8081AssignmentResourcePrenatal.pdf

INFANT BEHAVIOR AND DEVELOPMENT 9, 1X3-150 (1986)

Prenatal Maternal Speech Influences
Newborns’ Perception of Speech Sounds*.

ANTHONY J. DECASPER AND MELANIE J. SPENCE
University of North Carolina at Greensboro

Pregnant women recited a particular speech passage aloud each day during their
last 6 weeks of pregnancy. Their newborns were tested with an operant-choice
procedure to determine whether the sounds of the recited passage were more
reinforcing than the sounds of a novel passage. The previously recited passage
was more reinforcing. The reinforcing value of the two passages did not differ for
a matched group of control subjects. Thus, third-trimester fetuses experience
their mothers’ speech sounds and that prenatal auditory experience can in-
fluence postnatal auditory preferences.

prenatal sensory experience auditory perception newborn perception
fetal experience prenatal learning speech perception moternol voice

Human newborns do not act like passive and neutral listeners. They prefer
their own mothers’ voices to those of other females, female voices to male
voices, and intrauterine heartbeat sounds to male voices, but they do not prefer
their fathers’ voices to those of other males (Brazelton, 1978; DeCasper &
Fifer, 1980; DeCasper & Prescott, 1984; Fifer, 1980; Panneton & DeCasper,
1984; Wolff, 1963). Why should newborns prefer some sounds over others?
One hypothesis is that their auditory preferences are influenced by prenatal ex-
perience with their mothers’ speech and heartbeats (DeCasper & Prescott,
1984). Several considerations suggest this hypothesis is plausible.

Third-trimester fetuses hear, or are behaviorally responsive to, sound
(e.g., Bernard & Sontag, 1947; Birnholz & Benacerraf, 1983; Grimwade,
Walker, Bartlett, Gordon, & Wood, 1971; Johansson, Wedenberg, & Westin,
1964; Sontag & Wallace, 1935). Intrauterine recordings taken near term indi-
cate that maternal speech and heartbeats are audible in utero (Querleu &
Renard, 1981; Querleu, Renard, & Crepin, 1981; Walker, Grimwade, &Wood,
1971). Nonmaternal speech, for example male speech, is less audible because
of attenuation by maternal tissue and/or masking by intrauterine sounds
(Querleu & Renard, 1981; Querleu et al., 1981).

* This research was supported by a Research Council Grant from the University of North Carolina
at Greensboro and a generous equipment loan by Professor Michael D. Zeiler. We wish to thank
the medical and administrative staff of Moses H. Cone Hospital, Greensboro, NC and, especially,
the mothers and their infants for making this research possible. Thanks also to G. Gottlieb, R.
Harter. R. Hunt, R. Panneton. K. Smith, and, especially, W. Salinger for their helpful comments
on drafts of the manuscript. Portions of this paper were presented at the Third Biennial Interna-
tional Conference on Infant Studies, March 1982. Austin, TX.

Correspondence and requests for reprints should be addressed to Anthony J. DeCasper,
Department of Psychology, University of North Carolina at Greensboro, Greensboro, NC 27412.

133

134 DECASPER AND SPENCE

The newborns’ preference for their own mothers’ voices requires that
they had some prior experience with her voice, but there is no evidence that the
necessary experience occurred after birth. Fifer (1980) failed to find any relation
between maternal-voice preference and postnatal age, whether the newborns
roomed with their mother or in a nursery, or whether they were breast fed or
bottle fed. Since the maternal voice is audible in utero, and since third-trimester
fetuses can hear, perhaps the necessary experience occurred before birth. In
contrast, newborns show no preference for their own fathers’ voices, even if
they had explicit postnatal experience with his voice. Since male voices are not
very audible in utero, perhaps the absence of a paternal-voice preference indi-
cates the absence of prenatal experience with his voice (DeCasper & Prescott,
1984). The correlations between the presence or absence of specific-voice sounds
before birth, and the presence or absence of specific-voice preferences after birth
suggest that prenatal auditory experiences influence the earliest voice preferences.

Consider that complex auditory stimuli can function as positive rein-
forcers, neutral stimuli, or negative reinforcers of newborn behavior. Known
reinforcers include vocal-group singing, solo female singing, prose spoken by a
female, synthetic speech sounds, and intrauterine heartbeat sounds (Butter-
field & Cairns, 1974; Butterfield & Siperstein, 1972; DeCasper, Butterfield, &
Cairns, 1976; DeCasper & Carstens, 1981; DeCasper & Sigafoos, 1983). On
the other hand, male speech and instrumental music lack reinforcing value,
while white noise and faster-than-normal heartbeat sounds are aversive (Butter-
field & Siperstein, 1972; DeCasper & Prescott, 1984; Salk, 1962). The differen-
tial reinforcing effectiveness of these sounds seems to covary more with their
similarity to sounds that were present in utero than with any general acoustic
characteristic(s), which further suggests that prenatal auditory experience in-
fluences postnatal auditory perception.

Finally, prenatal auditory experience has been shown to cause postnatal
auditory preferences in a variety of infrahuman species (e.g., Gottlieb, 1981;
Vince, 1979; Vince, Armitage, Walser, & Reader, 1982).

The hypothesis implies that prenatal experience with maternal speech
sounds causes some property of the sounds to be differentially reinforcing
after birth. Speech sounds enable at least two kinds of discriminations; some
speech cues allow discrimination of language-relevant sounds, per se, or what
is said, and some allow discrimination of the speaker or source of the speech
sounds (Bricker & Pruzansky, 1976; Studdert-Kennedy, 1982). Thus, the pre-
natal experience hypothesis implies that newborns prefer their own mothers’
voices, regardless of what she says, because of prenatal experience with her
voice-specific cues. This implication, however, cannot be directly tested for
obvious ethical and practical reasons. The hypothesis also implies that new-
borns will prefer the acoustic properties of a particular speech passage if their
mothers repeatedly recite that passage while they are pregnant.

We directly tested the latter implication in the following way. First, preg-
nant women tape-recorded three separate prose passages. Then, they recited

PRENATAL AUDITORY LEARNING 135

one of the passages, their target passage, aloud each day during the last 6 weeks
of pregnancy. After birth their infants were observed in an operant learning
task where recordings of the target passage and a novel passage, one their
mothers had recorded but had not recited, were both available as reinforcers.
Then their relative reinforcing effectiveness was evaluated. If the prenatal ex-
perience with the target passage increases its reinforcing value then: (a) the
acoustic properties of the target passage will be more reinforcing than those of
a novel passage; (b) the differential reinforcing value of the target passage
should be carried by its language-relevant cues and, thus, should not require
the presence of the infant’s own mother’s voice cues; and (c) the reinforcing
values of the target and novel passages should not differ for control newborns
who had never been exposed to either passage.

METHOD

Prenatal Phase

Pregnant Subjects. Thirty-three healthy women approximately 7 Yz
months pregnant were recruited from childbirth preparation classes after being
informed about the project. All were experiencing uncomplicated pregnancies.

Prenatal Procedures. After becoming familiar with three short children’s
stories they tape-recorded all three. Recordings were made in a quiet room on an
Akai 4000 stereophonic tape rec . The tapes would be used as reinforcers
in a postnatal learning task. Each woman was then assigned one of the stories
as her target story. Assignment was made after all three had been recorded to
prevent them from biasing the recording of their target, for example, by ex-
aggerated intonation.

The women were instructed to read their target story aloud “two times
through each day when you feel that your baby (fetus) is awake” and to “read
the story in a quiet place so that your voice is the only sound that your baby
can hear.” They maintained a log of their daily recitations and were occasion-
ally checked by the researchers.

Story Materials. The stories were The King, the Mice, and the Cheese
(Gurney & Gurney, 1965), the first 28 paragraphs of The Cat in the Hat (Seuss,
1957), and a story we called The Dog in the Fog, which was the last 28 para-
graphs of The Cut in rhe Hut with salient nouns changed. The three stories
were about equally long, they contained 579, 611, and 642 words, respectively.
Each could be comfortably recited in about 3 min. Each was also composed
from equal size vocabularies of 152, 142, and 154 words, respectively. Salient,
high-frequency nouns common to at least two stories were changed. For exam-
ple, cat and hat in The Cut became dog and fog in The Dog, and cat and dog
from those stories became turtle and zebra in The King. The Cat contained 46
unique words (i.e., words that appeared only in The Cat), which accounted for
22% of the total word count; The Dog contained 57 unique words, which ac-
counted for 22% of the total word count; and The King contained 85 unique

136 DKASPER AND SPENCE

words, which accounted for 44% of the total word count. All three stories con-
tained common high-frequency words. For example, a, all, and, did, do, he, I,
in, like, not, now, of, said, that, the, to, with, and you occurred at least three
times in each. The common high-frequency words accounted for 43% of The
Cat, 38% of the The Dog, and 36% of the The King. The remaining words oc-
curred at least once in at least two of the stories. The stories also differed in
prosodic qualities, such as patterns of syllabic beats. Thus, they differed in the
acoustic properties of individual words as well as in prosody. The Cat and The
Dog sounded more similar to each other than either did to The King, but we
could readily identify the origin of short (several seconds) segments from all
three.

Postnatal Phase

Experienced Newborns. Sixteen of the 33 fetal subjects completed test-
ing as newborns. The 16 had been prenatally exposed to their target story an
average of 67 times or for about 3.5 hours in all. They were tested at an average
age of 55.8 hours (SD= 10). Each had to have had an uncomplicated full-term
gestation and delivery, a birth weight between 3500-3900 grams, and APGAR
scores of 8, 9, or 10 at 1 and 5 min after birth. If a subject was circumcised, he
was not tested until at least 12 hours afterward. Parents gave informed consent
for the testing and were invited to observe.

Seventeen infants were not tested or did not complete a test session: 5
because their mothers failed to return their logs, 4 because they encountered
intrapartum or postpartum difficulties, 5 failed to meet state criteria at the
time of testing or cried, and 3 subjects’ sessions were unavoidably interrupted.

Apparatus. Sessions occurred in a quiet, dimly lit room adjacent to the
nursery. The infants lay supine in their bassinets and wore TDH-39 earphones,
which were suspended from a flexible rod. They sucked on a regular feeding
nipple with the hole enlarged to 1 mm. Rubber tubing connected the nipple to
a Statham P23AA pressure transducer that was connected to a Grass poly-
graph and solid state programming and recording components. Each infant
heard a tape recording of his/her target story and a tape recording of a novel
story, one of the others their mother had recorded but not recited. Both stories
were recorded by the same woman, and each was played on separate channels
of the stereo rec . The tape ran continuously, and sound was electronically
gated to the earphones by the automated programming equipment. Intensities
averaged 70 dB SPL at the earphones.

Testing Procedures. Sessions began about 2.5 hours after a scheduled
feeding in to maximize the chance of obtaining an awake, alert, and
cooperative infant (Cairns & Butterfield, 1974). Each infant was brought to a
quiet-alert state before testing could begin (Wolff, 1966) and had to visually
fixate and follow an experimenter’s face when he/she spoke to the infant. (If

PRENATAL AUDITORY LEARNING 137

the infant was not alert and did not fixate or follow, he/she was returned to
the nursery, and another attempt was made after a later feeding.)

The infant was then placed supine in the bassinet and the earphones were
locked in place. One researcher, who could not be seen by the infant and who
was blind to the exact experimental condition in effect, held the nonnutritive
nipple loosely in the infant’s mouth. Another monitored the equipment. The
infant was then allowed 2 min to adjust to the situation and had to emit sucks
having negative pressures of at least 20-mm Hg, a pressure normally exceeded
by healthy infants. (If the infant failed to suck adequately, he/she was returned
to the nursery, and another attempt was made after a later feeding.)

Testing began with 5 min of baseline sucking during which no voices were
presented over the earphones. Unconstrained nonnutritive sucking occurs as
groups or bursts of individual sucks separated by interburst intervals of several
seconds. A sucking burst was defined as a series of individual sucks separated
from one another by less than 2 s; when 2 s elapsed without a suck the equip-
ment registered the end of the burst. Thus, interburst intervals (IBIS) began 2 s
after the last suck of one burst and ended with the onset of the the first suck of
the next burst. This criterion accurately captures the burst-pause pattern of
newborns’ nonnutritive sucking (see Figure 1). IBIS tend to be unimodally dis-
tributed for individual infants, and modal values vary between infants. The
baseline was used to estimate the distribution and median value of each infant’s
IBIS just before reinforcement began. Differential reinforcement of IBIS began
after baseline had been established. (Hereafter, if the infant stopped sucking

Figure 1. Polygraph record of a newborn’s nonnutritive sucking. Wide horizontal
marks indicate the onset and offset of o sucking burst. The time between the end
of one burst and the beginning of the next denotes on interburst interval. Onset
of the narrow event mork denotes that the time criterion, t seconds, has elapsed
since the end of the last burst. Vertical lines indicate time in seconds.

138 DECASPER AND SPENCE

for two 1-min periods for any reason, he/she was returned to the nursery and
not tested again.)

Reinforcement Contingencies. For eight randomly selected infants, suck-
ing bursts that terminated IBIS equal to or greater than the infants’ baseline
medians (t) produced the recording of a woman’s voice reciting the target story.
Bursts terminating IBIS less than the baseline median were reinforced with the
same woman’s recording of a novel story. Thus, only one of the two stories
was presented binaurally with the first suck of a burst and remained on until
the burst ended. Reinforcement contingencies were completely controlled by
the solid-state equipment. Reinforcement contingencies were reversed for the
other eight newborns, to control for the effects of any response bias that might
arise from either of the contingencies or from changes in the behavioral dis-
positions of the infants, for example, arousal or fatigue. Differential reinforce-
ment lasted about 20 min.

The same differential reinforcement procedures were used in earlier voice-
preference studies (DeCasper & Fifer, 1980; DeCasper & Prescott, 1984). The
rationale is based on well established reinforcement procedures that differenti-
ate the temporal properties of behavior: Differentially reinforcing a range of
IBIS causes the shorter differentially reinforced IBIS to increase in frequency
(see newborn studies by DeCasper & Fifer, 1980; DeCasper & Sigafoos, 1983;
as well as animal studies by Anger, 1956; Catania, 1970; DeCasper & Zeiler,
1977; Malott & Cumming, 1964).

Subject Controls. Twelve control newborns matched to a prenatally ex-
perienced counterpart on sex, race, and median interburst interval of baseline
were also tested. They met the same selection criteria and were tested under ex-
actly the same conditions as their counterparts, but their mothers had never
recited any of the three stories.

Other Experimental Controls. The influence of mother-specific voice
cues on the reinforcing effects of the target stories was controlled by having
nine newborns reinforced with recordings made by their own mother and seven
with recordings made by some other infant’s mother. Both stories heard by an
infant were recorded by the same woman to insure that their reinforcing value
could not be unequally influenced by the speaker’s voice characteristics. The
acoustic properties of any one story could not systematically influence the re-
inforcing value of the target because each of the three stories had served as the
target at least four times. No particular combination of target/novel pairings
could systematically influence the reinforcing vaiue of the target because five
of the six possible target/novel pairings occurred at least twice. Unpredictable
subject loss prevented precise counterbalancing of voices and target/novel
pairings (see Table 1).

TA
BL

E
1

Co
nd

iti
on

al

Pr
ob

ab
ilit

y
of

Re

sp
on

di
ng

wi

th

(O
.O

t
5

IB
I

<0 .4 t) an d (1 . Of 5 19 1 < 1. 4t ) Du rin g Ba se lin e an d Du rin g Re in fo rc em en t fo r Ex pe rie nc ed an d Co nt ro l Su bj ec ts Ex pe rie nc ed Su bj ec ts Co nt ro l Su bj ec ts sto ry Ta rg et N ov el Ta rg et N ov el C rit er ia c. P. c. P. c. P. C .P . c. P. Se x c. P. c. P. c. P. To rg ef /N ov el fo r To rg ef vo ic e Bo se . R ei nf . Bo se . R ei nf . Bo se . R el nf . Ba se . R ei nf . F C ot /D og <5 M .ll .0 6 .I8 .2 1 F D og /C ot <6 M .0 6 .2 6 .2 6 .3 2 F D og /K in g <3 M .1 2 .lO .lB .1 3 .1 6 .lO 36 M .1 6 Ki ng /D og <6 M .lO .1 6 .3 3 .2 8 .0 6 .0 5 .3 2 F .I1 C at /K in g <3 0 .I4 .1 6 .3 3 .0 6 .0 4 .0 7 .1 9 F .1 3 D og /K in g <4 0 .0 2 .1 5 .lO .0 7 .0 0 .lO .2 0 .2 3 M D og /K in g <4 0 .0 9 .1 2 .1 9 .2 1 .lO .1 7 .2 1 .1 6 M Ki ng /D og <6 0 .0 3 .0 9 .1 9 .1 1 .0 6 .0 4 .2 4 .1 1 M C ot /D og 23 M .2 9 .3 7 .1 2 .I3 M D og /C ot r5 M .2 3 .2 1 .0 6 .0 5 F D og /K in g r3 M .1 6 .1 7 .1 3 .ll .0 0 .2 2 .1 2 .I3 M Ki ng /D og 23 M .2 6 .3 2 .I7 .I8 .6 3 .3 5 .0 9 M .1 6 C at /K in g 25 M .I7 .3 1 .0 9 .lO F .4 9 .1 6 .1 2 .0 7 Ki ng /D og 23 0 .0 6 .0 0 .1 7 .0 8 .2 6 .0 5 .lO .0 5 M D og /K in g r3 0 .0 3 .2 0 .2 4 .2 9 .1 3 .I2 .1 4 F .I9 C at /K in g r6 0 .2 3 .2 8 .0 6 .1 8 .3 1 .1 6 .1 4 .0 3 140 DECASPER AND SPENCE Data Analysis. Interburst intervals were read off the polygraph records. Times between the event marks signalling the end of one burst and the begin- ning of the next burst were measured and rounded down to the nearest whole second (see Figure 1). Thus, the scorers (AJD and MJS), who were highly prac- ticed, did not have to make detailed judgments about IBI values that might bias the data. Interscorer reliability approached 100%. Next, each subject’s IBIS from the baseline and reinforcement phases were converted to a proportion of their time criterion (t). For example, if t =4 s then all 2-s IBIS had the value O.Sf, and if t = 6 s then 2-s IBIS had the value of 0.331. Converted IBIS were grouped into bins that were 0.21 s wide; Bin 1 con- tained IBIS between O.Ot and 0.2t s, Bin 2 contained IBIS between 0.2t and 0.4f S . . , and Bin 10 contained IBIS between 1.81 and 2.0t s. Bin 11 contained all &Is greater than 2.0f s. IBIS were assumed to be equally distributed within a bin. The conversion equates the relative size of IBIS across subjects and allows averaging over subjects. RESULTS Experienced Newborns The hypothesis asserts that in utero exposure to the acoustic properties of the target story will make it more reinforcing than the novel story. If so, the rela- tive frequency of short IBIS should increase over baseline when reinforced by the target stories in the IBI t condition.

Baseline IBI distributions were examined first in to determine
whether they differed between reinforcement contingencies. They did not dif-
fer: A mixed ANOVA of the relative frequencies of baseline IBIS, with Contin-
gencies (c t vs. > t) and Bin (l-10) as factors, indicated a significant effect of
Bin, F(9, 126) = 13.3, p< X101. The effect merely confirms that the IBIS were unimodally distributed. Most important, there was no Contingency effect, F(1,14) p c 1 .O, and no Contingency x Bin interaction, F(9, 126) = 1.67, p > .lO.

The predictions of the hypothesis were first assessed by examining the
differences between the relative frequencies of IBIS that occurred during base-
line and those that occurred during reinforcement. Difference scores were
entered into a mixed ANOVA with Contingencies (< t vs. > t) and Bin (l-10)
as factors. There was no effect of Contingency, F( 1,14) < 1 .O, and a signifi- cant effect of Bin, F(9,126) = 5.48, p-c .025. Most important, there was a sig- nificant Contingency x Bin interaction, F(1,126) = 2.07, pc .05. Planned tests of simple effects confirmed that the interaction occurred because with the IBI < t contingency the relative frequency of short IBIS increased over baseline levels, while those of all other IBIS either decreased or did not change. With the IBI>t contingency the relative frequency of IBIS slightly greater than f
seconds increased, while those of the others decreased or did not change. Any
IBI between 0 and t seconds would have produced the target story under the

PRENATAL AUDITORY LEARNING 141

IBI < t contingency, and any IBI L t seconds would have produced it under the IBI > t contingency. But only the relative frequencies of the shorter IBIS rein-
forced by the targets sytematically increased.

The differential reinforcement effects are more clearly revealed in the
analysis of IBIS between O.Ot and 0.4t (the shorter IBIS) and those between 1 .Ot
and 1.4t (IBIS slightly longer than t seconds). Conditional probabilities of
baseline and reinforced IBIS in these classes were obtained by dividing the rela-
tive frequency of IBIS in each class by the relative frequency of that class and
all longer IBIS (see Table 1). This is a sensitive measure of temporally differ-
entiated responding because: (a) it adjusts the inherently unequal opportunity
for infants to emit equal numbers of short and long IBIS in a limited period of
time; (b) it measures the probability that an infant will emit a particular class
of IBIS given the opportunity to do so (cf. Anger, 1956; DeCasper & Fifer,
1980); and (c) it renders the conditional probabilities of IBIS between O.Ot and
0.4[, and those between 1.01 and 1.41, arithmetically independent of one
another. The dependent variables for the target story and for the novel story
were their reinforcement ratios: (conditional probabilities of IBIS during rein-
forcement) divided by (conditional probability of IBIS during reinforcement)
plus (conditional probability of IBIS during baseline).

The average values of baseline conditional probabilities of target-story
IBIS and novel-story IBIS did not differ, t(15) = 1.37, p> .lO. However, their
reinforcement ratios differed as expected. A mixed ANOVA with Contingency
( t) and Interval (O.Ot-0.4t vs. 1 .Ot-1.4t) as factors, revealed no effect
of Contingency, F( 1,14) < 1 .O, and no effect of Interval, F( 1,14) = 1.59, p > .20,
but a significant Contingency x Interval interaction, F(l, 14) = 6.65, p< .025 (Figure 2). Target-story reinforcement ratios were larger than novel-story rein- forcement ratios, independent of the contingency and of the interval. The fact that 13 of the 16 infants had larger target ratios than novel ratios (p = .Ol 1 by the binomial test) and 13 of the 16 had target-story ratios greater than .50 indi- cates this result was typical. The individual-subject consistency implies that maternal voice cues were not necessary for producing the differential rein- forcement effect. Neither the target-story reinforcement ratios nor the differ- ence between the target ratios and novel ratios differed between the 9 infants who heard their own mothers’ voices and the 7 who heard unfamiliar voices, p-values of both f-tests > .lO.

Control Subjects

The following analysis of control-subject performances parallels that of the
experienced subjects. The relative frequency distributions of baseline IBIS did
not differ between reinforcement contingency conditions. A mixed ANOVA
with Contingency (c t vs. > t) and Bin (l-10) as factors revealed a marginal
effect of Bin, F(9,90) = 1.90, .lO

.05, but no effect of Contingency,
F( 1,lO) < 1 .O, or of the Contingency x Bin interaction, F(9,90) = 1.38, p> .lO.
The subsequent mixed ANOVA on the difference scores of IBIS that occurred

142 DKASPER AND SPENCE

.60

IBI BIN Ct – sec1
Figure 2. Mean reinforcement ratios of the target (hatched bars) and novel (open
bars) stories for Experienced infants in the 161 t condition (right side). The means are based on a total of 400 baseline and
1040 reinforced interburst intervals.

during the baseline and reinforcement phases revealed no effect of Contingency,
F(1) 10) < 1 .O, a significant effect of Bin, F(9,90) = 5.19, p< .OOl, and a signifi- cant Contingency x Bin interaction, F(9,90) = 3.48, p c .005. However, none of the follow-up tests of simple effects were statistically reliable; the inter- action seemed to result from unsystematic variation in the difference scores of the two contingency conditions in Bins 1-5. Subsequent analysis of conditional probabilities confirmed that the pre- ceding interaction did not result from systematic effects of target-story rein- forcement. The baseline conditional probabilities of target and novel stories did not differ, t( 11) < 1 .O; neither did their reinforcement ratios computed for the intervals O.Ot-0.4t and 1 .Ot-1.41. The mixed ANOVA with Contingency and Interval as factors revealed no reliable effects whatever, p values of all F statistics > .10 (Figure 3).

A comparison of the reinforcement ratios of matched-subject pairs re-
vealed that experienced newborns had larger target-story ratios than their
matched naive counterparts, t(l1) =2.68, p< .05, but that their novel-story ratios did not differ, t(1 1)~ 1 .O. DISCUSSION Three implications of the prenatal-experience hypothesis were confirmed: (1) For experienced subjects the target story was more reinforcing than the novel PRENATAL AUDITORY LEARNING 143 IX (0t - .4tmt-1.4t) (0t - .4tM-7.4t) IBI BIN Ct-set) Figure 3. Mean reinforcement rotios of the target (hatched bars) and novel (open bars) stories for Control infants in the I81 t
condition (right side). The means are based on 300 baseline and 800 reinforced in-
terburst intervals.

story when both were concurrently available (2) the greater reinforcing value of
the target story was independent of who recited the story; and (3) for matched-
control infants the target story was no more reinforcing than the novel story.
The only experimental variable that can systematically account for these find-
ings is whether the infants’ mothers had recited the target story while pregnant.
Subject characteristics also seem unable to account for the results; the differen-
tial-reinforcement effect did not occur within the matched-control group, and
the differential-reinforcing value of the target story differed between matched
subjects, but the reinforcing effect of the novel story did not. The results also
cannot be attributed to individual-subject and subgroup differences in baseline
patterns of responding. The most reasonable conclusion is that the target stories
were the more effective reinforcers, that is, were preferred, because the infants
had heard them before birth. The conclusion is consistent with earlier, inde-
pendent evidence that hearing becomes functional during the third trimester
and that maternal speech attains audible in utero …