Lexical Bias in Phonological Speech Errors: A Related Beply to Baars et al. (1975)

Martin Corley[*], Rob Hartsuiker[*] and Heike Martensen[+]
[*] University of Edinburgh
[+] Universiteit Antwerpen

Phonological speech errors tend to result in real words more often than chance would predict. There are two competing accounts of this 'lexical bias' effect: it is considered to be the result either of phoneme-to-wordform feedback in word production, or of pre-articulatory editing for lexical status. Support for the latter account comes from a study showing that the effect is modified by context (Baars et al., 1975); in a task designed to elicit exchange-type phonological errors, participants appeared to suppress nonword errors if real words were present in the context, but not otherwise. However, as the task relies heavily on (silent) reading, it remains possible that a third, perception-based, account of the findings can be offered.

In the present paper, we report two experiments designed to investigate the editing account in more detail. Experiment one used speeded naming of nonword pairs (such as "pood gath") in an attempt to elicit spoonerisms (responding "good path"). These target items were all nonwords and were designed such that exchange errors would either result in words ("pood gath") or not ("pooth gad"). Each critical pair was preceded by several filler items, which were not read aloud.
The fillers included pairs with onsets designed to elicit spoonerisms ("g... p...") as well as unrelated items. The filler lists were either all nonwords (nonword context), or a mixture of words and nonwords (mixed context).

In the mixed condition, nonword pairs with lexical outcomes ("pood gath") were spoonerized more often than nonword pairs with nonlexical outcomes ("pooth gad"). This lexical bias effect disappeared in the pure nonword condition, providing further evidence against the feedback account. However, inconsistent with the editing account, we
observed that lexical context increased the number of word outcomes, rather than decreasing the number of nonword outcomes. These findings are consistent with a perceptual account (such as one in which the lexical route plays a lesser perceptual role in a nonword context).

Experiment two was designed to investigate this alternative explanation, and consisted of a replication of experiment one with the naming task replaced by a speeded verification task. Early results suggest that it is indeed highly possible to make perceptual errors when reading target items. In the light of accumulating evidence, we suggest that error elicitation tasks such as that used by Baars et al. (1975) should not be considered as purely production-based tasks.