Approval and Disapproval: InfantResponsiveness to Vocal Affect inFamiliar and Unfamiliar Languages

Anne FernaldStanford University

FERNALD, ANNE. Approval and Disapproval: Infant Responsiveness to Vocal Affect in Familiarand Unfamiliar Languages. CHILD DEVELOPMENT, 1993, 64, 657-674. In a series of 5 auditorypreference experiments, 120 5-month-old infants were presented with Approval and Prohibitionvocalizations in infant-directed (ID) and adult-directed (AD) English, and in ID speech in non-sense English and 3 unfamiliar languages, German, Italian, and Japanese. Dependent measureswere looking-time to the side of stimulus presentation, and positive and negative facial affect. Noconsistent differences in looking-time were found. However, infants showed small but significantdifferences in facial affect in response to ID vocalizations in every language except Japanese.Infants smiled more to Approvals, and when they showed negative affect, it was more likely tooccur in response to Prohibitions. Infants did not show differential affect in response to Approvalsand Prohibitions in AD speech. The results indicate that young infants can discriminate affectivevocal expressions in ID speech in several languages and that ID speech is more effective thanAD speech in eliciting infant affect.

Research on the development of infants'responsiveness to emotional expressions hastaught us much less about how infants listento voices than about how they look at faces.Although early investigators suggested thatyoung infants smile earlier to voices than tofaces (Wolff, 1963) and respond differen-tially to positive and negative affect in thevoice (Buhler & Hetzer, 1928; Lewis, 1936/1951), recent research on infants' perceptionof emotion has focused primarily on the face.The majority of microanalytic studies ofmother-infant communication have codedmaternal facial expression and gaze whileneglecting vocal expression (e.g., Fogel,1982; Malatesta, Grigoryev, Lamb, Albin, &Culver, 1986), or limiting vocal measures toamount of speech, verbal content, or timingof vocalizations (e.g.. Field, 1977; Lamb,1978; Mayer & Tronick, 1985). With a fewnotable exceptions (e.g., Papousek, Pa-pousek, & Bornstein, 1985; Stern, 1985;Stern, Spieker, & MacKain, 1982), researchon early emotional development has largelyignored the role of vocalizations in affectivecommunication with the infant. The implicitassumption that young infants are more re-sponsive to the face than to the voice is also

reflected in research on the early discrimina-tion and recognition of emotional signals.This literature, too, is dominated by studiesof infants' perception of facial expressions(e.g., Barrera & Maurer, 1981; Nelson & Dol-gin, 1985; Schwartz, Izard, & Ansul, 1985).While a few studies have investigated in-fants' discrimination of facial-plus-vocal ex-pressions (e.g., Haviland & Lelwica, 1987;Walker, 1982; Walker-Andrews, 1986), onlyrecently have researchers begun to explorehow infants respond to vocal expressionsalone, without simultaneous facial affect (Pa-pousek, Bornstein, Nuzzo, Papousek, &Symmes, 1990; Walker-Andrews & Lennon,1991).

Studies showing that infants youngerthan 7 months cannot consistently categorizefacial expressions (e.g., Ludemann & Nel-son, 1988) have led several investigators toconclude that infants are unable to recognizeemotional expressions until late in the firstyear (Klinnert, Campos, Sorce, Emde, &Svejda, 1983; Nelson, 1987). However, re-cent research by Caron, Caron, and Mac-Lean (1988) suggests that studies relyingsolely on facial stimuli may underestimate

This research was supported by two grants from the National Institutes of Health (MH41511and HD24349). I am grateful to Joyce Dorado, Lora Eichner, Deanne Perez, Garla Herrera, andGerald McRoberts for assistance with data collection and analysis, and to two anonymous review-ers for their helpful comments. Requests for reprints should be sent to: Anne Femald, Depart-ment of Psychology, Stanford University, Stanford, GA 94305.

[Child Development, 1993, 64, 657-674. © 1993 by the Society for Research in Child Development, IncAll rights reserved. 0009-3920/93/6403-0016$01.00]

658 Child Development

infants' ability to discriminate naturalisticemotional expressions. Caron et al. foundthat 7-month-old infants could discriminatehappy from angry expressions when dy-namic facial displays were accompanied byappropriate vocal affect. When the voice wasremoved, however, infants no longershowed evidence of being able to make thediscrimination, leading Caron et al. to con-clude that infants rely more on the voicethan the face when discriminating betweenhighly animated emotional expressions. Re-search on the response of 6-month-old in-fants to auditory-visual compounds providesfurther evidence that young infants are in-fiuenced more strongly by auditory than vi-sual input (Lewkowicz, 1988). These find-ings motivate one question addressed by thepresent research: Are young infants able todiscriminate positive and negative vocal ex-pressions at an age when they are not yetconsistently able to discriminate positiveand negative facial expressions?

A second question addressed by this re-search is whether infants respond with dif-ferential emotion to vocal expressions dif-fering in affective tone. The numerousstudies of infants' perception of facial dis-plays have focused on discrimination andcategorization abilities, without askingwhether infants respond meaningfully to fa-cial expressions (see Nelson, 1987). Simi-larly, the few studies of infants' perceptionof facial-plus-vocal displays provide limitedevidence for discrimination of some vocalexpressions (Walker-Andrews & Grolnick,1983), and cross-modal matching of vocaland facial cues (Walker, 1982; Walker-Andrews, 1986; Walker-Andrews & Lennon,1991), with no evidence that affective vocalexpressions are meaningful to infants. Re-search on social referencing, however, doessuggest that infants around the age of 8months are able to interpret some facial ex-pressions meaningfully (Klinnert et al.,1983). The only evidence that infants at ayounger age are sensitive to the emotionalmeaning in facial expressions comes from astudy by Haviland and Lelwica (1987). Inthis research, 10-week-old infants, pre-sented with their mothers' facial and vocalexpressions of happiness, sadness, andanger, responded with matching facial ex-pressions in the happiness and anger condi-tions. However, since the mothers' emo-tional displays were simultaneously facialand vocal, it is unclear to what extent infantswere responding to the face or to the voice.The present study investigates whether

young infants respond with differential af-fect to emotional vocalizations presentedwithout simultaneous facial affect.

While previous studies of infants' re-sponse to affect have presented infants withexpressions of discrete emotions such as fearand happiness (e.g., Caron et al., 1988; Nel-son & Dolgin, 1985), the present researchfocuses on more naturalistic affective vocal-izations, typical of those used by adult inter-acting with infants. Adults in many culturesuse a special speech register characterizedby exaggerated intonation when speaking toinfants (Fernald et al., 1989; Grieser & Kuhl,1988). Although most studies of "mother-ese" have focused on its language-teachingfunctions (e.g., Fernald & Mazzie, 1991;Snow, 1972), several investigators have ar-gued that exaggerated prosody in speech toinfants may function more immediately toinfiuence infant affect and attention in theearly months of life (Fernald, 1984, 1991; Pa-pousek et al., 1985; Stern et al., 1982).

Empirical support for the role of exag-gerated intonation in modulating infant af-fect and attention comes from three differentsources in the recent literature on speech toinfants. First, experimental studies havefound a listening preference for infant-directed (ID) speech over adult-directed(AD) speech in preverbal infants (Cooper &Aslin, 1990; Fernald, 1985; Fernald & Kuhl,1987; Werker & McLeod, 1989). Second,several investigators report consistencies inthe relation of prosodic forms to communica-tive functions in speech to infants (Ferrier,1985; Ryan, 1978; Stern et al., 1982). For ex-ample, Fernald (1992a) describes cross-language similarities in the prosodic con-tours used to express approval or praise toinfants, in contrast to those used to expressdisapproval or prohibition. Approval vocal-izations, such as Good! are typically spokenwith exaggerated rise-fall Fg-contours, whileprohibition vocalizations, such as No! arespoken with lower pitch in short, sharp FQ-contours. These studies suggest that specifictypes of prosodic contour are differentiallyeffective in eliciting attention and communi-cating affect to the infant. Third, adult listen-ers are able to identify the communicativeintent of the speaker much more accuratelyin ID than in AD speech using only prosodiccues (Fernald, 1989). The present researchextends these findings by asking whether in-fants respond differentially to prosodic pat-terns in ID speech and whether ID speechis more effective than AD speech in elicitinginfant affect.

Anne Fernald 659

In the five experiments reported here,infants were presented with affective vocal-izations typical of those used by motherswhen either praising or scolding an infant.The rationale for using ID approvals andprohibitions, rather than vocal expressionsspecified in terms of discrete emotions, wasthat such ID vocalizations are more commonin infants' experience and are well suited foraffective communication with infants. Totest the hypothesis that infants respond dif-ferentially to positive and negative vocaliza-tions, two dependent measures were used.The first dependent measure was theamount of time infants looked to the side onwhich each type of vocalization was pre-sented. Using a modification of the auditorypreference procedure developed by Fernald(1985), approvals and prohibitions were pre-sented on opposite sides of a testing booth.However, unlike the original procedure,equal numbers of each type of vocalizationwere played in random order; on each trial,speech continued to play for 20 sec, or untilthe infant terminated the trial by lookingaway from the side of presentation. Thuswhile the type of vocalization presented oneach trial was under the experimenter's con-trol, the duration of the trial was under theinfant's control. This looking-time measurewas intended to index subjects' differentialattention to approvals and prohibitions.

The second dependent measure was in-fant facial affect in response to approvals andprohibitions. When a mother says Good girl!to her infant with positive vocal affect, herintention is to reward the child's behavior;when she says No, stop that! with negativevocal affect, her intention is to inhibit thechild's behavior, that is, to punish ratherthan to reward. The different communica-tive intentions motivating approval and pro-hibition vocalizations in ID speech lead toclear predictions about infants' affective re-sponses to these two categories of vocal ex-pression. If infants are indeed respondingdifferentially and appropriately, they shouldshow more positive affect in response to ap-provals and more negative affect in responseto prohibitions.

Approval and prohibition vocalizationsin four languages were used as stimuli inthese experiments, a cross-language designmotivated by two considerations. First, toeliminate the possibility that infants wereresponding to semantic content, affective vo-calizations were presented in unfamiliar lan-guages as well as in English. Second, severallanguages were included to investigate the

generality of the prosodic patterns used toexpress approval and disapproval in IDspeech in different cultures. While there issome cross-language evidence for acousticsimilarity among ID vocalizations similar incommunicative intent (Fernald, 1992a), thistells us nothing about the perceptual simi-larity of these prosodic patterns. The findingthat infants reared in English-speaking fami-lies respond differentially and appropriatelyto ID approval and disapproval vocalizationsin unfamiliar languages, as well as in En-glish, would provide evidence for the func-tional equivalence of these affective vocal-izations across languages.

Two other features of the experimentaldesign were intended to provide evidencefor response convergence and stimulusconvergence in this auditory preference pro-cedure. First, one objective in using bothlooking-time and facial affect as responsemeasures, following Werker and McLeod(1989), was to determine to what extentthese behaviors constitute convergent mea-sures of infant listening preference. Whilesmiling would seem to be a clear index ofpleasure in response to a vocal simulus,looking-time is a less straightforward mea-sure of auditory preference. The use of bothmeasures in a series of studies which in-clude a range of vocal simuli should be help-ful in comparing and evaluating looking-time and facial affect as indices of infantlistening preference. Second, of the six setsof approval and prohibition vocalizationsused as stimuli, four were prepared in thelaboratory from scripts (Experiments 1, 3,and 5), and two were composed of excerptsfrom field recordings of spontaneous moth-er-infant interaction (Experiments 2 and 4).While "posed" affective vocalizations can becontrolled for semantic content and otherrelevant linguistic and acoustic characteris-tics, they may lack critical features of morenatural ID speech. Recordings of spontane-ous speech, on the other hand, while morerepresentative of natural mother-infant in-teraction, inevitably introduce uncontrolledvariation in linguistic and acoustic structurewhich may or may not be relevant to the per-ception of affective quality. Since both ap-proaches have advantages, and neither is en-tirely satisfactory, both are used here inorder to explore the extent of stimulus con-vergence in this procedure.

Experiment 1

In Experiment 1, 5-month-old infantswere presented with Approval and Prohibi-

660 Child Development

tion vocalizations typical of ID speech inGerman. The speech of German parents toinfants has been described in several studies(Fernald & Simon, 1984; Fernald et al.,1989; Papousek, Papousek, & Haekel, 1987),and is similar to English in its global pro-sodic characteristics. Two hypotheses weretested: First, that infants would look longerto the side on which Approvals were pre-sented than to the side on which Prohibi-tions were presented, reflecting a listeningpreference for positive vocalizations; andsecond, that infants would show more posi-tive affect to Approvals and more negativeaffect to Prohibitions.

MethodBecause Experiments 1-5 used the

same subject population, apparatus, proce-dure, and methods for signal calibration anddata coding, these will be described in detailfor Experiment 1 only. The vocal stimuli dif-fered across experiments and will be speci-fied separately for each experiment.

Subjects.—The subjects in Experiment1 were 20 full-term, healthy infants (10males, 10 females), recruited from a predom-inantly white, middle-class populationthrough a university hospital. All infantscame from homes in which English was theonly language spoken and were not exposedregularly to any other language. Infants were5 months old (M: 148 days; range: 132-168days). Thirteen additional subjects wereeliminated from the study due to fussiness(6), failure to orient (4), or experimenter er-ror (3).

Stimuli.—The ID Approval and Prohi-bition vocalizations used as stimuli in Ex-periment 1 were spoken in German by a bi-lingual English/German adult female.Audio-recordings were made on a RevoxB77 MKII reel-to-reel tape recorder, using aSony ECM-23F microphone. The Approvalstimuli consisted of 14 vocalizations with ap-propriate ID semantic content (e.g.. So bravbist du [You're so good]). These positive vo-calizations were spoken with the exagger-ated rise-fall Fg-contours typically used byGerman mothers when praising preverbalinfants (Fernald, 1991). The Prohibitionstimuli consisted of 14 negative vocaliza-tions (e.g.. So bose bist du [You're sonaughty]), spoken with intonation patternstypical of those used by German motherswhen scolding preverbal infants.

In order to equate the two sets of speechstimuli for verbal density, each Approval vo-calization was matched with a Prohibition

vocalization equivalent in number of sylla-bles. The intervals between vocalizationsvaried from 450 ms to 1,200 ms for both Ap-provals and Prohibitions. However, becauseindividual Approval vocalizations tended tohave longer vowel durations than matchedProhibitions, the total duration of the se-quence of 14 Approvals was 24 sec, whilethe duration of the Prohibition sequencewas 20 sec. Multiple repetitions of thesequence of Approval vocalizations weredubbed continuously onto one stimulustape; Prohibition vocalizations were dubbedonto a second stimulus tape.

The vocal stimuli were acoustically ana-lyzed using the Micro Speech Lab (MSL)speech analysis routines on a Macintosh llfxcomputer. For each stimulus tape, the meanFQ was calculated across the entire sequenceof vocalizations. Fp-range was calculated bysubtracting the Fg-minimum from the Fg-maximium for each vocalization, and com-puting the mean Fo-range for Approvals andProhibitions. Both the standard deviation ofthe mean Fg and the mean Fg-range wereconverted into semitones, in order to facili-tate meaningful comparison of pitch vari-ability and range across stimuli differing inmean-Fg (see Fernald et al., 1989). As indi-cated in Table 1, the mean Fg was higher,the Fo-variability was greater, and the meanFg-range was wider for Approvals than forProhibitions. Examples of Fg-contours fromthe German stimuli are shown in Figure 1.As a manipulation check, six monolingualEnglish-speaking adults were presentedwith 10 positive and 10 negative vocaliza-tions in random order from the German stim-uli and were asked to categorize these as ei-ther Approvals or Prohibitions. Listenerscategorized 95% of the stimulus vocaliza-tions correctly in this task.

Apparatus.—Testing was conducted ina sound-treated laboratory room, acousti-cally isolated from an adjacent control room.The testing booth consisted of 1 m x 2 mcloth-covered panels on three sides and wasopen on the fourth side. A chair for parentand infant was centered in the booth, facingthe front panel. At the level of the infant'seyes, a green blinker light was mounted onthe front panel, operated from the controlroom. The lens of a Panasonic WV 3200video camera projected through an openingin the front panel below tbe green light. TwoJVC loudspeakers were mounted behindthe left and right panels, at 45° angles to eachside relative to the infant, at the level of theinfant's head. Two identical 10 cm x 12.5-

TABLE 1

MEAN FUNDAMENTAL FREQUENCY (FQ), STANDARD DEVIATION OF MEAN FQ, AND MEAN FQ-RANGE OFAPPROVAL AND PROHIBITION VOCALIZATIONS USED AS STIMULI IN EXPERIMENTS 1-5

APPROVALS PROHIBITIONS

Mean MeanLANGUAGE Mean FQ" SD' ' Fo-range"̂ Mean-Fo" SD'' Fo-range'

Experiment 1:German ID 233 5.1 19.1 216 2.7 8.1

Experiment 2:Italian ID 302 5.3 17.5 225 3.5 11.2

Experiment 3:''English nonsense ID 237 3.9 16.2 257 3.9 14.7

Experiment 4:Japanese ID 373 4.3 11.0 283 3.4 11.4

Experiment 5:English ID 300 5.0 16.4 230 3.2 8.9English AD 212 2.4 7.8 208 2.6 4.8

* Mean FQ calculated in hertz for each sequence of stimulus vocalizations.'' Standard deviation of mean FQ converted into semitones to adjust for differences in absolute FQ.° Fo-range calculated by subtracting Fo-minimum from Fo-maximum for each stimulus vocalization in the se-

quence, converted into semitones.•• In Experiment 3, Approvals and Prohibitions were approximately matched for Fo-characteristics.

500 -

400 -

300 -

200 •

100 •

BRAV

/

BIST DU

^

BRAV BIST DU

f V

JA

y

TOLL

1

\

APPROVALS

500-

400 -

300 •

200-

100 •

HOR DOCH AUF NEIN NICHT FUR DICH

PROHIBITIONS

FIG. 1.—Examples of fundamental frequency (FQ) contours (in hertz) of German infant-directedApproval and Prohibition vocalizations used as stimuli in Experiment 1.

662 Child Development

cm black-and-white photographs of awoman's face with an emotionally neutralexpression were mounted symmetrically onthe left and right panels, at locations corre-sponding to the two loudspeakers. In thecontrol room, a Panasonic AG-1800 video re-corder and monitor were connected througha soundproof patch panel to the camera inthe testing room. A time/date generator pro-vided a time code on the video record ofeach test session. Two Uher Report 4000tape recorders were used to present thespeech stimuli through loudspeakers in thetesting booth. Appropriate loudness levelsfor the two sets of speech stimuli were deter-mined by asking three adult subjects tomake loudness-matching judgments, using amodified version of the psychophysicalmethod of adjustment. Peak signal intensi-ties varied from 65 to 68 db (SPL, A-Scale),measured at the position of the infant's headwith a Quest (Model 215) sound level meter.Signal level was calibrated for each loud-speaker before each test session.

Procedure.—The infant sat on the par-ent's lap, facing the front of the testingbooth. The parent listened to vocal musicover AKG headphones, to divert attentionfrom the stimuli presented to the infant. Inthe control room, the infant's head positionwas observed on the monitor by a judge,blind to which set of speech stimuli was pre-sented on which side of the testing booth.The responsibilities of the judge were toturn on the blinker light in order to draw theinfant's attention to midline at the start of anew trial; to signal the experimenter whento start a new trial, as soon as the infant'seyes were centered; to judge when the in-fant made a criterion head-turn (>45°) to-ward the side on which the speech stimuluswas presented; to judge when the infantlooked away for more than 2 sec from theside on which the speech stimulus was pre-sented, signaling the experimenter to stopthe trial; to time the presentation of speechsignals, stopping the trial after 20 sec if theinfant had not yet looked away. Three indi-viduals served as judge during this series ofexperiments, all trained extensively in judg-ing criterion head-turns and look-aways.

The responsibilities of the experimenterwere to calibrate the loudness levels of thespeech stimuli before each test session; toassign each subject randomly to an experi-mental condition; to operate the tape record-ers, following the judge's signals to start andstop trials; to present each speech signal on

the appropriate side, following a predeter-mined random trial order; to monitor theaudio signals over headphones, making surethat sound production was terminated onlyduring the silent intervals between vocaliza-tions.

Each test session consisted of 16 trials,in which eight Approval and eight Prohibi-tion vocalizations were presented. On thefirst four trials. Approvals and Prohibitionswere presented in alternating order; duringthe remaining trials, stimuli were presentedin a different random order for each subject.Approval vocalizations were presented onthe right side of the test booth for half thesubjects, and on the left side for the otherhalf The initial order of presentation of Ap-provals and Prohibitions was also counter-balanced across subjects. If an infant failedto orient to the sound presentation within 10sec on four or more of the 16 trials, or on twoor more of the first four trials, the subjectwas excluded from the final sample.

Coding duration of looking-time.—Duration of infant head-turns wasmeasured off-line from the video record. Inpreparation for coding, a laboratory assistantrecorded the time code marking the onsetand offset of the audio signal on each trial.This protocol was used by a coder who wasblind to the hypothesis of the study andblind to the side of presentation of Approv-als and Prohibitions. Watching the video dis-play without the soundtrack, the coder usedthe slow-motion function of a Panasonic AG-1950 video editing deck to determine whenthe infant made a 45° head-turn, relative tothe onset of each trial, and how long thehead-turn was maintained. The time be-tween the onset of the audio signal and theinfant's head-turn toward the side of soundpresentation was coded as the response la-tency. The time from the infant's head-turntoward the side of presentation and the in-fant's return to midline, terminating the trial,was coded as duration of looking. If, how-ever, the infant returned to midline for lessthan 2 sec during a trial, the trial was notdiscontinued. Since this judgment had beenmade on-line by the judge during the testsession, the coder performed a post hoc ma-nipulation check to correct for errors that af-fected the duration of sound presentation. Inthe few cases where the judge had eitherterminated a trial after a look-away <2 sec,or continued a trial after a look-away >2 sec,that trial was dropped from the analysis. Ifsuch errors occurred on more than one trial

Anne Fernald 663

for a given subject, the data from that subjectwere not included in the analysis and wererecorded as experimenter error.

Coding infant affect.—Infant facial af-fect was coded off-line from the video recordby a second coder blind to the hypothesis ofthe study and blind to the side of presenta-tion of Approvals and Prohibitions. Positiveand negative affect were coded separately,following procedures described by Hirsh-berg and Svejda (1990). Positive affect wasrated on a three-point scale: 0 = neutral; 1= positive attention; 2 = smile. The three-point scale for negative affect was: 0 = neu-tral; 1 = tense brow, wary; 2 = frown. Eachtrial was divided into 5-sec intervals, withone positive and one negative affect ratinggiven for each interval. Mean positive andnegative affect scores were computed foreach trial.

Reliability.—Reliability was assessedby having two coders independently codethe head-turn and affect data for 40% of thesubjects. For the looking-time data, correla-tion coefficients ranged from .88 to 1.0 (M =.92). Gohen's kappa values ranged from .75to .87 (M = .82) for positive affect, and from.77 to .84 (M = .80) for negative affect. Giventhat infant affect could potentially provide acue to the side of presentation of Approvalsand Prohibitions, an additional reliabilitycheck was performed. A single trial from thevideotape for each subject was randomly se-

lected and dubbed onto a second videotape,so that no information about the previous re-sponses of a given infant could infiuencecoders' judgments on any trial. Two codersindependently coded duration of head-turnand facial affect for each of these 20 trials.The correlation coefficient for agreement be-tween coders on this composite tape was .96for duration of looking-time; kappa valueswere .81 for positive affect, and .79 for nega-tive affect.

Data analysis.—Infant looking-timeand affect data in Experiment 1 were ana-lyzed using 2 x 2 x 2 mixed ANOVAs inwhich sex and side-of-approval were be-tween-subject variables, and vocal expres-sion (Approval, Prohibition) was the re-peated measure. Gomparable analyses werealso used in Experiments 2—5.

Results and DiscussionTable 2 shows the mean durations of in-

fant looking-time toward Approvals and Pro-hibitions in German ID speech. A 2 x 2 x2 ANOVA yielded a significant main effectof vocal expression, F(l, 16) = 10.66, p <.005. No other main effects or interactionswere significant. While the finding that in-fants looked longer to Approvals than to Pro-hibitions could refiect a preference for Ap-provals, it could also refiect a difference inthe localizability of the two types of vocalexpression. If infants oriented to Prohibi-tions more slowly than to Approvals because

TABLE 2

MEAN INFANT LOOKING-TIME (in Sec) IN RESPONSE TO APPROVAL AND PROHIBITIONVOCALIZATIONS IN EXPERIMENTS 1-5

Approval/Prohibition

Group" Approvals Prohibitions Difference

Experiment 1:German ID 9.28 6.32 p < .005

Experiment 2:Italian ID 5.50 4.15 p < .02''

Experiment 3:"̂English nonsense ID 5.97 5.49 N.S.

Experiment 4:Japanese ID 7.15 6.43 N.S.

Experiment 5:English ID 5.05 5.56 N.S.English AD 5.61 5.06 N.S.

^ n ~ 20 for each group.^ Effect of vocal expression significant only for infants presented with Approvals on the

right side.*̂ In Experiment 3, Approvals and Prohibitions were approximately matched for Fo-

characteristics.

664 Child Development

they were more difficult to localize, this pat-tern could reduce the mean trial duration forProhibitions. To investigate this possibility,the mean response latencies to the first twoApproval trials and the first two Prohibitiontrials were computed for each infant. Apaired t test revealed no differences in in-fants' initial response latencies to the twotypes of vocal expression.

Infants responded with more positiveaffect to Approvals than to Prohibitions spo-ken in German, as shown in Table 3. Infantsalso responded with more negative affect toProhibitions than to Approvals, as shown inTable 4. Positive affect scores were analyzedin a 2 X 2 X 2 ANOVA, revealing a signifi-cant main effect of vocal expression, F(l, 16)= 7.10, p < .02. The same analysis per-formed on the negative affect scores alsoyielded a significant main effect of vocal ex-pression, F(l, 16) = 4.52, p < .05. No othermain effects or interactions were significant.

In Experiment I, 5-month-old English-learning infants responded with a differen-tial looking-time and affect to Approvals andProhibitions spoken in German, a languagewith which they were completely unfamil-iar. When presented with vocal expressionstypical of those used by German motherswhen praising or scolding a preverbal infant,infants looked significantly longer to Ap-provals than to Prohibitions. Infants alsoshowed significantly more positive affectwhen listening to Approvals and signifi-

cantly more negative affect when listeningto Prohibitions. It should be noted, however,that infants responded with both positiveand negative affect to both kinds of stimulusvocalization and that the amount of negativeaffect shown overall was small in magnitude,an issue to be considered in the "GeneralDiscussion." In the next experiment, this in-vestigation of infants' responsiveness to vo-cal affect in ID speech was extended to in-clude a second language unfamiliar toAmerican English infants, using more natu-ralistic vocal stimuli.

Experiment 2

The goal of Experiment 2 was to explorethe generality of the findings of the first ex-periment, by presenting 5-month-old infantswith Approval and Prohibition vocalizationsin Italian and by using as stimuli spontane-ous tokens of ID speech recorded from sev-eral different mothers. In Experiment 1, theGerman stimuli were spoken by a singlespeaker and were posed rather than natural,in order to equate the number of syllablesacross vocal expressions. The stimuli in Ex-periment 2, in contrast, consisted of sponta-neous Approvals and Prohibitions recordedduring home observations of Italian families.

Method

Subjects.—Subjects were 20 full-term,healthy infants (10 males, 10 females). In-fants were 5 months old (M; 141 days; range:

TABLE 3INFANTS' POSITIVE AFFECT SCORES IN RESPONSE TO APPROVAL AND PROHIBITION

VOCALIZATIONS IN EXPERIMENTS 1-5

Approval/Prohibition

Group" Approvals Prohibitions Difference

Experiment 1:German ID 60 .28 p < .05

Experiment 2:Italian ID 32 .19 p < .05

Experiment 3:''English nonsense ID 52 .32 p < .05

Experiment 4:Japanese ID 26 .26 N.S.

Experiment 5:English ID 50 .20 p < .01English AD 27 .34 N.S.

NOTE.—The values represent mean scores for infants' positive facial affect, rated on a three-point scale: 0 = neutral, 1 = positive attention, 2 = smile.

• n = 20 for each group.'• In Experiment 3, Approvals and Prohibitions were approximately matched for Fo-

characteristics.

Anne Fernald 665

TABLE 4

INFANTS' NEGATIVE AFFECT SCORES IN RESPONSE TO APPROVAL AND PROHIBITIONVOCALIZATIONS IN EXPERIMENTS 1-5

Approval/Prohibition

Group* Approvals Prohibitions Difference

Experiment 1:German ID 05 .17 p < .05

Experiment 2:Italian ID 09 .19 p < .01

Experiment 3:''English nonsense ID 06 .13 p < .05

Experiment 4:Japanese ID 13 .09 N.S.

Experiment 5:English ID 04 .09 p < .05English AD 10 .09 N.S.

NOTE.—The values represent mean scores for infants' negative facial affect, rated on athree-point scale: 0 = neutral, 1 = tense brow, 2 = frown.

' n = 20 for each group.'' In Experiment 3, Approvals and Prohibitions were approximately matched for Fo-

characteristics.

132—151 days). Twenty-five additional sub-jects were eliminated due to fussiness (13),failure to orient (8), or experimenter error(4).

Stimuli.—The positive and negative vo-calizations used as stimuli in Experiment 2were recorded during home visits to fourmonolingual Italian-speaking families inRome, Italy, following procedures describedby Fernald et al. (1989). From the recordingof each Italian mother, a 6-9-sec segmentwas chosen in which she praised her infantfor retrieving an object. An equivalent seg-ment was chosen in which each motherscolded her infant for touching a forbiddenobject or tried to stop the infant from ap-proaching the object. For each type of vocalexpression, the four individual segmentswere combined on one audio tape, with1,200-ms pauses between segments. Thepositive sequence consisted of 30 sec of IDApproval expressions from four Italian moth-ers, while the negative sequence consistedof 32 sec of ID Prohibition expressions bythe same four speakers. Multiple repetitionsof the Approval and Prohibition sequenceswere dubbed continuously onto two sepa-rate stimulus tapes.

As shown in Table 1, the mean FQ washigher, Fg-variability was greater, and themean Fg-range was wider for Approvals thanfor Prohibitions, as in the German stimuliin Experiment 1. Six monolingual English-speaking adults, presented with 10 positive

and 10 negative vocalizations from the Ital-ian stimuli, categorized 98% of the vocaliza-tions correctly as either Approvals or Prohi-bitions.

Results and DiscussionAs shown in Table 2, infants listening

to Italian mothers' speech looked longer toApprovals than to Prohibitions, although ef-fects of side were also found in this experi-ment. A 2 X 2 X 2 ANOVA yielded signifi-cant main effects of vocal expression, F(l,16) = 8.21, p < .02, and side-of-Approval,F(l, 16) = 9.50, p < .01, as well as a signifi-cant vocal expression x side-of-Approval in-teraction, F(l, 16) = 11.94, p < .01. Infantswho heard Approval vocalizations on theleft side of the testing booth listened longerto both types of vocal expression than didinfants presented with the opposite config-uration. Follow-up tests indicated that onlythose infants who heard Approvals on theright side looked significantly longer to Ap-provals than to Prohibitions, F(l, 18) =22.31, p < .0001. Subjects presented withpositive vocalizations on the right looked 6.4sec to Approvals and 3.9 sec to Prohibitions,while subjects presented with positive vo-calizations on the left looked 4.6 sec to Ap-provals and 4.4 sec to Prohibitions. Of the20 subjects in Experiment 2, 14 lookedlonger to Approvals than to Prohibitions; ofthe six subjects who showed the oppositepattern, four were in the condition in whichApprovals were presented on the left. A

666 Child Development

paired t test revealed no differences in in-fants' initial response latencies to Approvalsand Prohibition, indicating that infantscould localize both types of vocal ex-pression.

As shown in Tables 3 and 4, infants re-sponded with more positive affect to Ap-provals than to Prohibitions in Italian IDspeech and with more negative affect in re-sponse to Prohibitions than to Approvals.Positive affect scores were analyzed in a 2 x2 x 2 ANOVA, revealing a significant maineffect of vocal expression, F(l, 16) = 6.16, p< .05. The analysis of negative affect scoresalso yielded a significant main effect of vocalexpression, F(l, 16) = 12.17, p < .01. Noother main effects or interactions were sig-nificant in these analyses.

In Experiment 2, 5-month-old infantsresponded with differential looking-time topositive and negative vocalizations from thespontaneous ID speech of four Italian moth-ers, although this preference was significantonly when Approvals were presented on oneside of the booth. Thus the looking-timefindings of Experiment 1 were only partiallyreplicated in Experiment 2. American in-fants listening to Italian mothers' speechalso responded with more positive affect toApprovals and with more negative affect toProhibitions. Infants' greater positive affectin response to Italian and German Approvalsis consistent with the hypothesis that infantsprefer to listen to these positive vocal stim-uli. However, it is not clear from these datathat it was the affective quality of the Ap-proval vocalizations that elicited the lis-tening preference. Since the Approvals inExperiments 1 and 2 were characterized byhigh mean FQ and wide Fo-range, ,while theProhibitions had a narrower Fg-range, asshown in Table 1, it could be that it was thedifference in the extent of overall FQ-modulation that recruited longer looking-time and more positive affect to Approvalsin most infants, rather than other acousticfeatures uniquely associated with affectivequality. Infants might simply smile more towide-range pitch contours than to narrow-range pitch contours, regardless of the af-fective tone of the vocalization. This alterna-tive hypothesis is quite plausible, in light ofFernald and Kuhl's (1987) finding that 4-month-old infants listened more to wide-range Ffl-contours, typical of ID speech, thanto narrow-range Fg-contours, typical of ADspeech, even when other acoustic featureswere held constant. To investigate this pos-sibility, the next experiment focused on Ap-

proval and Prohibition vocalizations equiva-lent in Fg-modulation.

Experiment 3

The goal of Experiment 3 was to elimi-nate the confound between Fg-character-istics and affective quality which character-ized the ID Approval and Prohibitionvocalizations used as stimuli in Experiments1 and 2. In order to determine whether in-fants' responses to Approvals and Prohibi-tions were simply related to differences infrequency range and modulation, the posi-tive and negative vocal stimuli in Experi-ment 3 were approximately matched in Fg-range and Fg-variability. It was predictedthat infants would look longer and showmore positive affect to Approvals than toProhibitions, even when these vocal expres-sions were matched in Fg-characteristics.The positive and negative vocal stimuli con-sisted of American English prosodic patternsusing identical nonsense syllables, in orderto eliminate differences in segmental con-tent between Approvals and Prohibitions.

Method

Subjects.—The subjects in Experiment3 were 20 full-term, healthy infants (10males, 10 females). Infants were 5 monthsold (M; 141 days; range: 138-150 days).Twelve additional subjects were eliminatedfrom the study due to fussiness (5), failure toorient (5), or experimenter error (2).

Stimuli.—The stimuli in Experiment 3consisted of a sequence of 14 vocalizationseach consisting of one to four English non-sense "words," including only syllableswhich are phonologically acceptable in En-glish, (e.g., Bem nol kren and Zot). A nativespeaker of American English, experiencedwith infants, was instructed to produce thesame sequence of nonsense utterances intwo different ways, once as if praising a pre-verbal infant, and once as if scolding the in-fant. Prior to recording the stimuli, thespeaker practiced extensively, using visualfeedback from a Visipitch Fg-analyzer (KayElemetrics) to ensure that the positive andnegative vocalizations were matched inoverall Fg-range. As shown in Table 1, En-glish nonsense Approvals and Prohibitionswere similar in Fg-range and Fg-variability.The stimuli were recorded and dubbed ontotwo stimulus types. The total durations ofthe Approval and Prohibition sequenceswere 23 sec and 21 sec, respectively. Six En-glish-speaking adults categorized 95% of theEnglish nonsense stimulus vocalizationscorrectly as Approvals or Prohibitions.

Anne Fernald 667

Results and DiscussionTable 2 shows the mean durations of in-

fant looking-time toward English nonsenseApprovals and Prohibitions. A 2 X 2 X 2ANOVA yielded no significant main effectsor interactions. Paired t tests revealed no dif-ferences in initial response latencies to thetwo types of vocalization. Infants' mean af-fect scores to positive and negative vocal ex-pressions in ID English nonsense speechare shown in Tables 3 and 4. Positive andnegative affect scores in response to Approv-als and Prohibitions were analyzed in 2 x 2X 2 ANOVAs. Infants listening to Englishnonsense stimuli showed significantly morepositive affect to Approvals than to Prohibi-tions, F(l, 16) = 5.97, p < .03, and morenegative affect to Prohibitions than to Ap-provals, F(l, 16) = 5.63, p < .03, with noother significant main effects or interactions.

In the first two experiments, positive vo-calizations were higher in mean-Fg andwider in Fg-range than negative vocaliza-tions, as is typical of Approvals and Prohibi-tions in mothers' speech in a number of lan-guages (Fernald, 1992a). In Experiment 3the two types of vocal expression were moresimilar in Fg-range and Fg-variability (seeTable 1), thus eliminating the previous con-found between Fg-modulation and affectivequality. Even when positive and negativevocalizations were matched in pitch rangeand variability, infants listening to Englishnonsense speech responded with small butsignificant differences in affect. This findingsuggests that subjects in Experiments 1 and2 were not simply responding to differencesin pitch range and variability but were sensi-tive to other acoustic features differentiatingApprovals and Prohibitions, such as spec-trum, intensity, and rise time. Which ofthese acoustic features are critical in elic-iting affect in infants is a question for futureresearch.

While infants looked reliably longer topositive vocalizations in Experiments 1 and2, there was no difference in mean looking-time to Approvals and Prohibitions in Exper-iment 3. Thus when negative and positivevocalizations were similar in Fg-variabilityand Ffl-range, infants looked equally long toboth. Should the absence of differentiallooking to Approvals in Experiment 3 be in-terpreted as the absence of a listening pref-erence for positive over negative vocaliza-tions in English nonsense speech? Onepossible explanation for these results is thatinfants were unable to discriminate Ap-proval and Prohibition vocalizations which

were similar in pitch characteristics. How-ever, the finding that infants listening to En-glish nonsense speech responded with dif-ferential affect is not consistent with thisinterpretation. Subjects' affective respon-siveness to English nonsense speech sug-gests that the positive and negative vocaliza-tions were indeed discriminable. Moreover,infants' positive affective response to Ap-provals in English nonsense speech wouldseem to indicate a listening preference forthese positive vocal expressions even in theabsence of a difference in looking-time. Thediscrepancy between looking-time and af-fect in Experiment 3 suggests that these twomeasures may assess different aspects of in-fants' response to auditory signals. AlthoughEnglish nonsense Approvals and Prohibi-tions were equally effective in maintaininginfant attention, they were not equivalent intheir infiuence on infant emotion. Infantsshowed evidence of greater pleasure in re-sponse to Approvals and greater displeasurein response to Prohibitions. One implicationof these findings is that infants' looking-timeto German and Italian Approval vocaliza-tions in Experiments 1 and 2 may have beendetermined primarily by the wide Fg-rangeof these signals, rather than by other acousticfeatures uniquely associated with positivevocal affect.

Experiment 4

The purpose of Experiment 4 was to testthe generality of the findings of Experiments1-3 by presenting infants with Japanese IDspeech, a choice motivated by two consider-ations: First, Japanese differs in prosodicstructure from the languages used in thethree previous experiments, although IDspeech in Japanese is similar in its globalprosodic characteristics to ID speech in Eu-ropean languages (Fernald et al., 1989). Sec-ond, cross-cultural research on the percep-tion of emotional signals by adults hasshown that Japanese facial expressions (Shi-moda, Argyle, & Riccibitti, 1978) and vocalexpressions (Magno-Galdognetto & Kori,1983) are more difficult to identify than anal-ogous English or Italian emotional expres-sions.

Method

Subjects.—The subjects in Experiment4 were 20 full-term, healthy infants (10males, 10 females). Infants were 5 monthsold (M; 140 days; range: 133-152 days).Fourteen additional subjects were elimi-nated from the study due to fussiness (4),

668 Child Development

failure to orient (8), or experimenter error(2).

Stimuli.—The stimuli consisted of ex-cerpts from recordings made during homevisits to three native Japanese families whowere visiting the San Francisco area for ashort period. The mothers in these familieswere all monolingual speakers of Japanese.A Japanese research assistant recordedmothers interacting with their 10—14-month-old infants, following procedures describedby Fernald et al. (1989). From each record-ing, segments were chosen in which themother spontaneously praised and scoldedher infant. For each type of vocal expression,6-9-sec segments from each speaker werecombined on one audio tape, with 1,200-mspauses between segments. The Approval se-quence consisted of 22 sec of praise expres-sions, while the Prohibition sequence con-sisted of 24 sec of disapproval expressionsfrom the same three speakers. In these spon-taneous Japanese vocalizations, the mean-Fgwas higher and Fo-variability was greater inApprovals than in Prohibitions, although themean Fg-range was similar (see Table 1). Sixmonolingual English-speaking adults cate-gorized 82% of the Japanese stimulus vocal-izations correctly as Approvals or Prohibi-tions.

Results and DiscussionTable 2 shows the mean looking-times

to Japanese Approvals and Prohibitions. A 2X 2 X 2 ANOVA yielded no significant maineffects or interactions. Paired t tests re-vealed no differences in initial response la-tencies to the two types of vocalization.Mean affect scores in response to posi-tive and negative vocal expressions in Japa-nese ID speech are shown in Tables 3 and4. ANOVAs performed on positive and nega-tive affect scores revealed no significantmain effects or interactions.

In Experiment 4, English-learning in-fants failed to respond differentially to posi-tive and negative vocalizations in JapaneseID speech. Although differing in mean-Fg,Japanese Approvals and Prohibitions weremore similar in Fg-range and Fg-variabilitythan were the positive and negative vocalstimuli used in Experiments 1 and 2. More-over, Japanese Approvals were relativelynarrow in Fg-range, compared to the Germanand Italian Approvals. The attenuation of FQ-range in Japanese mothers' speech relativeto ID-speech in European languages wasalso noted by Fernald et al. (1989) and isconsistent with other research showing that

facial and vocal expressions of emotion tendto be much less extreme among Japanesethan American adults (Ekman, 1972). Thelimited prosodic contrast between positiveand negative vocalizations may have madeJapanese Approvals and Prohibitions moredifficult for infants to discriminate. It is in-teresting to note that the English-speakingadults who were asked to categorize the vo-cal stimuli in this series of experiments wereless accurate when judging vocal expres-sions in Japanese than in German, Italian, ornonsense English, consistent with the find-ing of Magno-Galdognetto and Kori (1983)that European listeners were often unable toidentify the intended affect in Japanese vo-cal expressions.

Experiment 5

In Experiment 5, infants were pre-sented with positive and negative vocaliza-tions in both ID and AD speech in AmericanEnglish. One goal of this experiment was toprovide a native-language control, since thestimuli used in Experiments 1-4 were all inlanguages unfamiliar to the infant subjects.The second goal was to test the hypothesisthat infants would respond differentially toaffective vocalizations in ID speech but notin AD speech.

Method

Subjects.—The subjects in Experiment5 were 40 full-term, healthy infants (20males, 20 females). Infants were 5 monthsold (M: 138 days; range: 131-147 days). Six-teen additional subjects were eliminatedfrom the study due to fussiness (5), failure toorient (8), or experimenter error (3). Twentyinfants were randomly assigned to the IDspeech condition, and 20 to the AD speechcondition, with equal numbers of males andfemales in each group.

Stimuli.—The stimuli in Experiment 5were spoken by three female native speak-ers of American English, all experiencedwith infants. Approvals and Prohibitions inboth ID and AD speech were recorded inrole-played scenarios. While the text was im-provised, speakers were instructed to usethe word Good frequently in Approvals di-rected to both infants and adults and to usethe words No and Don't frequently in Prohi-bitions directed to both infants and adults,in order to match the semantic contentwithin and across the ID and AD speechconditions. For the AD Prohibitions, eachspeaker role-played the following scenewith an adult colleague: The adult col-

Anne Fernald 669

league, while hanging a picture on the wall,started to step backward into an electricalcord, knocking over a lamp. The speakertried to prevent this accident with suchwarnings as No.' Don't step back! Don't putyour foot there. The AD Approvals wererole-played in a scene in which the speakercongratulated her colleague for completinga difficult task, using such praise expressionsas Good job! That's great, you're finally fin-ished. Good work. For ID Prohibitions,speakers spoke as if attempting to stop aninfant from touching an electrical cord,while for ID Approvals, they spoke as ifpraising an infant for putting a ring toy to-gether. From the recordings of each speaker,6-9-sec segments of Approvals and Prohibi-tions to each addressee were chosen,matched for the number of positive and neg-ative words. Two ID and two AD speechstimulus tapes were constructed, followingprocedures described in Experiments 2 and4. The durations of the ID Approval and Pro-hibition sequences were 26 sec and 24 sec,respectively; the durations of the AD Ap-proval and Prohibition sequences were 22sec and 23 sec, respectively.

Results and DiscussionTable 2 shows the mean duration of in-

fant looking-time toward Approvals and Pro-hibitions in ID and AD speech in English.Looking-times were analyzed separately forID and AD speech in 2 x 2 x 2 ANOVAs,which yielded no significant main effects orinteractions. Paired t tests revealed no dif-ferences in initial response latencies to thetwo types of vocalization for infants in eithercondition. Infants' mean affect scores to pos-itive and negative vocal expressions in IDand AD speech are shown in Table 3. Posi-tive and negative affect scores in responseto ID and AD stimuli were analyzed sepa-rately in 2 X 2 X 2 ANOVAs. Infants lis-tening to ID speech showed significantlymore positive affect to Approvals than toProhibitions, F(l, 16) = 9.69, p < .01, andsignificantly more negative affect to Prohibi-tions than to Approvals, F(l, 16) = 5.45, p <.05. In contrast, infants listening to ADspeech showed no reliable differences intheir affective responses to Approvals andProhibitions. The only significant finding toemerge in the analyses of infants' emotionalresponses to AD speech was that femalesshowed more negative affect than males,F(l, 16) = 5.18, p < .05. No other main ef-fects or interactions were significant.

The finding that infants did not looklonger to Approvals in either ID or AD

speech in Experiment 5 was surprising,given that positive and negative vocaliza-tions were spoken in a language familiar tothe subjects, and Approvals were wider inFg-range than Prohibitions. In Experiments1 and 2, which also used positive and nega-tive vocalizations differing in Fg-range, in-fants tended to look longer to Approvals.The interpretation proposed earlier, that thewide Fg-range of Approvals might recruitgreater attention, would predict a similar dif-ference in looking-time to the English stim-uli in Experiment 5. The major finding ofExperiment 5 was that infants respondedwith differential affect to positive and nega-tive vocalizations in ID speech but not inAD speech in natural English, consistentwith the view that affective communicationis an important function of the special pros-ody of ID speech in the preverbal period(Fernald, 1992b).

General Discussion

In this series of experiments investigat-ing infants' responsiveness to affective vo-calizations in different languages, four majorfindings emerged. First, 5-month-old infantsshowed an overall tendency to respond withmore positive affect to Approvals and morenegative affect to Prohibitions. Thus infantsshowed small but significant differences intheir emotional responses to affective vocalexpressions in the absence of affective facialexpressions. Second, ID vocalizations elic-ited differential emotional responses in in-fants, while AD vocalizations did not. Third,contrary to expectations, infants did not lookconsistently longer to Approvals than to Pro-hibitions, suggesting that looking-time andemotional responsiveness may be affectedby different acoustic features of vocal ex-pressions. Finally, infants familiar only withEnglish showed evidence of differential re-sponsiveness to affective vocal expressionsin German, Italian, nonsense English, andnatural English ID speech, but not in Jap-anese.

The finding that 5-month-old Americaninfants show more positive affect to Approv-als and more negative affect to Prohibitions,spoken in two unfamiliar languages as wellas in nonsense and natural English, providesthe first evidence that young infants responddifferentially to positive and negative vocalexpressions presented without facial affect.These results extend Haviland and Lel-wica's (1987) finding that adult facial-plus-vocal displays of happiness and anger in-duced matching facial affect in 10-week-oId

670 Child Development

infants. However, while Haviland and Lel-wica implied that adult facial expressionswere critical in inducing appropriate emo-tional responses in infants, the present find-ings call this interpretation into question. InExperiments 1, 2, 3, and 5, ID vocal expres-sions presented without affective facial ex-pressions were sufficient to elicit positiveand negative affect from 5-month-old in-fants. Research showing that 7-month-old in-fants fail to categorize happy and angry facialexpressions without concurrent vocal affect(Garon et al., 1988) lends further support tothe view that infants at this age attend moreto vocal than to facial cues when respondingto emotional expressions. An early relianceon auditory rather than visual cues in pro-cessing emotional displays would not be sur-prising, given that infants' auditory percep-tual abilities are more mature than theirvisual perceptual abilities in the first monthsof life (Aslin, 1987).

While the evidence for positive affect inresponse to Approvals was robust across allID speech samples except Japanese, the evi-dence for negative affect was similar but lessconvincing. Infants responded with mixedaffect to both kinds of vocalization, but whenthey showed negative affect it was signifi-cantly more likely to occur in responseto Prohibitions. The mean negative affectscores in response to Prohibitions, evenwhen reliably different from the negative af-fect scores for Approvals, were generallylow, refiecting an overall tendency towardneutral affect. However, the relatively lowfrequency of negative affect in these experi-ments needs to be understood in terms ofcertain basic limitations both of infants andof the experimental procedure. The proce-dure involved eight presentations each ofpositive and negative vocal stimuli in quasi-random order within a 7-10-min test ses-sion. While it is possible to make an infantsmile repeatedly, eliciting repeated alterna-tions between distress and smiling is lessfeasible. In fact, whenever a subject becamereally distressed in response to a stimulusvocalization, the test session was discon-tinued and eliminated from the sample.Moreover, Approvals and Prohibitions werepresented at equivalent loudness levels, al-though under natural conditions Prohibi-tions would typically be more intense.

It is therefore not surprising that indi-vidual infants in this procedure did not alter-nate between extremes of smiling and dis-tress. Instead, those infants who smiledfrequently to Approvals tended to become

sober in response to Prohibitions, but notto frown or cry, while those infants whoshowed considerable distress to Prohibi-tions tended to brighten, but not to smile, inresponse to Approvals. This effect may ex-plain the pattern of results in Experiment 2with Italian vocalizations. The English-speaking adult listeners who made categoryjudgments about the auditory stimuli consis-tently commented that the Italian vocaliza-tions were the most emotionally intense ofall the stimuli, while the Japanese stimuliwere the least emotionally intense. Experi-ment 2 had the highest attrition rate due toinfant distress, and many infants who con-pleted the test session responded with ex-treme negative affect to the Italian Prohi-bitions. Frequent negative responses toprohibitions may have resulted in an attenu-ation of positive affect in response to ItalianApprovals. Moreover, almost all of the sub-jects who reacted negatively to Italian Prohi-bitions also showed some negative affect toApprovals. Such carryover effects are wellknown in procedures involving presentationof negative emotional signals to infants.Given that infants cannot switch back andforth between positive and negative affectover repeated trials, a design better suitedfor eliciting both extremes might involveblocks of positive and negative trials, sepa-rated by a substantial break. As it was, in-fants' negative reactions to Prohibition vo-calizations were probably underestimatedhere, given the compromises necessary tokeep attrition to a minimum.

Infants' display of differential affect inresponse to positive and negative vocaliza-tions in these languages was not consistentlyaccompanied by differences in looking-time.When listening to German, infants lookedlonger and showed more positive affect toApprovals. When listening to Italian, infantsalso tended to look longer to Approvals, al-though this effect was qualified by an inter-action with side of presentation. However,both nonsense English and natural EnglishID speech elicited differential affect withoutcorresponding differences in looking-time.One interpretation consistent with the find-ings of Experiments 1—3 is that infants at-tend more to vocal stimuli with greater Fg-modulation, regardless of whether they"prefer" these vocalizations in the sense ofliking them better. Wide-range vocal signalsmay simply be more compelling, just as si-rens are compelling although not necessarilypleasant to listen to. Thus when Approvalsand Prohibitions were matched in Fg-range

Anne Fernald 671

and Fg-variability, as in the English non-sense and Japanese stimuli, the looking-timedifference disappeared. However, infantsstill responded with differential affect to En-glish nonsense Approvals and Prohibitions,suggesting that acoustic features other thanFg-modulation distinguished these vocaliza-tions as positive and negative in affective va-lence. Unfortunately, this interpretationdoes little to explain why the infants lis-tening to natural English ID vocalizationsfailed to look longer to Approvals. Given thediscrepancies between the looking-time andaffect measures across Experiments 1—5, itseems safest to conclude that these behav-iors may index different aspects of infants'responsiveness to acoustic signals. Looking-time may be determined by the relative"compellingness" of auditory stimuli, oftencorrelated with Fg-modulation, while infantaffect may be infiuenced more by otheracoustic features, such as Fg-contour, inten-sity, and spectrum.

Why did infants listening to ID-speechin German, Italian, and natural and non-sense English respond to Approvals withsigns of pleasure? And when they showeddispleasure, why was it more likely to occurin response to Prohibitions? The affectivepreference of 5-month-old American infantsfor Approvals over Prohibitions in EnglishID speech could be attributable to previousexperience with these types of vocalization.However, before discussing possible effectsof experience on infant listening prefer-ences, two alternative explanations will beconsidered. First, it could be argued that in-fants were innately sensitive to the meaningof emotional expressions, as Darwin (1872/1965) proposed. This explanation assumesthat expressions of vocal affect are similar inform across cultures, as has been docu-mented for facial expressions (Ekman, 1972).Research on universal features of vocal af-fect shows that adults are fairly accurate indecoding emotional meaning from vocalcues, although there is less consensus aboutwhich acoustic correlates differentiate emo-tional vocalizations (Scherer, 1986). How-ever, several studies indicate that increasesin mean Fg and Fg-range are typical of ex-pressions of enjoyment and happiness,while decreases in mean Fg and Fg-rangecharacterize expressions of irritation andcontrolled anger (see Scherer, 1986). Thusthe wide-range Approvals and narrow-rangeProhibitions used as stimuli in Experiments1, 2, 4, and 5 have prosodic features typicalof expressions of happiness and irritation, re-

spectively. Perhaps infants recognize theemotions expressed in these vocalizationsand respond, in Darwin's words, with "aninstinct of sympathy" arising from their "in-nate knowledge of expression" (Darwin,1872/1965, p. 358). However, the findingthat infants did not show differential affectto Approval and Prohibition vocalizations ei-ther in AD English or in ID Japanese arguesagainst a simple nativist account.

A second explanation for infants' emo-tional responsiveness to Approval and Prohi-bition vocalizations also derives from Dar-win (1872/1965), who observed that theacoustic features of animal calls are oftenwell designed to achieve particular effectson the listener. Perhaps infants smile moreto Approvals in ID speech because these ex-aggerated rise-fall pitch contours are intrin-sically pleasing to listen to. Prohibitions, incontrast, are startling in their abrupt onsetand harsher in tone quality and thus mayelicit more negative affect (see Fernald,1992a). According to this explanation, in-fants respond with more pleasure to pleasantsounds, and with more displeasure to un-pleasant sounds, without necessarily know-ing anything about the emotional states mo-tivating the production of these sounds bythe speaker. Of course, since pleasant-sounding Approval vocalizations are gener-ally accompanied by smiles and other indi-ces of the mother's positive affect, andProhibitions co-occur with other indices ofnegative affect, infants may soon come to un-derstand that particular vocal patterns are in-formative about the speaker's emotionalstate. Initially, however, these characteristicvocal patterns could function as uncondi-tioned stimuli, eliciting positive and nega-tive affect by virtue of their intrinsic acousticproperties, although not yet appreciated ascues to the emotional state of the speaker(Fernald, 1992b).

The argument that mothers' vocal ex-pressions have an immediate emotional ef-fect on young infants must be qualified inlight of the finding that American infantswere affectively responsive to ID vocaliza-tions in nonsense and natural English andtwo other European languages but not inJapanese. This finding suggests thataffective vocalizations in Japanese differacoustically from those in English, German,and Italian. The reduced Fg-range in Japa-nese mothers' speech may refiect the attenu-ation of affect typical of Japanese emotionalexpressions (Ekman, 1972). While the Japa-nese and European Approvals share an over-

672 Child Development

all rise-fall Fg-pattern, the Japanese vocal-izations have less Fg-variability, a smootheramplitude envelope, more pronouncedvowel elongation, and a distinctive tonalstructure, in addition to a narrower Fg-range.However, since prosodic contours are com-plex functions of graded changes along mul-tiple acoustic dimensions, it is not a simplematter to quantify such perceptual differ-ences in terms of acoustic attributes. Futureresearch is needed to identify which pro-sodic features are universal correlates of vo-cal affect and how culture-specific displayrules infiuence the acoustic structure of vo-cal expressions.

One interpretation for the finding thatAmerican infants are affectively responsiveto English, German, and Italian vocaliza-tions but not to Japanese vocalizations is thatby 5 months of age these infants have al-ready become accustomed to the exagger-ated affective displays typical in Europeancultures. For American infants as well asadults, the relatively exaggerated Germanand Italian vocal expressions may be per-ceived as similar to those in English, whilethe less exaggerated Japanese vocalizationsmay seem less similar. Japanese infants, onthe other hand, may become accustomed tothe narrower dynamic range of vocal expres-sion typical of their culture. According tothis explanation, infants in all cultures areinitially responsive to the same vocal cues.However, cultural differences in displayrules governing emotional expression maydetermine the levels and range of emotionalintensity to which the infant is routinely ex-posed and which the infant comes to expectin social interaction with adults. Such a pro-cess of cultural "calibration" would accountfor the finding that American infants weremaximally responsive to affective vocaliza-tions in ID German, natural English, andEnglish nonsense speech, perceived to bcmoderate in emotional intensity by Ameri-can adults. Italian ID vocalizations, per-ceived by adults as more intense than En-glish ID vocalizations, were clearly aversiveto many American infants, although infantsresponded in general with appropriate af-fect. Japanese vocalizations, in contrast, per-ceived as less intense by adult listeners,failed to elicit a differential emotional re-sponse from American infants. Future re-search focusing on infants from other cul-tural backgrounds will be necessary toexplore further the hypothesis that both uni-versal acoustic features and early experiencewithin a particular culture infiuence the re-

sponsiveness of young infants to affectivevocal expressions in parental speech.

ReferencesAslin, R. N. (1987). Visual and auditory develop-

ment in infancy. In J. D. Osofsky (Ed.), Hand-book of infant development (2d ed.). NewYork: Wiley.

Barrera, M. E., & Maurer, D. (1981). The percep-tion of facial expressions by the three-month-old child. Child Development, 52, 203-206.

Buhler, G., & Hetzer, H. (1928). Das erste Ver-standnis fiir Ausdruck im ersten Lebensjahr.Zeitschrift fiir Psychologie, 107, 50-61.

Garon, A. J., Garon, R. F., & MacLean, D. J. (1988).Infant discrimination of naturalistic emo-tional expressions: The role of face and voice.Child Development, 59, 604-616.

Gooper, R. P., & Aslin, R. N. (1990). Preferencefor infant-directed speech in the first monthafter birth. Child Development, 61, 1584-1595.

Darwin, G. (1872/1965). The expression of emo-tions in man and animals. Ghicago: Univer-sity of Ghicago Press.

Ekman, P. (1972). Universals and cultural differ-ences in facial expressions of emotion. In J.Gole (Ed.), Nebraska symposium on motiva-tion. Lincoln: University of Nebraska Press.

Fernald, A. (1984). The perceptual and affectivesalience of mothers' speech to infants. In L.Feagans, G. Garvey, & R. Golinkoff (Eds.),The origins and growth of communication,(pp. 5-29). New Brunswick: Ablex.

Femald, A. (1985). Four-month-old infants preferto listen to motherese. Infant Behavior andDevelopment, 8, 181-195.

Fernald, A. (1989). Intonation and communicativeintent in mothers' speech to infants: Is themelody the message? Child Development, 60,1497-1510.

Fernald, A. (1991). Prosody in speech to children:Prelinguistic and linguistic functions. In R.Vasta (Ed.), Annals of Child Development(Vol. 8, pp. 43-80). London: Jessica KingsleyPublishers.

Fernald, A. (1992a). Meaningful melodies inmothers' speech to infants. In H. Papousek,U. Jurgens, & M. Papousek (Eds.), Originsand development of nonverbal vocal com-munciation: Evolutionary, comparative, andmethodological aspects. Gambridge: Gam-bridge University Press.

Femald, A. (1992b). Human maternal vocalizationsas biologically relevant signals: An evolution-ary perspective. In J. H. Barkow, L. Gos-mides, & J. Tooby (Eds.), The adapted mind:Evolutionary psychology and the generationof culture. Oxford: Oxford University Press.

Anne Femald 673

Femald, A., & Kuhl, P. K. (1987). Acoustic deter-minants of infant preference for motheresespeech. Infant Behavior and Development,10, 279-293.

Fernald, A., & Mazzie, G. (1991). Prosody and fo-cus in speech to infants and adults. Develop-mental Psychology, 27, 209-221.

Femald, A., & Simon, T. (1984). Expanded intona-tion contours in mothers' speech to newborns.Developmental Psychology, 20, 104-113.

Femald, A., Taeschner, T., Dunn, J., Papousek,M., Boysson-Bardies, B., & Fukui, I. (1989).A cross-language study of prosodic modifica-tions in mothers' and fathers' speech to pre-verbal infants, youmai of Child Language, 16,477-501.

Ferrier, L. J. (1985). Intonation in discourse: Talkbetween 12-month-olds and their mothers. InK. Nelson (Ed.), Children's language (Vol. 5,pp. 35-60). Hillsdale, NJ: Erlbaum.

Field, T. M. (1977). Effects of early separation,interactive deficits, and experimental manip-ulations on infant-mother face-to-face interac-tion. Child Development, 48, 763-771.

Fogel, A. (1982). Early adult-infant interaction:Expectable sequences of behavior./ournaZ ofPediatric Psychology, 7, 1-22.

Grieser, D. L., & Kuhl, P. K. (1988). Maternalspeech to infants in a tonal language: Supportfor universal prosodic features in motherese.Developmental Psychology, 24, 14-20.

Haviland, J. M., & Lelwica, M. (1987). The in-duced affect response: 10-week-old infants'responses to three emotion expressions. De-velopmental Psychology, 23, 97-104.

Hirshberg, L. M., & Svejda, M. (1990). When in-fants look to their parents: I. Infants' socialreferencing of mothers compared to fathers.Child Development, 61, 1175-1186.

Klinnert, M., Gampos, J. J., Sorce, J. F., Emde,R. N., & Svejda, M. (1983). Emotions as be-havior regulators: Social referencing in in-fancy. In R. Plutchik & H. Kellerman (Eds.),Emotion in early development: Vol. 2. Theemotions (pp. 57—86). New York: AcademicPress.

Lamb, M. E. (1978). Infant social cognition and"second-order" effects. Infant Behavior andDevelopment, 1, 1-10.

Lewis, M. M. (1936/1951). Infant speech: A studyof the beginnings of language. London:Routledge & Kegan Paul.

Lewkowicz, D. J. (1988). Sensory dominance ininfants: 1. Six-month-old infants' response toauditory-visual compounds. DevelopmentalPsychology, 24, 155-171.

Ludemann, P. M., & Nelson, G. A. (1988). Gate-gorical representation of facial expressions by7-month-old infants. Developmental Psychol-ogy, 24, 492-501.

Magno-Galdognetto, E., & Kori, S. (1983). Inter-cultural judgment of emotions expressedthrough voice: The Italians and the Japanese.Paper presented at the Tenth InternationalGongress of Phonetic Sciences, Utrecht,Netherlands.

Malatesta, G. Z., Grigoryev, P., Lamb, G., Albin,M., & Gulver, G. (1986). Emotion socializa-tion and expression development in pretermand full-term infants. Child Development, 57,316-330.

Mayer, N. K., & Tronick, E. Z. (1985). Mothers'turn-giving signals and infant tum-taking inmother-infant interaction. In T. M. Field &N. A. Fox (Eds.), Social perception in infants(pp. 199-216). Norwood, NJ: Ablex.

Nelson, G. A. (1987). The recognition of facial ex-pressions in the first two years of life: Mecha-nisms of development. Child Development,58, 889-909.

Nelson, G. A., & Dolgin, K. G. (1985). The gener-alized discrimination of facial expressions byseven-month-old infants. Child Development,56, 58-61.

Papousek, M., Bornstein, M. H., Nuzzo, G., Pa-pousek, H., & Symmes, D. (1990). Infant re-sponses to prototypical melodic contours inparental speech. Infant Behavior and Devel-opment, 13, 539-545.

Papousek, M., Papousek, H., & Bornstein, M. H.(1985). The naturalistic vocal environment ofyoung infants: On the significance of homoge-neity and variability in parental speech. InT. M. Field & N. A. Fox (Eds.), Social percep-tion in infants (pp. 269-297). Norwood, NJ:Ablex.

Papousek, M., Papousek, H., & Haekel, M. (1987).Didactic adjustments in fathers' and mothers'speech to their three-month-old infants. Jour-nal of Psycholinguistic Research, 16, 491-516.

Ryan, M. (1978). Gontour in context. In R. Gamp-bell & P. Smith (Eds.), Recent advances in thepsychology of language (pp. 237-251). NewYork: Plenum.

Scherer, K. R. (1986). Vocal affect expression: Areview and a model for future research. Psy-chological Bulletin, 99, 143-165.

Schwartz, G. M., Izard, G. E., & Ansul, S. E.(1985). The 5-month-old's ability to discrimi-nate facial expressions of emotion. Infant Be-havior and Development, 8, 65—77.

Shimoda, K., Argyle, M., & Riccibitti, P. (1978).The intercultural recognition of emotional ex-pressions by three national racial groups: En-glish, Italian and Japanese. European Journalof Social Psychology, 8, 169-179.

Snow, G. E. (1972). Mothers' speech to childrenlearning language. Child Development, 43,549-565.

674 Child Development

Stem, D. N. (1985). The interpersonal world ofthe infant. New York: Basic.

Stern, D. N., Spieker, S., & MacKain, K. (1982).Intonation contours as signals in maternalspeech to prelinguistic infants. Develop-mental Psychology, 18, 727-735.

Walker, A. S. (1982). Intermodal perception ofexpressive behaviors by human infants./our-nal of Experimental Child Psychology, 33,514-535.

Walker-Andrews, A. S. (1986). Intermodal percep-tion of expression behaviors: Relation of eyeand voice? Developmental Psychology, 22,373-377.

Walker-Andrews, A. S., & Grolnick, W. (1983).Discrimination of vocal expressions by young

infants. Infant Behavior and Development, 6,491-498.

Walker-Andrews, A. S., & Lennon, E. (1991). In-fants' discrimination of vocal expression:Gontributions of auditory and visual informa-tion. Infant Behavior and Development, 14,131-142.

Werker, J. F., & McLeod, P. J. (1989). Infant pref-erence for both male and female in-fant-directed-talk: A developmental study ofattentional and affective responsiveness. Ca-nadian Journal of Psychology, 43, 230-246.

Wolff, P. (1963). Observations on the early devel-opment of smiling. In B. M. Foss (Ed.), Deter-minants of infant behavior, II. London:Methuen.