Dishabituation of Visual Attention b Infant- Versus Adult ... toward adults (Fernald & Simon, 1984; Femald ... 210 Kaplan, Goldstein, Huckeby, Owren, and Cooper fundamental frequencies

INFANT BEHAVIOR AND DEVELOPMENT 18,209-223 (1995)

Dishabituation of Visual Attention b Infant- Versus Adult-Directed Speech: Effects o r Frequency

Modulation and Spectral Composition

PETER S. KAPLAN, MICHAEL H. GOLDSTEIN, ELIZABETH R. HUCKEBY, AND MICHAEL J. OWREN

University of Colorado at Denver

ROBIN PANNETON COOPER Virginia Polytechnic Institute and State University

Dishabituation of visual attention by infant- and adult-directed (ID and AD) speech was investi-

gated in four experiments. Four-month-olds received 12 10-s presentations of a checkerboard pattern with a speech segment compounded only on the ninth trial. Recovery of visual attention was observed on the compound trial in response to both ID and AD speech, but only ID speech dishabituated visual attention during the following pattern-alone retest (a Thompson-Spencer dishabituation effect, observed in the first two experiments). Synthetic analogs of these speech segments’ fundamental frequencies (Fus) elicited equivalent increases in attention on the compound trial, but neither elicited Thompson-Spencer dishabituation (Experiment 3). A synthetic version of the intact ID signal elicited Thompson-Spencer dishabituation, but synthetic stimuli simulating the Fu only, the Fu plus the first harmonic above the Fo, and the harmonics only did not (Experiment 4). These data have implications for the acoustic characteristics of ID speech that increase infant attention and arousal.

infant-directed speech adult-directed speech dishabituation fundamental frequency frequency modulation harmonics sensitization

When addressing young infants, adults from a variety of human cultures adopt a distinctive manner of speech characterized by higher overall frequency, greater frequency variability, wider frequency excursions, greater amplitude, slower tempo, simplification of vocabulary and syntactic structure, more repetitions, and longer pauses between words relative to speech directed toward adults (Fernald & Simon, 1984; Femald et al., 1989; Grieser & Kuhl, 1988). Several studies have shown that infant-directed (ID) speech elicits stronger responding from infants than does adult-directed (AD) speech (Cooper & Aslin, 1990; Femald, 198.5; Pegg, Werker, & McLeod, 1992; Werker & McLeod, 1989). However, only a few studies have attempted to determine which acoustic dimen-

This research was supported by NICHD Grant HD19143 and by a Faculty Fellowship sabbatical year award from the University of Colorado at Denver. We thank Patrick Bashor, Amy Bergstrom, Michael Blank, Michelle Marsh, Mya Monhait, Chad Morris, Melinda Patterson, Tracey Peacoe, Jennifer Ryther-Duncan, Lauren Schlom, and Tina Suozzo for their assistance in testing infants.

Correspondence and requests for reprints should be sent to Peter S. Kaplan, Department of Psychology, University of Colorado at Denver, P.O. Box 173364, Denver, CO 80217-3364.

sions of ID speech are responsible for its greater attentional effects (Colombo & Horowitz, 1986; Cooper & Aslin, 1994; Femald & Kuhl, 1987), and these studies have produced some contradictory results. The experiments reported here had two goals. The first was to provide data concerning the acoustic characteristics of ID speech that are critical for eliciting infant attention. The second was to assess the utility of the habituation- dishabituation paradigm, with particular attention to the Thompson-Spencer dishabituation effect (see following discussion), for the study of infants’ responses to speech stimuli.

The exaggerated frequency modulation that is characteristic of ID speech has been pro- posed as a primary determinant of the attentional effects of ID speech on infants (Femald, 1992). Support for this hypothesis was found in an experiment by Femald (1985), in which 4- month-olds were trained using an operant head- turning procedure. Femald demonstrated that infants turned their heads more in the direction that resulted in contingent ID speech than in the direction that resulted in contingent AD speech. Femald and Kuhl (1987) went on to demon- strate that computer-synthesized stimuli that simulated the ID and AD speech segments’

209

210 Kaplan, Goldstein, Huckeby, Owren, and Cooper

fundamental frequencies (Fus) also produced preferential responding. These stimuli had none of the amplitude modulation or linguistic content of the natural stimuli but did retain the frequency “rise-fall” characteristics of the latter. In contrast, computer-based stimuli that simulated only the amplitude or duration characteristics of the ID and AD speech segments did not elicit differential responding. These data suggest that the Fu characteristics of ID speech account for infants’ attentional preferences.

Several other studies, however, have yielded results that are inconsistent with those of Fer- nald and Kuhl (1987). Colombo and Horowitz (1986), for example, obtained no differences in 4-month-olds’ responses to l-s bell-shaped tone sweeps designed to simulate the frequency ranges of either ID speech ( 150-5.50-150 Hz) or AD speech (15&275-150 Hz). Infants were shown to discriminate the two tone sweeps in a habituation-novelty experiment but revealed no preferential responding in that experiment or in one using an operant head-turning procedure, even when the ID-range sweep was tested against an unmodulating 150-Hz tone. Using a visual-fixation procedure to study auditory preference, Cooper and Aslin (1990) demonstrated that 2-day-olds and 1 -month-olds looked at a checkerboard pattern longer when the contingent result was the presentation of ID speech than when AD speech resulted. Cooper and Aslin (1994) later tested Fo versions of their ID and AD stimuli and found no differences in I-month-olds’ looking times except on the first trial, where there was significantly longer looking in response to the ID stimulus. They also demonstrated that low-pass filtered versions of their ID and AD speech segments did not elicit differential responding from the infants, even though the two low-pass stimuli were shown to be discriminable.

A number of factors differed between the Femald and Kuhl (1987) and Cooper and Aslin (1994) studies, including behavioral paradigm, age of participants, and the specific stimuli used. These differences make the inconsisten- cies in results difficult to interpret. However, an additional study suggests that frequency modulation alone may not be a strong determinant of infant attentional responses. Kaplan and Owren (1994) found that, following a fixed-trial habituation sequence with a checkerboard pattern, a 10-s train of either rising, falling, or bell-

shaped frequency-modulated vocalic sweeps taken from maternal ID speech elicited significant recovery of looking. Importantly, when the checkerboard pattern was retested alone after the termination of the sweep train, looking remained elevated only in the rising sweep condition. This effect, termed Thompson-Spencer dishabituation, refers to the renewed response to a familiarized stimulus when it is retested after the introduction of a novel stimulus (Kaplan, Werner, & Rudy, 1990) and is thought to be mediated by an increase in the infant’s state of arousal which is elicited by the novel (in this case speech) stimulus (Kaplan et al., 1990; Thompson & Glanzman, 1976). When the dishabituating properties of the natural rising sweep were compared to those of a computer-based synthetic analog to the Fu, significant differences in responding emerged, with mod- estly but significantly less responding to the Fu stimulus.

The conflicting reports on the role of frequency modulation in mediating effects of ID speech on infant attention point to the need for additional research, not only to further assess the effects of frequency modulation but also to identify other relevant stimulus characteristics. Colombo (1985) found that visual attention increased during presentations of tone complexes as the number of harmonics increased. Colombo and Horowitz (1986) suggested that frequency modulation and spectral composition combine to determine the effectiveness of ID speech. Along similar lines, Cooper and Aslin (1994) posited that frequency modulation com- bines with the presence of higher-frequency harmonic components and temporal discontinuities (i.e., during periods of no voicing) to determine the attentional response to speech stimuli.

A series of four experiments was carried out to replicate the differential attentional responses to ID and AD speech in the habituation-dishabituation paradigm and to investigate the roles of frequency modulation and spectral composition in mediating those effects. Use of the habituation-dishabituation paradigm is advan- tageous because it provides a quick and simple behavioral assay of infant visual responsiveness and is known to be sensitive to the effects of auditory stimuli (Colombo & Horowitz, 1986; Culp, 1975; Kaplan, Fox, Scheuneman, & Jenkins, 1991; Kaplan & Owren. 1994; Self, 1975).

Dishabituation by ID and AD Speech 211

Another motivation for employing the habit- uationdishabituation paradigm was to assess the effectiveness of ID and AD stimuli for eliciting Thompson-Spencer dishabituation, an effect that, as noted above, is distinguishable from the response to the novel stimulus itself (Kaplan & Werner, 1986, 1987). Because the renewed responsiveness in Thompson-Spencer dishabituation occurs after the novel stimulus has terminated, it must be mediated by some lingering consequence of the novel stimulus (e.g., sustained arousal; Thompson & Spencer, 1966). This form of dishabituation has been demonstrated in human infants a number of times in the past using novel checkerboard pat- terns or square-wave tones (Kaplan & Werner, 1986; 1987; Kaplan, Fox, Scheuneman, & Jenkins, 199 1, see General Discussion). Because it is routinely asserted that ID speech more effectively modulates infant arousal levels than does AD speech (Femald, 1984, 1992; Papousek, Papousek, & Symmes, 1991), it is of interest to test the effects of these types of stimuli in the Thompson-Spencer dishabituation paradigm.

EXPERIMENT 1

If an ID speech segment has greater effects on infant attention than does an AD speech segment, it can be expected to elicit greater recovery of looking when it is compounded with a familiar visual stimulus. In addition, however, an ID speech segment should produce greater Thompson-Spencer dishabituation following the single compound presentation. This predic- tion was tested in Experiment 1.

Subjects

Method

Thirty healthy, full-term, 4-month-old infants (12 males, 18 females), averaging 124 days of age (range = 117-132 days) participated. Infants were identified using newspaper birth announcements from the Boulder-Denver metropolitan area. Appointments were made with the parents by telephone. In all experiments reported here, infants came predominantly from white, middle-class families. Data from an additional 15 infants were not included in the analysis. Six infants were excluded due to continuous crying (defined as intense crying for at least 50 s consecutively at any point during the test), 5 showed no interest in the projection screen (defined as cumulative fixation of 0.0 s over a consecutive span of 50 s), 2 fell asleep, and 2 were excluded due to experimenter error. Lack of interest in the projection screen applied to infants who were calm but engrossed in some other detail of the testing situation, usually their own hands or feet, such that failing to fixate the projection screen can be viewed as a

competing response to other stimuli that possessed attributes of color and movement.

Apparatus

A standard infant car seat was located in front of a 10.8cm square rear-projection screen, itself mounted in the center of a large black wooden panel. The screen was roughly 42 cm from the infant’s head. A video camera (Hitachi Model HV 725U) was positioned on the other side of the wooden panel and provided a full-face view of the infant through a small round aperture cut in the panel, 1.9 cm to the infant’s left of the projection screen. The infant’s field of view was restricted by a black Plexiglas hood which was moved into position over the infant’s head, flush against the wooden panel, immediately before the start of a session. A loudspeaker (Sony Model SS95) was located at midline behind the car seat, angled up slightly toward the infant’s back.

One observer viewed the infant on a 16-in. black-and- white video monitor (Panasonic WV-5470) which was located in the experimental room. For a randomly selected 37% of the infants, a second, independent, observer was located in an adjacent room watching the infant on an identical video monitor. Although this observer was generally aware of the design of the experiment, he or she was blind with respect to treatment condition and could not tell when specific visual or auditory stimuli were presented. The other observer was not blind to condition and could see and hear the stimuli. Each observer used a hand-held microswitch to signal to a microcomputer an infant’s visual fixation. The cornea1 reflection technique was used to esti- mate visual fixation continuously throughout the test. Observers pressed the microswitch when the reflection of the projection screen was judged to be centered on the infant’s pupil. Interobserver reliability was calculated by correlating the total fixation time on each 10-s trial for the two observers, The mean interobserver correlation was .93 (SD = .05, range = .81-.99).

Stimuli were presented using two Kodak Carousel slide projectors (Model 760H) outfitted with Uniblitz shutters. One projector contained a slide of a check pattern and the other a clear slide of the same space-average luminance. The two projectors were situated at right angles to a beam-splitter plate, which allowed one beam to pass directly through to the projection screen, while the other projector’s beam was reflected by the mirror side of the plate such that it also fell on the projection screen. Shutters, which provided precise temporal control over visual stimulus presentations, were controlled by a Commodore-128 microcomputer, which also tabulated data concerning microswitch closure.

The visual stimuli were a 4 x 4 achromatic check pattern and a uniformly illuminated field of equal space-average luminance. Auditory stimuli consisted of two approximately 10-s speech segments recorded from an adult female. The ID segment was recorded using a MacRecorder microphone and software (Farallon Computing) and a Macintosh IICX computer while the adult female spontaneously talked to her 4-month-old infant. The MacRecorder has S-bit resolution and a sampling rate of 22 kHz, with anti-aliasing filtering above 11 kHz. The sentence used as the ID speech segment which was extracted from a 3-min recording was “round and around she goes, and where she stops, nobody knows.” This particular sentence was selected because it possessed several key features that are characteristic of ID speech, including high average frequency and exaggerated frequen-

212 Kaplan, Goldskin, Huckeby, Owren, and Cooper

TABLE 1 Characteristics of ID and AD Speech Fu Contours From Experiment 1

phrase ID/AD Max (Hz) Min (Hz) M Hz Duration (s) dB Range

“Round and around” ID 534 217 399.7 1.2 60-82 AD 260 210 243.6 0.7 70-73

“She goes” ID A34 260 305.6 0.4 62-70 AD 304 184 233.1 0.3 65-71

“Where she stops” ID 521 271 372.2 0.6 62-71 AD 260 206 240.4 0.3 66-70

“Nobody knows” ID 521 157 356.9 1.2 62-67 AD 228 171 199.2 0.5 67-70

cy sweeps. During a session after the ID phrase had been identified, the same woman was invited to record an AD version of the sentence with no infant present. Table 1 sum- marizes the minimum, maximum, and mean of the Fo, as well as the duration of each phrase and amplitude range of each phrase in dB (SPL). The ID sentence contained higher mean Fo, greater frequency modulation, greater word duration, and greater amplitude range than the AD phrase, consistent with past findings (Cooper & Aslin, 1990; Femald & Simon, 1984; Femald et al., 1989).

In order to equate the durations of the ID and AD stimuli at 10.2 s, the ID phrase was repeated twice, and the AD phrase, because its words were of shorter durations, was repeated three times. Background noise from slide projectors measured near the infant’s head was 58 dB (SPL).

At the start of the session, the infant was placed in the car seat, and the hood was put into place. The test sequence was initiated by the experimenter when it was judged that the infant was fixating the projection screen. Thereafter, stimulus presentations were under computer control. Test onset was therefore infant controlled, but all other stimulus changes were on a fixed-trial schedule. Infants were randomly assigned to one of two groups, with the constraint that the total number of infants be equal, and that there be an equal number of males and females in each group. All infants were given 12 10-s presentations of the 4 x 4 check pattern, with 10-s interstimulus intervals (ISIS) between check pattern presentations. During ISIS, the projection screen was uniformly illuminated. Auditory stimuli were presented on/y during the ninth 10-s check pattern presentation (with the check pattern and auditory stimuli occurring simultaneously). On this trial. one group of infants heard the ID speech segment, and the other group heard the AD speech segment. At no other time during the 240-s session did infants in either group hear this recording or any other recording. Speech stimuli were presented using the Macintosh computer by an experimenter who was isolated from either observer.

Results and Discussion

Figure 1 presents mean fixation times as a function of all 12 presentations of the 4 x 4 check pattern for infants in the two conditions. Mean fixation decreased from Trial 1 to Trial 8 (before the introduction of speech segments) in both conditions. A 2 (later speech segment

type) x 8 (trials) mixed two-factor analysis of variance (ANOVA) was carried out on these data. That analysis yielded a significant effect of trials (epsilon correction factor’ = 0.79), F(6,155) = 3.33, p < .Ol, but no effect of subgroups, F( 1,22) = 0.25, and no interaction between subgroups and trials, F(6,155) = 1.17, ps > .30.

Relative to the final habituation trial (Trial S), mean fixation of the check pattern increased on Trial 9. both when the ID and the AD

7 i

3

2 I1-.__ 1 ~__ A..,_ .I __L

1234567 8 9 10 11 12

TRIALS

Figure 1. Mean fixation time in seconds is plotted as a function of the 12 presentations of the 4 x 4 check

p”” tern. Speech stimuli were resented only on Tria tt

9. Closed circles denote data r infants presented the ID

I= h segment on Trial 9, and open circles denote

ata for infants presented the AD speech s ment on Trial 9. Standard error bars are included “B or Trials 9-10.

’ The epsilon factor corrects for an asymmetrical covatiance matrix in repeated-measures data and is used as a multiplier for the degrees of freedom. The reader will notice different degrees of freedom in different analyses in this article that contain the same number of groups, trials, and participants. These differences reflect differences in the epsilon value calculated for each ANOVA.


speech segments were presented. Mean fixation increased from 5.0 to 6.9 s in the ID condition and from 4.2 to 6.2 s in the AD condition. A mixed two-factor ANOVA produced a significant effect of trials (E = 1 .OO), F( 1,28) = 21.35, p < .OOl, but no significant effect of speech type (ID vs. AD), F( 1,28) = 1.35, p > .25, and no significant interaction between speech type and trials, F( 1,28) = 0.13. The results suggest that the ID and AD speech segments were equally effective at restoring fixation of the 4 x

4 pattern on Trial 9. However, when the 4 x 4 pattern was again

presented alone 10 s later, differences between the groups were evident. Comparing fixation on Trial 8 to that on Trial 10, there was an increase from 5.0 to 6.7 s in the ID condition and a slight decrease from 4.2 to 4.1 s in the AD condition. The mixed two-factor ANOVA performed on these data revealed a nonsignificant effect of trials (E = l.OO), F(1,28) = 3.86, p < .06, a significant effect of speech type, F(1,28) = 5.38, p < .03, and a significant interaction between speech type and trials, F(1,28) = 4.90, p < .04. This interaction was further investigated by performing within-subjects t tests comparing fixation on Trial 10 to that on Trial 8 in each group. Fixation increased significantly in the ID condition, t(14) = 3.35, p < .Ol, but did not change in the AD condition, t( 14) = 0.16.’

Because fixation time on Trial 10 was significantly greater for the ID than the AD speech segment when there had been no difference on Trial 9, fixation values during the ISIS before and after Trial 9 were examined to determine if ID infants had simply continued to fixate the projection screen after the voices terminated on Trial 9. Mean fixation values for the ISIS before and after Trial 9 were, respectively, 3.5 and 4.4 s for ID infants and 2.8 and 4.6 s for AD infants. Analysis of these data produced a significant effect of trials, F(1,28) = 10.88, p < .Ol, but no effect of speech type, F( 1,28) = 0.06, and no interaction between these two variables, F( 1,28) = 1.17, p > .25. Thus, IS1 fixation increased significantly after the speech stimuli

2 In every case in which significant increases in looking occurred between Trials 8 and 10, z tests showed that a significant proportion of the individual infants increased duration of looking by at least 0.1 s, p < .05. Individual test results are not reported here in order to save space.

were introduced, but similarly in the two groups. After Trial 10, mean pattern fixation in the two conditions converged, and neither group exhibited elevated fixation relative to Trial 8 values.

The results of this experiment show that 4- month-old infants who had been familiarized with a checkerboard pattern under a fixed-trial habituation arrangement exhibited significant recovery of looking during brief presentations of either an ID or an AD speech segment. However, Thompson-Spencer dishabituation was evident only following the ID speech segment.

A potential explanation for the differential Thompson-Spencer dishabituation is that “response inertia” was greater in the ID condition. That is, for whatever reason, infants in the ID speech condition were more likely to contin- ue to fixate the projection screen after speech segment presentations and, therefore, noticed the onset of the next pattern presentation sooner or fixated longer. Because latencies to the first fixation were not recorded in this study, it is not possible to say whether infants hearing the ID speech segment noticed trial onset faster. However, there was no difference in the amount of fixation in the IS1 after the termination of the speech segments and before pattern retest, offering no support for an explanation based on differential response inertia.

A possible concern with the experiment is the lack of a “no-change” control group which would insure that responses on Trials 9 and 10 were not attributable to spontaneous recovery. The absence of Thompson-Spencer dishabituation in the AD condition lessens the likelihood that spontaneous recovery played a strong role, but that possibility cannot be ruled out. Indeed, there appeared to be some spontaneous recovery on Trials 4 and 7 in the ID condition (Figure 1). Although the no-change control was not included here, it has been studied in previous experiments using the same procedure. Following familiarization with eight 10-s presentations of a 4 x 4 checkerboard pattern, no recovery of visual attention has been observed during or after a subsequent no-change trial, the omission of the scheduled ninth pattern presentation, or the presentation of a weak tone (Kaplan, Fox, Scheuneman, & Jenkins, 1991, Experiments 1 and 4). For reasons of economy, these control conditions were not repeated here.


EXPERIMENT 2 PPXdUW

The purpose of Experiment 2 was to replicate the differential Thompson-Spencer dishabituation effect using stimuli that had been shown by Cooper and Aslin (1990) to differentially affect visual responding in I-month-olds and 2- day-olds. Replication of differential responding to ID and AD speech with Cooper and Aslin’s stimuli would extend the generality of the effect and help rule out the possibility that the dishabituation and “preference” paradigms may produce inherently different results due to pro- cedural discrepancies. Experiment 2 also differed from Experiment 1 in that latencies to the first response on Trials 8, 9 and 10 were recorded.

Subjects

Method

The basic procedure was the same as in Experiment 1 except that the stimuli from Cooper and Aslin (1990) were substituted for those used in Experiment 1. In this experiment, a second observer was present on 100% of the tests. As in Experiment 1, this observer was in a separate room, was kept blind with respect to stimulus condition, and could not hear the auditory stimuli. Unlike Experiment 1, the data presented below were taken from this “second” observer and not from the “first” observer who was in the testing room with the infant. The change was designed to further reduce the possibility of experimenter bias influencing the results by having a blind observer who was not in the testing room and who was unaware of treatment condition or the moment-to-moment stimulus presentations. Infants were again randomly assigned to ID and AD conditions, with the constraints that the total numbers in each group be equal and that there be equal numbers of males and females in each group. The mean interobserver reliability score was .92 (SD = .06, range = .77-.99).

The participants were 40 healthy, full-term, 4-month-olds (IX males, 22 females), averaging 121 days (range = 1155132), recruited as in Experiment 1. An additional 26 infants were tested but not included in the analysis due to crying (n = 17). inattention (n = 8), or sleeping (n = 1).

Apparatus The apparatus was the same as in Experiment 1 except that a JBL Control 1 loudspeaker was used instead of the SONY speaker. The ID and AD recordings were taken from auditory preference studies with 2-day-olds and I-month-olds con- ducted by Cooper and Aslin (1990). The speech segments were recorded from an adult female who was instructed to speak the following four sentences as if she were either talk- ing to an infant or an adult: “Good morning. How are you today? What are you doing? Let’s go for a walk.” The recording was selected by Cooper and Aslin (1990) from among four others based on a test in which 10 adults rated the appropriateness of the segments as speech directed toward either an infant or an adult. Table 2 lists information about the ID and AD sentences’ Fo ranges, average ftmda- mental frequencies, sentence durations, and amplitude ranges. The ID segment was presented only once, with a l-s pause between each sentence (total duration = 9.2 s). The AD segment was presented twice, with a 0.6-s to 0.8-s pause between each sentence (total duration = 9.8 s).

Figure 2 shows the mean fixation times on all 12 trials of the experiment for infants in the ID and AD conditions. Mean fixation decreased from Trial 1 to Trial 8 (before the introduction of the speech segments). A 2 (later speech segment type) x 8 (trials) mixed two-factor ANOVA produced a significant effect of trials (E = 0.81), F(6,215) = 2.34, p < .05, but no significant effect of subgroups, F( 1,3 1) = 0.60, or of the interaction between subgroups and trials, F(6,215) = 0.62.

As in Experiment 1, introduction of either the ID or AD speech segment on Trial 9 was accompanied by a significant increase in fixation relative to Trial 8. Mean fixation increased in the ID condition from 3.2 s to 6.2 s, and in the AD condition from 3.1 s to 5.5 s. A mixed two-factor ANOVA on the fixation data for Trials 8 and 9 indicated a significant effect of trials (epsilon = 1 .OO), F( 1,38) = 25.5 1, p < .OO 1, but no effect of speech type (ID vs. AD),


TABLE 2 Characteristics of ID and AD Speech Fo Contours From Experiment 2

Phrase ID/AD Max (Hz) Min (Hz) M Hz Duration (s) dB Range

“Good morning”

“How are today?” you

“What are doing?” you

“Let’s for a walk” go

ID 637 185 366.7 1.4 65-79 AD 325 163 233.7 0.7 65-75 ID 650 158 260.9 1.4 66-82

AD 391 162 239.0 0.8 63-76 ID 640 164 336.7 1.4 63-79

AD 347 166 281.9 0.8 62-75 ID 653 155 266.9 1.8 62-74

AD 360 206 236.2 0.9 62-70

“Good Morning...” Stimuli

*ID on Trial 9

0 AD on Trial 9

I

Dirhabituation by ID and AD Speech 215

F(1,36) = 2.18, p > .15, or of the interaction of these two variables, F(2,71) = 1.67, p > .20. As in Experiment 1, mean pattern fixation on Trials 11 and 12 converged in the ID and AD conditions.

/ 1, / 1 J 1 , I

1 2 345676 9 10 11 12

TRIALS

Figure 2. Mean fixation time in seconds is plotted as a function of the 12 presentations of the 4 x 4 check tern. Speech stimuli were resented only on Tna 9.

I! * p”

Closed circles denote data r infants presented the ID h segment on Trial 9, and open circles denote

rfor infants who heard the AD speech Tent on Trial 9. Standard error bars are included r Trials 9-10.

F(1,38) = 0.19, or the interaction between speech type and trials, F( 1,38) = 0.76.

Again, as in Experiment 1, fixation values on Trial 10 remained significantly above those on Trial 8 only in the ID condition. Mean fixation times on Trials 8 and 10 were, respectively, 3.2 and 5.2 s in the ID condition and 3.1 s and 3.7 s in the AD condition. Analysis of these data yielded a significant effect of trials, F( 1,38) = 9.69, p < .Ol, and a significant interaction between speech type and trials, F( 1,38) = 5.30, p < .03, but no overall effect of speech type, F(1,38) = 1.22, p > .25. Within- subject t tests indicated that fixation increased significantly from Trial 8 to Trial 10 in the ID condition, t(19) = 4.28, p < .Ol, but not in the AD condition, t( 19) = 0.53.

Mean fixation times during the ISIS before and after Trial 9 were, respectively, 3.0 s and 4.1 s in the ID condition and 3.0 s and 3.7 s in the AD condition. There were no significant effects of trials, F( 1,38) = 2.28, p > .lO, speech type, F( 1,38) = 0.02, or their interaction, F( 1,38) = 0.04.

The findings from Experiment 2 replicated those from Experiment 1 and extend the generality of the differential Thompson-Spencer dishabituation effect to different ID and AD stimuli. They also corroborate Cooper and Aslin’s (1990) results, which were obtained with younger infants using a different behavioral paradigm but with the same stimuli.

As in Experiment 1, no differences were found in fixation during the IS1 between Trials 9 and 10, indicating that infants did not look longer on Trial 10 because their Trial 9 responding carried over across the 10-s ISI. Latencies to the first fixation were nominally lower in the ID condition, suggesting that, on average, ID infants responded more quickly after pattern onset. However, the change in latencies from Trials 8 to 10 did not differ significantly between ID and AD conditions, even though fixation times did change significantly.

Thompson-Spencer dishabituation may have been stronger in response to ID speech segments than AD speech segments because the AD speech segments had to be presented a greater number of times to equate total stimulus duration. It is therefore possible that, if 4- month-old infants respond to the linguistic content of speech, they may have habituated more (or sensitization may have decayed more) during the presentation of the AD speech segment than during the presentation of the ID speech segment on Trial 9. This, and not the differential attentional properties of ID and AD speech, may have accounted for the differential responding on Trial 10.

Mean latencies to the first fixation on Trials 8, 9, and 10 were 2.3, 1.5, and 1.6 in the ID condition, and 3.6, 1.4, and 3.3 in the AD condition, respectively. An analysis of variance carried out on these data produced a significant effect of trials (E = 0.94), F(2,36) = 4.02, p < .05, but no significant effect of speech type,

This hypothesis seems implausible for several reasons. First, 4-month-olds seem not to respond to linguistic content (Cooper & Aslin, 1994; Femald & Kuhl, 1987). Second, it seems unlikely that much habituation (or decay of sensitization) could have occurred as a result of a single additional presentation of a speech segment that was roughly 3 s in duration. Finally, the data argue against this possibility. Although comparisons between experiments should be made with caution, there were large differences in mean Thompson-Spencer dishabituation between the ID condition from Experiment 1


and the AD condition from Experiment 2, even though the number of repetitions was the same in those two conditions. Furthermore, there was, if anything, slightly less mean dishabituation in the ID condition from Experiment 2, in which the ID signal was presented only once, relative to that in Experiment 1, in which it was presented twice. The conclusion that the number of repetitions of speech segments cannot account for dishabituation failures finds support in studies by Kaplan and Owren (1994), who obtained significant Thompson-Spencer dishabituation in response to an ID stimulus comprised of five repetitions of a 0.5-s rising frequency-modulated sweep. In the same study, however, five repetitions of a falling or bell- shaped sweep did not elicit significant dishabituation.

As in Experiment 1, there was no no-change control group in Experiment 2. However, the results of this experiment cast further doubt on the possibility of spontaneous recovery in the ID condition because recovery occurred again on Trial 10, and not before or after. The fact that significant recovery in response to ID speech was obtained twice on Trial 10 suggests that recovery is tied to the presentation of the speech stimulus on Trial 9.

EXPERIMENT 3

In this experiment, a first attempt was made to isolate an acoustic characteristic of ID speech that might be critical in influencing Thompson- Spencer dishabituation. The dishabituating properties of the computer-synthesized Fu ID and AD stimuli from Cooper and Aslin (1994) were tested.

Method

Forty healthy, full-term, 4-month-old infants (16 males, 24 females), averaging 120 days of age (range = 116-125) participated. An additional 25 infants were tested but excluded from the analysis due to crying (n = 20) or inattention (n = 5).

The apparatus was identical to that in Experiment 2. The speech stimuli have been described by Cooper and Aslin (in press). Briefly, the stimuli from Experiment 2 were initially digitized at a sampling rate of 25 kHz, individually stored, and then analyzed using the Pitch Edit program from Micro Speech Lab software (which produces a series of Fo values for each utterance, with each value based on a 20.ms sampling window). Each waveform was also sub- ,jected to a detailed visual inspection to check the derived

Fo values. Fo values for each sentence were entered into a tone-generator program (with silence inserted whenever a period of no-voicing occurred) to create synthetic F() contours. The amplitude of the speech recordings, measured at a place in the apparatus corresponding to the location of the infant’s head, was held constant and equated at 6X dB (SPL).

The testing procedure was identical to that in Experiments 1 and 2 except that the Fu contours were substituted for the full signals on Trial 9. Two observers were present on all tests, as in Experiment 2. with a mean interobserver correlation of .91 (SD = .07, range = .76-.99).


Figure 3 presents the mean fixation values for infants in ID and AD conditions on all 12 trials. Mean fixation decreased from Trial 1 to Trial 8, and a mixed two-factor ANOVA revealed a significant effect of trials (E = 0.78), F(5,210) = 5.73, p < .OOl, but not of subgroups (ID speech vs. AD speech), F(1,30) = 2.20, p > .15, or their interaction, F(5,210) = 0.50.

Introduction of the Fu speech segments on Trial 9 elicited increases in looking times in both the ID and the AD groups, with mean increases from 4.5 to 6.3 s in the ID condition and from 3.8 to 5.1 s in the AD condition. Analysis of the fixation data for Trials 8 and 9 showed a significant effect of trials, F( 1,38) = 10.53, p < .O 1, but no effect of speech type, F( 1,38) = 1.50, I, > .20, and no interactions, F( 1,38) = 0.35.

When the checkerboard was retested alone on Trial IO, mean fixation remained roughly

88

1 2 3 4 5 6 7 I3 3 10 11 12 13

TRIALS

Figure 3. Mean fixation time in seconds is plotted as a function of the 12 presentations of the 4 x 4 check tern. Speech stimuli were

f! resented only on Tria t”” 9.

Closed circles denote data r infants presented the ID- Fo speech segment on Trial 9, and open circles denote data for infants who heard the AD-F0 speech segment on Trial 9. Standard error bars are included for Trials 9-12.


unchanged relative to Trial 8, with values on Trials 8 and 10 of 4.5 s and 4.4 s in the ID condition and 3.8 s and 3.4 s in the AD condition, respectively. Analysis of the Trial 8 versus Trial 10 data for the two groups produced no significant effects of trials, F( 1,38) = 0.03, speech type, F(1,38) = 0.70, or their interactions, F( 1,38) = 0.08, each p > .40.

As can be seen in Figure 3, mean looking on Trial 12 was higher in the ID condition than in the AD condition, t(38) = 3.41, p < .002. An additional mixed two-factor ANOVA was carried out comparing fixation on Trial 8 to that on Trial 12 in the two conditions, to determine whether, even though neither group exhibited significant Thompson-Spencer dishabituation, fixation habituated further in the AD condition but not in the ID condition. That analysis produced a significant effect of speech type, F( 1,38) = 4.86, p < .04, but no significant effect of trials, F( 1,38) = 1.06, p > .25, and, importantly, no significant interaction between speech type and trials, F(1,38) = 2.10, p > .15. Therefore, although mean fixation on Trial 12 was significantly higher in the ID condition than in the AD condition, there was no reliable difference in the amount by which fixation in the two groups changed between Trials 8 and 12. This suggests that, relative to responding on Trial 8, responding after Trial 10 was not better maintained in the ID condition than in the AD condition.

Consistent with Cooper and Aslin’s (1994) results with 1-month-olds using the visual-fixation auditory preference paradigm, these findings show that there is no significant difference in 4-month-olds’ visual dishabituation following ID and AD Fe contour stimuli.

EXPERIMENT 4

Experiment 4 was designed to study the effects of both frequency modulation and the number of harmonics on infants’ responses to ID speech. Four computer-based synthetic stimuli were modeled on the ID “Good morning . . .” stimuli from Experiment 2. The synthetic intact stimulus was comprised of sine-wave-based simulations of the Fe plus the first five harmonics (Ht to Hs). This stimulus was included as a baseline comparison for the other synthetic stimuli and to insure that a synthetic stimulus of appropriate frequency modulation and with five harmonics above the Fo could dishabituate

visual attention. The Fo-only stimulus was analogous to the ID-F0 stimulus from Experi- ment 3 and was included as a replication condition. The Fe plus Ht synthetic stimulus contained sine-wave simulations of the complete fundamental and first harmonic and was included to determine if a frequency-modulated stimulus with the Fo and only the first harmonic would dishabituate visual attention. Finally, the harmonics-only condition was one in which the five harmonics were present but the Fo was missing. This condition was included to determine if the Fe is even necessary for the Thompson-Spencer dishabituation effect. Clark- son (1992) has shown, using an operant head- turning procedure, that 7-month-olds can discriminate tone complexes with and without the Fo. They nevertheless group them in the same perceptual category.

All four of these synthetic stimuli were frequency modulated, but they varied in their spectral composition and linguistic intelligibili- ty. The Fo-only stimulus had no harmonics and was linguistically unintelligible. The Fo plus Ht stimulus had one added harmonic and was somewhat intelligible. The intact synthetic stimulus had the first five harmonics above the Fo and was the most intelligible. The harmonics-only stimulus also had the first five harmonics above the Fo and intelligible linguistic content but lacked the low-frequency component of the intact signal.

Subjects

Method

Eighty healthy, full-term, 4-month-olds (40 males, 40 females), averaging 124 days in age (range = 117-128 days) participated. An additional 33 infants were tested but their data were excluded from the analysis due to crying (n = 22) inattention (n = 8), or experimenter error (n = 3).

Apparatus

The testing apparatus was the same as that described in Experiment 2. The stimuli were synthesized using the SIG- NAL (Engineering Design, Belmont, MA) system imple- mented on a Gateway 2000 486/2X microcomputer with a SVGA Sony Trinitron monitor. The natural ID speech stimulus from Experiment 2 was first digitized at a IO-kHz sampling rate with low-pass filtering at 4 kHz. A separate digital spectrogram was then made of each phrase using 1024-point Fast Fourier Transforms (FITS) taken at 2-ms intervals with a resulting frequency resolution of 9.8 Hz. A Fo tracing was derived from each phrase by locating the peak amplitude value in each FFI within a cursor-bounded rectangular area of the digital spectrogram around the Fo contour. Time ranges were measured from a waveform rep- resentation displayed simultaneously with the spectrogram.


Atonal portions of each phrase were represented as silent segments of 0 Hz.

Amplitude measurements of the fundamental frequency and first five harmonics were made from the digital spectrogram at intervals of approximately 20 ms. Measurements were derived from power spectra based on 2%point FFTs. Values from segments in which there was no fundamental frequency energy were set to -60 dBv. After conversion to voltage values, each amplitude model was extended to the original phrase duration and smoothed with a 25.point window.

Synthesis was based on the fundamental frequency contour extracted from each phrase and the six amplitude func- tions. Frequency-modulated sine waves were first created from the fundamental frequency tracing and at integer mul- tiples of its values up to the fifth harmonic. Each sine wave was then amplitude modulated by multiplying it by the appropriate voltage-based model. The four synthetic stimuli were created through addition of the requisite harmonics and concatenation of the four phrases with 1 s of silence between each. To equate the amplitudes of the various stimuli, each synthetic phrase was first adjusted to an amplitude value of 1.0 v root mean square (RMS). Each overall synthetic stimulus was then similarly set to 1 .O RMS.

rc

9 10 11 12

As in previous experiments, sound pressure level read- ings were taken for the four synthetic stimuli at a place in the testing apparatus corresponding to the location of the infant’s head. Maximum decibel measurements for the four phrase stimuli were 70, 68, 67, and 67 dB with the synthetic intact stimulus, 71, 69.5, 68, and 69 dB with the Fo-only stimulus, 72, 70, 69, and 69 dB with the Fo plus HI stimulus, and 69, 68.5, 68.5, and 675 dB with the harmonics- only stimulus, respectively.

PlYlCdUl?

Figure 4. Mean fixation time in seconds is plotted as a function of the 12 presentations of the 4 x 4 checkerboard for infants in the synthetic intact, Fo-only, Fo plus HI, and harmonics-only conditions of Experiment 4. Speech stimuli were presented only on Trial 9.

factor ANOVA performed on the response data for Trials 8 and 9 showed a significant effect of trials, F( 1,76) = 38.28, p < .OOl, but showed no effect of synthetic stimulus type, F(3,76) = 0.43, or of the interaction between stimulus type and trials, F(3,76) = 0.55.

Twenty infants were randomly assigned to each of four groups, with the constraint that there be an equal number of infants in each group and equal gender distributions. The procedure was the same as in previous experiments except with respect to the synthetic speech-based stimuli presented on Trial 9. A second observer was present for all tests, with a mean interobserver correlation of .94 (SD = .OS. range = .73-.99).


Figure 4 presents the mean looking times over all 12 trials for infants in each of the four groups. A mixed two-factor ANOVA carried out on the data for the first 8 trials produced a significant effect of trials (E = 0.77), F(5, 410) = 4.22, p < .OOl, but neither the subgroups effect nor the subgroups x trials interaction was significant, F(2,59) = 0.34 and F(l6,410) = 0.79, respectively.

Differences did appear when looking times on Trials 8 and 10 were compared for each group. Mean looking times increased from 4.0 to 5.4 s in the synthetic intact condition, from 4.9 to 5.7 s in the Fo-only condition, and from 4.8 to 5.8 s in the Fo plus HI condition, but decreased from 4.6 to 3.5 s in the harmonics- only condition. A mixed two-factor ANOVA performed on these data yielded no significant effect of synthetic stimulus type, F(3,76) = 1.20, p > .30, or trials, F( 1,76) = 3.30, p < .08, but did produce a significant synthetic stimulus type x trials interaction, F(3,76) = 3.83,~ < .02. Subse- quent within-subjects t tests showed that gaze duration increased significantly in the synthetic intact condition, t(19) = 2.53, p < .05, did not change in the Fo-only condition, t( 19) = 1.40, p > .15, or the Fo plus Hi condition, t(19) = 1.52, p > .15, and decreased significantly in the harmonics-only condition, t( 19) = 2.3 1, p < .05.

The presentation of synthetic speech-based A follow-up ANOVA was carried out on the segments on Trial 9 was accompanied by difference scores between Trials 10 and 8 for increases in mean fixation in all four groups. each group, F(3.76) = 3.50, p < .02, and the Mean fixation increased from 4.0 to 6.9 s in the error term from this analysis was used to per- synthetic intact condition, from 4.9 to 6.9 s in form Newman-Keuls tests. These tests revealed the Fo-only condition, from 4.8 to 7.2 s in the significant differences between the harmonics- Fo plus HI condition, and from 4.6 to 6.3 s in only group and each of the other three groups. the harmonics-only condition. A mixed two- 17 < .05. No other differences were significant.


The preceding analysis did not show significant differences between the synthetic intact and Fo-only conditions, although the within- subjects r tests showed a significant Thompson- Spencer dishabituation effect in the synthetic intact condition but not in the Fo-only condition. To further analyze these between-groups differences, an additional ANOVA was performed by pooling the data from Trials 8 and 10 for the Fo groups from Experiments 3 and 4 and comparing them to the pooled data from the natural intact ID stimulus from Experiment 2 and the synthetic intact stimulus from Experiment 4. The justification for pooling the latter data was that the synthetic intact stimulus exactly mimicked the frequency and amplitude modulation of the natural stimulus (up to the fifth harmonic above the Fo) and elicited near identical responding. The pooled data revealed a significant effect of trials, F(1,78) = 14.94, p < .OOl, and a significant speech type x trials interaction, F( 1,78) = 6.16, p < .02, but no overall effect of speech type, F( 1,78) = 0.62. Subsequent within-subjects t tests indicated that looking increased significantly from Trial 8 to Trial 10 for the intact stimuli, t(39) = 4.70, p < .Ol, but not for the Fo stimuli, t(39) = 0.95.

Response latencies were analyzed using a 4 (synthetic stimulus type) x 3 (Trials 8, 9, and 10) mixed two-factor ANOVA. That analysis showed a significant effect of trials (E = 0.90), F(2, 68) = 4.80, p < .Ol, but no effect of synthetic stimulus type, F(3,68) = 0.27, or of the interaction between these two factors, F(5,137) = 0.79. Mean latencies to the first look on Trials 8, 9, and 10 were 3.3, 1.3, and 1.4 s in the synthetic intact condition, 2.4, 1.5, and 1.2 s in the Fo-only condition, 2.1, 1.9, and 1.7 s in the Fo plus Ht condition, and 2.5, 2.3, and 2.0 s in the harmonics-only condition, respectively.

These results replicate the ID-F0 results from Experiment 3 and suggest that the benefit of adding five harmonics to the Fo is modest but significant. Somewhat surprisingly, the harmonics-only stimulus was the least effective at eliciting infant attentional responses.

GENERAL DISCUSSION

In Experiments 1 and 2, Thompson-Spencer dishabituation was observed in response to ID but not AD speech segments. These findings add to the growing body of experiments demonstrating differential responding to ID and

AD stimuli in young infants. They were based on two different sets of ID and AD stimuli, including one shown to elicit differential responding in 1-month-olds when used in a visual-fixation-based auditory preference procedure (Cooper & Aslin, 1990). In contrast to the infants in previous studies, infants in these experiments did not show differences in responding while listening to the ID and AD stimuli (although, with the exception of Cooper & Aslin, 1994, infants in most previous studies did not exhibit ID-AD response differences during the first trial). Rather, differential responding was detected only on the test trial that started 10 s after the termination of speech stimuli. Implications of these data for the study of acoustic determinants of infants’ responses to ID and AD speech, for an arousal-based interpretation of Thompson-Spencer dishabituation, and for interactions in the caregiver- infant dyad are discussed below.

Acoustic Determinants of Differential Responding to ID and AD Speech

In contrast to the full-spectrum ID and AD speech stimuli, neither the ID-Fo nor the AD-F0 stimuli produced significant Thompson- Spencer dishabituation (Experiment 3). Further- more, although dishabituation was not significantly greater in the synthetic intact than in the Fo-only condition in Experiment 4, the difference in dishabituation between pooled Fo data (Experiments 3 and 4) and pooled natural and synthetic intact data (Experiments 2 and 4) was statistically reliable. These outcomes accord well with Kaplan and Owren’s (1994) results with rising frequency-modulated sweeps and also support both Cooper and Aslin’s (1994) and Colombo and Horowitz’s (1986) findings with Fo contour and pure tone stimuli. However, these results are contrary to Femald and Kuhl’s (1987) Fo findings. Overall, our data suggest that frequency modulation in a speech segment’s low-frequency components cannot fully account for the dishabituation elicited by ID speech.

Nevertheless, the Fo contour was an important component of stimuli that elicited significant Thompson-Spencer dishabituation. Not only did the harmonics-only stimulus fail to generate Thompson-Spencer dishabituation, it was also associated with a significantly smaller mean dishabituation effect than were any of the


three stimuli that included the Fu. The significantly greater dishabituation in the synthetic intact relative to the harmonics-only stimulus indicates that the Fo contour is a necessary (but, the Fo findings show, not a sufficient) determinant of the dishabituating power of ID speech.

Overall, although several investigators have demonstrated stronger responding to speech and nonspeech stimuli as harmonics are added to the Fo (Experiment 4; Colombo, 1985; Cooper & Aslin, 1994; Kaplan & Owren, 1994), our findings suggest that neither frequency modulation in the Fo nor the presence of the first five harmonics alone can account for the differential dishabituation in response to ID speech.3 Rather, both factors contribute to the perceptual salience of ID speech for infants. However, frequency modulation in low-frequency components may play a greater role, at least with 4-month-olds.

Although it has been suggested that differences in the ages of the infants studied by Femald and Kuhl (1987) and Cooper and Aslin (1994) may have accounted for differences in observed responding to Fo stimuli, the 4- month-olds tested in the current experiments responded more similarly to Cooper and Aslin’s I-month-olds than to Femald and Kuhl’s 4-month-olds. Perhaps the use of measures involving response elicitation (i.e., head- turning) rather than response elicitation and maintenance (persistent visual gaze) is more pertinent to the obtained discrepancy in findings. Because Thompson-Spencer dishabituation involves a lingering aftereffect of speech stimuli, those stimuli may have to be stronger or possess a more resonant sound (i.e., have the Fu plus harmonics) to affect responding. This interpretation is consistent with Cooper and Aslin’s ( 1994) recent demonstration of differential responding to ID- and AD-F0 stimuli on the first trial in their paradigm, but not over an entire session.

Differences in responding to stimuli varying in presence versus absence of the Fu contour or harmonics may be related to differences in perceived formant structure. A formant is a vocal

‘Of course, if the number of harmonics above the Fo was the primary determinant of responding, then AD speech segments and male voices should be the most effective dishabituators. Both have lower Fos than do female ID voices and hence a greater number of harmonics within a given frequency range.

tract resonance that reinforces sound energy in a particular frequency region and appears as a dark band on a speech spectrogram. The relative locations of the first two or three formants are considered to be particularly important in vowel production and perception (e.g., Joos, 1948; Peterson & Barney, 1952). Unfortunate- ly, no published studies have systematically examined the effects of variations in formant structure on infant responsiveness to ID or AD speech. Spectrographic analysis (not presented here) revealed that only the first two to three formants were in evidence in the synthetic intact stimulus, whereas the natural intact stimulus included three to four formants under 5000 Hz. Nevertheless, these two stimuli each elicited significant Thompson-Spencer dishabituation. Furthermore, the Fo-only stimulus, which had no harmonics and therefore no formant structure, nevertheless elicited only slightly (but significantly) less responding than the synthetic intact stimulus. The harmonics- only stimulus possessed at least two formants but produced no dishabituation. It is therefore not clear that variations in number of formants can easily explain the current results.

Another way to interpret the results of Experiments 3 and 4 is in terms of the perceived naturalness of each stimulus. Adult lis- teners in our laboratory judged the synthetic intact stimulus to be the most similar to the natural intact stimulus and the most natural sounding. In contrast, the harmonics-only stimulus had a “tinny” sound and was judged to be the most artificial sounding. Four-month-olds have considerable experience listening to natural ID speech and may already have formed speech prototypes by that age. The weak response to the harmonics-only stimulus may, therefore, have reflected a perceived dissimilarity to pro- totypical ID speech. On the other hand, the harmonics-only stimulus might have been expected to be particularly effective as a dishabituator. For instance, it included both frequency modulation and five harmonics. Furthermore, its novel, aprototypical sound could just as easily be predicted to produce stronger rather than weaker dishabituation of visual attention.

An Arousal-Based Interpretation of Thompson-Spencer Dishabituation

It has been posited that Thompson-Spencer dishabituation is mediated by stimulus-induced


changes in state, where the term state is intend- ed to include the “general level of arousal, acti- vation, tendency to respond, etc.” of the organ- ism (Groves & Thompson, 1970; see also Castelucci & Kandel, 1976; Thompson & Glanzman, 1976; Thompson & Spencer, 1966). Thompson-Spencer dishabituation of infant looking at a familiar 4 x 4 checkerboard pattern has been demonstrated when stimulation is changed to novel 12 x 12 or 20 x 20 checker- boards, 75 and 84-dB, lOOO-Hz square-wave tones, a checkerboard-tone combination, adult female faces, and rising ID frequency sweeps (Kaplan & Werner, 1986, 1987; Kaplan, Fox, & Huckeby, 1991; Kaplan, Fox, Scheuneman, & Jenkins, 1991; Kaplan & Owren, 1994). In contrast, Thompson-Spencer dishabituation has not been observed with novel but less complex visual stimuli and less intense auditory stimuli, falling ID frequency sweeps, a no-change control, or when the scheduled checkerboard pattern presentation was omitted (Kaplan & Werner, 1986; Kaplan, Fox, Scheuneman, & Jenkins, 1991; Kaplan & Owren, 1994). Thompson- Spencer dishabituation can be eliminated by increasing the time interval between the termination of the novel stimulus and the retest of the familiarized stimulus (Kaplan & Werner, 1986, 1987; Kaplan, Fox, Scheuneman, & Jenkins, 1991) as would be expected if the effect were mediated by a lingering but decaying process. Furthermore, a schema-comparison account of Thompson-Spencer dishabituation can be ruled out on the basis of an experiment in which infants familiarized with a 4 x 4 checkerboard pattern subsequently received varying numbers of presentations of a novel 20 x 20 checkerboard pattern. Although the amount of schema change should increase with more presentations of the initially novel (20 x 20) stimulus, the magnitude of Thompson-Spencer dishabituation decreased.

Also consistent with an arousal-based interpretation are experiments in which some of the stimuli that elicit Thompson-Spencer dishabituation have been shown to affect autonomic responding. For instance, Berg, Berg, and Graham (1971) found that 2-s, 75dB, lOOO-Hz square-wave tones with slow rise times elicited significant heart rate deceleration in 4-month- olds, and that the same stimuli with fast rise times elicited heart rate acceleration. Similarly, Femald (1992) briefly discussed an unpub- lished study showing that vocalic sounds with

frequency ranges corresponding to ID speech produced greater heart rate deceleration than those with frequency ranges corresponding to AD speech.

Research by Pegg et al. (1992) provides further support for the idea that ID speech is more arousing than AD speech. They used an infant- controlled habituation procedure in which a checkerboard pattern and speech segment were compounded on every trial. Following habituation to one speech segment (ID or AD), infants were presented the other speech segment. Pegg et al. found that looking times increased significantly when infants were shifted from ID to AD or from AD to ID, but that the amount of recovery was greater in the AD-ID testing order than in the ID-AD testing order. These findings are consistent with the hypothesis that discriminability and the sensitizing properties of the novel stimulus both contribute to the response when infants are shifted from AD to ID speech (or vice versa).

Several researchers have posited that ID speech more effectively modulates infant arousal levels than does AD speech (Femald, 1984, 1992; Papousek et al., 199 1). The results from Experiments 1 and 2 are consistent with this claim. Based on their observations of mother-infant dyads in different interactional con- texts, Papousek et al. (1991) hypothesized that rising ID frequency sweeps are used by parents to increase arousal levels, whereas falling ID frequency sweeps are used to decrease and sta- bilize arousal levels. Kaplan and Owren’s (1994) demonstration of Thompson-Spencer dishabituation in response to rising but not falling ID sweeps is consistent with this hypothesis. Taken in combination with the studies cited above, these findings support an arousal- based interpretation of Thompson-Spencer dishabituation.

Implications for Attention and Learning in the Caregiver-Infant Dyad

Whatever its cause, the lingering responsiveness following ID speech, although short-lived in the current studies, may have implications for face-to-face interactions between infants and caregivers. Increases in an infant’s visual attentiveness following ID speech may lead to facilitated information processing (Femald, 1984). Indeed, Ryther-Duncan et al. (1993) reported that an ID speech segment that was


reliably and immediately followed by a smiling adult female face acquired the ability to subsequently increase 4-month-olds’ attention to a simultaneously presented novel checkerboard pattern. In contrast, AD speech segments had no effect on looking at the checkerboard after paired presentations with the face. Similarly, neither ID nor AD stimuli presented randomly with respect to the face significantly affected responding to the checkerboard. These results suggest that ID speech better promotes associa- tive learning about a face that immediately fol- lows it than does AD speech.

On the caregiver’s side, the lingering responsiveness of the infant to ID speech may alter adult facial and vocal responses. Studies by Papousek et al. (1991) and Stem, Speiker, and MacKain (1982) have shown that the shape of ID frequency contours used by parents varied with infant state and behavior. This recipro- cal responsiveness may play an important role in early learning in general and early language learning in particular.

In conclusion, the current experiments employed the Thompson-Spencer dishabituation paradigm to replicate previous findings on differential responding to ID and AD speech stimuli, to address contradictory Fu findings in the literature, and to extend the study of acoustic determinants to stimuli that possessed harmonics but lacked the Fo. The results obtained, taken together with those of Kaplan and Owren ( 1994), encourage the arousal-based interpretation of Thompson-Spencer dishabituation. The ultimate value of this interpretation may be to stimulate an investigation of the effects of lingering attentional aftereffects on information processing in the caregiver-infant dyad.

REFERENCES

Berg, K., Berg, W., & Graham, F. (1971). Infant heart rate response as a function of stimulus and state. Psvc‘hophysiology, 8, 3@44.

Castelucci, V., & Kandel, E. (1976). An invertebrate system for the cellular study of habituation and sensitization. In T. Tighe & R. Leaton (Eds.), Habituation: Perspectives from child development. animal behavior, and neurophysiology. Hillsdale, NJ: Erlbaum.

Clarkson, M. (1992). Infants’ perception of low pitch. In L.A. Werner & E.W. Rubel (Eds.), Developmental psychoucoustics. Washington, DC: APA Books.

Colombo, J. (1985). Spectral complexity and infant attention. The Journal of Genetic PsycholoSqyv. 146. 5 19-S26.

Colombo, J., & Horowitz, F. (1986). Infants’ attentional responses to frequency-modulated sweeps. Child Development, 57, 287-291.

Cooper, R.P., & Aslin, R.N. (1990). Preference for infant- directed speech in the first month after birth. Child Development, 61. 158&1595.

Cooper, R.P., & Aslin, R.N. (1994). Developmental differences in infant attention to the spectral properties of infant-directed speech. Child Development, 65. 1663-1677.

Culp, R. (1975). The use of mother’s voice to control infant attending behavior. In F. Horowitz (Ed.), Mono- graphs of the Society for Research in Child Development, 39(5%X Serial No. 166).

Femald, A. (1984). The perceptual and affective salience of mothers’ speech to infants. In L. Feagans, C. Garvey, & R. Golinkoff (Ed%), The origins and growth of communication. Norwood, NJ: Ablex.

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

Femald, A. (1992). Human maternal vocalizations to infants as biologically relevant signals: An evolutionary perspective. In J. Barkow, L. Cosmides, & J. Tooby (Eds.), The adapted mind: Evolutionary pqchology und the generution of culture. New York: Oxford.

Femald, A., & Kuhl, P. (1987). Acoustic determinants of infant perception for motherese speech. Irtfant Behavior and Development, 10,279-293.

Femald, A., & Simon, T. (1984). Expanded intonation contours in mothers’ speech to newborns. Develop- mental Psychology, 20, 104-l 13.

Femald, A., Taeshcher, T.. Dunn, J., Papousek, M., de Boysson-Bardies, B., & Fukui, 1. (1989). A cross- language study of prosodic modifications in mothers’ and fathers’ speech to preverbal infants. Journal of Child Lan~uqe, 16, 477-501.

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

Groves, P., & Thompson, R. (1970). Habituation: A dual- process theory. Psychological Review, 77,4 19450.

Joos, M. (1948). Acoustic phonetics. Lan,qq~e. 24 (Suppl.), l-136.

Kaplan, P., Fox, K., & Huckeby, E. (1991, April). Are faces sensitizing? Paper presented at the meeting of the Society for Research in Child Development, Seattle, WA.

Kaplan, P., Fox, K., Scheuneman, D., & Jenkins, L. (1991). Cross-modal facilitation of infant visual fixation: Temporal and intensity effects. Infant Behavior and Del,e/opment, 14, 83-109.

Kaplan, P., & Owren, M. (1994). Dishabituation of infant visual attention by in 4-month-olds by infant-directed frequency-modulated sweeps. fnfant Behkor and Development. 17. 347-358.

Kaplan, P., & Werner, J. (1986). Habituation, response to novelty, and dishabituation in human infants: Tests of a dual-process theory of visual attention, Journal of Experimental Child Psychology, 42, 199-217.

Kaplan, P., & Werner, J. (1987). Sensitization and dishabituation of infant visual attention. Infant Behuvior cmd Development, IO, 1833197.


Kaplan, P.S., Werner, J.S., & Rudy, J.W. (1990). Habitua- tion, sensitization, and infant visual attention. In C. Rovee-Collier & L.P. Lipsitt (Eds.), Advances in infancy research (Vol. 6). Notwood, NJ: Ablex.

Papousek, M., Papousek, H., & Symmes, D. (1991). The meaning of melodies in motherese in tone and stress languages. Infant Behavior and Development, 13, 539-545.

Pegg, J.E., Werker, J.F., & McLeod, P.J. (1992). Preference for infant-directed over adult-directed speech: Evidence from 7-week-old infants. Infant Behavior and Development, 15.325-345.

Peterson, G., & Barney, H. (1952). Control methods used in a study of the vowels. Journal of the Acoustical Society of America, 24, 175-184.

Ryther-Duncan, J., Scheuneman, D., Bradley, J., Jensen, M., Hansen, D., & Kaplan, P. (1993, March). Infunt- versus adult-directed speech as signals for faces. Paper presented at the meeting of the Society for Research in Child Development, New Orleans, LA.

Self, P. (1975). Control of infant visual attending by auditory and interspersed stimulation. In F. Horowitz

(Ed.), Monographs of the Society for Research in Child Development, 39(5-6, Serial No. 166).

Stem, D.N., Spieker, S., & MacKain, K. (1982). Intonation contours as signals in maternal speech to prelinguis- tic infants. Developmental Psychology, 18,727-735.

Thompson, R., & Glanzman, D. (1976). Neural and behavioral mechanisms of habituation and sensitization. In T. Tighe & R. Leaton (Eds.), Habituation: Perspectives from child development, animal hehav- ior, and neurophysiology. Hillsdale, NJ: Erlbaum.

Thompson, R., & Spencer, W. (1966). Habituation: A model phenomenon for the study of the neural sub- strates of behavior. Psychological Review, 73, 1643.

Werker. J.F., & McLeod, P.J. (1989). Infant preference for both male and female infant-directed talk: A developmental study of attentional and affective responsiveness. Canadian Journal of Psychology, 43, 230-246.

15 January 1994; Revised 17 May 1994 n

Dishabituation of Visual Attention b Infant- Versus Adult ... toward adults (Fernald & Simon, 1984; Femald ... 210 Kaplan, Goldstein, Huckeby, Owren, and Cooper fundamental frequencies

Documents