How time shapes letter position flexibility: Testing positional uncertainty and open bigram accounts

One of the critical benchmarks for understanding orthographic processing during word recognition and reading is the transposed-letter effect (e.g., in lexical decision, CHOLOCATE [created by transposing two letters from CHOCOLATE] produces slower and more error responses than CHOTONATE). Two main theoretical frameworks explain this phenomenon: positional uncertainty models, which attribute the effect to uncertainty in letter position encoding that diminishes over time, and open bigram models, which propose a level of ordered pairs of letters between the letter and word levels that may be more resilient to decay. We designed two delayed lexical decision experiments to test whether the transposed-letter effect vanishes or persists at two time delays (750 ms and 1,500 ms). In Experiment 1, a robust transposed-letter effect in accuracy emerged at 750 ms (9.6%) but diminished to a small (2.9%) yet reliable effect at 1,500 ms. Experiment 2 replicated this pattern with a contrast manipulation on the critical letters (e.g., CHOLOCATE vs CHOTONATE), yielding a slightly smaller transposed-letter effect (2.0%) at 1,500 ms. These findings demonstrate that positional uncertainty diminishes over time, yet residual orthographic overlap persists, particularly for a subset of participants, supporting hybrid accounts that combine bottom-up perceptual refinement with top-down contributions from shared sublexical codes (e.g., open bigrams).