As previously noted in section three, /ju/ is formed by a palatal approximant and a close back vowel. However, the question remains whether the sequence is diphthongal /i#u/ or an example of an onset /j/ plus a nucleus /u/.
Two arguments in favour of a diphthongal analysis follow evidence from speakers who realise /j/ in items such as lewd where the onset /lj/ violates a constraint known as Minimal Sonority Distance, MSD, (Borowsky 1984:6, Roca and Johnson 1999:257) and the patterning of diphthong~simple vowel alternations in derivationally related forms.
(7.1) a. From Gimson (2001:97)
/aI/~/I/ divine ~ divinity deride ~ derision
/i:/~/e/ hero ~ heroine sheep ~ shepherd
/eI/~/Ï/ sane ~ sanity profane ~ profanity
/«U/~// joke ~ jocular phone ~ phonic
/aU/~/¿/ south ~ southern abound ~ abundant
/i#u/~/¿/ seduce ~ seduction reduce ~ reduction
However, if /ju/ is taken as a diphthong, its structure would contrast with the other recognised diphthongs in English whose sonority falls from the first to second element. Using the sonority scale proposed by Ladefoged (2001:228), we can observe that the sonority of /i#u/ rises from the first to second element, reversing the sonority ordering of other English diphthongs.
Further arguments in favour of /j/s onset status (in addition to those mentioned in section three) stem from observations that, although speakers may violate MSD through [lju:d], they tend not to violate the Binarity constraint (which restricts onsets to two consonantal segments) by producing *[blju:]. Since both onset slots are occupied by [bl], there is no available space for [j]. Thus, it will fail to be licensed and not surface (Roca and Johnson 1999:277). However, this situation is not necessarily applicable to, for example, the /slj/ or /stj/ clusters in [slju:] and [stju:] if it is assumed that the [s] is attached directly to, and therefore licensed by the syllable node or the prosodic word node bypassing the intermediate onset node.
However, this appears to be challenged by the form [blju: r«U] Blue Row which was produced by two older Mersea males. The phrase relates to an area of the island which featured prominently in their childhood memories. Although both speakers used this phrasal form, one did not render [blju:] as an adjective, preferring [blu:], while this comparison was not applicable to the second speaker. Therefore, the occurrence of [blju:] in this context may be attributed to an historical, or relic form within the dialect (as a remnant of EYD after /l/) and preserved by its social significance as a native islander marker, as opposed to evidence of an underlying diphthongal representation.
American Englishes and some British English dialects provide further evidence for the onset /j/ by allowing /j/ to follow labials and dorsals but not coronals, although some evidence of hypercorrection, such [njun] noon, can be detected (Wells 1982b:489). One explanation of this inconsistency is attributed to the effects of the Obligatory Contour Principle (OCP) which recognises the tendency to disallow adjacent similar elements from all tiers but the skeleton, within a common domain or constituent (Roca and Johnson 1999:695). Thus, in this context, the OCP ensures initial onset clusters do not share a similar place of articulation. If all vowels (and thus /j/~/I#/) are considered as a coronal rather than dorsal, it explains why these dialects disfavour the word-initial onset sequences in new, tune, dune etc. However, it is not the case for word-medial coronal+/j/ clusters particularly after sonorants and unstressed syllables:
(7.2) From Borowsky (1986:289)
nb Dialect A refers to American
Dialect B refers broadly to British
Although there were no examples in the Mersea data of preceding-/r/, this pattern was produced with respect to preceding-/l/. For example, [j] was retained in va!lue but not in dilu!te. However, out of twenty mono-morphemic tokens containing preceding-/n/ (with respect to the older generation), there were no items in which /n/ was the onset of a stressed syllable. Therefore, regarding unstressed syllables such as me!nu and ge!nuine, sixteen of the twenty tokens were realised as [nju] while four (all renderings of avenue) were produced with [nu].
An explanation advanced by Borowsky is through a rule of y-insertion (1984: 4-5 and 1986:281). Through this rule, the initial part of /I#u/ is detached and regarded as stray until later resyllabification takes place and assigns it to the onset as /j/ (or /y/), unless blocked by an intervening rule such as OCP, and it is subject to stray erasure. In addition, Borowsky (1986) concludes that there are no prohibitions to Cj at the beginning of unstressed syllables due to the resyllabification of the preceding C to the previous syllable. This provides the syllable in question with a coda, leaving the following onset free for the unattached /j/. By way of illustration, menu would originally be syllabified as [mE.nI#u]. The [I#] subsequently becomes stray before being reassigned to the onset after the [n] itself has been resyllabified into the coda giving [mEn.ju]. The resyllabification process is reinforced as stressed syllables tend to be bimoraic and since lax vowels have one mora and tense vowels have two (Hammond 1999:105), the appropriation of, for example, [n] and [l] in menu and value helps to achieve this target.
The effects of the OCP or *[α F] [α F] filter (Borowsky 1984:7) on preceding consonant rankings extracted from the Mersea data can be seen below:
(7.3) nb - /T/ will be excluded from discussion as its result is only based on one token.
(/T/>) /st/> /s/> /b S/> /f/> /n/> /v/> /h/> /p/> /m/> /k/> /l/> /g/
The presence of [st] and [s] as the most likely to promote yod-dropping is expected as the elements adjacent to the yod are coronal. Even though the [st] cluster features a preceding [t], these tokens are included in this data set due to /§/s inability to function as part of a complex onset. This is formalised by Hammond as the Affricate Proscription which states that affricates do not appear in complex onsets (1999:55). Hammond also notes that [s] can cluster with just about any following consonant except voiced-obstruents, affricates and [r] (1999:55). Indeed, while evaluating the data, there were no audible instances of *[s§u] validating [st]s inclusion in this ranking.
If we were to believe Trudgills (1974) proposed ranking given in (4.3) above, and the effects of the OCP, the next highest ranked consonant would be the coronal /n/ (as /t d/ are being disregarded in this ranking). However, although the Mersea data favours the coronal /S/ above /n/, it also promotes the labials /b f/ as more likely to cause yod-dropping. The position of /b/ is similar to that of Spurling (2004), although the two data sets contrast with respect to /f/.
Palatalisation of /j/
Further evidence for the onset status of /j/, regarding Borowskys (1984, 1986) analysis comes from the palatalisation of English coronal obstruent+/j/ clusters. Borowsky claims that palatalisation, as an independent rule, applies when the following vowel is unstressed and that the palatalised segment must be a syllabified onset /j/ and not a nuclear /i/ (1984:9). In addition to asserting that the palatalisation rule is only applicable to unstressed syllables, Borowsky claims that it may operate only when the coronal and the /y/ are heterosyllabic (1984:10) and thus, tautosyllabic coronal+/j/ (such as residue) would not be palatalised in British English. However, with respect to the Mersea data, this appears not to be the case, as forms such as due [½u:] and Tuesday [§us.deI] were produced.