I will discuss here an often misanalyzed part of the morphophophonology of English: the allomorphy of the plural suffix. Conventional (that is to say, boring) analyses hold that the suffix must be underlying /-z/ rather than underlying /-s/. An analysis with underlying /-s/ would require a rule of voicing after voiced segments, claimed to be undesirable based on words such as false. Merely describing this line of argumentation makes me physically sick.
It should be clear to anyone with a basic training in linguistics that underlying /-s/ is a preferable analysis for empirical, theoretical, typological, and metacontrarian reasons. /s/ is evidently the least marked of the English sibilants: /z/ has the wrong voicing value, /ʃ/ the wrong place features, and /ʒ/ combines the worst features of both.1 At least a handful of other linguists out there have had the sense to come to the same conclusion as me: in ![[redacted]](/images/b.gif "[redacted]")
2 et al.’s (2005) analysis of the European Portuguese plural suffix, they ingeniously argue for underlying /-s/ based solely on /s/ being the most frequent sibilant in the world’s languages. How refreshing to see some proper respect for markedness! Notice that their analysis is entirely unaffected by the fact that the plural can surface as literally any Portuguese sibilant ([z, ʃ, ʒ]) other than [s]. The authors evidently treat considerations like the actual surface forms as so unimportant that they feel no need to mention them. Needless to say, I am humbled to follow in the footsteps of work such as theirs.
Next we turn to a data point which has been overlooked in earlier work on the English plural. This oversight is surely not coincidental, as the datum is fatal to the underlying /-z/ dogma which plagues our field. Consider the perfectly ordinary English noun die, and the predictions made for dice based on the underlying shape of the plural suffix:
Singular
[daɪ]Plural with /-z/
**?!??!![daɪz]Plural with /-s/
[daɪs]
It is clear that only underlying /-s/ generates the correct pronunciation of dice.3, 4 Those who refuse to let go of the /-z/ analysis that has been forced down their throats may ask why day surfaces as day[z] rather than *day[s] in the plural. Evidently there is morphologically-driven /-s/ voicing in such words. The morphological status of the voicing explains why it admits exceptions such as die. Moreover, purely phonological voicing of /s/ to [z] in word-final position would be impossible. It is of course uncontroversial that typological facts about vocal fold vibration (which may seem to lend themselves to a physical explanation) are literally encoded in the genome of every member of our species.5
For the reasons outlined above, we can confidently conclude that the English plural is /-s/, and that any other analyses are fundamentally misguided. Those interested in further implications of this work are invited to The House6 in April, to witness me yelling my conclusions from a rooftop until the /-z/ ringleaders get me.
1 It is no accident that for many speakers, orthographic ʒ represents /z/ (but see Note 3).
2 The names of these brave linguists are not published here, in anticipation of the hate mail they would undoubtedly receive from /-z/ clan members.
3 Here, as is often the case, the orthography misrepresents the phonological facts (but see Note 1). Compare well-known cases such as kill instead of the actual makebeginnotbeingalive which sometimes surfaces in L1 acquisition (anecdotal evidence; due to the phonological system of the child in question, the actual surface form was [kʰɪɫ]).
4 The pattern is productive: witness mice and lice, although here, the connection with the words my and lie is not as obvious.
5 In forthcoming work, I argue that the crucial genetic mutation banning word-
6 You know the one. You KNOW the one. YOU KNOW THE ONE.