ABSTRACT
The ability to effectively communicate memories through language is uniquely human. It is also critical to our social lives, affecting how couples jointly remember their lives together, sports fans remember a season, or how jurors determine the credibility of witnesses during trials. This project help us understand the remarkable ability to share memories along with the information that those memories may, or may not, be reliable descriptions of the past. Self-reflection about the accuracy of your own memory is referred to as metamemory. Although cognitive scientists have studied this extensively, little is known about how people use language to communicate the quality of their memories or how others use this language to decide whether those memories are reliable. Because people are not only sensitive to what others say, but also how they say it, this project specifically examines how evaluating someone else?s memory depends upon characteristics of their speech beyond the words they are saying. Finally, because individuals often encounter others with accents different from their own, the project examines how accent differences may limit or bias the ability of individuals to determine if the memory reports of others are accurate or not.
To effectively share memories for collective action or decision-making, witnesses must convey not only the content of memories, but also signal their reliability. In turn, evaluators or judges of these memory justifications should rely upon them only to the extent they are deemed accurate. The current project examines how the content and manner of speaking operate during such communication. More specifically, it tests the hypothesis that judges process both content and prosody information when evaluating witnesses? memories, that they may do so relatively independently, and that the ability to effectively communicate and evaluate these information streams may be differentially impacted by important individual differences in recognition memory ability, self-evaluation of one?s own memories (metamemory), Verbal IQ, and working memory capacity (WMC). To objectively score the content communicated by witnesses, the project uses machine learning algorithms trained to discriminate the natural language accompanying accurate versus inaccurate memory justifications comparing their performance to human evaluators of the same statements. It also trains these machine learners to determine the words and phrases that human evaluators find particularly important and manipulates the format of communication, contrasting transcribed memory justifications (text only), the original audio recordings, and modifications of those recordings that greatly reduce either the content or the prosodic information that is available. This reflects the first memory research to examine how judges integrate these two streams of information when evaluating others? memory. Finally, this approach is extended to social metamemory communication between speakers with different language backgrounds by comparing native English speakers? ability to effectively evaluate the memory justifications of people who speak English as a first versus a second language. Because most English speakers world-wide have acquired English as a second language, this extension of the research has significant societal importance.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
Click here to read more.
