Before we continue on, there is a concept we need to understand in order to fully comprehend the research presented below. A pitch produced by a musical instrument or voice is "composed of a fundamental frequency and harmonics or overtones."( Doering [link]) These components of sounds establish the timbre of a particular instrument, based on the relative strength and number produced by the instrument.(Jones [link]) The reason a trumpet sounds different from a clarinet or a human voice, even when all are producing the same note, is due to the relative strength and number of harmonics (or overtones) present in the sound. Despite the presence of harmonics the listener perceives only a single tone, but the presence of these harmonics greatly affects our perception of harmonic intervals and chords as being consonant or dissonant.

The study by Tramo et al. (2001) compares the acoustic representation of several intervals and the trains of action potentials produced by the auditory nerve fibers of a population of 100 cats hearing the same intervals. It is important to keep in mind the information presented about the ascending auditory pathway . Auditory nerve fibers are the central axons of the spiral ganglion cells that transmit synaptic information to the cochlear nucleus neurons in the brainstem. When an interval is sounded the nerve fibers corresponding to the frequencies present in that interval will fire. Virtually all information about sound is transmitted to the brain through trains of action potentials produced by these synapses.[3]

First the researchers examined the acoustic waveforms and their corresponding autocorrelations (the cross-correlation of a signal with itself, to find disguised repeating patterns) for four different intervals. Two of the intervals, the perfect fourth and fifth, are considered consonant, while the other two, the minor second and tritone, are considered dissonant. When looking at the acoustic representations of the consonant harmonic intervals there is a clear pattern of peaks and the autocorrelation is perfectly periodic. The period of the autocorrelation corresponds to the fundamental of bass of the interval. Yet with the dissonant intervals, there was no true periodicity. Instead the acoustic spikes occurred inconsistently.

Figure 1: The acoustic representation of the sound waveforms and the corresponding autocorrelations for each interval. This image was created by Tramo, Cariani, Delgutte & Braida. (2001)

Using 100 cats, the experimenters measured the firings of the auditory nerve fibers on the cat's cochlear nucleus using electrodes implanted in the brainstem. Similar to the best frequencies in the section on tonotopic organization there is frequency selectivity present with the spiral ganglion cells and the nerve fiber "will only increase the number of action potentials it fires if it is sensitive to the frequencies present in the interval."[3]

Using the data collected, the researchers measured the the intervals between the spikes and then constructed an all-order histogram for the interspike intervals (ISIs). This all-order histogram is equivalent to the autocorrelation of the acoustic waveforms. In examining the histogram for the consonant intervals, it is obvious that the major peaks appear with clear periodicity corresponding to the fundamental bass. Dissonant harmonic intervals' ISIs are irregular and contain little or no representation of pitches corresponding to notes in the interval or the fundamental bass.[3] As was seen in the autocorrelation, there is again no clear periodicity of spikes. The neural response appears to mirror the acoustic representation of the intervals.

Figure 2: The all-order histogram produced by Liégeois-Chauvel et al. showing the the interspike intervals produced by the auditory nerve fibers of 100 cat subjects.

The reason that the consonant intervals' autocorrelations and histograms exhibit a clear periodicity and the dissonant intervals fail to demonstrate this relates back to the overtone series. Existing with each note of an interval are the inherent harmonics. The notes in consonant intervals, such as the perfect fourth and fifth, have some of the same overtones. Thus, the harmonic series of each note reinforces the other. However, the harmonics of the notes in the dissonant intervals, such as the tritone and minor second, do not reinforce each other in this way. This results in periodic amplitude fluctuations, known as beats, which make the tone combination sound rough or unpleasant. The audio example below presents a comparison of a perfect fifth followed by another fifth with the top note slightly lowered resulting in a clear example of beats.

In another study, a patient, MHS, who had bilateral lesion of the auditory cortices, was examined. In this study the experimenter presented the patient with either a major triad or a major triad with the top note slightly flattened, similar to the audio example presented above. After the chord was presented the patient was asked to indicate whether the chord was out of tune or properly tuned. The patient answered with an accuracy of 56% which is two standard deviations below the norm of 13 controls. [3]

There are two possible interpretations of MHS's results.

In addition to his inability to recognize the tuning of the chords, in another study he also demonstrated an impairment in his ability to detect the direction of pitch change between two notes. This was determined using a pitch discrimination test where the subject must identify whether the pitch is going up or down. The changes in frequency get progressively smaller. MHS "performed poorly in the final third of the test where the ΔFs [changes in frequency] were smallest."[3]

Samson et al. (2001) conducted an experiment that confirms that the left hemisphere does in fact process rapid temporal information in music. The objective of the experiment was to "test the effect of tempo on irregularity discrimination in patients with unilateral lesions to the temporal lobe."[5] The experimenters played two successive sequences of a single tone repeated five times. One of the sequences presented contained regular silent intervals between the onsets of the tones, but the other contained irregular silent intervals. The patients were asked to judge which of the two sequences was regular or irregular. The duration of the silent intervals between the tones of the sequence were referred to as the "interonset intervals" and the various durations used were 80, 300, 500, 800, or 1000ms. This procedure was carried out on a total of 24 subjects: 8 with right hippocampal atrophy, 10 with left hippocampal atrophy, and 6 normal controls.[5]

The test confirmed the results of previous studies. At the rapid tempo, the interonset interval of 80ms, irregularity discrimination thresholds were significatly reduced for patients with damage to the left hemisphere.[5] Of course the opportunity for error was higher at the faster tempos no matter whether the patient had cerebral damage or not, but the results show a marked deficit for patients with left hemisphere damage as opposed to the other two subject categories. Again in agreement with previous studies, this data shows that damage to left hemisphere affected only the processing of fast temporal information, while slow temporal information processing was spared. In addition, damage to the right hemisphere resulted in very little difference in the ability to process rapid temporal information, and slow temporal processing was also normal.