Double codas

Double non-suffixed English codas can be generalized in the following fashion:

(a) C1 is a nasal and C2 is an obstruent (no voiced obstruent permitted except /d, z, dZ/). Nasals (C1 ) combining with stops (C2 ) are invariably homorganic. (b) If C1 is /s/, then C2 is a voiceless stop. (c) C1 is a liquid (/l, @/) and C2 is any consonant except for /z, Z, D/. Also non-existent is the /lg/ cluster. (d) If C1 is a voiceless non-alveolar stop (/p, k/), then C2 is a voiceless alveolar obstruent (/t, s/). Also permitted is the /ft/ cluster.

Possibilities increase considerably if we add to these the clusters created by the suffixes with /t, d, s, z, θ/ (past tense, plural, possessive, ordinals, etc.).

As stated earlier, the sonority sequencing principle dictates the opposite of onset sequencing for codas. This means that optimal codas should have the sonority level dropping as we move from C₁ to C₂. Indeed, this is the case for the double codas we find in non-suffixed (monomorphemic) forms in English (e.g. arm[ɑɹ̣m], sharp[ʃɑɹ̣p], belt [bεlt]). Exceptions are (a) two-stop sequences, which are never homorganic (e.g. apt[æpt], act[ækt]), and (b) stop +/s/, which always agree in voicing (e.g. lapse[læps], tax[tæks]).