Conspiracy in Eleme Phonological Operations

1

Conspiracy in Eleme Phonological OperationsIsaac Eyi Ngulube

English studies, University of Port HarcourtEmail: isaaceyingulube@yahoo.com;Mobile: 08075220804

AbstractThe discussion of phonological processes affecting segments attempted here highlights thefunctional unity of phonological rules that are driven by the enforcement of Eleme syllabletemplate. I shall demonstrate here that in Eleme the syllable template prohibit coda:NOCODA *C] (‘syllables are open’), ONSET *[ V (‘syllables must have onsets’),*COMPLEXONS *[ CC (‘onsets are simple’), *M/V (V cannot associate with marginnodes), *P/C (C cannot associate with peak nodes) and *VVV sequences, while permitting:V, CV. This paper argues that the phonological operations discuss here are the differentrepair strategies which Eleme adopts in other to attain the output goals enforced by thetemplate. As Kisseberth (1970) observes ‘this reoccurrence of a common output factorwhich guides different rules, without being explicitly stated in the rules, is calledConspiracy’ (cited in Kager 1999, p. 56). In sum, Eleme data lends credence to the validityof the conspiracy theory with the implication that Eleme phonological operations areenforced by the syllable template.

1. IntroductionThe Eleme language is spoken to the east of Port Harcourt, the capital of Rivers State

in southern Nigeria. Eleme is located between 70 10 and 70 15ˈ E and 40 35ˈ and 40 60ˈ N(Ngulube 2011). Eleme is a language in the Ogonoid group, which Bond and Anderson(2005) classified as: Niger-Congo; Benue-Congo; Cross River; Delta Cross; Ogonoid;Eleme.

In writing this paper, I draw on Optimality Theory because it offers many validinsights. As the work is geared towards an understanding of the phonological operations ofEleme, I will try as much as possible to avoid many of the abstractions of this learnedtheory. The standard phonological theory during the latter part of the 20th century was therule-based derivational theory laid out in SPE (Chomsky and Halle, 1968). In 1993, Princeand Smolensky proposed a non-derivational approach called Optimality Theory (OT) andthis has attracted formidable following in phonology. ‘OT is a theory of the human languagecapacity’; the focal tenet is that the grammar of a language is imbued with rival forces(called constraints). These forces are ubiquitous i.e. occur cross-linguistically. Languagesvary predominantly in the way and manner they work out the dissensions between theclashing forces. Specific languages resolve the rivalry by ordering hierarchically indraconian supremacy the various competitors. It is the stratification of these rival forces thatestablish the occasion under which constraints are contravened. A typical language grammartherefore is a medium for puzzling out the friction amongst these cross-linguistic constraints.The ‘surface realizations’ of grammatical structures mirror settlements between theseclashing forces. An output form is best if it incurs the least serious violations of a set ofconstraints, taking into account their hierarchical position. Yip (2002, p. 78) sums up the

2

concept of optimality simply as: ‘the grammar that selects the optimal output consists of aset of ranked, universal constraints, which assess the desirability of each output candidate.Violations of higher-ranked constraints are fatal, and the winner is output candidate thatsurvives the winnowing. The constraint set is the same in all languages; language variationcomes from the ranking only’. A vital notion in OT is constraint, a grammatical descriptionthat can be obeyed or infringed upon by an output form. Kager (1999, p. xi) explains it thus:

Constraints are universal, and directly encode Markedness statements andprinciples enforcing the preservation of contrasts. Languages differ in theranking of constraints, giving priorities to some constraints over others. Suchrankings are based on ‘strict’ domination: if one constraint outranks another, thehigher-ranked constraint has priority, regardless of violations of the lower-ranked one. However, such violation must be minimal, which predicts theeconomy property of grammatical processes.

This universal character of constraints, which correctly predicts accepted cross-linguisticpreferences for specific segments and category of structures, is what makes OT appealing toits followers. Constraints are cross-linguistically or language specifically attested. It isbecause constraints are always clashing that outputs automatically defy one constraint or theother. OT’s position is that individual grammars supervise the clash and ensure thatdisregard for superior constraints are abstained from aggressively than defiance of lesser-ranked constraints. The primal point of OT is that ‘ungrammatical structures’ are not as aresult of flouting of constraints neither is downright conformity to constraints grammaticaldescription necessary to the grammar’s output. What produces the choicest grammaticalresult is the ‘least costly’ defiance of the constraint (Kager 1999, p. 3).

Markedness and Faithfulness are the two categories of constraints put forward in OT.Both constraints examine the output’s grammatical ‘well-formedness’. Markednessconstraints (MCs) govern output grammaticality by ensuring that outputs meet the necessarygrammatical description. Faithfulness constraints (FCs), on the other hand, demand thesafeguarding of the input features in the output form. In other words, FCs insists that outputforms ought to protect ‘segments’, ‘linear order’, ‘shared values for voice (or otherfeatures)’ and reflect it. FC therefore is bi-directional; it looks simultaneously at theunderlying and the surface forms. While, MC concentrates solely on the output ensuringthat it meets the expected grammatical description. So, no constraint considers uniquely theinput form. According to Yip (2002, p.79)

FC penalizes changes to the input form (insertion, deletion, or featural change). IfFC >> MC, where C1 >> C2 means C1 outranks C2, marked segments/structures arefound. If MC >> FC, they are not. For example, if *DELETE >> NOCODA, codawill not be deleted, and closed syllables will be found. If NOCODA >> *DELETE,codas will not surface, and all syllables must be open.

3

It can be inferred, therefore, that the function of FC is to:

Protect the lexical items of a language against the ‘eroding’ powers of MC,thereby…making it possible for languages to have set of formally distinct lexicalitems to express different meanings. Second, by limiting the distance between inputand output, FC restricts the shape variability of lexical items. Faithfulness thuskeeps the contextual realizations of a single morpheme (called its alternant) fromdrifting too far apart. This enhances the one-to-one relations of meaning and form.In sum, the overall function of Faithfulness is to enhance the phonological shape oflexical forms in the output, as sort of inertness limiting the distance betweenoutputs and their basic shapes (Kager 1999, p. 10).

Finally, the ‘architecture’ of OT is constructed around the Lexicon, which accommodates allthe ‘underlying forms of the morphemes’ and submits same to Gen which is a device in theoverall design of OT that spawns various output forms and in turn submits them to Eval forassessment and the selection of the most harmonic candidate.

2. Syllable integrityThis section develops an analysis of Eleme syllable structure. Its two-fold objective is todescribe the principles of syllabification and establish the Eleme syllable template.

Current literature (Kenstowicz 1994, Blevins 1995 and Kager 1999) reveals that allmain approaches to phonology have acknowledged the syllable as a basic unit inphonological analysis and its role in phonological theory as momentous. The syllable iscrucial in stating phonological processes as in (i), rules as in (ii) and phonotactic patterns asin (iii): (i) ‘There are phonological processes such as pharyngealization, stress, tone andballisticity, which take the syllable as their domain of application; such rules and constraintsare sensitive to a domain that is larger than the segment, smaller than the word and containsexactly one sonority peak’ (Blevins 1995, p. 209). (ii) ‘Phonological rules are often moresimply and insightfully expressed if they explicitly refer to the syllable’ (Kenstowicz 1994:250). (iii) ‘The syllable is crucial in defining phonotactic patterns: well-formed sequencesof segments, in particular of consonants and vowels. The syllable also governs patterns ofepenthesis and deletion. It supplies a level of prosodic organization between segments andhigher-level prosodic units: the foot and the prosodic word. … The syllable functions in thedemarcation of morpheme edges … and in defining the position and shape of affixes…’(Kager 1999, p. 91).

The syllable node must have a daughter nucleus (quintessentially a vowel or asyllabic nasal in Eleme or other syllabic sonorants in some other Cross River languages),and may have leftmost and rightmost daughters respectively, the nodes onset (the segmentbefore the nucleus) or coda (the segment succeeding the nucleus) or both. The constituentfollowing the onset is the rhyme. The nodes onset, nucleus and coda in turn, may eachdominate at most one terminal element X, which realizes, as C’s or V’s or they may beempty. Generally, not all syllables in all languages possess all three parts – onset, nucleusand coda. The only obligatory element is the nucleus. The distribution of the onset and codadepends on the phonological structure of each language (Yul-Ifode 1995, p. 70).

4

The syllable, which sometimes matches the mora, corresponds to a unit of timingdeployed in some phonological theories to explain the positioning of stress, accent and tonein certain languages. Although syllable onsets are weightless for the purposes of stress andtone placements, syllable rhymes differ in their moraic structure. For instance, syllablerhymes with a single vowel correspond to one mora (and are very often construed as lightsyllables in rhyme approaches to phonology) whereas syllable rhymes with double vowels(which is differently described as heavy syllables) have two moras (cf Browslow 1995). Inthis study, syllables in Eleme that have the structure CVV and CVCV will be designated aspolymoraic, while syllables with V, N and CV will be specified as monomoraic. The termsyllable or mora is employed in this study solely as an explanatory tool and no specifictheoretical perspective is advocated by using it.

Available literature on the Ogonoid languages has highlighted the differencesbetween syllable structures in the six languages that constitute the Ogonoid group(Williamson 1985, Faraclas 1986, 1989 and Nwolu-Obele 1998). For instance, Hyman(1983: 171) posits that Gokana has ‘phonology devoid of syllables’, and proceeds to explainthe phonological processes of Gokana using moraic weight. Vɔbnu (1991: 62) objects toHyman’s general approach to syllable analysis in Gokana and demonstrates that Gokana likeother Ogonoid languages has syllable. Faraclas (1989) has shown that the significantdistinction in the syllable structure of the western Ogonoid (Baan and Eleme) and the easternOgonoid (Gokana, Kana, Tai and Yeghe) languages is that the western bough has opensyllables, whereas the eastern bough admits a coda. These syllable structures occur inEleme: V, N, and CV; these syllable templates in Eleme contrasts with the archetypalsyllable templates for Cross-River languages, which are CVC, CGV, CVN and CVVC(Faraclas 1989, p. 385).

I shall use verbal reduplication, which is a prevalent phenomenon in Eleme(Williamson 1985: 428), to illustrate the significance of the mora in the language.Reduplication occurs in a number of disparate verbal constructions and diversely conveysnotions analogous to continuity of an action (Ex.2.1a), plurality (Ex.2.1d), futurity (Ex.2.1b)and negation (Ex.2.1c) as exemplified below.Ex.2.1

a. ŋ-gā ló-ló òku b.ɔ-mɔ-mɔ-sɛI SG-CONT-RPER-weave weave basket 2-SG RFUT-see see Q

‘I am still weaving/will weave basket’ ‘Will you see it?’c.àbà rɛ-dɔ-dɔ rī d.ò-wı è-wı-wı

3PL NEG 3- RNEG fail fail-3PL ‘Child’ ‘Children’‘They did not fail.’

Here, I shall consider the non-tonal phonological facets of reduplication before taking onboard the tonal contrast across reduplicated stems later. The phonological restrictions onreduplication in the language are similar in that it is the first mora of the verb root thatreduplicates rather than the first syllable. Now examine the constructions in (Ex.2.2) whereeach evinces disparate reduplicated stems. The data in (Ex.2.2) consist of roots with astructure that is monomoraic e.g.dè ‘eat’ in (Ex.2.2a) and sī ‘go’ in (Ex.2.2b).Ex.2.2

5

a.m-dé-dé b.é-sí-síI SG-RFUT-eat eat 3-SG RFUT-go go

‘I will eat.’ ‘He will go.’

If the verb root that is reduplicated has a polymoraic syllable structure (CVV or CVCV), it isthe first mora of the root that reduplicates. The data in (Ex.2.3a) kpáá ‘repair’ with a CVVsyllable structure illustrates the point; on the other hand, (Ex.2.3b) is kābí ‘weed’, which is averb with disyllabic structure (CVCV). What this example indicates is a partial reduplicationof the verb root and, as with the rest of the data in (Ex.2.2) and (Ex.2.3), the first mora of theroot reduplicates.

(Ex.2.3) a. ɛ-kpá-kpáá b.ɛ-ká-kábí3-RFUT-repair 3-RFUT-weed‘He will repair.’ ‘He will weed.’

The paradigm of first mora reduplication demonstrated above stands for all cases of verbalreduplication in Eleme. Characteristically, it is solely the root that undergoes this process; allother bound morphemes fall outside this process. Although the rules concerning whatphonological element of the root is replicated in reduplication processes are invariablethroughout the grammar of Eleme, in spite of the semantic import of the specificconstructions, reduplicated stems vary in the way tone is accomplished on the reduplicant androot, depending on the meaning of the construction and the tone class of the verb rootundergoing the process of reduplication. For instance, consider the verb bè ‘fight’ which is alow tone verb. A number of constructions apropos to reduplication of this root are providedin (Ex.2.4) below.Ex.2.4

a.ɛ-kā-bé-bé áɁɔ b. ɛ-ká-bé-bé áɁɔ3-SG CONT RPER fight fight war 3-SG MOD-RMOD fight fight war‘He is still fighting the war.’ ‘He should have fought the war.’c. é- bé-bé áɁɔ d. é- bé-bè áɁɔ3-SG RFUT fight fight war 3-SG RNEG fight fight war‘He will fight the war.’ ‘He doesn’t fight war.e.rɛɛ - bē-bé áɁɔNEG 3-SG RNEG fight fight warHe didn’t fight the war.’

From the above data, the tonal paradigm across the reduplicant and root are HH as in(Ex.2.4a), (Ex.2.4b) and (Ex.2.4c). And it appears that the tone is reduplicated with the verbroot but that is not the case. The pattern alters to HL in (Ex.2.4d) and MH in (Ex.2.4e).Observe that the constructions in (Ex.2.4c) and (Ex.2.4d) are distinguished solely by theirtonal paradigm. Having discussed the internal composition of the syllable, I shall now turnto syllabification.

6

2.1. SyllabificationHere, I shall argue that the Eleme syllable strictly conforms to the shapes V or CV and thatthere are no systematic differences between word-initial and word-final syllables. The CVsyllable form occurs universally (Jakobson 1962, Prince and Smolensky 1993 and Blevins1995). As Féry and Vijver (2003: 6) indicate:

First, if a language has syllables without onsets (V), it also has syllables with onsets(CV). Second, if a language has closed syllables (CVC), it also has opened ones(CV). Furthermore, if a language has syllables with complex onsets (CCV), it alsohas CV syllables. And finally, if a language has syllables with complex codas(CVCC), it also has CVC syllables and therefore also CV ones.

If these conclusions are accepted, then the initial generalisation correctly predicts the Elemesyllable structure. This can be explained by constraint interplay between Markednessconstraints (MCs) and Faithfulness constraints (FCs) (Féry and Vijver 2003: 6). Recollectthat MCs and FCs are the two constraints put forward in OT (McCarthy 2004: 4). Theseconstraints assess the output’s ‘structural well-formedness’. MCs govern outputgrammaticality by ensuring that outputs meet the necessary grammatical description.ONSET and NOCODA are MCs. FCs, on the other hand, requires the safeguarding of theinput features in the output form. MAX and DEP are FCs. While MC ONSET calls forsyllables to have onsets, and NOCODA forbids codas. In contrast, FC MAX and DEPprohibit deletion and epenthesis respectively. These two constraints are joined together hereunder the cover term FAITH following Féry and Vijver (2003: 6). I will discuss differentrankings and indicate that in all instances CV syllable is the optimal. And also show that itis in the last ranking that V syllable comes out as the marked Eleme syllable.

Before I proceed, it is pertinent to clarify some of the working tools that will be usedhere. In OT, tableau is made up of rows and columns. They show the evaluating process.The top cell from the left houses the input depiction that other candidate forms are to relate.The competing candidates are placed under this. The optimal candidate is point to by a handsymbol. The competing constraints are arranged across the top of the table. The highestranked constraint is shown first on the left and the lowest ranked constraint last on the right.The lines separating competing constraints denote crucial rankings; if the lines are brokenthis connote non-crucial rankings. Asterisks in the cells are indicative of constraintviolations. While an exclamation mark evinces fatal violation, a shaded area indicates aconstraint that is obsolescent because of the violation of a superior constraint (Féry andVijver (2003), Prince and Smolensky (2004), McCarthy (2004)).

In Tableau 1 where ONSET and NOCODA are ranked above FAITH, which expressesno insertion and deletion, CV syllable type emerges as optimal.

7

Tableau 1: ONSET, NOCODA>>FAITH/CV/ ONSET NOCODA FAITHCV

CVCV *!

*! **

/V/CV

CVCV *!

*!**

In the same vein re-ranking the constraints such that FAITH is classed above NOCODA butbelow ONSET, as in Tableau 4.2, CV syllable form also triumphs.

Tableau 2: ONSET >> FAITH >>NOCODA/CV/ ONSET FAITH NOCODACV

CVCV *!

*!*

*

/V/CV

CVCV *!

** *!

In Tableau 3, I have re-ranked the constraints such that FAITH dominates ONSET, which inturn governs NOCODA. This ranking (FAITH >> ONSET >> NOCODA) allows CV and Vsyllable types to emerge as preferred candidates.

Tableau 3: FAITH >> ONSET >> NOCODA/CV/ FAITH ONSET NOCODACV

CVCV

*!*! *

*

/V/CVCVC V

*!*!*

**

Tableaux (1) – (3) exemplify that regardless of the constraint ranking CV syllables emanatemaking it unmarked and universal. V types of syllables, in contrast, are only allowed in thethird ranking. As the Tableaux indicate, CV and V syllable types occur in Eleme. But thisis not all, while V syllables abound in Eleme, as evidenced in (Ex.2.5a), N syllables are alsofound in Eleme, as indicated in (Ex.2.5b) and Tableau (4) below. As example (Ex.2.5)

8

indicates Eleme nouns admit syllabic segments before onset consonants. These segmentsare syllabic because they bear tone just as the rhyme.Ex.2.5

a.ɔ-Ɂɔ moon b. m-gbé sunò-Ɂó leg m-mu waterè-Ɂó bush m-bó goatà-kà mother n-sa bookɔ-nɛ person n-na animal

Tableau 4 indicates that Eleme permits syllabic consonants thus candidate (a) passes on allconstraints, and therefore is accepted as the optimal candidate. Candidate (c) on the otherhand, is rejected because it violates the constraint that stipulates that a consonant must beadjacent to a vowel, that is, Eleme does not allow complex medial consonants. On thecontrary, candidate (b) is equally rejected because it violates the constraint that stipulatesthat onsets must be simple, that is, Eleme disallows complex onsets.

Tableau 4: ONSET, *COMPLEXONS >> *CC MEDIALCON

As Tableau 4.5 indicates, Eleme disallows complex nuclei as well as closed syllables. It isbecause candidate (b) and (c) violate these constraints that they are rejected in favour ofcandidate (a), which conforms.

Tableau 5: PEAK, ONSET, >> *COMPLEXNUC

To state the principles of syllabification clearly, I propose the syllable template below forthe basic syllable types of Eleme.

Ex.2.6

Ons Nuc

(X) X

/m-gb-é/ ONSET *COMPLEXONS *CC MC

a. N-C-Vb. C-C-V *!c. N-C-C-V *!

/jɔ-wɛ/ PEAK ONSET *COMPLEXNUC *NOCODAa. CV-CVb. CVC-V *! *!c. V-V-V-V *! *!*** *!***

9

The X positions in the skeletal tier are filled with C, V and ◌N segments. Consonants canconstitute the margin that is onset, while vowels and syllabic nasals form the nuclei. Theparenthesized element is optional. As the template indicates, Eleme syllable consists of anoptional onset, an obligatory nucleus and lack coda. As this template further indicates, thebasic Eleme syllable contains at most two segments. I shall assume that segmentalsequences are scanned by this template and are syllabified (or resyllabified) in accordancewith it from right-to-left, see the examples in (Ex.2.7).Ex.2.7

(i) è. Kī. kó.r ó (iii) e.ba.i

NONONON NONN

(ii) ku .ra (iv) t ā

ONON ON

Eleme canonical syllable shape can thus be express as in (Ex.2.8).

Ex.2.8 (C) V

2.1.1.1 NucleusAs the template and examples above indicate, Eleme syllables must have a nucleus; thenucleus is made up of a vowel or syllabic nasal. All vowels fill the syllable nucleus. Thenucleus is the central part coinciding with the peak of sonority. Word-medial and final VVstrings constitute disyllabic sequences obligatorily syllabified as V-V as in:

ó-bu-ō-bu ‘spring’

2.1.1.2 OnsetCross-linguistic investigations into syllable types have indicated that syllables prefer

commencing with a consonant, and ending with a vowel. Secondly, onsets are found in alllanguages, and no language has been known to forbid it (Jakobson 1962, Greenberg 1978,Kaye and Lowenstamn 1981, Itô 1986, Blevins 1995 and Kager 1999: 92-93). These twoassertions are grounded in the first generalisation in section 2.1.1, which states that ‘if alanguage has syllables without onsets (V), it also has syllables with onset (CV)’ (Féry andVijver 2003: 6). Kager (1999: 33) posits that this is the ‘implicational universal for syllableonsets’. In other words, with regards to onsets, the dichotomy is between languages thatpermit syllables without onsets and those that prohibit syllables without onsets. Of course,no language is recognized to forbid onsets; certain languages even provide onsets through

10

epenthesis if onsets are missing (Payne 1981, Itô 1989, McCarthy and Prince 1993b, Kager1999: 93). The occurrence of an onset is therefore not marked when compared to its non-occurrence (Kager 1999: 33). Itô (1989), Prince and Smolensky (1993) represent this as aconstraint thus:

ONSET*[ V (‘syllables must have onset’)

This constraint prohibits V syllables, and asserts that onsets are obligatory. According toKager (1999: 94) this onset undomination is docked ‘in the articulatory and perceptualsystems’ because vowels prefer ‘a preceding consonant rather than another vowel’.

In Eleme, the occurrence of an onset is not marked as compared to its non-occurrence. If this is correct, then Eleme allows V syllable word-initial, medial and final asin (i) and (iii) of (Ex.2.7) above. The presence of V-syllables in Eleme automaticallyexplains why there are no word-initial and word-medial and/or internal sequences of three ormore vowels. In other words, vowel sequences are not permitted in word-initial position e.g.*aiba and there are no vowel sequences longer than two vowels e.g. *auia exceptinterjections and/or an agreement particle; others are the result of consonant deletion inmorphophonemic combinations.

2.1.1.3 CodaWhile the domain of the onset is on the left margin of the syllable, the coda in

contrast operates on the right margin. Furthermore, whereas the presence of a consonant isobligatory and unmarked in onsets, the opposite is the case with coda, that is, the absence ofa consonant is cross-linguistically ‘preferred to the presence of one’ (Kager 1999: 94). Thisassertion is again grounded in the second generalisation in section 2.1.1, which states that ‘ifa language has closed syllables (CVC), it also has open ones (CV)’ (Féry and Vijver 2003:6). Kager (1999: 94) also claims that this is the ‘implicational universal for syllable-finalsegments’. Similar to the situation with onsets, some languages permit codas while othersprohibit them and of course, no language is known to insist on syllables having coda(Blevins 1995, Kager 1999: 94). Typologically, Kager (1999: 94) suggests that the preferredsituation is for syllables not to have codas. He represents this as a constraint thus:

NOCODA*C] (‘syllables are open’)

The above constraint prohibits syllables ending in consonants. Akin to the situation withonsets, this restraint is docked ‘in the perceptual system’, final consonants ‘…tend to beunreleased, and hence lack perceptual cues that are present in prevocalic consonants, whichare released’ (Kager 1999: 94 also cf Ohala 1990, Steriade 1995).

Eleme syllables have no coda as illustrated in (Ex.2.9), (Ex.2.10) and (Ex.2.11). Itfollows therefore that CV-CV is preferred to CVC-V in Eleme, that is, an intervocalicconsonant is to be syllabified as an onset rather than as a coda. Although, Sommer (1981)

11

and Borgstrøm (1937) have reported languages such as Oykangand and Barra Gaelic whereaccording to them prefer CVC-V to CV-CV, but Clements (1986), McCarthy and Prince(1993b), Blevins (1995) and Kager (1999) think that their analysis are doubtful.

2.1.1.4 *ComplexTypologically, while languages contrast because of the ‘presence or absence of onsets andcodas’ as the discussions above illustrates, languages equally exhibit dissimilarities withregards to the ‘complexity’ of syllable margins (Kager 1999: 95). He notes that certainlanguages may permit codas, and still prohibit ‘complex’ codas, namely, codas that are madeup of more than one consonant. In other words, these languages insist that coda remains‘simple’, specifically, ‘consist of one consonant’ (Kager 1999: 96, also cf Harris 1983). AsOT literature reveals, a lot of languages limit the ‘complexity of syllable margins’; thisimplies that complex onsets and codas are cross-linguistically less favoured in contrast tosimple onsets and codas, which are highly favoured (Kager 1999: 96).

These assertions are grounded in the third and fourth generalisations in section 2.1.1,which state that ‘if a language has syllables with complex onsets (CCV), it also has CVsyllables and if a language has syllables with complex codas (CVCC), it also has CVCsyllables and therefore also CV ones’ (Féry and Vijver 2003: 6). These are the ‘implicationaluniversals for onset and coda’ complexities respectively (Kager 1999: 96). Crosslinguistically no language has been reported which permits complex syllable margins, yetforbids syllables with simple margins. The tendency is for some languages to prohibitcomplex onsets or codas; other languages prevent complex onsets or codas by insertingvowel at the relevant site (Lynch 1974, Itô 1986, Blevins 1995, Kager 1999, p 96). Thefollowing constraints (i) and (ii) were proposed (Kager 1999, p. 97) to account for theseprocesses:

(i) *COMPLEXONS

*[ CC (‘onsets are simple’)(ii) *COMPLEXCOD

*CC] (‘codas are simple’)

The first constraint states that a language wherein *COMPLEXONS is undominated preferssimple onsets. While the second formulation states that a language where *COMPLEXCOD

is undominated prohibits complex codas in preference for simple codas, that is if it permitscodas at all (cf Harris 1983, Popjes and Popjes 1986, Kager 1999, p. 97).

Eleme disallows complex syllable margins as (2.1.1.4 - 6) indicates, that is, there areno complex onsets and nuclei. In other words, no more than one C or V may associate toany syllable position node, that is, Eleme syllable structure does not permit sequences ofCC, CCC and CCCC as in *jogj, *jjga, *Ɂwwj as can be deduced from the data in(Ex.2.18). There are no syllables larger than CV and Codas are not permitted. Since,complex onsets and nuclei are ruled out and there cannot be more than one segment in anyonset or nucleus position. Consequently, VV is obligatorily syllabified as V.V, CVV asCV.V, VCV as V.CV and CVVCV as CV.V.CV. What this implies is that in Eleme CV is

12

the preferred syllable form, while V is less preferred because it lacks an onset as in (Ex.2.9)adapted from (kager 1999: 95).

Ex.2.9

2.1.1.5 Margin nodes*M/V

It can be inferred from the above discussion that V is barred from associating with marginnodes (onset). Eleme also prohibits N associating with margin nodes.

2.1.1.6 Peak nodes*P/C

Just as either V or N cannot associate with onsets, so C cannot associate with peak (nucleus)nodes (Payne 1981). In other words, Eleme disallows consonants from associating withpeak nodes unless they are syllabic nasals.

There is another dimension to the Eleme syllable, which I shall address presently.Eleme has syllabic nasals which bear tone, but onset nasals, which do not, as exemplified in(Ex.2.10).Ex.2.10

ǹdʒu mà

NON ON

The syllabic nasal as nucleus occurs only before consonants and it is restricted to themorpheme initial position. These are both necessary and sufficient condition for the syllabicnasal. There is no such constraint when a vowel is nucleus. The fact that the nucleus is theonly obligatory part of the syllable as demonstrated in (Ex.2.11) in conjunction with themaximal syllable size restriction, will correctly account for Eleme syllabification.

Following Everett (1993: 5), attach V’s or N’s to the nucleus, attach C’s to the onset,attach the remaining items to nucleus or onset, as determined by the syllabic constraints orsyllabic maximality, attach all subsyllabic material to (syllable node), beginning with the

rightmost nucleus, begin a new with each nucleus. From the viewpoint of phonology,authentic Eleme nouns begin with a harmonically conditioned syllabic vowels /a e ɛ o ɔ ɔ/ ora syllabic nasal [m n ŋ] homorganic with the following consonant (Williamson 1985, p.428). Nouns in Eleme are thus disyllabic; finite verb forms are also often disyllabic. Thefact that a greater number of words are disyllabic suggests a minimality constraint onprosodic words in Eleme (see Hall (1999) for discussion on minimality). Note that there are

Syllable type Onset No codaV - +

CV + +

13

constraints on which nasal may occur as syllabic element, note also that besides V or Nsyllable, the lexical items in (Ex.2.10) is composed of a syllable with the structure CV,which is the optimal syllable shape in Eleme as Tableaux (1) – (3) illustrates. Cluster ofconsonants are not admissible in Eleme and syllables are without coda. Trisyllabic (V-CV-V) nouns in Eleme, apart from having prefixal V syllable, the vowels are either identical asin (Ex.2.11a), or disparate, as in (Ex.2.11b) below.Ex.2.11

a. ò-ló-ō antelope b. à-wí-à siblingà-dá-ā tooth ò-lō-ī lawè-kpì-ì money m-gbā-ó dogè-tʃū-ū louse ɛ-bā-ī we

Verb roots in Eleme quintessentially constitute CV syllable structure, as in (Ex.2.12a). Fewhave the CVCV structure as exemplified in (Ex.2.12b), or CVV syllable structure, as shownin (Ex.2.12c). Roots with glottal stop and a vowel at times weaken phonetically to a singlevowel as in (Ex.2.12d). Infrequently, verbs manifest as vowel sequences, and consequentlyhave VV structure as in (Ex.2.12e).Ex.2.12

a. b. c. d. e.fò plant būtū break maa laugh (ʔ) à go aamu dig dàlà lift sɛī economizema lie down dʒìrá agree kpáá repairdʒù come kpara seek kpɛɛ tell

It is pertinent to note here that a greater number of Eleme verb roots commence with aconsonant and only a small proportion deviate from this trait. In contrast, nouns invariablystart with vowel or syllabic nasal as in (Ex.2.10) and (Ex.2.11) above.

As the discussion in section (2.1.1 – 2.1.2.2) and examples (Ex.2.5- Ex.2.10)indicate, Eleme has syllables without onsets (V) and also syllables with onsets (CV). Usingconstraint interaction between Markedness and Faithfulness constraints as indicated onTableaux 1 - 4, it has been shown that in each case the unmarked CV syllable emerge asoptimal. It is also noted that only in the third ranking that V emerges making it the markedEleme syllable structure.

3.0 Phonological operations3.1 Nasal insertionNasal insertion is a type of syllable structure process; it causes an alteration in the originalsyllable structure. Schane (1973) asserts that it may or may not lead to a simpler structure.In Eleme, nasal epenthesis occurs. A homorganic nasal develops between an underlyingnasal vowel and the following consonant as a means of reinforcing and retaining the nasalquality of the original word; this is a case of nasal strengthening. This has been observed tobe common when the following consonant is an obstruent. This data might give credence toa phonetic CVC syllable structure in Eleme. In others words, this data appears to nullify the

14

conspiracy theory at least at the surface level. But, as is evident at the underlying level, thetheory holds.

Ex.3.1 /sɔbí/ [sɔmbí] fart

/sapɔ/ [sampɔ] sponge

/sɛgí/ [sɛŋgí] massage

This process can be represented by the rule:

[] + nasal / +nasal ___ +cons

F +syll F

It is pertinent to note here that Yul-Ifode (1999: 74) considers this an NC sequence. ‘NCsequence in the analysis of doubtful segments is a cover term for (i) Homorganic nasalswhich may or may not be syllabic, in a sequence with other consonants. (ii) Prenasalisedconsonants’. If I interpret NC sequences as sequences of nasal and other consonants, thensome of the item in my data will yield CVC or CCV syllable pattern in Eleme, for instance:

sɔm.bí or sɔ.mbí.CVC.CV CV.CCV

On the basis of the syllable structure constraints:NOCODA

*C] (‘syllables are open’)*ComplexONS

*[ CC (‘onsets are simple’).This analysis is rejected. On the other hand, interpreting NC sequences, as phoneticallycomplex phoneme will yield CV, which is a clear-cut syllable structure in Eleme. However, Iwill not adopt it for Eleme. This is because this interpretation does not reveal the processinvolved in the realization of the NC segment. The homorganic nasal will be treated asoriginating from the phonemic nasal vowel before the consonant that is being anticipated.NC sequences are as a result of nasal consonant insertion.

3.2. DeletionDeletion is a syllable structure process as well since it alters the structure of the syllable.There are two types of deletion in Eleme, C2 consonant deletion and prefix vowel deletion.The domain for the deletion process in Eleme is the underlying CVCV structure. The secondconsonant in a CVCV syllable structure is deleted, giving rise to a CV-V structure. Thevowels retain their inherent tones in such a process. This process is common in Eleme as anintermediate stage of the change from CVCV to CV-V. The conspiracy theory also holds.The Eleme syllable template enforces the processes.Ex.3.2

mgbógó mgbáótògì tòìlōgì lōì

15

This process is captured by the following rule, which states that the voiced plosives aredeleted in non-identical vocalic environment.

[+cons] / [+syll] __ [+syllV1V2

The prefix vowel is often deleted in rapid speech and the result is a CV-V from an underlyingV-CV. This process has been observed in nouns (see example Ex.3.3). The vowels, whichare often deleted, are the mid vowels -e ɛ a. I have posited that the underlying form is the V-CV, because the process by which a word-initial vowel gets deleted is more natural than for itto add. Moreover the formulation of the rule in terms of distinctive features is well motivatedand easier. The question of when to postulate deletion in contrast to elision remains animportant one in the analysis of Ogonoid languages but I shall not go into that argument here.

Ex.3.3

aba baa-jɛ his cooking pot

àkà kàà-jɛ his mother

èté téé-jɛ his tree

ɛbɛrɛ bɛrɛɛ-jɛ his bag

3.3 LiaisonThe process of liaison is common in Ogonoid languages (cf Ikoro 1996). Whereas in Kanaliaison is strictly phonologically triggered, in Eleme liaison is both morphosyntactically andphonologically motivated. Secondly, while in Gokana the operation of liaison is triggered bythe imperative suffix, in Eleme the operation of liaison is triggered either by the perfective orimperative suffix. Thirdly, while in Tai and Baan the presence of nasality at the liaisoninsertion site is opaque, in Eleme the insertion site is sensitive to nasality as it determineswhether r or n is to be inserted. In a nasal environment n is the preferred liaison; elsewhere ris the preferred liaison.

Eleme has identical (CV1V1) and non-identical (CV1V2) vowel sequences. It is theamalgamation of these vowel sequences and verbal suffixes that trigger liaison. Liaisonoccurs between CVV verb roots with the perfective and imperative plural suffixes –a and –aa. In the process of liaison, ‘an epenthetic r or n is inserted in the verb root, depending onthe presence of nasality at the insertion site’. The consequence of r or n insertion is either‘vowel shortening’, ‘coalescence’ or ‘glide formation’. It is the type of vowel sequence thatdetermines which of these other processes occurs with liaison. Second, the structure CV1V1

trigger liaison with both the perfective or imperative suffix and the only exceptions are aa, ɔɔ,uu and sequences with LM tonal pattern, liaison and shortening applies with all other CV1V1

sequences. For instance, the Eleme data in (Ex.3.4) has four items; these items all haveidentical vowel sequences. The four items divides into two parts based on the observed on-going processes. While (Ex.3.4a) the identical vowel sequences shorten following theinsertion of either r or n at the insertion site, in contrast (Ex.3.4b) no other processaccompanies the process of liaison. As already noted the reason is that the sequences of aa,

16

ɔɔ, oo, and uu do not permit further processes in the trail of liaison. It is because the inherenttone class of the verb root which is LM that items numbers (viii) and (ix) with ii and eesequences also failed to undergo further processes. Moreover, the data in (Ex.3.4) below tworules are in operation, an epenthetic consonant is inserted intervocalically, and the consonantis realized as r if the insertion site is [-nasal]. On the other hand, n is realized if the insertionsite is [+nasal]. Schematically, the rules are stated thus: (i) r / v-v and (ii) n / v-v

Ex.3.4Root PERF IMP.PL

(a) (i) dʒìì è-dʒìr-à dʒīr-āā climb(ii) tɔɔ ɛ-tɔn-a tɔn-aa stay

(b) (iii)bāā ɛ-bāār-ā bāār-āā tear(iv) gbɔɔ ɛ-gbɔɔn-a gbɔɔn-aa separate

If the structure is CV1V2 and V1 is [+round] while V2 is not ɔ or a, liaison, vowel shorteningand coalescence apply. But where shortening operates, coalescence does not.

Ex.3.5 Root PERF IMP.PL(a) bùì è-bùr-à bùr-āā read {shortening}(b) dʒūē è-dʒòr-à dʒòr-āā come{coalescence}(c) dʒùà è-dʒùr-à dʒūr-āā bring{shortening}(d) sūɛ è-sɔr-ā sɔr-āā fart {elision}

In the above data (Ex.3.5b) and (Ex.3.5d) vowel coalescence take place with the ue or uɛsequences. Vɔbnu (2001: 130) postulates that the vowels ue or uɛ merge into a single vowelbecause of roundness assimilation. I rather suggest that what is happening in examples(Ex.3.5b) and (Ex.3.5d) is captured by the rules below.

V V

-ATR r + ATR as in (b)

+RND

-High

In (Ex.3.5b) two processes are ongoing as captured by the rule above. First, the non-ATR vowel uis erased due to the differences in the height of the vowels. Secondly, the remaining +ATR vowel eis realized as +ATR vowel o due to roundness assimilation, where the first process is a requirementfor the second process.

17

V V

+ATR / r - ATR as in (d)

+RND

-High

In (Ex.3.5d) as captured by the above rule two processes are equally on going. First, thenon-ATR vowel u gets deleted due to differences in the height of the vowels just as in(Ex.6.2b). Secondly, the remaining –ATR vowel ɛ is realised as –ATR vowel ɔ due toroundness assimilation, where the first process triggers the second process. This is a clearerexplanation than just postulating the merging of two vowels with height differences withoutresolving the height issue. The differences in height are dealt with through deletion and thenthe stage is set for the process of roundness assimilation.

Apart from the resolved height differences, the vowels in both (Ex.3.5b) and(Ex.3.5d) also differ on the front and back dimension. Whereas e and ɛ are front vowels, uis a back vowel. As this data exemplifies: ɔwā ʔūwī ‘you fetch’ (emphatic expression) fromʔībī ‘fetch’ (non-emphatic expression), the close front vowel -i- is realised as the close backvowel -u- in similar environment. This suggests that the front and/or back distinction aresometimes neutralized in certain phonological context. This appears to be the case here in(Ex.3.5b) and (Ex.3.5d) above considering that the final vowels during the processes are oand ɔ, which are the back vowels.

Vɔbnu (2001, p. 130) asserts that in Gokana and by extension Ogonoid coalescenceis obligatory with the uɛ sequence and optional with the ue sequence. My Eleme datavalidates this claim. Furthermore, coalescence as observed in (Ex.3.5d) differs from glideformation as seen in (Ex.3.6i). With regards to (Ex.3.5a) and (Ex.3.5c) these are simpleprocesses of shortening. I have opted for the term ‘shortening’ following Ikoro (1996, p. 39)instead of Yul-Ifode’s (1985, p. 76) ‘contraction’ because, in Eleme, the process ofcontraction, as I have demonstrated in Ngulube (2008) is subsequent to another phonologicalprocess, that of assimilation.

3.3.1 Liaison and the perfective suffixLiaison is triggered if the perfective suffix (-ā) attaches to a disyllabic (CVV) stem. Thisprocess alters the structure of the verb stem to CV (V) CV. The process of liaison is -r- or -n- depending on the insertion site [±nasal] is compulsorily embedded between root andsuffix. Verb roots ending with [-nasal] vowels accept -r- conversely [+nasal] vowels admit -n-. Vowel lengthening follows liaison specifically if the verb root has the imperative M toneclass; this is represented as V VV__ [r n insertion] as in (Ex.3.6).

Ex.3.6 maa-pɛɛr-ā <pɛ maa-wīīr-ā < wī maa-tʃıın-a < tʃı1-SG PF PRE-fly-PS 1-SG PF PRE-call-PS 1-SG PF PRE-reject-PS‘I have flown.’ ‘I have called.’ ‘I have rejected.’

18

Ex. 3.7 gbárā ɔnɛ è-kór-à < kòò gbárā ɔnɛ ɛtɔn-ā < tɔɔman SPEC: SG PF PRE-fold-PS man SPEC: SG PF PRE-stay-PS‘The man has folded it.’ ‘The man has stayed.’

If the structure is CV1V1 as in (Ex.3.7), vowel reduction takes place especially if the verbroot lacks the imperative LM tone class. This is represented as V1 + V1 V/__ [r ninsertion]. Note that whereas in (Ex.3.6) pɛ ‘fly’, wī ‘call’ and tʃı ‘reject’ that are normallymonosyllabic in structure are realised as disyllabic, in (Ex.3.6) kòò ‘fold’ and tɔɔ ‘stay’ thatare normally disyllabic are realised as monosyllabic. One can therefore conclude that theperfective suffix (-ā) licences liaison to trigger different phonological processes withdifferent classes of verbs. Vowel reduction as in (Ex.3.6) or lengthening as in (Ex.3.7) isblocked after liaison if the structure is CV1V1 with a verb stem having a LM tone pattern inthe imperative as in (Ex.3.8). In the same vein, vowel reduction is blocked if a CV1V1 verbstructure is either uu or aa with LL tone pattern as in (Ex.3.9). Where the vowel sequence isCV1V2, vowel coalescence and/or reduction follows liaison only if V1 is u or ɔ, and V2 iseither e or ɛ as in (Ex.3.9).

Ex.3.8ɛ-gbéér-ā < gbèē māā léér-ā < lèē3-SG PF PRE-lost-PS 1-SG PF PRE-keep-PS‘He has lost it.’ ‘I have kept it.’

Ex.3.9ǹsar-a ɛ-bààr-ā < bāā è-kùùr-à < kùùbook PS PF PRE-tear-PS 3-SG PF PRE-crawl-PS‘My book is torn.’ ‘He crawled.’

Ex.3.10ɔfáá è-kūr-ā < kùɛ ɔwājɛ à-mɔr-ā ǹtító <mɔɛvehicle PF PRE-hit-03SG wife 3-SG PF: PRE-see-PS work‘A car knocked him.’ ‘His wife has found a job.’

Note again that in example (Ex.3.5) the monosyllabic verb roots: pɛ ‘fly’, wī ‘call’ and tʃı‘refuse/reject’ all have the imperative M tone class. Conversely, the disyllabic verb roots inexample (Ex.3.6): kòò ‘pack/fold’ and tɔɔ ‘stay’, both have the LL tone pattern not the LMtone class. So, while monosyllabic verb roots with M tone pattern trigger liaison with vowellengthening in its trail, disyllabic verb roots with LL tone pattern trigger liaison with vowelreduction in its trail. On the other hand, vowel shortening, reduction and lengthening areblocked if the verb roots are either disyllabic with LM tone pattern or the verb roots haveeither uu or aa sequence with LL tone pattern. Note that the same verbs that allowconsonantal liaison in the perfective as illustrated in examples (Ex.3.5) – (Ex.3.9), alsolicence liaison in the imperative. These same verbs authorize vowel reduction with theintensive suffix –i as the discussion below will reveal.

19

6.1.2 Liaison and imperativesThe imperative mood specifies an order initiated by a speaker at the moment of speaking. InEleme, there is no obvious subject marker for the first, second and third person singular; incontrast, the subject marker is necessary for the first, second and third person plural. Post-verbally, the nominal and pronominal objects are mandatory; this is also true of otherOgonoid languages (cf Vɔbnu 1991, Nwolu-Obele 1998, Isaac 2003 and Bond 2006). Thereare different ways of realizing affirmative imperative in Eleme. The first method is throughtonal variation while the second is using suffixation. These approaches are complementary.If the addressee is alone a single tonal morpheme is used. In this connection, three toneclasses are discusses here.

In (Ex.3.11), the verbs in isolation are assigned low tone pattern. When these verbsaccept object NPs, their inherent low tone patterns did not alter. In item (c), the object of theverb is mbó ‘goat’, here the bilabial nasal m and its low tone are deleted but its nasality ispreserved on the last syllable of the verb. The situation in item (d) is different. It is thevowel ī that is deleted but its mid tone is preserved and spreads, delinking the low tone ofɔbó ‘arm’.

With regard to (Ex.3.12), where the verbs in isolation bear mid tone, and theirvarious objects in isolation bear LMM as in è-kūū ‘cloth’, LM as in è-nū ‘something’ andLHH as in è-kpíí ‘money’. When the verbs combine with the objects, they are togetherrealised as HH – MM as in (f), HH – M as in (g) and HH – HH as in (h).

Finally, in (Ex.6.12), the inherent LM tonal patterns of these verbs in isolationremain unchanged even when the various verbs accept object NPs. Meanwhile, the mid toneof the last syllables of the verbs spread beyond their boundaries to the next adjacent vowelsdelinking the prefixal low tones of these nouns, which are usually low tones as in ɔtɔɔ‘house’, àkpà ‘bag’, ǹlé ‘beauty’, ètò ‘load’ and ɔtɔ ‘ear’. Only one inference can bededuced here, that Eleme uses tonal alternations in the realization of imperativity.Ex.3.11

Low (isolation) verb + object(a) dɔ fall dɔ gbá fall (into the) mud(b) bùì read bùì ǹsa read a book(c) dàlà lift dàlabó lift goat(d) dʒìdʒààrī spread dʒìdʒààrɔbɔ spread yours arms(e) dʒùùrà pull dʒùùrabó pull goat

Ex.3.12Mid (isolation) verb + object(f) bā tear bá-é-kūū tear a cloth(g) nā do ná-é-nū do something(h) gbɛ throw away gbɛ é-kpíí throw away money

Ex.3.13 Low-mid (isolation) verb + objectfìrɛ pull down fìrɛ ɔtɔɔ pull down a housesùā go with sùā ākpà go with a bagtòwɛ show tòwɛ nlé exhibit beautybɛɛ destroy bɛɛ ētò destroy the loaddʒìbī pinch dʒìbī ɔtɔ pinch an ear

20

If the order is meant for a group of people, the suffixation method becomes necessary andthe following suffixes are used in Eleme āā, ɛɛ or aa. The use of any of these is predicatedon the type of verb, and the [±nasal] nature of the preceding vowel. Plural verb roots areassigned mid tones. In order to explain this symmetric tone assignment a spreading rule isproposed for the suffix. Therefore, the M tone of the suffix spreads leftwards to the nextvowel. Where the prior syllable already has M tone, tonal coalescence occurs, on thecontrary if the previous vowel is non-mid, the non-mid tone is deleted. In (Ex.3.14) andbeyond ɔbàà ‘you (pl)’ is employed.

Ex.3.14 ɔbàà-dɔ-āā fall ɔbàà-tōwɛ-ɛɛ teachɔbàà-dʒùùrā pull ɔbàà-dʒù-āā come

If the imperative suffix attaches to a CV1V1 sequence, liaison occurs and no other processfollows. But, if the imperative suffix is attached to CV1V2 where V is u, e, ɛ or i, liaisonoccurs triggering vowel coalescence. In contrast, with the perfective suffix liaison occurs

triggering vowel reduction in its trail (see Ex.3.6). Vowel reduction is blocked in (Ex.3.14)in contrast it is licensed in (Ex.3.15), while vowel coalescence triggered by liaison is evident

in (Ex.3.16). The operation of liaison is prohibited if the verb root is CV1V2 and the lastvowel is a or o. The reason for this is not very clear as in (Ex.3.17).

Ex.3.15 ɔbàà-tʃīr-āā pin IMP.SG tʃīīɔbàà-tɔn-āā stay IMP.SG tɔɔ

Ex.3.16 ɔbàà-mɔr-āā see IMP.SG mɔɔɔbàà-uīr-āā press IMP.SG ūīɔbàà-mɔn-āā dig IMP.SG mùɛɔbàà-dʒūr-āā cover IMP.SG dʒùè

Ex.3.17 ɔbàà-būā-āā greed IMP.SG bùàɔbàà-pīō lick IMP.SG pìò

The analysis of the plural imperative and the perfective marking raises the question of whyliaison is permitted in certain cases and prohibited in others. It is pertinent to note that thesame verbs that licence liaison in the imperative also trigger liaison in the perfective. Thismay be a sign that liaison is active in all instances.

The analysis and morphophonemic transcription adopted here indicate that theepenthetic r and n are not treated as belonging to the verb suffix āā or aa. On the contrary,they are analysed as belonging to the verb root. A survey of data from related languages alsoindicated that ‘in proto-Ogonoid, r and n often occur in syllable final positions’ (Ikoro 1989:124). Vɔbnu (1991) asserts that in Gokana, l has replaced r in most contexts, but n is stillprevalent. In contrast, Hyman (1982, 1985) has demonstrated intervocalically that therelationship between l and r in Gokana is one of free variation. Dimmendaal (1978),Williamson (1985) and Bond (2006) have indicated that like Kana, Eleme has lost l and rsyllable finally. These studies equally suggest that in contemporary Eleme, all consonantshave been lost in syllable final position. It is difficult to say precisely the situation in Baan

21

for paucity of available data. Akin to Eleme, Kana has lost l and n in similar context butkeeps other consonants in this position (cf Ikoro 1989). Nevertheless, as the data below(Ex.6.18) illustrates Eleme have vowel sequences (either identical or non-identical) whereGokana has l or n. The tonal system of the roots is observed to be the same wherever thissound change occurs. The Gokana and Kana data are from Vɔbnu (1991), andN.Williamson (1990) respectively.Ex.3.19 Gokana Kana Eleme

víl ábíé òwí grassvín ŋwíí òwıı childfùl fìì fù cutkìn kıı tʃıı reject

In this data (Ex.3.19), in Gokana l and n are in syllable final spot, although confined tomonomoraic structures. In Eleme, these lexical items with the exception of fù arepolymoraic structures. The way the plural imperative and the perfective aspect behavesuggest that in Eleme final r and n can be recovered in certain contexts where they occur asliaison consonants though not in word-final positions. Synchronically, one may even arguethat in contemporary Eleme, r or n occur in syllable final position but are compulsorilydeleted word-finally, similar observation is put forward for Kana (Williamson 1985,Faraclas 1989). Consider the behaviour of CV1V1 in the perfective provided in the databelow for easy understanding.Ex.3.20

nɛ-gbárà ɛ-pɛɛr-ā IMP.SG: pɛɛSPEC: SG man PF: PRE-jump-PS‘The man has jumped.’

nɛ-gbárà ɛ-tɛn-a IMP.SG: tɛɛSPEC: SG man PF: PRE-sharpen-PS‘The man has sharpened.’

māā-bāār-ā nsa IMP.SG: bāā1-SG PF: PRE-tear-PS book‘I have torn a book.’

The liaison r or n in the data (Ex.6.19) above occur with monomoraic structure as in pɛ‘jump’ or polymoraic structures as in tɛɛ ‘sharpen’ and bāā ‘tear’. If r or n gets elided theoutcome of this elision will depend on whether it is mono- or polymoraic structure, withmonomoraic structure, automatic compensatory vowel lengthening accompanies the elisionprocess, and in contrast nothing happens with polymoraic structure, Ikoro (1989) attested tothis in his study of Kana.

The analysis of consonant liaison presented so far varies from Williamson (1985) inthat she thinks that the CVV structure that trigger liaison was derived from CVCV whose C2

was l. Moreover, she argues that C2 is preserved in Gokana as r before –i suffix and thatEleme has lost l before –i suffix and substituted it with r before the perfective suffix -a. In

22

support of her argument she insists that l never occurs as C2 synchronically in Elemewhereas only r is establish in such positions. But my data attest to the fact that l does occuras C2 as in (Ex.6.20):Ex.3.21

mbàlàʔɔ full moon mbòlòʔó thighmbálà fish trap mbāló soupmbálā marital status mbēlá likeness

Diachronically, Eleme had r and n in word-final positions, which are now lost (Dimmendal1978, Williamson 1985, Nwolu-Obele 1998 and Bond 2006). These sonorants can berecovered through liaison in other contexts. Data from other related languages have shownclearly that modern Eleme no longer preserves the original proto-Ogonoid final r, althoughaccording to Vɔbnu (1991) Gokana still retains it as l only in CVC structures (see exampleEx.6.18). The common ground between Williamson (1985), Ikoro (1989), Isaac (2003), andthe present study is that the epenthetic r or n is never analysed as part of either the perfectiveor the plural imperative suffix.

The view of the plural imperative developed here can be extended to Gokana. Thislanguage has the plural imperative suffix as -īī (data from Williamson (1985: 434) and thesingular imperative suffix as -ī). Hyman (1985) calls this singular imperative suffix a‘meaningless grade suffix’. In Gokana, deletion of r, l and n word-finally occurs with CVVstructure. This process is blocked with CVC structure. If the ultimate C of the CVCstructure is a final voiced stop, this weakens to a fricative in conjunction with an imperativesuffix. If the ultimate C in a CVC structure is l, this is realised as r in juxtaposition with theimperative suffix -īī, compare the data in example (Ex.3.22) – (Ex.3.24) (Gokana data fromVɔbnu (1991), while Kana data is from N.Williamson (1990).Ex.3.22

Gokana Kana ElemeSG PL SG PL SG PLbìì bīīr-īī bìì bīr-āā bìì bīr-āā squeeze ittɔɔ tɔɔn-īī tɔɔ tɔn-āā tɔɔ tɔn-āā staykòò kōōr-īī kòò kōōr-āā kòò kōōr-āā gather

Ex.3.23bàn bān-īī bàrā bārā-āā ba bar-āā begbɔb bɔr-īī bɔb bɔb-āā bɔ bɔ-bí tie

Ex.3.24īl īl-īī īī īr-āā ī īr-āā marry of manūl ūr-īī ūī ur-āā ūrīī ūr-īāā boil of soupɔl ɔr-īī ūɛ ɔr-āā ɔ ɔr-āā singfùl fùr-īī fìì fīr-āā fù fùr-āā cut

23

That the plural imperative suffix is -āā in Eleme and Kana, and -īī in Gokana is substantiatedhere (Ex.3.22) – (Ex.3.24). Vɔbnu (1991: 52) treats l or n in the above data as part of thesuffix, this is not feasible and that is why he resorts to costly ad-hoc rules. He rejects thepresent analysis for Gokana based on the following arguments:

kool or faal does not confirm to the syllable structure of the language. Even whenCVVC is divided as in CV-CV [CV-VC? SMI], irregular syllable shapes stillappear. kool or faal has a CV1 V2C shape in which V1 is identical to V2. InGokana, what occurs is CGVC syllable shape. When it occurs, G is always j as inbjom ‘wasp’, pjob ‘tsetse fly’, kjag ‘speck in the eye’. kool or faal cannot standalone.

Vɔbnu asserts that, the suffix in the data given in (Ex.3.22) – (Ex.3.24) is –li, which he calls‘transitivizing suffix’. But according to Ikoro (1989: 197) Vɔbnu’s analysis only generatesmore questions not answers, as a variety of his derivations illustrate. One of suchderivations is exemplified in (Ex.3.25) for ‘call’ and ‘tie into a bundle’.Ex.3.25

Vɔbnu (1991: 62) Gokana: ‘call’ ‘tie into a bundle’Compensatory lengthening kōl-lí bɔb-bíC-elision kō-lí bɔ-bíL-r intervocalically kō-rí ----C-weakening ---- bɔ-víSurface form kō-rí bɔ-ví

My analysis as presented above provides the necessary justification for the gap in thedistribution of Eleme l and r. Moreover, it provides a non-ad hoc explanation forcompensatory lengthening.

4. ConclusionThe data on liaison as presented in this paper appears problematic for the conspiracy theorythis is because the syllable template did not predict a coda. I suggest that the only solutionto this challenge is to posit that phonologically the template re-syllabifies the data butphonetically it allows the strong bond in the verb root to keep the liaison segment.

24

ReferencesBlevins, Juliette. 1995. The syllable in phonological theory. In The handbook of

phonological Theory ed. Goldsmith, John A. Blackwell.Bond, Oliver, and Anderson, Gregory D.S. 2005. Divergent Structure in Ogonoid languages.

Paper presented at 31st Annual Meeting of the Berkeley Linguistics Society, Berkeley, CA.Bond, Oliver. 2006b. Aspects of Eleme verbal morphosyntax. PhD thesis: University of

Manchester.Borgstrøm, C. Hj. 1937. The dialect of Barra in the Outer Hebrides. Norsk Tidsskrif for

sprogvidenskap 8. 71-242.Broselow, Ellen. 1995. Skeletal positions and moras. In Goldsmith. 175- 205.Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English, Harper and Row,

New York.Clements, George N. 1986. Syllabification and epenthesis in Barra dialect of Gaelic. In K.

Bogers, M. Maus, and H. van der Hulst eds. The phonological representation ofsuprasegmentals. Dordrecht: Foris. 317-36.

Dimmendaal, Gerrit J. 1978. The consonants of Proto-Upper Cross and their implications forthe classification of the Upper Cross languages. Doctoraal Scriptie (MA thesis) LeidenUniversity.

Donwa-Ifode, S.O. 1985. Glide formation, Assimilation and Contraction: A reassessmentevidence from Isoko. Journal of West African Languages: XV 2, 41-55.

Donwa-Ifode, S.O. 1995. Basic phonetic. Sunray Publications Limited, 224 PortHarcourt/Aba Expressway, P M B 5089, Port Harcourt, Rivers State, Nigeria.

Everet, Daniel and Barbara and Ernest Buller. 1993. Stress placement, syllable structure, andminimality in Banaw, The International Journal of American Linguistics, 59: 3, pp 280-293.

Faraclas, N.G. 1986. Cross-River as a model for the evolution of Benue-Congonominalclass/concord systems. Studies in African Linguistics 17, 39 – 53.

Faraclas, N.G. 1989. Cross River In John Bendor-Samuel (eds.), The Niger-CongoLanguages, Pp. 377 – 399. Lanham.

Fery, Caroline and Ruben van de Vijver. 2003. The Syllable in Optimality Theory.Cambridge University Press.

Greenberg, Joseph H. 1978. Some generalizations concerning initial and final consonantclusters. In J. Greenberg ed. Universals of human language, 11: phonology. StanfordUniversity Press, 243-80.

Hall, T. Alan. 1999. The phonological word: A review. In Studies on the phonological word,eds. T. Alan Hall and Ursula Kleinhenz, 1-22.

Harris, James. 1983. Syllable structure and stress in Spanish: a non-linear analysis.Cambridge, Mass.: MIT Press.

Hyman, Larry M. 1982a. The Representation of nasality in Gokana. In The structure ofphonological representations. Part 1, eds. Harry van der Hulst and Norval Smith eds. 111– 130. Dordrecht: Foris.

Hyman, Larry M. 1982b. The Representation of length in Gokana. In Proceedings of thefirst West Coast conference on formal linguistics, (eds.) Daniel P. Flickinger, MarlysMacken and Nancy Wiegland, 198-206.

25

Hyman, Larry M. 1983. Are there syllables in Gokana? In Current approaches to Africanlinguistics 2, 171 – 179 eds. Jonathan Kaye, D. Sportiche and A. Dugas [Publications inAfrican Languages and Linguistics 5] Dordrecht: Foris

Hyman, Larry M. 1985. A theory of phonological weight. Dordrecht: Foris.Ikoro, Suanu M. 1989. Segmental phonology and lexicon of Proto – Keggoid (=Ogonoid).

MA. Thesis: University of Port Harcourt, Nigeria.Ikoro, Suanu M. 1996. The Kana language. Leiden: CNWS Publications.Isaac, Baridisi Hope. 2003. Studies in Gokana Grammar, PhD thesis: University of Port

Harcourt, Nigeria.Ito, Junko. 1986. Syllable theory in prosodic phonology. PhD thesis: University of

Massachusetts, Amherts, [Published 1988, New York. Garlard].Ito, Junko. 1989. A prosodic theory of epenthesis. Natural Language and Linguistics Theory

7. 217-60.Jakobson, R. 1962. Selected writings 1. The Hague and Bloomington: Mouton and Indiana

University.Kager, Rene. 1999. Optimality theory, Cambridge University Press.Kaye, Jonathan and Jean Lowenstamn. 1981. Syllable structure and markedness theory.In A.

Belletti, L. Brandi, and L. Rizzi (eds.) Theory of markedness in generative grammar:proceedings of the 1979 GLOW Conference. Pisa: Scuola Normale superiore di Pisa. 287-315.

Kenstowicz, M. 1994. Phonology in generative grammar. Oxford: Basil Blackwell.Lynch, John D. 1974. Lenakel phonology. PhD thesis: University of Hawaii.McCarthy, John J, and Alan Prince. 1993a. Generalized alignment. In G.E. Booij and J.Van

Marle (eds.) Yearbook of morphology 1993. Dordrecht: Kluwer. 79-153. McCarthy, JohnJ. 2002. A thematic guide to optimality theory. Cambridge: Cambridge UniversityPress.

McCarthy, John J, and Alan Prince. 1993b. Prosodic Morphology 1: constraint interactionand satisfaction. MS., University of Massachusetts, Amherts and Rutgers University.[Appear as Technical report no.3, Rutgers University Center for Cognitive ScienceCambridge, Mass.: MIT Press.]

McCarthy, John J. 2004. Headed spans and autosegmental spreading. University ofMassachusetts, Amherst. URL.

Ngulube, Isaac E. 2011. ‘The Eleme orthography’ in Orthographies of Nigerian LanguagesManual IX Ozo-Mekuri Ndimele (Ed.) Nigerian Educational Research & DevelopmentCouncil. Abuja.

Nwolu-Obele, Dada. 1998. Foundation Studies in Eleme. Outreach Publication. PortHarcourt, Nigeria.

Ohala, John J. 1990. The phonetic and phonology of aspects of assimilation. In J. Kingstonand M. Beckman eds. Papers in Laboratory phonology 1: Between them grammar andphysics of speech. Cambridge University Press. 258-75.

Payne, David L. 1981. The phonology and morphology of Axininca Campa. Arlington:Summer Institute of Linguistics Publications in Linguistics.

Popjes, Jack and Jo Popjes. 1986. Canela-Kraha. In D.C. Derbyshire and G.K Pullum eds.Handbook of Amazionia languages1. Berlin: Mouton DeGruther. 128-99.

26

Prince, Alan and Paul Smolensky. 1993. Optimality Theory: Constraint interaction ingenerative grammar. Report. News Brunswick, NJ: Rutgers University Center forCognitive Science. [Reprinted] Optimality Theory in Phonology: A Reader, ed. By JohnJ. McCarthy, Malden, MA and Oxford, Blackwell. 2004.

Schane, S.A. 1973. Generative Phonology. Holt, Rinehart & Winston, New York.Sommer, Bruce. 1981. The shape of Kunjen syllables. In D.L. Goyvaerts ed. phonology in

the 80’s Ghent: E. Story Scientia.Steriade, Donna. 1995a. Underspecification and markedness. In Goldsmith (1995), 114-74.Steriade, Donna. 1995b. Positional neutralization. MS., University of California, Los

Angeles.Vɔbnu, Stephen Kanen. 1991. Phonological Processes and Syllable Structures in Gokana,

MA thesis: University of Port Harcourt, Nigeria.Yip, Mora. 2002. Tone. Cambridge: Cambridge University Press.Yul-Ifode, S. O. 1999. A Course in phonology with an Appendix on: Williamson’s

distinctive features. Riverside Communications, P O Box 7390, Federal Secretariat, PortHarcourt, Nigeria.