FRANCK RAMUS
10. J. Mehler, et al., Cognition 29, 143 (1988).
11. T. Nazzi, J. Bertoncini, J. Mehler, Journal of Experimental
Psychology: Human Perception and Performance 24, 756 (1998).
12. H. J. Neville, D. L. Mills, Mental retardation and develop-
mental disabilities research reviews 3, 282 (1997).
13. D. Perani, et al., NeuroReport 7, 2439 (1996).
14. G. Ehret, B. Haack, Naturwissenshaften, 68, 208 (1981).
15. R. A. Wyttenbach, M. L. May, R. R. Hoy, Science 273, 1542
(1996).
16. P. K. Kuhl, J. D. Miller, Science 190, 69 (1975).
17. P. Kuhl, in Categorical perception: The groundwork of cogni-
tion S. Harnad, Ed. (Cambridge University Press, 1987) pp. 355-
386.
18. M. D. Hauser, The evolution of communication (Bradford
Books/MIT Press, Cambridge, MA, 1996).
19. T. Dutoit, V. Pagel, N. Pierret, F. Bataille, O. van der
Vrecken, The MBROLA Project: Towards a set of high-quality
speech synthesizers free of use for non-commercial purposes, IC-
SLP’96, Philadelphia (1996). This software is freely available from
http://tcts.fpms.ac.be/synthesis/mbrola.html.
20. The production of this kind of stimuli is described in great
detail in F. Ramus, J. Mehler, Journal of the Acoustical Society of
America 105, 512 (1999).
21. The experiment takes place in a sound-attenuated booth, with
only the baby and the experimenter inside. The experimenter is
blind to the experimental condition and listens to masking noise
during the test. Newborns are randomly assigned to the control or
to the experimental group. Order of presentation of the languages
is counterbalanced across subjects. During a given phase, sentences
corresponding to the condition are played in a random order. The
habituation phase lasts at least 5 minutes. The habituation criterion
is a 25
22. Subjects are full-term healthy newborns, aged between 2 and
5 days, recruited at the Port-Royal maternity hospital in Paris.
Forty-two additional babies were tested, and their results were dis-
carded due to: rejection of the pacifier (1), sleeping or insufficient
sucking before the switch (12), crying or agitated (9), failure to meet
the habituation criterion (9), sleeping or insufficient sucking after
the switch (6), loss of the pacifier after the switch (4), computer
failure (1).
23. The slight tendency for babies in the control group to suck
more overall than those in the experimental group, though visible
on Figure 1A, is not significant [F(1,31)=2.6, p=0.12].
24. In the English-Japanese discrimination experiments by Nazzi
et al. (11), the variability due to the 4 voices was much reduced by
low-pass filtering the stimuli. In other experiments (10), a single
bilingual speaker was used.
25. P. W. Jusczyk, D. B. Pisoni, J. Mullenix, Cognition 43, 253
(1992).
26. While the resynthesis process reduces all voices to one, other
(prosodic) characteristics of the different speakers are preserved.
We didn’t test discrimination of the natural sentences played back-
wards, due to newborns’ failure to discriminate these sentences even
when played forward.
27. Twenty additional babies were tested, and their results were
discarded due to: sleeping or insufficient sucking before the switch
(6), crying or agitated (4), failure to meet the habituation criterion
(1), sleeping or insufficient sucking after the switch (3), loss of the
pacifier after the switch (6).
28. Seven additional babies were tested, and their results were
discarded due to: crying or agitated (1), sleeping or insufficient
sucking after the switch (4), loss of the pacifier after the switch (2).
29. The tendency for babies in the control group to suck more
than those in the experimental group during the habituation phase
is not significant [F(1,31)=2.7, p=0.11].
30. Before running these experiments, all tamarins had partic-
ipated in a habituation-dishabituation experiment involving their
own, species-typical vocalizations. Thus, all subjects were famil-
iar with the general test set up. Subjects sit in the test cage within
the acoustic chamber, and do so without stress. They also sit or
hang on the front panel during testing, thereby allowing relatively
unambiguous observations of head turning behavior. Experiments
were run by transporting a tamarin from their home room to a test
room, acoustically and visually isolated from all other tamarins.
Two observers watched the session from a monitor outside of the
test room. Stimuli were played back from a concealed speaker only
when the subject’s head and body were oriented approximately 180
degrees away from the speaker . A positive response was scored if
the subject turned and oriented toward the speaker within the play-
back period. If the response was ambiguous, we ran the trial again,
but with a different exemplar from the habituation series. A subject
was considered habituated if he or she failed to respond on two con-
secutive trials. Following habituation, the test stimulus was played.
The final trial of the session was a post-test playback, presenting the
long call of a tamarin. Given the salience of the long call, we ex-
pected the tamarins to respond. If they failed to do so, we excluded
the entire session, under the assumption that failure to respond to
all stimuli represents general habituation to the test set-up. Inter-
trial interval within a session was set at a minimum of 15 sec and
a maximum of 60 sec. All trials were videotaped. After running
a session, trials were digitized onto a computer and subsequently
scored blind with respect to test condition by stepping through the
experiment frame-by-frame. Two observers scored each test trial;
inter-observer reliability was 0.92.
31. Each tamarin was tested in the four conditions, with the order
of presentation of languages counterbalanced across subjects. Inter-
session interval was no less than 4 days, with a median of 7 days.
When a session’s data were excluded due to an ambiguous response
to the test sentence, or failure to respond to the post-test trial, the
tamarin was tested again in the same condition after 1-3 weeks, in
order to complete all the conditions. The significance of the pro-
portion of monkeys reacting to a given change is assessed through a
binomial test, and the difference between two conditions is assessed
through a 2X2 chi-square test.
32. The fact that newborns fail under this very condition in Ex-
periment 1A is likely due to their immature auditory system, since
susceptibility to speaker variability seems to resolve a few months
after birth (25).
33. Work supported by the Delegation Generale pour l’Armement
and a National Science Foundation Young Investigator Award to
MDH. We thank the Port-Royal maternity hospital for providing
access to the newborns, and the parents for their participation. All
parents gave informed consent. We thank the New England Re-
gional Primate Research Center (P51RR00168-37) for providing
the tamarins; our research was approved by the Animal Care and
Use Committee at Harvard University (92-16) and the CCPPRB
Paris-Cochin.