The architecture and processes underlying visual word recognition represent some of the most intricate systems in human cognition. The seemingly simple act of reading a word involves not only a complex interplay between cognitive layers but also relationships between the word's spelling, phonology, and meaning. Over time, research has revealed that our mental systems rely on specific, consistent mappings between these properties to perform fast and accurate word recognition.
In particular, multiple studies have shown that orthographic-semantic (O-S) consistency—how similar in meaning a word is to its orthographically similar neighbors—plays a crucial role in how quickly we identify written words. Words with neighbors that share both similar spelling and meaning are recognized faster in lexical decision tasks, in which participants have to decide if a displayed string of letters is a real word or not as fast and as accurately as possible. However, until now, researchers haven't explored which specific types of orthographic neighbors drive this effect. Are neighbors formed by adding a letter (CAT to CATS), substituting a letter (CAT to BAT), or deleting a letter (SEAT to SAT) the most relevant?
A team led by Professor Yasushi Hino from Waseda University, Japan, including Debra Jared and Stephen J. Lupker from the University of Western Ontario, Canada, sought to answer this question. Their research, available online in Volume 143 of the Journal of Memory and Language on May 5, 2025, aimed to pinpoint which types of spelling relationships primarily contribute to the observed O-S consistency effect, offering deeper insights into the mechanisms of word recognition.
To unravel this puzzle, the researchers employed a multi-faceted approach. They first conducted extensive analyzes on lexical decision data taken from previous studies, aiming to replicate their results. To offer new insights, they calculated O-S consistency from large datasets of English words based on different types of orthographic neighbors, such as addition neighbors, substitution neighbors, and deletion neighbors. They leveraged an advanced vector space model (i.e., word2vec) to represent the meanings of words as multi-dimensional vectors, allowing for precise measurement of semantic similarity. To further validate their findings, they conducted a controlled online lexical decision experiment with human participants, carefully designing stimuli to isolate the effects of these different orthographic neighbor types.
Across both their analyses of existing data and their new experiment, a consistent pattern emerged: the O-S consistency effect was predominantly driven by addition neighbors, as well as broader "target-embedded" neighbors, which contain the entire target word's spelling pattern. In contrast, substitution and deletion neighbors showed little to no significant impact on lexical decision performance. This suggests that when we process a word, our brains are particularly sensitive to words that could be formed by adding letters to it.
Further investigation into the role of these target-embedded neighbors (including addition neighbors) revealed a fascinating dynamic tied to their morphological relatedness, i.e., whether they share a common root meaning. The researchers showed that the O-S consistency effect largely arises from two opposing forces. First, there's a processing facilitation for words that are morphologically-related addition neighbors (e.g., 'CREAMY' speeds up the recognition of 'CREAM'). Second, there's a processing inhibition for morphologically-unrelated addition neighbors ('SCREAM' slows down the recognition of 'CREAM'). "This type of process appears to be developed to read words more quickly and accurately," states Hino. "When there are unrelated words sharing similar spellings, such processing would be inhibited to prevent our mental systems from activating incorrect meanings."
Overall, these findings build upon a growing body of research demonstrating that our reading performance is profoundly shaped by the statistical relationships between orthography, phonology, and semantics that words possess. The results underscore once more the critical role of spelling-to-meaning consistency. Worth noting, understanding these fundamental mechanisms has broad implications for language acquisition and education. "With this newfound statistical knowledge, we may be able to offer new ways to read words more quickly and accurately," adds Hino. "It may also be possible to think of innovative strategies to teach words to children."
This study serves as a stepping stone towards a more comprehensive picture of how our brains work and the mechanisms they rely on to make reading more effective.
