most consonant interval, the perfect fifth, we see a regular pattern of major and minor
peaks (Figure 9.1H). The pattern with one major peak and three minor peaks repeats every
4.55 ms (1/x220 Hz). This periodicity corresponds to the missing F 0 of a harmonic series
containing energy at the second harmonic (440 Hz, A 4 ) and third harmonic (660 Hz, E 5 ),
the F 0 s of the notes actually present in the stimulus. Rameau’s concept of the ‘basse fon-
dametale’(fundamental bass) in his Treatise on Harmony^53 is related to the missing F 0 of a
harmonic interval.
Autocorrelation functions provide another representation of temporal regularities and
irregularities embedded in acoustic waveforms (Figure 9.1I–L). Autocorrelation functions
are computed by multiplying the waveform with a delayed copy of itself and integrating
over time. A large value at a given delay indicates the presence of a dominant periodicity in
the waveform whose period equals the delay. Like pitch percepts, but unlike acoustic
waveforms, autocorrelation functions are stable despite changes in the relative phases of
frequency components. In the autocorrelation functions plotted in Figure 9.1I–L, the
periodicity at the upper limit of the xaxis (50 ms) corresponds to 20 Hz, the lowest
audible frequency.
In the autocorrelation function of the perfect fifth (Figure 9.1L), the first major peak
again corresponds to the missing F 0 ,A 3 (220 Hz). The second major peak occurs at
9.09 ms, which corresponds to A 2 (110 Hz), the bass note an octave below. In fact, all the
major peaks up to 50 ms correspond to the fundamental bass and its subharmonics
(undertones) at A 2 ,D 2 ,A 1 , and on down to A 0 , the lowest note on the piano (F 0 27.5 Hz).
In between the major peaks is a set of three, evenly spaced, minor peaks. The first of
these minor peaks occurs at 1.51 ms, which corresponds to E 5 (660 Hz), the upper note of
the interval. The second minor peak occurs at 2.27 ms, which corresponds to A 4 (440 Hz),
the root of the interval. The third minor peak occurs at 3.03 ms, which corresponds to E 4
(330 Hz), the octave below the fifth and the fifth above the fundamental bass at A 3.
Temporal regularities are also seen in the waveform and autocorrelation of the perfect
fourth, the other consonant interval in our stimulus set (Figure 9.1F and J). Here the major
peaks are at 6.82 ms (D 3 ,F 0 146.7 Hz) and 13.6 ms (D 2 ,F 0 73.3 Hz). Thus, in addition
to a representation of the fourth, there is a representation of its inversion as a fifth with the
implied root at D 3 and the fundamental bass at D 2. The autocorrelation function of the
fourth is a bit more complicated than that of the fifth, as there are two more peaks between
each pair of major peaks. In the first set of minor peaks, the following notes are repres-
ented: A 4 (the root), D 4 (the interval), A 3 (the octave below the root), and G 3 (the fifth
below D 4 , and the fourth of an interval rooted at D 3 ). Thus, we find representations of
notes that function as fourths and fifths in the major and (all) minor scales of A and D.
In summary, the temporal fine structure of the perfect fifth and fourth contains repres-
entations of the two notes constituting the interval, plus harmonically related bass notes
that are implied by the interval. In music, these bass notes support the deep structure of
harmony. Parncutt^18 demonstrated experimentally that listeners associate major triads
with pitches that are harmonically related to note F 0 s, including the fundamental bass, plus
the pitches of note F 0 s actually in the stimulus. These pitches cannot be accounted for
simply on the basis of combination tones (for review, see Ref. 54). Houtsma and
Goldstein^55 showed that musicians can use missing F 0 pitches to identify melodic intervals
134