What is Usage-based Linguistics?
Usage-based linguistics studies language as a dynamic, constantly changing system. We see the main goal of linguistics to be explaining why languages are the way they are by explaining why languages change in the ways they do. We search for these explanations in cognitive, perceptual and motor processes that operate during language learning and language use. Every time you hear a language, the mental representation of the language in your brain changes. Every time you speak, you change both your brain and the brains of everybody in hearing range. Every time you write, you change your own brain and the brains of your readers. We are interested in how one’s mental representation of language change as a result of any given experience with language, either in hearing, speaking, reading, or writing. What do you take away from the experience? How will your future use of language change as a result of this minute change to your memory of language? If it changes at all, then this experience will have planted a seed of change, available to be carried around the community by the winds of fashion.
To read about the approach in more detail, check out this paper
We believe it to be self-evident that language change can occur only through micro-changes in observable behavior (which happen as a result of language experience passed through the cognitive filter of the learner’s brain). Thus, the work of this lab is at the level of observable behavior. We pursue several research strategies used in concert to attain the overall goal.
One research strategy is to identify patterns in real languages and then see whether those patterns are picked up on by the learners of those languages. In particular, we can compare patterns that vary in how frequent they are, and thus in how much experience speakers of the language have with them. We can also examine patterns that vary in their statistical robustness, or in how much of a demand they place on the learner’s cognitive resources like short-term memory. Another important strategy is to expose learners to various language patterns, some of them modeled on the patterns found in real languages, in a controlled laboratory environment. Here, we can manipulate both the patterns and the way the learners experience them. Then we can see not only what patterns the learners pick up, but also what kinds of experiences are crucial for picking up the patterns.
Finally, another strategy is to model documented language changes and try to infer the neuromotor mechanisms that are necessary to explain them. The behavioral consequences predicted by the proposed mechanisms can then be tested in the laboratory. The work described so far views experience, or repetition, as helping the learner achieve the target behavior. However, according to usage-based linguistic theory, this is not the only effect of repetition. In particular, repetition leads to automatization of production, which makes the production process streamlined and less accessible to conscious intervention. Repetition of a non-target behavior may also lead the learner to perseverate on that behavior. Thus repetition may make productions deviate from the community target as well as shifting the community target over time. This is the mechanism thought to be responsible for reductive kinds of sound change (deletions, shortenings, etc), which are so prevalent in languages. Examining the effects of frequency, predictability and repetition on the production target forms another major line of work for the lab.
A lot of the work we do involves working with corpora, large collections of text or speech. Corpora are invaluable for identifying patterns in language use that serve as the target of acquisition as well as for examining influences on real-life speech and language production where language change happens. However, corpora are also difficult objects for statistical analysis. Language is highly redundant. Most linguistic features are highly collinear and enter into complex interactions. Language is also subject to pervasive rich-get-richer dynamics, in which using something makes it more and more likely to be re-used in the future. Most words in a corpus occur only once, while a few (like the) occur all over the place. Identifying best practices for properly dealing with corpora in statistical analysis is therefore also a major concern of our lab.
For summaries of our papers and links to related work (how it all fits together), see here
Published and in-progress papers can also be seen in reverse chronological order here