Professional Documents
Culture Documents
CITATIONS
DOWNLOADS
VIEWS
13
29
18
1 AUTHOR:
Lola L Cuddy
Queen's University
90 PUBLICATIONS 1,551
CITATIONS
SEE PROFILE
107
Our notion of the prototypic nature of the major triad is consistent with a
description provided by Krumhansl (1990): "The basic idea is that within categories certain members are normative, unique, self-consistent, simple, typical, or the best exemplars of the domain.... They are reference points to which
other category members are compared.... [Certain patterns] seem somehow
'better'than others, because they are simpler, more regular, or more symmetric" (Krumhansl, 1990, p. 17). Moreover, the major triad appears to fulfill a
criterion for the tonal-harmonic scheme described by Jones (e.g., 1981, 1982).
as representative of an ideal prototype. Jones (1982) notes that the "tonalharmonic scheme reflects a listener's sense of the stable harmonic context and,
in particular, of the tonal center of a piece" (p. 2). The priority assigned to the
major triad is consistent with empirical evidence (e.g., Krumhansl & Kessler,
1982; Roberts & Shaw, 1984), with psychoacoustical theories of sensory consonance (following Helmholtz, 1863/1954), and with music theory in the Schenkerian
tradition (Schenker, 1906/1954).
In recent commentaries, however, Butler (1989, 1990) has contested the
Cuddy and Badertscher (1987) results on several grounds. First, he argues that
the probe-tone technique they used to assess tonal structure is unreliable. He
argues that it is not meaningful to identify probe-tone judgments of pattern
completion with perceived tonal structure. According to Butler (1989, 1990),
the instructions for the typical probe-tone procedure are so vague that the listener is free to set any one of a number of response criteria. Butler (1989)
suggests, however, that probe-tone judgments might reliably demonstrate effects such as primacy and recency effects traditionally associated with free
recall.
Second, Butler (1989, 1990) has accompanied his criticisms of the probetone technique with criticisms of current approaches to the study of the perception of tonal structure. In their place, Butler offers an alternative account based
on logical analysis of the interval content of music patterns, the "recognition of
critical intervallic relationships as they unfold throughout the musical performance" (1989, p. 233).
Butler (1989) cites, as compatible with his own approach, Browne's (1981)
analysis of the diatonic pitch set in terms of its interval-class content. Browne
(1981) points out that within the diatonic pitch set of seven pitch-classes, there
are 21 interval classes. There are two minor seconds (or major sevenths), five
major seconds (or minor sevenths), four minor thirds (or major sixths), three
major thirds (or minor sixths), six perfect fourths (or perfect fifths) and one
tritone (augmented fourth or diminished fifth). The frequency of occurrence
with which each interval class occurs in the diatonic set may be summarized by
the vector <254361>.
Next, Browne (1981) points out several properties of the vector, the most
important of which is the principle of unique multiplicity. "The diatonic set
contains a full range of intervallic ubiquity. The six interval-classes occur from
one to six times, and each of them a unique number of times. This constitutes a
full spread of possibilities from 'rarity' to 'common-ness'a maximum possible hierarchization" (p. 6). Note that "rarity" and "common-ness" in this analysis
refer to frequency of occurrence in the interval vector, not to frequency of
108
109
The interval content of the minor triad logically implicates three diatonic keys;
the interval content of the fifths pattern logically implicates five diatonic keys.
From the perspective of logical analysis of interval content, neither pattern
should 'elicit girtunambiguous sense of tonal center and accompanying structutalrKiera#ehymnd, therefore, neither should be as strong an indicator of tonal
sff^iur&tothe diminished triad.
(o Ihdre is, however, evidence suggestive that the minor triad pattern and the
fiflfc pattern might yield a reasonably strong sense of tonal structure. With
respect to minor triads, it has been shown with the probe-tone technique that
harmonic minor triads strongly instantiate the tonality of the root of the triad
(Krumhansl & Kessler, 1982); minor triads suggested a single key to listeners
to a greater extent than did diminished triads, which tended to be interpreted in
terms of both the major and minor keys in which the chord functioned harmonically. Cohen (1991) reported a study in which listeners were asked to listen to
excerpts from the Bach Preludes and to sing the scale of the key suggested by
the excerpt. For each of the six Bach Preludes in the minor key, the four opening
notes outlining the minor triad signaled the minor tonality for listeners.
With respect to fifths patterns, the evidence is indirect. Cross, Howell, and
West (1985) reported experiments in which listeners heard three-note patterns,
including fifths patterns (pitch-class sets of the type <5, 0, 7>), and rated the
goodness-of-fit of a single tone following each pattern. The single tone was
either a member of one of the scales logically implicated by the pattern or was
an out-of-scale note. For the fifths patterns, listeners readily rejected, as "wrong,"
the note that did not fit within any of the five scales implicated by the patterns.
Moreover, the authors found that fifths patterns were more effective contexts
for producing rejections of nonscale notes than were three-note patterns containing the rare interval of the tritone (patterns implicative of only one diatonic
scale). Cross, Howell, and West (1985) concluded that "the lower the logical
scale specificity, the stronger the scalar schema" (p. 137). The authors are
careful to explain that scale identity may not be the same as key identity. However, to the extent that scale identity contributes to perceived tonal structure,
the implications of their conclusion appear to be opposite to Butler's (1989,
1990) proposals.
A final point deals with the order in which the notes of the pattern were
presented. In the present experiments, all patterns were realized either as three
notes with ascending pitch contour or three notes with reversing pitch direction, i.e., a change in pitch contour. The reversing pitch contour for the diminished triad exemplified an ordering that should, according to Butler (1989), be
especially conducive to the recovery of a tonal center. On the other hand, the
application of the Gestalt principle of "good continuation" to music patterns
(Deutsch, 1982, p. 101) leads to the prediction that tonal structure will be more
readily detected for patterns with a unidirectional pitch contour than for patterns with more complex contours. Cuddy and Cohen (1976) found that major
triad patterns with unidirectional pitch contour were more easily recognized
under transposition than were major triad patterns with a contour that changed
110
direction. The main aspects of the method for the two experiments will be
described next, followed by the specific description of each.
General Method
Listeners
Listeners were volunteers from the university community who had attained
at least Grade VIII Royal Conservatory level of performance in voice or an
instrument. This level is comparable to the practical component of the Grade 12
music curriculum in Ontario schools. A typical listener had received music
training at the level of a junior in an undergraduate music program. No listener
was a professional musician. Listeners ranged in age from 16 to 30; the ratio of
females to males was 2:1. They were paid $3.00 per session for participation.
Patterns
Test patterns were eight diatonic patterns of three successive notes spanning six or seven semitones. For the first four patterns, the pitch contour ascended; for the remaining four, the pitch contour reversed direction. Within
each contour (ascending and reversing), there were four patternsthe major
triad, a pattern of fifths (as described above), the minor triad, and the diminished triad. The patterns with ascending contour are exemplified by C4E4G4, C4
F4G4, D 4 F 4 A4, and B3 D 4 F 4 , respectively. The patterns with reversing contour
are exemplified by E4G4C4, G4C4F4, A4 D 4 F 4 , and D4F4B3, respectively.3 In the
experiments, the patterns were transposed to different, randomly selected, frequency locations within the overall range B3 to D5.
Fourteen practice patterns were also constructed. They were three-note
diatonic patterns within the range of an octave, and all were different from the
test patterns.
Apparatus and General Procedure
Sine-tones for each pattern were produced by a DMX-1000 signal processor under control of a LSI 11/23 host computer. The sampling rate was 19.3
kHz. The duration of each tone in each pattern was .33 s with 25 ms rise time and
fall time. Frequency values for the tones were determined according to the
system of equal temperament, with A4 = 440 Hz. The amplitude of each tone
was set according to the Fletcher-Munson loudness contours. The overall level
was adjusted to that judged comfortable by the listener, about 65 dB SPL.
In the experiments, the experimental conditions were randomly ordered
across trials, and the entire set of trials presented as a single block. The order of
trials, and the frequency location of the pattern presented on each trial, were
randomized independently for each listener. The patterns were delivered through
Sennheiser HD-424 headphones to the listener seated in a sound-proof booth.
Responses were entered on a Zenith Z19 console located in the booth and stored
in the host computer for analysis. Trials were self-paced; after responding to a
trial, the listener pressed the "enter" key on the terminal to initiate the next trial.
In both experiments, listeners were told that there were no "right" or "wrong"
answers to the tests. They were asked to respond in terms of their own auditory
Cuddy
111
"feelings or impressions." Therefore, no feedback was provided on either practice or test trials.
Experiment I
In the first experiment, listeners were asked to judge the suitability of each
of the 12 notes of the chromatic scale as a tonic or key center for each of the test
patterns.
Method
Thirteen listeners were tested. On each trial, a test pattern was presented
and was followed, after a delay of Is, by a 1-s presentation of a single tone
called a key-probe. There were 12 key-probes for each test pattern. Each keyprobe was coded in terms of pitch-class distance, from 0 to 11, from the first
note of the pattern. Each pitch-class distance was tested once for each pattern.
The timbre of the key-probes was that of a "Shepard tone" (Shepard, 1964;
see also Cuddy & Badertscher, 1987), a complex of octave equivalents for
which the amplitude envelope is shaped to obliterate a clear sense of pitch
height. Listeners were asked to rate the suitability of the key-probe as a tonic or
key-note for the pattern on a scale of " 1 " to "6," where " 1 " represented "very
good " and " 6 " represented "very poor." They were told that the timbre of the
key-probe would be different from that of the pattern, the purpose being to
remind them not to rate the key-probe for melodic continuity or completion but
as an abstract key-center for the pattern.
Nine practice patterns were each paired with each of five randomly selected key-probes. Practice trials consisted of the 45 pairings of patterns and
probes presented in random order for ratings by listeners. Practice trials were
followed by test trials in which each of the test patterns was paired with each of
12 key-probes. In the test trials, two patterns and their pairings with the 12 keyprobes were replicated (major triad pattern with ascending contour and fifths
pattern with reversing contour). The replication trials were randomly interleaved with all other trials and allowed an assessment of internal reliability.
Five additional practice patterns were embedded among the test patterns in
order to encourage the listener to expect a variety of diatonic patterns. The
order of pairings of patterns and probes, and location of the pattern within the
frequency range, were independently randomized across trials for each listener.
Each session consisted of 45 trials in the practice phase followed by 180
trials in the test phase and lasted about one hour.
Results and Discussion
For each test pattern, a set of mean ratings across the 13 listeners was
obtained for each of the 12 pitch-classes of key-probes. This set of ratings will
be called the key-probe profile for the pattern.
The eight panels of Figure 1 show the set of mean ratings obtained for each
of the eight test patterns. For purposes of illustration, the various transpositions
of each pattern have been collapsed to a single frequency locationthe note112
C 4 E 4 G4
E4 G4 C4
2-I
34
5H
Major Triad Ascending
I 1I I I I I I I I I I
I I I I I I I I I I I I
DbAbEbBb F C G D A E B F #
DbAb EbBb F C G D A E B F #
CFG
Q4C4F4
y\.
5-
i i i i i i i i i i i i
C Eb G
<
45"
kA^.
i i i i i i i i i i i i
DbAb EbBb F C G D A E B F I
DbAb EbBb F C G D A E B F #
4
G 4 CEb 4
23"
5-
i i i i i i i i i i i i
I I I I I I I I I I I
DbAbEbBb F C G D A E B F #
DbAbEbBb F C G D A E B F #
eEb 4 Gb 4
Eb4Gb4C4
WW-
4-
4-1
5"
Diminshed Triad Ascending
I I I I I I I I I I I I
i i i i i i i i i i i i
DbAbEbBb F C G D A E B F #
DbAbEbBb F C G D A E B F #
113
names exemplified in each panel. The vertical axis of each panel is mean rating.
The horizontal axis is key-probe ordered according to the cycle of fifths. The
note-names for the key-probes are assigned with respect to the exemplar of
each pattern, described further below. Exemplars for the left- and right-hand
panels are paired so that in each row key-probe profiles refer to the same pitch
content. Two panels, representing the profiles for the major triad pattern with
ascending contour and the fifths pattern with reversing contour, include the
data for replicated patterns.
The left-hand column of Figure 1 shows the mean ratings for patterns with
ascending contours. Each pattern is illustrated by an exemplar that begins on
C4, and the names of the key-probes on the horizontal axis correspond to the
note-names of the particular exemplar. The key-probe C corresponds to C4 in
the exemplars and thus corresponds to the first note of the patterns in the lefthand column. For example, for the major triad pattern with ascending contour,
both replications, the key-probe assigned the highest rating for the exemplar
C4 E4 G4 was the key-probe C. Across the transpositions of the pattern, the keyprobe assigned the highest rating for the major triad pattern with ascending
contour was the pitch class of the first note (first serial position).
The right-hand column shows the data for patterns with reversing contours.
The exemplar for each panel in the right-hand column is the reversal of the
exemplar in the corresponding left-hand column. The key-probe C in each panel
in the right-hand column corresponds to C4 in the exemplar, and, as well, to the
serial position at which C4 occurred in the exemplar. Thus it can be seen, for
example, that for the major triad pattern with reversing contour, the key-probe
assigned the highest rating for the exemplar E4G4C4 was the key-probe C. Across
the transpositions of the pattern, the key-probe assigned the highest rating for
the major triad pattern with reversing contour was the pitch class of the last note
of the pattern (third serial position).
Inspection of Figure 1 reveals that for major triad patterns, fifths patterns,
and minor triad patterns, ratings for key-probes on either side of the key-probe
given the highest rating tended to decrease in a fairly regular fashion. Ratings
for key-probes appeared systematically related to distance on the cycle of fifths;
ratings tended to decrease as distance of the key-probe increased, in either
direction, from the key-probe given the highest mean rating. For the diminished
triad patterns (bottom panels of Figure 1), ratings did not appear to be systematically related to distance on the cycle of fifths.
The ratings collected for each test pattern were subjected to ANOVA (onefactor repeated measures, conducted separately for each pattern). Table 1 shows
the results of the ANOVA. The rows in Table 1 represent the different patterns;
the columns represent the two different orders of presentation. The upper entry
in each cell is the value of the F-ratio, based on 11 and 132 df. The middle entry
is 82, the component of the numerator of the F-ratio that is an estimate of the
variability of the treatment (population) means (Myers, 1979, p.84). The lower
entry is the estimate of error variance. The second entries on each line for the
major triad pattern with ascending contour and fifths pattern with reversing
contour are the results for the replication of the pattern represented by the cell.
114
PsychomusicologyFall 1991
The correlations between profiles for replicated patterns were highly significant (for the major triad pattern with ascending contour, r (10) = .87; for the
fifths pattern with reversing contour, r (10) = .90, both/? <.001).
The entries in Table 1 show significant differences among key-probe ratings for all test patterns except the diminished triad patterns. Inspection of the
table from the top to the bottom row reveals a gradual decrease in 92. The
systematic differences among key-probe ratings were greatest for the major
triad pattern with ascending pitch contour. The results for this pattern were
followed, in order, by the major triad pattern with reversing contour and the
fifths patterns, then followed by the minor and diminished triad patterns.4 These
Table 1
Summary statistics for key-probe profiles for eight stimulus patterns (Experiment 1). The upper entry in each cell is the F-ratio for the profile data. The
middle entry is 92, an estimate of variance among the mean ratings. The lower
entry is the MS error
Pitch contour
Pattern
Major triad
Fifths
Minor triad
Diminished triad
Ascending
6.57****, 6.54
Reversing
5 g7****
.82, .82
.55
1.91, 1.92
1.47
479****
5 44**** 5.89****
.45
.50, .57
1.54
1.45,1.51
3.30***
2.64**
.28
.20
1.67
1.58
1.37
1.01
.05
1.71
.002
1.77
115
Table 2
Correlations between key-probe profiles and standardized profiles around the
cycle of fifths (Experiment 1). The correlation for the best-fitting
standardized
profile is underlined
Pattern: Major triad ascending
Standardized major key profile
Db
Ab
Eb
Bb
F
C
G
D
-.13
.04
.06
.30
.79
.80
.20 -.24
-.23 -.20
.26
.31
.65
.79
.27 -.33
Pattern: Major triad reversing
Standardized major key profile
Db
Ab
Eb
Bb
F
C
G
D
-.11
.13 -.03
.07
.69
TL
.11 -.22
Pattern: Fifths pattern ascending
Standardized major key profile
Db
Ab
Eb
Bb
F
C
G
D
-.09
.14 -.36
.60
M
.63
.12 -.33
Pattern: Fifths pattern reversing
Standardized major key profile
Bb
F
C
G
D
.20
.21
.32
.74
J9
.26 -.15 -.39
.34
.32
.27
.54
J3
.24 -.37 -.48
Standardized minor key profile
db
ab
eb
bb
f
c
g
d
,51 -.35 -.09
.47
J5
.26
.32
.48
-.26 -.29 -.08
.43
.18
.09
.34
.85
Pattern: Minor triad ascending
Standardized minor key profile
db
ab
eb
bb
f
c
g
d
-.08
.07 -.21
.06
.49
J%
.03 -.30
Pattern: Minor triad reversing
Standardized major key profile
Db
Ab
Eb
Bb
F
C
G
D
.38
.84
,86
.37
.12 -.01 -.32 -.68
Pattern: Diminished triad ascending
Standardized minor key profile
db
ab
eb
bb
f
c
g
d
.01
.22
.21 -.09
.09
.52 -.40 -.55
Pattern: Diminished triad reversing
Standardized major key profile
Db
Ab
Eb
Bb
F
C
G
D
.60
.63
.09 -.14 -.09 -.16 -33
-.35
C4 E4 G4
Exemplar:
A
-.43
-.52
E
-.54
-.42
B
-.49
-.41
E4G4C4
Exemplar:
A
-.23
E
-.31
B
,40
E
-.72
B
-,58
Exemplar:
F#
-.46
C4F4G4
Exemplar::
A
-.63
F#
-.34
,56
F#
-.34
G4 C4 F4
A
-.60
-.43
E
-.64
-.45
B
-,52
-,50
F#
-.22
-.21
a
-.05
.06
e
-.38
-.36
b
-,41
-.66
f#
-.48
-.31
Exemplar: C4 Eb4 G4
a
.06
e
.08
b
-.30
f#
-.53
Exemplar: G 4 C 4 Eb 4
A
.69
E
-.53
Exemplar:
a
.06
e
-.08
Exemplar:
A
.27
E
-.26
B
-.29
F#
-.06
C 4 Eb 4 Gb 4
b
.03
f#
-.02
Eb 4 Gb 4 C 4
B
.02
F#
-.25
116
The next analysis examined the relationship between the key-probe profiles reported in Figure 1 and the standardized profiles for major and minor keys
derived by Krumhansl and Kessler (1982). The standardized profiles are a set of
ratings obtained for the major and for the minor key by averaging the probetone profiles for several key-defining contextschords and chord cadences.
Krumhansl and Kessler (1982) demonstrated that an abstract representation of
key relationships may be recovered from the standardized profiles and expressed
as two circular configurationsthe cycle of fifths and the cycle of thirds, respectively.
A best-fitting standardized profile was sought for each of the key-probe
profiles in Figure 1. The ratings for each key-probe profile were correlated with
the ratings for the standardized profiles for each of the 12 major and 12 minor
keys. The highest positive correlation was determined, and the key of the standardized profile yielding the highest correlation, if significant, was selected as
the best-fit. Because of the large number of correlations involved, the criterion
for significance was set at .005, higher than the conventional level.
Table 2 shows the set of correlations for the entire cycle of keys around the
best-fitting key. The correlation for the best-fitting key is underlined. The assignment of key-names in Table 2 corresponds to the exemplar of each pattern
given in the table and also in each panel of Figure 1.
Two sets of correlations are provided for the major triad pattern with ascending contour and fifths pattern with reversing contour, one set for each
replication of the pattern. In addition, for the fifths pattern with reversing contour exemplified as G4C4 F4, F major was the best-fitting key for one replication, f minor was the best-fitting key for the other. Both cycles, therefore, are
given in Table 2. It can be seen that both F major and f minor are possible
selections for best-fitting key for this exemplar.
No best-fitting key was found for either diminished triad pattern according
to the above criterion for significance, probably because, as noted above, differences among key-probe ratings for these patterns were not reliable. The
cycles containing the nearest fitting keys for the diminished triad patterns are
included in the table. It may be noted that the key of Db, the diatonic key
logically implicated by the pattern Eb4 Gb4 C4, is weakly implicated in the
correlational data.
For major triad patterns, fifths patterns, and minor triad patterns, the variance shared between the key-probe profile and a standardized profile was greater
than the residual variance. There was, however, a systematic effect in the residual variability of the key-probe profiles that lends itself to musical interpretation. For profiles best fit by a major key, the key-probe corresponding to the
subdominant was rated higher than the key-probe corresponding to the mediant.
The direction of this difference is opposite to the direction of the difference in
the standardized profile for the major key. The finding may mean that keyprobe ratings are slightly drawn along the cycle of fifths toward the key in
which the tonic note performs a dominant function. (This result is also evident
in Table 2.) The profile for the major triad chord contains a similar tendency
(Krumhansl & Kessler, 1982).
Cuddy
117
Fifths
Minor triad
Diminished triad
Ascending
Reversing
.47, .58
.45
.60, .65
.49
.89
.68, .63
.81
.72, .68
.58
.62
.68
.62
.67
.32
.68
.41
Table 3 reveals a slight advantage of the weighted version of the patterncontent model over the unweighted version for eight of the ten pairs of correlations. (Correlations were higher, on the average, by about .05). Even so, the
weighted version of the pattern-content model did not suggest a more promising account of the key-probe profiles than an account based on the standardized
profiles. Correlations for the weighted version of the pattern-content model
were lower than those obtained for the best-fitting standardized profile in all
cases except the diminished triad pattern with ascending contour. Moreover,
for the weighted version of the pattern-content model, only two correlations
were significant beyond the .005 level.
A comparison between the underlined correlations in Table 2 and the correlations in Table 3 suggests that tonal hierarchy predictors usually accounted
for a greater amount of the variability in key-probe ratings than did predictors
based solely on pattern-content. This suggestion was followed by conducting
hierarchical regression on the entire set of key-probe profiles (10 patterns x 12
probes). The weighted version of the pattern-content model was entered first
into the regression and accounted for 36% of the variance (p < .001). The bestfitting standardized profile was entered second and accounted for an additional
20% of the variance (p < .001). Thus, there was systematic evidence that the
key-probe ratings were not merely indicative of pattern-content, but also contained information about tonal structure.5
The results of the first experiment suggest that the major triad pattern with
ascending contour occupied a privileged position among the patterns tested.
This pattern was associated with the greatest amount of differentiation among
the key-probe ratings. It was followed, in amount of differentiation among keyprobe ratings, by the major triad with reversing contour, and the fifths patterns.
Much weaker differentiation among key-probe ratings occurred for minor and
diminished triad patterns. This ordering did not correspond to the ordering of
the number of keys logically implicated by the patterns.
Experiment II
In the first experiment, it was found that key-probe ratings for major triad
patterns, fifths patterns, and minor triad patterns implicated tonal centers and
tonal hierarchies. The major triad pattern with ascending contour was a particularly effective context in that it yielded the strongest differentiation among
key-probes and the clearest indication of tonal structure. In the second experiment, converging evidence for this finding was sought. Rather than assessing
structure through the analysis of key-probe ratings, the second experiment collected direct judgments of structural quality or "goodness" for the patterns
(Cuddy, Cohen, & Mewhort, 1981; Garner, 1974), and judgments of major/
minor quality.
Method
There were two successive parts to the experiment; 18 listeners participated in both parts. Listeners had not participated in the first experiment. In the
first part, listeners were asked to rate the perceived structural goodness of each
Cuddy
119
Fifths
Minor triad
Diminished triad
Ascending
Reversing
1.83,2.17
3.00
.23, .27
.26
2.83
2.56,2.61
.27
.26, .25
2.61
3.78
.27
.26
3.22
4.06
.28
.27
from the set of practice patterns. As noted above, the order of presentation of
the patterns, and the frequency location within the range, were both independently randomized for each listener.
Each session, therefore, consisted of 9 trials in the practice phase and 15
trials in the test phase for ratings of structural goodness, then 9 trials in the
practice phase and 15 trials in the test phase for ratings of major/minor quality.
The entire session lasted less than half-an-hour.
Results and Discussion
Table 4 shows the mean rating of structural goodness for each of the eight
test patterns (upper entry in each cell) and the standard error of the mean (lower
entry in each cell). The format of Table 4 is the same as that of Table 1. Inspection of Table 4 yields the following observations, supported by tests of orthogonal contrasts within the ANOVA.
120
Major triad patterns and fifths patterns received higher ratings of structural
goodness than minor triad and diminished triad patterns, F (1,17) = 19.59, p <
.001. Differences within the set of means for major triad patterns and fifths
patterns depended on pitch contour. For patterns with ascending pitch contours, the major triad pattern received higher ratings than the fifths pattern, F
(1, 17) = 6.54, p < .02; for reversing contours, the means were reversed, but
were not significantly different, F (1, 17) = 1.22,p > .20.
Differences within the set of means for minor and diminished triad patterns appear to favor higher ratings for minor triads than for diminished triads,
but the differences were not significant (for ascending contours, F (1, 17) =
2.37, p > .10; for reversing contours, F (1, 17) = .60 ns). For both minor and
Table 5
Ratings of majorI minor quality for eight stimulus patterns (Experiment 2). The
upper entry in each cell is mean rating; the lower entry is standard error of the
mean. On the rating scale, "1" represented "very strong major quality" and
"6" represented "very strong minor quality"
Pitch contour
Pattern
Major triad
Fifths
Minor triad
Diminished triad
Ascending
Reversing
1.39, 1.39
2.83
21, .22
.32
2.89
2.50,2.17
.29
.29, .26
5.33
4.83
.25
.25
5.33
5.06
.22
.24
diminished triad patterns, patterns with ascending pitch contour received higher
ratings than patterns with reversing contours, F (1, 17) = 15.30, p < .001. The
two means collected for each replicated pattern were similar (for the major triad
patterns with ascending contour, F (1,17) = 3.40, p > .05; for the fifths patterns
with reversing contour, F (1,17) = .06 ns). The set of standard errors was stable;
moreover, standard errors were not systematically related to mean ratings.
Table 5 shows the results for the ratings of major/minor quality. The format of Table 5 is identical to that of Table 1. Inspection of Table 5 yields the
following observations, supported by tests of orthogonal contrasts within the
ANOVA.
Major triad patterns and fifths patterns were judged to convey a strong
sense of major key; minor and diminished triad patterns were judged to convey
a strong sense of minor key. The difference in ratings was highly significant, F
(1, 17) = 90.18,/? < .0001. The differences between major triad and fifths patCuddy
121
terns were similar in direction to those found for ratings of structural goodness.
For patterns with ascending pitch contours, the major triad received higher
ratings of major quality than the fifths pattern, F (1, 17) = 20.55, p < .001; for
reversing contours, the means were reversed, but were not significantly different, F (1,17) = 1.78,/? > .15).
The slight differences between means for minor and diminished triad patterns were not significant (all contrastsp > .15). The two mean ratings collected
for each replicated pattern were similar (means for the major triad patterns with
ascending contour were identical; means for the fifths patterns with reversing:
contour were not significantly different; F (1, 17) = 1.06, p> .30). The set of
standard errors was stable; moreover* standard errors were not systematically
related to mean ratings.
In both parts of the experiment; the major triad pattern with ascending
contour was assigned a privileged position. Of all patterns tested, this pattern
was associated with the highest degree of structural quality in the first part of
the experiment and the strongest conveyor of major quality in the second part
(both overall, and in comparison to other major and fifths patterns).
Differences between the two parts of the experiment may also be noted. In
particular, the perceptual structure of minor and diminished triad patterns was
judged to be weak to moderate but the minor quality of these patterns was
judged to be clear and strong.
The reliability of the replications; and the stability of the standard errors in
both parts of the experiment, are indicators of the consistency of listeners^
response strategies for both tasks. Had minor and diminished triad patterns
elicited a greater variety of individual strategies than major triad and fifths
patterns, significantly greater between-subject variability would be expected
for the former patterns than fbr the latten
General Discussion
The initial motivation for the present experiments was to search for converging evidence to support the distinction drawn between tonal structures
conveyed by the major and the diminished triad (Cuddy & Badertscher, 1987).
Cuddy and Badertscher's (1987) data showed that a major triad pattern was
more effective in recovering the tonal hierarchy than a diminished triad pattern.
In the first experiment reported here, key-probe profiles for the major triad
patterns revealed a clear sense of key center, and contained information about
the tonal hierarchy of the key. Key-probe profiles for the diminished triad patterns did not yield reliable evidence of a sense of tonal structure. In the second
experiment, both structural quality and the sense of major modality were rated
significantly higher for the major triad patterns than for the diminished triad
patterns. These differences between the major triad and diminished triad patterns are convergent with the differences reported by Cuddy and Badertscher
(1987). Altering the order of the notes in the triads, in the present experiments,
did not reverse these findings, but, rather, preserved the direction of the differences.
122
123
The point here, however, is that not all keys are implicated to the same degree.
Rather, the implication of key relationships involves a hierarchical structure.
Multiple tonalities or key-centers may exist simultaneously, but with perceptual priority assigned according to hierarchical principles of organization.
A number of issues remain to be addressed. Questions remain concerning
the cues used to detect tonal centers for patterns deviating from the prototype,
and the role of music experience in the implementation of these cues. Listeners
may have multiple strategies availablefor example, strategies to listen for
specific interval distributions, interval sequences, or implied harmonic progressions. It is promising to note, in answer to Butler's (19S9,1990) criticism
that strategies to rate probe tones are unreliable, that listeners in our experiments produced reliable responses to a given test pattern. Thus, although listeners may flexibly adapt strategies in order to deal with a specific pattern,fhey do
so consistently.
Questions remain concerning the effects of serial order of the tones ^f
melodic patterns. The data suggest that patterns with ascending pitch contour
are perceived as more structurally coherent than patterns with reversing contour (with the possible exception of fifths patterns). This finding may reflect a
general perceptual process described by the Gestalt principle of "good continuation," but given that a limited number of serial orders was tested, no firtn
conclusion can be made.
Further questions concern the role of larger contexts. According to the
present evidence, the rare interval of a tritone is not a reliable cue to tonal
structure. This finding, of course, leaves open the possibility that the tritone
may serve to consolidate or to disambiguate tonal information conveyed by
longer melodic patterns (Cuddy et al., 1981) or folk-melodies (Boltz, 198$).
However, other "predictable, prototypical aspects of a tonality" may be of greater
importance in establishing the tonal strength of a melody (Dowling, 1991, p.
307).
The present work was addressed to Butler's criticisms of Cuddy and Badertscher
(1987). It provides no support for the notton that the results obtained by Cuddy
and Badertscher (1987) were simply Artifacts of a flawed methodology. Instead, the theoretical interpretation of these results in terms of the representation of the tonal hierarchy (Krumhansl, 1990) or the stability conditions of the
basic pitch space (Lerdahl, 1988) is further strengthened by the converging
evidence of the present experiments.
References
Bharucha, J. J. (1984). Event hierarchies, tonal hierarchies, and assimilation: A reply
to Deutsch and Dowling. Journal ofExperimental Psychology: General, 113,421425.
Boltz, M. (1989). Perceiving the end: Effects of tonal relationships on melodic completion.
Journal of Experimental Psychology: Human Perception and Performance, 15,
749-761.
Brown, H., & Butler, D. (1981). Diatonic trichords as minimal tonal cue-cells. In.
Theory Only, 5 (6 & 7), 39-55.
Browne, R. (1981). Tonal implications of the diatonic set. In Theory Only, 5(6 & 7), 321.
124
125
Author Notes
This research was supported by the Natural Sciences and Engineering Research
Council of Canada and the Advisory Research Committee of Queen's University. Karen
Smith and Alan Marr provided excellent technical assistance. Valuable comments on
an earlier draft were provided by A.J. Cohen, C.L. Krumhansl, and M.G. Wiebe. Editorial comments of M.R. Jones and four anonymous reviewers are also gratefully acknowledged. Requests for reprints should be sent to L.L. Cuddy, Department of Psychology, Queen's University, Kingston, Ontario, Canada, K7L 3N6.
126