This example implements a rich theory of harmony. More specifically, it implements a whole collection of harmony CSPs.
The example implements a considerable amount of Schoenberg's tonal Theory of Harmony. The different harmony CSPs implement increasingly complex exercises which stem from Schoenberg's textbook. For example, it implements Schoenberg's recommendations on convincing chord root progressions (there are multiple variants of increasing complexity), and Schoenberg's suggestions for treating minor.
Nevertheless, the example does not follow Schoenberg faithfully. For example, the example is implemented in 31-tone equal temperament (31-ET), instead of the 12-tone equal temperament which Schoenberg stunningly exhausted in his compositions. 31-ET has been chosen, because it allows for convenient pitch class calculations (much like 12-ET, but with 31 pitch classes) and at the same time supports enharmonic notation (e.g., C# and Db are different pitch classes). You can play all examples by simply clicking on the score images and you will hear that this tuning also sounds lovely (it is a variant of Meantone temperament, the prevailing tuning of Renaissance music and early Baroque).
The example also diverts from Schoenberg in the actual rule sets implemented. Sometimes, constraints are added which stem from a different source, and sometimes rules suggested by Schoenberg have been generalised.
The full source for creating the solutions presented below is available.
Oops: the examples and the code of this release produces parallel fifths/octaves (between voices more than an octave apart). This has meanwhile been fixed in the SVN repository, and I will update this site some time later...
We skip much introductory material of Schoenberg's textbook and start with one of the first musically more interesting exercises: compose a harmonic progression of diatonic chords where consecutive chords are connected by what Schoenberg calls a "harmonic band", that is, common pitch classes between chords. In case there are common pitch classes, then these are placed in the same voice. The other voices move in small intervals (Bruckner's "law of the shortest path").
The first two staffs show the usual four-voice setting of the harmonic progression. The third staff presents analytical information: the chord root is notated with the duration of the actual chord, and other chord pitch classes as grace notes. Giving this information is a bit overkill in this simple exercise, but I find it useful for more complex cases, and the program creates it therefore automatically. Finally, also the Roman numerals of the chord roots are shown.
click the score for sound (mp3)
A second solution to this first exercise is given, as for many other exercises below as well.
The first exercise does not necessarily end with an authentic cadence (see above). This second exercise constraints that chord progressions end in a cadence. The constraint used here does not stem from Schoenberg.
This constraint requires that the union of the pitch classes of the last N chords equals the pitch class set of the underlying scale. In the following solution, the scale here is C-major, and N=3. Together with Schoenberg's harmonic-band constraint this results in an authentic cadence.
In order to better understand the effect of this cadencing constraint, N is set to 4 for the following solution (the scale is changed to Bb-major). Note that this "cadence" of 4 chords does not necessarily result in an authentic cadence.
In the previous exercises, all chords where is root position. Schoenberg suggests that 6th chords can also be used freely, which is done in the following solution.
Later in his textbook, Schoenberg revises his prior recommendation on chord progressions (harmonic band). This case indicates Schoenbergs teaching approach in general. At first he introduces rather strict rules which (as he says) exclude many useful solutions but at least allow mostly convincing results. Later, he repeatedly relaxes and refines his rule set.
Schoenberg's recommendations are briefly summarised in the following, please consult his textbook for the full details. He introduces the notion of so-called ascending progressions (e.g., V-I or III-I), descending progressions (e.g., I-V or III-V), and super-strong progressions (e.g., V-VI). In my generalisation of Schoenberg's concept, ascending progressions have a harmonic band, but the root of the second chord does not occur in the first (Schoenberg says something like "the king of the first chord is overruled by a new king"). Descending progressions also have a harmonic band, but the root of the second chord was already a non-root tone in the preceeding chord. Finally, super-strong progressions have no harmonic band. (There is a fourth case not mentioned by Schoenberg were two consecutive chords have the same root.)
Ascending progressions can be used freely, super-strong progressions should be used sparely for special emphjasis, but descending progressions should commonly be treated with particular care. Schoenberg recommends that in a sequence of three chords C1, C2, C3 the sequence C1, C2 can only then be descending if C1, C3 is ascending (e.g., III-V-I).
The following example contains a super-strong progression between the 2nd and 3rd chord (V-IV), followed by a decending progression (IV-VI), all other progressions are ascending. These solutions are already a bit longer, because we meanwhile have a more rich harmonic vocabulary.
The next solution consists solely of asdencing progressions.
All solutions presented so far consisted of consonant triads only. When introducing seventh chords discussing dissonance treatment for the first time, Schoenberg recommends that these dissonances should be resolved by the strongest chord root progression available — "a fourths upwards the fundament" — and that the seventh (the dissonant tone) should be resolved by a step downwards.
The following exercise implements a rule which is already more liberal (and which was not proposed by Schoenberg). Dissonant chords are always resolved by an ascending progression. The following example contains a few seventh chords. For example the 4th to 5th chord (I7 - IV) is an instance of a "fourths upwards the fundament".
Again, multiple solutions are presented.
This exercise introduces the use of the minor key. The minor key is particular in that the set of seven diatonic scale degrees is extended by two raised pitches. This is the first step in Schoenberg's textbook towards opening up tonal harmony, a development which finally culminates into the free use of 12 pitch classes.
For Schoenberg, it is important to make clear the characteristics of minor. He therefore at first imposes relatively strict rules on the "special degrees" VI and VII which can occur either in their natural or their raised form. Schoenberg suggests that these "turning points" are always resolved stepwise in the direction of the accidental (e.g. natural accidental downwards toward natural accidental, and sharp upwards to next sharp if available).
In the solutions given below, this resolution may however happen in a different voice. For example, the F# in the 3rd chord moves upwards to G# in the 4th chord, but this G# happens in a different voice. Again, two solutions are given, both in A-minor.
The following two examples combine the treatment of "turning points" in minor with the dissonance treatment introduced before.
TODO
Sketch: Not only a single underlying scale is used, but chords are related two their simultaneous scale. Just before a new scale starts (or where scales overlap in time?), "neutral" chords are used which only use pitch classes which occur in both consecutive scales (the size of this "window" is controlled by some variable N). The first chord after the "old" scale ends uses pitch classes which do not occur in the previous scale (modulation chord). Optionally, a cadence in the new scale follows immediately.
Finally, some exercises are shown which apply Schoenberg's rule set to unconventional scales and chords made possible by the selected tuning (31-ET).
This exercise uses the septimal minor scale. The meantone note names of this scale are as follows. Remember that in this tuning enharmonic notes are actually different pitches. For example, the interval C-Eb is the minor third (6/5) whereas C-D# is similarily consonant subminor third (7/6).
C D D# F G G# A# (C)
TODO: pict of septimal minor
The exercise uses only scale notes and only chords in root inversion. The rules of connecting chords is again the more simple harmonic band constraint, and four chords form a cadence. The analysis now shows the ratios approximated by the tuning. The notation first shows the ratios of the chord in a column surrounded by parentheses, and a transposition factor for all these chord ratios. Two solutions are shown.
Whereas the septimal minor scale still shows much similarity with the common diatonic scales, the next example uses an even less familiar scale, namely the following pentachordal scale proposed by Paul Hahn.
C Db D# E F G Ab A# B (C)
TODO: pict of Hahn's pentachordal
Again, only scale notes and only chords in root inversion are used in this exercise. All chord progressions are ascending progressions (following the generalised definition given above, Schoenbergs definition does not apply anymore for this scale). Again, four chords form a cadence.
The next solution uses chords in both root inversion and 1st inversion.
All the exercises above are defined by the same CSP. The different exercises above are realised by giving different arguments to this CSP in the solver call.
Moreover, this example is implemented in such a way that the creation of the abstract harmony (e.g., the analytical harmonic information, in contrast to the actual notes) is defined in a modular way. In other words, the harmonic CSP without the four-voice voiceleading rules can be reused in other constraint problems, for example harmonic counterpoint problems where the polyphonic structure depends on some underlying harmony.