Self-Monitoring in Speaking: In Defense of a Comprehension-Based Account.

Ardi Roelofs

doi:10.5334/joc.61

Abstract

Speakers occasionally make speech errors, which may be detected and corrected. According to the comprehension-based account proposed by Levelt, Roelofs, and Meyer (1999) and Roelofs (2004), speakers detect errors by using their speech comprehension system for the monitoring of overt as well as inner speech. According to the production-based account of Nozari, Dell, and Schwartz (2011), speakers may use their comprehension system for external monitoring but error detection in internal monitoring is based on the amount of conflict within the speech production system, assessed by the anterior cingulate cortex (ACC). Here, I address three main arguments of Nozari et al. and Nozari and Novick (2017) against a comprehension-based account of internal monitoring, which concern cross-talk interference between inner and overt speech, a double dissociation between comprehension and self-monitoring ability in patients with aphasia, and a domain-general error-related negativity in the ACC that is allegedly independent of conscious awareness. I argue that none of the arguments are conclusive, and conclude that comprehension-based monitoring remains a viable account of self-monitoring in speaking.

Highlights

In verbal communication, speakers occasionally make slips of the tongue, which may be detected and corrected
I showed that the cross-talk problem is dealt with by the activation-verification account proposed by Levelt et al. (1999)
I made clear that, on the comprehension-based account proposed in Roelofs (2004), self-monitoring uses the speech comprehension system and involves a comparison process, which may be differently impaired

Summary

Introduction

Speakers occasionally make slips of the tongue, which may be detected and corrected. A related account has been advanced by Kröger et al, who proposed a biologically inspired model of comprehension-based internal monitoring In their model, about 500,000 spiking neurons implemented structured symbolic representations and condition-action production rules, which achieved semantic and phonological error detection in computer simulations of picture naming. According to the comprehension-based account, effects of uniqueness point and linear position should be present in the latencies of internal phoneme monitoring but not of picture naming This is what Özdemir et al observed, supporting comprehension-based monitoring of inner speech. To deal with cross-talk between speaking and listening, the comprehension-based monitoring model of Levelt et al (1999) and Roelofs (2004, 2005) assumes an activation-verification mechanism, which creates processing threads that allow speakers to distinguish between production, internal monitoring, and external monitoring streams (cf Salvucci & Taatgen, 2008, 2011). Crosstalk interference is taken care of by the activation-verification account of Levelt et al (1999)

A Double Dissociation in Aphasia

Summary and Conclusions