News

This paper has just come out in the Quarterly Journal of Experimental Psychology.

The research was led by Rebecca Norman, a PhD student in the lab who is co-supervised by Jo Taylor and Jenni Rodd.

Link to paper

From mid-childhood onwards, most new words are learned through reading. The precise meaning of many words depends upon the linguistic context in which they are encountered, which readers use to infer the appropriate interpretation. However, it is unclear what features of these linguistic contexts best support learning of new word meanings. We investigated whether learning words in contextually diverse sentences benefits word form and meaning learning in adults (n = 239). Participants learned meanings for 8 pseudowords through reading 10 sentences about each. Four pseudowords were learned in a diverse condition (10 sentences on different topics) and four were learned in a non-diverse condition (10 sentences on the same topic). An old-new decision post-test indicated that diversity did not influence word form learning. In a second post-test, participants chose which trained pseudoword completed a sentence from either an unfamiliar, untrained context, or a familiar, trained context. For familiar contexts, accuracy was higher for pseudowords learned in the non-diverse condition, but for unfamiliar contexts, accuracy was higher for pseudowords learned in the diverse condition. These results suggest that diverse contexts may promote development of flexible, decontextualised meaning representations that are easier to generalise to new contexts. Conversely, non-diverse contexts may favour extraction of context-bound representations that are more easily used in the same context.

Lena Blott and Jennifer Rodd have published a new paper together with undergraduate researcher Oliver Hartopp and Professor Kate Nation from the University of Oxford.

The paper can be accessed here.

Previous work from the Word Lab has demonstrated that we can prime people such that relatively infrequent meanings of ambiguous words (e.g. the “animal enclosure” meaning of “pen”) become temporarily more readily available. This has been termed “word-meaning priming” (see e.g. Betts et al., 2018; Gilbert et al., 2018; Rodd et al., 2013; Rodd et al., 2016).

For the present experiment, we designed our priming stimuli to be short narratives in which cues to disambiguation were relatively weak, and distant from the ambiguous word itself. We replicated the previously observed word-meaning priming effect with these stimuli, which are, arguably, more naturalistic than the types of sentences typically used in psycholinguistics studies. We hope that this experiment can be a step towards using more naturalistic and varied forms of disambiguation to study how comprehenders can flexibly update their lexical knowledge to aid comprehension.

The study was preregistered. Data and code are available here.

Lena Blott and Jennifer Rodd have published a new preprint together with undergraduate researcher Oliver Hartopp and Professor Kate Nation from the University of Oxford.

The preprint can be accessed here.

Previous work from the Word Lab has demonstrated that we can prime people such that relatively infrequent meanings of ambiguous words (e.g. the “animal enclosure” meaning of “pen”) become temporarily more readily available. This has been termed “word-meaning priming” (see e.g. Betts et al., 2018; Gilbert et al., 2018; Rodd et al., 2013; Rodd et al., 2016).

For the present experiment, we designed our priming stimuli to be short narratives in which cues to disambiguation were relatively weak, and distant from the ambiguous word itself. We replicated the previously observed word-meaning priming effect with these stimuli, which are, arguably, more naturalistic than the types of sentences typically used in psycholinguistics studies. We hope that this experiment can be a step towards using more naturalistic and varied forms of disambiguation to study how comprehenders can flexibly update their lexical knowledge to aid comprehension.

The study was preregistered. Data and code are available here.