Information Technology Reference
In-Depth Information
of songs used for this experiment can be found
in Table 1.
To prevent a subject's pitch selection for a given
trial from being influenced by the previous trial, a
sequence of 14 pseudo-random interfering notes
was played in between each trial, as was done by
Halpern (1989). The songs were presented to the
subjects in one of two different orders, which were
assigned randomly and equally to the subjects.
The
Autoscore
transcription software was
run in its
Constrain
mode for this experiment.
In this mode the program identifies the closest
semitone to the user's hummed frequency for
a given note, recording the note's pitch as that
value. One consequence of using this mode is that
occasionally a subject may hum two notes that
sound very similar in pitch, but their frequencies
differ enough that they straddle the cutoff between
two semitones and end up being reported as two
different pitches. We saw in our collected data
several instances where this occurred.
them knew all but two. Subjects did not attempt
songs they did not know. In all but a few cases,
subjects hummed exactly as much of each tune as
was represented by the lyrics in front of them.
Table 2 illustrates the aggregate values of the
starting pitches chosen by subjects for each of the
performed songs. We report the data first averaged
by song, then by subject. The reported average
pitch values are given in the MIDI note scale (e.g.,
middle C is represented as 60 and consecutive
semitones are separated in value by one). Since
Halpern noted that women typically sing about
an octave higher than men for the same song
(1989), all of the pitch values for female subjects
(denoted by italics in the table) were reported by
their actual MIDI values but were reduced by 12
(semitones) when calculating averages to allow
direct comparison to those of male subjects.
Both sections of the table show that the starting
pitches chosen by MUSI subjects were consistently
higher than those of the other two groups, whose
values are roughly the same. When examining the
values grouped by song, the MUSI group starting
pitches were an average of 2.1 semitones higher
than the BTWN group and 2.3 semitones higher
Results
Nine of the subjects knew all 12 songs used in this
experiment; 4 of them knew all but one, and 2 of
Table 2a. Average starting pitch selection and standard deviation, tabulated by stimulus song, from
Experiment 1
Subject Group
Song
MUSI
BTWN
NONM
ALL
Pitch
S.D.
Pitch
S.D.
Pitch
S.D.
Pitch
S.D.
d01
45.75
2.87
43.60
1.52
44.25
0.50
44.46
1.94
d02
46.00
2.74
44.67
1.21
45.25
1.26
45.27
1.83
d03
44.75
2.22
44.00
2.97
43.75
0.96
44.14
2.21
d04
52.80
2.05
48.83
3.06
49.00
1.63
50.20
2.96
d05
43.50
1.29
44.20
1.79
42.50
1.73
43.46
1.66
d06
48.00
1.73
44.20
1.92
43.00
0.82
45.21
2.67
d07
45.80
3.27
44.83
2.56
43.75
2.63
44.87
2.75
d08
49.00
1.00
47.50
1.87
46.75
2.22
47.80
1.86
d09
50.60
1.52
47.67
1.51
49.00
3.16
49.00
2.30
d10
46.60
2.61
45.33
2.94
44.75
1.26
45.60
2.44
d11
49.50
5.45
45.83
2.14
46.50
1.91
47.07
3.47
d12
46.80
3.63
43.25
1.71
42.75
1.50
44.46
3.07
Note: Pitch values are given as MIDI note numbers; Female subjects are shown in italics.
Search WWH ::
Custom Search