Research
The bulk of my research explores sources and ramifications for two opposing tendencies evident in language change: a tendency toward pattern-coherence, and a tendency to preserve semantically relevant contrasts.
Computational simulation is my method of choice to explore the interactions of these factors in pattern formation in language. For a quick example of this research method exploring the evolution of constrained phoneme inventories, run this short powerpoint presentation (requires Quicktime player).
Pattern-coherence arising through similarity-driven feedback
Two general observations about the distribution of sound patterns in language serve as the central organizing problems in the field of phonology.
- Although sound patterns vary between languages, individual languages tend to exhibit given sound patterns categorically across the entire lexicon. For example, if a language exhibits categorical final-obstruent devoicing in some forms, it is likely to do so in all forms. A central strand of my work provides a novel account for this observation based in evidence for feedback-driven pattern-extension in language change (Wedel 2002, 2006, 2007). You can explore a simple simulation of feedback-driven pattern extension on the simulation page of this website.
- Many pattern-types recur in unrelated languages. Continuing our example, final–obstruent devoicing occurs with minor differences in many genetically unrelated languages such as Russian, Turkish and Maltese. My work strongly supports an account for this observation provided by the Evolutionary Phonology framework (Blevins 2004), by articulating a mechanism that distinguishes factors responsible for the initiation versus the entrenchment of phonological patterns.
- More background and an overview of results
Preservation of semantically-relevant contrasts through category competition, or how do sounds stay different over the course of language change?
A long-standing problem in phonology involves the long-term maintenance of sound contrast within a language. We tend to progressively cut corners in highly-practiced speech, and sounds do often merge together over time as predicted by a bias toward similarity and pattern-coherence (see previous section). Given these processes, how do languages avoid eventually devolving to complete homophony? In a functional sense, the answer is obvious: if all our words sounded the same then we wouldn’t be able to convey meaning. However, the actual mechanism for keeping sounds distinct over the course of language change has not been clear.
Theoretical models over the last century have all, in essence, postulated some innate 'watchdog' which overtly monitors sound contrast (e.g., Martinet 1955, Flemming 1998). In recent work (Wedel 2004, Blevins and Wedel to appear, Wedel in prep, Wedel and Fatkullin in prep), I have articulated an alternative model in which the maintenance of sound contrast over time is indirectly driven by the functional role of sounds in conveying distinct word meanings. This model is based on the recognition that clear pronunciations are more consistently mapped to particular word category than ambiguous pronunciations. Because of this difference, contrastive pronunciations exert a cumulatively greater influence on the long-term evolution of the sound system as a whole, which indirectly results in a tendency to preserve contrast between most sounds over the long-term.
On the simulations page of this website you can explore factors influencing contrast maintenance via this mechanism in simulations over a one-dimensional space or two-dimensional space.
In one form of this model, this mechanism for contrast maintenance is functionally parallel to the biological phenomenon of niche-specialization in which two similar species that jointly colonize a new environment quickly evolve away from one another, driven by pressure to split the available resources rather than to compete (Wedel 2006).
This work shows that mechanisms supporting linguistic contrast maintenance may lie in cognitively-general properties of categorization behavior. Even while these mechanisms are ‘innate’, this model for contrast maintenance cannot be described solely in terms of a physical structure residing within a single person’s mind – it is in important ways dependent on communicative goals, history, and chance, and therefore must be understood as being distributed both over time and speakers. I am currently collaborating with an undergraduate student in Psychology and Cognitive Science, Jim O’Donnell, to develop a web-based experiment to test predictions of this model in an implicit category learning task.
A few news stories have recently appeared on more work of mine on contrast maintenance in language, framing this work in terms of the innate/emergent debate on language patterning. To clarify, I provide another summary of this work here with specific reference to the ongoing innateness debate.