|
Computational Cognition
Cognition & Communication
Research Studies & Opportunities
Downloads
|
Cognition and Communication Laboratory
Informal Research Description
General Questions
Current Projects
Links To More Info
How do people manage to speak? That is, how do they take
some idea that isn't expressed in words and express it out loud? Right now we
in the Cognition and Communication lab are particularly concerned with the timing
of it all. Many people have assumed that fluent speech requires advanced planning
- that speakers decide pretty far in advance which words to use and the order
in which they will say them. Recent work in the lab has focused on showing that
people don't prepare words far in advance of saying them, even when they sound
perfectly fluent (like news-anchors). Instead, people often make decisions
about what to say next and which words to use at the last possible moment. If
they are lucky, the timing works out so they sound fluent. Often though, speakers
are disfluent - meaning that they pause in the middle of sentences, stretch
words out, restart sentences, or say "uh" or "um". When speakers are particularly
concerned about sounding fluent, they seem to estimate how little advanced planning
they need to do to have words ready to say when needed. These estimates increase
the likelihood that speech will be fluent, but these estimates can be sometimes
wrong.
A fun and challenging aspect of the research is that we don't know what ideas
look like, so we don't know what speakers start with when they speak. However,
to get interpretable results, we often have to make sure that multiple speakers
start off with the same ideas and end up saying the same sentences. This makes
it necessary to come up with new ways of creating and changing the ideas that
speakers express, and new ways of measuring the processes we think are involved
in speaking. One of the new methods is to have people describe pictures while
we monitor when they look at different objects in the pictures. Under these
circumstances, people tend to look at objects during the second before mentioning
them. In other words, their eyes are about a second ahead of their mouths. Monitoring
what speakers look at seems useful for predicting what they will talk about
and when they prepare the names of objects. We can then use this information
to figure out more about other aspects of speaking.
General Questions
So general questions of interest in the lab are: How does
language production work? How are conceptual representations mapped on to words
and phrases? What determines the grammatical roles that words and phrases occupy
in a sentence? What are the computational properties of the production system
that underlie relatively error free speech? How might processes of language
production and comprehension be related? What features might language production
share with other cognitive processes? How is the sequencing and timing of word
production in speech related to the sequencing and timing of motor movements
in other activities (such as dancing)? How do individual differences such as
age and dialect affect word choice and the timing of speech?
Current Projects
Current projects investigate the time course for selecting
words and making syntactic decisions when producing sentences, how speakers
monitor their speech preparation and vary their degree of planning, how the
cognitive changes associated with normal aging affect language production, and
cross-linguistic studies of production, particularly word order.
We study these questions and others using traditional measures of performance
such as response latencies and response probabilities, in addition to eye movement
data. Computational modeling of existing data provides predictions for future
research.
Additional Readings
Click on these links for more on
|
