Third Mediterranean Meeting on Morphology (MMM3)

Abstract

J. Trommer (University of Osnabrück)
The Role of Syntax and Morphology in Affix Order

Most syntactic accounts of affix order in inflection are actually »hybrid«: Closeness of affixes to stems reflects syntax, but the position of affixes w.r.t stems is determined by morphological stipulation. In my talk I argue that affix order is in fact subject to two different sets of principles, but the borderline is quite different: Affix order of contentful items like Tense and Aspect is computed in syntax, while the order of agreement heads is the result of of universal alignment constraints applying in a post-syntactic morphology module according to the Principles of Optimality Theory. The basic observation that leads to this model, is that crosslinguistically the order of agreement markers patterns quite differently from that of other inflectional affixes. For example, Julien (2000) observes that Aspect affixes always follow Tense markers in linear order, as long as Tense is not a suffix, in which case the order is reversed:

	both Prefixes	Mixed	both Suffixes
Tense > Aspect	Tense Aspect Verb	Tense Verb Aspect	*Verb Tense Aspect
Aspect > Tense	*Aspect Tense Verb	*Tense Verb Aspect	Verb Aspect Tense

Conversely, in languages where subject agreement is split in affixes marking Person and Number, a typological survey shows that the Number marker always follows the Person affix, regardless of their position w.r.t. the stem:

	both Prefixes	Mixed	both Suffixes
Person > Number	Person Number Verb	Person Verb Number	Verb Person Number
Number > Person	*Number Person Verb	*Person Verb Number	*Verb Number Person

A similar pattern can be found for the relation between (portmanteau) inversion marking and subject agreement in language families like Algonquian or Chukotko-Kamchatkan.

I'll call the pattern found for Aspect and Tense the »Movement Pattern« and the one found with agreement markers the »Alignment Pattern«. The Movement Pattern naturally follows from a Kaynean approach to syntax, where affix order results from uniformly leftward movement. Given standard assumptions about the universal architecture of the verbal extended projection the orders Tense > Aspect > Verb and Tense > Verb > Aspect reflect the base position of Tense and Aspect while Verb > Aspect > Tense results from the movement of a complex containing the Verb and Aspect to the left of Tense.

On the other hand, the Alignment Pattern is supposed to follow from the ranking of violable constraints in the sense of Optimality Theory , which align person features to the right and number features to the left edge of morphological words. Given the possibility of different constraint rankings, this also allows to derive the position of agreement affixes w.r.t. stems without any further stipulation and explains the fact that the most frequent order for (subject) person and number affixes in the languages of the world is Person > Verb > Number. Note that the respective order of Person and Number is analyzed here as an epiphenomenon of more specific constraints.

Taken together, the accounts for the Movement and the Alignment Pattern obviate the need for any arbitrary stipulation in affix order. Crucially, it is the splitting up of affix ordering mechanisms in two separate components of the grammar (syntax and morphology) that allows to reduce it to gene-ral principles. Since agreement nonetheless interacts with other inflection (e.g. in port-man-teaus) and typically appears »outside« of it, this is naturally captured in a model, where morphology applies after syntax as in Distributed Morphology (Halle & Marantz, 1993).