14 variants f d 20 variants f d 17 variants f d 23 variants f 0d 17 1 variants 0d 106 f 1 t (1)m n3 f3 rf1 n2 f2 rf1 variantsn2 f f2 rf1 n f0 rf .n f0 rf n f0 rf variants7n f7 frf1 variants f dA 2 variants f d 2 dz 11 d/z 51 Table 3 (cont'd) 4.4 Grade Variants Total OMB8n 8f 1rf OM*9n 9f rf1 n18 f18 rf1 n16 f16 rf1 01.1Mn7 7f ONVANSINDrf1 5n 5f ..1111MEMrf1 n63 63f rf1 (1)1 variants b 71 f variants b 9 f variants b 15 f 1 variants b 16 f variants b11 6 f variants b 5 f variants b 58 f w mfw 1 1 m1fw 21 8n 8f 1rf n2 2f rf1 n1 f1 rf1 2n 2f rf1 n f rf1 n f rf n f rf1 (1)m variants f variants f variants WOW f variants f variants2 2 f 0 15 15 f b 8 wb 1 b 1 0b 11 b 2 variants 0wb 13 1 Table 3 (cont'd) E-4 1 2 3 Grade 4 5 6 Variants Total n1 af 1 rf1 n5 f5 rf1 4n 4f rf1 an6 f6 rf1 nvariants8 f 8f rf1 avariants4n f 4f rf1 variants28n 28f frf1 v (1)f variants f b 1 variants f wb0 31 variants f 0f 31 variants f b0 51 f0 44 0f 2 bf0 17 371 MOM IMO 01.1100rf n 01f 0111111 41~n ONOf 01.11111111111rf WOO MOD rf ,rf 111111M. MIMI 1 OM*f 1 rf1 =IOW n1 MM. I1 rf1 w (1)i 0 0 rf 0 0 0 variants f 1 variants f 1 n - f1 rf1 n I0 rf n f0 rf n af0 rf n I0 rf n f00 rf 1 0 1 rf1 y (1)1 variants f z 1 variants z 4.00 1 Table 3 (cont'd) 1 2 3 Grade 4 5 6 Variants Total 3n MMOf3 IMMON*0rf1 IINM 3n f3 1rf ENNIOn3 PMf3 rf1 .....n1 f 1 rf1 .....5n 5f rf1 .....4n ....4f rf1 n19 f19 rf1 .. (1)1 variants f 1s 21 variants f s 3 variants f 3v$ 21...- variants f s 1WIMP variants f ds 41 variants f s 13 variants §s 15 f1 tkiN, (1)rn variants4111111/11111.11017111.2 f 2 rf1 n 0I rf n 0I rf n f0 rf n1 I1 rf1 1 1 rf1 4n 4f rf1 dist $ 1 ...... variants f d 1 variants f dist 1 variants dist sd 2f1 Table 3 (cont'd) 1 2 3 Grade 5 6 Variants Total n5 f10 . 50rf ....n6 f rf n7 f rf 2n f4 rf .....4n f rf n6 f rf n38 f39 . 25rf z (4)f variants $ 37 f variants f 07.29 25 variants 07.25$ 51 f variants f 02.25 2 variants f 0$5.31 23 variants f 8.33s 26 va r iants s 15 f 0 s /11.111.0.11.10IMINIMOIMI dist 1 0 dist 0 ...I23 n o ....f ONNO n1 ONIMf 1 n 0 ....f n 0 NOMNOf n 0 MOROf n1 ONWO fi OM*/ 2n .0f2 , b variants f g I variants f P 1 variants f 13g 1 br r variants f n2 2f variants f n2 f2 variants f n1 f1 n 0 f variants f 2n 2f variants f n2 2f variants f n9 9f Table 3 (cont'd) E-I cd 1 2 3 Grade 5 6 Total n f n f n 4 f 1 Variants 1 bt A variants 1 f 0. 0 0 0 0 variants f 1 0 NOM OMNI 1111M IMMO 411=1/0 n 0 f UFOn1 MN=f1 n1 f1 n 0 f 2n 2f n1 f1 IMODn05 WPM5f C,'") d variants f variants 01.1110 f s: g 1 g 1 variants f 0g 1 variants f J 1 variants 95fjg 13 dr 4n 4f ...... n2 VINO 2f MOM n1 MOMf1 n 0 MM.. f MMAn3 f3 MOM4n OMNI 4f 14 n 14 f r variants Yw0 21f variants w 2f variants w 1f variants f Yw 21 variants f 0w 31 variants f Yw0 10 2 44 Table 3 (cont'd) E-1 ktdtoa 1 2 3 Grade 4 5 6 Variants Total f variants 3 3 f n 0 f 4MM 0 01111a MEMO 0 IMMO Vella 0 IMMO n 0 f variants MIRO 3 IIMND 3 f : fft vPb 1 Pbv 11 ..,1 OEM 1 INO 1 2 2 IPMN 2 2 0 0 0 5 5 A variants f 0 1 variants f 0 1 variants f 0 1 variants f0 23 n1 1 n f n f n f r r1 1 gr r variants f 1 0 0 0 0 0 variants f w 1 fr V Table 3 (cont'd) 1 2 3 Grade 4 5 6 Variants Total MOM 1 1 OMNI 1 MOM 1 1 1 n 0 *NM f n 0 f n 0 f 4111* 3 =OM 3 CIO k variants f variants f 1 variants f t 1 variants f 3 kr 5n f5 IWOn6 f6 3n f3 n1 f 1 2n f2 n2 2f 19 n 19 f r variants f 0 41 variants f w0 24 variants f w0 01,10 21 variants f 0 IF111811 1 variants f w 11 variants f w 1 variants f 0 13 6 3 3 n 0 OMM f n 0 OMMf n 0 f n0 0 f n0 0 f .NES 3 3 variants f 0 2 I variants f 21 Table 3 (cont'd) 1 2 3 Grade 4 5 6 Variants Total faNO n3 f3 n3 f3 n 0 f IMMn1 f1 nWNW. 1 f 1 OMIMn1 OW.f 1 n9 f9 kl / variants f 3 variants f 3 variants f I variants f w 1 variants f MOM 1 variants f 81 0n f 0n f n f n0 f n f 0n f wn01 f1 cr)1-1:' P 0 0 variants f 1 1 1 0 0 0 variants f _ 1 PA n f n f nf 1 f1 n1 f1 n1 f1 n f _fn f 1 0 0 variants f w , 1 variants f vir 1 variants f w 1 0 variants fw3 3 3 Table 3 (cont'd) 411011101 1 2 3 Grade 4 5 6 Variants Total 1 1 n 0 011111111.f n 0 f n 0 f n 0 MINASf n 0 f 1 1 Cr. variants 011111. 1 f variants f 1 IMO 20 4110111 2 6111160 1 IMMO 1 OW. 1 1 n 0 f n 0 f n 0 f 04 4 skw k variants ft 1f variants t 1f variants f 0 1 variants f0 21 ....4n ....4f 7n 7f 8n 8f n1 f 1 2n 2f n1 f1 23 n 23 f w variants 0 4 f variants 0 7 f variants f 0 8 variants f 0 1 variants f 0 2 variants f 0 1 variants f 0 23 Table 3 (cont'd) 1 2 3 Grade 4 5 6 Variants Total 4.0 1 1111.110 1 n 0 f n 0 f GIONIP 1 0111111 1 411111111n 0 f n 0 f 2 2 s variants f t 1 variants f i 1 variants ft.s 11 VI1,---% 2 2 1 1 0 0 2 2 0 5 5 / variants f v0 11 variants f 0 1 variants f w 2 variants f0wv 221 Table 3 (cont'd) H4.4 idtoa14 1 2 3 Grade 4 5 6 Variants Total 1 1 3 3 4MM 1 1 OHM 0 11111110 2 2 WINO 0 0 7 Totaltr r variants f 342 1 variants f 334 3 variants w339 OW. 1f 218 variants f w357 2 261 = variants f 1851 ) SubstitutionsOmisDistortions s ions 169172 1 170162 2 181154 4 122 96 0 207147 3 143117 1 = 987853 11 *Fortargeteach.indicatorofgrade; each (see Omissions f targetisFootnote (1, the in, onnumber the 1, thisGoldman.Fristoe of section,variants forproduced;each cell entry, n is the number of subjects or f). are indicated by the null sign, 0. Each cell entry the observed variants test is indicateda detailed by discussion the numeral of theto the rf figure.)left of that The target's number ofrf indicates the relative frequency ofwho produced a variant for the are listed, with the number of occurrences of row phoneme in the columnthe total variants for each occurrences position Ui Variants Occurring in the Free Speech of Speakers of Black English Table 4 to 1 2 3 Grade 4 5 6 Variants Total OMn7 Maw8f 13 . 62 *MO 15 33 awe95 .35 6 10 39 .26 5 19 40 .48 9 32 50 .64 9 24 40 .60 moo 51 126 277 .45 rf rn variants orf nf orf nf orfNOMf nf orf nf orf nf orf variants0.101101M1101nf o MOOf variants f OM1114111101111M.011.410.0variants f variants .NPf variantsONNIMIIMIIIIIIMON .11. f va riants elm f O 8 0 33 0s 91 0 19 0 32 0 24 s0 125 1 't n f .0. OTINIMO WAS IIINNIMO NM 011. 01110 ONO WNW woo .1 mow rf f variants14 O76 149 .51 orf nf orf nf orff 76nf orf nf orf nf variants25 186 322 .58 0 186 f variants17 132 188 .70 0 132 f variants23 191 296 .65 0 191 f variants26 142 243 .58 0 142 f variants26 117 214 .55 0 orf nf o 117 f 131 844 variants1412 .60 0 844 f n f .... IMMO OM 01111111111 MeV 41,11.01* OM ONIII WM AMMO SOO INIO rf I 1 13 .08 1 2 12 .17 ...... 1 1 1 1 ..:. 0 0 0 orf nf o variants3 4 26 . 15f i variants orf nf orf nf orff nf orf nf orf nf variants f variants f 1 b w 1 h0 1 P 1 h0P 1 0111110411111111 1 Table 4 (cont'd) 11 3 Grade 4 5 6 Variants Total 01 o NM* f o n f o rf n f MOo SWIM.rf n f 1110o CONN.rf ONOn NNWf o ...Wmrf ON.n f MM.o rf n 1 f 1 1 rf1 ii 0 rf 2 2 4 . 50 0 0 0 3 3 5 . 60awe in variants 0 f1 variants OW* f variants 0 f 1 01110 IMMO OM 11101,0 01M o n f 3 NOMo3 rf1 6 f7 o emar.rf n3 f8 o rf f 0 o rf n 1 f 1 o 1 rf1 n f 0 o .12n f19 23 . 83rf a f variants2 variants 9 .78 variants 10 .80 f5 ENIIMvariants p f 1 variants 0 I 10 21 4 2 "", 1 0 p09 21 0p 1 z n o . n f o rf YOE IMO On. wowo emir=rf n f o rf1 . f rf rf 0 2 2 2 c i variants2 2 2 1 orf nf orf variants 111 woommo tfi Table 4 (cont'd) .0 E4-c.1r:tss. I 2 3 Grade 4 5 6 Variants Total MOD MIND rf n 11Yf o 411011rf OM* rf 11 o rf f o rf 110 o rf n allf o 01rf xn variants2 4 4 1 0 0 0 0 variants2 4 4 41 ; 4 On* o 111.1 en*n f rf 0 o rf nfvariants2 orf nf orf nf2 6.34 f variants1 1 1 f1 0 orf nf orf 0 f 0 rf variants3 3 7 .43 0 1 1 21 n f o rf o rf MOOn 1 f3 o rf rf .1D 0 .n0 o rf ON. 0 11 1 3 rf 0 variants 5.60 3 nf orf nf o variants 0 5.60 3 Table 4 (cont'd) fri7:4)rCr 1 2 Grade 4 5 6 Variants Total n2 f 3 o8 rf n6 f 10 o rf n3 f 3 o rf IMOn3 f5 o11 .45 rf WOOn12 f 15 o29 .52 ...MOrf 5 6 12 . 50 31 42 109 .OIONIIIIrf 38 m variants .38 f variants 37 .27 f variants 12 .25 f variants MEMf variants f nfvariants orf nf o .M0f variants FMf O1 21 t09 81 09 21 M111 0t9 21 09t 13 1i 0t 24 0t9 30 75 f nfvariants12 orf nf orf 36 69 .S2 f variants23 89 144 .62 f nfvariants20 orf nf orf nf70 132 .53 f variants18 73 116 .63 f variants20 53 orf82 .65 nf orf nf o 1f variants21 48 67 .72 .114 369 610 .60 f variants rf O9t 30 5I 0 9t 2161 7 09t 223612 09t 213913 09t 202310 0 1t 20 6 09t 188130 51 II1 f3 o8 . 38rf f o WWI!rf ORMn f o rf 4111.110n f 01100o 11rf en.n f o rf allle oMM. rf n 1 f3 8o . rf38 variants VOW I variants 3 Table 4 (cont'd) 3 Grade 4 6 Variants Total f rf ONM o ftMIIMrf 4/MDo WIIMNOrf n NOf 011M o AMMUNDrf n .4Mf o rf 0111.1. VIM OMBCe rf o . rf OM 0 000o NNWM W. a 3 3 7 . 43 3 3 7 43 1 dr: f f variants variants CP 21 0 21 ON. rf n f o rf elm Oweo =MOM ,10.0n WOONNf ORMo NOINIVOIMrf n f o rf n f ODo rf MOMn OA.f o 001rf n 1 f 1 o 1 1 0 0 0 variants1 1 2.50 0 variants2 2 3.67 m 101.M.varlant, 1 0 1 11101.011. f IMMO f rf n f o rf 4111M f rf VINO f oWNW rf *MOn ex.. o rf n o n f GSMo 0011111..rf nOM. 1 o 4111ri MN,n o n 3 i 3 1 4 5 6 . 83 27 32 41 . 78 2 variants 2 2 f2 8variants 8 15 . 53 f variants7 9 10 . 90 f variants 4 4 ONO f IIIIIIMOISIOIROvariants 4 4 f variants MOM f variants 0110f k ..... 2 lc0 _53 lc0 45 0lc 22 0k 31 0k 4 1 0lc 2111 55

co 4 I pm P4 41 44 N 04 tit 04

04 1 Pos to 1 04 bu 43 PAti 11 11

al OS

lI 1: I 04 1.45 N124

05 VI IA 4641

441

04 r4 04 445 N 04 441 o.

(.3 co 44 I Ico alm g

j k 00 4 I st,awl

130

r4 Table 4 (cont'd) 2 3 Grade 4 5 6 Variants Total n 1 f 1 0I rf1 n 1 WEIfI o4 25Ta 010 n 1 000fI o10 . 10 n 1 f1 o6 . rf17 n f 0 o rf WID4n 4I o21 . rf19 variants 1 variants 1 variants O f1 variants 1 variants .2 1 0 / 0/t 1 OM. 11,0 INS o f n4 f4 o13 . 31 frf va5 riants 5 o27 . rff19 variants3 f3 o17 . rff18 variants1 1 o21 . rff05 variants3 f3 o a . 25 rff 4variants f6 o18 . 33 rff variants20 f22 108 . 20 rff variants 0 4 0gt 131 0t 2 1 0 1 0o 21 0 6 0gt1 17 221 NSW TI aVON rf rf n1 f 1 o rf 0 rf I fI o2 . 50 0 0 variants 2.50 1 0 variants 1 al.40& wt.. Table 4 (conted) 1 2 3 4 5 6 Variants Total n f /IOW MO OM. 44140 MVO OW/ nmaw OIWO o rf in variants6 7 orf27 .26 nf orf nf f variants8 15 50 .30 f variants5 6 orf22 .27 nf orf nf orff variants3 6 33 .18 . variants2 5 10 .50 f nfvariants7 orf 11 26 .42 11 f variants31 50 168 .30 f Owd 41I1 0 15 0 6 0d 24 0 5 0 Tw0d 45 I31 /OW9n mowf28 MM. 40/1114/ 17 WNW 46 WOO 41141414 41/14. 20 414141044f 41441,o 4/1411111,rf ON/4n f16 . o21 .76 rf /IOWn16 46 /Oaf 73 .63 rf n 11 21f o39 .54 rf n77 201 362 . 56 f o rf 97 .45 41444114444WIRWROWN4441W 41 .68 91 .51 41MP 4.44. 4414144.4144040441114444114variants orf nf orf MO.f 4114011141141414111141414/14440variants WIN,f variants f variants MIN/f variants f variants f variants f f O 28 0 46 w0 43 1 0 16 0 46 0 21 0w 200 1 o /Wel.rf MOW 1 1 7 .14 4141414414 f 0 o rf DL o 0 rf n f 0 o rf n f 0 o rf nf1 o 1 7 .14 rf va r iants O 1 variants O 1 CT) e 8c

shi 1P4

le o'UQ

I tv

las

0 1 a Pit

12.

tot tm.

10 ID

stb 1".

11% 114%

0 -

0 1

1".

1"1 er. 12, PI N 1". . 1

0 1 a s

ti 1 19 Target Table 4 (coned) 1 2 3 Grade 4 5 6 Variants Total m variantsn4 7f orf29 . 24 nf orf nf o f variants6 9 SO . f18 va4100 riants5 5 32 .16 frf ONO WOOf 0 MOPo rf OOPvariants4 MOO 5 o20 . f25rf MOnitvariants2 orf nf o SNP2 12 .17 f variants21 28 143 .20 rff 0 7 0 9 0? 41 0? -.14 0 2 0? 26 2 f variantsnf8 o 29 125 . 23 frf variantsnf17 orf nf orf nfONO 51orf 353 nf orf nf orfWOO nf o . 00111110f14 OM,variants13 OMR 34 320 011.11IP . ONO.f11 MOOvariants3 5 89 .06 f variants9 46 149 .31 f variants7 23 138 .17 f variantsSr 188 1171 .16 rff 0 29 0? 45 6 0 ? 27 7 0? 4,. 4 1 0? 44 2 0 ? 20 3 0? 169 19 3 f3 o6 . 50rf 5 0011. 5 o10 . rf50 AMMO 4 4 o10 .40 rf 0 rf ONO 0 rf 1MOI 0 0 rf n12 f12 26 .46 o rf p i variants b 41Mf3 variants bO 4f 1 variants b 4f variants -. b0 11 f1 Ut 09

I", 11.4

UI 10 N

Cr so " pd

Pss N

-to

10 UI 1"% 1N%

14.

10 10 ash 1K+

10 to

0.1

:14 I

10 Us

I N

Us 13% sp) s-s% us r J 10 =8

S. 64% I Ce 4

sm. 0.4 0 -to

Target $1 4.+ Table 4 (cont'd) 6.4 03001.4 2 3 Grade 4 5 6 Variants Total n OM, f 0 IMMOo ONION.rf OM4n GIN4f 21o . 19 rf ON.2n 2f MOOo rf 00110 4111M 0 4111. 011611111111 NOW 1 MOW 1 ONO 1 1 ONO 41111PID 0 11/00 7 MOO 7 27o . 011=11026rf r Zn variants N.v ems* f va riants 5.40 f nf orf nf orf nf orf nf vaziants f variants I f 0 31 0 2 0 1 0yr 43 211 .... 1f o12 . 08rf 3n 5f 27o . 19 rf 2n 7f o11 64rf n f 0 o rf n1 ORMf1 111. rf va riants f1 variants f variants . f variants orf2.50 nf orf nf o f 0 variants7 14 52 . 27f i t tC 21 dist 1 3 9 1 0Z 2 s 0 0 1 dist 9ti 313 C.11 Table 4 (cont'd) f-+ c. 1 2 3 Grade 4 5 6 Variants Total 10.n _f o rf n f o rf n.... ONOf rf 0 rf 1 f 1 0 rf ea. 0 ewe rf sm.n7 9f o28 . rf32 m variants2 3 8.38 OMf variants4 5 14 . f36 variants 6.17 variants --...... dist I 21 0h? 31 1 dist 4211 n2 f3 o9 . rf33 n3 ONOf6 o15 . 40 rf ONO2 2 o3 . rf6? MOn2 4WDf3 4111110o13 . 23rf 1 f2 o4 . rf50 n f 0 o rf n 10 16 44 . 36 f variants 4110/1.I3 variants 6 variants f2 variants f3 variants 2 variants 0 16 f 0 0 0 0 411111P WNW 01110. n2 f2 o3 . ....rf6? n 1110f 0 o rf n2 3 5 . 60 0 f 0 o 0 0 rf 4nvariants f5 o8 . r162f i variants.01111111 C$ f 1 variants f variants dist a$ f 11 zs 22 ' , " " ' " " Table 4 (cont1d) - 1 3 Grade 4 5 6 Variants Total In MM. 0 rf 4 6f ON. o12 50rf M..n3 MOM 3f @AIM40 . 75rf n IMOD1 0 MM.o rf NM/n OMf 0 OS*o IA!rf n IND 1 0 6100o rf 4n 51 8o . 62rf variants 5I1 variants CP 3 variants C 45 MOND 0 rf n1 401. f1 o 1 1rf 2n f3 o3 *YENS1rf II 0 rf 0 11 rf 0 rf 3 am4 4 1rf variants I1 variants 6 I1 variants 1 0 dist 5 1 dist 0sz." 11 n1 f1 o13 . 08rf n f 0 MOD o rf n1 01f1 o1 rf1 @AIM n f 0 o Mrf 2n 2f o6 . 33rf 2.n AMPf3 o11 .27 M..rf n6 am.7f o31 . 23rf i variants 0 f1 variants 0 -1 1 variants 0 2f variants 0 f3 variants 0 ....f

. . . .4. 40,16aaht 1 oe' -4.4.4-41 - .1. L.: _I: L - Table 4 (cont'd) E-, 3.e.to 1 2 3 Grade 4 5 6 Variants Total as.n7 Mem .:10 Ofte 2 NM 2 7 11 00,0 5 OWN. 17 17 MAO 38 9 MOP 17 59 .29 47 95 320 .30 rf 81 .21 88 .43 OMIW in variants orf31 .32 i ?ifealeINOIMINIvariants orf nf orf 11 .18 f variants 50 .22 f nfvariants orf nf orf f variants AMV f nfvariants orf ni o f variants f 10 0 2 0 83 0 12 5 0 34 4 0 17 0 8312 n12 f o rf 21 14 90 262 . 34 6 52 139 .39 16 66 194 .34 15 54 119 .45 84 387 1264 .31 rf variants 31 220 .14 f nfvariants orf nf orf nf94 330 .29 f variants f OlINIMIIIIMIIIMMIIMINDvariants orf nf orf nf orfOftDf variants f variants Malb f variantsnf o r f 0? 29 2 0d? 6923 2 k0? 4247 1 09 3814 0 ? 5214 0? 4113 kd07 271113 21 n5 9f 4 MOO 4 10104 411, 6 9 . 67 1 OM 1 OM*4 . ~MOP25 2n 4010f2 2o rf1 n 1 4.0f 1 o 1 10rf1 +6n17 agar 23f o53 . aminew43rf 23 . 17 41.11D 411101 14 . 64 MOO variants nfvariants orf variants WNW 11.0111101.011variants variants variants orf nf orf nf orf variants 1. 1 13 24 1 17 5 =WA, Table 4 (cont'd) E44,, teto140 1 2 3 Grade 4 5 6 ...... Variants Total n6 f9 o14 .64 rf 14 20 25 .80 14 23 25 .92 7 21 4 6 6 1 5 5 6 .83 50 rf rn variants f variantsnf orf nf orf nf orf nf orf nf orf nf o f variants OMBf 011111~~~10variants 23 .91 f MIIIIMINIMMINIIMI1111110variants f variants f WINIONvariants 84 99 .85 ONO f POf 431 t0$ 12 231 P0f 15 251 f0 20 Y 0 6 f0 23 P0f 6012 4 t pf? 1 ? 11101100111 .110 ts? 24 f variants15 f37 orf57 .65 nf orf nf orf nf orf nf orf nf orfvariants 21nf o 60 92 .65 variants19 49 62 .79 variants21 52 65 .80 f variants16 31 34 .91 variants15 22 26 .85 f 107 251variants 336 .75 rf df 13 89 ft 2914 9 ft 2317 7 ft 43 81 ft 17 48 ft 16 4 1 d0 286419 0t ? 61 d0sz 61 sd0 1 0 d0 2 d0 1 t$fz? 136 2 1 Co Table 4 (contld) E....os.Oco 1 2 3 Grade 4 5 6 Variants Total 22 222 306 ON.f WOOo . 73rf n f o rf n f o . 92 rf n f o rf n f o rf n f o rf albn f o41/0 rf i variants Od 201 21f variants28 413 494 .84 d 359 54f variants28 479 521 d 425 f variants28 189 264 .72 d 136 f variants25 143 242 d 96. 59 f variants28 133 160 .83 d 102 f 159 1579variants 1987 260 . 79 f n o =MOW 0 0 54 0 53 0 47 0 31 d0 1319 Zn f rf 4n 11f 28o .39 rf 3n 6f o17 .35 rf 2n 4f o18 .91 rf n.f.1 0 rf n f 0 o 9n 21f 63o .33 rf variants d 11 f variants vd 24f variants vd f13 variants vd 16 5f 3rt ....3i 8o . 38rf 2n eura2f 7o . 29 rf 2n 2f 7OMo . 29 rf n way f 0 ....o rf 2n 2f o 50rf ON.2 .I3 11 12 o rf in va riants f varants i f varans f1 variants 4 . f variantsnf orf nf 5 . f60 variants 31 . 39 f v b 3 0 2 it 0f 1 0 2 0 3 fb0 381 Table 4 (cont'd) S 1 2 3 Grade 4 5 6 Variants Total 4n f7 9o . 78rf n11 f17 o62 rf.27 5n f7 o15 .47 rf n3 f3 o6 . 50rf 4n f4 o10 .4rf 2n 2f o5 . 40rf 29n f40 107 o . rf37 v f variants O 7f variants 0 17 f variants 0 7f variants 0 3f variants 0f OM,Z2f variants 0f f1 variants 0 f 37 f3 MM.n MSSf 0 o 01101110rf MOn OMf 0 o 00111rf n3 f3 34 o . 09 rf n f 0 OMBo MINNSrf 1 .10 1 1 rf1 *MOn f 0 o .111.rf n4 4f o35 . rf11 variants variants .1111, variants 01111. vO 21 r 1 0vr 21 MOM f 0 o WO.=rf n1 VIM f1 o 1 001111rf ONEn ONOf 0 MOOo rf 41111.1 0 0 OMB 0 MIIM rf 40110 NM 0 OMB 3 .111111111rf n1 MIDf 1 o1 rf 1 z variants s f1 variants s 1f Table 4 (cont'd) 4..w E-4 tes.es 1 2 3 Grade 4 5 6 Variants Total n f IMOo /MUM Omn MOOf o n OWo =MO WM* OM NNW 01010 MO OlIM alIMINO MODn %IND OMo MINIM, n OM o M 0 rf variants1 2 3 . f67rf f 0 rf nf orf nf orf 0 0 f 0 rf variants1 I2 3 . .-.6?rff z w0 1 w0 I1 n5 9f 32 .28 10 16 165 .10 9 14 151 .09 2 3 38 .08 6 10 92 .11 GINO2n 111.002f o1? .12 rf ems n34 mumf54 495 o .11 rf f 111011MMIUMMIAMOONIMvariants orf nf orf nf orf41Mf nf orf nf orf variants f variants f variants MIIMf variants f variants f variants Mf 0s 36 0d$ 11 23 dist 0s 257 0 3 0s 5 0s 1 dist $d0 3317 2 n1 sw.f 1 o1 rf1 ellIWn OMf 0 OMo 41111111111 rf OMBn OMf 0 NMo rf - 111Mn f 0 wooo rf n NMIf 0 ONOo rf IOWn ONOf 0 WMo 0111411110 rf n 1 OMf 1 o 1 =MOMrf 1 b A A variants 0 f 1 variants 0 OM1f Table 4 (cont° d " : 2 1 2 3 Grade 4 5 6 Variants Target n ONO OM MO MIONOO IIIIIM MOW 011111.0 ORM b r 3 4f 11 . 36 6 14 18 . 78 I 2 2 1 0 2 Mlle2 2 3 3 1 14 25 40o . 62rf r variants orf nf orf -.4f variants 8f variantsnf orf nf f orf nf orf variants 6.33 f 001111101111.1111!variantsnf orf nf IMOf variants OMBf 0 0 6 w 2 w 2 0 3 0 1015 MOW IMMO 0.1.10 41110 WINVO 11011. 1 3 3o 1rf 3 :s o rf 1n 1f o1 ONION.rf MOO INNO era.o rf VINOn f WNWo MOO!rf o GINIONNIrf n 41110f oalga SWAMPrf ONIP 6 .50 1 0 0 0 5 7 10 . 70 d r variants 3f variants 3f variants IMP1f variants 101.0 .1111110 6 n f o 6100110 MY111111 MINO 01.1. MIND .01.10 alell OM OINIO - 0 rf 2n 2f 5o . 40 rf o rf n f 0 o rf n f 0 o rf n f 0 aimo MINIM.rf 102n 2f GOO 5o rf ,MINMMV00INOvariants 0 2 variants 0 .40 2f CA.1 Table 4 (cont'd) :- 1 3 Grade 4 5 6 Variants Total 2 WPM 2 M.2 1rf f 0 o ON111110 rf MINNn OMf 0 o MINNOWrf WPMn f 0 o rf OMn WAD f 0 OMBo rf n f 0 MMo rf OM.ii2 ONOf2 o2 rf 1 ir r 4101411101111001110.11111MIWOvariants MOM2 variants 2 1 1 o1 1rf ONOn f 0 oWee 01/1rf n f 0 o rf 0111110n f 0 OMBo rf n 4101010f 0 ONOo rf ONO ONO 0 WO. o ...MOrf 4WDit 1 f1 Weeo 1 rf 1 gl variants 1 variants 0 n f o o WNW OM IMO 111.11= .110 WOO 4110 MO OM MO .110 NININDIA M. gr r variants1 1 1 frf1 variants1 6 6 frf1 0 orf nf orf nf orf nf orf nf o 0 0 0 variants2 7 7 rff 1 0 1 w0A 4 1 w0A 421 4411.11111110.11114110/1111011010.1 Table 4 (cont'd) 1 2 3 Grade 4 5 6 Variants Total 10 MOO 111.11.00 MIRO n1 1f 01 1rf f 0 o --rf n f 0 o ...... rf n am f 0 moo o Immorf Ma.n MOM f 0 oewe 41111...rf =MOn f 0 ONOo *Momrf n1 aeaf1 o1 rf 1 va riants 1 va riants 1Miraf WO MOW 010 IMO ...... OM 110110 MOW NIMONI/O 4.1M SIM SM. 111141 OINID 61101111110 WM 4111/0 n1 f o 1 1rf f 0 o rf n f 0 o rf n f 0 o rf n f 0 o rf n f 0 o rf NNWn1 *MOf1 MOWo1 rf 1 variants IMO va riants O 1 f n IMMO 4IWO M. IMO OMB *MOOD ellw WO MOO COMM MO OM 11110 Olga 1111M OINOMINt o .,- n. WM ONO 4011/101M k r f rf f 0 o .....rf n f 0 o rf f 0 o rf 1n f1 o1 1rf n f 0 o rf n1 f1 o1 rf1 r 1variants Mop f1 variants f Table 4 (cont'd) *MPI.'OTC.as 1 2 3 Grade 4 5 6 Variants Total n f NMI *VS 2 2 *MP 33 ORM MOD 1 MEW one 1 mown f 0 o rf OM n f 0 o rf n3 6f o10 60rf p/ A 0 orf nf orf nf orf nf orf variants 6 . f 0 variants 4 4 f . variants . 1f 1111101101111MINONSINY0 Ow 1 w 4 0w 5 n 1 f 1 1 1 0 *ND 1 OM. 0 SOD NINO NOM 0 SIM MYre f 0 o rf OMn f 0 iNMo rf n 1 /1 f o1 rf 1 variants orf nf orf nf orfMIDf nf orf variants 1f pr r 0 1 0 n 1 OMf 1 o 1 rf1 n 1 f1 o 1 1rf WY n MIDf 0 MDo rf SIMn MNf 0 OM0 rf OMn MM.f 0 Mlleo rf OMn f 0 1Io rf n2 2f 2o ...NMrf 1 sk s variantsONN11MIND O f 1 variants11.1.111110. 0 f1 variants 0 2f Table 4 (cont'd) 1 2 3 Grade 4 5 6 Variants Total n OMOD f ame OINM 011M .01, 0110/1100 OM. 41111M IMONIMI n WMo ...0n 41/100 1011o 0110111 n IMO MOMo aolOOMID skw va2 riants 2 orf2 nf orf nf orff nf1 orf variants1 1 1 1 f1 variants3 3 3 f1 variants1 1 1 f1 f 0 rf variants1 f1 1 f1rf variants8 8f 8 rff1 0 2 0 1 0 3 0 1 0 1 0 8 n f 0 0 MO 1 ONO 1 1 1 sae 0 NM OMn WIN f 0 111110o rf n f 0 mawo rf emn1 f1 ammoo1 rf 1 $ orf nf orf nf orf nf orf variants semf variants mowf O 1 0 1 $t n f o o 410 WW1 ONO 0110 MN, NNW 4110 n 41111110o 0 rf 0 rf variants1 1 1 1rf nf orf nf o 0 0 rf nf orf 0 variants1 f1 1 rf1 1 0 1 0%) 4.4 Table 4 (coned) :-.... Ii..:4-. I 2 3 Grade 4 5 6 . Variants Total OM, n IMOf 0 .1o rf OM.n1 OMf1 WI. oI rf1 MOnI Mef1 o1 1rf ONOn .10f 0 o rf IMOn .IIIID f 0 OMo rf WMn MVOf 0 o rf n2 WOO2f o2 rf1 tr r variants w f1 variants 0 1f variants w0 1f 1 2 2 1 0 4111111MO ONO 0 MOOo OWNINNO rf n WOf 0 MOo rf 111111. *NM 0 OM rf n f 0 o 1111=11011 rf n 1 =NO 2f 2o rf 1 9 1101111111111=100111111MINIMvariantsnf orf nf orf OMBf variants f 8 r t 2 t 2 cow Table 4 (cont'd) i4IIr:saZt 1 2 3 Grade 4 5 6 Variants Total ONIDn1 NEM3f o3 1rf ON2 SOW5 5o MOM,1r: SIMn ONOf 0 o rf MOW MOO 0 SIAMo rf n 0110111 0 41100 o rf 01100 Ma 0 WM 3 ONIM8 8 rf 1 Or r variants AMP f variants nf or: nf o 4111111111161111111110MMINININDvariants MIND 0 3 w 41 4 VariantsTotal 610 01212 1069 677 676 = 0w 4 SubstitutionsOmisDistortions s ions 30305 3 2 573639 0 628434 7 280397 0 472204 0 338181519 0 -= 21692.1.;854763 9 *For ofproducedoccurrences variant isand indicated oithe the numbertarget by thephoneme of occurrences in the speechof eacheach samples is cellnoted. n isfrom the thatnumber grade of ischildren indicated in the by columnthe grade who produced value. Relative frequency (r1) is obtained by dividing f by a variant of the row target phoneme. The number of o. In the lower part of each cell, the typeo value, and the number of variations Ui 76 4.1. 2. Summary of most stable variants.Table 5 is a summary, derived from Table 3, of all the variadons wIich occurred five times or more on the Goldman-Fristoe test, summed across grades.Target phonemes are rank ordered from the one (/8/) which had the most variants assoc iated with it down to the target with the fewest variants.2 For each target phoneme, the adjusted frequency of total variants is given in the first frequency column.Then, the adjusted frequency is noted for each variant which had an adjusted frequency of more than five.It will frequently be the case, then, that the sum of the frequencies of the variants of a particular target will be smaller than the total frequency for that target.This is because other variants occurred with a frequency less than fiveand were not reported on the summary table. Table 5 also includes the features changed in the case of substitutions. Feature changes are noted by change in +1- feature value and also by change in the marking value of each feature.(Recall that Appendix A presents the linguistic background for the analysis of substitution errors in terms of phonetic features. ) The feature data of Table 5 will be referred to in the following chapter where a linguistic analyses of the substitutions will be presented.

4. 1. 3.Goldman-Fristoe variants com ared with those of Free Speech samples. When an articulation test, such as the Goldman-Fristoe, is administered to an individual there is the tacit assumption on the part of

2The frequencies reported on Table 5 and elsewhere in this section are adjusted frequencies.This means that if the taxget phoneme occurred more than once in the Goldman-Fristoe test, the raw frequencyof variation was divided by the number of occurrences on the Goldman-Fristoe.This was done so that all frequencies would be comparable to each other, asthough each target appeared only once on the test.For example, initial /z/ appears once on the Goldman-Fristoe test, but final /z/ appears fourtimes.Thus, row 13 of Table 5 reports the raw frequency of variation forinitial /z/, but for final /z/ reports the raw frequency (39 -- see Table 3) divided by four. Thus, row 18 of Table 5 reports the raw frequency of the most stable variant of initial / z/,15 substitutions of /s/. For final /z/ the frequencies of the most stable variants (/s/ substitution and omission, with raw frequencies of 15 and 23, respectively) are divided by four to produce the reported adjusted frequencies of 3.75 and 5.75.Thus, the reader can examine row 13 of Table 5 and compare, for example,/8/ for /z/substitutions in initial and final positions, with no confounding due to frequency of occulrence.

7'9 Summary of Frequencies and Feature ChangesProduced for Those by Speakers Variants of BE on the Goldman-Fristoe Test Table 5 * Over-all Frequency with a Frequency Greater than Five Frequency of Change Target Position of Variationof Target 37 IndividualVariants f IndividualVariants 18 +1-FeaturesChange +strin-COT FeaturesU corin M/U Co d 14 5 + voi-cnt MU voi cntcor (273) m 126 omit f 2198 +str-cor U cor 110 omit 9611 +str-cor U cor Table 5 (cont'd) Target Position i Over-all Frequency of Variationof 107Target IndividualVariants d Frequency of IndividualVariants 106 +1-FeaturesChange in-cnt FeaturesChangeinU M/U cor (114) 5 -cnt U cntcorstr i 63 b 58 - -str-cnt MU cor cntstr (106) 2815 omit b 1317 -str-cnt MU Ucor cnt str (95) 95 omit t 7515 7 -voi VOi U voi c. Table 5 (cont'd) Change Target Position Over-all Frequency of Variationof Target 78 IndividualVariants Frequency of individualVariants 51 +1 - FeaturesChange in-voi Features in WUU voi (78) 41 omit 264012. 5 +voi M voi (55. 5) 111 14.10. 5 5 omitomit 41.10. 58 (52. 3) (52) 41.52 8 omitomit 3515 U voi (47. 33) (44) A 47.44 33 omitomit 46.36 67 6 -voi U voi Target Position Over-all Frequency of Variation Table 5 (cont'd) IndividualVariants Frequency of Individual +1- FeaturesChange in Changein M/U 0CO (38. 75) of Target 19 9. 75 omit s Variants 15 5. 75 -voi Features U voi 14. 5 14 3. 75 -cns-voi-lat-cor-ant-voc U-cns voi CA:,) (34) m 19. 5 15. 5 -voc-cns-cor-ant (30.4) 24 6. 4 omitomit 21 5. 4 -lat

.1540.11.204 .00.0.406. Table 5 (cont'd) Target Po s ition Over-all Frequency of Variationof Target IndividualVariants Frequency of IndividualVariants +1- Features Change in FeaturesillC hange M/U (27) Zri 11 8 C 35 -cnt-cnt U cnt (12)(13) 1213 8 omit tC 12. 53 5 +ant-cnt U cntU ant (11. 5) (9) 11. 9 5 omitomit 86 -str U corstr (6. 33) ever,totalTarget frequencyonly phonemes those associated variants are rank ordered according to frequency with each target 6. 33 appears in parentheses below the targetomit symbol; of variants associated5. with each. 33 how- The testquencies (see Footnoteare adjusted 2, thisfrequencies section, are reported which had a frequency of if thefor target a detailed appeared discussion more than once in the Goldman-Fristoeof the adjusted frequency figure). greater than five. An fre- 00 82 the a Lministrator that the phoneticperformance of the individual on the test will be essentially similar to thatof his everyday speech behavior.Table 6 p esentsinformation which can be used to evaluate thevalidity of that assumption. On that table the rows areeither particular target phonemes nr irdividual variantsof a target.(Thus, row 15 deals with all variants associated with the target phoneme/0/ in initial position, while row 18 is concerned only with/f/for /0/ substitutions in all positions.) For each target (or variant) defined on a row,the total number of children producing that variant (in eitherGoldman-Fristoe or Free Speech) is noted in the first column. Other columns presentthe proportion cf the total number of children producing thatvariant in (1) Both speech modalities; (2) Free Speech only; (3) Goldman-Fristoeonly.The third category is further divided among those who failed toproduce the variant in Free Speech because the target neveroccurred in their speech sample -- the N/O (no opportunity) column -- andthose who produced the target with no variation in the Free Speech -- theN/V (no variant) column. The ideal situation (from the pointof view of test accuracy) would be for 100% of the subjects to fallin the "Variant in both GF andFS" and the "No opportunity for variation"categories. Both the other columns represent test inaccuracies:Entries in the "No variation" column represent instances of misprediction,where a child was expected (on thebasis of the Goldman-Fristoe test) toproduce a variant in Free Speech whichhe in fact did not produce.Entries in the "Free Speech only" column represent instances of the failureof the articulation test to predict FreeSpeech errors. For thefollowing three variants, more than halfthe total subjects producing that variant did so on theGoldrnan-Fristoe test and failed to produce the same variant in Free Speech:/b/ substitution for /p/ in initial position; /f/substitution for /v/ in final position; andomission of final s/. For four variants, more than half the totalchildren who produced that variant did so only in Free Speech.These are omission of final /d/; /d/ substituting for /t /in medial position; It/ substitutingfor /0/ in all positions, and omission of/0/ in all positions. These data seem to indicate thatthe GoldMan-Fristoe test ofarticulation alone probably does not present anaccurate phonetic profile ofindividual subjects for all phonemes.However, many of the variants tabulated on Table 6 show a high degree ofstability across speech modalities. 4.1.4.Variants attributable to dialect asagainst variants attributable to developmental factors.One of the major goals was todetermine which phonetic variations seem tobe associated with Black dialectand which seem to be articulatoryvariations attributable to developmental factors.Variants which do not decrease infrequency across the six grades investigated can be assumed to perseverein the adult community and canbe assumed to be va riantsattributable to dialect. On the other hand,variants 83

Table 6 Relationships Among Goldman-Fristoe (G-F) and Free Speech (F-S) Variants in BE Group

* Produced Variant on Produced Produced Variant G-F, but not on F-S Variant Variant on both on F-S, ****G-F and F-Sbut not on G-F n** N/V***N/0

/p/ for /b/ 53 4 45 2 2 (f) .07 .85 .04 .04

Omit /b/ 27 2 19 -1 (f) .07 .70 .04

/t/ for /d/ 97 27 6 42 22 (f) .28 .06 .43 .23

Omit /d/ 85 2 2 10 71 (f) .02 .02 .12 .84

/k/ for /g/ 42 6 19 10 .7 ( f) .14 .45 .24 .17

Omit /g/ 14 2 8 4 o (f) .14 .57 .29 .00

/6/ for /39/ 36 6 28 1 1 (1) .17 .78 .03 .03

Omit /j/ 9 0 4 2 3 (f) .00 .44 .22 .33

/f/ for /v/ 9 5 1 2 (f) .56 .11 .22 84 Table 6 (cont'd)

Produced Variant on Produced Produced G-F, but not on F-S Variant Variant Variant on both on F-S, G-F and F-Sbut not on G-F N/V N/O

/b/ for /v/ 59 9 50 (i) .15 .85 .00 .00

Omit /v/ 39 8 4 3 24 62 (f) .21 .10 .08 .

/8/ for /z/ 21 8 0 3 10 .48 (f) .38 .00 .14

Omit /z/ 35 13 3 3 16 . (f) .37 .09 .09 46

Omit /s/ 30 16 4 3 (f) 53 .13 .10 .23 All /0/ Variants 48 11 20 6 11 .23 (1) .23 .42 .13 All /0/ Variants 134 17 67 41 9 (m) .13 .50 .31 .07

All /0/ 148 12 28 70 38 Variants .08 .19 .47 .26 (f) /f/ for /0/ 136 55 32 38 11 all positions .40 .24 .28 .08

/t/ for /0/ 91 6 3 8 74 all positions .07 .03 .09 .81 ;35 Table 6 (cont'd)

Produced Variant on Produced Produced Variant G-E', but not on F-S Variant Variant on both on F-S, G-F and F-Sbut not on G-F n N/V N/O 11111111 Omit /8/ 61 11 6 8 36 all positions .18 .09 .13 .59

/d/ for /4 /165 5 2 100 59 (1) .03 .01 .61 .36

/d/ for /S / 13 5 o o 8 (m) .38 .00 .00 .62

Omit /A/ 38 6 4 27 1 (m) .71 .03

Omit IL/ 89 8 3 59 19 (f) .09 .03 .66 .21

Omit /1 / 57 5 1 25 26 (m) .09 .02 .44 .46

Omit /1 / 141 13 0 88 40 (f) .09 .00 .62 .28

/b/ for /p/ 45 24 10 6 (f) .53 .22 .13

***N/V indicate the number (and proportion) ofsubjects who produced the target phoneme in FreeSpeech, but did not produce the indicatedvariant. ****N/0 indicates the number andproportion of subjects whose Free Speech sample did not include the targetphoneme (thus, there was, for these subjects, no opportunity toproduce the variant). 87 which decrease over time probably do not exist in any significant degree in the adult speech community and can be assumed to have a developmental origin.Henceforth, a general lack of decrease or a general increase of variants over grades will be referred to as a "dialect pattern" and a general decrease will be referred to as a "developmental pattern." 4.1.4.1.Word-final obstruents.The initial data analysis (as presented in Tables 3 and 4) involved the computation of the frequency of individual variants for each grade level.Initial computations revealed that the two patterns mentioned above could be discerned in the frequency by grade relationships of the most vulnerable target phonemes. In particular, for all the voiced final obstruents (/,/,/z/,/b/,Id/,/g/, and /inone type of variation (omission) followed the developmental pattern, while the other common type of variation (devoicing) followed a dialect pattern.Figure13 graphically depicts these relationships.The frequency of omissions of final voiced obstruents decreased across grade levels. The substitution of the voiceless final obstruent /s/, /p/, /k/, and /eV, respectively) dipped slightly in the middle grades, then increased in the fifth and sixth grades.While Figure 1 depicts the patterns associated with all the word-final obstruents, Table 3 shows that these patterns were valid for each of the word-final obstruents individually. SOW - To ascertain whether the frequencies revealed in Figure 1 do in fact differ from chance, an analysis of variance was performed on the data. For each subject a "developmental" score was derived by com- puting the average number of obstruent-final omissions in the Goldman- Fristoe data of that subject. A "dialect" score was also derived by com- puting the average number of obstruent-final devoicings in the Goldman- Fristoe data of that subject.These scores were then subjected to a six (grades) by two (omissions vs. devoicings) analysis of variance.The analysis revealed that the combined frequency of both types of variants differed across grades, F (5,188) = 2.773, p (.05.There were, overall, significantly more devoicings (dialect variants) than omissions (developmental variants) F (1,188) = 49.758, p ( .01.The mean dialect score per subject was .19, while the mean developmental score was .07.The most interesting aspect of the analysis, however, is the interaction of grade by type of variant.This interaction is also highly significant, F (5,188) = 5.165, p ( .01.

30nthis and all Figures depicting target phonemes which occurred more than once in the Goldman-Fristoe test, the frequency measureof the ordinate is an adjusted frequency (see footnote #2, this section). 88 Figure I Omissions (Developmental Variants) and DevoicingSubstitutions (Dialect Variants) by Grade for BE Group

211 OMISSION 26 MOWING 24 22 20 IS

6 4

6 GRADE

;-1 89 Duncan Multiple Range post-tests were carried out to ascertain exactly which grade differences were responsible for the significant results of the analysis of variance. None of the means of omissions differ significantly from each other.That is, the bottom line of Figure 1, while it reveals a general decrease in incidence of omission of final voiced obstruent across grades, the grade differences are not significant.Examination of the substitution of the voiceless counterpart of the final voiced ubstruent (the upper line of Figure 1) reveals the significant aspects of the analysis.The incidence of devoicings in the fifth and sixth grades is greater than in the first, third and fourth grades.Thus, the differential incidence of devoiced variants across grades is responsible for the significance of the 'grade' main effect.Furthermore, the significance of that variable is attributable to a real increase in frequency in the later grades, ratherthan to the inexplicable drop in devoicings in the third and fourth grades. The post-test demonstrates also that the major component of the significant interaction between type of error and grade is attributable to the divergence of frequency scores in the fifth and sixth grades. Thus, the anticipation that the two different types of variations associated with voiced obstruents in word-final position pattern differently is tentatively confirmed.The importance of this finding will be developed in more detail in the following section. Variants of other target phonemes were found to pattern differ- entially such that in each both a developmental and a dialect pattern could be discerned. Each of these four phonemes,/13/,/d/, /SI, and /2/, will be considered separately.Again, a discussion of the linguistic significance of these results will be postponed until the following section. 4.1.4. 2.Omission of /X I.Figure 2 displays the differential distribution of IT 1 omissions, dependent upon position of the target phoneme. (By "omission of /1 /" we mean, of course, omission of the consonantal aspect of that phoneme, leaving only a schwaquality vowel.) Omissions of a /in final position increase over grades and clearly show what is identified as a dialect pattern. Omission of /11/ in medial position, on the other hand, shows a general decrease in frequency over grades in accordance with a developmental pattern.Chi squaretests4were carried out on omissions in each position toascertain

4Differential frequencies across grades for /f /,/0/,/d/, and /A/ were analyzed by the Chi Square statistic.The reason for this is that for each phoneme, a subject's developmental and dialect scores were either "0" or "1", since only one instance of each type of variant was possible for each phoneme.

4,...elk V 4 90 Figure 2 Incidence of Omission of/t. /in Medial and Final Position by Grade for BE Group

20 amen- MEDIAL 26 . .41. FINAL 24 22 -- 20

1$ 16- 14

12 100

6-a-0 4 2.-- 0

1 2 3 4 3 6 GRADS 91 Figure 3 Incidence of Omission of ///in Medial and Final Position by Grade for BE Group

21 MEDIAL 26 .11,... FINAL 24.. 22

10I.

6 4

2 0 11011. 4,-

1 2 3 4 5 6 ORADI 92 if the frequency distributions acrossgrades differed from chance. For final omissions of/0"/, X 2 = 24.7 (df = 5), p =.004 and for medial omissions,7c,2= 20.6 (df =5), p =.001.The more interesting statistic is the two-way chi square,which demonstrates the significant divergence of the twofrequency distributions.X2 for the two-way test = 37.1 (d). = 5), p =.0001. The significance of the two-wayChi Square indicates an interaction ofomissions in the two positions with thegrade variable. The greater than chance divergenceof the two frequency distributions demonstrates that two separate patterns,corresponding to the position of the target phoneme, have beenisolated. 4.1.4.3. Omission of /2/.Like /1 /,omissions of final / A/ follow the dialect pattern, as shownin Figure 3, while omissions of initial and medial /2/ follow thedevelopmental pattern (note that/g. / does not appear in initialposition, so only medial/1 / was considered).7(2for final omissions = 23. 9 (df = 5), p =.0002 andX2 for initial and medial omissions = 11.7 (df = f), p = . 04.The two-way chi square reveals7,2 5.7 (df = 1) p = . 02, indicating a greaterthan chance divergence.Thus, omission of medial /2/ decreases acrossgrades and omission of final /A/ increases across grades. Theinteraction of the position variable with the grade variable (revealed bythe significant two-way Chi Square) demonstrates that two distinct patterns,corresponding to the position of the target phoneme, have beenisolated. 4.1.4.4. All types of variants of/0/. Figure 4 shows the developmental pattern of variantsassociated with /0/ in initial position(1,2= 12.8 (df = 5), p = .02) and thedialect pattern of /0/ variants occurringin medial and finalpositions5 (7C.2= 9.5, df = 5, p = .08).The two-way chi square produces7.2= 9.4 (df = 5), p = .09, indicating that the divergence of the two distributions is not greaterthan might have been expected by chance. 4.1.4.5. Substitution of /d/ for /to I.Figure 5 presents the frequency of the /d/for /t /substitution in two positions -- initial and medial. Substitutions in initial position appear tofollow a dialect pattern

5Frequencies are also adjusted on.graphs (such asFigure 4) which compare incidenceof variation in one position to the incidenceof variation in two other positions.Thus, the frequencies of variation of/0/in initial position are raw frequencies(since there was only one /e/ occurring in initial position on theGoldman-Fristoe test.Frequencies associated with the medial final variations areadjusted by simply adding medial raw frequencies and final raw frequenciesand dividing by two. 93 Figure 4 Incidence of Variants (of all types) Associated with /0/ in Initial and Medial + Final Positions by Grade for BE Group

28 26 24 22 30

16 Iv 14 aa. se 12

6 4

1 2 3 4 3 6 GRADS 94 Figure 5 The Substitution of Id/ for//in Initial and Medial Position by Grade for BE Group

25 26 24 22 20

16 14

11 6 4

1 2 5 6 95/ with an overall increase across grades. However, a Chi Square test on this distributionindicates only chance distribution across grades substitution of /d/ for medial /i /is a very low frequency event, but the few instances of that variation do occur in the earlygradel, indicating a slight developmental pattern, but not beyond a chance variation. Thus, differential dialect as against developmental patterns are discernable in ten of the target phonemes for the Niagara Falls speakers of Black English. The validity of these patterns, as well as their theoretical significance will be discussed in Section 5.

4. 2.San Antonio: The Mexican-American Sample 4. 2.1. Raw data presentation.Tables 7 and 8 report the data from the San Antonio sample. The former table is for Goldrnan-Fristoe data, the latter for Free Speech.The rows, columns, and cell entries are exactly the same as tables 3and 4 for Niagara Falls sample. (See page 20 for a detailed descriptionof cell entries.) 4. 2.2. Summary of most stable variants.Table 9, which reports variations which occurred in the San Antonio Goldman-Fristoe data with a frequency greater than five isanalagous to Niagara Fall's Table 5. Target phonemes are rank ordered from the one (/z/) which produced the most variants associated with it, down to the target with the fewest variants.Table 9 (like Table 5) also includes the features changed in the case of substitutions. Feature changes are noted by a change in +1-feature value and also by change in the marking value of each feature. Data reported on Table 9 are adjusted frequencies where relevant (see footnote 2, this section, for a definition of adjusted frequency). The linguistic significance of these variants will be discussed fully in the following chapter. 4. 2.3.Goldman-Fristoe variants compared with those of Free Speech samples.Table 10 presents, for the San Antonio sample, the same information that Table 6 reports forthe Niagara Falls sample. This table provides a basis for evaluating the stability of phonetic performance across speech modalities. In general, it appears that the Goldman-Fristoe better predicts phonetic performance in the free speech of the Mexican-American child than it does for the Black child.Only two variants,/f/ substituting for /8/ in final position and /b/ substituting for /v/ in final position, show more than half the subjects failing to produce in Free Speech the variant they produced on the Goldman-Fristoe test.The Goldman- Fristoe test failed to predict Free Speech variations for more than half (+ex+ cont;nmedon page I(ea)

1.4 Rd glop Table 7 Variants Produced by Speakers of MAE on the Goldman.Fristoe Test of Articulation* Grade Total 1 41111/01101111111141/1111111. 2 3 4 5 tt 6 Variants n f rf n1 f1 rf n f rf IMP 1NIMM rf n f rf n f rf n f rf (4)rn variants5 f 10 . 50 variants f . 251 variants2 f 4 . 450 0 variants4 f 5 . 531 variants1 f 1 . 251 variants13 21 21. f40 0 0 010110 CoC. n12 f250 10 . 42rf n11 f12 . 22rf UMW4n 7f . 35rf n1 4f . 80rf n10 f019 . 38rf 4n 7f0 . 35rf n42 0 74f . 35rf (5)f variants f 0 25 variants f 0 12 variants f 0 7 variants f 0 4 variants f 0A 18 1 variants f 0 7 variants 0I 73 f 1 rf rf OM. 0 1111111.rf 110110 n f0 rf 0111IMn1 IMMO f1 rf1 2n 2f rf1 variants f v 1 variants vP f1 Table 7 (cont'd) co 1 2 3 Grade 4 5 6 Variants Total n16 f16 rf1 n14 f14 rf1 n12 f12 rf1 5n f5 rf1 n11 f11 rf 1 4n 4f rf 1 n62 f62 rf1 b (1)f variants f p 14 1 variants f P 13 1 variants f p 11 1 variants f p 5 variants f p 11 variants f P 4 variants f f31 INA n 0f 1 n f n f n n f n f n P0 f 58 c;)c (2)1 variants16 f f235 .72 rf19 variants20 f 29f .72 26rf variaL.ts16 f fg25 .78 22rf variants15 f 201 .67 rf18 variants11 f 11 . 50rf11 variants10 f 14 . 70rf13 variants88 s 122 . f69rf5 d ttss 21 ti$ 21 dist t 21 dist s 1 i ti 1 dist tsti 109 431 , 4.4 Table 7 (cont'd) 1-4 :.Rsto41) 1 2 3 Grade 4 5 6 Variants Total n10 f10 rf1 n14 f14 rf1 n12 f12 rf1 n11 f rf n11 f rf 8n 8f 1. 0 rf n66 f66 rf1 (1)rn variants f 5 7 variants f 5 12 variants f 5 11 variants f $11 1. 0 10 variants f 511 1. 0 11 variants f A 8 variants s .... f1 ttss 1 t 2 t 1 dist 1 tst; 59 41 n20 20f rf1 n15 f15 1. 0 rf n20 20f 1. 0 rf ..8n. 8f 1. 0 rf n13 f13 1. 0 rf ..n6 f6 1. 0 rf 82n dist 82f rf1 (1)1 variants f te. 19 1 variants f i 15 variants f dist g 19~A 1 variants f dist I 71 variants f 5 13 variants f A 6 variants tg 79 2f 1 n 0f rf n 0f rf n2 2f rf1 n f0 rf n f0 rf n f0 rf 2n dist 2f rf1 d (1)1 variants f t 2 variants t 2f .0 44a) Table 7 (cont'd) E-4 kesto 1 2 3 Grade 4 5 6 Variants Total 0f rf n 0f rf n 0f rf 0f rf 0f rf MAIM 1 f1 rf1 1 1 rf1 (1)m a variants 0 1 f variants 1....4 22n f22 rf1 n12 f12 rf1 n16 f16 rf1 ..-8n 108f 111111010rf1 n12 f12 rf1 n19 f19 rf1 n89 f89 ...... rf1 I- (1)f variants f t 20 2 variants f t 11 1 variants f t 15 1 variants f t 71 variants f t 10 2 variants f t 19 variants 0 82 7f n3 f03 rf n1 f01 rf n1 0f rf n f0 rf n1 0f rf n 0f rf n6 t 7f rf (3)1 variants f .33 3 variants f p .33 1 variants f p2.67 2 0 variants f p1.33 1 variants p . f739 Table 7 (cont'd) 4.4 1 2 3 Grade 4 5 6 Variants Total 3 3 1 1 1 1 1 1 1 0 0 1 1 1 6 6 1 f (1)f variants s 3 f variants f 0 1 variants 0 1 f variants f s 1 variants s0 4f2 II" MOM n1 f1 rf1 n1 f 1 rf1 SON.n 0f IIIIIIMINIIrf 111000n NEMO0f 110111111.1111rf ON.n f 011011Mrf WIMPn MEMOf rf n W.0f rf1 C/..(.. (1)i variants d 1 f variants f d I 0 0 variants2 d 2 f2 2 2f rf1 1 fI 1rf 0 rf OMIND GOMM 0f rf WNW ,, -f. IMMOf0 rf 3 3f rf1 (1)m variants kd 1 f variants k 1 f variants kd f21 11 ilmeme M.+ 44 Table 7 (cont'd) P AsIx014 1 2 3 Grade 4 5 6 Variants Total n19 f19 rf1 n13 f13 rf1 n12 f12 rf1 n7 Nowf7 rf 1 n10 f10 rf 1 3n elm3f rf 1 n64 f64 rf1 g (1)f variants f k 9 16 1 variants f k 12 1 variants f k 11 1 variants f k 7 variants f k 9 91 variants f k 3 variants k0 58 f2 1....% dist t 1 0 0 dist t? 21 14;**- _n 101f0 rf n2 2f 1111011100Drf1 n2 f2 rf1 ONOn MOODf0 rf n1 f1 rf1 OM*n 0110f rf sawn5 f5 rf1 3 (1)i variants f Zd 1 variants f dz., 1 1111111111110001101111111100111variants f d 1MIND variants Cd1 f31 Table 7 (cont'd) 1 2 3 Grade 4 5 6 Variants Total 11n3 3f rf1 0.11.n f 1rf n f IMINSIMIIrf1 WM.n1 fa"f1 rf 1 n f MONIONIDrf 1 MOOn OPMf 4110/00rf n f 1rf (1)m variants f variants6 6 f variants2 f 2 1 variants f 1 variants2 f 2 0 variants14 e" 14 f C/1,...Fmit cd., 21 e'Yd 21 yd 1 Y d 2 V'd 4721 j n f rf1 n f0 rf1 n f rf1 n f 1rf n f rf 1 n f rf n 0 f 1rf (1)f 11 11 17 ...17 f 11 ...11 9 9 11 11 4 4 63 63 variants f sC.v 371 variants cz... 16 1 variants f sc ... 29 variants f cv 9 variants f C'z 91 variants f a' 4 variants scs 54 5f1 0 0z 1 0LAI1.1I Table 7 (cont'd) E..4)aste144) 1 2 3 Grade 4 5 6 Variants Total . I 1 0 rf WOO 0 rf MOMn2 f rf IMOn 0f ONMINlowrf Orb7n f7 rf (3)i variants4 f 4 33 variants f .33 MON variants f 2.33 I variants .33 f al1111110 1 2 7 n ft 4rf n ft 1 rf n f1 rf n1 f1 rf n ft rf f rf t (2)m variants2 f 2 . 501 variants1 f . 501 variants1 f .50 1 variants f .50 1 0 0 variants5 5 .50 f t ? 1 0 f 0 09 211 (1)f ONO 3 3f 01101101.rf1 IBM 0f 0111111110rf 01110 1 gam 1 ONIONNINDrf1 OEM 1 IMO 1 Oil!rf1 1111.1111Myrf 0 .101111110 rf variants5 f5 rf1 variants f 0g? 1 variants f 0 1 variants f 0 1 0 31 Table 7 (cont'd) E-4 1 2 3 Grade 4 5 6 Variants Total (2)m 0 variants2 f 2.50 variants1 f 1.50 0 0 0 variants3 3 . f50 (3)f variantsn18 f 41f .76 rf variants20n f 34f0 .57 rf2 variantsn11 f 20f0 . 61rf1 variants8n f f13 . 54rf variantsn17 f 31f . 61rf variantsn12 f f18 . 50rf variants86n 0 157 f . f61rf3 1411111.01011=1.111MT 0 33 8 0 34 0 20 0 13 0 31 w0 16 2 w0 147 /IMMO. 10 (2)rn 2n 2.50f rf n 0f rf n1 f1.50 rf n 0f rf n 0f rf n f0 rf n3 3f . 50rf variants f 0n 11 variants f Y 1 variants Yn0 f11 0cri11 Table 7 (contld) cdbe$3.) 1 2 3 Grade 4 5 6 Variants Total IONIMn3 f3 . 33rf OM* n1 f 1 . 33rf 0 rf 1 f1 . 33rf MINDn AIM. 0f ...MOWN.rf OWN. n MM. 0f rf ownn5 *am.f5 . 33rf m (3)f variants f n0 21 variants f 0 1 variants f 0 1 variants 0 4f 1 rf rf MON 1 1 rf1 WOO 011.0 .11110110.11rf OM. rf (1)1 0 0 variants f d 1 0 variants 1 1 f1 rf 0 MMIMIONOrf MINIMn1 f1 rf INI.n OWN.0-f rf 01.n.1 f rf OWNSn ONIMMf0 rf MVOn3 ...3f rf (5)m variants f 0 .20 1 variants f 0 .20 1 variants f 01.20 1 variants 0 .20 f3 : Table 7 (cont'd) E44.) kedtotJ 1 2 3 Grade 4 5 6 Variants Total n10 f11 . 22rf 9n f14 . rf31 n6 7f .23 rf n3 f3 .20 rf n1 f 1 .20 rf 4n 4f. .20 rf 33n 40f .24 rf n (5)1 variants f 0? 29 variants f 0? 13 1 variants f 0 7 variants f 0 3 variants f 0 1 variants f 0 4 variants 0? f37 3 (Cs.)f.a.t .. 6n f6 rf1 4n ONOOD4f rf1 ....n6 f6 411010110 1rf nMIMS6 6f rf1 41110W n OOMO0f ONOMMEWOrf MONOn 1MINOf0 rf 22n 22f rf1 (1)1 variants f bw 15 variants f b 4 variants f b 6 variants f b 6 variants wb 21f 1 P n1 f rf n2 f rf 2n f rf n1 f rf n f rf n f rf n 6f rf (3)in variants f b1.33 1 variants f b2.33 2 variants f b2.33 2 variants f b1.33 1 0 0 variants6 b . f336 -J1..40 00s- 4.1 Table 7 (cont'd) 0 E-t kAiazv 1 2 3 Grade 4 5 6 Variants Total p (2)1 variants6 f f6 . rf50 variants6 f f6 . rf50 nvariants1 f f1.50 rf variantsn 1 f f1.50 rf n _f0 rf variantsn3 f f3.50 rf variantsn17 f17 . frf50 n6 8fO . 667rf 4n 04f . 650rf n5 0f rf1 2n 0f rf1 n10 f11 . rf55 n _f00 rf3 n27 0 f30 .17 56rf H.Y.'r.... (2)i variants f w 8 variants f b 4 variants f w5.50 5 variants f w2.50A 1 variants f w 11 variants wb2 25 4f1 4n f rf n3 f rf n2 f rf n1 f rf n1 f rf n 0f rf n11 f13 . rf59 (2)rn variants f wO6.75 51 variants f w03.50 21 variants f w2.50 2 variants f w1.50 1 variants f w1.50 1 variants w0 11 f2 Table 7 (cont'd) 1 2 3 Grade 4 s. 6 Variants Total (2)1 n3 f3 .50 rf 4n f5.62 rf n2 2.50f rf n 0f rf n2 f3 .75 rf n3 f3.50 rf n14 16f . 57rf iwzd variants f d0 1 variants f ez 41 variants f z 2 variants f dist z 21 variants Zz 1 f variants C0 2f1 - 1....1 dist dist dist dz 48 s .011011 111111118=101 6 6 . 50 3 4 . 67 MOM 0 rf OM= 0f rf n2 2f rf n2 f rf n f rf (2)rn variants f 3 variants f variants f . 501 variants f 2 .50 variants13 14 . OM. 54 dist 0? 1 tIS0 21 0Z 1 0 2 90tC 3161 dist t ? 11 Table 7 (cont'd) tO1.4 E- C 1 2 3 Grade 4 5 6 Variants Total SNOWn5 f rf ...an1 f2 1. 0 rf n1 f rf n1 f rf .....n1 f rf n 01.f0 rf n9 f10 . 56rf s (2)f variants f 5.50 3 variants f 9 2 variants f 1.50 1 variants f 1.50 1 variants f es1.50 1 variants OEN. f51 dist 0I 1 0 0 00Z1 21 n f rf1 n f rf1 n f rf n f rf1 n f rf1 n f rf1 n dist f rf1 1"1.r.<.) (1)1 variants16 f 16 variants12 f 12 1 1 variants9 f 9 I variants16 f C16 variants12 f 12 variants14 Z14 f variants79 v 79 f g tCs 14 1 as 1 3. 81 16 Z 12 14 I3tcs 75 21 Table 7 (cont'd) E4 1.4 1 2 3 Grade 4 5 6 Variants Total n10 f10 1rf 4n 4f rf1 _9n 9f rf1 n10 f10 rf1 4n 4f rf1 n6 6f rf1 n43 43f rf1 (1)m variants e# 91 f variants f 4 variants f 9 variants f c 10M. variants f 4 variants f 6 variants 42 f 1 faj4. i'' n f0 rf1 n f rf nIMO f rf n f rf n f rf n f rf n 0 f rf C-%k 11 11 9 9 1 9 9 1 15 15 1 9 9 1 9 9 1 62 62 1 - (1)f variants f variants a 8 f variants f variants f variants f variants f variants f sC 10 1 dist 1 t6 81 C' 15 $e' 81 dist C 81 dist tas 57 21 1 f1 . 50rf MOO MOO 0 rf n1 f rf n1 f rf n 0f rf n1 f rf .....n4 =NO4f . 50rf (2)m variants f 0 1 variants f 01.50 1 variants f d1.50 1 variants f 01.50 1 variants 0 f 31 Table 7 (cont'd) Some 0 1 2 3 Grade 4 5 6 Variants Total 4 7 .88 2 2 . 50 OWN 0 rf n1 f1 .50 rf n2 f2 . 50rf n2 2f . 50rf n11 f14 . rf64 t (2)f variants 0? 43 f variants f ? 2 variants f k 1 variants f 0 2 variants f 0 2 variants 09 8f1 n18 f18 rf1 n16 f16 rf1 n15 f15 rf1 n10 f10 rf1 n16 f16 rf1 n5 f5 rf1 n80 80f rf1 (1)i variants 13 f variants ,., 14 f variants f 14 variants f 10 variants f 16 variants f 5 variants f 4f1 0 dtfs 31 ftc 1 t$ 1 t t t tZds 72 21 44 V Table 7 (cont'd) E 4 caa,to 1 2 3 Grade 4 5 6 Variants Total (1)m n10 f10 rf1 NIVVO7n OM. 7f rf1 n12 f12 rf1 MONO9n NNW9f rf1 n5 5f rf1 ONIMP 9n GOIM9f 1rf n52 f52 OINN1rf variants f ds 71 variants f 1$ 421 variants f fs 246 variants f t 33 variants f 0$ 21 variants f f$ 24 variants d0 11 f1 0t 1 0 0 s0 ft 1 0 9 21 t$f .7 2311 51 (1)f MOMvariants8n f 8f rf1 variantsONn5 f MOM 5f ~IMODrf1 variants8n f 8f rf1 variantsn3 f f3 rf1 variants4n f 4f rf1 variantsn3 f f3 rf1 variantsn31 f31 frf1 ts 26 0ft 21 t$f 251 0ts 11 ft 2 1s 21 t0sf 1011 82 Table 7 (cont'd) F.+ E-4 to 1 2 3 Grade 4 5 6 Variants Total n19 f19 rf1 n21 21f rf1 n23 f23 rf1 n15 f15 rf1 n13 f13 rf1 n13 f13 rf1 104 n 104 f ri1 (1)1 variants d 19 f variants d 21 f variants d 23 f variants d 14 f 1 variants d 13 f variants d 13 f variants d0 103 f 1 4 4 rf1 2n 2f rf1 n1 f 1 rf1 00 rf n1 f1 rf1 n1 f 1 rf1 9n 9f rf1 VsiC'.. (1)m variants d 4 f variants dz 11 f variants d 1IMIIIIM f variants f d UMW1 variants f d 1 variants dz 01 8f1 n16 f16 rf1 n13 f13 rf1 n16 f16 1rf WINOn7 f7 rf1 n13 f13 rf1 MV. 8n MOM8f rf1 n73 f73 frf1 (1)i variants 0b 10 f 24 variants b0 481 f variants 10b 10 f 13 variants f b0 25 variants f w0b 391 variants b0 35 f variants 0b 4719 2 f r wm 1 wmfrA 21I. Table 7 (coned) 1 2 3 Grade 4 5 6 Variants Total n18 f18 rf1 n10 f10 1rf n11 f11 rf1 441n f rf1 n f rf1 WOOn f 1rf n f 0111rf (1)rn variants f 0 variants f 0 7 variants f variants5 f 5 variants4 f 4 variants4 f 4 variants52 52 f 1 wb 10 71 b 3 b0 47 0 41 0 4 0 4 0b 3614 1 n f n f n 011014111MD n n (1)1 variants18 f 18 11rf1 variants21 f 21 18rf1 variants12 f f12 11rf1 variants11 f f11 rf1 variants13 f f13 1rf variantsn13 f f13 1rf variants88n 88f frf1 0b0f 2.41 0f 21 0f 1 0f 10 1 f 13 bf 12 1 b0 75 373 11111111111111M f MIN 1 WOO 01.11101110 rf SIMn NINOI rf WOO rf nMi. WNW f OMINIMO rf ftlMn MMOf MMUMrf IMOn OMOf rf n1 f1 rf1 w (1)i variants f variants GOISIEID Table 7 (cont'd) M.+ 44beo Grade Total ONN111001111101110111.111111011E4 kns 1 2 3 4 5 6 Variants n11 f11 rf1 4 41000 4f rf1 6 6 rf1 4 0010 4 rf1 ONO 6 6 rf1 8 f8 rf1 n39 f39 rf1 (1)i variants f s 81 variants f $ 21 variants f s 6 variants f $ 31 variants f $ 51 variants f $1 71 variants gd f21 dist dA 1 dist g 1 v d v$A 31 11 F-4 01111111MIN dist z 2 8n 8.,f rf1 n11 f11 rf1 IMMO4n 4f rf1 nIMO GM.0f rf nIMO2 f2 rf1 GRIM4n 4f rf1 n29 29f rf1 (1)m variants f $ 61 variants f s 3 variants f $ 4 variants f s 11 variants f s 4 variants d 18 f2 4 d 1 dist 4i 251 d dist tisZ 261 Table 7 (cont'd) E.'4.)it14no0 1 2 3 Grade 4 5 6 Variants Total 5 5 £ 4n f4 n2 f2 n 1 f1 01110n9 IIf9 OWNSn1 f1 n22 f22 , br r variants dw0 21 variants f w0 31 variants f w0 1 variants f 0 1 variants f w0 81 variants f w 1 variants f wd0 15 61 MOMn2 f2 n 0 f n 0 f n 0 f n1 f1 OVINO n 0 WNWf ONNIDn3 f3 tC.... d variants f 0 11 variants f b 1 variants b0 f1 dr _n6 6f n1 f1 n5 f5 n1 f1 n 1 f1 n 0 0111111. f 411.110n014 OM, f14 r variants f Yw 15 variants f w 1 variants f w0 23 variants f 0 1 variants f w 1 variants Yw0 10 f31 Table 7 (contld) 1 2 3 Grade 4 5 6 n Variants Total z (4)1 variants24 f $81 .84 73 variants31 100 f.81 $ variants33 113 f .86 rf variantsnf32 102 f .80 rf variantsnf31 f $96 .77 rf variantsnf29 f 92 .79 rf 180 variants 584 f .81 frf 1.a.s nfrf nf rfnf 51 99 1 s 112 1 s 102 94 1 $ 91 1 0$ 7 CU dist t0 11 dist i dist 0 1 0 0ti 571 11 OWN dist 3 n 0 f n 0 f n1 f1 n 0 f 2n I2 n 0 f MN= 3 =MN* 3 bi variants f 0 1 variants f 2 variants f 3 011.0 SIMS 01111 n0 0 WOWf 1 1 40 4 br b variants f 1 variants f 0 1 variants f 0 1 variants f 0 1 variants f 0 13 Table 7 (cont'd) 1 2 3 Grade 4 5 6 Variants Total .1.1111 n 1 GM. f1 011110n1 4181Im £1 n 0 f n 0 WINOf 11111011n 0 MI11110f n f WPM.n f variants P 1 f variants P 1 f 0 variants fP2 2 2 n 0 f n1 f 1 OMBn 1 GOND f1 n 0 f n 0 f 1IMn2 4IMM 2f 4n 4f variants w 1f variants f 0 1 variants f w 1 variants f w0 2 n1 WNWf 1 OM*n IMMOf n 4000f n f OMNIn 41100f n0 MO=f 011.116n1 MM.f1 gr r variants W 1f 0 0 0 0 0 ve riants w f1 4., Table 7 (cont'd) E-4 ittof-to 1 2 3 Grade 4 5 6 Variants Total n ON. f n f n f kL A 0 variants f 1 variants f 1 variants f 2 2 2 variants f 4 4 4 ealee 41 OM* n f 0n f n0 f 0n f 0 1 1 k variants f 1 1 1 0 0 0 0 0 variants f 1 n3 f3 n3 3f n1 f1 n2 f2 n3 f3 n 0 f n12 f12 r variants f 0.4o variants f variants f . WPM variants f variants f variants f . 1 . 1 . 0w 21 w 3 0 w0 1 w0 21 w0 75

1 .*-n44.14 Attii!..,.4.8t1;.:4446:+stros:4'44#44416.ktit:;14:0:#: . Table 7 (cont'd) 2 3 Grade 4 5 6 Variants Total n f n =Maf n f n f n 01111.0 f variants f 1 variants f 1 pL 2 2 0 0 0 02 2 1 1 05 5 / variants f 0 1 variants f 0 2 variants f w 1 variants f 0 23 3 MOO 3 IMO 0 OHIO n 0 NOMf 0 NNW 01111. 1 OMNI 1 n 0 f IMMO4 OIMMO4 skw s ----a.,.-ariants a 0 3 f variants f dist 1 distvariants f 0 13 Table 7 (contld) Grade Total n 1 n 2 f n 3 f n 4 f .....n 5 .....f n 6 f nVariants f skw w variants f 13 f13 13 variants f 8 8 8 variants 7 7 7f variants f 6 6 6 variants f 4 4 4 variants 4 4 4f variants f 42 42 42 GOMIS 0 03 3 01 1 n 0 f n0 0 OMNIf 0 1 1 0 0 05 5 s variants fdist 0 VIM 21 variants f t MUNI 1 variants f dist 1 variants distf t0 21 sl IMMO 1 1 OMB 1 MI= 1 OWNS 1 IMMO 1 n 0 f n 0 f 1 1 Maw4 MUMS4 A variants f 1 variants f OM* 1 variants f w 1 ONIIM variants f 1 variants f 1 dist t 0 10 dist wt 11 Table 7 (cont'cl) :1) 1 2 3 Grade 4 5 6 Variants Total n 0 OMO f OWN. 0 OM, f UM/ 1 .1.10 1 n 0 f 01.6 1 IIIMMOD 1 MOM 1 411.10 1 1101 3 3 variants f 0 1 variants fdist 1 variants fdist 1 variantsdist0 2f1 st 1 1 n 0 f n f n f n MOMf n f 4111M 1 MONO 1 t variants Owe f 0 0 0 0 variants f 1 0 11111111110 n 0 f n 0 f n 0 OEMf n f n 0 IMMO f 0 tr t variants f 0 IMO 1 variants f 0 1 ia Table 7 (cont'd) E.+ be 1 2 3 Grade 4 5 6 Variants Total r 7n f7 2n f2 2n f2 n1 f1 3n f3 n 0 f variants a 15 .....f15 f variants f w0 585 _61 variants f w481 2 variants f 0451 2 variants f w326 1 variants f w0390 21 304 = w25370 10 5 SubstitutionsDistortionsOmisTaal s ions 437 140 8 388 89 4 381 67 3 279 44 3 297 87 6 252 49 3 = 2034 476 27 oftargetgrade;Foreach.indicator each each (see fOmissions targetis cell the(i,Footnote entry, m,onnumber orthe are nf). 1,Goldman.Fristoeis indicatedof thisthe variants numbersection, by produced; of thefor subjectstest nulla is sign, rfindicated whoindicates 0. produced by the the relative anumeral variant frequency tofor the the left row of of thephoneme that total target's variants in Each cell entry thc observed variants are listed, with the number of occurrences detailed discussion of the rf figure.) The number of occurrences the columnposition for each of Table 8 Variants Occurring in the Free Speech of Speakers of Mexican .11401.1% American English* I.to 1 2 3 Grade 4 4101.111111111110101 5 Variants Total 5n eNNOf12 21o . 57 =Nam.rf ea"4n we.9f 27o . 34rf war.n3 f3 o18 . rf16 n 1 f3 o rf 8n e/0/024f o63 . 38 rf n 1 f 1 ISM80 .12 4111410.06 rf alman22 wagb 52f 146 awn o . ...mew36rf ..<- m variants 0 12 f dforamarawawmomevariants 0 9f variants 0 f3 variants 0 9.33 are 3f variants 0 24 onof variants 0 woof1 variants 0 52 wow f nit f37 75o .49 rf 13 35 143 .24 7 18 65 . 27 ONO2 6 10 . 60 WOW 16 ONO51 281 .18 M.M11110 n5 f13 68o . 19rf Mal*53Ti 160 642 MOMf o . ? i rf f variants 0 37 Oa*f variantsNINIVIIIIMIIIINONIIIIIIIIIMnf orf nf orf nf0 orf nf 3E. MOMf variants 0 18 f variants 0 f6 01.1111variants 0 orf 51 f variants ..... f variants 0 160 mawf MVID 0 13 i variantsn1 MIOf1 o3.34 frf variantsn2 f2 o10 .20 ONININIOfrf n f 0 o rf variantsn1 f1 8o .12 WM/1Mfrf n f 011,110o 011110rf n MIPf o41 rf variants4n eNNO4f o21 . 19 rf b 0 1 -- m0 1 0 1 0 3 Z1 9

Ns oh 1 w 1

oft 4

1 No 11

Int

IN% Mt

10 II

IPM, a I Pit I. %14 0 st I M4

104 Me

' r 04 .1%I o 14% I 1.11 04% 1 46

054

10 N

0 1

1015 45,.

10 114

Target 4 II 127 a '44 1 41 1: I01 al 441 iv, *1 1 a 1 11g 41+1 smi 441 1: 1 . a 1OIN a 1 a 1 1 I otIt ;IN 1 Ifl o 1 IN 1 tar 1 di a 1 I41 4. I as a to 1.4 1 01 4 05 4t: I %4 CO 01 0% a 1 eft stoi 4.4 I 01 a 1 1 4,co .ga 441a 1 tot 11 nos o$4 1 a.2 0441 Co 1: 01 fts aI. 11a 441a 441 .1 o 1 I %pi 1 0% /4ftfri in 44 a In 44 1 01 3911.1ra. u 144 'CI T SZ

Ow Pie dh IN% gt IV Pr 1,0i PEI IMP V I0 10 .42

Is% 1641 WS 4

113 ola 11

1hs *

10 tr. 0 0 Pit

la 0 IP

04 0+ 141

10 vs

ens N

IP 4.

11 NID 11 IP

"6 1 !I 1 44 1PS' 16% N

Im% e fat 11

In g to. .0 1.0 o N VI

et Ihs+ 1:0

10 P t 0 I N

114. Ow, 1.a - N

10 N

1PP% N et sot 1" tht of P IS N 4 ta 1 4 O'T

rget Ta ia. 129

441 0,1 N N 401 sal 011 ..4 I ta

O case) sa4

01 co. 11

'tI eft

01 01 di %co 0 ce41 ect

at N

eg I 414 CO 441 ItI

01 l 01 In

%41 le4 1 PIS 11 Ght IN1 ehi

01 a g al 10

0 %0 01

441 4.4 144 111 %el

01 a 01

I roe we

INS

a0 ew 1 N IP4 etti S d a 1 N . 01

4.41,1 tr.1 94 I Ico 441

01 01 co lb oh

eftbe ek: %41

01 do O

ite cs 144 I 1 we t'41 tft %.4 I Of 4r: I 441 ct N

16 01 01 ger 0 N 6.4 441 e

44 tt4 108.11,34 130

144 NV% sill 444 4,44 di«4

twiti ft1 !IA

al.0

off 01 ea 11

pis

of 01 46*

"SI r". po 4441 411

*41 ON'S

01 11

%el N

4.4

of of

4.414I 1141 N N I"4

In 4.4 1.4 in .I

al 4(4

I I8.4 044 p.4

01 4.4

I41 4441

$ ben ICI0 s.41 .4 P4 N %II M«4

01 I

A el %HI 0loo

01 : 11 vi

Oi 01 0 0 1 *HI

01 41 vl 1.4 I owl .4

01 o 441 v.41 1. Oi 01

VI 0N %41 oil vl

col in 01 01 NM 0 .4

01 01

VIill vl N 441 P4

of 01 Is o 41 441 to

01 01 « 11 vi vi

01 01 o o 441

01 01

VIilt 6.4 N 441 e4 P4

-4 411 e 0

o.4 ^ Table 8 (cont'd) 1 2 3 Grade 4 5 6 Variants Total n 1 i 1 o 1 1rf n fMOW o oONO 0011.1111, rf ONOn f o MIN.o VONONIMrf OWn NWf o rf MOPn ....o rf n f o rf n f o rf va riants Y IIIf 1 variants y 1f 1 1 18 1 1 3 1 1 5 8 2 1110. 2 11 .18 rf A m nf orf2 . f50 variantsnf4 8 orf .44 OM.f nfvariants orf .34 fWM nfvariants orf nf otif 3 .33 OWf variants 26 .31 OWf nfvariants orf nf o f _.-Arariants 14 21 63 .33 _f ....1.011111111101101variants 0 1 IMMO 0 8 0 1 0 1 0 8 0 Z 0 21 f nfvariants18 23 orf54 . f42 nfvariants18 30 orf81 . 37f nfvariants10 33 orf51 .64 f nfvariants9 orf nf orf 21 92 .23 f ...variants18 63 148 .43 f ...... nfvariants13 orf nf o 43 127 .34 f variants86 213 553 .39 rff wO9 15 71 0w 29 1 0 33 0 21 d0 62 1 0w 42 1 d0w 202 9 1 I ££

UI tn4 PSi 1 a Ia 10 IP 10 0 , 8 P 1 NIKS to - 10 .11

4,11 RIb law a IP% oh

104

RIb

10 tri IP%

12. Ib

iPs a PI+

0 113 N

1mi

10 Its p.

10 1.4

0 Target r C I C')ctime Table 8 (conted) 3 Grade 4 5 6 Variants Total 1 2 8o . 25rf n f 0 mooo rf nen=3 f 3 o15 . 20rf nONO fIMO 0 110o rf IMO 0 o 010.11/11.1rf WOWn OlINDf 0 IMOo rf 4n arm 5f 23o . 22rf m variants 0 2 variants 0 f3 variants 0 5 OM.n6 f7 116 ONO . 06 VIM4 lials 6 75 . 08 1 1 6 . 16 n f 0 ~IMO MOO 1 ewe 4 allib 22 .18 rf 3n f7 o rf n15 25f 298 .08 o rf f variants orf MOf 0.100.11111MONINIIIIIIIMnfvariants orf 01111f variantsnf orf MMf orf nf o variants MDf variants 79 .09 ONO f variants f 09 16 0 6 0 1 0 4 0 7 09 24 1 1 1 o rf ONO. 4n 4f o rf nIMMO3 f3 o11 . 27 rf NOONDn f 0 IONo rf 411Mn f 0 Maeo rf UMWn 411.11f 0 o rf n8 8f o26 . 31rf i variants 0 6.16 ONOf1 variants b 9,44 4f 01.1.111111011.1010.variants b ORMf3 variants flb 7 1f Table 8 (cont,d) 3 Grade 4 5 6 Variants Total 1 2 o3 . 67rf n1 f1 3o . 34rf n MADf INIMo rf MUDn f 0 SIMo rf MUD n f 0 o rf 61110n f 0 o rf 2n f3 6o 50rf rn variants 2 variants b 1 variants b 3 MOP ONOo .111111100 rf ONOn f ONOo rf OMI n f UMWo rf 1d .1If o SUNNISrf WIN n VIM1 o 61.1rf INIMP Ye/ WM*o ri n @MVP f o rf f variants2 06 6 6f 1 0 0 0 variants1 02 3 . 2f67 0 Ivariants3 08 9 .89 8f IMOn1 NMf1 IINDo1 1rf MVO2n 2f 2o 01rf1 MA,n1 f 1 NM.o1 rf1 0111n f 0 *NM o rf nMVO 1 11Mf1 o1 rf1 ONOn1 f1 o rf MVOn6 f6 9o . 67rf r i variants w f1 variants w f2 variants w f1 variants w f1 variants w 4 .25 f1 variants w 6If e. ci Table 8 (cont'd) s.eso Grade Variants Total r n1 OMf1 o 1 rf1 *MO n ...f 0 o rf n1 2f 3o . 67rf NOPn MIDf 0 o rf OMa MUMf 0 o rf n f 0 04WD rf 2n lef3 dm.4o . 01.1M75rf rn variantsn f1 variants w 2f variants w 3f i variants2 .3f orf7 nf orf .42 f variants2 6 11 .54 f n f 0 o rf n f o o . rf variants2 3 16 .19 f variants2n 2f o16 .12 frf variants8n f14 SOo .28 rff dist c 21 dist 0 51 dist 0 21 2 dist 0 261 rn _variantsn3 411.0f NNWorf4 nf orf nf orff1 variants3 3 10 .30 f variants1 1 1 f1 elmn f o Nabo rf variants2n f10 o10 frf 1 variantsn 1 f1 o7 .14 frf variantsn10 f19 o32 . rf59f ' 0 3 dist 8 21 0 1 dist 0 46 0 1 dist 90 93 Table 8 (coned) 1 2 3 Grade 4 5 6 Variants Total f variants5n 8f orf12 nf . f67 variants1 1 orf3 .34 =OMBf variantsnf3 3 ONO41orf .07 011111110 ONOn OMf 0 MOo rf variants2n WOO4f SIMo13 .31 frf 0111M 2n .11104f OWN 15o .27 IMINVO.rf n13 20f 84 .24 o OISONOrf 0 43 dist 1 z f3 4 variants dist 4f variants dist 8 f dist i 1 t I0tz 413 i variants7n f15 o19 .78 frf variantsnf13 29 orf58 .50 i variantsnf13 32 orf56 .57 f variants23nf MOO55 120 .46 =MD orf 0111110.f OMBnfvariants18 orf nf orf 411.11161 103 .59 41111108Mf 4111101 15mriants 47 410111.74 .64 f41111 variants89n 239 430 .56 f 41E10o rff i Z 15 ----- tZ 28 1 c. 32 sc. 54 1 tZ 60 1 tZ 46 1 Zst 235 31 COC d- 0. 0111111111110, C.) Table 8 (cont'd) 1 2 3 Grade 4 5 6 Variants Total 9 9 9o 1rf 6 6 9o . 67rf 7n f10 o rf n1 1I oI 1rf 3 3 7o rf 01.0 n1 1 I 1o rf1 27n 30f o rf m variants c 6f variants f5 variants Z 13 .76 variants Z f1 variants Z .43 3I variants Z f 1 variants 36 .83 dist sd. 1 dist c. 1 s 9 1 dist d$c. 25 21 n3 3f 4o . 75rf n5 f5 8o . 62rf IMOn3 M.4f OMo5 .80 rf nOMIM5 f8 o9 .89 rf 4WDn 1 OMf1 o 1 rf1 n f 0 o rf n17 21f 27o .78 rf f NIIIIIMIIIIIIIINIIIIIIIIMIIIIvariants C WMf3 variants C f3 variants Z 4f variants C MIDf5 variants c. f1 variants s MAWf 1 0s II dist z 21 dist 0ZI 16 21

. t , - - .....S41...+1, .,,,, -40. Table 8 (contod) 44 120 1 2 3 Grade 4 5 6 Variants Total n f 0 o rf n1 f1 o5 .20 rf n IMOf 0 MOWo MIMI*rf n ONOf 0 OMo rf n1 f1 o6 . rf17 WSWn1 NNW f1 o11 .09 rf 01110n3 f3 22.o .14 rf i variants f1 variants O ...... f 1 fINIIIIIIIIIIIINOVAIIIIMvariants 0 MODf 1 101111110!variants 0 3f t m variants1n 2f 8o '.25 f rf variantsn 1 f1 4o .25 frf nf.orfvariants1 e 6 .16 variants8n 39f 113 .34 o rf nf4 o 11 32 ..34 rf a 1 f1 8o .12 rf n16 f55 171 o . 32rf O 2 0 1 ...... ----- t0 f1 0 2910 f variants 0 11 f variants 0 f1 variants d0 4510 f n17 f 14 8 15 11 411100 AMID OIMOM ONO 12 MIND OM 011101151 MOM OMB rf f variants 47 161 .29 orf nf orf nf orfow.f nf orf variants 27 183 .14 f variants 19 135 .80 amf variants 34 143 .24 sobf nfvarIants orf nf orf nf o30 151 .20 f variants 38 148 .27 41f variants77 195 921 .21 f O9 3314 09 23 4 09 18 1 09 33 1 0 30 09 37 1 09 174 21 Table 8 (cont'd) 0Do 1 2 3 Grade 4 5 6 Variants Total 8MOn WMf14 ONOo15 . 93 rf IMOn11 00021f ONO32 .65 9 11 21 .52 1110 10 23 OM 31 .74 Ilia, 15 18 34 .53 rf 11 25 35 .71 64 112 168 .67 i variants f OIONNIMMEMINIIINIMIIIIIMvariants orf nf orf nf orfOM/f nf variants f variants f variants f variantsnf orf nf orf. f variants f t 14 st 20 1 t 11 st 22 1 t 18 dft 22 21 dfst 107 21 variants7 f8 o13 . f61rf variants9n f13 o19 .68 frf nfvariants15 orf nf orf nf orf22 27 .81 f variants9 10 11 .91 f variants10 12 16 .75 f variants3 8f 9o .89 frf variantsn53 73f 95o .77 rff 24 at 36 s 16 t 81 t 10 2 t 71 a 3725 0 2 0a 21 0t 4 1 a0 1 0 a aOt 81

a,:ek;.144.1h,A;0444.;4+41,'44;;4 Table 8 (contld) I 4..I$4ccw Grade Variants Total ..e n14 29f 53o . 66 rf n 12 22f o59 .37 rf n15 47f o66 .71 rf n 13 23f o39 .59 rf n13 f19 27o .70 rf 9n f15 24o .62 rf 76n 155 268 .58 f o rf f variants s 16 f variants t 56f variants sz 2510 f variants s 10 8f variants tz 36f variants t 8f variants 8 63 f dft 10 21 0fzs 415 ft 57 0at 21 ds 226 0a 43 dft 451617 4 20n 153 201 .76 f 31 319 444 .71 d 1 f f 0z 10 - i variants d orf nf orf nf143 f variants f variants31 385 614 .62 f variants31 396 608 .65 f variants32 343 567 .60 f variants29 324 528 .61 f 174 1920variants 2962 .65 orf f d 297 orf nf orf nf372 orf d d 344 d 289 nf orf nf 290 10 22 9 52 d d 1735 0 0 zt0 13 0 t0 53 1 0 34 - - 0at 180 4I .

2.:Onoll Table 8 (contld) ea 0OD 1 2 3 Grade 4 5 6 Variants Total 4n f 11 orf23 .47 nf orf nf orff variants5 5 26 .19 f variants2 2 4 .50 f nfvariants2 orf nf orf nf3 10 .30 f variants6 16 29 .55 f variants3 4 orf16 .25 nf o f variants22 41 108 .38 rff m variants d0 10 1 d0 23 0 2 0 3 d0 15 1 08 31 8d0 2614 1 f o rf f o rf 1 1 o 1 rf1 1 2 o2 rf1 1 1 o 1 rf1 2 3 o3 rf1 n5 7f o7 rf1 variants 0 f1 variants f2 variants b f 1 variants 0b 2f1 variants a0 43 n3 f3 o8 . rf37 n3 f6 o8 .75 rf n7 f9 o17 . 52 rf n5 f7 o18 . rff39 variantsn11 f17 o32 . frf53 nvariants6 f11 o19 . 58 rf variantsn35 f53 102 . 52 o rff m variants b0 f2 1 variants 00 f3 variants f00 f423 variants 0f 43 0f0 836 0f f263 Ob0f 162114 2 Table 8 (contd) at, E- ito.3) 1 2 3 Grade 4 5 6 Variants Total n 1 f1 3 .....34 Mlle 1 MID 1 INNO 1 MINI! 1 ONOn MOf 0 oNM rf OMn f 0 MOo rf 411n3 4WDf3 o3 rf1 OEMn ONOf 0 OMo rf n5 NIMf5 7o .71 =NMIrf tn variantsMO. SIMIIIMMOMMOND orf nf orf . f variants f variants f variants MOPf 1 a 1 3 8 4 1 f variantsn26 222 297 .74 .10f MIND orf nf orf nf orfOINNIMf nf IMP variants29 265 601 .44 MI. Mlle .1~0f variants32 433 691 .62 f variants32 433 784 .55 orf nf orf al...ortn.f.f variants30 438 756 .58 f 32 variants 556 817 .68 f 181 2347variants 3946 .59 orf f a 188 253 a 432 s 424 s 555 $ 2281 dist Od9 1212 28 dist 0s 11 1 0 429 4 0s 1 dist t0 54 0 1 dist t0d9 12 3413 25

, ..000 Table 8 (coated) 1 E4 t0se n 1 o 2 3 Grade 4 5 6 Variants Total v f va7 riants f16 22 . f72rf variants7nf orf nf 16 25 . f64 variants6 9 I 5 orf nf . f60 variants5 6 orf14 nf . f43 variants10 12 44o . f27rf variantsnf7 orf 13 23 . 57 variants42n 72f 143 MOOo . 50 OMMODrf f I 0 1 8 7 f f 8bO 141 $0 26 0f 2 f0 24 0f 39 0f 58 b00f 2642 21 1111111 S I ...., o o sm. 11111M f 0 rf variantsa1 1f 2 . 50 frf nf orf 0 ....af orf 0 0 1 2 orf2 1 .N2a 31 4o . 75 rf 1 1 af orf af variants f variants .. f 2 3 n asmf aMS.o 011111110rf OMn WOf OMB MINIM OM OM 011111o 24 ONO a IOWf n o NM MIDj MMn j i 0 variants2 2 orf2 nf f 1 0 0 orf f 0 rf nf orf 0 2 2f 2o 1rf e dist 5 1 variants dist s 1If Table 8 (cont'd) 1 2 3 Grade 4 5 6 Variants Total WAD f 0 o rf 11111,n f 0 o rf CAW n f 0 rf IMO 0 rf n f 0 a rf n1 f1 3o .33 rf 1 1 =No3o . 33rf variants f1 variants 0 1 br ONO MOW 4WD variantsn1 f1 orf1 nf o f1 variants4 8 10 .80 rf OMvariants4n f6 9o . 67rf OMn f 0 orf variantsnf10 16 orf21 .76 f MOvariantsnf1 1 orf3 .33 f variantsnf20 32 o44 .73 rff w 1 w0 7f1 w0 f51 0 16 0 1 w0 29 3 ONO n1 f1 o 1 rf1 OMn2 2f 2o rf1 SOOn 1 f 1 o1 rf1 IIIIIDn NWf 11111o rf ONOn f 0 o rf Olen f 0 o rf MI*n OMI IMOo rf 1 d variants f variants f variants f 0 variants4 4 4 MINf dr b 1 08 1 3 1 13ibO 1 Table 8 (cont'd) 3 Grade 4 5 6 Variants Total dr nvariants2 f3 o3 rf1 nvariants2 f3 4o .75 frf ...nvariants3 NOM4f 0...o5 .80 ...mafrf ...n f 0 o rf nfvariants1 orf nf orf nf o1 1 f1 0 ONO 010.1111 8variants 11 13 .85 rff 0w 21 w 3 0 31 0 1 0w 56 ~MI .10 OM WI MOO eINWN* o n o n f orf nf orf variants1 1 2 . 50 n f 0 o rf rf rf a rf variants f rf 1 0 1 n 1 f 1 oI 01.110110rf1 n f 0 o rf 0 rf n f o rf uL o rf 0 I 1 1 I fr r variants O 1 variants O 1

S. Table 8 (cont'd) I 2 3 Grade 4 5 6 Variants Total ORMn f 0 n f 0 o rf NOM MIO 0 W1110 SOD 0 WOO 2n f o rf gr r nf orf nf orf variants w2 orf6 nf orf nf. 2f33 0 variants2 w2 6 . 233f n f 0 1 MID2 15 .13 111111111110 MI. 0 IMO 0 OINIMMO IOW2 WOO3 4 ilIMP 0 OMB 13 5 o rf ke A orf nf orf nf variants 0 2f orf nf orf nf variants 0 orf nf orf nf.75 f3 variants 19 .26 f5 *AMMO MAO OEM NOM 011110 0 n f o rf n f o OMn o n WWI WIND MINIM OM IOW ONO MOM NNW OSIMOOM kr r 0 variants1 1 1 f1rf f 0 rf f 0 orf nf orf nf 0 0 o rf variantsn1 f1 o1 rff 1 11"22trimC.) Table 8 (cont'd) 03671. 0 1 2 3 Grade 4 5 6 Variants Total n1 .....f1 o1 ...rf1 n ....f 0 o rf ORMn OMf 0 o rf n f 0 o rf n OWN f 0 o rf n f 0 o rf n1 f1 o1 rf 1 p2 2 011INININVIvariants 0 f 1 variants 0 f1 n f 0 o rf n1 dna,f1 o1 .1.0MMrf1 Maen f 0 o rf .....n f 0 NMo rf 11n f 0 MOo rf n f o rf n f o rf skw a variants 0 algaf 1 . ariants 0 1 11 41 rf1 Mew oMOD rf1 10. rf1 flMee f o rf n f o rf an f o rf n .-f o rf 1 w variants5 06 6 6 variants3 03 3 3 variants3 03 3 3 0 0 0 variants11 O12 12 12 f

O...... 0400,"' Steg&VWISCIPAU4441444411. etill14NrgA S." W 6171

tol to% - I" N

10 10 so a PP.

11 0

10 10 to cr. 0 1"I 1"6

10 10 10

10 10 10 Us Mb 1"I Mb

10 10 10

lob las

10 10 10 0 1"s 1mi

10 ID 10

lab

10 10 0 0 1", 1"s

10 10 10

10 g)

/41 tft 14 .3 1141 N 114* N

0 Pe+

UI et rg Ta 4 40 Table 8 (cont'd) ("4 ta 1 2 3 Grade 4 5 6 Variants Total n 1 f1 1 rf1 rf rf flIVO OMB 1 1 VIM 1 rf1 rf MOD2 4111102 2 rf 1 variants 1 variants dist 1 variants O 1 st rf rf rf rf n rf 3 dist 3 o5 t variants3 O3 5 . 603 0 0 0 0 0 variants . 60f3 3 3 3 rf1 1 f 1 o2 . 50rf 0 rf 0 rf 0 rf f 0 o rf 4 4f o5 .80 rf str variants 3 variants 0 1 variants 4

4,4ie Table 8 (cont'd) 1 2 3 Grade 4 5 6 Variants Total n1 f1 o1 rf1 n f 0 NOMo rf Maen f 0 MIPo rf MIDn f 0 OMo rf 0n SIMf 0 OMo rf MOPn f 0 NNWo rf n 1 Mewf1 o 1 rf 1 OL t variants 9 f1 variants 0 1f variants2n 2f 2 f 1 variants2n 2f orf3 nf orf nf.67 f 0 0 0 ONO 0 OEM variantsnf4 o 4 5 .80 rff 1 orf nf orf nf orf 0 2 w0 1 w0 31 SOWn ONO f 0 OMo rf ISMn1 WOWf1 ONOo1 ~11/0rf1 n1 ONO f1 o1 rf1 NMIn f 0 01111. o IMMOrf nONO 1 ONOf1 WOOo1 rf1 OMn f 0 o rf *MOn3 f3 NOMo3 rf 1 t r variants W ....f1 variants 0 f1 variants 0 OMf1 variants w0 2f1 CA)C :1 to 1 w- Table 8 (cont'd) 1 2 3 Grade 4 5 6 Variants Total 4n f5 o6 .83 rf n5 f5 o6 .83 rf n1 f 1 2o . 50rf n8 f12 o18 .67 rf n7 8f o11 .73 rf n7 8 o9 .89 rf n32 f39 o52 .75 rf Or variants dist 23 variants t f5 variants t f1 variants t 12 f variants t 8f variants t 8f 0.11101MOIMMI11111.0variants dist t 37f 2 OMBn1 VINOf1 SIM2 III OM 1 MOM 1 1 1 OMID 0 11/1M 0 MUD 0 0 2 2 3 . 67rf variants4111100101111011 . I50 variants orf nf orf nf orf nf o variants orf nf orf nf orf nf 1 O 1 dist 1 dist 0 1 SubstitutionsOmissionsTotals 555760188 948747194 7 10611193 132 1169 989178 2 1274 949313 12 12321049 179 4 ANO 657653501184 42 Distortions*For each cell n producedoccurrencesisof the variant number is andindicated of the children number by the in ofthef value.occurrences column grade who or the target phoneme17 in the speech samples from that Relative frequency (rf) is obtained byof each is noted. 0 produced a gradevariant is of indicated the row bytarget the o value, dividing f by o. In the lower part and the numberphoneme. of variations The number of of each cell, the type ac.41:050.4k.:440...,strAptaiabithifteiaii-40 klIziae.444444::Xi,liii-;::r9 1?5, fr Summary of Frequencies and FeatureProduced Changes by Speakersfor Those of Variants MAE with a Frequency Greater than Five Table 9 on the Goldman-Fristoe Test Target* Position Over-all Frequency of Variation of Target39 IndividualVariants $ Frequency of IndividualVariants 31 +/- FeaturesChange-voi in FeaturesinChange M/UU voi (214) i 146 7329 bss 142.754718 -cnt-str-voi MU Ucor voicnt voi (213) v m 52 b 1419 +dist-cnt-str ** MUM Ucor cntdist str 88 omit f 7536 +dist-voi MU dist voistr Target Position Over-all Frequency of Variation Table 9 (cont'd) IndividualVariants Frequency of Individual +/- FeaturesChange in inChange M/U of Target 61 5 Variants 54. 5 2 +ant+cnt Features UUM ant corcnt (209) in 66 59 4 -str-str a.ritcnt UUM ant corstrcnt ird(`'CTJ f 82 tN 79 1 +ant+cnt-str UMU antcorcnt str (184) i 437962 C 427557 -cnt-cnt UU cnt cnt Table 9 (conttd) fr v Target Position i Over-all Frequency of Variationof Target 80 IndividualVariants Frequency of IndividualVariants 72 +1- FeaturesChange in FeaturesinChange WU m 52 t$ 23 +str-cnt U cntcorstr I d 11 +voi-cnt U cntcorstr 4.4f--* (163) 8 omit 11 5 -cnt UM corvoi f 31 ft 11 +str-COI' UU corstr cnt ts 10 8 +str-cnt U cntcorstr Table 9 (cont'd) Target Position Over-all Frequency of Variation of Target IndividualVariants Frequency ofIndividualVariants +/- Features Chang e in FeaturesinChange IOU is 104 d 103 -cnt U cntcorstr (113) (89) 89 9 omit t 82 8 -voi-cnt UU voiU cor; str U cnt 211 6114 d 52 7 +ant-voi-str UU voi antcorstr (75)(64) 3 64 kI 58 5 -voi+cnt-voi UMU voicnt voi (52. 33) (62) A 6252. 33 omit p 4958 -voi U voi Target Position Over-all Frequency of Variation Table 9 (cont'd) Individual Frequency ofIndividual Change in Change (21.5) of Target 15 Variants w Variants 12. 5 +1- Features -ant-voc-cns Featuresin-cns M/U CP (20. 05) 5. 25 omit 5. 25 -lat-cor (8. 5) (8) 14. 8088. 5 omit 14. 74 7.8. 85 thisjustedtotalonly*Target section, frequency thosefrequencies variantsfor associateda detailed if arethe reportedtarget discussionwith appearedeach which target of had themore appears adjusteda frequency than in frequency parentheses of greater figure). below than five. the phonemes are rank ordered according to frequency of variants associate with each. The once in the Goldman-Fristoe test (see Footnote 2, targetAll symbol; frequencies however, are ad- which**Thefeaturetinguish is "distributed"[-Fdist] any -phonemes (such refers as ofdist] /13/) English to thehas length afrom longer each of constrictionthe other, constriction thus than is not associateddoesfeature with which distinguishes /13/ from /v/ is "distributed" (dist), a distinctive feature of English. a sound (such as /v/) which is a feature which does not dis- a fricative -- a sound The 158 Table 10 Relationships Among Goldman-Fristoe (G-F) and Free Speech (F-S)

Variants in MAE Grou

Produced Variant on Produced Produced G-F, but not on F-S Variant Variant Variant on both on F-S, G-F and F-Sbut not on G-F n**N/ V*** N/dig*"

/t/ for /0/ 104 10 32 30 32 (1) .10 .31 .29 .31

/t/ for /0/ 29 0 3 1 25 (m) 0 .10 .03 .86

/t/ for /9/ 37 5 1 3 28 (±) .14 .03 .08 .76

/s/ for /0/ 4 1 1 0 2 (i) .25 .25 0 .50

/8/ for /0/ 39 5 11 6 17 (m) .13 .28 .15 .44

/s/ for /9/ 41 1 3 6 31 (f) .02 .07 .15 .76

/f/ for /0/ 6 3 1 (1) .50 .17 0 .33

/f/ for /EV 12 2 9 1 (in) 17 .75 .08

/f/ for /0/ 17 4 2 5 6 (f) .24 .12 .29 .35 159 Table 10 (cont'd)

Produced Variant on Produced Produced G-F,. but not on F-S Variant Variant Variant on both on F-S, G-F and F-Sbut not on G-F n N/V N/O

/d/ for /i /177 4 0 98 75 (i) .02 0 .55 .42

/f/ for /v/ 2 0 2 0 0 (i) 0 1.00 0 0

/1/ for /v/ 16 0 0 0 16 (m) 0 0 0 1.00

/f/ for /v/ 85 29 27 19 10 (f) .34 .32 .22 .12

/13/ + /b/ for /v/ 67 4 58 4 (1) .06 .87 .06

/0/ + /b/ for /v/ 57 22 23 5 7 (m) .39 .40 .09 .12

/13/ + /b/ for /v/ 8 4 1 2 (f) .50 .13 .25

/s/ for /z/ 33 1 31 0 1 (i) .03 .94 0 .03

/s/ for /z/ 22 3 14 1 4 (m) .14 .64 .05 .18 160 Table 10 (cont'd)

Produced Variant on Produced Produced G-F, but not on F-S Variant Variant Variant on both on F-S, G-F and F-Sbut not on G-F N/V N/0

/s/ for /z/ 183 4 0 175 4 (f) .02 0 .96 .02

/g/ for /a/ 84 8 51 19 6 (i) .10 .61 .23 .07

/i/ for /a/ 71 2 48 9 12 (m) .03 .68 .13 .17

/i/ for /a/ 95 7 34 38 16 (f) .07 .36 .40 .17

/a/ for /g/ 113 16 10 46 41 (i) .14 .09 .41 .36

/a/ for /I/ 54 15 16 11 12 (m) .28 .30 .20 .22

/C/ for /i/ 58 14 29 12 3 (f) .24 .50 .21 .05

/a/ for /I/ 54 3 47 2 2 (f) .06 .87 .04 .04

/p/ for /b/ 62 4 47 6 5 (f) .07 .76 .10 .08

/t/ for /d/ 119 19 10 46 44 (f) .16 .08 .39 .37 161 Table 10 (cont'd)

Produced Variant on Produced Produced G-F, but not on F-S Variant Variant Variant on both on F-S, G-F and F-Sbut not on G-F N/V N/O

/k/ for ig/ 70 13 29 15 13 (f) .19 .41 .21 .19

Omit /2 / 97 8 0 78 11 (f) .08 0 .80 .11

*Substitution variants are indicated by noting that one phoneme was produced 'for' another.Position of the variant is indicated in parentheses below the specification of the variant.Thus, the first row indicates the substitution of /p/ for /b/ in final position. ** n indicates the number of subjects who produced the indicated variant on either Goldman-Fristoe or in Free Speech. ***NIV indicate the number (and proportion) of subjects who produced the target phoneme in Free Speech, but did not produce the indicated variant. ****N/0 indicates the number and proportion of subjects whose Free Speech sample did not include the target phoneme (thus, there was, for these subjects, no opportunity to produce the variant). 162 the subjects in the caseof the /t/ for /0/ substitutionin medial and final positions, the /s/ for/0/ substitutions in initial andfinal positions, and the /1/ for /v/substitution in medial position.Thus, every prediction error of a 50% magnitude orgreater involved either the/0/ or /v/ target phonemes. By contrast,recall that six targets recordedthis magnitude of misprediction in theNiagara Falls sample. (SeeSection 4.1.3. ) 4.2.4.The frequency distributionsof individual variants across grades.It was impossible todiscover for the Mexican Americansubjects clearly describable dialect vs.developmental patterns of frequencydis- tributions across grades as wasrevealed in the Niagara Falls sample. Since there seem to be noempirical or linguistic groundsfor grouping any of the targetphonemes together, they aredisplayed in individual graphs in Figures6 through 16.The II target phonemes (see Table 9) which most frequentlyproduced variants are representedin the graphs. Chi Square tests wereused to ascertain the degree towhich each frequency distribution acrossgrades differed from chance.No142 is reported which has a probabilityvalue greater than .10. Whereappropriate the ordinate represents adjustedfrequencies (see footnote 2).The information contained in the figuresis the following: Figure 6: This graph showsthe distribution of all variants(in all positions) of the target phoneme/0/. Variants of /0/ in generaldecrease across gradelevels. 'X2 for this distribution = 10.1, p = .07. Figure 7:This graph shows the distributionof the three most common variantsof the /0/ phoneme -- thesubstitutionoi /t/, /8/, and /f/.The frequency distributionof the /t/ for /0/ substitution, which does not differ from chance,is relatively flat except for adip at the sixth grade. The/s/ for /0/ distribution has ale.4 of 20.4, p = .002, and the /f/for /0/ substitution has a1.2= 11.8, p =.04. Even though the statistics show thelatter two substitutions to bedistributed differently from what would beexpected by chance, the patternof the distributions do not seem to indicate anyoverall trends. Figure 8:In this figure the distributionof the Id/ for /I / substitution is displayed for the twopositions in which /1 / occursin English, initial and medial. Chi Square testsshow that neither of thesedistributions differ significantly from chance,but inspection of the graphsreveal a definite decreasing trend in theincidence of the /d/ for/i /substitution in initial position across grades. 163 Figure 6 Incidence of Variants Associated with the Target Phoneme /0/ by Grade for the MAE Group

211 26 24 l- 220-.

20 Pm.

16 14

1 1 2 3 4 1 6 GRADS 164 Figure 7 Incidence of /t/,/s/, and /f/ Substitutions for/e/ by Grade for MAE Group

v. 28 - sere....It/FOR/el .46.11. 4/FORha 26 If/fOR / 24 221-- 20 ,-- 18 - 16 - 14 -- 12 -

1 4. 6 - 4 - 2 - 0

1 2 3 4 6 GRADE 165 Figure 8 Incidence of /d/ for /4 / Substitutions in Initial and Medial Positions by Grade for MAE Group

.0.41 INITIAL

....gip... MEDIAL

Air&

4111111111Mmt Li. 4 i I 1 2 3 4 3 6 GRADE

1C7 166 Figure 9:This figure presents theincidence of the omissionof final /A/ across grades. A1G2 of 17.5,p = .004,partially reflects the erratic increase in omissionsin the fifth grade group.There is, however, a generaldecrease across the othergrades. Figure 10: The frequencyof devoicing of final/d/ (It/ for /d/ substitution) is shown in this figure.The X` of 9.9, p = .08,reflects an erratic U-shaped distribution. Figure 11: The devoicingof final /g/ (Ik/ for/g/ substitution) shows a completely different patternthan the U-shaped distributionof /t/ for /d/.There is a definite overalldecrease across grades.142 for this distribution = 10.3, p = . 07. Figure 12: The devoicingof final /b/ (/p/ for/12/substitution) shows a pattern similar tothat of the devoicing of/d/ in that the increase in fifth grade gives the curve aslight U-shaped appearance.It is also, however, similar to the distributionof final /g/ devoicing inthat except for the increase in fifth grade,there is a general decreasein frequency of devoicing. Figure 13: This graph presentsdevoicing of final /j /,the /Z/ for /j /substitution. Except for alow frequency in firstgrade, incidence of this variant generally decreasesacross grades.The%2 of 9.4, p = .09 reflects both the low first gradefrequency and the overalltrend. Figure 14: In this figurethe devoicing of /z/is graphed, both in final position and in combinedinitial and medial positions.(Recall that all frequencies are adjusted so thatthey are comparable to eachother.) The devoicing of final /z/is by far the mostfrequent variant found in theSan Antonio sample.Thele2 of 9.0, p =10, probably reflectsthe increase in frequency associated with themiddle grades. Althoughthere is a decreasing trend in the later grades,note that the frequencyremains higher than in first grade. The oppositeis true of the devoicing of/z/ in initial and medial positions.Thel..2 of 9.8, p = .08, for this distributionreflects a general dip in frequencies inthe middle grades. Whilethere is an increase in the later grades, the finalsixth grade frequency islower than that of the first grade. Figure 15: This figure presentsvariants associated with/v/, the second most frequently variedtarget phoneme in SanAntonio. Two types of variants of /v/ weretypically produced. In finalposition, the familiar devoicing phenomenon wasobserved, resulting in /f/ for/v/ substitutions.Except for an increase insecond grade this distributionis almost flat, and, in fact, does notdiffer from a chance distribution.The variant of /v/ observed in initialand medial positions, is thesubstitution of 167 Figure 9 Incidence of Omission of Final /A/ by Grade for MAE Group

28 -- 26 240- 22 20

1$0... 16, 14

12 10.-- $ 46 2 0 I 5 6 1 2 $ 4 GRADE 168 Figure 10 Incidence of /t/for /d/ Substitution in Final Position by Grade in MAE Group

20 26 24 22 20

16 14

12 10 8 6 4 2 0

1 2 3 4 5 6 GRADE 169 Figure 11 .4 Incidence of /k/ for /g/ Substitution in Final Position by Grade for MAE Group

20 26 24 22 20

1 0 6 4

2.. 0

1 2 3 4 3 6 01111011 171 170 Figure 12 Incidence of /p/ for /b/ Substitution in Final Position by Grade in MAE Group

28 26 24 22 20

16 14

12 10

6 4

1 6 GRADE 171 Figure 13 Incidence of /C/ for /f/ Substitutionin Final Position by Grade for MAE Group

28 26 * 24 22 20

16 14

12 10

2 0

4111111MIUMINP 111111111Prft.MIMMINIIIMIIIIMMIUMIIL LIMPOIMUMMOIJ 2 3 4 5 6 GRADE 172 Figure 14 Incidence of /s/ for /z/ Substitution Initial + MedialPosition and in Final Position by Grade for MAE Group

,;11

1.1

28 1, 26 24 22 20

16 14

10 8 6 4

2 0

6 1 2 3 4 5 GRADE 173 Figure 15 Incidence of /f/for /v/ (Final Position) and /b/ (or /On for /v/ (Initial and Medial Positions) Substitutions by Grade for MAE Group

28 26 24 22 20

18 s 16 V w a3 14 as as 12 0. 10 8 6 4

2 0

1 2 3 4 5 6 GRADS 174 either /b/ or /13/(a voiced bilabial fricative which occurs inSpanish, but not in English).The subjects sometimes produced full stops(/b/) and perceptible fricatives (/13/), but many of the substitutedsounds had qualities of both.The result was that coding (by native speakersof English) was rather arbitrary.Therefore, /b/ and /13/ substitutions have been added together for presentation here.Substitutions of /b/ and /0/ for /v/ in initial and medial positions show a definite decreasing trend across grades.This trend is reflected in0.2 of 16. 6,p = . 006. Fijure 16: The almost identical overall frequency of/E/ for /1/ and /I/ for /a/ substitutions (in all positions) is shown inthis figure. Statistical tests reveal, however, that while the distributionof /E/ for /I/ substitutions does not differ from chance, the /§/ for/C/ distribution does((.2= 14. 4, p = .01).This difference is manifested by a pattern of general decrease in frequency across grades.Thus, while the overall frequencies of these two common substitutions are similar, they seemto have differential patterns across grades. In the next section the linguistic and socio-linguisticimplications of results reported above will be discussed. 175 Figure 16 Incidence of /Z/ for /i/ and /i/ for /e7 Substitutions in All Positions by Grade for MAE Group

29 26 24 22 20

111

16 14

11 6 4 2 0

1 2 3 4 5 6 GRADE SECTION 5

DISCUSSION

The purpose of this section is three-fold.First, predictions of variants derived from an earlier study (Williams,et al.,1971) and from published literature on BE and MAE will be evaluatedin the light of the present findings.Second, the implications of the present results for sociolinguistics, or dialect study, will be discussed.Third, specula- tions about the relationship between these data and generallinguistic theory (as presented in Appendix A) will be presented. 5.1.Evaluations of Empirical Predictions 5.1.1.Substitutions occurring in SAE, BE, and MAE speech samples. The types of substitutions found in the three speechgroups involved in this two-part studyare compared in Table 11.Fourteen substitutions were found to becommon to all three groups -- SAE, MAE, and BE. Another five substitutionsare unique to the MAE and BE groups. One substitution.is uniqueto the SAE group and another is unique (with one exception) to the MAEgroup. There are no substitutions unique to the BE group. With one exception (the substitution of /b/ for /f/) all substitutions which appeared in the SAE sample also appearedin both the BE and the MAE samples. However, three of thesubstitutions which occurred with a relatively high frequency in the SAEand MAE groups (/s/ for /0/, /d/ for /j /, and /i/ for /Z/)were rare among the BE speakers.The following five substitutionswere found to be unique to the BE and MAE groups (i. e.,these five substitutions did notoccur at all in Marshalltown): four devoicing substitutions--/k/ for /g/, /f/for /v/,/t/for /d/, and /p/ for /b/-- and one other /d/ for /0/. One very low frequency substitution,/5/ for /j /, occurred in the SAE and MAE samples, but not in the BEsample. Because /ii/ for/j / occurred onlyonce in the Marshalltown data it was not considered a stable substitution in Williams,et al. (1971).Thus, it is almost unique to the MAEgroup. With two exceptions, then, (na/ for /f/ and /ii/for /j /), Friedlander's (1965) findingsare replicated. He reported that BE speakers produce all the variants produced bySAE speakers and MAE speakers produce all the variants producedby the BE speakers.This state of affairs means that thereare no variants unique to BE.It would

177 178 Table 11 Comparison of Substitutions Occurring in SAE, MAE, and BE Groups

Substitutions Occurring in All Three Groups

/s/ for /z/ /6/ for /3/ /8/ for /A/ /d/ for/7j/ /f/ for /8/ /d/ for /1; / /t/ for /9/ /b/ for /v/ It/ for /6/ /13/for /p/ /6/ for /6/ /w/ for /r/ /6/ for /6/ /w/ for /2/ Substitutions Occurring in both MAE and BE Groups, but not in SAE Group /d/ for /9/ /f/ for /v/ /k/ for /g/ /t/ for /d/ /p/ for /13/ Substitution Occurring only in SAE Group /b/ for /f/ Substitution Occurring only in MAEGroup* /6/ for fp Substitution Occurring only in BE Group

None

There is one exception to this statement. One such substitution was recorded in the SAE group. , 179 also mean that there are no variants unique to SAE, either, but one of the two exceptions (/b/ for /f/) does appear to be unique to the SAE sample. Only one substitution (/i/ for /j/) emerges as unique to the MAE sample (although it does occur once in the SAE group).There are, however, five substitutions which are unique to the non-SAE groups. These variants will be discussed further below. 5. l 2.Predictions derived from published literature on BE and MAE. Section 2 of this report presented relevant findings from a number of recent studies of BE and MAE. These findings were then used as a basis on which to expect certain phonetic variants in the BE and MAE samples. Before the outcome of these expectations is discussed, it should be pointed out that within BE and MAE there are "sub-dialects", corresponding to geographical location and, in the case of MAE speakers, to the dialect of Spanish which is their native language.Thus, findings based on studies of individual geographical areas should not be expected to yield perfect prediction for other geographical areas. 5. 1. 2, 1.Predictions for BE group. The omission and devoicing of final consonants, especially stops, was correctly predicted for the BE group.In BE there are two sources of word-final omission.First, there is the omission of the last segment of a word-final cluster in which both segments of the cluster share the same voicing value.Thus, "hand" (in which both segments of the final cluster are voiced) is pronounced in BE as "han" and "desk" (in which both segments of the final cluster are unvoiced) is pronounced "des".Unfortunately, there are no such final clusters on the Goldman-Fristoe test.Therefore, the Goldman-Fristoe data for word-final omission of stops does not include omissions in this phonological environment. (Omissions of this type were observed frequently in the Free Speech samples of the BE speakers.) Omission of post-vocalic consonants has been emphasized much less in the literature, but is mentioned in Section 2 as a possible characteristic of BE. All omissions of word-final stops reported from the Goldman-Fristoe test are, in fact, post vocalic.Thus, this expectation is borne out by the data. Had final clusters been present in the Goldman-Fristoe test, however, the omission of final stops would undoubtedly have been much higher.It is important to note, with reference to Figure 1, that the final stop omissions included in that developmental pattern are post- vocalic stops.It is quite possible that the omission of stops in final clusters would have been found to follow a dialect pattern. Omispion of /4- / and / 2/ were found, as anticipated.These omissions were particularly frequent in final position, in which they followed a dialect pattern (see Figures 2 and 3).The substitution of /f/for /A/ occurred frequently, especially in medial and final positions, as predicted.The substitution of /d/ for A / was also prevalent in the 180 Niagara Falls data.The only predicted variant which did not occur in the Goldman-Fristoe data was the substitution of /v/ for/t /in medial position.This substitution did occur, however, five times in the Free Speech sample.

5. 1. 2. 2.Predictions for MAE group.Turning to the MAE sample, the predicted substitution of /b/ and /0/ for/v/ occurred with great frequency, but the substitution of /v/ for /b/ was almostnon-existent. The articulatory confusion of /Z/ and/5/ and the substitution of /s/ for /z/ occurred as predicted.There were some /i/ and /y/ substitutions for /A as predicted, but /I/ for /j/ appeared in allpositions, not just in medial position.The /y/ substitution occurred in medial position,rather than in initial position, as predicted. No variants of/h/ were found, though the substitution of /x/ (an unvoiced velarfricative) was predicted. Neither did the coders misperceive an unaspiratedinitial voiceless stop as a voiced one. No substitutionof a voiced for a voiceless initial segment was recorded. The devoicing of final obstruents which wasprevalent in the MAE sample was not predicted by the literature reviewed inSection 2; neither was the substitution of/5/ for /j/.This latter substitution, while not frequent, is interesting because it is the only substitutionunique to the MAE sample.

5. 2.Sociolinguistic Implications Hitherto in this report we have been referring to thosephonernena characteristic of BE and/or MAE as "dialect" phenomena.To be more precise it is necessary to distinguish betweenthose phenomena which are the result of phonetic interference from a primarylanguage and those which are truly dialect phenomena, in which case thechildren are pre-. sumably mono-lingual. The data concerning obstruentdevoicing, which is characteristic of both the Black andMexican-American sample, illustrate this distinction most clearly.As mentioned above, previous authors have noted that word-final obstruent devoicing ischaracteristic of Black dialect.The present data show that the incidence ofword-final obstruent devoicing increases with grade level in the NiagaraFalls sample for all word-final obstruents. On the other hand, the frequency patternof obstruent devoicing across grades is erratic in the SanAntonio group. The incidence of some devoicing substitutions seem todecrease with grade level (see Figure II), others are erratic (see Figure13), and still others show a U-shaped distribution (see Figure10). This patterning difference across grades is indicativeof the distinction between dialect and interference phenomena.Data from the BE group show that the children areacquiring a characteristic of their dialect 181 as they age.Presumably the correct linguistic description of this state of affairs is that this dialect containsa phonological rule which devoices word-final obstruents. The increasing incidence ofthe substitution of voiceless for voiced obstruents in word-finalposition, then, would reflect the individual child's acquisition of this rule of hisdialect. It should be pointed out thata word-final devoicing rule is very common among languages of the world.It occurs in a number of other Indo-European languages (e.g. , German and Russian), as well as in unrelated languages.This rule may be a reflection ofa tendency noted by some authors (Greenberg, 1969) toward weakening ofsegments in final position.In any case, it is entirely natural fora dialect to develop such a rule. On the other hand, devoicing phenomena in the MAE sample, since they do not show a consistent increase in frequencyover time do not seem to be the result of rule acquisition. We suggest, instead, that it may be due to the child's uncertaintyas to the appropriate voicing values associated with particular lexical items.This is a plausible hypothesis in view .A the fact that thereare phonetic reasons why the perception of the voiced-voiceless distinction in Englishmay be difficult for speakers of Spanish.Although this distinction exists in both Spanish and English, the cues for the perception of voiced and voiceless obstruents differ in the two languages.This is true because the phonetic correlates of voicing in English and Spanish differ from each other.Thus, for example, in English, vowels which precede voiced obstruentsare typically longer and more tense, whereas in Spanish all vowelstre tense. Furthermore, in English, voiceless phonemesare aspirated in some environments, whereas this is not the case in Spanish.The above is by no means a complete description of the differences between the English and Spanish manifestations of the voicedvs. voiceless distinction in obstruents. Suffice it to say, however, that the phonetic rules for realizing this distinction are considerably different in the two languages. If the Spanish-speaking child were to apply the perceptualcues appropriate for Spanish when attempting to distinguish between English voiced and voiceless obstruents, he would make many perceptual misjudgments. Thus, the random pattern of obstruent devoicing in MAE may be due to the speaker's uncertainty as to the appropriate phonetic value for the word being produced. Another example of variants not deriving from rule differences, but, instead, from phonetic interference from the native language is the confusion between /5/ and /C/ in the MAE group. In all positions the substitution of /U for /5/ as well as of /5/ for /C/ is very common. Spanish does not distinguish between /5/ and /Z/-- in fact,/5/ does not occur in native Spanish words. As the Spanish-speaking child learns 182 English, he learns to produce the foreign phoneme /g/ Furthermore, he learns that the /g/ - /eV distinction exists in English. Since Spanish remains the child's native language, however, he may never acquire the phonetic skills necessary for consistently correct application of the distinction. From the point of view of the MAE speaker, the English speaker is perceived as indiscriminantly producing sometimes /5/, sometimes /a/. Since there are no rules to govern the distribution of these sounds, the MAE speaker may also indiscriminately substitute one for the other. When he substitutes an /g/ for a /a/ we have an example of over- correction. Figure 16 illustrates that as the child exposure to English increases the incidence of over-correction decreases, although the substitution of /a/ for /g/ (which may be interpreted as a lapse into the Spanish phonological pattern) decreases far less.The frequency levels of both these substitutions remain high, however, and it is known that these substitutions are extremely common in MAE spoken by adults. The substitution of /5/ for ii (which only occurs in final position) is an example of the same mechanism. The obstruentvdevoicing which is common in MAE results in the interpretation of /j/ as a/a/ and, thus, subject to the same alternation with /g/ as is /6/.

5. 3.Implications for Linsuistic Theory A major goal of the pair of studies consisting of the presentstudy and Williams, et al. (1971) has been to relate facts about thearticulatory performance of selected groups of children to general linguistic theory. In this section aspects of the present data which appear relevent to such an interpretation will bediscussed. While the pattern of substitutions for /0/ in the MAE and BE data is in general consistent with the interpretation presented inWilliams, et al. (1971, page 70), the BE data of the present studyis perhaps most revealing on this point.The substitution of /f/ for /0/ is by far the most common variant in the entire BE sample. This isconsistent with the claim made in Williams, et al. (1971) that the features ofnon-stridency and continuance dominate the more vulnerable feature [coronal].Thus, the features [-strident] and [+continuant] are maintained in the /f/ for/0/ substitution, while the value of the feature [coronal] changes from + to This feature change is also consistent with the general hypothesisderived from markedness theory (see Appendix A) that as featureschange to effect substitutions, there is a tendency for the change from their marked to their unmarked value.In the /f/ for /0/ substitution, the vulnerable feature [coronal] changes from its marked to itsunmarked value. 183 The behavior of /0/in MAE does not fit this patternas neatly as does the /0/ in the BE group.In the MAE data, in particular, the frequent substitution of /s/ and /t/for /0/ can be largely accounted for by phonetic interference.Similarly, the substitution of /s/ for /z/in both sets of data is consistent with the hypothesis put forthto explain the data in Williams, et al. (1971), namely, that thereis a strong tendency toward preserving the stridency feature valueat the expense of features which might attenuate the perceptibility of stridency. Thus, voicing, by reducing intraoral airpressure, detracts from the noisiness of a fricative, making it less strident than its voicelesscount- erpart. Another conclusion from the Williams, et al. (1971) study which is substantiated in the present study is that the class ofsubstitutors is distinct from the class of substitutees. There isone apparent class of exceptions to this, however: the voiced obstruents /4/, and /b/, which are common substitutors,appear in this study to be sometimes substituted for by their voiceless counterparts (in final position). How- ever, we have suggested above that this voiced/voiceless alternation in BE is a result of the acquisition ofa regular phonological rule and in the case of MAE the result of phonetic interference. Sincein both cases we have independent evidence for arguing that the voiced/voiceless alternation is not a result of immaturity of the articulatoryapparatus, the general conclusion that the substitutorsare drawn from the class of sounds which does not include the substitutees remains valid forsubsti- tutions having a developmental origin.

Examination of the present data in the light ofa theory of phonology which incorporates the ideas of markedness (see Appendix A)reveals that almost all substitutions not accounted for by first language interference are effected by the change of a feature from its marked to its unmarked value. One apparent counter-example to this is the substitution of /b/ for /v/1 which entails the change of the feature [coronal] from Uto M. Note that this does not entail the change of the +/- value of[corona. t is considered that bilabial, non-strident fricativesare unmarked for [coronal], whereas bilabial stopsare marked for [coronal]. We might qualify the above statement, then, by stating thatevery change in the +1- value of a feature is always in the direction ofthe features becoming

41101111001MO 1 The /b/ for /v/ substitution in the MAEgroup would not be a counter-example to the markedness claim because in thatgroup the substitution can be accounted for by phonetic interference.It is only among the BE and SAE groups that the /b/ for /v/ substitutionconstitutes an apparent counter-example to the markedness prediction. 184 unmarked. The only counter-example to thismodified claim is the substitution of /b/ for /p/ in initial position.It should be noted that this occurs only in the word "pajamas" in theGoldman-Fristoe test.In this word the initial segment precedes an unstressed vowelwhich, in turn, precedes a voiced obstruent in a stressed syllable.It is likely that in the production of this word the phonetic phenomenon known asco-articulation (see Ohman, 1966 and Perkell, 1969) is taking place. On this view,the speaker wculd be setting his larynx for phonation at the beginning of the wordin antici;ation of what is required for the production of the stressed syllable.Had the test included a word such as "potato" or "palace" we suspect thatthe initial /b/ for /p/ substitution would not have occurred. In summary, it appears that markednesstheory is useful in explaining some of the substitutions which occur in MAEand BE.It is important to note, however, that markedness theoryis of no utility in explaining the substitutions which result from phonetic interference.It appears that the clearest example of dialectal variation, namelyword-final obstruent devoicing, is compatible with markedness theory becausevoicing is marked in obstruents.This compatibility, however, may be illusory,since not all language change can be explained by such astraightforward application of markedness theory.Therefore, it appears that three basic factors can explain the data in this study: (1) Phonetic interference;(2) Dialect variation; and (3) Maturational factors which seem tobe consistent with markedness theory. REFERENCES

Cairns, C. E.Markedness, neutralization, and universal redundancy rules.Language, 1969, 45, 863-885. Chomsky, N. and Halle, M. The sound pattern of English. New York: Harper & Row, 1968. Fasold, R. W.Distinctive linguistic characteristics of Black English. In J. E. Alatis (Ed.) Monograph series on lanjuages & tics, 22, Twentieth Annual Georgetown Round Table, 1969. 1 1 Faso ld, R. W. and Wolfram, W. Some linguistic features of Negro i . dialect.In R. W. Faso ld and R. W. Shuy (Eds. )Teaching i , , standard English in the inner city.Washington, D. C.: , Center for Applied Linguistics, 1970. Friedlander, G. H.Report on the articulatcry nd intelligibility status of socially disadvantaged pre-school children.Publication of U.S. Office of Health, Education, and Welfare: ERIC-ED- 014321, 1965. Greenberg, J. H. Some methods of dynamic comparison in linguistics. In J. Puhvel (Ed.) Substance and structure of language. Berkeley and Los Angeles: University of California Press, 1969. Halle, M. Phonology in a generative grammar. Word, XVIII, 1962, 54-72.Reprinted in Fodor, J. and Katz, J. (Eds. ), The structure of language: Readings in the philosophy of language. Englewood Cliffs: Prentice Hall, 1964. Labov, W. D.The social stratification of English in New York City. Washington, D. C.: Center for Applied Linguistics, 1966. Labov, W. D.The stud_y of nonstandard English. Champaign, Illinois: National Council of Teachers of English, 1970. Natalicio, D. S. and Williams, F.Repetition as an oral language assessment technique.Austin, Texas: Center for Communi- cation Research, 1971. 6hman, S. E. G.Coarticulation in VCV utterances: Spectrographic measurements. Journal of the Acoustic Society of America, 1966. 185

.6 6 186

Perkell, J. S.Physiology of speechyroduction: Results and implications of a quantitative cineradiographic study.Cambridge, Mass.: M. I. T. Press, 1969.

Reed, D. W.,Lado, R.,and Shen, Y. The importance of the native language in foreign language learning.Journal of Language Learning,1948, 1, 17 -23. Stewart, W. A.Urban Negro speech: Sociolinguistic factors affecting English teaching.In R. W. Shuy (Ed.), Social dialects and langu age learning. Champaign, Ill.:National Council of Teachers of English, 1964. Stockwell, R. and Bowen, J. D.The sounds of English and Spanish. Chicago: University of Chicago Press,1961. Weinreich, U. On the description of phonic interference.Word, 1557, 13, 1-11. Williams, F.Language and povern. Chicago: Markham Publishing Co.,1970. Williams, F. (Ed.), Cairns, H. S. , Cairns, C. E. , and Blosser, D. F. Analysis of production errors in the phonetic performance of school-age standard-English-speaking children.Austin, Texas: Center for Communication Research, 1971. Wolfram, W. A. A sociolinguistic description of Detroit Negro speech. Washington, D. C.: Center for Applied Linguistics, 1969. Wolfram, W. A.Social dialects from a linguistic perspective: Assump- tions, current research, and future directions.In R. W. Shuy, I. Feigenbaum, & A. Grognet (Eds. ), Sociolinguistic theory, materials and trainiz21ree related studies.mWash- ington, D. C.: Center for Applied Linguistics, 1970. APPENDIX A

Appendix A of the present report consists of a reproduction of Section 2, "Linguistic "erspectives," from Williams, et al. (1971),. the report of the initial study of this series of two studies. Notice that the tables of this A- -endix (since it was originally a section from another report) are numbered as they were in the original report.The Appendix refers only to its own internal tables.

187 LinguisticPerspectives'

Although the present study involved identification and tabulation of articulatory substitutions, a major theoretical focus was upon the interpretation of the substitutions relative to contemporary phonological theory.Since theories of phonology purport to explain regular aspects of the sound structure of language, it is natural to expect that regular aspects of articulatory errors could be explained in terms of a phonological theory.For reasons which will be discussed below, the theory of generative phonology, as adapted to include the concept of markedness, is judged to be most adequate for this purpose (Chomsky and Halle, 1968, Chapter 9).

1.Phonetic Features

1. 1.Phonetic features iL phonological theory.Phonological theory seeks to characterize, for any language, the inventory of phonological segments (phonemes) in that language. Contemporary phonologists generally accept the view that the ultimate unit of phonological analysis should not be the individual phoneme, but rather the phonetic features of which the phonemes are constructed.Table 1 presents brief descriptions of the features used to describe the phonetic content of English phonemes. Notice that many of the features correspond to traditional parameters of phonemic description, such as place and manner of articulation.Table 2 presents a matrix displaying the feature values associated with each consonant nalyzed in the present study. These features frequently refer to thispresence or absence of articulatory properties, such as nasality or vaicing or the involvement or noninvolvement of the corona of the tongue.In other cases, the +/- values reflect the extreme values of a feature (such as backness for the vowels) which range over a continuum and serve to classify phonemes relative to other phoaemes. For example, both /u/ and /a/ are classified as E+ back], although /a/ involves more tongue

1This section was prepared by C. Cairns and H. Cairns.

189

169 r90

.Table 1 Phonetic Features

Feature Description

A speech sound is consonantal if it is produced with a constriction along the center line of the Consonantil oral cavity.Only the vowels and the glides (ens) (/w//h/, and lyl) are nonconsonantal.

Vocalic sounds are those which have a largely unobstructed vocal tract.The liquids /1/ and /r/, which are consonantal, are also vocalic. This is true because while there is a central obstruction for the liquids, there is a large Vocalic unobstructed area to either side of the tongue. (voc) Although there is no central obstruction for glides, the most narrow area in the vocal tract during the production is not large enough to qualify them as vocalic.The glides, therefore, are nonvocalic.

A sound is anterior if the point of articulation is as far front in the oral cavity as the alveolar Anterior ridge.Thus, all the labial and dental sounds (ant) are anterior, while sounds produced farther back are nonanterior.

A sound is coronal if its articulation involves the front (or corona) of the tonguei A sound is non- Coronal coronal if another part of thejbngue is used (such (cor) as in /k/) or if the tongue is not involved in the production of the sound at all (such as in /p/). 191

Table 1 (cont'd)

Feature Description

A sound is noncontinuant if it is produced witha complete obstruction in the oral cavity. Only the Continuant nasals, stops and affricates are noncontinuant. (The nasals are considered to be noncontinuant (cnt) because while there is an opening in the nasal cavity, the oral cavity is completely obstructed.)

A strident sound is produced byan obstruction in the oral cavity which forces the air througha relatively long, narrow constriction. As the air Strident rushes out of the opening of this construction, its (str) turbulence serves as a primary noisesource. This turbulent air is then directed againsta second obstruction which causes a secondary noisesource.

Voiced sounds are those in which phonation (vibra- Voice tion of the vocal folds in the larynx) takes placeas (voi) the sound is articulated.

A lateral sound is one which involvesa contact between the corona of the tongue andsome point on the roof of the mouth, along with a simultaneous Lateral lowering of the sides of the tongue.In English, /1/ (lat) is the only lateral sound, and this feature differen- tiates it from /r/, the only other liquid, which is nonlateral.

191

:7"; 192 Table 1 (contld)

Feature Description

Nasals are characterized by a loweringof the Nasal velum, which opens the nasal cavityfor sound (nas) resonation. The above features are used for thedescription of consonants in English.Hence, they are the only features used for analysis inthe present study (cf. Tables 1 & 3).For the sake of com- pleteness, however, the followingfeatures for vowel classification are alsopresented.

High vowels are those which involvethe highest tongue position, and thus the narrowestconstric- High tion in the oral cavity./u/ and /i/ are the only high vowels; all others are nonhigh.

Low vowels are those which involvethe lowest tongue position,/ae/, /a/, and /0/.All Low other vowels are nonlow. Note how the so- called middle vowels.are classedin this system; /e/ and /o/ are nonlow and nonhigh.

The traditional back-front distinction isaccounted for the back/nonback distinction.Thus, /u/. Back /0/, and /0/ are classed as back,while /Up /ae/, and /e/ are nonback.

As in traditional classifications, therounding of the lips is a feature for vowel differentiation. Round Thus, /u//at and /*/ are round. Others are nonround. Feature Content of Phonemes Table 2 Features + + + + + + + + + + Phonemes+ + + + + + + + + VocalicConsonantal UMzsbOdt v OM MD OM WO SO OD fWO bprITZgk WO WO OD OM IN whyl rmn + + as Anterior + + + + OM OM WO GO GIB - Ck.) Coronal + + + /olio Continuant + GI* .111101110Strident + + + GO OM OM SR + + + + OM + WO OM S +00 +Oa + + Voiced WO WO WO SO OD OM UM OD UM OM + + NasalLateral SO GO WO WO ,111 SIP S WO OM WO WO WO - 194

retraction than /u/ does./u/is classified as [+ back] to distinguish it from /i/ (the only other [+ high] vowel), while /a/ is classifiedas [+ back] to distinguish it from /ae/ (the only other (+low] vowel). The major point here is that for the purposes of classifying all the phonemes of a language, +/- valuesare referred to, rather than degrees of realization of individual features. 1.2.Phonetic features in the present study.Phonologists have shown that analyzing phonemes as bundles of features leads to more theoretically parsimonious and adequate phonological analyses. A more pertinent question, however, is--how does the feature approach aid a study such as the present one, which deals with articulatory errors? Suppose that two children misarticulated the phoneme /z/.Child A substituted /s/ for /z/ and Child B substituted /t/ for /z/.If the basic unit of analysis is the individual phoneme, then there is no principled basis for claiming that one substitution is any "closer" to the target phoneme (/z/) than the other, nor is there any general way available to classify the substitutions.Analyzing the substitutions as feature changes, however, is much more informative. Turn to Table 2 and examine the feature content of /z/,/s/, and /t/.The column of pluses and minuses below /8/ differs from that of /z/ by only one cell a-voice] for /s/; [+voice] for /z/).This means that /s/ and /z/ have the same value for every feature other than (voice]./t/ and /z/, on the other hand, differ by three feature values. Like /s/,/t/ is (-voice], but, in addition,It/is [-strident] while /z/ is [+strident] and /t/is (-continuant] while /z/ is [+continuant].(For a description of the features involved, see Table 1.) Therefore, it can be said that /8/ differs from /z/ by only one feature, while /t/ differs from /z/ by three features.The /s/ for /z/ substitution could be said to represent a smaller degree of error than the /t/ for /z/ substitution. Note that a /d/ for /z/ error is in between the other two substitutions in distance from the target sound, as /d/ differs from /z/ by only two features ([continuant] and [strident]).Thus, use of features seems ti provide a metric by which we can judge the degree oi an articulatory sulbstitution error. Another advantage in using the feature approach to interpret articulation errors is that it provides a definition of general classes of errors. (These error classes correspond to "natural classes" of

2This discussion will not include phonological arguments for the use of feature systems. The interested reader is directed to a classic article on the subject, Halle, 1964. 195

phonemes in phonological theory, which will be discussed later.) There are, for instance, eight substitutions which involveonly a change from [+voice] to [-voice]. They are /s/ for /z/;Ie for /V; /t/for /d/; /1/ for /v/; /p/ for /b/; /Cr/for /j/, /k/ for /g/, and /h/ for /w/.It is readily apparent thata very large number of classes oferrors can be defined using the feature system.Thus, the featitre approach providesan informative, quanti- fiable framework Avithth ahich to investigate substitutionerrors. 2. 112.e.Carkedness As was discussed above, the use ofa feature analysis of individual phonemes allows the researcher to distinguishclasses of phonemes (and classes of substitution errors).For example, one can refer to the class of "all voiced sounds," I. e. all those phonemes which are (+voice].In the terminology of phonological theory this grouping by features is referred toas the grouping of the phonemes into "natural classes." Phonologists expect that the phonemes grouped into "natural classes" by features will also evidencesimilar behavior in phonological processes.That is, phonological rules will apply to all the phonemes whichare members of a natural class. A phonological rule would never apply only to /lc/, forinstance, but 43 all other voiceless stops as well. 3 Analogously,one would expect that classes of phonemes would be similarly affected bysubstitution errors.Unfortunately, however, the use of +/- values forfeatures often leads to unsatisfactory groupings into natural classes.

A large natural class defined by +/- feature values isthe class of all voiced segments (all phonemes whichare [+voiced]). This class includes all voiced obstruents (obstruentsare phonemes which are [+consonantal], (-vocalic], and (-nasal],cf. Table 4), all nasals, glides (/w/ and /h/), liquids (/r/ and /1/) andvowels. (Nasals, glides, liquids and vowelsare referred to as "sonorants.") It is expected, then, that the voiced obstruents and thesonorants,

3An ex,:ellent example of the unitary phonological behavior of members of a natural class is word-final devoicing in German.All members of the class of voiced obstruents change-from (+voice]to [-voice] when theyappear in word-final position.Thus, the morpheme "Bad" (bath) becomes /bat/ when there isno inflectional ending.The presence of a stem-final /d/ in the underlying representation of this stem is revealed by the phonetic form when there isan inflectional ending, e. g., /baden/ (plural form). 196

since they form a natural class of all voiced segments, would function similarly in phonological processes and in articulation errors.This is not the case, however.Phonologically, voiced obstruents behave as a class and participate in phonological processes not involving the sonorants (cf. footnote 2).In articulatory errors, devoicing (a change from E+voice] to [-voice], as in the/ s/for /z/ substitution) is only associated with the voiced obstruents.Vowels, glides, nasals and liquids are never devoiced.In fact, very few languages of the world contain voiceless sonorants in their phonemic inventories. A system is needed which will allow the voiced obstruents and the voiced sonorants to be separated into two different natural classes.It would not do, however, to eliminate the +/- feature system entirely, because the+/- values are necessary to describe the phonetic content of phonemes. The solution is to add another system which 4escribe the relationship between a particular feature and the other lures in the phoneme within which it occurs. Consider again the feature [voice]. Articulatorily, the production of voicing in obstruents involves more articulatory complexity than does the production of voicing in sonorants. This is so because for phonation (voicing) to take place there must be a stream of air passing through the larynx.The production of an obstruent, however, involves the creation of an obstruction in the oral cavity, coupled with the closure of the velum, (recall that obstruents are (+consonantal], [-vocalic], and [-nasal]) thus creating an impediment to the stream of air.Since the introduction of an obstruction in the oral cavity and the closure of the velum causes supra-glottal air pressure to rise as air passes through the larynx, phonation during the production of an obstruent requires extra articulatory effort.In the case of the soraorants, however, there is a large open cavity above the glottis (the oral cavity in the case of vowels, glides and liquids -- the nasisl cavity in the case of nasals), so supra-glottal air pressure does not increase as air passes through the larynx, and voicing occurs spontaneously, with no extra effort on the part of the speaker.In fact, the suppression of voicing requires extra articulatory effort.Thus, a voiced obstruent is a more coniplex phoneme than a voiced sonorant. The system which has been devised to capture relationships of this sort is the system of marking (Cairns, 1969; Chomsky and Halle, 1968, Chapter 9).In this system each feature is assigned a marked (M) or unmarked (U) value.Table 3 presents the same phonemes as those presented in Table 2, with M/U feature values specified, rather than +/- values.Notice that obstruentra whicl. are E+voice] on Table 2 (/z/, 3/,Id!,/v/,/b/,/)'/,and /g/) are CM voice] on Table 3. On the other hand, the sonorants which are (+ voice] on Table 2 (/w/, /j/, /1/,/r/, /m/, and /n/) are Et1 voice] on Table 3.(If vowels were

6 Features WU Feature TableValues 3 of Phonemes Phonemes VocalicConsonantal a U+zsi + + + +dt v + + +f +bplir + + + Zgk+ + + whylrrnn = - + + + + ContinuantCoronalAnterior U UMMUUUUMMMMMUUUUMUUMUUUUUUUUUUMMMMMUUUUUUUUUUUUUUUUUUUUUMMMMMUU LateralVoicedStrident MUMUMUMUMUUMUMUUMUUUUUUMMMMUUMMUUMUUUUUUUUUUUU UMMUUMMUUUUUUUUUUUUUU ComplexityOINIMMIPNasal 2U UUUUUUUUUUUUUUUUUUUMMIUUUUUUUUUUUUUUUUUMUUU4 3 1 0 3 2 2 1 3 3 2 2 I I 2 2 2 'consonantal'.because the conversion Sec a Note that the feature the Interpretative of M's to Uls for 'consonantal' does Conventions presented all features is not take on marked and dependent upon the +/- in Table 4 (page 201). unmarked values. value of the featureI 21 This is -4 198

included in these two tables, they would beC+voice] on Table 2 and EU voice) on Table 3.)Table Z, then, reflects a +1- value system which describes the phonetic content of thephonemes while Table 3 reflects a marking system which expresses therelationship of the individual features to the phcnemes in whichthey occur.The assignment of CMvoice), therefore, is made whenever the articulatoryad- justment to produce or suppress voicinginvolves more articulatory effort on the part of the speaker.(Recall that voiced obstruents and voiceless sonorante ars CM voice), whilevoiceless obstruents and voiced sonorants are CUvoice). )9 The marking system of Table 3 also provides a natural class (allthose phonemes which areCM voice)) including voiced obstruents and anothernatural class (al those phonemes which are EU voice)) which includesvoiced sonorants. Consider the previous example of an/5/ for /a/ substitution. Table 2 revealed that the substitution represents achange of one feature value ((+ voice] to C-voice3).Consultation of Table 3 shows that the value change is also fromCM voice) to CU voice); thus, the +1. system indicates that the substituted phoneme isonly one feature distant from the target phoneme, while the M/U systemindicates in addition that the substitution is a simplifyingsubstitution.This I. true because /5/ has fewer M's in its total inventory thandoes /a/ and also because the feature of [voice) is changed fromits more complex to its less complex value.It is important to notice that thesubstitution is not simplifying just because the feature changesfrom a 4.to a value.If an /h/ had been substituted for a/w/, this would also be a change of(+voice) to [-voice), but Table 3 revealsthat such a substitution would not be a simplifying one, as it would involve achange from voice] to CM voice]. (Recall that /h/, which is voiceless, ismarked for voicing because it is a sonorant; seefootnote 4. )

2. 1.A perceptual basis for marking.Marking may also reflect increased complexity attributable toperceptual factors.Three classes of phonemes are relevant to thisdiscussion; (1) strident fricatives -- 1. tlese are obstruents which are[+ continuant) and C+ strident) (/s/,/z/, /s/); (2) nonstrident (mellow) fricatives these are obstruents which are[+ continuant) and [- strident) (/*/,/0/,/ll,and /v/); and (3) pure (nonaffricated) stops --these are obstruents which areC. continuant) and (- strident) (/d/,/t/,/b/,/p/,/g/, and /k/).The main perceptual cue indicating the presenceof a pure stop seems to be a brief

4Notice that the one voiceless sonorantunder consideration,/h/, is in fact markedCM voice] on Table 2. 199

period of silence during the time of closure.Fricatives, on the other hand, are signaled by thepresence of noise during the time of the consonantal obstruction, and stridentfricatives are noisier than nonstrident fricatives.Therefore, the perceptual distinction between strident fricatives and stops is greater thanthat between nonstrident fricatives and stops.Thus, one might say that strident fricativei are optimal fricatives, whereasnonstrident fricatives are nonoptimal because theil nonstridency tendsto attenuate the perceptual distinction between the fricatives and thestops (Jakobson, 1941). Note that in Table 3 the nonstrident fricativesare EM strident], while the strident fricativesare EU strident).This marking reflects the fact that the nonstrident fricativesare nonoptimal and are more perceptually complex. The pure stops, ofcourse, are Ev strident); their lack of stridency makes them perceptually optimal.2 Among the nonstrident fricatives,/f/ and /9/ (and their voiced counterparts /v/ and /i1/)are distinguished from each other by point of articulation. According to the featuresystem employed here /9/ and /d / are C+coronall while /f/ and /v/are (-coronal). /f/ and /v/ are clasgified hereas EM coronal), whereas /9/ and /*/are EU coronal3. ° Although this assignment of Ms and Us is not as certain as those assignments discussed above, there doesseem to be some evidence for this marking. Among the evidence is the fact that several languages exist_ which containan /f/ and a /v/ but do not contain a /9/ and a /41, whereas few languages have.a /9/and a /1/, but not /f/ and /v/. Another considerationis the general opinion of speech therapists that labialsare easier to articulate than lingual consonants. 2. 3.Complexity. Since the assignment ofan M value to a feature reflects relative articulatory or perceptual complexity, the

5This illustrates an important point which must be borne in mind as Tables 2 and 3 are compared with eachother. A feature which has a +. value for a particular phonemeon Table 2 will not necessarily be marked with an M on Table 3 (nor will- values on Table 2 correspond invariably toa U on Table 3).If this were the case then there would be no point in having the marking system. 6 The feature values serve to distinguishthe phonemes from each other. The reader will notice thatin both the M/U and the +/- matricesno two phonemes have the same set of feature values. 200

total number of Ms assigned to a particular phoneme can be said to characterise the total complexity of that phoneme. Moreover, the relative complexity of phonemes can be determined by comparing the number of M assignments of each, so that a phoneme with more Me is more complex than one with fewer Me. The bottom row ofTabf; 3 displays the complexity value for each phoneme on the chart, derived by simply counting the number of Ms assigned to that segment.

2. 4.Universal interpretative conventionsAs was discussed above, the +1- systemanci the M/U system zor assigning feature values are designed to reflect different kinds ot facts about the individual features.It is apparent, however, that the two systems are not independent of whether its descriptive value is + or .In phonologicial theory, therefore, there must be some description of the relationships betweezi the two systems.That description must hold for all languages of the world (that is, it must be universal) or the M/U assignments could be made up separately for any individual language which was being studied.Table 4 presents the set of universal interpretative conventions which relate the M/U values of Table 3 to the +/- values of Table 2. The conventions are presented both in the formal notation of phonological theory mainly for the purpose of illustrating that notation.They are also verbally described in the same table.

3.Phonetic Features and Markedness as Related to the Present Study It should be valuable to view articulatory substitutions as feature changes rather than as the substitution of one unitary phonenve for another.If this is the correct approach, an analysis of errors should indicate that many features of the target phoneme are retained and only a few are changed when a substitution occurs.This was discussed above as "degree of error" of the substitution.Using the feature as the primary unit of analysis allows the description of an error to include a statement about the "distance" of the substituted phoneme from the target phoneme. The use of the M/U system to characterize feature values should make it possible to explain many substitution errors. One would expect that within phonemic classes more errors would be associated with the more complex (more marked) phonemes than with the less marked ones.It also seems reasonable to expect that when a substitution occurs, the substituted phoneme will be lees marked (have fewer Ms) than the target phoneme which it replaces. When the individual features of the substitution are considered, the theory of complexity presented here predicts that the valuechanges which effect the substitution will be from M to U.

4%) 201

Table 4 Interpretative Conventions

41011111111111111111111MIIIMMINIMINIIMII1111111MINIMINIIIMEMINSPIO=IIIMIMMIIINIONIMINIIIMINSION111111MIMINIMISMIIIIIIME

Conventiona NMI Formal Notation I. The unmarked value of 'vocalic' CU voc3 tactvoc 3/ 16..1 is the opposite value ofconsonantal. cns j

2. The unmarked valueof 'nasal' EU nag] is 0 [-nag]

3. For all consonantalsounds, the Cu ant] unmarked value of 'anterior'is -40C+antY[Tgrai .

4. For trueconsonants (i.e. all CU cnt3 -4 [cnt C+cons3 E-voc3sounds), theun- [4.cns} marked value of 'continuancy'is -voc

5. Theunmarked value of 'stridency' Eu sti).-4 is the sameas the continuancy Ca str3/ree feature for all nonnasal,true con. - voc sonants (i. e.,all obstruents).It - nas is for all other sounds. ct cnt

6. The unmarkedvalue of 'coronar is- for continuant, nonstrident sounds and for nonanteriorsounds. It is + for all othersounds. (That is,/8/ and /d/are marked for coronal,although /t/. /d/,/8/, and /z/are not./f/ and /v/are also not marked for corona]. The point isthat /f/ and /v/ are theunmarked nonstrident continuants. 202 Table 4 (cont'd)

issafileMINIMMOOMMerbionwasdianuenimmorISIMIIIIIMIIIIIIIINNIIIIIIIIIIMOS111=1111111110111111

Conventions Formal Notation

'1111111111M 7. An obstruent (1. e., at+cns) C-voc3 t-nas3 sound) is unmarked for voicing when it is E-voi3 all -voc other sounds are t+voi) in the une. enas marked state. 1t+voi) ij

411111111111111MMINIIII 8. The unmarked value of 'lateral' CU lat) au. (-la!) is .

MIBBIONIONIIW 111111111111MINPOMMINIMIIIMIIM11110111111111111111111111111MIMENIMMIIIMINIMINOMMIONIMISIMOr aNote that these conventions must apply in the order in which they are written.For example, 'I' must have applied to give a '+, or a '-' value for the feature 'vocalic' before conventions 1, ''5, ' or ,7' can apply. The interpretative conventions are written in terms of the U value for each feature.The M value assumes exactly the opposite +/- value from that assumeay the U value.Thus, for exar=tp e, in convention #8, CU lat3 is related tor-lat3 and CM lat) is related to C+ lad]. 203

A point which has not been mentioned is the importance cat the in: midual featiares.All the discussion up to this point has seemed to indicate that Estridentl, for instance, is as important s feature as EanteriorlActually, very little is known about the :Ave impoYtance of the features, but one might speaulate that re like stridency, vohkh is very distinctive perceptually, might be more important (or carry more information) than the features which spocify place of articulaticisuThis is the sort of information which could be provided by a study of articulatory substitutions.If some features seem more vulnerable to value changes than others, then the more stabla ones can be characterised as more important. ,,The vulnerability of a feature will probably be related to whether it is marked (since feature changes are expected to be from value to a U value) and also to the values of other features with which it interacts.

;

203 APPENDIX B

Frequency of Occurrence of Target Phonemes on the 1 Goldman-Fristoe Test of Articulation

Target Position f Words

i 1 zipper

a m 1 scissors f 4 scissors, matches, pencils, Claus 1 i 2 scissors, Santa s in 2 pencils, Christmas f 2 house, Christmas

i 1 vacuum

v m 1 shovel

f 1 stove.

i 1 shovel

I m 1 fishing

f 1 brush

i 3 bathtub, bath, bed

b rn 1 rabbit

f 1 bathtub

i 1 this (or that) i In 1 feather f 0

i 1 jumping I 211 1 pajamas f 1 orange

205

s.ovik0, . 7 206

Target Position I Words

i 1 thumb

0 m 1 bathtub

f 1 bath

i 1 pencils m 3 zipper, jumping, sleeping f 2 cup, lamp

1 gun

rn 1 wagon

1 flag

i 1 duck

d m 1 window

f 1 bed

i 2 chicken, church

Z m 1 matches

f 1 church

i 1 telephone t m 2 bathtub, Santa f 2 rabbit, carrot

3 fishing, feather, finger

1 telephone

1 knife

i 3 cup, car, carrot k rn 2 chicken, vacuum

f 1 duck knife 5 window, cleaner, pencils, orange, Santa 5 telephone, gun, wagon, chicken, plane

2=35 207

Target Position Words

1 matches 2 lamp, Jumping 3 vacuum, thumb, drum

i 2 rabbit, ring r m 2 carrot, orange f 0

4 scissors, church, squirrel, airplane 5 zipper, cleaner, car, feather, finger

i 1 lamp

m 2 telephone, yellow ' f 3 wheel, squirrel, pencil

1 house In 0 0

2 window, wagon 0 0

1 yellow

TY1 0 0

Below are the clusters appearing on the Goldman-Fristoe. All are morpheme initial and there is one instance of each.

pL airplane br brush 208