Abstractions
The first studies investigated the interval perception using tone pairs with a whole-tone fre-
quency change within a tone pair.46,47To determine whether the direction of a sound change
within a pair is encoded by the auditory cortex irrespectively of the absolute sound fre-
quency, the pitch of the standard tone pairs was randomly varied between five different fre-
quency levels (AB, CD, BC,... , EF, DE, AB,...).^47 Deviant pairs had either reversed
frequency change (BA,...,FE) or a frequency repetition (AA, BB,...,EE), also at five dif-
ferent frequency levels. It was found that despite this complexity in stimulation and the
parallel reading task given to the subjects, the deviant sound pairs elicited MMN, indicat-
ing that the change in the direction of the interval was discriminated in the auditory cor-
tex. This line of research has been continued by Paavilainen and his colleagues.^48 They
found that in addition to the direction of the tone interval, also the interval size is auto-
matically encoded in the auditory cortex. This was evidenced by the MMN elicitation by
smaller or larger intervals among minor seventh standard intervals (all intervals were ran-
domly presented at 10 frequency levels). Furthermore, the cortical interval representation
may also be formed on the basis of converging information from both ears as indicated by
the MMN elicitation when the two successive sounds of an interval are delivered to sepa-
rate ears.^49
In sum, the present results show that the human auditory cortex can automatically
encode highly complex and abstract auditory information (see Ref. 50 for a review).
Expectations
In the two studies to be summarized below, the regularity of the sound change itself cre-
ated a rule about the sound continuum, in other words, an expectation for the next sound
of the sequence either in terms of pitch or sound location. An MMN elicitation by abrupt
changes in such sound continuums would strengthen the above-documented conclusions
about the ‘intelligence’of automatically formed neural sound representations in extracting
invariant rules from the continuously varying sound information.
Tervaniemi et al. presented reading subjects with regularly spaced sound sequences con-
sisting of regular pitch decrease of one semitone as the standard event.^51 In two conditions,
the sounds were Shepard sounds and in two conditions they were pure sinusoidal tones.
Randomly, a pitch repetition or an ascending pitch interrupted this line of regularly
descending pitch. The MMN was elicited by ascending sounds, be them Shepard or sinu-
soidal sounds, and by pitch repetition when they were Shepard sounds. These data sug-
gested the existence of primitive expectation formation even without conscious attention
being directed to the sounds.
To c o nfirm and expand the previous data obtained by using sound sequences based on
regular pitch change, the follow-up experiment was conducted in which the regularity was
created by virtual sound movement.^52 To create a subjective feeling of sound movement,
sounds were presented via nine loudspeakers extending in a horizontal symmetrical arc in
front of the subject. Short noise bursts were presented about three times a second, contin-
uously moving back and forth between the two extreme loudspeakers. Occasionally the
regular virtual sound movement was broken by the sound skipping two loudspeakers,
300