In this talk I will compare how different approaches to gender assignment account (or not) for the data with high variability with a special focus on the role of the surface noun form – the surface similarity and frequency of form patterns. We will test Distributed Morphology, Relational Morphology, Network Morphology and Optimality Theory, in particular Hierarchical MaxEnt.
The data present a series of corpus and experimental studies for every group of words : we will look at common gender animate nouns (hybrid nouns), expressive nouns - animate and inanimate, inanimate indeclinable nouns and, finally, at the gender agreement attraction errors. All these nouns have ambiguous or contradictory cues for gender assignment.
For the expressive nouns we will look at the different structural and phonological factors affecting variation in gender agreement and discuss approaches to the problem of gender markedness in Russian. I will also introduce a new approach to this problem, relying on the influence of the surface form. In a nutshell, my idea is that neuter is the unmarked gender language-wide, while masculine is the greedy class due to its frequency and diversity.
Considering hybrid nouns, we analyze how gender and case are interconnected: why it is only in the nominative that some nouns are acceptable with a feminine agreement. We also look at how their acceptability in oblique cases is affected by syncretism. It appears that they are especially problematic in locative and the reason for that are the syncretic patterns of inflectional affixes.
Indeclinable nouns are of special interest, as they do not have stem-final inflexional affixes, hence phonological influence is the influence of the last one or two phonological segments of the root, which makes the word look to belong to one of the declension classes. And the root is morphologically indivisible in all structural approaches.
Finally, we will look at the gender agreement attraction errors - compare the influence of head syncretism and the dependent noun syncretism on attraction effects in processing.