This semester I decided to include a "reggae segment" to my undergraduate historical linguistics course, where we look at interesting linguistic changes in both Jamaican Creole and in "Dread Talk" (Rastafarian English), as exemplified by audio clips from various reggae songs. So, in addition to going over the traditional examples of sound change like Grimm's Law and Verner's Law, we also looked at sound change in Jamaican Creole. After discussing Verner's Law as a conditioned sound change, I introduced the phonologically-conditioned "likkle law" of Jamaican Creole English: the sound change /t/ > /k/ which occurs in certain words, such as likkle (< little):
“...Well, me said ‘no, no, don’t be like dat’, me likkle heart a pitter-pat...”
My initial characterisation of the conditions on this change was that /t/ became /k/ intervocalically. My guess was that Jamaican Creole English had been based on a vernacular British dialect in which intervocalic /t/ was realised as a glottal stop (/ʔ/); and that glottal /ʔ/ had been approximated in Jamaican Creole by velar /k/. Thus vernacular British English /liʔḷ/ had been approximated in Jamaican Creole by /likḷ/.
However my students--though the class is smaller than usual, they're a bright batch of students--quickly pointed out that pitter didn't undergo the change, and that therefore the conditioning environment appeared to be more restricted.
I went back and checked Cassidy :40 and found that in fact the conditioning environment does indeed seem to be restricted to instances in which /t/ immediately precedes /ḷ/ or /l/. Examples include /sekḷ/ (settle), /brikḷ/ (brittle), /wakḷ/ (wattle), and /wiklo/ (whitlow) [there is also an apparent example of /t/ > /k/ word-finally, in buck for butt (as in what goats do, not the body-part), but maybe this simply reflects semantic change of buck rather than sound change]. But this now looks like a rather unusual sound change. Why should dental/alveolar /t/ become velar /k/? And why particularly before /ḷ/ or /l/?
Even more perplexing is the fact that Cassidy :40 also states that /d/ becomes /g/ in the same environment: /figḷ/ (fiddle), /rigḷ/ (riddle), /ni:gḷ/ (needle), /kyaŋgḷ/ (candle), /haŋgḷ/ (handle), /dwiŋgḷ/ (dwindle) etc. Now the parallelism of voiced dental stop /d/ becoming voiced velar stop /g/ just as voiceless dental stop /t/ becomes voiceless velar stop /k/ is clear, but the reason for these two changes still seemed rather mysterious.
I think I now might have the beginnings of an explanation. Hallé & Best discuss the fact that French listeners tend to perceive illegal /tl/ and /dl/ clusters as /kl/ and /gl/, respectively, suggesting that /dl/, /tl/ undergo “phonotactic perceptual assimilation” to the phonetically most similar permissible clusters. Could this be what happened in Jamaican Creole?
[Update: My advisor Hans Henrich Hock informs me of some parallel changes in other languages involving /tl/ > /kl/, including:
Cassidy, Frederic G. 1971. Jamaica Talk: Three hundred years of the English language in Jamaica. London: Macmillan Caribbean, 2nd edn.
Hallé, Pierre A. & Catherine T. Best. 2007. Dental-to-velar perceptual assimilation: A cross-linguistic study of the perception of dental stop+/l/ clusters. Journal of the Acoustic Society of America 121/5: 2889-2914. [draft version here]
Hock, Hans Henrich. 1991. Principles of historical linguistics. Berlin: Mouton de Gruyter, 2nd edn.