Agreement , history , and Obolo : A reply to Connell

The Obolo nasal agreement pattern first observed by Faraclas (1984) appears to be ‘static’ in nature: we do not see alternations arising from it, which makes it difficult (if not impossible) to determine the direction of the agreement. The generalisation is that if a closed syllable has a nasal consonant, then its coda must also be nasal. Although Obolo has the requisite morphology to produce hypothetical underlying forms with /...NVC.../ sequences, the language has a complex array of morpheme-boundary phenomena, including deletion and neutralisation of consonants, which obscure such sequences in ways that are orthogonal to the syllable-internal nasality agreement. As such, though we know that disharmonic NVT syllables are illicit on the surface, it is not clear whether the phonology of Obolo would respond to such an input by changing it to NVN, or to TVT (both of which are allowed).


Circumstantial support for left-to-right directionality
Connell's comparative survey reveals key support for one point of conjecture in the original proposal: the directionality of agreement.
The Obolo nasal agreement pattern first observed by Faraclas (1984) appears to be 'static' in nature: we do not see alternations arising from it, which makes it difficult (if not impossible) to determine the direction of the agreement.The generalisation is that if a closed syllable has a nasal consonant, then its coda must also be nasal.Although Obolo has the requisite morphology to produce hypothetical underlying forms with /…NVC…/ sequences, the language has a complex array of morpheme-boundary phenomena, including deletion and neutralisation of consonants, which obscure such sequences in ways that are orthogonal to the syllable-internal nasality agreement.As such, though we know that disharmonic NVT syllables are illicit on the surface, it is not clear whether the phonology of Obolo would respond to such an input by changing it to NVN, or to TVT (both of which are allowed).Bennett (2014) argues for the former, NVT→NVN, on theory-internal grounds.The ranking conditions required to prevent agreement from obtaining in TVN syllables, where disagreement is tolerated, also dictate that faithfulness for [+nasal] outranks faithfulness for [-nasal].As such, the ranking posited to analyse Obolo does not allow nasality to be removed, even for the sake of agreement: disharmonic NVT syllables must surface as NVN, and not TVT.Though alternations are never observed to confirm it, the agreement must operate from left to right, from onset to coda.Connell's (2015) comparison to more distantly-related Upper Cross River languages provides further support for this conclusion, on entirely independent grounds.Forms like *m d 'swallow', reconstructed as disharmonic in Upper Cross River, have Obolo cognates that show the result of left-to-right nasal harmony: m n (not *p t).Thus, the historical context helps illustrate a generalisation lurking behind the data, rather than evident directly in it.The left-to-right directionality is not a mere theoretical gimmick; left-to-right agreement appears actually to have happened historically. 1

Heterogeneity in the origins of long-distance agreement?
In light of Obolo's static nasal agreement, it is extremely tempting to infer a historical pathway of assimilation.Obolo has no NVT syllables now; but some historical precursor surely did.So what happened to those disharmonic words, in the course of a shift to a language where harmony is lexically uniform?The most intuitive explanation is that there was a historical change: sequences like *m t (a possible pre-Lower Cross intermediate form from proto-Cross River *m d 'swallow') shifted to agreeing ones like [m n].
But there is another possibility: the pattern could also have arisen from accumulation of lexical items that obey the harmony generalisation, combined perhaps with the piecemeal elimination of disharmonic forms.This idea was recently suggested by Coetzee (2014) for Afrikaans.Afrikaans has virtually no underlying CVC roots with two obstruents that disagree for voicing, and Coetzee demonstrates that L1 speakers of Afrikaans extend this pattern to novel nonce words; it is a part of their synchronic grammatical knowledge.But the history of Afrikaans does not show evidence for systematic long-distance voicing agreement.Rather, a constellation of other changes (gliding of intervocalic /d/, loss of voiced /ɣ z/, sporadic addition of final vowels) caused the elimination of many disharmonic roots, while others were replaced by borrowings.The result is that a gap emerged almost coincidentally, and speakers are able to detect and learn it as part of the grammar.
Could the same thing have happened in Obolo?It is quite difficult to say without more extensive data, particularly from more historical sources.But there are some whispers of this in the data that Connell (2015) presents The word 'dance', for example, has the shape NVT in many ower Cross languages (e g kit n k) If the phonology of Obolo (or some antecedent thereof) applied nasal agreement productively, this root would be a good candidate to show it in action But the Obolo word for 'dance' is entirely unrelated: dʒ p And this is not a lone example For all of the NVT forms that Connell identifies as 'pattern 3'the exceptions to nasal agreement -Obolo seems to have non-cognates instead.This would seem to be consistent with a model in which Obolo's agreement is not only the result of a gradual preservative nasal spreading process in the area (as Connell suggests), but rather was led in part by the grammar.If sporadic cases of nasal agreement accumulated to the point that pre-Obolo speakers began to learn it as part of their phonologies, that could provide an impetus for replacement of remaining disharmonic forms, even through non-phonological mechanisms (such as shifting to other lexical items, including perhaps loanwords).A lexical accumulation account thus has the potential to offer a more comprehensive explanation than historical gradualism: it can explain not just why Obolo lacks NVT syllables, but also why it has non-cognates in place of disharmonic forms in related languages -and perhaps even why the language's morpheme boundary alternations obliterate situations that could show the harmony in action.
1 In this respect, a number of Lower Cross River forms are puzzling, as Connell (2015) points out.In Ebughu and Enwang, we find [_p_k], seemingly cognate with [_m_k] in Ekit and Ibibio.Connell (2015) asserts that these are false cognates, on the grounds that [p] in the former two languages normally corresponds to [kp] elsewhere.But it is also a fact that most ower Cross languages do not have labiovelar nasals except in homorganic nasal-stop clusters (like mk p ): the distinction between bilabials and labio-velars is neutralised among nasals, leaving [m] as the most likely product of nasalising *kp.