Learning and Long-Term Retention of Large-Scale Artificial Languages

Frank, M. C., Tenebaum, J. M., Gibson, E. (2013). Learning and long-term retention of large-scale artificial languages. PLoS ONE 8(1): e52500. doi:10.1371/ journal.pone.0052500

Studying the way that a person learns an artificial language, a made up language never seen before, is a useful tool in helping researchers understand the important cues to natural language learning. A shortcoming with typical artificial language learning studies is that the languages have been quite different from natural languages – usually the number of words in the language is quite small, and each word occurs equally often. In order to “scale up” the artificial language used in the current experiment, the researchers adopted an artificial language with a 1,000-word vocabulary. Word frequency was also manipulated: Words occurred as few as 10 times, to as many as 8,000 times. Unique from the typical lab experiment, the participants had the artificial language downloaded on personal iPods so that their 10 hours of exposure could occur throughout their everyday activities, like during their daily commute or exercising. Importantly, the only cue to segment or learn words from the artificial language was the probability of syllable co-occurrences – syllables that belonged together within a word were more likely to occur together than syllables that spanned a word boundary. This cue exists in natural languages, and may help language learners learn word units over in addition to other cues such as pauses or stress patterns. 

Following 10 hours of listening to the large-scale artificial language, participants were tested on their ability to identify words from the language immediately after the 10 hours had been completed, 1-2 months after, or 3 years after. Participants were able to identify words from the language immediately after listening for 10 hours, and scored just as well 1-2 months after, with higher scores for high than low frequency words. At the end of 3 years, participants still were able to identify high frequency words from the artificial language. Although they did not show retention of low frequency words, this is an incredibly impressive finding as the words from the artificial language were nonsense, meaningless words. This study demonstrated that language learners could successfully segment words from an artificial language with a large vocabulary, and that the retention of newly learned words depended on word frequency. These processes might support the learning of second languages. For example, you might remember words from a second language you’ve studied in the past, especially the ones you heard most often. The results suggest, too, that listening to a new language for several hours might help you learn something about the words in that language.

Blogger: Nicolette Noonan, PhD student with Drs. Lisa Archibald and Marc Joanisse, and coordinator of the Canadian SLP blog.