|
Lab personnel: Jenny R.
Saffran, Director |
|
The Learning Languages Lab is located in the Psychology Department, UW-Madison. Adults are tested in the Psychology Department, children are recruited and tested at elementary schools with permission from their parents. We also test children in the Infant Learning Lab, located in the Waisman Center.
|
|
Our lab studies how children and adults learn, specifically, what mechanisms they use to acquire a new language. To do this, we make up artificial languages and expose adults and children to them and see how much they can learn. These languages consist of nonsense words, shapes or nonlinguistic sounds. To hear a sample of one of our languages, please press here. Selected references can be viewed by clicking here.
Here are some of the questions that our lab tries to answer:
Where does syntax come from?
To what extent are the statistical properties of languages useful for the acquisition of other aspects of language? We are currently investigating the acquisition of phrase structure, an important component of syntax, and asking whether children and adults can detect structure via statistics. Children and adults listen to a tape containing a nonsense language that for kids we call Silly Speak. To distract children and adults from thinking too hard about the made-up languages they play with Legos or engage in a coloring task; below are some of the artistic byproducts of this research.
We have found that learners, including both adults and children, are able to capitalize on statistical cues to phrase structure in language learning -- the dependencies between different types of words in phrases (Saffran, 2001). To reveal constraints on the particular types of computations performed by language learners we contrasted the acquisition of two types of grammars by adults and children (Saffran, 2002). One of the languages used in this research was a small phrase structure grammar (Language P, for predictive), in which dependencies between word categories afforded predictive cues to phrases, as in natural languages ). The second language was equally complex in terms of its size and formal characteristics, but did not contain predictive dependencies marking phrases (Language N, for non-predictive). If predictive dependencies between categories serve as a cue to phrase structure, then Language P should be learned more successfully than Language N. Both adults and children showed this pattern of learning, suggesting that predictive dependencies play a role in statistical learning of syntactic structures, and raising the interesting possibility that natural languages possess this type of structural organization in part because human learning mechanisms utilize predictability to discover units like phrases.
To determine whether the same types of regularities are detected and used in both linguistic and non-linguistic learning tasks, we also exposed adult learners to non-linguistic versions of these languages, implemented in vocabularies of non-speech sounds and visual shapes (Saffran, 2002). The same pattern of results emerged for the non-linguistic auditory materials as seen with linguistic auditory materials: Language P learners outperformed Language N learners. These results are consistent with the hypothesis that predictive dependencies enhance learnability. The mode of presentation of visual materials affected performance, with sequentially presented visual materials not showing the predictive dependencies effect, but with simultaneously presented visual materials showing the predictive dependencies effect. The results suggest a complex interaction between modality, mode of presentation, and statistical learning that may be tied to the manner in which information is characteristically encountered in the visual versus the auditory environment.
In related research, we are beginning to ask how learners acquire "languages" when they contain no linguistic information at all? These languages contain "sentences" of shapes or nonlinguistic sounds. We can then ask whether learners acquire these non-linguistic "languages" in the same way as they acquire linguistic languages. Currently, Melissa Lehman is assisting in the design of a new tactile version of this experiment. We want to know how people learn artificial languages through their sense of touch. We intend to recruit both sighted and blind participants to take part in our study. We expect blind participants to outperform sighted participants in this language learning task. If you are interested in participating, please call us at 265-0592.
Why do children outperform adults in language learning?
Across all of these research enterprises, we are interested in asking why children are better at learning languages than adults, despite the fact that adults are generally better learners than children in other domains. Are there other learning tasks other than language on which children outperform adults? We hope to further explore how adults and children learn differently en route to better understanding the critical period for language acquisition.
How do people learn & process multiple languages?
One of the most fascinating questions in language acquisition is how learners acquire and use multiple languages. In her honors thesis, Florencia Anggoro is studying the relationship between words in bilingual adults' first and second languages. We are also developing paradigms for exposing learners to multiple artificial languages at once, distinguished by their accents, to ask whether learners can acquire more than one set of statistics at the same time.
Research Opportunities: Psychology Research Experience Program (PREP)
Rachel Morikone, an undergraduate student from California State University, Long Beach is pictured below presenting her research, "Using word segmentation and accent to learn languages" conducted in the Saffran lab to other students and faculty involved with the PREP program at UW-Madison. The PREP program gives undergraduate students an opportunity to engage in a summer research project with guidance of faculty members.
References
Saffran, J. R. (2002). Constraints on statistical language learning. Journal of Memory and Language, 47, 172-196.
Saffran, J. R. (2001). The use of predictive dependencies in language learning. Journal of Memory and Language, 44, 493-515.
If you are interested in our other research, visit the Infant Learning Lab page, or Jen Saffran's research home page. Comments about this page can be directed to Rebecca Seibel at seibel@waisman.wisc.edu.