Fri September 27, 2013
Dr. Jodi Tommerdahl, University of Texas at Arlington – Analyzing Language Samples
In today’s Academic Minute, Dr. Jodi Tommerdahl of the University of Texas at Arlington explains efforts to understand how children use language in different social settings.
Jodi Tommerdahl is an associate professor of curriculum and instruction at the University of Texas Arlington where she works in the Southwest Center for Mind, Brain, and Education. Her research interest lies at the intersection of language and cognition, with a particular focus on the acquisition of linguistic and cognitive skills in childhood as well as their loss in cases of brain damage. The research project discussed here was conducted in partnership with Dr. Cynthia Kilpatrick, an assistant professor of linguistics at the University of Texas Arlington.
Dr. Jodi Tommerdahl – Analyzing Language Samples
When scientists study groups of people, they use statistics to define a representative sample of the population. But how do we do that with language? What is a ‘sample’ of language? Many language studies record short snippets of a person’s language, maybe one or two hundred sentences, but almost nothing is known about how representative these samples are of a person’s typical language. For instance, if I record Jimmy talking to his mom at the same time and place two days in a row, will his language be pretty much the same or different? And how might it differ? Will the grammar change? This is what we’re asking.
To find out, we recorded 25 2 and 3-year-olds while they played with a parent. We recorded them twice, and kept the context similar both times to isolate the natural differences as opposed to grammatical differences that might result from changing contexts. We chose several grammatical items such as the past tense and compared how many times each was produced in each sample. And we looked at different lengths of samples. The results showed a generally low degree of reliability, even when we looked at longer samples.
So why do this? We want to build tools that can help to recognize Specific Language Impairment, or SLI, where children have normal IQ but weak language skills. SLI affects a lot of children but it is hard to identify and language samples are important in diagnosis. Knowing whether a 200-sentence sample tells you much more than a 100-sentence one is very important for clinicians working with child language, as recording, transcribing and analyzing samples is a very time consuming process. Also, the grammatical items we look at are those, which may be indicators of SLI. We want to find ways to diagnose language impairment earlier so children can get the help they need.