/str/-->/str/: Assimilation at a Distance?

In this journal, Michael Shapiro (1995) has argued that the environment of an innovating change of [s] to [∫] in American English is /str/, with the change not observed in /st/ clusters lacking /r/, and that /t/ undergoes no change QUA PHONEME in the cluster (i.e., is not affricated to /č/). From this he concludes that "/s/ changes to /∫/ owing to the presence of /r/" (103) and observes that "if assimilation it be then it is not an example of contact assimilation but of assimilation at a distance." In this paper I shall demonstrate that this innovation, although assimilation, is not assimilation to /r/ and that it need not be viewed as assimilation at a distance.

The sound change /str/ → /∫tr/ is, as pointed out by Shapiro (1995), geographically widely distributed. It is found in New Zealand (all examples herein are from New Zealand English speakers), and for speakers who evince the change it applies not only word-internally (straight, strategy, stronger, instruments, Australia) but also across word boundaries (race track, last race).¹

Shapiro (1995, 103) notes that "there is a relevant phonetic similarity between [∫] and [r]," in that they are both palatal. Such a similarity is what accounts for /s/ becoming [∫] when it precedes /r/ in the same syllable, as in, for example, Sri Lanka.² In the speech of some, this process has been generalized to apply over syllable boundaries, giving such pronunciations as I[ʒ]rael (Shapiro 1995, 105 n. 5) and cla[∫]room.

An innovative pronunciation of /s/ as [∫] is also attested before [t∫]. This applies to all sequences of /stj/, typical examples being moisture, question³ (word-internal) and last year, best yet (straddling a word boundary), these being /stj/ → /st∫/ → /∫t∫/, both historically and synchronically.⁴ The speakers whom I have observed using /∫tr/ also use /∫t∫/. I have observed one speaker who consistently uses /∫tr/, but notably he does not palatalize the first [s] in last race, although other speakers do. What is significant here is that in this speaker's pronunciation, the /t/ is elided, giving la[sr]ace. This speaker palatalizes /s/ in the sequences /st∫/ (from /stj/) and /str/, but not in heterosyllabic /sr/. Because we would not expect to have assimilation at a distance (i.e., across /t/) without also assimilation of the same target when adjacent to the trigger, this shows that /str/ → /∫tr/ cannot be conflated with /sr/ → /∫r/. Can /str/ → /∫tr/ be conflated with /st∫/ → /∫t∫/? I argue that it can be.

Shapiro (1995) argues that, in the sequence /str/, /t/ does not affricate. This claim is based in part on the fact that Olive, Greenwood, and Coleman (1993, 278–83) do not mention affrication in this environment. This fact, however, is not relevant, as the /str/ sequence analyzed by Olive, Greenwood, and Coleman is not an instance where /s/ has become [∫]. What should be considered is the pronunciation of /t/ in the sequence /∫tr/ in the innovating pronunciation. My own observation is that /t/ is always affricated in cases where /∫tr/ is used, and this is confirmed by comments by an informant who uses the innovating pronunciation.⁵ Shapiro (1995, 103) also makes the observation that "[str-] does not become because of the well-known effect of initial-sibilant-plus-stop to deaspirate the stop." However, Olive, Greenwood, and Coleman (1993, 288) note that unaspirated /t/ frequently becomes alveolar-palatal and affricated when followed by the glide /j/, showing that lack of aspiration does not necessarily inhibit affrication.⁶

We now have innovative pronunciations of /str/ → /st∫r/ → /∫t∫r/ (e.g., race track) and /stj/ → /st∫/ → /∫t∫/ (e.g., last year). Both of these processes include the step /st∫/ → /∫t∫/.

In Russian there is a regular process which changes /sč/ to /šč/ [∫^jt^j∫^j] (or in the pronunciation of many speakers, to [∫^j∫^j]) (Boyanus 1935, 92; Jones and Ward 1969, 203). This process applies across word boundaries, as in bez četverti 'minus a quarter', which becomes beščetverti (Panov 1968, 90), and across clitic-word boundaries, as exemplified by s-čaem 'with tea' becoming ščaem. Jones and Ward class this as...