Anchoring does not boost the benefit of contextual diversity for learning word forms or generalising learned meanings to new contexts

Jiayin Li, Louise Wong, Eusebia Catarina Almeida Rodrigues, Rachael C. Hulme, Holly Joseph, Fiona E. Kyle, Joanne S. H. Taylor

Research output: Working paperPreprint


Words that appear in many contexts/topics are recognised faster than those occurring in fewer contexts (Nation, 2017). However, contextual diversity benefits are less clear in word learning studies. Mak et al. (2021) proposed that diversity benefits might depend on anchoring a new word meaning before introducing diversity. In our study, adults (N = 288) learned meanings for eight pseudowords, four experienced in six topics (high diversity) and four in one topic (low diversity). All items were first experienced five times in one topic (anchoring phase), and results were compared to Norman et al. (2022) which used a similar paradigm without an anchoring phase. An old-new decision post-test (did you learn this word?) showed null effects of contextual diversity on word form recognition accuracy and response time, as in Norman et al. A cloze task involved choosing which pseudoword completed a sentence. For sentences situated in a previously experienced context, accuracy was significantly higher for pseudowords learned in the low diversity condition, whereas for sentences situated in a new context, accuracy was non-significantly higher for pseudowords learned in the high diversity condition. Anchoring modulated these effects. Low diversity item accuracy was unaffected by anchoring. However, for high diversity items, accuracy in familiar contexts was better in the current experiment (anchoring), but accuracy in new contexts was better in Norman et al. (non-anchoring). Inconsistent with Mak et al.’s proposal, these results suggest that anchoring facilitates meaning use in familiar contexts, but not generalisation to new contexts, nor word recognition in isolation.
Original languageEnglish
Publication statusPublished - 3 Jul 2023

Cite this