Levenshtein distance contributes to analyzing spelling as a measure of lexical quality

Lexical quality, which includes the precision of the orthographic information in a person’s stored lexical representations, is often assessed using spelling tests. Productions are typically scored in a binary fashion, as either correct or incorrect. Such scoring treats errors that are closer to the correct spelling and errors that are more distant from the correct spelling as equally wrong. Levenshtein distance, a computer-scored metric that reflects the number of additions, substitutions and deletions that are needed to transform an incorrect spelling into the correct spelling, provides a way of differentiating among misspellings. Using data from 15 previous studies with a total of 1,834 adult participants, we asked whether use of Levenshtein distance measures allows for a more sensitive measure of spelling ability than simple binary correctness. Our results showed that performance in tests of vocabulary and reading comprehension correlated significantly more highly with the average Levenshtein distance score across a set of spellings than with spelling correctness. For these skills, as well as for performance in a spelling recognition test, the average Levenshtein distance score across a participant’s spelling errors added significantly to the prediction of performance after correctness on the spelling production test was taken into account. This was not the case for reading speed, perhaps because precise orthographic knowledge is less important for rapid reading. The results suggest that nonbinary scoring provides useful information about individual differences in spelling ability.