This article first appeared in The Schubertian,
No. 26, September 1999, pp. 2-3.
The Schubertian is published by the Schubert Institute of the United Kingdom (SIUK) http://franzschubert.org.uk/intro/ (link OK 26 September 2009).
A statistical overview of Schubert's keys will be presented here. This may be of some value in helping us to stand back for a moment from our more usual absorption in particular cases. Several limitations must, however, be acknowledged, as E. G. Porter discussed in his book Schubert's Song Technique, Dobson, London, 1961, chapter 3. In particular, many songs are not in any one key, but rather in roaming keys, mixed keys, or fused major/minor keys. The emotional significance of the keys has been well indicated by John Reed in his book The Schubert Song Companion, Manchester University Press, Manchester, 1985, Appendix II and so need not be considered here.
My tabulation includes all versions of a given movement (which term here covers not only instrumental movements but also individual songs); this causes negligible bias and, in any case, Schubert transposed some versions to different keys. Only the movements where the beginning and ending keys are the same are included, and minuets with trios are represented only by the key of the minuet. The included movements constitute 77% (1701/2221) of the whole; but in 208 cases the ending key was not known (because the movement was unfinished or fragmentary), so the effective coverage is instead 85% (1701/2013). The counts are shown in the graph and summarised in the table.
|Major||617 (45%)||195 (14%)||553 (41%)||1365 (100%) [ 80%]|
|Minor||194 (58%)||74 (22%)||68 (20%)||336 (100%) [ 20%]|
|Total||811 (48%)||269 (16%)||621 (36%)||1701 (100%) [100%]|
Four conclusions may be drawn:
 The range of keys is large. The most extreme cases may be mentioned: in C# major the Deutscher D139 and the vocal quartet Der Entfernten D331 (compare perhaps the distant key with the song's meaning, 'To the distant beloved'), and in ab minor the Deutscher D790/8. (I use upper case letters for major keys, lower case for minor.) Five cases begin or end in one of those keys but change keys and so are not included here (in ab-Ab: Auf dem Wasser zu singen D774 and the Minuet of the Impromptu D899/4; in c#-C#: Das Grab D569, Thekla D595v1 and the Piano Sonata D960/2).
 Major keys are more frequent than minor (80%: 1365/1701).
 Flat keys are more frequent than sharp (57%: 811/1432).
 The predominance of flat keys over sharp is greater in minor than in major (74% v 53%).
Do these conclusions hold for all genres? I divided the works into three groups according to the Series of the Bärenreiter edition: I-III (larger vocal works), IV (lieder) and V-VII (instrumental). The result is that all the conclusions hold in each group separately (full details are available on request). The value in conclusion  becomes, in the respective genres, 84%, 70%, 86%, in  62%, 59%, 52% and in  83% v 58%, 75% v 53%, 65% v 50%.
Do the conclusions hold throughout Schubert's life? I divided the works into three stages: D001-D500, D501-D800 and D801-D998. These divisions correspond approximately to the years 1810-1816, 1817-1823 and 1824-1828 and might loosely be called early, middle and late. The result is that all the conclusions hold in each period separately, excepting only that in the middle period sharps keys slightly predominate over flat (but still not in minor). The values in conclusion  become, in the respective periods, 80%, 86%, 74%, in  60%, 49%, 59% and in  84% v 55%, 61% v 47%, 66% v 56%.
A further conclusion from the graph is tempting: zero or two sharps or flats are favoured, compared with just one. Indeed, the first three places are taken by Bb, C and D major, with F and G major noticeably less used, and a similar comparison holds in the minor keys. Thus 'dips' occur at one sharp or flat in what would otherwise be fairly smooth bell-shaped curves. However, that conclusion refers to a more specific pattern and, as might be expected, holds rather less uniformly over genres and periods (again the details are available on request).
The four listed conclusions, then, appear to hold quite generally. Do they have broader musical significance? That might be more easily discussed when such tabulations are compared with those for other composers, some of which I am presently preparing. In the meantime, comments and suggestions from readers would be welcome.