HaslUns LaboratorieB Sfatlu Report on Sp«ch Research1991, SR·I07/108, 41-62

From Babbling Towards the Sound Systems of English andFrench: A Longitudinal Two-Case Study*

Andrea G. Levittt and Jennifer G. Aydelott Utmantt

The utterances of one French and one American infant at 0;5, 0;8, 0;11, and 1;2 weretranscribed and acoustically analyzed for syllable duration and vowel formant values.Both general and language-specific effects emerged in the longitudinal study. Initialsimilarities in the consonantal repertoires of both infants, increasing control in producingtarget Fl and F2 values, and developmental changes in babbling characteristics over timeseem to reflect universal patterns. Yet the babbling of the infants differed in ways thatappear to be due to differences in their language environments. Shifts in the infants'sound repertoires reflected phoneme frequencies in the adult languages. The English.learning infant produced more closed syllables, which is characteristic of English, than theFrench-learning infant. The French-learning infant tended to produce more regularly­timed nonflnal syllables and showed significantly more final-syllable lengthening, bothcharacteristic ofFrench, than the English-learning infant.

The study ofvery young children's acquisition ofthe sound patterns of their native language hasfocused, in particular, on whether or not there iscontinuity between babbling and early speech andon the roles of the linguistic environment and ofgeneral developmental processes in the child'searly productions. Current research has producedmuch evidence in support of continuity betweenbabbling and early speech and has suggested theimportance of physiological maturation to thechild's growing sound repertoire. However, empir­ical investigations of language-specific influenceshave had somewhat mixed outcomes, perhaps,in part, because of differences in the types of

The research reported here was supported by grantDCOO403 to Haakina Laboratories. Recording of the Frenchinfant was begun when the first author was a NATOpostdoctoral fellow at the CNRS in Paris, France. Analyses ofthe data were conducted by the second author as part of asenior honors thesis at Wellesley College. We thank thefamilies of our American and our French infants for theircooperation. We are grateful to the Sloan Foundation and toWellesley College for additional support for this project.Portions of this paper were presented at the April 1990conference ofthe Intemational Conference on Infant Studies inMontreal, Canada. Catherine Best, Alice F. Healy, MichaelStuddert-Kennedy, and Doug Whalen provided helpfulcomments on earlier drafts of the paper.

41

language-specific influences that were in·vestigated (e.g., segmental versus prosodic) or inthe techniques used (e.g., phonetic transcriptionversus acoustic analysis). Therefore, in our pre­sent investigation of continuity and languagespecificity, we both phonetically transcribed andacoustically analyzed a number of prosodic andsegmental characteristics in the babbling develop­ment of an English-Ieaming and a French-leam­ing infant at four points between the ages of 0;5and 1;2.

The best-known claim about the discontinuitybetween babbling and early speech comes fromJakobson (1968: 25). He suggested that 'thephonemic poverty of the first linguistic stages', ascompared with 'the phonetic abundance ofbabbling', is indicative of a fundamentaldistinction between the two stages, and that thebabbling period is therefore unrelated to theacquisition of language. He even proposed thatinfants pass through a silent period during thetransition from babbling to speech, in which thechild produces no sounds whatsoever.

This claim of discontinuity between babblingand early speech has not been supported by recentstudies (Ferguson, 1978; Kent & Bauer, 1985;Menyuk, 1971; Menyuk & Menn, 1979; Oller,Wieman, Doyle, & Ross, 1976; Stark, 1980;

42 Levitt 'mil Utman

Vihman, Macken, Miller, Simmons, & Miller,1985). For example, Vihman et a1. (1985) exam­ined the babbling and early word productions ofnine children from 0;9 to 1;4 and found that bab­bling continued to occur after the onset of firstwords, without evidence of a silent period, andthat the phonetic properties ofbabbled utterancesboth influenced and were influenced by word pro­ductions. Babbling increased in complexity follow­ing the acquisition of words, and some sounds ap­peared in babble only after words containing thosesounds were learned. In addition, an infant's'babble repertoire' was reflected in his or her laterwords.

The evidence that supports the role ofphysiological maturation in children's early vocalproductions comes largely from studies that lookat the developmental stages of babbling or atpatterns in children's early segmental inventories.Oller (1980) and Stark (1980) defined severaldevelopmental stages of speechlike sound makingwithin the babbling period that reveal babbling tobe a logical and organized precursor to adult-likespeech.

The first stage identified by both Oller andStark begins at birth and lasts a month to sixweeks. During this period, termed the 'phonation'or 'reflexive vocalization' stage, infants producecry and fussing sounds as well as some partiallyresonant and a few fully resonant vocalizations.At the next stage, called the 'goo' or 'cooing' stage,roughly between 0;2 and 0;4, increasing vocalcontrol gives rise to the infants' production ofsounds that are typically heard as 'coos' by adults.At about 0;4 infants enter the 'expansion' or 'vocalplay' stage, during which they produce a widevariety of vocalizations, including squeals, growls,and nasal murmurs. Towards the end of thisperiod, they produce sounds that begin to

.approximate a number of adult consonants andvowels. Next, at about 0;6 or 0;7, infants begin aperiod of 'canonical' or 'reduplicated' babbling, inwhich they produce strings of identical consonant­and vowel-like elements, e.g., 'bababa.' The timingof these elements is much closer to that of adultspeech than before. At the next stage, 'variegated'or 'nonreduplicative' babbling, which begins atabout 0;10 or 0;11, infants are able to producestrings of syllables that contain a variety ofconsonantal and vocalic types, e.g., 'adega,' as wellas a greater variety of syllable types. This laststage immediately precedes the infant's earlyword productions. Each stage generally reflectsthe onset and increasing incidence of a new type ofvocalization, but the infant does not abandon

earlier types of vocalizations. Thus, a child who iscapable of variegated babbling may still produceutterances of the reduplicative type, just as a childwho has begun to produce words also continues toproduce babbled utterances.

Others have found additional evidence in sup­port of the roles of continuity and maturationalprocesses through the examination of infants' ac­quisition of phone types. Thus, Locke (1983) re­views the evidence for universal patterns in chil­dren's acquisition of speech sounds and argues infavor of anatomical and physiological maturationas the common source of these patterns.

Evidence for maturational processes has alsoemerged in studies of the acoustic properties ofchildren's babbling and early speech. Buhr (1980)measured the vowel formant frequencies of thetranscribed utterances of an American infant from16 to 64 weeks of age and plotted the results on anF1/F2 plane. In describing the longitudinal devel­opment of the child's vowel space, Buhr attributesthe early appearance of the sounds m, [e], and [£]to the early development of the control ofjaw posi­tion. The sound [u], on the other hand, requirescontrol of three elements: jaw position, which ap­pears early on, longitudinal tongue movement,which arises early on in connection with swallow­ing, and lip rounding, which is also associatedwith sucking behavior. According to Buhr, the lateemergence of [u] is due either to the difficulty ofcoordinating the three types of oral movement, orto the fact that voluntary control of lip roundingemerges considerably after its involuntary use insucking. Finally, the development of neuromuscu­lar ability in the lips, jaw, and front tongue isknown to precede the physical descent of the lar­ynx and rear tongue. Thus, Buhr attributes theemergence of the acute axis of the vowel triangle,which includes front vowels such as [i] , before theemergence of the grave axis, which includes thelow back vowels such as [0], to this difference inphysiological development.

Other acoustic studies also chart basic develop­mental progress towards more adult-like produc­tions. Bond, Petrosino and Dean (1982) measuredthe first and second formant frequencies of thespeechlike productions of an infant from 1;5 to 2;2and plotted the results on an F1/F2 plane. Theirdata suggest a refinement of the vowel space overtime toward more adult-like formant frequencypatterns, which presumably reflect the develop­ment of increased articulatory proficiency withage. Mack and Lieberman (1985) measured thebabbling and words of an infant from 46-149weeks for vowel length and consonant closure.

From Babbling Towards the Sound Systems ofEnglish lind Fr~h: A LongitudinJll Two-Cllse Study 43

They found that specific segments of babble in­creased greatly in duration during the course ofdevelopment, demonstrating the child's growingability to approximate the longer adult-likenorms.

But even though current research providessupport for the ideas that there is continuitybetween babbling and speech and that maturationplays a role in the development of children's vocalproductions, these factors do not preclude thepossibility that the specific language environmentmay begin to influence vocalizations duringinfancy. The reports on language-specificinfluences have been mixed, however. A number ofstudies suggest an influence of the characteristicsof the native language on the child's prelinguisticutterances, whereas others have failed to Uncoverlanguage differences. One of the earliestsuggestions for language-specific influences comesfrom Brown (1958: 199), who hypothesized thatbabbling 'drifts in the direction of the speech aninfant hears', until the infant has mastered theadult phonology of the parents' language.

Some perceptual evidence appears to supportBrown's notion of 'babbling drift.' Dinger andBlom (1973) asked 50 Dutch and 50 Americanwomen to listen to 90-second recordings of Dutchand American infants taken at 33, 43, 53, 62, and70 weeks (approximately 0;8, 0;10, 1;0, 1;2, and1;4), and to identify whether the babbling wasDutch or American. The women correctlyidentified the language background of the infantssignificantly often for ages 1;0, 1;2, and 1;4months suggesting that after 1;0, languageidentification was possible. However, Dinger andBlom acknowledged that some words may havebeen included in the babbling samples, so that itis not clear on what basis the listeners made theirjudgments. Boysson-Bardies and her colleagues(Boysson-Bardies, Sagart, & Durand, 1984) alsofound that French adult listeners were able todistinguish at better than chance levels thebabbling of French infants from the babblings ofChinese and Arab infants at 0;6 and 0;8, thoughthey were no longer able to do so when listening tothe babbling of ten-month-old infants (see alsoWeir, 1966). The stimuli consisted of 15 secsamples of babbling from a number of differentchildren from each of the language backgrounds.Unlike the productions at 0;6 and 0;8, which hadclearer melodic patterns, the stimuli from theproductions of the 10-month olds were all highlyarticulated reduplicative babbles, and were thuspoorer in intonation contour.

On the other hand, a number of other studiesfailed to find support for adults' ability to identifyperceptually the linguistic background of the childfrom babbled productions (e.g., Atkinson,MacWhinney, & Stoel, 1970; Olney & Scholnick,1976; Thevenin, Eilers, Oller, & Lavoie, 1985).Thus, Thevenin et al. (1985) examined the abilityof adult monolingual speakers of English andbilingual speakers of English and Spanish todiscriminate the babbling of English-learningfrom Spanish-learning infants at 0;7-0;10 and at0;11-1;2. The stimuli consisted of 8 utteranceseach of 1- to 3-second stretches of canonicalbabbling from 14 infant subjects (7 English- and 7Spanish-learning children). Adult listeners werenot able to make correct language backgroundidentifications for the infants at either 0;7-0;10 orat 0;11-1;2, although they did show certainconsistent patterns of response to particularinfants and utterances.

One possible explanation for the discrepancy be­tween the Thevenin et al. (1985) study and that ofBoysson-Bardies et al. (1984) is that adults' abilityto judge the linguistic background of a child maydepend on the prosodic characteristics of the hairbling, and thus may require long samples withpronounced melodic contour. It may also be thatspeech development, like the development of thechild's phonological system, shows evidence of re­gression (Menn, 1971). Thus, children may in factproduce intonation contours that mirror those ofthe language of their environment at an early ageand then apparently cease to produce such pat­terns consistently as they turn their attention tothe production of more segmentally complex ut­terances. It is then perhaps not surprising thatthe French judges in the Boysson-Bardies et al.(1984) experiment were unable to identify the lan­guage background of the 10-month-old children,although they had done so for the younger chil­dren, because the prosodic cues were apparentlypoorer in the productions of the 10-month olds.

More recently, however, studies of the acousticand phonetic properties of infants' babbling fromspecific language backgrounds have providedevidence in support of the babbling drifthypothesis. Boysson-Bardies, Halle, Sagart, andDurand (1989) found that the distributions ofvowel formant frequencies of five 10-month-oldinfants each from French, English, Cantonese, orArabic backgrounds reflected the differentformant-frequency pattern distributions of adultvowels in the target languages, a result thatdemonstrates the early influence of the language

44 Leuitt Ilnd Utmiln

environment on vowel production in infantbabbling.

Similarly, Boysson-Bardies (1989) reportsevidence of babbling drift in a cross-linguisticstudy involving infants from French, American,Japanese, and Swedish backgrounds. Fivechildren from each linguistic community wererecorded in six sessions, starting at 0;10, whenthey could only babble, and continuing until the25-word stage. In this study, transcriptions of theconsonantal repertoires of the infants showedevidence both of universal developmentalprocesses and of language-specific influence fromabout age 0;10 on. Thus, stops represent, by andlarge, the largest proportion of consonants forinfants from all language backgrounds,presumably because they are easier to producethan other consonants. However, from the sessionat 0;10 onwards there was a clear stop frequencydifference between the French infants, on the onehand, and the American and Swedish infants, onthe other. The latter groups regularly produced ahigher mean percentage of stops than did theFrench. This difference is consistent with thehigher frequency of stop consonants in the target­words attempted by the American infants andthose attempted by the Swedish infants whencompared to the frequency of stop consonants inthe target words attempted by the French infants.

Most studies examining the effect of the lin­guistic environment have generally taken a cross­sectional rather than longitudinal approach. Onthe other hand, most of the previous research in­vestigating the role of maturational factors inspeech development has focused on a single lan­guage longitudinally. Even in the rare caseswhere a study of the effect of the linguistic envi­ronment was longitudinal, as in the case ofBoysson-Bardies (1989), the data were usuallyonly phonetically transcribed and not acousticallyanalyzed. Ifwe are to determine the time course ofmaturational and language-specific factors onspeech development, we need to make longitudi­nal cross-linguistic comparisons. Furthermore,acoustic analyses may reveal effects that are notapparent in perceptual or transcription studies. Inan acoustic analysis of voice onset time (VOT),Macken and Barton (1980) found that some chil­dren, who were heard by adults to be using initialvoiced stops for voiceless targets, were actuallyproducing longer VOTs for their voiceless targetsthan for their voiced ones. The difference betweenthe two types of stops was, however, inaudible tothe unaided ear. Therefore, the present study ex­amined the role of maturational and language-

specific effects longitudinally by comparing boththe transcriptions and the acoustic analyses of thevocal productions of a French-learning and anAmerican English-learning infant at 0;5, 0;8, 0;11,and 1;2. A case study approach to this problem isappropriate as a means of generating specific hy­potheses about the speech development of French­and English-leaming infants that can later betested on a larger sample of subjects from eachlanguage background.

French and American English infants were cho­sen for comparison because of specific known seg­mental, syllabic, and prosodic differences betweenFrench and English. A comparison of the soundinventories of the two languages reveals thatEnglish has diphthongs, which are not found inFrench, more fricatives and affricates thanFrench, and one more nasal consonant thanFrench (frY), while French, on the other hand, hasnasal vowel phonemes, which are not found inEnglish (Delattre, 1965). Syllabically, the two lan­guages also differ in the most common syllablestructures and the typical word lengths (in num­bers of syllables). English contains more closedsyllables and more monosyllabic words thanFrench (Delattre, 1965). Finally, in terms ofprosody, French and English differ in fundamentalfrequency (Fo) contours and durational patterns.For the present study, we focused on acousticmeasures of durational pattern differences be­tween the adult languages. French and Englishare classified respectively as syllable-timed andstress-timed languages (e.g., Pike, 1945, but cf.Wenk & Wioland, 1982). French, the syllable­timed language, has a rhythmic structure charac­terized by nonfinal syllables generally equal inlength, or isosyllabicity, and by regular breath­group-final lengthening. English, the stress-timedlanguage, has phrase-final lengthening, but thisinteracts with lengthening due to lexical stress,and therefore does not show the relative nonfinalisosyllabicity of French.

Therefore, if the language environment influ­ences infant babbling, we should expect to seegreater nonfinal isosyllabicity, more phrase-finallengthening, and perhaps more polysyllabicutterances in the babbling of the French infant,whereas we should see more monosyllabicproductions and more utterances with closedsyllables in the babbling of the American infant.Furthermore, if the infants' productions areaffected by the segmental inventories of theirlinguistic environments, their productions should,over the course of time, reflect that influence inthe proportions of specific vowel and consonant

From Babbling Towards the Sound Systems ofEnglish and FreJU:h: A LongitudinJJ1 Two...OJse Study 45

types. Thus, language-specific effects wereassessed through the examination of segmentalfeatures, syllable type, and timing, both in termsof the transcriptions of the utterances (consonantand vowel inventories) and in terms of theiracoustic properties as well (the F1 and F2 formantfrequency values of the vowels and durationalmeasures of the syllables).

General effects, or language-independent simi­larities in the developmental patterns of the twoinfants, should be revealed through the study ofsyllable types and segmental features. The bab­bling sequences produced by both infants overtime should evolve in keeping with the stages de­scribed by Stark (1980) and Oller (1980). Ifphysiological constraints are the maindeterminants of the types of sounds that veryyoung infants produce, the vocal repertoires of thetwo infants should be more similar to one anotherat earlier stages rather than at later stages, whenother determinants (from the surroundinglanguage) have also come into play. Finally, overtime, both infants should improve in their abilityto produce the target formant frequency values oftheir vowels and thus show less variability inthose measurements.

MethodSubjects. The babbling of a male American

English-learning infant (E-CR) and that of afemale French infant (F-YC) were recorded intheir homes by their parents on a weekly basisfrom 0;5 to 1;5. Both infants were second children,each with a single sibling of the same sex at leastfour years older. The American infant, recorded ina household in Providence, RI, heard regionallytypical American English and the French infant,recorded in Paris, France, heard Parisian French.

Materials. E-CR was recorded using a Panasoniccassette tape recorder with a Realistic supercar­dioid 33992A microphone. YC's utterances wererecorded using a Panasonic RQ 3145 cassette taperecorder with a high-quality cardioid microphone.

ProcedureData collection. Individual recording sessions

lasted approximately 10-20 minutes. The parentswere instructed to choose a time for recordingwhen the infant was likely to be alert and unlikelyto cry (e.g., in the morning after eating). The par­ents were to hold the microphone at about 20 emfrom the baby, avoiding external noise, and to saythe date at the beginning of each recording toidentify individual sessions. The parents could ei­ther elicit babbling by talking and gesturing(making sure to stop talking while the baby vocal-

ized) or record spontaneous babbling as it oc­curred. A comment sheet was filled out for eachtape, including date, time, and situation (e.g., 'inbath' or 'watching sister), and the parents wereasked to indicate whether babbling was sponta­neous or elicited by a specific person.

Utterance sampling. The utterances used in thisanalysis were taken from four separate months ofrecordings beginning when the children turned0;5, 0;8, 0;11 and 1;2. These ages were chosen tocapture four of the stages of infant vocalizationthat were described earlier: vocal play,reduplicative babbling, variegated babbling, andthe period of the onset of the child's first words(Oller, 1980; Stark, 1980). Only utterancescontaining V, CV, VC and CVC syllables withvocalic durations greater than 100 ms wereselected for analysis. No other productions typicalof the vocal play period (i.e., squeals, growls, etc.),emotive sounds (i.e., crying, laughing, etc.) orvegetative noises were used.

Phonetic transcription. Utterances were dividedinto separate strings or breath groups, which weredefined as an individual syllable or sequence ofsyllables separated from other utterances by asilent interval of 400 ms or more. Each utterancewas transcribed by two phonetically trainedlisteners, both of whom are fluent in English andFrench. Utterances for which the transcribersdisagreed were retranscribed until agreement wasreached. l

Acoustic analysis. The utterances were input athalf speed and digitized with a Macintosh Pluscomputer at a sample rate of 10 KHz, using aMacADIOS digitizer and a TTE 411AFS amplifier.The utterances were measured for vocalic Fl andF2 frequencies (Hz) at the peak. amplitude portionof each vowel. Vowels for which either F1 or F2measurements were indeterminate were droppedfrom the formant frequency analysis. For F-YC,15.9% and for E-CR, 15.8% of the vowels overallgave indeterminate measures. Eleven utteranceswere randomly selected from each infant'sproductions for remeasurement. For E-CR theseutterances contained a total of 41 syllables, andthe overall mean discrepancy for Fl and F2measurements was under 15 Hz. For F-YC theseutterances contained a total of 69 syllables, andthe overall mean discrepancy for F1 and F2measurements was under 20 Hz.

Syllabic durations (ms) were also measured. Inmultisyllabic utterances, measurements weremade by visually inspecting the waveform and bylistening to each of the measured portions. Aconservative criterion for measurement of vowel-

46 LeuiH lind Utman

final syllables was adopted: only visibly voicedportions of the final syllable were included.Remeasurement of the durations of the samerandom sample of E-CR's utterances was within23 ms on average, while remeasurement of thedurations of F-YC's eleven utterances was within13 ms on average. A total of 881 syllables for theFrench infant (114,317,201, and 249 syllables at0;5, 0;8, 0;11, and 1;2, respectively) and 701syllables for the American infant (178, 218, 218,and 87 syllables at 0;5, 0;8, 0;11, and 1;2,respectively) were measured.

ResultsThe results will be described in terms of

segments, syllable structure, and timing, in thatorder. The transcriptions provide the data for ananalysis of consonant and vowel frequencies ofoccurrence and of the syllable structure. Theacoustic analyses provide the data for formantanalyses of the vowels and for the timingcharacteristics of the children's utterances.Segmental Features

Transcription analysis. The sound systems ofEnglish and French will be discussed first interms of the consonant and vowel frequencies andthen in terms of the distribution of consonantsaccording to manner and place of articulation. Inorder to examine longitudinal patterns of changein segment frequencies, charts of percentdistribution of vowels and consonants werecreated for each infant at each age and then listedin order of percentage of occurrence (see Tables 1­4). The charts were compared to lists of the soundsof French and English in order of percentage ofoccurrence (see Table 5; Delattre, 1965).

As can be seen from Table 5, English andFrench share 18 consonants of the 25 listed. Sixconsonants appear in English but not in French(generally fricatives and affricates and the nasal/rY>, and one appears in French but not in English(lq/, a labial-palatal approximant). Of the 21vowels, 8 are shared by the two languages, 5 occurin English but not French (mostly diphthongs)and 8 occur in French but not English (frontrounded and nasal vowels).

Across all ages observed, the infants had 27segments in common in their productions, andeach had an additional 8 sounds that the other didnot use. Sixteen of the common segments wereconsonants, thirteen of which appear in bothFrench and English. The English-learning infantproduced an additional 7 consonants, mostlyfricatives and affricates, two of which are onlyfound in English ([et] and [If]), and one ([x]) that isfound in neither English nor French. The French

infant produced· only the shared 16 consonants,three of which occur only in English and not inFrench. As for the vowels, both infants producedthe same 11 vowels, though with differentfrequencies of occurrence. The eleven vowelsincluded two non-French vowels ([Ie] and [U) andone non-English vowel ([0]).

Table 1. List of consonants in order ofpercentage ofoccurrence at 0;5, 0;8 , 0,'11, and 1;2 for E-CR(Percentage base: total number of consonants insyllables analyzed at each age: 178, 218, 218, and 87 at0;5,0;8,0;11, and 1,'2, respectively).

0;5 0;8 0;11 1;2

x 24.4 b 30.4 d 17.0 d 27.4

II 15.9 w 15.9 7 14.8 1 25.8h 14.8 h 10.9 w 8.2 ( 113w 12.7 j 8.9 45 7.7 b 8.17 11.7 II 6.5 j 7.7 45 8.1j 9.5 v 6.5 h 7.1 j 6.5m 3.2 x 6.5 1 6.6 Cl 3.21 3.2 1 5.4 b 6.0 w . 3.2

f 2.1 7 33 1 4.4 f 1.6

P 1.0 d 2.2 t 33 h 1.6v 1.0 45 1.1 II 3.2 s 1.6

m 1.1 D 2.7 t 1.6Cl 1.1 ( 2.2z 1.1 8 1.6

3 1.6Cl 1.1v 1.1x 1.1f 0.5m 0.5r 0.5z 0.5

Because French and English share many of thesame segments, it is not clear whether the infants'productions are similar because of the similarsounds to which they were exposed, or because, asLocke (1983) would argue, certain sound patternsare particularly common in the languages of theworld because they reflect both the child's and theadult's natural phonetic proclivities. However, an

. examination of the longitudinal changes in eachinfant's productions suggests target-languageinfluence.

Most of the consonants in the American infant'sbabble repertoire (see Table 1) which showed anincrease in relative frequency over time ([<1, 6]) orwhich began to appear at 0;11 ([t, n, r. S, z)) aremore frequent in the English language than thosewhich diminished in relative frequency (lb, h, w, j])Qr disappeared altogether from the inventory ([g,

From Babbling Towards the Sound Systems ofEnglish and French: A Longitudinal Two-Case Study 47

p, v]). The American infant, CR, also introduced orincreased the relative frequency of voiced andunvoiced alveolar and palato-alveolar fricativesand affricates ([11, ", I, So 3]) at 0;11. It is possiblethat CR developed the neuromuscular ability toproduce these sounds at about this time, resultingin an increase in their production. The only non­English sound in CR's inventory is [x], whichshowed a gradual decrease over time anddisappeared entirely by 1;2. There are only twoconsonants in English Uri and !e!) that CR did notproduce.

CR's vowel productions (see Table 2) show asimilar effect. CR displayed a preference for [~],

[z], and [a] at all months, and percentages foreach remained relatively stable over time, consis­tent with their high frequency of occurrence inEnglish. The sound [e] was relatively infrequentin CR's babble repertoire longitudinally, and isalso infrequent in English. Dramatic increases inthe relative frequency ofmand [e] at 0;11 and, at1;2, increases in the frequency of [i] and decreasesin [:>] and [u] are also consistent with the relativefrequency of occurrence of these sounds inEnglish.

Table 2. List of vowels in order Of percentage Ofoccurrence at 0;5, 0;8 , 0;11, and 1;2 for E-CR(Percentage base: total number of vowels in syllablesanalyzed at each age: 178, 218, 218, and 87 at 0;5,0;8,0;11, and 1,'2, respectively).

0;5 0;8 0;11 1;2

~ 38.1 ~ 27.9 ~ 31.1 ~ 36.4Ie 19.0 a 21.4 a 27.4 Ie 17.0u 13.5 u 20.0 £ 9.0 a 12.5:) 9.2 II: 11.6 \ 8.0 \ 9.1a 8.0 £ 7.4 II: 5.7 £ 9.1£ 43 i 33 :) 5.2 i 5.7i 1.8 e 33 3.8 e 3.4

1 1.8 .. 2.8 u 3.8 :) 230 1.8 \ 1.9 e 1.9 0 23e 1.2 :) 0.5 0 1.9 u 23.. 1.2 .. 1.9

Similar language-specific patterns are alsoapparent in the repertoire of the French infant,YC. YC showed a definite preference after 0;8 forthe sound [d], which has a relatively highfrequency of occurrence in French (see Table 3),but this sound is also frequent in English andoccurred often in CR's productions as well.Nonetheless, in general, most consonants whichshowed increased frequency of occurrencelongitudinally ([b, m, D, j]) or appeared in the

inventory at 1;2 ([p]) occur more frequently inFrench than the consonant which decreased infrequency ([g]). In addition, YC initially produceda number of non-French sounds in her inventory,but these sounds diminished greatly in frequencyby 1;2 ([h]), disappeared gradually over time mm,or appear only at 0;5 or 0;8 without recurrence ([1].1, r, xl). YC did not display the increasedproduction of alveolar fricatives and palato­alveolar fricatives and affricates shown by CR at0;11, although [,,] and the palatal glide Ul didappear in her inventory at about this time.However, affricates do not exist in French, and [~did not reappear in YC's speech sample at 1;2.

Table 3. List of consonants in order of percentage ofoccurrence at 0;5, 0;8 , 0;11, and 1;2 for F-YC(Percentage base: total number of consonants insyllables analyzed: 114, 317, 201, and 249 at 0;5, 0;8,0;11, and 1;2, respectively).

Oj5 0;8 °ill Ij2

1 36.6 d 66.4 d 58.4 d 30.8d 29.3 1 12.6 j 10.3 b 16.8h 18.3 m 5.3 w 9.9 j 14.4

II 3.7 h 3.2 b 8.1 m 9.6

IJ 3.7 j 3.2 1 5.4 n 8.2m 2.4 n 2.4 oJ 3.2 P 7.7b 1.2 b 1.6 II 2.2 1 5.8r 1.2 g 1.6 h 1.1 w 3.4t 1.2 t 1.2 n 1.1 t 1.4II 1.2 II 1.2 m 0.5 g 1.0v 1.7 w 0.4 t 0.5 h 1.0

x 0.47 0.4

Table 4. List of vowels in order of percentage ofoccurrence at 0;5, 0;8 , 0;11, and 1;2 for F-YC(Percentage base: total number ofvowels in syllablesanalyzed: 114, 317, 201, and 249 at 0;5,0;8,0;11, and1,'2, respectively).

0;5 0;8 0;11 1;2

II: 44.2 a 39.6 II: 40.5 a 33.5a 26.5 Ie 27.7 i 14.1 II: 24.6~ 19.5 ~ 17.3 ~ 11.2 ~ 19.8£ 2.7 £ 4.5 a 10.7 u 8.9e 1.8 .. 3.0 £ 7.8 .. 5.6u 1.8 u 2.4 e 4.4 \ 2.0.. 1.8 \ 2.1 u 3.4 1.6

0.9 0 2.1 \ 3.0 £ 1.6

\ 0.9 i 1.2 .. 3.0 e 0.8:) 03 0 2.0 :) 0.8

0 0.8

48 Leoitt and Utman

In general, YC produced the same vowels as CR,in terms of the transcriptions assigned them. Herpreference for low front vowels ([a] and [11:]) wasstable at each of the four months sampled.However, the mid central vowel consistentlyfavored by CR ([~]), which is the most frequentvowel in English (see Table 5), remained in thirdplace for YC, and is the fourth most frequentvowel in French. The French sounds Iyl and laland the French nasal vowels did not occur inYC's productions. It is possible that these soundsfailed to emerge, and that the children's vowelinventories were similar, because of physiologicalconstraints on vowel production, such as inabilityto produce the necessary voluntary lip-roundingor the appropriate lowering of the velum.

Table 5. Lists of consonants and vowels in order ofpercentage of occurrence in English and French(Delattre, 1965 (Percentage base for vowels: totalnumber of vowels in the sample for the relevantlanguage. Percentage base for consonants: totalnumber of consonants in the sample for the relevantlanguage). Sounds not shared by the two languages areindicated by underlined type.

English Frenchconsonants vowels consonants vowels

12.7 ~ 24.1 r 15.0 e 19.3n 11.4 1 15.1 I 10.6 a 16.7r 8.3 £ 9.9 t 9.7 i 12.4

7.6 i 8.9 s 8.8 ~ 7.6s 7.4 a 7.3 p 8.0 § 7.6d 6.1 E 7.2 d 7.2 E 6.7z 4.9 u 5.9 k 6.4 u 6.4m 4.7 ai 5.7 m 6.0 :) 5.1

~ 4.6 0 5.2 n 5.2 y 4.7k 4.3 e 4.1 v 45 ~ 3.8w 3.7 !!! 2.3 j 3.2 0 2.6b 3.4 :) 2.1 3 3.0 ~ 2.4

h 3.3 ~ 2.1 f 2.6 !! 1.8v 3.2 ill .2 z 2.3 ! 1.7f 2.8 w 2.3 !! 1.0

P 2.3 b 2.3 51 .1

D 2.2 II 1.1j 2.0 I .9

II 1.6 !l .8

~ .9

I .8

m .8

I .6

3 .2

Patterns of change in relative frequency ofoccurrence for vowel sounds are not as clear forYC as they are for CR, with the exception of an

increase in the production of [u] over time, adecrease in [0], and the stable high frequency of[a] and [~], all of which correspond with therelative distributional frequency of each sound inthe Delattre data.

Overall, these longitudinal patterns ofproduction, for both infants, particularly in thecase of consonants, seem to suggest the possibilityof transition into the adult language. Indeed,when we calculate the percentage of the totalconsonantal inventory at 0;5, 0;8, 0;11, and 1;2that the children share, we see that it starts at35% at 0;5, increases to 69% at 0;8 and thendiminishes to 50% at 0;11 and 35% again at 1;2.This gradual decrease from a high of 69% overlapat 0;8 and each child's increased production of theconsonant sounds in the ambient languagesuggest language-specific developmental effects.

Some further support for language-specificpatterns emerges if we consider the children'sconsonant productions at each age in terms ofmanner (Figure 1) and and their vowelproductions in terms of vowel height andfrontness (Figure 2). Each figure also contains thedata for the appropriate adult language fromTable 5 collapsed into the same categories.

AB can be seen from Figure 1, English usesfricatives, affricates and nasals more than French,whereas approximants are more frequent inFrench. Surprisingly, both infants showedperhaps their closest match to the frequencydistributions of the manner categories of theirnative language in their vocal productions at 0;5(Figure 1). Both infants also showed a strongpreference for producing stops at 0;8 and 0;11,though more so in the case of the French infantYC. At 1;2, the infants appeared to be headingback towards a pattern that closely matches thatof their native language, except that the Frenchinfant, YC, produced very few fricatives, and theAmerican infant, CR, produced somewhat morefricatives and affricates than stops (as he did at0;5) and very few nasals. The results for CR at 1;2months are based on the smallest number ofsyllables (87), and should perhaps be regardedwith caution.

The early preference for stops on the part ofboth infants is consistent with the universaltendency for early consonant productions to bestops (Locke, 1983). Although our French infantproduced more stops, consistent with theirdistribution in adult French, than our Americaninfant, Boysson-Bardies (1989), on the other hand,found that French infants produced fewer stopsthan American infants. However, her study began

From Babbling Towards the Sound Systems ofEnglish lind French: A Longitudinll1 Two-CIIse Study 49

with older infants (0;10), followed them throughthe 25-word stage, and compared the children'sdistribution of consonants with the consonants inthe adult models of the children's target-wordsrather than the language as a whole. Thediscrepancy may thus be due to the fact that ourtwo infants were largely prelinguistic,2 whereasthose of Boysson-Bardies had made entry into thelinguistic systems of their environments.

In the case of the vowel categories, the infants'productions were divided into seven categories,front high (fh), front mid (fm), front low (£1), mid

central (mc), back high (bh), back mid (bm) andback low (bI). As can be seen from Figure 2, F-YCfavored front vowels, in particular, front low,which is consistent with the higher frequency offront as compared to central and back vowels inFrench. On the other hand, E-CR produced con­siderably more mid central vowels, which is con­sistent with their higher frequency in English.Furthermore, over the period of investigation, E­CR showed an increase in front high vowels,which is also consistent with their relatively highfrequency in adult English.

E·CR

80 • stop• friclaffric

60 • nasal~ approx.

40

20

00;5 0;8 0; 11 1;2 English

Age

F-YC

80

60

40

20

00;5 0;8 0;11 1;2 French

Age

CDUCCD~~

::JUUo

CDUCCD~~

::JUUo-oCDC)as-cCDu~

CDQ.

-oCDC)as-cCDU~

CDQ.

Figure 1. Distribution of the consonant productions of the American (E-CR) and the French (F·YC) infant as a functionof manner class at 0;5, 0;8, 0;11, and 1;1. (?) and [h) were not included in the graph. The bars at the right of each graphindicate the manner cbss distribution of consonants for each of the adult languages.

50 LeuiH and UtmDn

CDUCCD....~UUo-oCDenas-cCDU..CDD.

80

60

40

20

o0;5

E-CR

0;8 0;11Age

F-YC

1;2 English

• fh• fill• fl£'&I me0 bh• bmIi bl

CDUc!..~

UUo'5CDenas-cCDU..CDD.

80

60

40

20

o0;5 0;8 0;11

Age1;2 French

Figure 2. Distribution of the vowel productions of the American (E-CK) and the French (F-YC) infant as a function ofseven categories, front high (fh), front mid (fm), front low (£I), mid central (me), back high (bh), back mid (bm) andback low (bl) at 0;5, 0;8, 0;11, and 1;2. The bars at the right of each graph indicate the vowel category distribution foreach of the adult languages.

The percentage of each infant's consonant andvowel productions in terms of the categoriesestablished in Figures 1 and 2 was correlated withthe percentages of those categories in the adultlanguages. Table 6 shows the results of thosecorrelations. Most of the correlations themselvesare not significant (except one for F-YC at 0;5 andtwo for E-CR at 1;2), as would be expected given

the fact that the infants' consonantal and vowelrepertoires are more limited than the adults'because of developmental constraints. However, itis striking that in 14 out of 16 comparisons (p <0.002, by a binomial distribution test), the Frenchinfant's productions correlate more strongly thanthe American infant's productions with the Frenchconsonant and vowel categories, whereas the

From Babbling towards the Sound Systems ofEnglish and French: A Longitudinal Two-Case Study 51

American infant's productions correlate morestrongly than the French infant's productions withthe English consonant and vowel categories.

Table 6. Correlations between the infant consonantproductions and those ofEnglish and French in termsof manner categories and between infant vowelproductions and those ofEnglish and French in termsofvowel height andfrontness.

Formant analyses ofthe· vowels. Mean F1 and F2frequencies and standard deviations for all vowelsat each age were plotted for both infants (seeFigure 3) and compared with the findings ofBoysson-Bardies, Halle, Sagart, and Durand(1989) for English and French infants. Of the 178,218, 218, and 87 syllables analyzed at 0;5, 0;8,0;11, and 1;2, respectively, for E-CR, 26, 38, 33,

Age

0;50;80;111;2

Age

0;50;80;111;2

.p <0.05

French ConsonantsF-YC E-CR

0.99· 0560.67 0.940.81 0.780.61 034

French VowelsF-YC E-CR

032 -0.03031 0.26050 0320.25 031

English ConsonantsE-CR F-YC

0.88 0.470.81 0.490.94 0.520.96· 0.23

English VowelsE-CR F-YC

0.51 0.370.53 0.360.72 0.500.82· 0.39

and 14 syllables gave indeterminate F1 or F2measurements and were excluded from theanalyses. Similarly, of the 114, 317, 201, and 249syllables analyzed at 0;5, 0;8, 0;11, and 1;2,respectively, for F-YC, 9, 31, 35 and 65 syllables ateach of the four ages gave indeterminate F1 or F2measurements and were excluded from the analy­ses. Mean F2 for the English-learning child, E-CR,remained relatively stable until 0;11 when it be­gan to increase. Mean F2 for the French-learningchild, F-YC, remained relatively stable until 1;2when it began to decrease. At 1;2 the mean F2 forE-CR was higher than that of F-YC. This patternis consistent with the higher mean F2 found byBoysson-Bardies et al. (1989) in the productions ofEnglish-learning infants when compared to thoseof French-learning infants, although this differ­ence appeared in their data at 0;10.

Mean F1 remained fairly stable for bothchildren, with the mean F1 value of the Americanchild, E-CR, increasing slightly over the fourmonths sampled, and the mean F1 value of theFrench child, F-YC, decreasing slightly over thesame period. Overall, the mean F1 was lower forE-CR than for F-YC, but the average valuesappear to be converging, while in the case of F2,they seem to begin to diverge. Boysson-Bardies etal. found essentially identical measurements formean F1 for their sample of English and Frenchinfants, so this convergence is perhaps to beexpected. Although language-specific differencesoccurred later in this analysis than in that ofBoysson-Bardies et al., they nonetheless appear tosuggest the influence of the language environmenton the acoustic properties of vowels for bothinfants.s

3000

t t 1>";c F1 (E-cR)

N 0 F2 (E-CR):x:: -.- F1 (F-YC)c --..- F2 (F-YC)->- 2000u

tc

t ~- ~CD~

cr 1000CD...U.

00;5 0;8 0;11 1 ;2

Age

Figure 3. Mean F1 and F2 and standard deviations for the American (EoCR) and the French (F-YO infant at 0;5, 0;8, 0;11,and 1;2.

52 Ler1iH and Utman

Boysson-Bardies et al. (1989) have argued thatfrom about 0;10 of age, infants have sufficient ar­ticulatory control to produce vocal tract positionsthat result in sound productions that reflect theambient language. We explored the possibility ofsuch an influence by examining the formant fre­quency patterns for ten vowels produced by thetwo infants. Mean and standard deviation of F1and F2 for the ten vowel sounds were plotted for

each infant at each age (see Figures 4 & 5). In theadult languages, F1 generally varies inverselywith vowel height, whereas F2 is generally higherfor front vowels and lower for back vowels. F1 canthus often be correlated with correct jaw control(or tongue height), whereas F2 usually corre­sponds to tongue fronting (but cf. Ladefoged(1975) for some qualifications on inferring tongueposition from F2 for back vowels).

E-GR

0;5 0;84000 4000

3250

IIIHI II3250·

!I f I2500 T 2500 -

I I IN •I I !

t::I Fl

1750 1750 • F2

1000

~ ~ f 2 ! 1000 £ ~ ~ ~ !f!j ~ 2 ~ m ~t::I

D250 250

1. ecrea;;l::Jou i 1. e c re a ;;I ::J 0 U

0;11 1;2

i1. ecrea;;l::Jou

•1000 -

1750

4000 -,------------,

:::-II I i I I I . !! j !I ;j; iI 1m

4000 -,-------------,

i1.ecrea;;l::Jou

1750

1000

2500 -

3250

250

NI

Vowel Vowel

Figure 4. Mean F1 and F2 and standard deviations for 10 transcribed vowels ( [8] was not included) for the American (E·em infant at 0;5, 0;8, 0;11, and 1;2.

From Babbling towards the Sound Systems ofEnglish and French: A Longitudinal Two-Case Study 53

F-YC

0;5 0;84000 4000

3250

I II !3250

I I! II I I·!2500 2500N

l:l F1I •1750

m if H t1750 • F2

III I Iff ! ~ f • f1000 1000

250 250i 1 e c ~ a a ~ 0 U 1 e c ~ a a ~ 0 U

0;11 1;24000 4000

3250

LII I I I !3250 pI!

2500 •! I II I!

2500N

I1750 -

I J t ! ! i !1750

f ~1000 - t ~ ~. ~ q d! ~1000

250 2501 e c ~ a a :l 0 U 1 e c Ee a a :l 0 U

Vowel Vowel

Figure 5. Mean F1 and F2 and standard deviations for 10 transcribed vowels ( (8] was not included) for the French (F·YO infant at O;S, 0;8, 0;11, and 1;2.

54 Leuitt and Utman

Given the appropriate patterning of F1 inFigure 4, and in particular the low variance in F1at the first three ages, CR, the English infant,may have developed better jaw control first. Theincreased variance in F1 at 1;2 for CR appears tobe associated with a reduction in variance in F2,which suggests that CR may have begunmastering the articulatory patterns for F2 at thattime. On the other hand, given the regularpatterning ofF2 in Figure 5 from 0;11 months on,YC may have mastered tongue fronting first. Thiswould be reasonable, given the importance oftongue fronting in French (Delattre, 1965).Furthermore, there is a notable reduction in thevariance in F1 and/or F2 in YC's productions from0;11 to 1;2 in the four vowels that are mostfrequent in French (Ie, a, i, ~.

In order to test whether F-YC produced lessvariable F2's and E-CR less variable F1's, weconducted an analysis ofvariance on the log trans­forms of the standard error of the means for eachvowel's F1 and F2 measurements at each time pe­riod. The analysis contained one between factor(infant language source), with each of the tenvowels treated as subjects, and two within factors(time period and formant number). There was asignificant effect of formant number [F(l,18)=27.525, p =0.0001] with F1 means overall lessvariable than F2 means. There was also asignificant interaction of formant number bylanguage source [F(l,18)=9.099, p = 0.0074],which showed that F1 measures for E-CR wereless variable than those for F-YC, whereas F2measures for F-YC were less variable than thosefor E-CR. There were no other significant effectsor interactions.

Thus, the analyses of the formant patternssuggest that both infants appeared to developmore adult-like vowel formant frequency patternsover time, presumably as a result of theirincreasing articulatory control in producingvowels. There also is some evidence for theinfluence of language-specific articulatory effectsin the pattern of variability for F1 and F2 , whichreflects the particular ways in which each infantbegan to demonstrate increasing articulatorycontrol.

Syllable characteristics. All of the two-syllable orlonger utterances that contained at least twoconsonant- and vowel-like segments (e.g., CVCVor VCVC) were tallied for each infant at each age.Figure 6 provides some support for the stages ofStark (1980) and Oller (1980) in terms of the onsetof reduplicative babbling (for CR) and the periodduring which variegated babbling becomes more

frequent (for both infants). Thus, the Americaninfant, CR, produced reduplicative or canonicalbabbling at 0;8 but not before. Both Stark andOller suggest that such babbling should beginbetween 0;6 and 0;7, so CR's performance isconsistent with this stage in their developmentalschemes. On the other hand, the French infant,YC, actually produced some reduplicativeutterances in the samples at 0;5, but also showeda great increase in production of such babblesbetween 0;5 and 0;8. The speech samples of bothchildren yielded a fairly level number ofreduplicative babbles from 0;8 on, with the Frenchinfant producing a consistently higher number.

Stark and Oller predict that variegated babblingshould begin between 0;10 and 0;11 months. Bothinfants produced variegated babbles before the 11­month sample, but they both showed the greatestnumber of variegated babbles at 0;11. The factthat both reduplicative (in the case of YC) andvariegated babbling (in the cases of both infants)appeared earlier than expected is consistent withthe report of Smith, Brown-Sweeney, and Stoel­Gammon (1989), who also found that both redu­plicative and variegated productions co-occurredin the babbling of the ten American infants thatthey studied at four month intervals from 0;6 to1;6 months of age. (See also Mitchell & Kent,1990, for similar observations).

We conducted an analysis of variance on thenumber of reduplicated and variegated babbleswith one between factor (infant language source)and one within factor (babbling type - reduplicatedor variegated), treating each of the four time peri­ods as subjects. There was a significant effect ofinfant language source [F(1,6)=7.670, p = 0.0324],reflecting the higher incidence of multisyllabicbabbling in the French infant's speech. There wasalso a significant effect of babbling type [F(1,6) =10.826, p = 0.0176], since there were more varie­gated than reduplicated utterances overall. Therewas no infant language source by babbling typeinteraction. Thus, although the French infantproduced more multisyllabic babbles, the patternof variegated and reduplicative babbling was es­sentially the same for both children.

The transcription data were also used to deter­mine the number of open and closed syllablesproduced by both infants at each age. Thepercentage of closed syllables in the productions ofthe American infant (CR) dramatically increasedat 0;11, while percentages for the French infantremained low and relatively stable (see Figure 7).A chi-squared test on the frequency distributionsof open and closed syllables was significant in the

From Babbling tOWtlrds the Sound Systems ofEnglish and French: A Longitudinal Two-Cilse Study 55

case of E-CR (X2 (3)=33.648, P = 0.0001) but not inthe case of F-YC (X2 (3)=5.983, p=0.1124). Thisdiscrepancy is consistent with the greater fre­quency of closed syllables in English than in

French (Delattre, 1965) and suggests that by 0;11CR had begun to produce a distribution of syllabletypes that is characteristic of his languageenvironment.

E·CR

50fI) • ReduplicatedCDuc 40 • VariegatedCG..CD... 30...::;)-0 20..CD.Q

10E~z

00;5 0;8 0; 11 1;2

Age

F·YC

50U)CDu 40cCG..CD... 30...::;)-0 20..CD.Q

10E~z

00;5 0;8 0; 11 1;2

Age

Figure 6. Number of reduplicated and variegated utterances produced by the English (E-CK) and the French (F-YOinfants at 0;5, 0;8, 0;11, and 1;2. The English infant (E-CK) did not produce any reduplicated utterances at 0;5.

58 Leuitt and Ulman

E·CR- • 1st PositionflO600E • 2nd Position-c 500 • 3rd Position

0;:as 400..;:,Q

• 300:c

200.!!>.0 100cas 0•:E 0;5 0;8 0; 11 1;2

Age

F·YC-en 500E-c 4000;:as..

300;:,Q

• 200:c.!!->- 1000

cas 0Q)

:E 0;5 0;8 0;11 1 ;2Age

Figure 9. Mean syllable duration in rns for syllables in the first. second, or third position in the utterances produced bythe English (E-CR) and the French (F-YO infant at 0;5, 0;8, 0;11, and 1;2.

Final syllable lengthening in the present studywas examined in terms of mean syllable length forfinal and non-final syllables in reduplicativestrings, and in terms of the number of reduplica­tive strings which showed final syllable lengthen­ing. Reduplicative babbling was used in order toeliminate differences in syllable lengths due to theeffects of segments of intrinsically differentlengths. Final syllables for both infants werefound to be longer than non-final syllables at 0;11and 1;2 (see Figure 10).

However, when the percentage of reduplicativetokens which have longer final than penultimate

syllables is plotted for each infant at each age, theFrench infant showed a gradual increase in thepercentage of utterances with final syllablelengthening, while the American infant showed adrop at 1;2 (see Figure 11). The difference in thenumber ofutterances with final-syllable lengthen­ing for the two children at each of the four ageswas significant (t(3) = 2.377,p = 0.049, one-tailed).Regular final syllable lengthening is a more strik­ing characteristic of the French language, wherethe final syllable is the only one lengthened,than English, where syllables that bear stress, aswell as those that are final, can be lengthened.

From Babbling towards the Sound Systems ofEnglish lind French: A Longitudinal Two-Case Study 59

E-eR600 • Final

-0-- Non-Final-en'tJ-S 500CDCa;ou;

400=asa. ...~ ~'CoCDCI)a:- 300c'ico=CD>-::ECI)

2000;5 0;8 0;11 1;2

Age

F-YC

600-cnE,,- 500CDC;0

u;=coa. ... 400~~,,0CDD:.!

300e.gco.!!CD-::E~

2000;5 0;8 0; 11 1;2

Age

Figure 10. Mean syllable duration in DIS of final venus nonfinalsyllables of the reduplicated utterances produced bythe English (E-eR) and the French (F·YO infant at 0;5, &,8, 0;11, and 1;2. The English infant (E-eR) did not produce anyreduplicated utterances at 0;5.

These data on final syllable lengthening are con­sistent Konopczynski's (1986) finding of final syl­lable lengthening in the variegated and reduplica­tive babbling of French children between the agesof 0;8 and 2;0 months and with Oller and Smith's(1977) finding of little final syllable lengthening in

the reduplicative babbling of six American infantsaged 0;8 to 1;0. However, as is evident from Figure6, the American infant produced very few redu­plicative utterances overall, so that the presentresults, while consistent with previous research,must be interpreted with caution.

60 LerriH and Utman

10 -+----.-__-....-__-_..__.........r---__

E-eRF·YCI~

1;20;8 0;11Age

0;5

50

40

30

20

70

60

Figure 11. Percenuge of reduplicated utterances with finallyllable lengthening produced by the English (EoCR) andthe French (F-YO infant at 0;5, 0;8, 0;11, and 1;2. The English infant (EoCR) did not produce any reduplicated utterancesat 0;5.

GENERAL DISCUSSION

Analyses of the babbling of the two infants inthis study showed clear evidence both of similardevelopmental patterns, presumably due touniversal developmental processes in the twoinfants, and of the prelinguistic influence of theadult language environment.

Evidence of general developmental effects is ap­parent in both the acoustic analyses and in theexamination of segmental inventories. Bothinfants appeared to produce target F1 and F2 withless variability as they got older, suggestingincreased articulatory control. Both infantsproduced 16 of the 23 consonant-like sounds andall 11 vowel sounds which compose the totalcorpus, and they generally favored stopconsonants over other types. These findings areconsistent with Locke's (1983) hypothesis ofuniversal phonological development. Finally, theanalysis of the multisyllabic babbling of bothchildren provided some evidence in support of thedevelopmental stages of babbling as described byboth Stark (1980) and Oller (1980), although thepresence of reduplicative and variegated babblingprior to their predicted onsets is consistent withthe findings of Smith, Brown-Sweeney & Stoel­Gammon (1989) and Mitchell and Kent (1990).

The evidence of language-specific effects is alsoclear, although in most cases these effects appearin longitudinal analyses at the ll-month sampleor later for both infants. Each infant's phonetic

inventory began to resemble that of the adult lan­guage both in composition and frequency, so thatover the course of time, the two infants' phoneticinventories, in particular their consonantal inven­tories, diverged. Vihman, Ferguson, and Elbert(1986) found that the phonetic tendencies of 10American infants aged 0;9 to 1;4 tended to con­verge over the period studied. Both the Vihman etal. finding of increased similarity for infantslearning the same language and the divergencefound in the present analysis for infants learningtwo different languages suggest the influence ofthe language environment on the infantproductions.

Measurements of the acoustic properties of thevowels also revealed some language-specific ef­fects. Mean F1 for all syllables for both infantsappeared to converge over time, whereas mean F2for the American infant, CR, displayed an increaseof about 200 Hz at 1;2, corresponding with a dropof about 300 Hz in mean F2 for the French infant,YC, at that time, so that mean F2 is higher for CRthan for YC at 1;2. These F1 and F2 patterns areconsistent with the results of a similar analysis byBoysson-Bardies, Halle, Sagart, and Durand(1989), who found that mean F1 measurementswere essentially the same for English-learningand French-learning infants, whereas mean F2measurements were higher for the former group.Furthermore, the patterns found for F1 and F2 inthe infant groups reflected similar F1IF2 relation­ships in the data from the adult languages.

From Babbling towards the Sound Systems ofEnglish and French: A Longitudinal Two-Olse Study 61

Although this pattern was evident for the babblingoften-month-old infants in the Boysson-Bardies etal. study, it only emerged in the 14-month samplein our data.

In terms of syllable characteristics, theAmerican infant produced a larger percentage ofclosed syllables than the French at 0;11 and 1;2,and showed a dramatic increase longitudinally inthe percentage of closed syllables produced, whilepercentages of closed syllables for the Frenchinfant remained stable over time. Because Englishhas more closed syllables than French, this resultsuggests the influence of English on CR'sbabbling.

Analyses of timing patterns in the babbling ofthe two infants revealed evidence of the character­istic isosyllabicity of French for F-YC at all ages:when the durations of nonfinal syllables were av­eraged for each position in the utterance at eachage, mean syllable duration tended to be less vari­able for the French infant than for the Americaninfant. In addition, although both infants showedsome final syllable lengthening at 0;11 and 1;2, alarger percentage of the French infant's redu­plicative utterances showed final syllable length­ening, suggesting a more pronounced influence ofthe French rhythmic pattern on F-YC's babbling.

In general, our results agree with recent find­ings (Vihman et al., 1986; Boysson-Bardies, 1989;Boysson-Bardies et al., 1989 ) of language-specificinfluences on the segmental characteristics ofbabbling towards the end of the first year.However, some of the timing differences thatemerged in this study, particularly the suggestionof an early development of isosyllabicity in thebabbling of the French infant, suggest that thelanguage environment may affect the prosody ofbabbling even earlier than its segmental charac­teristics (see also Whalen, Levitt, & Wang, 1991).Clearly the prosody of infant babbling warrantsfurther cross-linguistic research.

REFERENCESAtkinson, K. B., MacWhinney, B., Ie Stoel, C. (1970). An

experiment in the recognition of babbling. Pllpos lind Reports inChild LAngutlge 1JerJdopment, 1, 71-76.

Bond, Z. S., Petrosino, L., Ie Dean, C. R. (1982). The emergence ofvowels: 17 to 26 months. JOU17IIIl ofPhonetics, 10,417-422.

Boysson-Bardies, B. de (1989) Material evidence of infant selectionfrom the target language: a cross-linguistic phonetic study.Paper presented at the Conference on PhonologicalDevelopment, Stanford University, September, 1989.

BoyssorrBardies, B. de, Halle, P., Sagart, L., Ie Durand. C. (1989).A cross-linguistic investigation of vowel formants in babbling.Jountlll ofChild LAngutlge, 16, 1-17.

Boysson-Bardies, B. de, Sagart, L., Ie Durand, C. (1984).Discernible differences in the babbling of infants according totarget language. JOlIntIII ofChild LAngutlge, 11, 1-15.

Buhr, R. (1980). The emergence of vowels in an infant. JOU17IIIl ofSp«dI "nd Hearing Resmrdl, 23, ~9'.

Brown. R. (1958). Words tmd things. New York: Free Press.De1attre, P. (1965). Qnnpllring the phonetic felltures of English,

GmrIm, Spiltlish "nd Frmch. Heidelberg: Julius Groos.Dinger, M. C., Ie 810m, J. G. (1973). An investigation of infant

babbling. Proc«dings frum the Institute of Phonetic Scienas(University of Amsterdam), 3, 42-50.

Ferguson. C. (1978). Learning to pronounce: The earliest stages ofphonological development in the child.ln F.D. Minifie &: L. LUoyd (Eds.), CommuniCiltion lind cognitive IIbilities-fArlybeIuroiorrd_to Baltimore: University Park Press.

Jakobson, R. (1968). Child IllngUllge, "phllsill, lind phonologiCilllDIimwls (A. Keller, Trans.). The Hague: Mouton.

Kent, R. D., Ie Bauer, H. R. (1985). Vocalizations of one-year-olds.JOUntIIl ofChild LAngutlge, 12, 491-526.

Konopczynsld. G. (1986). Vers un modele developpemental durhythme Fr~ Problemes d'isochronie reconsideres a lalumiere des donnees de l'acquisition du langage. Bulletin del'Institut de PhonItique de Grmoble, 15, 157-190.

Ladefoged. P. (1975). A aJUTSe in phonetics. New York: HarcourtBrace Jovanovich.

Locke, J. (1983). PhonologiCIIlllcquisition lind chllnge. New York:Academic Press.

Mack, M., Ie Ueberman, P. (1985). Acoustic analysis of wordsproduced by a child from 46 to 149 weeks. JOlIntIIl of ChildulngUllge, 12, 527-550.

Macken, M. A., Ie Barton, D. (1980). The acquisition of the voicingcontrast in English: A study of voice onset time in word-initialstop consonants. JOlIntIIl ofChild l.IInKUIIge, 7,41-75.

Menn, L (1971). Phonotactic rules in beginning speech. UnKUII, 26,225-241.

Menyuk. P. (1971). The acquisition and development of language.Englewood Cliffs, NJ: Prentice-Hall.

Menyuk. P., &: Menn, L. (1979). Early strategies for the perceptionand production of words and sounds. In P. Fletcher Ie M.Garman (Eds.), l.IIngutlge IIcquisition. New York: CambridgeUniversity Press.

Mitchell, P. R., .Ie Kent, R. D. (1990). Phonetic variation inmultisyllabic babbling. Jountlll ofChild LAngutlge, 17, 147-165.

Oller, D. K. (1980). The emergence of the sounds of speech ininfancy. In G. H. Yeni-Komshian, J. F. Kavanaugh, Ie C. A.Ferguson (Eds.), Child p1lmology, Vol. 1: Produdion. New York:Academic Press.

Oller, D. K., &: Smith,. L. (1977). Effect of final-syllable position ofvowel duration in infant babbling. JOlIntIIl of the AcoustiCillSociety ofAmericII, 62, 994-997.

Oller, D. K., Wieman, L. A., Doyle, W. J., Ie Ross, C. (1976). Infantbabbling and speech. Jounu" ofChild Ungutlge, 3,1-11.

Olney, R., Ie Scholnick, E. (1976). Adult judgments of age andlinguistic differences. in infant vocalizations. Jountlll of Childulngutlge,3,145-15O.

Pike, K. (1945). Intonlltion of AmmClln English. Ann Arbor, MI:University of Michigan Press.

Smith,. B. L., Brown-Sweeney, S., Ie Stoel-Gammon, C. (1989). Aquantitative analysis of reduplicated and variegated babbling.First UnKUIIge,9, 175-190.

Stark. R. (1980). Stages of speech development in the first year oflife. In G. H. Yeni-Komshian, J. F. Kavanaugh, Ie C. A.Ferguson (Eds.), Child plxmology, Vol. 1: Produdion. New York:Academic Press.

Thevenin, D. M., Ellers, R. E., Oller, D. K., Ie Lavoie, L. (1985).Where's the drift in babbling drift. A cross-linguistic study.Applied Psycholinguistics, 6, 3-15.

Vihman, M., Macken, M., Miller, R., Simmons, H., Ie Miller, J.(1985). From babbling to speech: A re-assessment of thecontinuity issue. UngUIIge, 61, 397-443.

62 LerJiH and Ulman

Vihman, M., Ferguson, C., ok Elbert, M. (1986). Phonologicaldevelopment from babbling to speech: Common tendenciesand individual differences. Applied Psycholinguistics, 7, 3-40.

Weir, R. (1966). Some questions on the child's learning ofphonology. In F. Smith ok C. A. Miller (Eds.), Tile genesis uflIIngutlge. Cambridge, MA: MIT Press.

Wenk, B. J., ok Wioland. F. (1982).15 French really syllable-timed?JOIlnvll ufPhondial, 10, 193-216.

Whalen, D. H., Levitt, A. C. ok Wang, Q. (1991). Intonationaldifferences between the reduplicative babbling of French- andEnglish-learning infants. JOllnllll ufChild Umgutlge, 18, 501-516.

FOOTNOTES·JOIlnvII ufCIriId umgutlge, in press.tAlso Wellesley College.

ttwellesley College. Now at Brown University.lThe two transcribers of the vocalizations used in this analysis

are both native speakers of English who are fluently bilingualin French. However, it is possible that when the data for the

French infant were amsidered, the transcriptions of ambiguoussounds, particularly vowels, might be interpreted in favor ofEnglish targets. For example, the vowel sound of 'beet,' whenshort in duration, might be interpreted as the vowel sound of'bit: • frequent sound in English, but one which does nor existin French. Of course, any such bias would lead to conservativeestimates of Ianguage-specifc di.ffere:nces between the infants.

2nle data at 1;2 may have included some words with the babble,because it was not possible to identify and exclude words onthe basis of the parents' written comments during that month.However, we observe that the pattern here, and othersdesaibed later, in general begin at 0;11 or earlier, and infantswere clearly prelinguistic then. Thus, we do not feel that thepossible inclusion of some words in the data at 1;2substantially changes our pattern of results.

30Jhe mean formant frequencies for English in Boysson-Bardieset al. (1989) were calculated for British English. We areassuming that the values for American English would bearessentially the same relationship to the other languages as didBritish English, although the actual means might differ.