Sharing data with other fields yields important discoveries

Public databases on the Internet allow different disciplines to share each other?s wealth.

By Beth Azar
Monitor staff

Linguist William Snyder, PhD, has saved a lot of time and energy thanks to a database of child-language samples collected by an interdisciplinary mix of researchers and compiled by two psychologists.

Without Carnegie Mellon University?s Child Language Data Exchange System (CHILDES), which maintains a massive collection of child- and adult-language transcripts, Snyder would have had to spend years collecting longitudinal samples of language from an international set of small children.

Instead, like hundreds of other researchers, he downloaded language samples from the Internet and analyzed them in the comfort of his office.

Such interdisciplinary data pooling is becoming more common and easier than ever with access to the Internet, says Wil-liam Bainbridge, PhD, a program officer in the Social, Behavioral and Economic Sciences Directorate at the National Science Foundation (NSF). NSF supports the creation of these public-access databases.

CHILDES and another social science database, the General Social Survey (GSS), are the two most successful examples of the power of interdisciplinary data sharing, says Bainbridge. When disciplines pool their resources, their wealth increases exponentially because different researchers will analyze the same data in unique ways, expanding the overall knowledge base.

CHILDES

'Use of CHILDES has become increasingly multidisciplinary over the past four or five years,' says Carnegie Mellon University psychologist Brian MacWhinney, PhD, who, along with psychologist Catharine Snow, PhD, of Harvard University, directed CHILDES? creation. At first, mostly language development researchers, representing mainly psychologists and linguists, accessed the database, but users now include sociolinguists, discourse analysts and speech pathologists.

'I?ve been surprised at the degree to which it?s been useful to people in different disciplines,' says Snow.

For example, some researchers use the database to study the psychological construct of 'Theory of Mind,' which represents the developmental stage when children have awareness of mental states. CHILDES provides longitudinal data that can be searched for utterances that display evidence of Theory of Mind.

Indeed, CHILDES contains transcripts?and some audio and videotape?of children?s conversations collected from nearly a hundred major research projects in 20 languages. The data include utterances from a wide range of ages and situations, and each utterance is coded in several ways. One coding system marks conversational features, such as false starts, interruptions and errors, while others code grammar and phonetics. Researchers can search and manipulate the database using a tool kit of computer programs that perform linguistic analyses, including searches for specific words or strings of words, morphological analysis of sentences and phonetic analysis.

CHILDES makes working with human language data much easier, says University of Maryland psychologist William Hall, PhD. He had a large sample of language from 5-year-old children?which he?s donated to CHILDES?but wanted to learn more about first words. CHILDES allowed him to benefit from first-word data collected by other researchers.

Snyder and his colleague Karin Stromswold, PhD, of Rutgers University, used CHILDES to examine when certain linguistic constructions develop. In a recent study, they found that certain exotic constructions in English develop at the same time.

For example, once a child learns what?s called the verb-particle construction?we can say 'pick up the book' or, just as correctly, 'pick the book up' where 'pick' is separated from 'up'?he or she soon learns the double object dative construction?'John sent Mary a letter'? where there are two objects of the verb. Different children learn these at different ages, but in all children, they seem to be learned as a package, says Snyder.

'It would have taken a lot of time to collect a longitudinal data set large enough to do this analysis on my own,' says Snyder.

General Social Survey

The GSS is another excellent data source for social and behavioral researchers, says Bainbridge. The National Opinion Research Council (NORC) has conducted the GSS since 1972 to gauge American opinions on a range of social issues, including attitudes about capital punishment, mental health, and women and work. The 22nd GSS is under way and will be available on the Internet free along with all the data from previous surveys.

GSS is a nice example of how data collected by one discipline can be useful across disciplines, says Bainbridge. Although the survey is created and conducted mainly by sociologists, much of its data are of interest to psychologists, says GSS principal investigator David Smith, PhD. It has measures of several social psychology constructs, such as alienation and well-being. And, in 1996 and now again in 1998, the GSS includes a set of questions on the sociology of mental health, with questions on people?s attitudes about the mentally ill and opinions on whether they should receive government assistance.

The Internet interface is designed to provide easy and comprehensive access to the data, says Smith. Every survey variable is handled by the database as a separate unit. So a search on 'capital punishment' will provide links to all the data from every year of the survey that relates to the GSS question about people?s attitudes toward capital punishment as well as every bibliographic entry, project report and appendix reference that mentions that question.

Access to GSS on the Internet has dramatically increased its use by researchers, says Smith. And over the years, as more researchers become aware of the database, the composition of users has become more diverse. Twenty years ago, most GSS users were academic sociologists, according to user data collected by NORC. Now, about 60 percent of GSS users are academic sociologists and 40 percent are from other disciplines, including psychologists, political scientists and statisticians, says Smith.

Such large longitudinal data samples allow psychologists to move beyond traditional small-sample, low-statistical-power psychology studies, says psychologist Leonard Saxe, PhD, who uses GSS as well as other large databases.

'I try to use several different sources, including data from large databases to look at the bigger picture, comparing people?s perceptions and their actual behavior,' says Saxe, a psychology professor at the graduate school of the City University of New York and adjunct professor at Brandeis University. 'Large databases provide a nice antidote to the typical small-scale psychology study and let researchers look at the bigger picture, beyond the laboratory.'

Educational tools

Professors from many different disciplines are also using online databases such as CHILDES and the GSS to teach students research techniques and how to analyze data.

Although the GSS has always been used in the classroom, use has about doubled since it went online two years ago, says Smith. Professors can give students exercises based on GSS data, and the students have access not only to the data but to bibliographies of previous research.

Online data access will change the way professors teach, says Snow. CHILDES has for her. Before, if a student in her language development class asked at which age conditionals?'if' statements?develop, she would have answered with a brief description of the literature. Now, she can pull up CHILDES in the classroom and show her students how to find the answer themselves. 'It?s more than just a research tool,' says Snow. 'It makes child language come alive for students.'Y

You can access CHILDES at http://www.childes.psy.cmu.edu and GSS at http://www.icpsr.umich.edu/gss.