of testing and heard English spoken at least 90% of the time. Twenty-four infants ... vowel was presented 5 dB(A) louder than the background. Thus, both the ...
Fundamental Frequency Influences Vowel Discrimination in 6-Month-Old Infants Renée N. Desjardins and Laurel J. Trainor Department of Psychology, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4K1 In tro d u c tio n T he ability to discrim inate different vow el tokens found in o n e ’s native language is a fundam ental skill necessary fo r the acquisition o f language. P revious research suggests that young infants are able to discrim inate readily betw een vow el categories (e.g., P olka & W erker, 1994; S w oboda, M orse & Leavitt, 1976; Sw oboda, Kass, M orse & L eavitt, 1978) as y oung as 8-w eeks o f age. H ow ever, previous research has relied prim arily on the use o f tokens from a m ale speaker, possibly b ecau se these are easier to analyze using com m ercial speech analysis program s and to produce synthetically. M ore o v e r, vow el discrim ination studies have typically em ployed tokens in w hich the fundam ental frequency (F0) rem ains constant o v e r the token. O ne exception, is an experim ent by Kuhl and M ille r (1982) in w hich they used both m o notone and falling pitch contours w ith a m ale speaker. Infants w ere able to ignore variations in pitch w hile attending to changes in the vow el category. U n like in the vow el perception experim ents, speech to infants v aries significantly in F 0 (e.g., F e m ald & Sim on, 1984; Papousek, P apousek & Haekel, 1987) T h ese variations in F0 as w ell as an overall increase in F0 are notable characteristics o f infant-directed speech. Infants show a strong preference, both attentional and affective, for infant-directed speech over adult-directed speech even w hen the passages are in a non-native language (e.g., W erker, P e g g & M cL eod, 1994). T his experim ent w as designed to study the effects o f a fem ale FO on the discrim ination o f tw o vow el tokens, HI and III, both fo und in the English lan g u ag e and both k now n to be discrim inable to very young infants (e.g., Sw oboda et al, 1976; 1978). W e chose to exam ine three different fundam ental frequencies: the low steady FO o f a typical fem ale voice in adult-directed speech, 250 Hz; the FO m ore typical o f speech directed to infants, 350 Hz; and a falling FO rem iniscent o f infant-directed pitch contours, 350 H z to 250 Hz. M e th o d S u b je c ts. Seventy-tw o infants (36 m ale and 36 female) b etw een 6 m onths 0 days and 6 m onths 30 days w e re tested. All infants had no reported hearing problem s, w ere healthy at the tim e o f testing and heard E nglish spoken at least 90% o f the time. T w e n ty -fo u r infants w ere tested at each o f three fundam ental frequencies; 12 infants in each condition had /i/ as the background token and III as the change token and 12 infants had III as the b a ckground and III as the ch an g e token. S tim u li. Synthetic tokens o f tw o E nglish vow els h i (as in “h e e d ”) and III (as in “h id ”) w ere produced using SenSyn (Sensim etrics C orporation) fo r the M acintosh on a M acintosh Ilci com puter. V alues for the first 3 form ants (F I , F 2 & F3) w ere taken from the average values for w o m e n ’s voices as m easured by Peterson and B arney (1952). V alues for F 4 w ere determ ined using S y rd a l’s (1985) form ula. B andw idths w ere taken from D unn (1961). B oth vow el tokens w e re synthesized at three different fundam ental frequencies. In the falling condition the F 0 began at 350 H z and fell linearly to 2 5 0 H z over the duration o f the vow el (500 m s). In the high steady condition, the F 0 rem ained at 350 Hz for the duration o f the vow el and in the low steady condition, the F 0 rem ained at 250 H z for the duration o f the vowel. T okens w ere equalized for loudness by 3 adult listeners and presented at an average o f 62 dB (A) in a quiet environm ent. A p p a r a tu s . Infants w ere tested in a sound-attenuating ch am b e r (Industrial A coustics Co.). A M acintosh Ilci com puter w ith an audiom edia card presented the stim uli. A Straw berry T ree I/O card connected to a custom interface box w as used to operate a -
button box, lights, and m echanical toys located in the soundbooth. Stim uli w ere presented via a D ennon am plifier (P M A 480) to a single G SI loudspeaker located 90° to the in fan t’s left. A box w ith a sm oked P lexiglass front w as located u n derneath the speaker. D uring reinforcem ent (see below ) lights and a toy in one o f the four com partm ents in the b o x be ca m e visible; otherw ise the contents w ere hidden. P r o c e d u r e . Infants w ere tested individually in a go/no-go conditioned head turn procedure. T h e infant sat on h is/h er p a ren t’s lap across from an experim enter w ho sat b e h in d a sm all table. B oth the parent and the experim enter listened to m ask in g m usic through headphones for the duration o f the experim ent. W h en the infant was attentive and facing forw ard (tow ards the experim enter), the experim enter pressed at bu tto n to indicate that a trial w as to begin. D uring both training and testing phases the b a ckground vow el was presented continuously repeating e very 2 s. D uring the training phase, the ex p erim en ter fam iliarized the infant to the contingency b etw een a head turn tow ards the speaker w hen a change in the vow el occurred and the anim ated toy reinforcer. O nly change trials occurred during training and the vow el w as presented 5 dB (A ) lo u d er than the background. T hus, both the vow el token and loudness w ere cues that indicated that a change had occurred. Infants w ere required to m ak e four consecutive correct responses w ithin 20 trials in o rder to proceed to the testing phase. A head tu rn was considered correct if it occurred w ithin 2 s o f the onset o f the c hange and if the in fan t turned at least 45° to the left to face the speaker. D uring the testing phase, a total o f 24 trials w ere presented (12 change and 12 no change control trials, in random order). O nly correct head turns w ere reinforced; the e xperim enter, w ho was unaw are o f w hat the infant w as hearing, recorded by a bu tto n press w henever a head turn occurred. R e su lts F o r each infant an A ’ score w as com puted. (A ’ is a nonparam etric equivalent o f d ’ and m ay b e b etter suited fo r infant studies in w hich only a sm all n u m ber o f trials is u se d (G rier, 1971).) A ’ scores w ere subm itted to an A nalysis o f V arian ce w ith 2 betw een-groups factors: F 0 (falling, high steady, low steady) and D irection o f vow el change (/I/ to /i/, and III to III). A significant m ain effect for F0 was obtained, F(2, 12) = 3.76, p < .03. Protected /-tests w e re perfo rm ed in o rd e r to determ ine w hich conditions differed. Infants in the low steady and falling F0 conditions obtained h igher A ’ scores on average than did infants in the high steady conditions, p < .02, p < .03, respectively. T hese results are sum m arized in F ig u re 1. T he m ain effect o f D irection o f ch an g e w as n o t significant. T h e interaction betw een F 0 and D irection o f C h a n g e w as also not significant. T-tests w ere also com puted in order to determ ine w hether in fan ts’ perform ance in each o f the three F 0 conditions differed significantly from chance (A ’ o f .50) in their A ’ scores. In the falling fundam ental condition, infants w ho had III as the background token perform ed significantly above chance, t (11) = 2.949, p < .007. H ow ever, infants w ho had I'll as the b a ckground token perform ed at chance, / ( l 1) = 1.072, p > .15. A ’ scores for these tw o groups are plotted in F igure 2. In the low steady F0 condition, infants obtained A ’ scores significantly above chance in both b a ckground vow el conditions: III background, t( 11) = 3.197, p < .005, and HI backgro und, t (11) = 2.874, p < .008. A ’ scores for the low steady groups are plotted in F igure 3. In the high steady F0 condition, perfo rm a n ce did n o t exceed c hance for either o f the b a ckground vow el conditions: III
96
-
background, both p s > .15. The A ’ scores for the high steady groups are plotted in Figure 4. D iscussion Taken together the results o f this experiment suggest that discrimination o f /i/ and HI vowel tokens for 6-month-old infants is possible when the FO is low but steady (250 Hz), and when the F0 is falling (from 350 Hz to 250 Hz)— but only when /I/ is the background token. However, with a high steady FO (350 Hz), infants perform at chance suggesting that they are unable to discriminate the two vowel tokens. These findings are surprising given that the /i/-/I/ distinction is one that infants as young as 8 weeks o f age can discriminate (e.g. Swoboda et al, 1976; 1978). However, these studies employed only steady FO tokens in the range of a typical male speaker (i.e., approximately 120 Hz). The infants in the lowest F0 condition in our experiment — 250 Hz— were also able to easily discriminate the two tokens. The 250 H z fundam ental is within the range of a w om an’s F0 in normal adult conversation. The falling fundamental and high steady F0 conditions correspond m ore closely to the pitch o f a female speaker’s voice when directed to an infant. Maternal speech to infants can range in pitch up to at least 600 Hz at times (Fem ald & Simon, 1984; Papousek, et al, 1987), yet our results suggest that a high steady F0 makes it difficult for infants to discriminate two vowel tokens. The wide pitch excursions found in infant-directed speech correspond m ost closely to the falling F0 condition in this experiment. Infants were able to discriminate the falling fundam ental tokens o f /I/ and H I, but when H I was the background token. Perhaps, then, the wide pitch excursions of infant-directed speech com pensate for the difficulty o f discrimination o f vowels posed by the use of high fundamental frequencies. In sum, these findings highlight the necessity for examining more closely the effects o f FO on vowel discrimination in infants.
The results suggest that high FO has a negative effect on 6-monthold infants’ ability to discriminate two vowel tokens, III and HI. Further research is necessary in order to determine whether features such as jitter (frequency variation from cycle to cycle of the wave form) normally occurring in natural tokens, would make discrimination o f high steady tokens less difficult. R eferences Fem ald, A. & Simon, T. (1984). Expanded intonation contours in m others’ speech to newborns. Developm ental Psychology. 20. 104-113. Grier, J.B. (1971). Nonparametric Indexes for sensitivity and bias; Computing formulas. Psychological B ulletin. 7 5 . 424-429. Kuhl, P.K. & Miller, J.D. (1982). Discrim ination o f auditory target dimensions in the presence or absence o f variation in a second dimension by infants. Perception & Psychophysics. 31. 279292. Pasousek, M., Papousek, H. & Haekel, M. (1987). Didactic adjustments in fathers’ and m others’ speech to their 3-monthold infants. Journal of Psvcholinguistic R esearch. 16. 491-516. Polka, L & Werker, J.F. (1994). Developm ental changes in perception of nonnative vowel contrasts. Journal of Experimental Psychology: Human Perception and Perform ance. 20, 421-435. Swoboda, P.J., Kass, J., Morse, P.A., & Leavitt, L.A. (1978). M em ory factors in vowel discrimination of normal and at-risk infants. Child Developm ent. 49, 332-339. Swoboda, P., Morse, P. & Leavitt, L. (1976). Continuous vowel discrimination in normal and at-risk infants. Child Development. 47, 459-465. Werker, J.F., Pegg, J.E. & McLeod, P.J. (1994). A cross-language investigation o f infant preference for infant-directed communication. Infant Behavior & D evelopm ent. 17. 323-333
0.8 0.7
0.6 0 .5 1— [ f o i # # # )------ --------------------------’ A '
--------
chance
0 4
0.3
0.2 0.1 0.0
Falling
Low Steady
High Steady
Pitch
Figure 3. A' scores in the low steady pitch condition.
Figure 1. A' scores by pitch condition o.i
0.7
p < .007 n.s.
I 0.6
m m Lm
0.5
chance
0.4 0.3
0.2 0.1 0.0
/I/
m
IM
in Background vowel
Background vowel
Figure 4. A' scores in the high steady pitch condition
Figure 2. A' scores in the falling pitch condition
-
97
-
ISSN 0711-6659
Canadian acou/tia
acou/tique canadienne SEPTEMBER 1998__________________________________________________________ SEPTEMBRE 1998 Volume 26 -- Number 3__________________________________________________ Volume 26 -- Numéro 3 M e s s a g e f r o m t h e pr e s id e n t / Un m o t d u Pr é s id e n t
1
P r o c e e d i n g s o f A c o u s t i c s W e e k i n C a n a d a 1998 / A c t e s d e l a S e m a i n e C a n a d i e n n e d 'A c o u s t i q u e 1998
Active Noise Control / Contrôle actif du bruit
4
Architectural Acoustics / Acoustique architecturale
14
Acoustics in Telecommunications/Acoustique dans les télécommunications Vibrations / Vibrations
29 38
Environmental Acoustics/Acoustique de l'environnment
42
Musical Acoustics / Acoustique musicale
52
Hearing/Audition
64
Jérémie Voix
*
Auditory Physiology/Physiologie du système auditif
76
Speech Production / Production de la parole
82
Speech Perception / Perception de la parole
90
N e w s / In f o r m a t i o n s
101
PROCEEDINGS IL i l . t J
W T D ' i L ^ v . A JL^I
_
A
AJLJA
