Past Event

A Whole Word Typology

Dr. Luc Baronian
(UQAC)

Hockett (1954) famously divided the world of morphological models into Item and Arrangement (IA), Item and Process (IP) and Word and Paradigm (WP) . In the decades that followed, the term “word-based morphology” started being using ambiguously. It can now refer to a traditional WP model or to a form of IP model where the item, strangely, is sometimes a stem, sometimes a word (cf. Aronoff 1976, Stump 2001 and Blevins 2003 and many others). Ford and Singh (1991) proposed a minimalist (with lower case <m>) model of morphology that is truly word-based, yet does not rely on paradigms. The model, in conjunction with its Lexicase companion (Starosta 1989), came to be known under the rubric of Seamless Morphology (Singh and Starosta 2003). In more recent years, the label Whole Word Morphology (WWM) has been adopted for Ford & Singh’s model, which is not only compatible with Lexicase Theory, but also with other lexicalist syntax models (see Baronian 2003 for HPSG). The single mechanism proposed by WWM is termed the strategy, and it minimally relates two words as follows:

/ X /a <–> / X’ /b

where (to simplify) “a” and “b” are morphosyntactic categories represented as feature bundles, while X and X’ are related word forms that are also related in meaning. At first sight, one may legitimately wonder whether this is too loose a characterization of morphological patterns. We shall see that, in practice, the notions of variable and constant defined in the theory constrain the predicted morphological patterns in a surprisingly accurate way. Two constraints are necessary to do so: 1) constants can be discontinuous, but must be made fully specific; 2) variables must be made (at least partially) specific when they are discontinuous.

When we define the notions of variables and constants correctly, WWM predicts a typology of affixal morphology, including prefixing, suffixing, circumfixing, infixing, ablaut and transfixing. The point of this talk is not to claim that WWM solves every problem or even that it is superior to other more fashionable models. Rather, the point is to invite the audience to reflect on how a sizeable amount of data can be explained by a relatively reduced set of assumptions on the nature of a linguistic module.

About the Speaker

Dr. Luc Baronian received his PhD from Stanford in 2006. He has contributed to a broad range of subfields: theoretical morphology, historical phonology of Armenian, French dialectology in North America, the study of Louisiana Creole, and the grammar of poetry and textsetting.