Disentangling the Effect of Letter Frequency, Bigram Frequency and Inter-item Similarity in Recognition Memory: A Global Matching Model Approach
Lyulei Zhang, Adam F Osth · 2025
Previous studies have reported a memory advantage for words with rarer letters; however, the specific sources of this feature frequency effect remain unclear. The current study aimed to disentangle the relative contributions of letter frequency, bigram frequency, and inter-item similarity to recognition memory performance through both empirical experimentation and computational modelling. Across three recognition memory experiments, we manipulated orthographic feature frequency at the level of letters (Experiment 1), bigrams (Experiment 2), and their factorial combination (Experiments 3). While both letter and bigram frequency produced mirror effects, such that words composed of rarer features showed higher hit rates and lower false alarm rates than words composed of more common features, we found bigrams as the primary driver of the observed feature frequency effects. Moreover, the relative roles of feature frequency and inter-item similarity were characterized through formal modelling. Results were fit with global matching models with words’ orthographic representations where the underlying features were either equally weighted or with weights proportional to their frequency in language. Models with differential feature weights outperformed the equal weight models, specifically in capturing target performance. Moreover, models incorporating bigram frequency, rather than letter frequency captured the feature frequency effect on hit rates in Experiment 3. The necessity of differential feature weighting was further validated through fitting with a recognition memory mega-study where differential feature weighting improved the model’s ability to capture variability in hit rates across individual words.