Temporal aspects of self-monitoring for speech errors

Is repairing speech errors an automatic or a controlled process? Insights from the relationship between error and repair probabilities in English and Spanish

“…It is worth noting that even though the constraints on repairing lexical and segmental errors may be different because of the different nature of representations and the different time-constraints of lexical selection and phonological encoding processes (Nooteboom, 2005); the current findings in lexical repairs have support in phonological repairs as well. Consonant errors are more likely in the onset position than in any other position in the word (e.g., Nooteboom & Quené, 2015;Shattuck-Hufnagel, 1992) and onset position also happens to have higher repair rates, as well as shorter cutoff-to-repair times, than other positions (Nooteboom & Quené, 2019). These findings mirror those reported in this experiment for lexical errors, and point to more efficient repair processes where the error probability is the highest in both lexical and segmental domains.…”

“…This kind of process can account well for findings such as conduite d'approche discussed in the earlier sections, and why speakers may inadvertently switch a correct response to a wrong one, as repairs are done fairly automatically, as long as alternatives with high activation are available (e.g., Nozari, 2019). This purely automatic account, however, cannot by itself explain the relationship observed between the error rates and proportion of corrected errors found here, and also reported in Levelt (1983) and Nooteboom and Quené (2019). The most straightforward prediction is that repair attempts in the purely automatic account should remain fairly insensitive to error probabilities.…”

“…The answer is that although the overall production process has a goal, one could envision a repair process that is partially unintentional. An example of such a process would be simply selecting the next most highly activated item once the produced item is deemed to be an error (see Nooteboom & Quené, 2019, for a similar view). In such a case, the repair process did not set and follow a new goal.…”

“…Repairing speech errors certainly meets the criteria of fast and efficient processing: errors can be repaired very quickly with very little temporal gap between the point of error interruption and the start of the repair (sometimes as short as 0 ms; Blackmer & Mitton, 1991;Hartsuiker & Kolk, 2001;Nooteboom & Quené, 2019). While not incompatible with a very efficient controlled process, fast operations are often taken to indicate some degree of automaticity (Fodor, 1983;Nozari, 2018).…”

“…The second class views the repair process as a form of revising the previous plan (e.g., Boland, Hartsuiker, Pickering, & Postma, 2005). A version of this account was introduced in the Introduction: detection of an error triggers a repair process which simply entails selecting the most highly activated item that was not produced (see also Nooteboom & Quené, 2019). Note that this process does not require resetting or recommitting to a goal, as it picks up production from the middle of the process.…”