On Salience – What do We Hear?

Introduction: On Salience
My musical training was unusually thorough, but as a young composer I still found it inadequate. And as a beginning teacher, despite knowing all the standard “rules”, I often lacked plausible explanations as to why certain situations in harmony, counterpoint, and orchestration sounded convincing,  or not.
The standard explanations are often limited to specific styles. Also, much training in writing skills is based on rules of thumb, designed to avoid common beginners’ errors. But the young composer needs a broader conceptual framework, which he can transfer to his own music.
While some violations of the standard  rules are audibly flagrant, others disturb less, and still others, not at all. Indeed, the intrepid student will often find examples of “broken rules” in the standard repertoire, which however sound perfectly convincing. Telling the student his hearing will eventually improve is a poor substitute for a real explanation. Better aim at a more refined understanding of many, varied musical situations, too often lumped together under these oversimplified rules of thumb.
A composer strives to communicate meaningfully with his listeners. To achieve this, he must succeed in organizing what is easily audible: If the listener’s first encounter with the work is uninteresting, he will likely not seek further acquaintance. But arousing immediate musical interest requires careful attention to the relative salience of musical phenomena. In theory and in teaching however, questions of relative salience are too often dismissed as unimportant.
Three common situations illustrate my point.
1) Species Counterpoint
Species counterpoint is generally accepted as an introduction to many idioms of tonal writing.  Here it will be a useful starting point, precisely because it is commonly taught, and because its rules are well known.
In species counterpoint parallel fifths are prohibited. Depending on the pedagogical tradition, there is some difference of opinion about the details of this prohibition, but let us begin with a clear-cut case.

There is no teacher of species counterpoint who would allow these fifths. They are chord tones, on strong beats, and in addition, the return to the E at the end of bar 1 creates immediately consecutive parallels with the first beat of bar two.  These could not be more obvious.

This example is subtler. Here the parallels sound between the last beat of bar 1 and the 2nd beat of bar 2. The parallels are audible, but less prominent, despite being less than four beats apart. (Some textbooks suggest the “four beat rule”: Parallels must be more than four beats apart to be acceptable.) This is because the parallel thirds on the strong beats enrich the harmony, and the differing melodic contours in bar 1 and in bar 2 do not direct the ear to the parallels.  Also, the F in bar 2 does not trigger a change in melodic direction, but is rather part of a larger movement down to C in bar 3.

Here we see the same parallels as in ex. 2, but with an added third part. This added part attenuates their effect even more, since the listener’s attention is distracted by the attack of the low G in bar 1, the suspension it forms into bar 2, and the rich sixth it adds on the 2nd beat  of bar 2.
Many counterpoint teachers would ban all of these parallels as unacceptable.  But does this make sense? Surely it is more useful to help the student to understand such distinctions between degrees of audibility; after all, examples like the last one abound in the repertoire. I am here reminded of Brahms’ well known catalogue of parallels found in the works of the masters, assembled precisely so he could see under what circumstances the rules of thumb were no longer adequate guides.
2) Set Theory
Many theorists use set theory as a tool for harmonic analysis of non-tonal twentieth century music, since it provides a fairly straightforward way of comparing harmonic events. Some composers also use it, to generate harmonic events. But set theory unallied to clear notions of salience can seriously mislead the young composer.

Let us begin with ex. 4a. This simple chord, set #3-2, scored here for divided violins, is heard as a harmonic unit, and it has a distinct intervallic character.
Now let us compare this chord with two others: ex. 4b and ex. 4c. Ex. 4b shares two common tones with ex. 4a, as well as its timbre and spacing. Its interval content is different; it is not the same set. Ex. 4c returns to set #3-2, now transposed and respaced, scored with heterogeneous timbres. I submit that if one asks even a musically sensitive listener to rate the three chords in order of audible similarity, a and b will be considered as more similar than either a and c, or b and c.  My point is not that there is no relationship between 4a and 4c, but that this purely intervallic relationship is very subtle, compared to relationships of timbre, common tones, and spacing.
Why are common notes, common timbre, and similar spacing easier to perceive as “related” than common intervals, particularly when 8ve transpositions are involved? There is clearly some perceptual prioritization going on here. This prioritization probably grows out of the way human hearing evolved: There is more survival value in distinguishing the roar of a lion from the twittering of birds (where rough distinctions of register, simple pitch comparison, timbre and rhythmic recognition suffice) than in detecting complex interval relationships presented simultaneously.  (Incidentally, interval relationships are much easier to see than to hear; visual and auditory prioritization do not favor the same elements.)
Yet another example: 4d. Here set #3-2 is embedded in a more complex chord. For simplicity I leave the timbre – strings – and the rhythm the same. Is it reasonable to expect even a fairly sophisticated listener to notice this family relationship, particularly if the two chords are not heard in close succession? The problem here is one of memory, and of distraction. There are so many simultaneous interval combinations being heard in this chord that without some sort of “help”, e.g. special articulation, timbral separation, etc., the listener has no reason to focus on the very subtle similarity between ex. 4a and 4d during the second or so during which this chord would sound. In fact the listener’s priorities are directed elsewhere. A little experimentation reveals that as chords contain more notes, their intervallic components get harder to distinguish, apart from a few special cases (e.g. chords whose component, adjacent intervals are all the same). If the composer wanted to underline the intervallic resemblance in this example, he would need to highlight it in some way, using other aspects of the music.
Once again: The composer must learn to distinguish and organize what it easily heard before worrying about such refinements of intervallic association. Only when correlated with salience do such intervallic relationships become audibly meaningful.
3) Motives
Another confusion between what is heard and what is seen occurs in motivic analysis. In his theoretical writings, Schoenberg often uses the phrase “remote motive forms”. While I have never found a precise definition for this term, there is indeed a perceptual  hierarchy between motive variants.
For example, students are taught that retrograde is an elementary form of motivic variation. But except for a few, trivial cases, rhythmic retrogrades contrast radically with the original forms; the listener is usually struck by the difference between the original and the retrograde, rather than by their similarity – assuming he notices the relationship at all.  What, then, does it mean to refer to a motive and its retrograde as being related? What about retrograde augmentation, in different sections of the work? To what extent are such resemblances audible at all, and therefore useful to the composer in organizing musical form?
Here again, we need a scale of perceived relatedness, rather than a simple black and white distinction.  As Schoenberg’s term implies, some motive variants – e.g. simple transposition with rhythm intact – are much more easily recognizable than others. My point here is that unless such a scale of relatedness is part of a discussion of motives and their use, both the composer and the analyst risk mistaking insignificant musical coincidences for formally significant connections.
Conclusions from these examples
We see that the same phenomena have different effects according to the general musical context. The composer needs to distinguish these effects, or else he risks misjudging the effect of his piece. Great composers never make such misjudgments. Their music is forceful, and  such perceptual impact requires successful coordination and aural prioritization of musical elements.
A composer, trying to create meaningful continuities and contrasts, or a theorist, trying to understand them, needs, at the minimum, the notions of perceptual foreground and background. Thus, one might say that the intervallic similarity between Ex. 4a and Ex. 4c is a common element in the background, due  to the more salient differences between the two chords.
Such distinctions about salience allow the student to make the transition from simplistic rules of thumb to more powerful, general principles. Refined observation of degrees of salience makes analysis more sensitive and relevant to how music is actually heard.
It may be objected that not all listeners are the same, and this is true. But it is also true that some observations correspond better to the music’s reality than others; otherwise all analyses become equally valid – a ridiculous proposition.
Another objection, namely that salience cannot be precisely quantified, does not undermine its importance.  Even a rudimentary ranking of salience leads to greater insight.
The need for a theory of salience
Ignoring the relative salience of musical elements can seriously mislead the student. Salience is not always simple to analyze, since it can change, even within a phrase. Also, no aspect of music exists in isolation. Every sound is heard in a given timbre, in a specific register, and combined with other sounds simultaneously and/or successively, creating harmony and counterpoint. When attempting to sensitively describe and explain musical phenomena, prioritization is necessary, because that is how the human brain works. Again, evolution: An organism with no priorities of perceptions risks ignoring danger until it is too late.
Musical salience is anchored in aural perception. As such it connects musical analysis with the field of cognitive psychology. Important experimental work has already been done in this area; see for example Albert Bregman’s pioneering book, Auditory Scene Analysis. Bregman enumerates criteria used by the human brain to divide auditory signals into meaningful streams of information. Many of his observations speak directly to issues of salience.
A theory of salience could thus be a bridge between the fields of analysis and music cognition. And by ensuring that musical theory respects the facts of  how human hearing works, it could better serve the needs of composers and performers.