The linguistic constraints of precision of verbal working memory.

Working memory (WM) precision has received little interest in the verbal WM domain, contrary to the visual WM domain. The aim of this study was to assess the precision with which words can be maintained in verbal WM. A probe-recognition task was used, in which the amount of phonological overlap between target and probe items was varied (25-75%). In Experiment 1, we assessed WM precision in monolingual, French-speaking young adults. In Experiment 2, we assessed WM precision in multilingual speakers with variable levels of language proficiency in German (L1), French (L2), and English (L3) in order to determine the extent to which WM precision is determined by the quality of underlying language representations. In Experiment 1, we observed that WM precision at the single-phoneme level was limited, with recognition performance decreasing considerably for the 75% overlap trials in which the probe and the target differed by a single phoneme. At the same time, recognition performance stayed above chance. In Experiment 2, we showed in multilingual speakers that discrimination accuracy was further reduced when tested in a second language. This study shows that phonological WM precision at the single-phoneme level is limited and furthermore depends upon the precision of the lexical representations in the linguistic system. Overall, this study provides support for theoretical models considering that verbal WM is grounded in the language architecture.