English onsets with /s/ and sonority hierarchy

Question

Why are clusters such as /sk/ (as in sky), /st/ (stop), and /sp/ (spill) allowed as onsets in English? The sonority decreases in these clusters and does that not violate the phonotactic rules?

On a related note, it seems intervocalic clusters such as /ks/ are divided into two syllables instead of counted as the onset of the succeeding syllable: Anaximenes is usually syllabified as /ˌæ.nækˈsɪ.məˌniːz/ instead of /ˌæ.næˈksɪ.məˌniːz/, for example.

Perhaps the latter observation could be explained by rejecting the maximal onset principle, but when combined with the first observation, it seems to show that there is something special about /s/ (and perhaps sibilants in general) in English. I had trouble finding any explanation for this phenomenon, however.

Any insight is appreciated! Thanks for reading my question.

Answer 1

Syllables do not have to conform to the notion of "sonority sequencing", which is only a rough approximation of crosslinguistic syllable structure tendencies. It is generally recognized (following decades of argument by Ohala, also see Wright 1996) that abstract "sonority" is not particularly relevant in explaining segment sequence tendencies, and what matters is whether a sequence is easily parsed. The desirability of onset [tr] relative to [rt] or [tw, nw] relative to [wt, wn] can be explained by the fact that the former lead to acoustic landmarks which make it easy to identify the segments (in an onset), but the latter are not (for instance, [w] has very weak acoustic cues, no striking release, low amplitude, but a strong influence on vowel formants, so [w] "should" be next to a vowel). Sequences of the form [sC] where C is not a fricative are fairly easy to parse acoustically – though not trivial, so they might be allowed, or they might be disallowed. Sequences of the form [ʃC] would also be good, but having a constrast between [sC] and [ʃC] is not so good (because it is harder to distinguish the difference between [s] and [ʃ] before a consonant).

Since [ks] is not a possible onset in English, there is only one possible syllabification of /VksV/: [Vk.sV]. The maximum onset principle does some work for you, in explaining why /VstV/ syllabifies as [V.stV] in English, not [Vs.tV].

Answer 2

The status of /s/ and sC clusters does indeed pose a problem for phonotactics. /sp, sk/ violate the sonority profile; /sl/ violates constraints against identical place of articulation (POA) in onset; /st/ violates both; and /sn/ violates three phonotactic constraints: sonority, POA, and sonority distance of onsets (other Cn onsets are not permitted in English). Furthermore, sC clusters permit different POAs in 2nd position e.g./sp, st, sk/; whereas POA is restricted in other onsets clusters, mostly coronal /r,l/ e.g. /pr,pl,tr,kr,kl/. In all these ways, /s/ behaves in a special way.

This has led some people to propose that /s/ is an appendix linked to a level higher than the syllable in prosodic structure (e.g. Goldsmith 1990). In the Government Phonology framework, /s/ in sC is considered to be a coda. This view is recently developed by Goad (2012) "sC Clusters are (almost always) coda-initial". Goad argues against the perceptually-based account by pointing out that cross-linguistically s+stop are the most common sC clusters. The perceptual cues to C in sC will be compromised by /s/; therefore, we would expect the most common sC to have a C with robust internal cues such as s+liquid.

Regarding /ks/: according to English phonotactics, this would be syllabified as /k.s/; otherwise, it would violate the sonority distance principle. In English, onset profiles cannot consist of stop+fricative or stop+nasal.