The Abstract Nature of the Bondu Vowel System: Evidence from [ATR] Harmony

Abbie Hantgan, Stuart Davis


This paper argues for an abstract analysis of the vowel system of Bondu, a Dogon language of Mali. Data come from fieldwork and have not been previously published. Phonetically, Bondu has seven vowels: two [+high, +ATR] vowels ([i], [u]), a [+low, –ATR] vowel [a], and a [±ATR] contrast in the mid vowels: front ([e], [?]) and back ([o], [?]). We argue, however, that underlyingly certain high vowels are [–ATR] while some low vowels are [+ATR]; the contrast is neutralized so that high vowels surface as [+ATR] and low vowels as [–ATR]. Evidence comes from the realization of the perfective suffix /–??/ which alternates between [è] ~ [??], depending on the underlying [ATR] value of the vowel in the verb root. The data in (1) illustrate root-controlled [±ATR] harmony. (Forms show 3rd person singular).

(1) a. [nòj–è] sleep c. [d???–??] leave
b. [nèmbìl–è] beg d. [k???–??] cut

The data in (2) are more complex.

(2) a. [bìj–è] lie down c. [?ìj–??] dance
b. [sù?–è] go down d. [?ù?–??] recognize

While all the root vowels in (2) are phonetically [+high, +ATR], we analyze those in (2c-d) as having an underlying [–ATR] feature. Here we follow Archangelli and Pulleyblank (1994) who view the feature combination [–ATR], [+high] as antagonistic: phonetically unrealized, though phonologically present. Similarly, while there is only one surfacing [+low, –ATR] vowel in Bondu, verb roots with low vowels are divided between those that take a [+ATR] suffix (3a) and those taking [–ATR] (3b).

(3) a. [bàr–è] help b. [pà?–??] tie

We analyze the surface low vowel in (3a) as abstractly [+low, +ATR], (3b) as [+low, –ATR] with the underlying [ATR] feature of the root spreading to the suffix.
A further argument for the abstract feature analysis comes from the complex alternations found with the imperative suffix in (4) (same roots from above).

(4) a. [nój–ó] f. [dó?–á]
b. [némbíl–ó] g. [ké?–á]
c. [bíj–ó] h. [?íj–á]
d. [sú?–ó] i. [d?ú?–á]

e. [bár–á] j. [pá?–á]

We analyze the realization of the imperative suffix vowel by spreading of the underlying [ATR] feature of the root vowel as in the perfective, but with an additional assimilatory process raising the underlying [+low] suffix vowel to [–low] when preceded by a vowel that is underlying [–low, +ATR]. We view this as an instance of parasitic harmony which applies in (4a-d). The unexpected realization of [+ATR] on the stem vowels in (4f-i) is analyzed as the docking of a floating [+ATR] feature that comes with the imperative suffix. We argue that our abstract feature analysis is superior to an alternative under-specification analysis since there is no consistency whether [+ATR] or [–ATR] is underspecified given the behavior of high vowel roots in (2) and (4); our analysis of the imperative is also consistent with Wolf's (2007) observation that floating autosegments avoid docking on morphemes that sponsor them.
Archangeli, D. and D. Pulleyblank (1994) Grounded Phonology. MIT Press: Cambridge, MA.

Wolf, M. (2007). For an Autosegmental Theory of Mutation. In L. Bateman, M. O’Keefe, E. Reilly & A. Werle (Eds.), University of Massachusetts Occasional Papers in Linguistics 32: Papers in Optimality Theory III (Vol. 32 pp. 315-404). Amherst: GLSA.

Full Text: