A paper by Professor Andries Coetzee (in collaboration with Professor Shigeto Kawahara from Rutgers University) just appeared in Natural Language and Linguistic Theory. Andries has been working for the several years on developing generative models of phonological grammar that can account for variable phonological phenomena. Given that most standard models of generative grammar are by design categorical in nature, this presents quite a challenge. However, in a series of publications over the past six years, Andries has shown that constraint-based models of generative grammar (both Optimality Theoretic and Harmonic Grammars) are actually quite well suited to handle variation. These models can retain many of the core results that the generative phonological endeavor has achieved over the past several decades while also allowing the grammar to generate variable outputs. In this paper, Andries and Shigeto go one step further and propose a hybrid model of the phonological component of grammar that incorporates the best of both traditional generative grammars and usage-based/exemplar models of phonology. These two approaches to phonological grammar are often presented as incompatible, and yet the patterns observed in the linguistic behavior of speakers and listeners show that there are aspects of truth to both approaches. In this paper, Andries and Shigeto develop a generative model of phonology (in the Harmonic Grammar framework) that allows rich memory encoding of experience with actual language use (a core component of exemplar and usage-based models of grammar) to impact the functioning of the basically generative grammar. They apply their model to variable phenomena in both English and Japanese, and show that their hybrid model accounts significantly better for the data than a model that is exclusively generative or exclusively exemplar/memory based. The full bibliographic information of their paper, including a title and abstract, is given below.
Coetzee, Andries W. & Shigeto Kawahara. (2013) Frequency biases in phonological variation. Natural Language and Linguistic Theory, 31(1):47-89.
In the past two decades, variation has received a lot of attention in mainstream generative phonology, and several different models have been developed to account for variable phonological phenomena. However, all existing generative models of phonological variation account for the overall rate at which some process applies in a corpus, and therefore implicitly assume that all words are affected equally by a variable process. In this paper, we show that this is not the case. Many variable phenomena are more likely to apply to frequent than to infrequent words. A model that accounts perfectly for the overall rate of application of some variable process therefore does not necessarily account very well for the actual application of the process to individual words. We illustrate this with two examples, English t/d-deletion and Japanese geminate devoicing. We then augment one existing generative model (noisy Harmonic Grammar) to incorporate the contribution of usage frequency to the application of variable processes. In this model, the influence of frequency is incorporated by scaling the weights of faithfulness constraints up or down for words of different frequencies. This augmented model accounts significantly better for variation than existing generative models.