Categorization of musical structures by 6-to-10 month old infants
Marc Mélen (1, 2, 3, 4) & Julie Wachsmann (1, 2)
(1) University of Liège
(2) Unit of Research in Psychology of Music
(3) Senior Researcher at the "Fonds National de la Recherche Scientifique
(4) "Centre de Recherches Musicales de Wallonie"
Introduction
Categorization is a fundamental process by which the subject decreases the
complexity and the diversity of the social or physical environment by organising it. A
category can be defined as a class made up of a number of different items that are
considered equivalent, yet discrete. Thus categorization consists in bringing together
objects or events into classes, and responding according to their status of member of a
class rather than to their peculiarities. Categorization is central to the development of
flexible, parcimonious information processing; it is involved in the identification and
recognition of objects and in the assimilation and organization of new knowledges.
The fonction of category systems is to provide maximum information with the least
cognitive effort. If all the perceptual representations formed for the multitude of objects
and their relations encountered during a lifetime were independent of each other, the
outcome would be mental disorganisation and intellectual chaos. Cognition and its
development would be difficult, if not impossible, under such circumstances. The
information-processing advantages of perceptual categorization include organised
storage of information in memory, efficient retrieval of this information, and the
capability of responding equivalently to an indefinitely large number of exemplars from
multiple categories.
Given the adaptative importance of categorization, there has been a remarkable effort
to study categorization skills in infants and toddlers during the last two decades (for a
recent review, see Quinn, 1998). Most of the studies were inspired by Rosch's theory
(see Rosch and Lloyd, 1978) and its extension to children's development by Rosch and
Mervis (for a synthesis, see Mervis, 1987). A classic distinction in this context is made
between subordinate-level (chihuahua, setter, etc.; or, velvet throusers, jeans, etc.),
basic-level (dogs; throusers) and superordinate-level (mammals; clothing) categories.
The subordinate-level represents the less abstract level of categorization and brings the
largest quantity of information about the exemplars it contains. The superordinate-level,
the most abstract level, is less informative but more distinctive, in that it gives few
information about its examplars but allows easy distinction between categories. The
basic-level represents a compromise between these extremes: it gives enough
information to distinguish efficiently members of the category, i.e. it is informative,
while preserving distinctiveness, i.e. the distinction between the different classes of
objects.
Following the seminal paper by Rosch, Mervis, Gray, and Boyes-Braem (1976), most
of the studies made with very young children (less than two years of age) aimed to
describe their ability to form categories at the basic- and the superordinate-levels. These
studies have mainly focused on categorization of geometric forms (see, for example,
Quinn, 1987; Lécuyer & Poirier, 1994); natural objects, both physical (see Behl-Chadha,
1996 for categorisation of furniture by 3-4-month-old infants) and living objects (see,
e.g., de Schonen and Deruelle, 1991 for human faces; Mandler, Bauer, McDonough,
1993 for animals). By contrast, categorization of auditory events has been barely
explored, with an exception for speech stimuli (see, notably, studies by Kuhl and coworkers, e.g. Kuhl, 1991).
The present paper is concerned with categorization of musical motifs, a class of
auditory events whose categorization is less studied (but see Trehub & Thorpe, 1989 for
categorization of rhythm and Clarkson & Clifton, 1985; Trehub, Endman, & Thorpe,
1990, for categorization of timbre). Moreover for lack of reference to a general model of
music perception, the few studies concerned with musical categorization do demonstrate
categorization of music stimuli, but give few information about the functional role of
this capability.
Deliège suggested a model of music perception in which categorization plays a
central function (see Deliège, 1987 for the seminal paper, and Deliège & Mélen, 1997,
for a recent synthesis). Indeed, as explained elsewhere (Deliège, this symposium), the
model of sameness and difference sees music perception as a vast process of
categorization which begins with rhythmic grouping and grouping of groups thanks to
the extraction of cues, proceeds with perception of similarities between motifs located at
different places in a piece based on the recognition of the cues they share, and
culminates through the organisation of the traces left by the cues, the imprints, in
categories motifd around prototypes (see Deliège, this symposium, for further details).
Deliège makes use of Rosch's theory to understand categorization in music, and
suggests that the cues constitute the basic-level of categorization, i.e. each cue is the
heading of a category located at the basic level. The motifs which share a same cue
while varying it represent the subordinate-level. Finally, at the superordinate-level the
different cues share a common reference: the style of the work.
Melen (1999 a & b), using an operant conditionning of head-rotations, demonstrated
the relevance of this model for the development of music perception in infancy. He
showed that 6-to-10-month-old infants organize rhythmic grouping according to the
principles of sameness and difference. Therefore, these studies were concerned with
simple demonstration of categorization. This paper represents a first step toward study
of more complex types of categorization in infancy. It aims to shed light on the
processes through which babies collect different motifs which share a common cue, i.e.
are variations of a same cue, to form a category distinct from another one, both being
situated at the basic-level.
Experiment 1
Methods
Participants
The participants were 24 healthy, full-term infants ranging from 8 to 10 months of
age. Four infants were excluded from the sample, because they failed to meet a
predetermined training criterion (see below). The final sample comprised 9 males and
11 females, with a mean age of 8 months, 17 days.
Apparatus
The experiment was controlled by a computer (Macintosh SE 30 4/20), which
monitored, through a program designed with the Max software, the electronic
equipement via a MIDI interface (Studio Opcode 2). The stimuli, composed with the
software Performer, were generated on line by a synthetizer (Yamaha TG-100) and
presented via a stereo amplifier (Sony TA-F117R) and one loudspeaker (Yamaha Studio
NS10-M). A French cartoon for small children (Bouli) served as reinforcement and was
displayed on a video monitor (Commodore 1702) thanks to a Sony SLV-325 video
recorder which was controlled by a home-made interface. A 15 watts green lamp was
also controlled by that interface. The assistant and the experimenter wore headphones
(AudioTechnica ATH610) carrying music to mask the nature of the stimuli presented to
the infants.
Musical material
The Ländler n° 10 D145, op. 11 by F. Schubert served as stimulus. This piece is 16
bars long and can be divided in two periods of 8 bars, comprising each two 4-bars
phrases. The first two-bars expose 1 motif (1st bar) and a variation of this first motif
(2nd bar), whereas the second motif is two-bars long. However, a symetrical grouping
can be considered as more parcimonious and preferable for methodological reasons.
Thus, in the present experiment, the piece has been considered as made up of two 2-bars
motifs. In other words, the piece involves two motifs (a and b), whose models (parent
motifs a and b) are exposed at the beginning, the rest of the piece being composed of 3
variations of these parent motifs. Thus, the piece corresponds to the following schema :
Parent A - Parent B - A1 - B1 - A2 - B2 - A3 - B3. Consequently, the piece can be
described as involving two categories, each comprising 4 sequences (1 parent and 3
variations).
Procedure
During the session the infant was seated on the assistant's lap. The experimenter was
seated in another room and monitored the experimental session through a one-way
mirror. The infant was presented repeatedly with a standard sequence (S), the parent
motif A, separated by 500-msec silence intervals. When the infant was quiet and facing
directly ahead to the flashing green lamp, the experimenter initiated a training trial, at
which time a contrasting sequence was presented once at an intensity 6 dB higher than
the previously heard standard sequence. During the training phase, the contrasting
sequence consisted in the parent motif B. If the infant turned his/her head at least 45 d°
toward the loudspeaker during the presentation of the training sequence (T) or the 500ms silent interval that followed (total response time : 5 seconds), the experimenter
recorded the turn and a cartoon was displayed (6 sec) after the end of the contrasting
sequence. The video monitor was under the louspeaker, located 1 meter from the infant
at an angle of 90 d°. Head-turns made at other times were not reinforced, nor were headturns of less than 45 d°. If the infant responded correctly on two consecutive trials, the
intensity of the contrast stimulus was made equivalent to that of the standard sequence.
Subsequently, the infant was required to respond correctly on four consecutive trials in
order to meet the training criterion. Infants who failed to turn to the change on the first
two trials were presented with the contrasting sequence 12 dB higher than the standard
on subsequent trials until they responded correctly on two consecutive trials, at which
time the intensity of the contrast stimulus was lowered 6 dB, and so on. The session was
abandoned if the training criterion was not met within 20 trials. During the testing
phase, the standard sequence remained the same as in the training phase. The infant was
presented with change and no-change trials, the latter of which provided a measure of
spontaneous rotation in the direction of the loudspeaker during the test session. The test
session comprised 36 trials, 12 no-change trials [Standard vs Standard, (S)] and 24
change trials [S vs Comparison, (C)], in random order. Among the change trials, 3
consisted in variations of parent A and were considered as not-to-be-responded-to
change trials (C-, i.e. C- 1, C- 2, C- 3) and 3 in variations of parent B which were
considered as to-be-responded-to change trials (C+, i.e. C+ 1, C+ 2, C+ 3), each
repeated 4 times. The infant was expected to turn the head more frequently for the
change trials that involved variations of parent motif B. The infants were divided in two
groups in order to counterbalance the categories (in the first condition: parent A, A1,
A2, A3 were, respectively, S, C- 1, C- 2, C- 3, and parent B, B1, B2, B3 were, T, C+ 1,
C+ 2, C+ 3; in the second condition: parent B, B1, B2, B3 were, respectively, S, C- 1,
C- 2, C- 3, and parent A, A1, A2, A3 were respectively, T, C+ 1, C+ 2, C+ 3).
Results and discussion
Raw data consisted in the number of head rotations to change and no-change trials.
Globally, the infants responded to 35.83 % of no-change trials, to 48.75 % of change
trials involving variations of parent A (C- trials) and to 57.92 % of change trials
involving variations of B (C+ trials).
As suggested by Thorpe, Trehub, Morrongiello, & Bull (1988), the head-rotation
procedure can be assimilated to a go-no go task, since for each stimulation, the
participant has one alternative: to turn or not the head. They further suggest that the data
fit with the framework of the Signal Detection Theory (SDT). Consequently, responses
to change trials were considered as hits, whereas responses to no-change trials were
considered as false alarms. The proportions of hits and false alarms were converted into
d', i.e. the index of the SDT, using a normal distribution table following the
conventional formula.
A d' was calculated for each type of change trials. Each actual d' was compared to the
expected value in case of no-discrimination, i.e. zero. A one-sample Student t test (1tailed) was run out and lead to the following results : for the C- trials, d' = .73,
t(19) = 3.349, p < .002; for the C+ trials, d' = 1.186, t(19) = 3.361, p < .002. Thus both
types of change trials have been discriminated by the participants.
The d' observed for both types of change trials were also compared to each other.
Both analyses were run with a paired Student t test (2-tailed): t(19) = -1.48, p > .14.
Thus, both types of change trials were equally discriminated.
The experiment involved two conditions in order to counterbalance the role of the
sequences. In the first condition the standard was parent A, the training sequence parent
B, the C- trials were variations of parent-A, the C+ trials were the variations of parentB. In the second condition, the role of the parent motifs were reversed, and consequently
the sequences involved in C- trials and C+ trials. A 2 (conditions) X 6 (items) betweenwithin ANOVA taking item as within factor was run out on the number of head turns.
Both main effects were non significant (respectively, F(1,18) = 3.40, p = .08; F(5,18) =
1.84, p > .10). Interestingly, the interaction between condition and items was significant
: F(5,90) = 3.41, p < .007. The Newman-Keuls realised as post-hoc showed that all the
variations elicited similar rates of responses in the first condition, whereas a3 elicited
significantly more head-turns than the other to-be-discriminated sequences in the second
condition. And, this percentage of responses was large enough to make the effect of
condition nearly significant (see figure 1A). Moreover, a paired Student t test (2-tailed)
between the number of responses to the parent A and to each variation in the first
condition, indicated a significant difference between parent A and A3 (t(9) = -1.466,
p < .002) but not between parent A and A1 or A2 (respectively, t(9) = -.166, .p > .60;
t(9) = -.366, p > .30).
The pattern of results presented sofar is far from ideal. One would have expected a
non-significant d' for the not-to-be-responded-to trials in the first analysis and a
significant result in the second analysis showing a better discrimination for the to-beresponded-to trials. Nevertheless, the fact that the different variations A and B were
discriminated from their parents, indicates that the infants might distinguish the
members of both categories. In fact, the second condition seems to be closer to the ideal
profile, since according to a paired t test (2-tailed, t(9) = -1.987), the d' in this condition
tended (p = .0782) to be larger for the to-be-responded-to sequences (d' = 1.937) than
for the not-to-be-responded-to sequences (d' = .999).
Consequently, the first experiment does not allow us to conclude to the existence of a
categorization process, since the infants responded virtually equally to both types of
change trials. Moreover, the sequence A3 seems to have a special status, since it tended
to elicit more responses in the first condition and did provoke more responses in the
second one. This can be referred to the high register of this passage, that might have
attracted infants' attention and made this sequence more distinctive than the other.
Mélen (1997, 1999b) showed that register change were powerful cues for 6-to-10-month
old infants, and these results give further support to this suggestion. One could also say
that the categories were not easy to discriminate, since the features which distinguish
them did not lead to higher proportions of responses for the to-be-responded-to
sequences. Even, if the d' for the to-be-responded-to sequences would have been
significant in the second condition, it could have been attributed to one particular
sequence rather than to categorical features as such.
Two questions arise at the end of this experiment : 1) is A3 really dissimilar to the
other sequences ? 2) Would categorization be manifest if the difference between
categories was increased ? Experiments 2 and 3, respectively, were run to address these
questions.
Experiment 2
Methods
Participants
Twenty female and 20 male young adults (mean age = 23 years, range = 15-29 years)
took part to the experiment. They never received any music training.
Apparatus
The experiment was controlled by a computer (Macintosh Classic 4/20), which
monitored, through a program designed with the Max software, the electronic
equipment. The stimuli, composed with the software Performer, were generated on line
by a synthetizer (Korg 05R/W) involving a built-in MIDI interface and presented via two
self-amplified loudspeakers (Altec Lansing ACS31).
Musical material
The Ländler n° 10 D145, op. 11 by F. Schubert served as stimulus.
Procedure
The participants were tested individually and presented with 45 pairs of sequences,
representing all the possible arrangements of sequences Parent A, A1, A2, or A3, on the
one hand, and all the possible arrangements of sequences Parent B, B1, B2, or B3, on
the other hand. Three control pairs were added, i.e. A vs A, B vs B, A vs B. Thus, there
was a total of 15 pairs, each presented three times. The task was to evaluate the degree
of similarity between the elements of a pair. They used a Likert-type scale ranging from
1 (totally dissimilar) to 7 (identical) made up of seven strokes of the keyboard. They
responded by pressing the corresponding stroke. No time pressure was imposed. A pair
of segments replaced the other, as soon as the participant had given his/her response.
The participants were allowed to respond only after the second element of the pair. The
experiment, followed by a debriefing, lasted about 10 minutes.
Results and discussion
The three control pairs showed that the participants performed consistantly, since the
mean rating was 6.8 (SD = .34) for the pair Parent A - Parent A and 6.56 (SD = .51) for
Parent B - Parent B, whereas it was 1.67 (SD = .80) for the pair Parent A - Parent B.
These ratings are very similar to those obtained by non-musician subjects for similar
task in other experiments (Deliège, 1996).
Two one-way Anovas considering the pairs, excepted the control pairs, as withinfactor was run out on the mean similarity ratings and led to a very significant effect (for
the A motifs: F(5,195) = 16.66, p < .0001; for the B motifs: F(5,195) = 27.51,
p < .0001). A Newman-Keuls post hoc analysis revealed that the participants considered
all the pairs containing the sequence A3 as being dissimilar from the other pairs (p <
.05). The sequences B and B1 were judged quite similar since the participants gave an
average rating of 6.1 to this pair, and this rating was higher than any other rating
(Newman-Keuls significant at p < .05). The similarity ratings for the other pairs were
ranging from 4.2 to 4.66 (see figure 1B).
This experiment confirms that sequence A3 appeared dissimilar to other sequences
derived from the parent A. The results of the present experiment must also be compared
to those of the experiment by Mélen, Praedazzer and Deliège (this symposium),
showing that 7-9 year-old children did not consider this part of the Ländler as being a
member of the category derived from the parent A. The high register of this sequence
could make it so particular that young listeners could have been misled and isolated it
from its category. The infants, in the first experiment, could have experienced the same
misleading impression.
Experiment 3
Methods
Participants
The participants were 11 healthy, full-term infants ranging from 8 to 10 months of
age. One infant was excluded from the sample, because she failed to meet a
predetermined training criterion. The final sample comprised 4 males and 6 females,
with a mean age of 7 months, 6 days.
Apparatus
See experiment 1.
Musical material
The parent motif A from the Ländler n° 10 D145, op. 11 by F. Schubert in the first
experiment and the Finale Rondo of Diabelli's Sonatine n°2 served as stimuli. The latter
is described in Koniari, Mélen and Deliège (this symposium). It was chosen because it
presents the same structure as the piece by Schubert. The parent motif B and its
derivatives (B1, B2, B3) were extracted from the piece and opposed to the parent motif
A and its derivatives from the Schubert's piece. We thought this could enhance the
distinctivity of the categories, since the sequences of the Rondo are clearly different
from the sequences of the Ländler except in terms of length.
Procedure
The same procedure as in experiment 1 was followed in the third experiment.
Results and discussion
As in experiment 1, raw data consisted in the number of head-rotations to change and
no-change trials. Globally, the infants responded to 21.62 % of no-change trials, to
25.83 % of change trials involving variations of parent A (hereafter, C- trials) and to
61.67 % of change trials involving variations of B (hereafter, C+ trials).
A d' was calculated for each type of change trials. Each actual d' was compared to the
expected value in case of no-discrimination, i.e. zero. A one-sample Student t test (1tailed) was run out and led to the following results : for the C- trials, d' = -.021,
t(9) = -.202, p >.40 ; for the C+ trials, d' = 1.526, t(9) = 3.097, p < .007. Thus only
change trials involving variations of B have been discriminated by the participants. The
d' observed for both types of change trials were also compared to each other using the
Wilcoxon test, Z = 2.7, p < .007. Thus, infants responded differently to both types of
change trials.
Therefore, it can be concluded that infants' responses to C- trials were not different
from what was expected by chance even when it involved the sequences A1, A2 or A3,
and that they responded preferably to the C+ trials.
Nevertheless, the sequence A3 still had a special status as demonstrated by a
Kruskall-Wallis Anova comparing the d' associated to the C- and the C+ trials
according to the condition, i.e. according to the type of sequences that composed these
change trials (variations A or B). The d' for the C- trials did not differ from one
condition to the other (Chi2(1) = .48, p > .45). Thus, the d' did not differ whether it
involved the variations A or the variations B. By contrast, the d' for the C+ trials was
larger in the second condition, i.e. when the extract of Diabelli's Rondo played the role
of standard sequence and Schubert's Ländler played the role of C+ trials (Chi2(1) =
4.29, p < .04). Once again, the sequence A3 of Schubert's piece is mainly responsible
for this pattern of results, since it ellicited significantly more head-turns than the other
when it belonged to the to-be-responded-to sequences (see figure 1C).
In conclusion, the infants appeared to be more efficient in this experiment than in the
first one, since they responded mainly to the C+ trials, whereas the C- trials provoked
less responses, even when it comprised the A3 of Schubert's piece.
The results of this experiment are more readily interpretable in terms of
categorization as those of the first experiment. Indeed, the infants neglected the
variations of the standard sequence and responded to the variations of the comparison
sequence. It could be concluded that they detected the common feature to the sequences
- the cue - and responded to the sequence that shared the same cue when it was
reinforced by the cartoon.
One could argue against the categorization hypothesis from the non-significant d' for
the C- trials: this could mean that infants could simply not discriminate between the
parent motif and its variations because they could not perceive the differences between
the items of the same category. The answer to this argument is a matter of empirical
evidence. This experiment is still to be done. In the meantime, our results give several
indirect supports to the categorization hypothesis. In the first experiment, a d'
significantly different from zero was observed for the C- trials of the first condition.
Yet, in this condition the C- trials were made up of variations a of Schubert's piece. It
revealed, therefore, that infants did actually discriminate the variations of the Standard
sequence from the Standard sequence itself. In the third experiment, when Schubert
served as the Standard category, infants categorized each sequence appropriately.
However, for the second condition using Diabelli's sequences as Standard category, it
cannot be concluded that infants categorised so well, i.e. that they perceived each
member of the Standard category as distinct exemplars of that category. Nevertheless,
we can suppose it was really the case, since there was no difference between the two
conditions for the d' associated to C- trials. If the infants categorized Schubert's
sequences, there is no reason to conclude they did not categorize Diabelli's sequences.
General discussion
On the whole, the present study suggests that infants aged from 6 to 10 months were
able to form musical categories. Yet, the results of the first experiment were
unconclusive, since it failed to demonstrate discrimination of the categories and
suggested, by contrast, excessive discrimination between the members of one category
(those derived from the Parent A of Schubert's piece). We speculated that infants could
succeed if the task was made easier. Rather than to decrease the distinctiveness of the
members of the categories, the distinctiveness of the categories themselves was
increased by presenting motifs coming from different pieces (Parent A and derivatives
from Schubert's piece vs Parent B and derivatives from Diabelli's piece). Tested with
such items, infants proved to be able to attribute to each category its examplars, while
remaining able to discriminate the derivatives of Schubert piece's parent motif.
Discrimination between the exemplars of the second cue, the variations of the parent
motif B of Diabelli's piece is not demonstrated directly. Nevertheless, we think we have
enough indirect evidence to conclude that, probably, the infants did discriminate
between members of both categories.
Referring back to Deliège's model of music perception and to Rosch's model of
categorisation, the present results suggest that infants can build categories by collecting
structural features that define these categories. Infants were able to extract the structural
features of the parent motifs - the cues - and to perceive new items as variations of these
cues. They extended the size of the category by assimilating to the parent motif the
sequences that derivated from that motif.
Of course, this process is not errorfree. One constant in this series of experiment is
the particular status of the sequence A3, the third derivative of the Ländler by Schubert.
Both infants (experiments 1 and 3) and adults (experiment 2) found this sequence very
distinctive. In experiment 1, the distinctivity of that sequence was even larger than the
distinctivity between categories, prevented to conclude to categorization. It must be
underlined that older children experienced similar difficulties in classification of this
part of Schubert's piece (Mélen et al., this symposium). Moreover, this can be referred
to data from Jolicoeur, Gluck, and Kosslyn (1984) showing that atypical members of a
category (like penguins or ostriches for the birds) are more readily classified at their
own "entry level" than at the basic-level (they are classified as, respectively "penguins"
or "ostriches" rather than as birds). They are so different from the other exemplars that
they seem to form the first examplar of a new category, which turns out to be very
restricted. Thus, even the errors made by the infants manifest the underlying
categorization process.
Further studies are needed to precise the structural characteristics on which the
infants decide to accept or to reject a particular examplar as a member of a particular
category and to precise the internal structuration of the categories. Nevertheless, this
paper can be considered as a first demonstration that principles of sameness and
difference underlye music perception in infancy at more complex levels than rhythmic
grouping.
References
Behl-Chadha, G.(1996). Basic-level and superordinate-like categorical representations
in early infancy. Cognition, 60, 105-141,
Clarkson, M. G. & Clifton, R. K. (1985). Infant pitch perception: Evidence for
responding to pitch categories and missing fundamental. Journal of the Acoustical
Society of America, 77, 1521-1528,
Deliège, I. (1987). Le parallélisme, support d'une analyse auditive de la musique : vers
un modèle des parcours cognitifs de l'information musicale. Analyse Musicale, 6, 7379.
Deliège, I. (1996). Cue abstraction as a component of categorization processes in music
listening. Psychology of music, 24, 131-156,
Deliège, I., & Mélen, M. (1997). Cue abstraction in the representation of musical form.
In I. Deliège, & J. Sloboda (Eds.), Perception and Cognition of music (pp. 387-412).
Hove: Psychology Press.
Jolicoeur, P., Gluck, M. A., & Kosslyn S. M. (1984). Picture and names: Making the
connection. Cognitive Psychology, 16, 243-275.
Kuhl, P. K. (1991). Human adults and human infants show a "perceptual magnet effect"
for the prototypes of speech categories, monkeys do not. Perception and
Psychophysics, 50, 93-107,
Lécuyer, R., & Poirier, C. (1994). Categorization in the five-month-old infants. Cahiers
de Psychologie cognitive, 13(1), 193-509,
Cognitive, 13(1), 193-509.
Mandler, J.M., Bauer, P., & McDonough, L. (1993). Separating the sheep from the
goats: Differentiating global categories. Cognitive Psychology, 23, 263-298.
Mélen, M. (1997). Les principes du même et du différent comme organisateurs du
groupement rythmique chez les nourrissons de six à dix mois. Thèse de doctorat en
psychologie non publiée, Université de Liège, Liège.
Mélen, M. (1999a). Les principes du même et du différent comme organisateurs du
groupement rythmique chez le nourrisson: Arguments théoriques. Musicae Scientiae,
3(1), 41-66.
Mélen, M. (1999b). Les principes du même et du différent comme organisateurs du
groupement rythmique chez le nourrisson: Une étude empirique. Musicae Scientiae,
3(2), 161-191,
Mervis, C. B. (1987). Child-basic object categories and early development. In U.
Neisser (Ed.), Concept and conceptual development (pp. 201-233). Cambridge (UK):
Cambridge University Press.
Quinn, P. C. (1987). The categorical representation of visual pattern information by
young infants, Cognition, 27, 145-179.
Quinn, P. C. (1998). Object and spatial categorization in young infants: "what" and
"where" in early visual perception. In A. Slater (Ed.), Percetual development: Visual,
auditory, and speech perception in infancy (pp. 131-165). Hove: Psychology Press.
Rosch, E., & Lloyd, B. B. (1978). Cognition and Categorization. Hillsdale, NJ:
Lawrence Erlbaum Associates.
Rosch, E., Mervis, C. B. Gray, W. D., & Boyes-Braem, P. (1976). Basic objects in
natural categories. Cognitive Psychology, 8, 382-439.
Schonen, S., & Deruelle, C. (1991). Spécialisation hémisphérique et reconnaissance des
formes et des visages chez le nourrisson. L'Année Psychologique, 91(1), 15-16.
Thorpe, L.A., Trehub, S.E., Morrongiello, B.A., & Bull, D. (1988). Perceptual grouping
by infants and preschool children. Developmental Psychology, 24, 484-491.
Trehub, S. E., & Thorpe, L. A. (1989). Infants' perception of rhythm: Categorization of
auditory sequences by temporal structure. Canadian Journal of Psychology, 43, 217229.
Trehub, S. E., Endman, M. W., & Thorpe, L. A. (1990). Infants' perception of timbre:
Classification of complex tones by spectral structure. Journal of Experimental Child
Psychology, 49, 300-313.