Did you know that αλλα occurs 557 times in the whole of the LXX (nearly four times the size of the NT); compared to 638 times in the NT? That number includes the alternate MSS text in Joshua, Daniel, etc.

And the LXX books with the most frequent instances of αλλα (instances per 1000 words in book) are books likely composed in Greek, stuff like Wisdom of Solomon (3 αλλα/1000 words) and 4 Maccabees (6.3/1000)? Tobit and Judith are also high (both 2/1000), but so are Malachi (~2/1000) and, believe it or not, Job (~3/1000)!

Comparatively, Holmes' second edition Apostolic Fathers Greek text has 352 instances. That's over half of the number of NT instances, but the corpus size is just over 1/3 of the NT.

That'll make a guy think.

Update (2008-03-12): On the idea of whether αλλα can be an indicator of Aramaic in the gospels, or translation Greek; let's not forget that it could just be a diachronic thing. LXX => early, NT => later, AF => even later. Maybe αλλα usage increased over time, particularly since it is a development from αλλος. (It may therefore be useful to compare use of αλλος/αλλα between LXX/NT/AF to get a better picture; I'll leave this for someone else to dig into).

On the Aramaic question, perhaps the place to start would be with Raymond A. Martin's Syntactical Evidence of Semitic Sources in Greek Documents (amazon.com). He thinks that minimal use of certain prepositions may be evidence of a translated text; also και/δε in some contexts, but doesn't say anything about αλλα (from my quick re-skim). Martin presents a lot of evidence, but I don't know that I buy it (call it my "correlation does not prove causation" skepticism).

Anyway, here's the chart of αλλα usage in the NT. I generated this with Logos Bible Software's Graph Bible Search Results feature, which is an option on the right-hand side of all Bible search results menus (Bible Speed Search, which I used for this; Bible Search; and Syntax Search). The numbers on the right of the bars are # of αλλα per 1000 words in a book. So Matthew has 2.0162 αλλα per 1000 words; Mark has 3.9788 per 1000; etc. This sort of distribution leads me to think that use of αλλα is perhaps more stylistic than translational; though the LXX numbers reported (that graph is below the NT graph) are much less frequent. This may say more about the translators and the time they translated in than the translation itself.

And here's the LXX graph: