The Weekly Newsletter of MIT Linguistics

MorPhun 4/29 (at 12:30!): Hagen Blix (NYU) on Arabic agreement

Speakers: Hagen Blix (NYU)
Title: (Dis)continuous Bleeding: Cyclicity and Spans
Time: Monday, April 29th, 12:30-2pm (note the unusual time)
Location: 32-D831

Abstract: Modern Standard Arabic verbal agreement shows two properties that I argue to have important implications for the theory of interpretative/post-syntactic morphology. Firstly, it shows affixes whose distribution cannot be described as natural classes in the SPE-style subset principle of Distributed Morphology (Halle & Marantz 1993). In example (1), we see that the prefixal y/tencodes a gender contrast in the third person singular of the subjunctive.

(1) a. 3m.sg: y-aktub-a 
b. 3f.sgt-aktub-a

While this clearly suggests that at least one of the affixes spells out gender, both affixes occur in both genders if we look beyond the third person singular. In the third person plural, both genders occur with y (2), whereas the second person plural shows both genders with t (3), with both forms marking the gender distinction at the suffixal position.

(2) a. 3m.pl: y-aktub-uu
b. 3f.pl: y-aktub-na

(3) a. 2m.pl: t-aktub-uu
b. 2f.pl: t-aktub-na

This rules out an analysis of the initial contrast in (1) in terms of the subset principle alone. Halle (2000) offers an analysis in terms of impoverishment, with t as an elsewhere marker and neither affix spelling out gender, but these impoverishment rules violate otherwise established markedness constraints (Noyer 1992, Nevins 2011). In particular, Noyer (1992) argued that in case of a “feature clash”, only the feature lower on the hierarchy in (4) can be impoverished:

(4) Person > Number > Gender

Secondly, various affixes, including the aforementioned t occur either as prefixes or as suffixes on the verb, depending on the Tense/Aspect configuration (5-7). Puzzlingly, while the prefixal forms of 2m.sg and 3f.sg are identical (6a), (7a), their suffixal counterparts in the perfect are distinguished not by different affixes, but rather by the order in which they occur (6b), (7b).

(5) a. 1pl.sbjv: n-aktub-a
b. 1pl.perf: katab-n-aa(6) a. 2m.sg.sbjv: t-aktub-ab. 2m.sg.perf: katab-t-a(7) 3f.sg.sbjv: t-aktub-a3f.sg.perf: katab-a-t

This is unexpected from the perspective of any theory that argues that prefixhood and suffixhood are part of the vocabulary item/rule itself (Halle & Marantz 1993, Noyer 1992, Halle 2000).
In this talk, I will argue that the hierarchy Noyer (1992) described as governing Impoverishment actually describes an inverse of the f-seq of Arabic agreement, and that spellout operating on spans, regulated by a superset principle (Starke 2009), can capture the distribution of these affixes without recourse to Impoverishment. I then show that a movement-based, antisymmetry approach to linear order can derive the positional effects at PF (Kayne 1994, Kayne 2017, Koopman 2018, Julien 2002, Blix under review) in a way that characterizes the affixes that may appear in prefixal position as a uniform set (as opposed to an arbitrary collection of affixes, as in previous work), and allows us to connect their paradigmatic distribution with their linear one.