Doer’s Profile—Evan Smith Lingua Pranca Contents Note on the peH3 Root in French—Andrée Borillo, et al.

On the Go ~ Went Alternation

A Contribution (?) to the Generative Phonology of English*

Bernard Comrie
University of Cambridge

Most theories of phonology (morphophonology, morphology) allow for suppletion, whereby the allomorphs of a morpheme are completely unrelated phonetically. In discussing English phonology, one example of suppletion that is frequently cited is go [gōw] ~ went [went]. Within generative phonology (taken here essentially in the sense of Chomsky & Halle (1968)), one attempts to set up a single underlying representation for all allomorphs of a morpheme. The fact (if it is a fact) that suppletion exists, i.e. that certain allomorphs of a single morpheme do not have a single underlying representation, can have a cataclysmic effect on this principle of generative phonology, since once one allows distinct underlying representations for the root of go and went, one is already on the slippery slope leading to such heresies as denying the underlying representations for such obviously related pairs as gonad ~ nativity or fart ~ petard. In this paper, we shall argue that the root morpheme of go/went does indeed have a single underlying representation. The rules required to derive the surface phonetic forms are either (i) rules that are independently required by the analysis of English phonology presented in Chomsky & Halle (1968), or (ii) natural phonological rules, such as occur in a wide range of genetically, areally, and typologically unrelated languagesindeed these rules are so natural that we would be surprised if English did not have them. At several points, we shall note that a generative phonology of English incorporating our analysis of the go ~ went alternation is superior, on a variety of independent grounds, to one that does not incorporate our analysis.

The two aspects of the go ~ went alternation that might seem to give the most difficulty are the alternation of the initial consonants [g] ~ [w], and the alternation [ōw] ~ [en]; the final [t] of [went] is readily analyzable as the past tense ending. We propose to handle the [g] ~ [w] alternation by providing an underlying segment that combines features of both [g] and [w]. (Compare Chomsky & Halle’s analysis of the [i] ~ [ǣy] alternation with underlying /ī/, which combines the quality of [i] with the tenseness of [ǣy].) The obvious candidate is [gw]. This segment is already required in English phonology (Chomsky & Halle 1968:223-4), in the underlying forms of words like language; moreover, it forms a voice pair with /kw/ of queen and delinquent. Chomsky & Halle mention only the development of /gw/ to [gw], so it might seem that some kind of exception mechanism must be introduced to develop /gw/ to [g] in go and [w] in went. But this is illusory. In English, phonetic [kw] (from /kw/) occurs both initially and medially (queen, delinquent); however, [gw] (from /gw/) occurs only medially (language), and never initially, apart from obviously unassimilated loans like Gwen and guano, which must be treated as exceptions. We must obviously look for an explanation for the asymmetry in the distribution of [kw] and [gw]; there is no reason why this should reflect an asymmetry in the distribution of underlying /kw/ and /gw/, rather it suggests that underlying initial /gw/ undergoes some development other than [gw]. The pair go ~ went shows us what this other development is: either to [g] (with loss of distinctive roundness, before rounded vowels) or to [w] (with loss of the velar component elsewhere):

(1)   gw → 
g / # ----  
w / # ----
    [gw Development]

In analyzing the [ōw] ~ [en] alternation, we shall first concentrate on the fact that there is an alternation between a diphthong (deriving, of course, from an underlying tense vowel) and a sequence vowel–nasal, i.e., at a slightly more abstract level of analysis, between V and V N. The obvious solution is to posit an underlying nasal vowel, such that /Ṽ/ will give [V N] in certain environments and [V] in others. On the basis of [gōw] ~ [went], we might suggest, provisionally, that nasalized vowels give a sequence of oral vowels plus nasal before a plosive (the nasal being, of course, homorganic with the plosive), and simply an oral vowel elsewhere. It might be objected to this analysis that nasalized vowels do not occur elsewhere in the underlying representation of English morphemes. But this objection is illusory. There is independent evidence from the alternation stand [stænd] ~ stood [stud] for V N ~ V alternations in English, i.e. independent evidence for underlying nasalized vowels realized phonetically as either an oral vowel plus nasal or simply as an oral vowel. The introduction of stand ~ stood, however, calls into question our original statement of the environment (V N before plosives, V elsewhere), since the vowel stands before d in both forms. Note, however, that both stand and went contain a front vowel, whereas both stood and go contain a back vowel. This can hardly be coincidence, and suggests that one relevant factor is the quality of the vowel; the following plosive is, of course, also relevant, since a homorganic nasal can only appear if there is a following plosive, from which it takes its place of articulation. This suggests the following rules, in schematic form:

 → V N / 
     (Nasal Epenthesis)
(3)   V → [-nasal] (Denasalization)

We may note that both the processes discussed so far—loss of the labial or velar component of a labiovelar/labialized velar, and development of a nasalized vowel to either an oral vowel or an oral vowel plus nasal consonantare natural phonological processes, and we should not be surprised to find that English has them. Quite the reverse.

We have already suggested that the final [t] of [went] is the past tense ending. However, except after voiceless consonants (with [-t]) and alveolar plosives (with [-id]), the past tense ending in English is usually [d], not [t], so it might seem that here we have another ad hoc explanation to the hitherto formulated rules of English phonology. Again, this is illusory. There are several verbs which have, in the present tense, the final sequence tense vowel plus nasal, and in the past tense the final sequence lax vowel plus nasal plus [t], at least as a possible form, e.g. dream [drīm] ~ dreamt [dremt], lean [līn] ~ leant [lent]; examples with sonants other than nasals are likewise found, e.g. kneel [nīl] ~ knelt [nelt], feel [fīl] ~ felt [felt]. The short vowel and final [t] of went is just another manifestation of this phenomenon, and therefore merits the same analysis. At an underlying level, and for the moment just using V as a symbol for the vowel, we have /gwV̄n # d/. A lexically specified set of verbs undergoes a readjustment rule, Boundary readjustment, which (a) replaces the word boundary # by a formative boundary + (Chomsky & Halle 1968:369-70), and (b) replaces /d/ by /t/. Thus /lēn # d/ becomes /lēn + t/, and /gwV̄n # d/ becomes /gwV̄n + t/. This same readjustment rule applies to such past tense forms as kept, though here the devoicing of the past tense formative is vacuous, as [t] would in any case be the result after a voiceless consonant other than an alveolar stop.

The sequence V̄n + t now provides input to a general rule, Laxing, which laxes vowels before certain consonant clusters, including nt across a formative boundary (Chomsky & Halle 1968:172), giving /gwVn + t/.

The only remaining problem is the qualitative alternation between [ōw] of go and [e] of went. Even this has a ready explanation, utilizing only rules required independently, if we assume that this segment is underlyingly /ɔ̃̄/, which by Denasalization will give /ɔ̄/, and by Laxing /ɔ/ in the past tense, i.e. intermediate representations /gwɔ̄/ and /gwɔn + t/. The regular development of /ɔ̄/ is [ōw], by Diphthongization and Vowel Shift (Chomsky & Halle 1968:188), so that we have now fully accounted for the derivation of [gōw] from /gwɔ̃̄/. In the derivation of went, we still have to change /ɔ/ into [e]. The solution to this apparently arbitrary alternation is contained in Chomsky & Halle (1968:201-5, 209-10), who discuss two minor rules of English phonology (i.e. which are restricted to certain lexical items), namely Backness Adjustment and, in its application to lax vowels, Vowel Shift. (Vowel Shift is, of course, a major rule in its application to tense vowels.) Backness Adjustment accounts for alternations between back and front vowels as in bind [bǣynd] /bīnd/ ~ bound [bāwnd] /būnd/ or foot [fut] /fōt/ ~ feet [fīt] /fēt/. They formulate the rule as follows (209):

(4)   V → 
-α back
-α round
α back

This would shift underlying /gwɔn + t/ to /gwæn + t/, the low back rounded vowel /ɔ/ becoming the low nonback unrounded vowel /æ/.

Vowel Shift of lax vowels accounts for such alternations as sit [sit] ~ sat [sæt]: just as /ī/ gives /ǣy] by Diphthongization and Vowel Shift, so /i/ would give /æ/ by Vowel Shift, though with lax vowels Vowel Shift applies only to a handful of forms. One of these forms is went: just as /ǣ/ gives [ēy] by Diphthongization and Vowel Shift, so /æ/ would give [e] by Vowel Shift. Thus input /gwæn + t/ to this rule would give output /gwen + t/, precisely as required. The reader might feel that the application of two minor rules (Backness Adjustment, Vowel Shift) in the derivation of a single form is just a notational concealment of an ad hoc analysis. However, Chomsky & Halle 1968:209) advocate just this analysis for ran, from underlying /run/, becoming /rin/ by Backness Adjustment, then /ræn/ by Vowel Shift. It can scarcely be coincidence that this is also a past tense form (moreover, also of a verb of motion)! In Chomsky & Halle’s discussion, the application of both rules in the derivation of ran may seem strained, but we have now provided independent motivation for accepting the validity of this kind of analysis. SPE plus our analysis of the go ~ went alternation is clearly a more coherent and cogent whole than SPE without it.

Finally, we summarize the proposed derivations of go and went:

(5)    gwɔ̃̄       gwɔ̃̄ # d      
gwɔ̃̄ + t Boundary Readjustment
gwɔ̃ + t Laxing
gwɔ̃n + t Nasal Epenthesis
gwɔ̄ gwɔn + t Denasalization
gwæn + t Backness Adjustment
gwɔ̄w Diphthongization
gwōw gwen + t Vowel Shift
gōw wen + t gw Development


* Michael Covington, who brought to the author’s attention the proposed anthology of linguistic levity and ludicrosity, is hereby thanked and absolved of all responsibility.


Chomsky, Noam & Halle, Morris. 1968. The Sound Pattern of English. New York: Harper and Row.

* * * * * * * *

A New Approach to Economy in Generative Phonology

A → B / C ___

I → J / V ___
   E → Bj / Č ___

cf, Skt., Fr. aie.

Doer’s Profile—Evan Smith
Note on the peH3 Root in French—Andrée Borillo, et al.
Lingua Pranca Contents