Previous analyses of [jū]

As Halle and Mohanan (1985: 89) point out,

It is well known that the sequence [Cy] in English is regularly followed by the vowel [uw] or its unstressed reduced reflex. Thus, although [kyuw] Kew, [kyut] cute, as well as [kwiyn] queen, [kwæk] quack, [kwam] qualm, [kwowt] quote, etc., are well-formed, *[kyiyn], *[kyæk], *[kyam], *[kyowt], etc., are not.

There are two possible ways of dealing with this observation in a phonological description: either [j] is nuclear, making [jū] a diphthong; or it is inserted by rule in the onset, before the vowel or vowels which eventually surfaces as [ū] (or [uw]). SPE, Halle (1977), Rubach (1984) and Halle and Mohanan (1985) all adopt the j-Insertion approach; we turn to the possibility of a diphthongal analysis later.

Before considering the SPE analysis of [jū] in detail, I should point out that the sample words in (3.22) above can be split into four subsets. Some forms with surface [jū], like tabular and angular, alternate with base forms, in this case table and angle, in which there is no vowel corresponding to [jū] in the derived forms. In SPE, a rule inserting /ʊ/ in tabular, angular was proposed (see (3.24)); this procedure has generally been followed in subsequent studies.

In the second set of [jū] words, which includes ambiguous, ambiguity, credulous, credulity and habitual, the vowel surfacing as [jū] belongs underlyingly to a morpheme distinct from the stem. In Chomsky and Halle (1968: 195), this morpheme is taken to be the `stem-forming augment' [+ʊ], which is stored with certain lexical items and subsequently deleted word-finally but retained before affixes. The remaining words in (3.22) fall into two further classes; those in which [jū] alternates with [Λ], as in reduce ~ reduction or study ~ studious, and non-alternating forms like cube, argue, venue, huge and duke.

In SPE, surface [jū] always corresponds to underlying high back lax rounded /ʊ/ (=/u/ in SPE). /ʊ/ undergoes a rule producing tense, unrounded [Ɨ̄] (Chomsky and Halle 1968: 195), which provides the context for /j/-Insertion before being unconditionally re rounded; since [Ɨ̄] is [+back, -round], it will not undergo Vowel Shift. In order to meet the structural description of this rule, reduce, cube, huge, venue and so on have to be represented underlyingly as /re=dʊkε/, /kʊbε/, /hʊgε/ and /vεnʊε/, with the final /ε/ being disposed of later in the derivation. In tabular, where [jə] may surface rather than [jū], a further rule laxing unstressed / Ɨ̄/ is also necessary. In addition, to account for [Λ]in reduction and study, Chomsky and Halle are forced to allow lax /ʊ/ to undergo Vowel Shift, and to extend the structural analysis of the Rounding Adjustment rule to convert the resulting [o] to [Λ]. The same derivation, involving Vowel Shift, applies to [Λ] in profundity, although here the underlying vowel is tense /ū/, which undergoes Vowel Shift, Rounding Adjustment and Backness Adjustment to [āw] in profound, but laxes, shifts and unrounds in profundity.

Chomsky and Halle still encounter problems with [jū], [Λ]and[ʊ].The extension of Vowel Shift to lax/ʊ/ will convert all underlying cases of this vowel (unless they are first tensed and unrounded to[Ɨ̄]) into surface[Λ]; and indeed, this strategy is used in SPE to derive putt, fund, pun and so on. However, in push, pull, cushion, put and soot, which have surface[ʊ], a complex `lay-by' rule (ChomskyandHalle1968:204) unrounds certain cases of /ʊ/ to [ɨ̄] until Vowel Shift has operated, where upon [ɨ̄] is re rounded. Lay-by rules of this type have attracted a good deal of criticism (GoyvaertsandPullum1975);and quite a part from such general objections, the proposed rule `does not cover several exceptional cases of unrounding' (ChomskyandHalle1968:204), including put, pudding and cushion.

The SPE analysis of [jū] and related vowels suffers from one final problem; [j] has to be deleted by a later rule in certain dialects after dental sand palate – alveolars (ChomskyandHalle1968:231), giving [nū] new, [dūk] duke, etc. Here, however, Chomsky and Halle are missing a generalization; while some American English accents do indeed lack[j] after coronals (unless [ū] is unstressed), [j]never surfaces after/r w ʤ ʃ/, for instance, in any dialect. Some sample SPE derivations are given in (3.25).

Halle(1977)is largely are vision of the SPE analysis of[jū],[Λ]and[ʊ]. Halle restricts the VSR to tense vowels, although these need not be stressed, and reformulates j-Preposing to operate before ,or lax/Λ/ in an open syllable. Sample derivations are shown in(3.26).

Halle's underlying representations are in some cases more surface-true than those of SPE; in addition, he no longer requires the SPE Rounding Adjustment rule, and his derivations make more use of independently necessary tensing and laxing rules rather than specially formulated ones. On the other hand, he introduces two additional absolute neutralization rules and two non-surfacing, abstract underlying vowels, /ɨ̄/ and /Λ̄/ (which are additionally suspect in belonging to the cross-linguistically rare category of back unrounded vowels), in order to derive the reduce ~ reduction, study ~ studious and profound ~ profundity alternations via the Vowel Shift Rule. Halle also assumes that both ambiguous and ambiguity have underlying /Λ/, which in both cases undergoes Prevocalic Tensing and Vowel Shift. This derivation is possible only if the Vowel Shift is generalized to all tense vowels, regardless of stress, since [jū] is stressed in ambiguity but not in ambiguous. However, Halle's revised formulation of Vowel Shift has one major drawback; this concerns forms like various and managerial. The SPE derivations for these are given in (3.27).

As SPE restricts VSR for tense vowels to those which are also [+stress], Chomsky and Halle have no difficulty with the failure of /ו̄/ to shift in both various and managerial. Halle, on the other hand, does not indicate how these vowels are to be stopped from shifting. A late tensing rule might be suggested, but some cases of tensing must be ordered before Vowel Shift to provide a suitable input, as in Canadian or variety, and it does not seem feasible to extract any context from the main Tensing Rule and order it after Vowel Shift.

However, if the VSR is restricted to stressed vowels, Halle cannot derive [jū] from /Λ/ in ambiguous. His account is further compromised by the difficulty of deriving [jū] in words like habitude, credulity and credulous. These have the same augment as ambiguous and ambiguity, so that the same underlying representation, /Λ/, should be appropriate. However, neither CiV Tensing nor Prevocalic Tensing can operate in credulity, etc., so that /Λ/ cannot be tensed and shifted. Nor can Halle deal adequately with items like angular and tabular. Here, /Λ/ is inserted by rule and the second expansion of Halle's y-Preposing rule, which inserts /j/ (= /y/) before lax /Λ/ in an open syllable, will then operate. Since no tensing rule is appropriate in such cases, and the /Λ/ vowel is unstressed, Vowel Reduction subsequently produces [jə]. However, as Chomsky and Halle (1968: 197) observe, the pronunciation [tæbjələ(r)] is only one variant: we must also allow for `fairly careful speech, in which the medial vowel is rounded'. Yet Halle has no way of deriving phonetic [tæbjʊlə(r)].

Finally, Halle's y-Preposing rule itself (Halle 1977: 621) is problematic. This rule inserts /j/ (Halle's /y/) before all instances of tense /Λ̄/, and before lax /Λ/ in an open syllable. The restriction to open syllables is intended to exclude pun, luck, but and so on from y-Preposing. However, butter, fussy and mussel arguably have /Λ/ in an open syllable but no [j]. Rubach (1984: 36) observes that `the only way to exclude these words from j-Preposing is to posit underlying geminates. This is hardly a solution, since the geminates would serve no purpose other than to block j-Preposing.'

As Halle (1977) based his treatment of [jū] and related vowels on SPE, so Rubach (1984) in turn attempts to improve on Halle's study. Rubach retains some elements of Halle's analysis, such as the underlying /ɨ̄/ vowel in profound ~ profundity, but also makes some significant departures from the earlier work.

Like Halle, Rubach proposes /Λ/ as the underlying vowel in study ~ studious and Lilliput ~ Lilliputian, but /Λ̄/ in reduce ~ reduction, punish ~ punitive. Rubach consequently formulates his j-Preposing rule (1984: 32) to operate before tense /Λ̄/, inserting /j/ in reduce, studious, Lilliputian and punitive, but correctly excluding reduction, study, Lilliput, punish, pun, cut and so on. In addition, Rubach assumes that this rule will insert /j/ in certain non-alternating forms like mute, cucumber.

Rubach's main innovation concerns the augment in ambiguous and ambiguity and the inserted vowel in tabular, angular. Halle considers the augment to be /Λ/; this will undergo Vowel Shift and High Rounding. For tabular, Halle proposes /Λ/-Insertion, open-syllable y-Preposing, and Vowel Reduction. We have seen that the derivation of ambiguous and tabular cause difficulties for Halle: he must extend Vowel Shift to unstressed vowels to account for surface [jū] in ambiguous, and cannot produce a rounded medial vowel in tabular. Rubach acknowledges these problems, and proposes that VSR be once again restricted to stressed tense vowels. However, he is then forced to assign underlying /ʊ/ to ambiguous and ambiguity, and to insert /ʊ/ in tabular, where Vowel Reduction may then optionally apply to give [jə] or [jʊ]. These uses of /ʊ/ rather than /Λ/ present Rubach, in turn, with two problems. First, he must exclude /ū/ (and consequently /ō/) from the domain of VSR, to stop tensed, stressed /ʊ/ from shifting in ambiguity. The exclusion of /ū/ and /ō/ from Vowel Shift is of no great consequence: Rubach assigns profound ~ profundity underlying /ɨ̄/, and the lose ~ lost, shoot ~ shot, and fool ~ folly, school ~ scholar, food ~ fodder and poor ~ poverty sets of alternations are extremely small, and arguably unproductive. Further-more, experimental evidence considered earlier (Jaeger 1986, Wang and Derwing 1986) suggests that Modern English speakers no longer perceive these [ūw] ~ [ɒ] alternations to be part of the synchronic Vowel Shift pattern.

Secondly, because Rubach's j-Preposing rule only applies before tense /Λ̄/, [j] is generated in ambiguous, ambiguity and tabular by an additional rule of j-Insertion (Rubach 1984: 36) which applies before lax /ʊ/. The cyclic nature of this rule means it does not apply in underived put, push, bullet, soot and the like, but will insert /j/ in ambiguous and ambiguity, where /ʊ/ is an augment; in architecture, where /ʊ/ is part of the suffix /-ʊr/; and in tabular and angular, where /ʊ/ is inserted earlier in the derivation. Rubach's derivations are given in (3.28).

However, it is not clear how Rubach is to derive blue, rude, etc., which have the same surface [ūw] as ambiguous, reduce and cube but lack [j]. Conversely, Rubach admits that he is unable to generate [j] in words like copula and population (1984: 37), and has to assume that the glide is present lexically in these forms. Rubach also requires two rules, j-Preposing and j-Insertion, to perform what seems intuitively to be a single process, and his analysis still relies on absolute neutralization and the non-surfacing vowels /ɨ̄/ and /Λ̄/ in the derivation of profound ~ profundity, reduce ~ reduction, mute, tutor and cucumber (and presumably also cube and venue).

Halle and Mohanan (1985) retain substantially the same derivations as Rubach for the profound ~ profundity, reduce ~ reduction and study ~ studious alternations. They also derive [Λ] in gun, but, etc. directly from /Λ/, and [ʊ] in put, push from /ʊ/. However, their treatment of the [yūw]/ [jū] sequence in non-alternating forms like cube, music, residue, avenue, statue and venue departs considerably from previous analyses, primarily because their version of Vowel Shift is restricted to long, rather than tense vowels. Like Halle (1977), Halle and Mohanan drop the requirement that vowels should be stressed in order to shift; to account for various ~ variety, impious ~ pious and maniac ~ maniacal, they consequently propose an ad hoc rule of Prevocalic Lengthening (a process quite distinct from the remarkably similar Prevocalic Tensing), to lengthen the stressed vowel in certain lexically marked words. Halle and Mohanan also propose that the English Main Stress Rule should be made sensitive to vowel length: it follows that the presence or absence of stress can be one indicator of underlying vowel length, and therefore of the eligibility of a vowel for Vowel Shift (recall that Halle and Mohanan order VSR on Level 2, where it is not subject to DEC). Halle and Mohanan rely on this supposed interdependence of vowel length, stress and VSR to argue that the vowels which surface as [jū] in (3.29) and (3.30) `cannot be identical in underlying representation, but become identical (save for stress)' due to Vowel Shift (1985: 90).

The argument which leads to this unexpected conclusion runs roughly as follows. In (3.29), the word-final vowels are stressless and must therefore be underlyingly short; [jū] cannot, therefore, be derived via Vowel Shift, and the underlying vowel must be [+high], since the surface vowel is [+high]. Halle and Mohanan propose underlying /ɨ̄/, which will subsequently undergo Stem-Final Lengthening and Tensing. However, in (3.30), the vowel surfacing as [jū] `is long and must therefore have undergone Vowel Shift. Since [yūw] is [+high], its pre Vowel Shift source must be [-high]' (Halle and Mohanan 1985: 90). They conclude that, in (3.30), [yūw]/[jū] is derived from /Λ̄/, which will shift to [ɨ̄]. y-Insertion (Halle and Mohanan 1985: 90) is formulated to operate before high back unrounded [ɨ̄] and [ɨ]. Lax [ɨ] must then be lowered in closed syllables, to give surface [Λ] in sulphur, profundity and so on, while lax [ɨ̄] in open syllables and tense [ɨ̄] in all cases are rounded. One final extra rule of ɨ-Lengthening, which applies to stressed short /ɨ/, is also posited to account for [jū] in sulphuric. Derivations for the profound ~ profundity, reduce ~ reduction, study ~ studious and sulphur ~ sulphuric alternations, and for cube, revenue and venue, are given in (3.31).

Halle and Mohanan's account of [jū] and the alternations in which it is involved must surely be the most complex and least satisfactory of the post-SPE studies considered here. Halle and Mohanan's underlying vowel system contains more non-surfacing vowels, i.e. /ɨ/, /ɨ̄/ and /Λ̄/, than those of either Halle (1977) or Rubach (1984), and Halle and Mohanan also require more additional rules, in the form of ɨ-Lowering, ɨ-Lengthening and ɨ-Rounding, to dispose of these non-surfacing segments. Their logic in assigning different final underlying vowels to revenue, avenue and residue on the one hand, and venue and statue on the other, also seems flawed, for two reasons. First, there seems no distinctive difference in stress between the final vowel of venue and that of avenue, yet stress is Halle and Mohanan's major motivation for arguing that the first is underlyingly short and the second long. Secondly, although Halle and Mohanan assert that the final vowels of venue and avenue, as well as the stressed vowel of cube, `become identical (save for stress)' (1985: 90) during the course of the derivation, a careful consideration of their ordered list of rules (1985: 100) shows that this cannot be so: [yūw] can indeed be derived from /Λ̄/ in cube and avenue, via Vowel Shift, y-Insertion, Diphthongization and ɨ-Rounding, but there is no way of deriving [yūw] in venue, statue, etc.

The venue vowel can, however, surface in two different ways, according to dialect. In Halle and Mohanan's Dialect D, final /ɨ/will undergo y-Insertion and postlexical ɨ-Rounding. However, since Dialect D shows no evidence of Stem-Final Tensing (Halle and Mohanan 1985: 59), /ɨ/ cannot be tensed. Nor can it be lengthened stem-finally, since Stem-Final Lengthening (Halle and Mohanan 1985: 61) affects only tense back vowels in dialects other than B. In Dialect D, then, the word-final vowel in venue will surface as short high lax [jʊ]. In Dialects A, B and C, /ɨ/ in venue will have [j] inserted, and will then be eligible for Stem-Final Tensing and Lengthening and post-lexical ɨ-Rounding. However, although this will allow for surface [jū], the vowel cannot then undergo Diphthongization to produce Halle and Mohanan's [yūw], since Stem-Final Lengthening is a Stratum 3 rule but Diphthongization, which applies to long vowels, applies on Stratum 2.

It is clear, then, that Halle and Mohanan cannot derive [yūw] vowels, `identical (save for stress)' (Halle and Mohanan 1985: 90) in venue, statue, cube and avenue. It seems also that they will find difficulty in deriving [yūw] in ambiguous and ambiguity (which they mention only very briefly) and in tabular (which they do not mention at all). To take tabular first; if /Λ/ is inserted, this cannot undergo Vowel Shift to [ɨ] since the medial vowel is unstressed and must therefore be underlyingly short. If /ɨ/ is the vowel inserted, it can attract /y/ and undergo ɨ-Rounding, but cannot be lengthened, tensed or diphthongized. As for ambiguous and ambiguity, the only possible underlying vowel is again /ɨ/ (see (3.32)).

Again, [yūw] cannot be derived, since Diphthongization affects only long vowels, and Halle and Mohanan propose a rule of Prevocalic Lengthening only in a few lexically marked words such as variety, maniacal and pious. Even if Prevocalic Lengthening were permitted, ambiguity would require underlying /Λ/, since the tensed, stressed, long vowel otherwise resulting could not be excluded from Vowel Shift. It seems that the best we can do in Halle and Mohanan's system is to derive [jʊ] in tabular and [jū] in ambiguous and ambiguity, but as the surface facts demand [jū] (Halle and Mohanan's [yūw]) obligatorily in ambiguity and at least optionally in ambiguous and tabular, the best is clearly not good enough.