Content uploaded by Simone Dalla Bella
Author content
All content in this area was uploaded by Simone Dalla Bella
Content may be subject to copyright.
Perception & Psychophysics
2003, 65 (7), 1019-1028
In everyday life, listeners rapidly switch from one radio
station to the other, assessing within a few seconds whether
or not they are hearing their preferred song. Similarly, in
the radio game “Name-That-Tune,” when asked to recog-
nize popular musical excerpts, contestants can often pro-
vide a correct guess within a few notes. From these exam-
ples, recognizing well-known melodies may appear to be
a fairly automatic activity that can be performed with lit-
tle effort and with a minimal amount of information. How-
ever, despite its apparent simplicity, music recognition in-
volves several cognitive operations. In order for a melody
to be recognized, a structural representation of the pre-
sented musical stimulus has to be computed on the basis
of its perceptual features. Then this representation has to
be matched to a representation stored in memory. Recog-
nition occurs when there is a correspondence between the
perceptual and the stored representations. The study of the
musical features involved in this matching process has re-
ceived particular attention. For instance, researchers have
shown that changes in melodic contour (i.e., the sequence
of ups and downs in a melody, regardless of interval size;
Dowling & Fujitani, 1971; Dowling & Hollombe, 1977;
Idson & Massaro, 1978; Kallman & Massaro, 1979) and
in tone chroma (i.e., the position of the notes relative to
the musical octave; Idson & Massaro, 1978; Kallman &
Massaro, 1979; for a different view, see Deutsch, 1972)
are both detrimental to recognition. The temporal dimen-
sion is less important for recognition than the pitch di-
mension is (Hébert & Peretz, 1997; White, 1960).
Surprisingly, none of these studies has attempted to ex-
plore the dynamic
process
of melody recognition. There-
fore, there is currently no model that explains which prop-
erties contribute to the recognition process while the
melody unfolds over time. One model that has proved use-
ful in auditory word recognition is the cohort model
(Marslen-Wilson, 1987; Marlsen-Wilson & Welsh, 1978;
for reviews, see Frauenfelder, 1996; Protopapas, 1999).
This could serve as a good departure point for modeling
the time course of melody recognition. In the most recent
version of the cohort model (Marslen-Wilson, 1987), the
overall process of word recognition is divided into three
stages: access, selection, and integration. In the access
stage, the first 100–150 msec of the incoming acoustic sig-
nal (i.e., the beginning of the word) activate a series of po-
tential lexical candidates by multiple access. This set of
candidates is referred to as the
word-initial cohort
and is
determined solely on the basis of bottom-up information
deriving from the sensory input. In the selection stage, the
cohort is progressively reduced with the increase of avail-
able acoustic information about the word. During this dis-
crimination phase, the level of activation of the candidates
is raised or lowered depending on their compatibility with
the sensory input, until the best-fitting match is selected.
Recognition is achieved when the most activated candi-
date surpasses a threshold value, as compared with its near-
est lexical competitors. Finally, the integration stage im-
1019 Copyright 2003 Psychonomic Society, Inc.
The research was supported by a doctoral scholarship from the
Groupe de Recherche en Neuropsychologie Expérimentale (University
of Montreal) to S.D.B., a fellowship from the Fonds Concertés de l’Aide
à la Recherche to N.A., and a grant from the Natural Science and Engi-
neering Research Council of Canada to I.P. We thank Glenn Schellen-
berg, Bruno Repp, and one anonymous reviewer for helpful comments
on an earlier draft of this article. Correspondence concerning this arti-
cle should be addressed to S. Dalla Bella, Department of Psychology,
Ohio State University, 1885 Neil Ave., Columbus, OH 43210 (e-mail:
dalla-bella.2@osu.edu).
Time course of melody recognition:
A gating paradigm study
SIMONE DALLA BELLA, ISABELLE PERETZ, and NEIL ARONOFF
University of Montreal, Montreal, Quebec, Canada
Recognizing a well-known melody (e.g., one’s national anthem) is not an all-or-none process. Instead,
recognition develops progressively while the melody unfolds over time. To examine which factors gov-
ern the time course of this recognition process, the gating paradigm, initially designed to study auditory
word recognition, was adapted to music. Musicians and nonmusicians were presented with segments
of increasing duration of familiar and unfamiliar melodies (i.e., the first note, then the first two notes,
then the first three notes, and so forth). Recognition was assessed after each segment either by requir-
ing participants to provide a familiarity judgment (Experiment 1) or by asking them to sing the melody
that they thought had been presented (Experiment 2). In general, the more familiar the melody, the fewer
the notes required for recognition. Musicians judged music’s familiarity within fewer notes than did
nonmusicians, whereas the reverse situation (i.e., musicians were slower than nonmusicians) occurred
when a sung response was requested. However, both musicians and nonmusicians appeared to segment
melodies into the same perceptual units (i.e., motives) in order to access the correct representation in
memory. These results are interpreted in light of the cohort model (Marslen-Wilson, 1987), as applied
to the music domain.
1020 DALLA BELLA, PERETZ, AND ARONOFF
plies that the syntactic and semantic information of the
recognized word is integrated within the higher level rep-
resentation of the utterance.
Based on the cohort model, the gating paradigm intro-
duced by Grosjean (1980; for a review, see Grosjean, 1996)
has represented one of the major procedures used to study
the time course of spoken word recognition (see Marslen-
Wilson, 1987). Participants are presented with segments
of a word of increasing duration (for instance, the first
30 msec, then the first 60 msec, and so forth). At each pre-
sentation, the length of the acoustic segment is increased
by a constant (e.g., 30 msec). Then, after each presentation,
the participants have to guess the identity of the word, and
they are requested to indicate how confident they are
about their guess (e.g., on a scale from
very sure
to
very
unsure
). The point at which a word is identified is esti-
mated in two ways. The isolation point (IP, measured in
milliseconds or percentage of the word) is the point at
which the word is correctly identified, without any change
in response thereafter. The term
isolation
refers to the se-
lection of a unique element within a cohort. Isolation does
not imply that the participants are totally confident about
their guess. Indeed, maximum confidence is reached at
the recognition point (RP, also called the
total acceptance
point
, measured in milliseconds or percentage of the
word), which is the point at which the participants are cer-
tain of having identified the word. Hence, IP does not nec-
essarily coincide with RP.
The usefulness of the gating paradigm for uncovering
the time course of spoken word recognition has been largely
validated by replicating well-established effects formerly
obtained with other psycholinguistic paradigms (for a re-
view, see Grosjean, 1996). A robust phenomenon that has
been successfully and consistently obtained with the gat-
ing paradigm is the word frequency effect (Grosjean,
1980; Tyler, 1984; Walley, Michela, & Wood, 1995). High-
frequency words are recognized more rapidly than low-
frequency words, in that their IP and RP occur earlier. This
phenomenon is of particular interest to us, insofar as an
analogous effect is likely to occur when recognition of
highly familiar and moderately familiar music is com-
pared, as will be clarified below.
Despite the successful application of the gating para-
digm in psycholinguistic research, it has been argued that
the unnatural sequential format of presentation and the
fact that participants have unlimited response time may
induce a task-specific strategic behavior. Hence, the gat-
ing paradigm might provide an unrealistic image of the
word recognition process. However, both objections have
been discarded. The results of the original study (Gros-
jean, 1980) were replicated when each participant was
presented with only one fragment of a word (Cotton &
Grosjean, 1984; Walley et al., 1995) and when participants
were encouraged to respond as rapidly as possible (Tyler
& Wessels, 1985). Thus, neither the presentation format
nor the unlimited response time appears to detract from
the validity of the gating paradigm.
Our main purpose in the present study was to explore
the time course of the melody recognition process through
the gating paradigm. This procedure appears to be well
suited for the study of music recognition. Gated presenta-
tion of music has been previously adopted in music cog-
nition research (e.g., Bigand & Parncutt, 1999; Palmer &
Krumhansl, 1987a, 1987b; Peretz, Gagnon, & Bouchard,
1998). Moreover, gated presentation has been successfully
applied to the recognition of acoustical properties of mu-
sical recordings (Schellenberg, Iverson, & McKinnon,
1999). Therefore, on the basis of these previous applica-
tions to the study of musical phenomena, the gating para-
digm appears to be a good choice for studying the process
by which well-known tunes are recognized.
A second aim of our study was to examine whether the
time course of music recognition is affected by musical
training and by music familiarity. Training and familiarity
are major factors that typically influence music process-
ing (for reviews, see Dowling & Harwood, 1986; Smith,
1997). For example, musicians perform better than non-
musicians on many musical tasks, such as recognizing the
exact transposition of a melody (e.g., Bartlett & Dowling,
1980; Halpern, Bartlett, & Dowling, 1995), detecting
whether two melodies are the same or different (e.g., Dow-
ling, 1978, 1984), or rating to what degree a certain tone
fits with an established musical key (e.g., Krumhansl &
Shepard, 1979; for a review, see Krumhansl, 1990). Sim-
ilarly, performance in musical tasks is usually better with
familiar than with unfamiliar material. Familiarity im-
proves nonmusicians’ performance, for instance, when they
are asked to transpose a melody (Attneave & Olson, 1971),
recognize the exact transposition of a melody (Bartlett &
Dowling, 1980), or categorize musical intervals (Smith,
Nelson, Grohskopf, & Appleton, 1994). In sum, the promi-
nence of the effects of training and familiarity on music
processing indicates that these factors may influence the
time course of music recognition as well.
A further objective was to explore the possibility that
structurally salient units inherent to the melody itself gov-
ern the recognition process. A melody is unlikely to be
perceived in a note-by-note fashion; instead, grouping
processes are thought to occur (Deliège, 1987; Lerdahl &
Jackendoff, 1983; Peretz, 1989; for a review, see Handel,
1989). Listeners segment a musical sequence according
to its surface properties (e.g., pitch, intensity, duration,
and so forth), following the bottom-up principles laid
down by Gestalt psychologists. A change in any of these
parameters leads to the perception of a break in the se-
quence and to the creation of groups separated by the
changes (i.e., at the group boundaries). These groups are
typically referred to as
motives
. Clustering notes into mo-
tives allows the storing of more information in working
memory, since the events can be grouped into perceptual
units that are larger than the individual notes. A typical
motive is about three to four notes long and is considered
to be the basic segmentation unit of a musical sequence
(e.g., Lerdahl & Jackendoff, 1983). Since motivic struc-
ture has been found to affect the way musicians and non-
musicians segment melodies (Deliège, 1987; Peretz, 1989),
it might also play a relevant role in memory access during
the recognition process.
MUSIC RECOGNITION PROCESS 1021
For these purposes, unfamiliar and familiar synthesized
tunes taken from the traditional song repertoire were played
with gated presentation to musicians and nonmusicians.
Familiar tunes were divided into highly familiar and mod-
erately familiar in order to parallel the opposition between
frequent and less frequent words in psycholinguistics re-
search. Instead of using fixed duration gates, as adopted
in Grosjean’s (1980) original study, we employed a vari-
able size gate corresponding to the duration from onset to
offset of each note in the melody. Indeed, the variability in
note duration and tempi from one melody to the next
makes the use of a fixed gate inadequate. For instance,
choosing the shortest note gate would result in tediously
long experimental sessions, whereas larger gates would
lack precision. Moreover, it is likely that the number of
notes is more relevant to melody recognition than is the
duration of the segment. In Experiment 1, participants were
requested to judge whether the tunes were familiar or un-
familiar. This judgment reflects the general feeling of
knowing evoked by the presented tune without necessar-
ily identifying it—for instance, by providing its title. The
feeling-of-knowing judgment is treated as an estimate of
the retrievability of the melody from memory (see, e.g.,
Koriat, 1993; Koriat & Levy-Sadot, 2001) and has been
found to be a good predictor of melody recognition
(Peynircio†glu, Tekcan, Wagner, Baxter, & Shaffer, 1998).
In Experiment 2, recognition was assessed by asking par-
ticipants to sing the tune that they thought had been pre-
sented. Sung responses allowed a more reliable verifica-
tion of recognition than did the usual naming or title
verification tasks. Indeed, a tune can be well recognized
even though its title cannot be retrieved from memory.
The following predictions were made. Because the feel-
ing of knowing can emerge in the absence of recognition
(e.g., in the
tip-of-tongue
effect; see Brown, 1991), it was
expected that fewer notes would be necessary to judge the
familiarity of a musical excerpt than to sing its continua-
tion. Moreover, by analogy with the language domain (i.e.,
the word frequency effect), it was predicted that highly fa-
miliar tunes would lead to a feeling of knowing and would
be identified earlier than moderately familiar tunes. Fa-
miliar tunes would, in turn, require fewer notes to be cor-
rectly judged than unfamiliar tunes, since to provide a
nonfamiliarity
judgment, an exhaustive search in memory
is more likely to be performed. In addition, since musi-
cians typically perform better than nonmusicians on mu-
sical tasks, we predicted that the former would be more
confident in their responses, thus showing earlier recog-
nition than the latter. Finally, if motives contribute to
recognition, it was expected that the emergence of a feel-
ing of knowing or the isolation of a melody would occur
at motives’ boundaries.
EXPERIMENT 1
Method
Participants. Twenty-three students from the University of
Montreal volunteered to participate in the experiment. The partici-
pants were assigned to one of two groups defined by musical train-
ing. Eleven (6 males and 5 females) were referred to as musicians. They
had at least 4 years of formal musical training (M 5 7.7 years). Their
ages ranged from 20 to 27 years (M 5 22.3 years). The other 12
(6 males and 6 females) were classified as nonmusicians, since they
had no formal education in music. Their ages ranged from 19 to
27 years (M 5 20.8 years). All the participants were francophone
and had lived in Quebec since the age of 2 years or earlier. These cri-
teria served to promote a homogeneous musical culture among the
participants (Peretz, Babaï, Lussier, Hébert, & Gagnon, 1995). All
the participants were remunerated.
Material. Eighty songs (40 familiar and 40 unfamiliar) were se-
lected to be used in the experiment. For each song, solely the melody,
without lyrics and accompaniment, was utilized as a stimulus. All
the melodies were taken from a repertoire of French traditional
songs (Berthier, 1979) for which norms of familiarity had previously
been established (see Peretz et al., 1995). Seven amateur musicians
were asked to segment each of the 88 tunes considered in our norms
into their natural parts by placing slashes on scores. The first and sec-
ond motive boundaries were determined by compiling their seg-
mentation data. Only those 40 melodies whose boundaries were
marked by the majority were considered appropriate for the study.
Half (20) were the most familiar melodies (mean familiarity rating 5
4.9, range 5 4.7 –5.0, with 1 meaning unfamiliar and 5 meaning
very familiar; see Peretz et al., 1995), and the other half (20) were
the least familiar melodies (mean familiarity rating 5 4.1, range 5
3.2– 4.6). Throughout the article, these two subsets of familiar
melodies will be referred to as high-familiarity tunes and moderate-
familiarity tunes, respectively. The number of notes in the high-
familiarity tunes ranged from 10 to 30 (M 5 16.1) and, in the
moderate-familiarity tunes, from 11 to 31 (M 5 20.9). Duration of
tune excerpts varied from 4.0 to 17.4 sec. The length of each melody
was the same as that used in the normative study (Peretz et al., 1995)
and provided ample opportunities for identification. Forty unfamil-
iar tunes, taken from the same repertoire of traditional songs
(Berthier, 1979), were also used in the experiment. These were con-
sidered unfamiliar by more than 200 students tested in a previous
study (Peretz, Gaudreau, & Bonnel, 1998). The number of notes in
the unfamiliar tunes ranged from 11 to 33 (M 5 18.8), and they
lasted between 4.5 and 13.2 sec.
All the melodies were played with a piano timbre through a Yamaha
TX-81Z synthesizer under the control of an IBM-compatible com-
puter. The analog ouput was sampled onto a Macintosh PowerPC
computer, using SoundEdit II software. The beginning and the end
of each note in a melody were determined manually through visual
inspection of the waveform. Boundaries were carefully determined
in order to minimize spectral splatter at the offset of each note. When
rests occurred in a tune, they were treated as part of the previous
note. The information on notes’ boundaries was used by the Psyscope
software (Cohen, MacWhinney, Flatt, & Provost, 1993) to play each
tune with gated presentation, as will be described below.
Procedure. Musicians and nonmusicians were tested individu-
ally, using the same set of 80 tunes, presented in one of two random
orders. Each tune was played with gated presentation. An example
of gated presentation is illustrated in Figure 1. For the first presen-
tation, only the first note of the melody (i.e., the first gate) was
played. On the following presentation, the first two notes of the
melody (the first two gates) were played, and so forth. After each
presentation, the participants were asked to say whether the tune was
familiar or unfamiliar and to rate their level of confidence on a 7-
point scale (with 1 5 very unsure and 7 5 very sure). The partici-
pants were asked to provide a response even though, at first, they
were unsure. Presentation of each melody lasted until the participant
gave three successive consistent judgments with a confidence rating
of 7. The responses were provided verbally and were noted by the ex-
perimenter.
The participants were informed that they would be hearing a set
of familiar and unfamiliar songs, and the nature of the presentation
format was explained. Moreover, a practice trial was given in order
1022 DALLA BELLA, PERETZ, AND ARONOFF
to familiarize the participant with the task. The participants were
tested in a quiet room, and the stimuli were presented through Labtec
(Spatializer) loudspeakers in free field, at a comfortable intensity level.
The test was divided into two sessions of about 1.5 h each. The exper-
iment was run on PsyScope, using a Macintosh computer PowerPC.
Results
One familiar melody and one unfamiliar melody were
judged to be unfamiliar and familiar, respectively, by more
than 50% of the participants. These were discarded from
the analyses. Similarly, when a participant did not provide
three successive maximum ratings before the end of the
melody, his or her responses were not considered in the
analyses. The overall rate of rejection was 14.7% of the re-
sponses (corresponding to 270 out of 1,840 judgments).
For each of the 78 remaining tunes, two measures were
taken to express the number of notes needed to develop a
feeling of knowing: The familiarity emergence point
1
(FEP), which was the point at which the participant correctly
began to consider that a melody was familiar, corresponded
to the note (or gate) number at which the correct response
familiar
was given for the f irst time and was never
changed thereafter. The familiarity point (FP) was the
point at which the participant was totally confident that
the melody was familiar or unfamiliar. FP corresponded to
the note (or gate) number at which the participant pro-
vided a correct answer and at which he or she reached
maximum confidence for the first of three consecutive
presentations. Note that FP is used with both familiar and
unfamiliar melodies, even though, in the latter case, the
term
unfamiliarity point
would be more adequate. Never-
theless, for sake of simplicity, only FP will be used through-
out the article. In addition, it is worth mentioning that al-
though FEP applies well to familiar melodies, this measure
seems unsound with unfamiliar melodies. Typically, a
melody is considered unfamiliar when its first notes are
presented, so it is impossible to determine when a
feeling
of not knowing
does emerge. Thereby, only FEPs obtained
with familiar melodies were submitted to the following
analyses.
In order to assess whether FEP and FP varied as a function
of the melodies’ familiarity and whether these measures
were affected by musical training, separate 2 (group)
3
2
(familiarity) repeated measures analyses of variance
(ANOVAs) were run by taking either subject (
F
1
) or item
(
F
2
) as the random variable. When subject was the ran-
dom variable, group (musicians or nonmusicians) was the
between-subjects factor, and familiarity (familiar or unfa-
miliar) was the within-subjects factor. When item was the
random variable, group (musicians or nonmusicians) was
considered as the within-items factor, and familiarity (fa-
miliar or unfamiliar) as the between-items factor. Further
analyses were performed in order to compare the results
obtained with high-familiarity and moderate-familiarity
tunes. In this case, the degree of familiarity (high famil-
iarity and moderate familiarity) was used, instead of fa-
miliarity, as a factor.
Unfamiliar versus familiar melodies. Throughout the
study, only the number of notes (e.g., at FP) were submit-
ted to the analyses. The mean stimulus duration will also
be reported, but for descriptive purposes. The mean FPs
obtained with unfamiliar and familiar melodies for the
musicians and the nonmusicians are presented in Table 1.
It was found that FP occurred earlier with familiar than
with unfamiliar melodies. This difference was affected by
musical training, as was revealed by a significant group
3
familiarity interaction [
F
1
(1,21)
5
5.66,
p ,
.05;
F
2
(1,76)
5
14.15,
p ,
.001]. Post hoc pairwise comparisons (by
Figure 1. Example of gated presentation as applied to the melody of “Happy Birthday,” showing Presentations 1–4.
The motive boundary of the melody is also displayed.
MUSIC RECOGNITION PROCESS 1023
Tukey HSD, corrected for unequal samples) revealed that
musicians attained FP earlier than did nonmusicians.
However, this effect was larger with unfamiliar melodies
(
p ,
.001, by subject and by item) than with familiar
melodies (
p ,
.001, by item, but nonsignificant in the
analysis by subject). Nevertheless, for both groups, FP oc-
curred later with unfamiliar melodies than with familiar
melodies (with
p ,
.001 in the analysis by subject and in
the analysis by item).
High-familiarity versus moderate-familiarity
melodies. The mean FEPs and FPs obtained with high-
familiarity and moderate-familiarity tunes for the musicians
and the nonmusicians are shown in Table 2. Both FEP and
FP were found to occur earlier with high-familiarity tunes
than with moderate-familiarity tunes, as was attested by a
significant main effect of degree of familiarity [at FEP,
F
1
(1,21)
5
71.17,
p ,
.001, and
F
2
(1,37)
5
11.85,
p ,
.01; at FP,
F
1
(1,21)
5
117.38,
p ,
.001, and
F
2
(1,37)
5
21.23,
p ,
.001]. Moreover, the musicians attained FEP
and FP earlier than did the nonmusicians; yet this main ef-
fect of group was significant solely in the analysis by item
[at FEP,
F
2
(1,37)
5
21.17,
p ,
.001; at FP,
F
2
(1,37)
5
18.36,
p ,
.001]. Finally, the group
3
degree of familiar-
ity interaction did not reach significance. In order to ex-
amine whether the musicians were more confident than
the nonmusicians about their responses at the FEP, a sim-
ilar 2 (group)
3
2 (degree of familiarity) repeated mea-
sures ANOVA was performed in which the mean confi-
dence rating provided at the FEP was considered the
dependent variable (for the mean confidence values, see
Table 2). The musicians appeared to be more confident than
the nonmusicians; however, the group effect was again
limited to the analysis by item [
F
1
,
1;
F
2
(1,37)
5
8.38,
p ,
.01]. Note that confidence was higher for moderate-
familiarity tunes than for high-familiarity tunes, but this
main effect of degree of familiarity reached significance
only with the analysis by subject [
F
1
(1,21)
5
7.81,
p ,
.05]. The group
3
degree of familiarity interaction did not
reach significance.
One of the purposes of the present study was to assess
whether the emergence of a feeling of knowing (i.e., at
FEP) coincides with the boundaries of motives. To do so,
the mode of the FEPs and of the FPs (i.e., the gate num-
ber at the FEP or the FP occurring most frequently) was
taken for each melody. Then the percentage of melodies in
which the FEP or the FP coincided with the boundary of
the first or the subsequent motive was computed. These
percentages were compared with a chance level. If the FEP
or the FP occurred on the motives’ boundaries, it would be
expected that the FEP or the FP would fall significantly
more often on this position than on other notes of the
melody preceding or following the FEP or the RP. For our
purposes, we considered the four notes around the mode
(i.e., two preceding and two following the mode), so that
the chance probability was 20%. This comparison re-
vealed that the percentage of melodies at which FEP cor-
responded to a motive boundary was significantly above
chance for the nonmusicians (42.6%,
p ,
.05, by a bino-
mial test; 17.9% for the musicians). Conversely, the FPs
significantly coincided with motive boundaries for the
musicians (35.9%,
p ,
.05 , by a binomial test; 28.9% for
the nonmusicians).
Discussion
The results show that 3 to 6 notes (i.e., about 2 sec) of
a familiar melody were sufficient to evoke a feeling-of-
knowing judgment (at the FEP). However, to gain full con-
fidence about this first estimate, approximately 2 addi-
tional notes were needed (FP). As was predicted, decisions
were made later with unfamiliar melodies than with fa-
miliar melodies: 8 to 10 notes were needed to decide that
melodies are unfamiliar. Moreover, a clear effect of fa-
miliarity was shown when high-familiarity tunes were
Table 1
Mean Familiarity Points in Experiment 1 With Familiar
and Unfamiliar Melodies for Musicians and Nonmusicians,
With the Corresponding Mean Duration
Notes
Group No. SE Duration (sec)
Musicians
Familiar tunes 6.3 0.2 3.3
Unfamiliar tunes 8.5 0.4 3.4
Nonmusicians
Familiar tunes 6.9 0.3 3.6
Unfamiliar tunes 10.2 0.5 4.1
Table 2
Mean Familiarity Emergence Points (FEPs) and Mean Familiarity Points
(FPs) in Experiment 1 With High-Familiarity (HF) and Moderate-
Familiarity (MF) Tunes for Musicians and Nonmusicians,
With Mean Durations and Mean Confidence Ratings
FEP FP
Notes Duration Notes Duration
Group No. SE (sec) Confidence No. SE (sec)
Musicians
HF tunes 3.2 0.3 1.8 3.4 5.4 0.2 2.9
MF tunes 4.8 0.4 2.6 4.0 7.3 0.3 3.8
Nonmusicians
HF tunes 4.0 0.4 2.3 3.2 5.9 0.2 3.1
MF tunes 5.7 0.5 3.0 3.4 8.2 0.4 4.2
1024 DALLA BELLA, PERETZ, AND ARONOFF
compared with moderate-familiarity tunes. Fewer notes
were needed to judge high-familiarity tunes as familiar, as
compared with moderate-familiarity tunes. This finding
is unlikely to have stemmed from the participants’ being
more confident with high-familiarity than with moderate-
familiarity tunes. Indeed, the opposite pattern was found
(i.e., the participants were more confident when judging
moderate-familiarity melodies than with high-familiarity
melodies). Overall, these results reveal a robust effect of
familiarity on the music recognition process.
As was expected, musical training was found to affect
the familiarity judgments, since the musicians reached the
FEP and the FP earlier than did the nonmusicians. How-
ever, the musicians’ advantage was less pronounced for
familiar than for unfamiliar melodies. This finding is con-
sistent with prior evidence showing that the use of famil-
iar music tends to reduce musicians’ advantage, as com-
pared with nonmusicians (Smith et al., 1994). In addition,
the musicians were more confident in their judgments
than were the nonmusicians at the FEP. This result may
explain why the two groups obtained different FPs, inso-
far as the musicians’ higher level of confidence at the FEP
might have led them to reach maximum confidence within
fewer notes (i.e., at the FP). Further support for this inter-
pretation can be found in the observation that motive
structure had a larger and earlier impact on the judgments
of the nonmusicians than on those of the musicians. In
conclusion, these findings indicate that musical training
affects the efficiency of the recognition process.
EXPERIMENT 2
In the second experiment, we sought to explore the time
course of melody recognition by asking the participants to
sing melodies, instead of providing familiarity judgments.
For this purpose, the participants were presented with
gated melodies and had to sing the tune that they thought
had been played.
Method
Participants. Twenty-four participants who did not participate in
Experiment 1 were recruited, mostly from the university commu-
nity, to participate in the experiment. Half (6 males and 6 females)
formed the musicians’ group. They had at least 4 years of formal
musical training (M 5 9.4 years). Ten had degrees in music, and all
had specific training or experience in singing. Their ages ranged
from 21 to 45 years (M 5 27.0 years). The other half (3 males and
9 females) formed the nonmusicians’ group, since they had no for-
mal musical training. Despite their lack of musical education, how-
ever, the nonmusicians were willing to sing. Their ages ranged from
20 to 42 years (M 5 28.3 years). As in Experiment 1, all the partic-
ipants were francophone and had lived in Quebec since the age of
2 years or earlier. All the participants were remunerated.
Materials and Procedure.The materials and the presentation
format were the same as those in Experiment 1, except that only the
40 familiar songs (20 high-familiarit y tunes and 20 moderate-
familiarity tunes) were considered here. The participants were in-
structed to listen to each presentation of a melody, to sing the melody
that they thought had been presented, and to indicate their level of
confidence on a 7-point scale (with 1 5 very unsure and 7 5 very
sure). They were encouraged to sing whatever came to mind and, in
addition, to provide a title when one was available. The trial lasted
until the participant was able to sing the correct melody and had pro-
vided three consecutive confidence ratings of 7. A melody was con-
sidered correctly sung when the participant produced at least three
recognizable notes beyond the end of the gated stimulus, as judged
by the experimenter, a professional singer. At the end of each trial,
feedback (including the title) was given. The sung responses were
tape-recorded, and the confidence ratings were noted by the exper-
imenter.
The participants were informed that they would be hearing a set
of familiar songs, and the nature of the presentation format was ex-
plained. Moreover, there was a practice trial. The experiment lasted
approximately 2 h. The experiment was run on the Experimenter
software (Altmann, Wathanasin, Birkett, & Russell, 1992), using a
Macintosh computer IIfx.
Results
The data were scored according to the same criteria as
those used in Experiment1. Two moderate-familiarity tunes
were dropped from the data set, since they were not rec-
ognized by more than 50% of the participants. In all, 129
judgments out of 960 (13.4% of the entire set) were dis-
carded.
2
The number of notes needed to recognize a melody
was expressed by the IP and the RP. The IP was the point
at which the participant demonstrated a correct insight
into the identity of the melody, by singing it for the first
time. The IP corresponded to the note (or gate) number at
which the participant first successfully sang three notes of
the target melody beyond the presented melodic segment,
Table 3
Mean Isolation Points (IPs) and Mean Recognition Points (RPs)
in Experiment 2 With High-Familiarity (HF)
and Moderate-Familiarity (MF) Tunes for Musicians
and Nonmusicians, With Mean Durations and Mean Confidence Ratings
IP RP
Notes Duration Notes Duration
Group No. SE (sec) Confidence No. SE (sec)
Musicians
HF tunes 5.2 0.2 2.5 5.8 6.0 0.2 3.0
MF tunes 6.9 0.3 3.6 5.8 8.0 0.2 4.1
Nonmusicians
HF tunes 5.0 0.1 2.5 4.5 6.5 0.3 3.2
MF tunes 6.6 0.3 3.7 4.6 8.4 0.4 4.4
MUSIC RECOGNITION PROCESS 1025
without any change in response thereafter. The RP was the
point at which the participant was certain of having sung
the corresponding melody. The RP corresponded to the
note (or gate) number at which the participant correctly
sang the melody and at which his or her confidence rating
reached 7 for the first of three consecutive presentations.
The mean IPs and RPs obtained with high-familiarity
tunes and moderate-familiarity tunes for the musicians
and the nonmusicians are shown in Table 3. The average
IP ranged from about five to seven notes, and the RP
ranged from six to eight notes. IPs and RPs were submit-
ted to 2 (group)
3
2 (degree of familiarity) repeated mea-
sures ANOVAs by considering either subject (
F
1
) or item
(
F
2
) as the random variable. When subject was the ran-
dom variable, group (musicians or nonmusicians) was the
between-subjects factor, and degree of familiarity (high
familiarity or moderate familiarity) was the within-subjects
factor. When item was the random variable, group (musi-
cians or nonmusicians) was considered the within-items
factor, and degree of familiarity (high familiarity or mod-
erate familiarity) was considered the between-items factor.
The analyses revealed that both the IP and the RP oc-
curred later for moderate-familiarity tunes than for high-
familiarity tunes, as was attested by significant main ef-
fects of degree of familiarity [at IP,
F
1
(1,22)
5
110.69,
p ,
.001, and
F
2
(1,36)
5
9.60,
p ,
.01; at RP,
F
1
(1,22)
5
147.20,
p ,
.001, and
F
2
(1,36)
5
11.78,
p ,
.01]. More-
over, the musicians isolated melodies earlier than did the
nonmusicians, but this effect was reversed at the RP. These
effects of musical expertise were statistically significant
by item analysis only [at IP,
F
1
,
1, and
F
2
(1,36)
5
6.84,
p ,
.05; at RP,
F
1
(1,22)
5
1.64, n.s., and
F
2
(1,36)
5
4.62,
p ,
.05]. No group
3
degree of familiarity interactions
reached significance. In order to examine whether the mu-
sicians were more confident than the nonmusicians about
their responses at the IP, a similar 2 (group)
3
2 (degree
of familiarity) repeated measures ANOVA was performed
considering the mean confidence rating provided at the IP
as a dependent variable (for the mean confidence values,
see Table 3). The musicians were more confident than the
nonmusicians, as was shown by a significant main effect
of group [
F
1
(1,22)
5
7.40,
p ,
.05;
F
2
(1,36)
5
105.86,
p ,
.001]. No main effect of degree of familiarity and no
group
3
degree of familiarity interaction reached signif-
icance.
To examine the contribution of motives to isolating
melodies, the percentage of melodies whose IP coincided
with the boundaries of the first or the subsequent motive
was compared with chance level, computed following the
same formula as that in Experiment 1. For both the musi-
cians and the nonmusicians, the IP coincided with the mo-
tives’ boundaries above chance (52.6% for the musicians,
p ,
.01; 67.6% for the nonmusicians,
p ,
.01, by a bino-
mial test). The RP corresponded to the motives’ bound-
aries above chance only for the musicians (47.4%,
p ,
.01, by a binomial test; 15.8% for the nonmusicians).
In order to assess the effects of task demands on the
melody recognition process, the results obtained in this
experiment were compared with those obtained in Exper-
iment 1. The results showed that fewer notes were neces-
sary to
judge
the familiarity of a musical excerpt than to
sing
its continuation, for both the musicians and the non-
musicians [IP
.
FEP, with
t
mus
(37)
5
9.89,
t
nonmus
(37)
5
5.55,
p
s
,
.001; RP
.
FP, with
t
mus
(37)
5
4.34,
t
nonmus
(37)
5
2.07,
p
s
,
.001].
Discussion
Within five to seven notes (i.e., about 3 sec), listeners
are able to isolate the correct melody. To reach full confi-
dence in their guess, however, they need approximately
one additional note. As in Experiment 1, listeners isolated
and recognized high-familiarity tunes earlier than moderate-
familiarity tunes. The effect of familiarity on recognition
was consistent and robust.
The results were less consistent when the effect of mu-
sical training were considered. The musicians’ RP occurred
earlier than the nonmusicians’ RP, hence replicating the
results of Experiment 1. This difference, again, might
have resulted from the musicians being more confident
than the nonmusicians about their judgments. However,
when IP was considered, the reverse situation occurred.
The nonmusicians isolated the correct melody earlier than
did the musicians. Although the IPs in the two groups
were not far apart, the difference was statistically signifi-
cant. This finding cannot be accounted for by differences
in confidence, since the musicians were more confident in
their production. This apparent paradox is best understood
in the framework of the cohort model, to be discussed
hereafter.
GENERAL DISCUSSION
Time Course of the Music Recognition Process
As is shown in Figure 2, summarizing the results of the
two experiments, both musicians and nonmusicians seem
to proceed in the same manner when recognizing a tune.
At first, the familiar music evokes a general feeling of
knowing (at FEP); then, after a few notes, a match with a
memory representation is found, and a tune is selected
(at IP). Only later do the listeners gain full confidence in
their judgments (at FP followed by RP). These results con-
firm that recognition occurs later than the feeling of
knowing. Much more information is needed to judge a
tune as unfamiliar. In general, nonmusicians seem to re-
quire more notes than do musicians to achieve recogni-
tion. However, nonmusicians isolate a melody more
rapidly than do musicians.
These findings can be easily interpreted using a cohort
model (Marlsen-Wilson, 1987). By analogy to word recog-
nition, the music recognition process can be divided into
two stages: access and selection. In the access stage, an ini-
tial cohort of melodies stored in memory is activated on
the basis of the first notes of the presented melody. Dur-
ing the selection stage, the members of the initial cohort
that do not match the incoming musical information are
dropped from the cohort. At the end of this stage, only one
1026 DALLA BELLA, PERETZ, AND ARONOFF
candidate will reach a threshold level, thus leading to iso-
lation. The model accounts for the fact that more notes are
needed to judge melodies as unfamiliar than to judge them
as familiar. To judge a tune as unfamiliar, the listener has
to acknowledge that the melodies stored in memory do not
correspond to the presented melody. To this end, all the mem-
bers of the initial cohort must be examined. Since none of
the melodies reaches the threshold needed for isolation,
listeners have to infer, by default, that the melody is unfa-
miliar. It is likely that both the search for a candidate
within the cohort and the need for an additional inference
process concur to delay the judgment for unfamiliar music.
Furthermore, the model accounts for the difference be-
tween high-familiarity and moderate-familiarity tunes by
referring to the concept of level of activation (as in Marslen-
Wilson, 1987). In the model of spoken word recognition,
it is assumed that the initial level of activation varies as a
function of frequency (i.e., the more frequent a word is,
the higher its activation). Analogously, high-familiarity
music may present a higher initial level of activation than
does moderate-familiarity music. It follows that less acti-
vation is needed to reach the threshold for isolation, which
is assumed to be the same for all melodies, thus account-
ing for the earlier recognition of high-familiarity tunes.
Finally, the cohort model can account for the differ-
ences due to musical training. Musicians have probably
memorized a larger number of melodies than have non-
musicians, following extensive musical exposure. Accord-
ingly, musicians would access a larger initial cohort of
melodies than would nonmusicians during the access
stage of the recognition process. Since the feeling of
knowing is based partly on the total amount of accessed
information (see Koriat, 1993; Koriat & Levy-Sabot, 2001),
a larger initial cohort size should entail a higher feeling of
knowing, thus leading musicians to reach FEP earlier than
do nonmusicians. Conversely, the larger size of the initial
cohort may hinder the isolation of the correct melody dur-
ing the selection stage, insofar as a larger number of non-
matching candidates has to be examined. This may, in
turn, account for the more precocious IP in nonmusicians.
The other differences occurring between musicians and
nonmusicians are best explained by differences in confi-
dence. Both the FP and the RP require maximal confidence.
Since musicians are generally more confident than non-
musicians when providing a judgment about music, it is
not surprising that they will attain FP and RP earlier than
will nonmusicians.
Variables Mediating the Recognition Process
It is likely that musical motives affect the time course
of the recognition process. Indeed, musicians and nonmu-
sicians are found to isolate the correct melody in accor-
dance with motive boundaries. Moreover, motive bound-
aries seem to play a role in increasing the confidence level
(i.e., at the RP), although only for musicians. Similarly,
motive boundaries affect the familiarity judgments of
nonmusicians at an earlier point (at the FEP) than for mu-
sicians (at the FP). This pattern of results suggests that the
motive structure has an impact at different stages of the
recognition process in musicians and nonmusicians. It
seems that the motive unit facilitates memory access in
nonmusicians and tune selection in musicians. This dif-
ferential impact might reflect the difference in melody co-
hort size, alluded to earlier, that would distinguish the
memory of musicians and nonmusicians.
Further analyses of the data were performed in order to
uncover other bottom-up processes, such as the principles
defined by Narmour (1990; for a review, see Schellenberg,
Figure 2. Summary of the results obtained in Experiments 1 (square symbol) and 2 (triangle sym-
bol) for musicians and nonmusicians. The mean values of each point (FEP, familiarity emergence
point; IP, isolation point; FP, familiarity point; and RP, recognition point) are ordered as a function
of the number of notes. The first four points all refer to judgments provided with familiar melodies,
whereas the last point, as indicated, has been obtained with unfamiliar melodies.
MUSIC RECOGNITION PROCESS 1027
1996), that may affect the time course of the recognition
process. However, these attempts proved unsuccessful. No
particular property could be identified.
CONCLUSIONS
In the present study, we sought to explore the time course
of the music recognition process by means of the gating
paradigm. Altogether, the results reveal that approxi-
mately 3 sec (e.g., at the IP) are required for a melody to
be recognized. Both familiarity and musical training are
found to affect the recognition process. However, the di-
vergence between musicians and nonmusicians does not
seem to arise from qualitatively different recognition sys-
tems. Both musicians and nonmusicians seem to segment
melodies into motives to facilitate recognition. These re-
sults can be easily accounted for by a cohort model. The
differences in initial activation levels of melody represen-
tations in a cohort according to their prior frequency of
experience can account for the observed effects of famil-
iarity. The effects of musical expertise can be attributed to
the larger melody cohort that musicians are likely to have
in memory, as compared with nonmusicians, and to the
higher degree of confidence in their musical skills. This
difference may, in turn, explain why the motive structure
of melodies has an earlier impact on the recognition process
in nonmusicians than in musicians. The motive would
serve as an index to the melody cohort for nonmusicians
and as a selection criterion among alternatives in the
melody cohort for musicians. Hence, the cohort model
represents a good point of departure for modeling the time
course of the music recognition process. However, the
model needs further specification of its inherent proper-
ties to provide a detailed account for the time course of
music recognition. The cohort model was developed orig-
inally to account for spoken word recognition and, more
recently, has been applied to visual word recognition (see
Johnson & Pugh, 1994). Insofar as music and language
present remarkable differences, it is obvious that only the
general properties of the original model (e.g., the multiple
access to the candidates of the cohort) can be transferred
to the music domain. For instance, we found that the seg-
mentation of melody into motives orient and affect the
recognition process. Which other bottom-up processes gov-
ern the recognition process? How do these processes in-
teract with top-down information, derived from the acti-
vation of memory representations? Moreover, at what stage
of the recognition process does this interaction occur?
These and similar questions cannot be answered with the
current general conceptualization of the cohort model and
must await further investigation.
In conclusion, the gating paradigm has proven to be an
adequate tool for studying the music recognition process.
This paradigm should compensate for the lack of a method-
ology well suited to capture the evolution of recognition
over time, thus stimulating more research into the nature
of the information that governs the time course of music
recognition.
REFERENCES
Altmann, G., Wathanasin, S., Birkett, T., & Russell, P. (1992). Ex-
perimenter program. Brighton, U.K.: University of Sussex, Network
Analysis Ltd.
Attneave, F., & Olson, R. K. (1971). Pitch as a medium: A new ap-
proach to psychophysical scaling. American Journal of Psychology,
84, 147-166.
Bartlett, J. C., & Dowling, W. J. (1980). Recognition of transposed
melodies: A key-distance effect in developmental perspective. Jour-
nal of Experimental Psychology: Human Perception & Performance,
6, 501-515.
Berthier, J. E. (1979). 1000 chants [1,000 songs]. Paris: Presses de
l’Ile-de-France.
Bigand, E., & Parncutt, R. (1999). Perceiving musical tension in long
chord sequences. Psychological Research, 62, 237-254.
Brown, A. S. (1991). A review of the tip-of-the-tongue experience. Psy-
chological Bulletin, 109, 204-223.
Cohen, J., MacWhinney, B., Flatt, M., & Provost, J. (1993). PsyScope:
An interactive graphic system for designing and controlling experi-
ments in the psychology laboratory using Macintosh computers. Be-
havior Research Methods, Instruments, & Computers, 25, 257-271.
Cotton, S., & Grosjean, F. (1984). The gating paradigm: A compari-
son of successive and individual presentation formats. Perception &
Psychophysics, 35, 41-48.
Deliège, I. (1987). Grouping conditions in listening to music: An ap-
proach to Lerdahl and Jackendoff`s grouping preference rules. Music
Perception, 4, 325-360.
Deutsch, D. (1972). Octave generalization and tune recognition. Per-
ception & Psychophysics, 11, 411-412.
Dowling, W. J. (1978). Scale and contour: Two components of a theory
of memory for melodies. Psychological Review, 85, 341-354.
Dowling, W. J. (1984). Musical experience and tonal scales in the recog-
nition of octave-scrambled melodies. Psychomusicology, 4, 13-32.
Dowling, W. J., & Fujitani, D. S. (1971). Contour, interval, and pitch
recognition in memory for melodies. Journal of the Acoustical Soci-
ety of America, 49, 524-531.
Dowling, W. J., & Harwood, D. L. (1986). Music cognition. Orlando,
FL: Academic Press.
Dowling, W. J., & Hollombe, A. W. (1977). The perception of melodies
distorted by splitting into several octaves: Effects of increasing prox-
imity and melodic contour. Perception & Psychophysics, 21, 60-64.
Frauenfelder, U. H. (1996). Computational models of spoken word
recognition. In T. Dijkstra & K. de Smedt (Eds.), Computational psy-
cholinguistics: AI and connectionist models of human language pro-
cessing. London: Taylor & Francis.
Grosjean, F. (1980). Spoken word recognition processes and the gating
paradigm. Perception & Psychophysics, 28, 267-283.
Grosjean, F. (1996). Gating. Language & Cognitive Processes, 11, 597-
604.
Halpern, A. R., Bartlett, J. C., & Dowling, W. J. (1995). Aging and
experience in the recognition of musical transpositions. Psychology
& Aging, 10, 325-342.
Handel, S. (1989). Listening: An introduction to the perception of au-
ditory events. Cambridge, MA: MIT Press.
Hébert, S., & Peretz, I. (1997). Recognition of music in long-term
memory: Are melodic and temporal patterns equal partners? Memory
& Cognition, 25, 518-533.
Idson, W. L., & Massaro, D. W. (1978). A bidimensional model of
pitch in the recognition of melodies. Perception & Psychophysics, 24,
551-565.
Johnson, N. F., & Pugh, K. R. (1994). A cohort model of visual word
recognition. Cognitive Psychology, 26, 240-346.
Kallman, H. J., & Massaro, D. W. (1979). Tone chroma is functional
in melody recognition. Perception & Psychophysics, 26, 32-36.
Koriat, A. (1993). How do we know that we know? The accessibility
model of the feeling of knowing. Psychological Review, 100, 609-639.
Koriat, A., & Levy-Sadot, R. (2001). The combined contributions of
the cue-familiarity and accessibility heuristics to feelings of knowing.
Journal of Experimental Psychology: Learning, Memory, & Cogni-
tion, 27, 34-53.
1028 DALLA BELLA, PERETZ, AND ARONOFF
Krumhansl, C. L. (1990). Cognitive foundations of musical pitch. Ox-
ford: Oxford University Press.
Krumhansl, C. L., & Shepard, R. N. (1979). Quantification of the hier-
archy of tonal functions within a diatonic context. Journal of Experi-
mental Psychology: Human Perception & Performance, 5, 579-594.
Lerdahl, F., & Jackendoff, R. S. (1983). A generative theory of tonal
music. Cambridge, MA: MIT Press.
Marslen-Wilson, W. D. (1987). Functional parallelism in spoken
word-recognition. Cognition, 25, 71-102.
Marslen-Wilson, W. D., & Welsh, A. (1978). Processing interactions
during word-recognition in continuous speech. Cognitive Psychology,
10, 29-63.
Narmour, E. (1990). The analysis and cognition of basic melodic struc-
tures: The implication-realization model. Chicago: University of
Chicago Press.
Palmer, C., & Krumhansl, C. L. (1987a). Independent temporal and
pitch structures in determination of musical phrases. Journal of Exper-
imental Psychology : Human Perception & Performance, 13, 116-126.
Palmer, C., & Krumhansl, C. L. (1987b). Pitch and temporal contribu-
tions to musical phrase perception: Effects of harmony, performance
timing, and familiarity. Perception & Psychophysics, 41, 505-518.
Peretz, I. (1989). Clustering in music: An appraisal of task factors. In-
ternational Journal of Psychology, 24, 157-178.
Peretz, I., Babaï, M., Lussier, I., Hébert, S., & Gagnon, L. (1995).
Corpus d’extraits musicaux: Indices relatifs à la familiarité, à l’âge
d’acquisition et aux évocations verbales. Canadian Journal of Exper-
imental Psychology, 49, 211-239.
Peretz, I., Gagnon, L., & Bouchard, B. (1998). Music and emotion:
Perceptual determinants, immediacy, and isolation after brain dam-
age. Cognition, 68, 111-141.
Peretz, I., Gaudreau, D., & Bonnel, A.-M. (1998). Exposure effects on
music preference and recognition. Memory & Cognition, 26, 884-902.
Peynircio Ïglu, Z. F., Tekcan, A. I., Wagner, J. L., Baxter, T. L., &
Shaffer, S. D. (1998). Name or hum that tune: Feeling of knowing
for music. Memory & Cognition, 26, 1131-1137.
Protopapas, A. (1999). Connectionist modeling of speech perception.
Psychological Bulletin, 125, 410-436.
Schellenberg, E. G. (1996). Expectancy in melody: Tests of the
implication-realization model. Cognition, 58, 75-125.
Schellenberg, E. G., Iverson, P., & McKinnon, M. C. (1999). Name
that tune: Identifying popular recordings from brief excerpts. Psy-
chonomic Bulletin & Review, 6, 641-646.
Smith, J. D. (1997). The place of musical novices in music science.
Music Perception, 14, 227-262.
Smith, J. D., Nelson, D. G., Grohskopf, L. A., & Appleton, T.
(1994). What child is this? What interval was that? Familiar tunes and
music perception in novice listeners. Cognition, 52, 23-54.
Tyler, L. K. (1984). The structure of the initial cohort: Evidence from
gating. Perception & Psychophysics, 36, 417-427.
Tyler, L. K., & Wessels, J. (1985). Is gating an on-line task? Evidence
from naming latency data. Perception & Psychophysics, 38, 217-222.
Walley, A. C., Michela, V. L., & Wood, D. R. (1995). The gating par-
adigm: Effects of presentation format on spoken word recognition by
children and adults. Perception & Psychophysics, 57, 343-351.
White, B. W. (1960). Recognition of distorted melodies. American Jour-
nal of Psychology, 73, 100-107.
NOTES
1. New terms are introduced that better reflect the isolation and the
recognition points in the development of a feeling of knowing in a fa-
miliarity decision task.
2. In a few cases, although the participants correctly produced three
notes beyond the presented melody, it became clear later that they did not
recognize the melody but were, rather, improvising on the basis of their
musical knowledge. In similar cases, confidence ratings usually re-
mained very low. Recognition was verified by questioning the partici-
pant at the end of the trial, in order to decide whether to include the re-
sponse in the analysis.
(Manuscript received July 25, 2001;
revision accepted for publication February 9, 2003.)
