Disassembling an experimental artifact in aphasia: Why phonemic errors with words and semantic errors with numbers?

There is broad consensus as to the significance of speech errors in aphasia. The analysis of errors is understood to provide clear clues for clinical diagnosis, the identification of those cognitive-linguistic processes affected, and the corresponding impaired cerebral structures. However, Stimulus Type Effect on Phonological and Semantic errors (STEPS), a phenomenon in which a person with aphasia produces more phonological errors with words (e.g., "tamle" for "table") but more semantic errors with number words (e.g., "thirteen" for "forty-two"), casts doubt on this consensus view. In this paper two studies are described, in which we explore whether STEPS is in fact a result of the lack of rigorous control over the materials compared (words versus numbers) and the evaluation conditions. Two persons, one with a reproduction conduction aphasia and the other with a repetition conduction aphasia, participated in the studies. Study 1 explored the role of memory load in the emergence of STEPS by eliciting the repetition of pairs of semantically-unrelated words. In Studies 2a and 2b, our participants were asked to produce sequences of high- and low-frequency words from one semantic category (colors), and this was compared to the performance in multi-digit number production tasks. The results showed that sequences of high-frequency colors, like multi-digit numbers, were produced mainly with semantic errors, whereas sequences of low-frequency colors showed a mixed pattern with many phonemic and semantic errors. It seems that the production of semantic errors and the absence of phonemic errors in multi-digit numbers that give rise to STEPS is an experimental artifact caused by the combination of several factors: the use of semantically-related high-frequency words, produced cyclically under high-memory-demand conditions. These findings contribute substantially to the current discussion of language production models and allow for a deeper understanding of the neurocognitive processes that underly speech errors in aphasia.