Precis of: Space and Sound, a two-factor approach to pitch perception

This is an invited guest-post by Adam Morton--ES.]

Perception, consciousness, language, and memory interact in our awareness of musical pitches and keys differently than in visual perception. For example there is good evidence that we remember pitches more accurately than we are aware of, even when we are vividly conscious of them in our experience. It would be good to understand this better, partly as an antidote to the way vision dominates the discussion of perceptual appearances. The topic has a practical side, too, as anyone interested in music will wish they were better at recognizing notes, intervals, keys. The essay this summarizes reports on my experiments with pitch during the past year, their degree of success, and some speculative conclusions. It is definitely unorthodox, though it connects with some data in(For the full details see my essay here; it has a bibliography [see also these papers--ES].

History, aims. My aim was to separate access from accuracy. People with absolute pitch vary in how precisely they can discriminate notes, so as a limiting case imagine someone who can verbalize whether a note is within a half-octave of some tone. They would be able to access information that most people cannot, but they would not have a very accurate pitch sense. I wanted to see if I could develop access, letting accuracy fall however it might. This would be like our grasp of colour, where we have a rough approximate vocabulary that almost everyone can apply, but few can use with great precision. Working towards this, I stumbled on the spread/wobble distinction (to be explained below). Now I can verbally classify pitches, independently of their register, into several rough overlapping classes. I often have intuitions now about keys. I make mistakes but I am usually within a tone or so. I can now do very simple sight-singing from written music.

spread/wobble 

The central move is to split the hard-to-remember pitch qualia into two components. They both present as spatial extents, though these are unconnected with where the sources are really located. This can be done in two ways: the training described in the essay, and by accepting the graph below of my hard-earned classifications and insisting that notes have the required spatial qualities, until they take them on spontaneously. They are best combined.

The first component is wobble. When you hear a note you can encourage it to wander or oscillate from one location to another, wandering in space before you or oscillating from side to side. The pattern of wandering is characteristic of a given pitch, independently of its octave. The location alternates between locations, stopping at one before you hear it at another. One technique here is to hear notes as beginning at a definite point in time but having two endings. The note begins with a percussive or consonantal onset sound, which soon ends, and the end of this onset has a sound quality itself. The on-sound and the off-sound occupy different apparent locations.

The other component is spread. Simultaneous with wobble, or beginning an instant later, there is a continuous spread or slide, to fill an extent that can end either within or beyond the wobble. It seems to occupy all of this horizontal interval. It is simultaneous, occupying the whole angle between its limits, and moving smoothly between them.

It is far from obvious that pitches should have these features. The only way to convince yourself they do is to make them appear in the sounds of the notes. Separating them is the hardest part.

Rough graph  Here are the wobbles and spreads that the twelve semitones present to me.

If you hear notes as if they have these S and W, they will eventually begin to sound intrinsic to the notes. The next step is to produce random notes and classify them, at first with the emphasis on getting roughly accurate W and S rather than naming the note. At this time you can begin to work with the atonal ear-worms below.

atonal earworms  Tonality gets in the way of pitch recognition. But you can fix in mind sequences without tonal centers and memorize them as having their components. My procedure was to begin with the “four corners” a, b, g, d above, get them reliably into my head, then extend them to a whole tone scale: not in ascending order, say F#, A#, G#, D, E, C. I imagined and sang these at will in arbitrary orders, to make little Debussy-like tunes. Then I fixed in mind simple tonal melodies of only a few notes in contrasting keys, from which I jump part-way through to the six reference notes or to other melodies. A lot of jumping at random points is helpful, to keep the tonalities at bay.

Why?  An attractive possible explanation connects the two components with phase-based location, a primary auditory-spatial mechanism in birds and possibly other animals. A pure note of a given pitch will interfere constructively with itself at one ear and destructively at the other, if it is at a series of locations with respect to the hearer. So information about which directions of one's head make a sound loudest generates information about possible locations of the source. For pure notes there is a multiplicity of locations, which is what is relevant now, but this is much reduced for a complex sound, more important biologically. (But, surprisingly, I find that instead of moving my head I can swivel attention in imagination. And it can work with one ear! I think I can reconcile these with the phase-based model.) Explanations of how it might work are less important at this stage, though, than finding ways of making it vivid that it does work: producing separate judgements of wobble and spread. (Phonetic and motor connections can help here, as with people who can classify pitches by mentally singing them, or instrumentalists who can see their fingers reproducing notes they cannot consciously name. But I am omitting these from this precis.)