Speech monitoring as action control

Final Report Summary - SMAC (Speech monitoring as action control)

Summary Report
The general theoretical objective of this project was to advance our knowledge of the processes involved in speech monitoring, and to test whether these are similar to the processes involved in the monitoring of other actions. This objective has been pursued through three different but related approaches: Part A aimed at exploring modulations of self-induced sound auditory suppression (SISAS) as a function of speech monitoring demands, examining in this way the potential use of internal modeling during speech production. Part B aimed at exploring the role of the cerebellum for the monitoring of speech production. Part C aimed at exploring to what extent domain general cognitive control and language production monitoring are achieved through the same processes. In what follows the progress within each of these parts will be concisely described.
Part A
The goal of this part of the project was to search for modulations of self-induced sound auditory suppression (SISAS) as a function of speech monitoring demands, examining in this way the potential use of internal modeling during speech production. Previous EEG studies have found modulations of participants’ auditory responsiveness (as indexed by an N100 component) as a function of their ability to predict their own auditory feedback when producing sounds. This has been taken to index the presence of internal modeling of our self-produced actions (e.g. Heinks-Maldonado et al., 2005). Furthermore, certain theories of language production have proposed that this kind of predictive internal modeling is used to avoid errors in speech production (e.g. Pickering & Garrod, 2013). If so, a situation known to put greater demands on speech production monitoring can be expected to modulate the auditory responsiveness to self-produced speech, and thus show up as a modulation of the N100. In the experiment we carried out to test this, 29 participants engaged in a speeded speech production task designed to elicit errors while recording their EEG. We included two types of conditions in our task, namely a situation priming for word errors, and another situation priming for pseudo-word errors. From the behavioral literature, we know that –all else being equal- non-lexical errors are easier to intercept that lexical errors. This lexical bias was also replicated in our own data where the priming related errors resulting in new words represented 5% of the data, while those resulting in pseudo-words only represented 1% of the data. We are currently carrying out response-locked analysis of the EEG data of the correct trials to examine whether this lexical bias, reflecting a difference in difficulty of speech production monitoring, will result in a modulation of participants auditory responsiveness to their own speech. This would be indexed by differences in the N100 elicited by the participant's own speech between trials that primed word versus psuedo-word errors. Such a result would be in favor of the hypothesis that the kind of internal modeling that is used for the supervision of more basic motor skills is also operative for the supervision of such a highly complex cognitive-motor skill as language production.

Part B

The goal of this part of the project was to test speech production monitoring performance after disrupting cerebellar functioning, examining in this way the potential use of internal modeling during speech production. Previous studies have established that the cerebellum is important to achieve motor control, and more concretely the cerebellum has been proposed as a center where the internal modeling of our actions is computed. In addition, several studies have shown that the cerebellum is involved in the processing of language, and this involvement seems to be especially confined to the right cerebellum. Here we wanted to link these findings by asking whether an impaired right cerebellum would result in impaired language production monitoring. To this end, we resorted to repetitive low frequency TMS, a technique which is known to disrupt the functioning of the stimulated area during a relatively extended period of time (during the whole task). The stimulation was neuro-anatomically guided, targeting Crus I and Crus II in the right cerebellum; these areas implicated in a wide range of motor control and language tasks. The same areas?? in the left cerebellum was stimulated in a separate experimental block within the same subjects, as a control condition. After stimulation, participants carried out a speeded speech production task designed to elicit both word (50%) and pseudo-word (50%) errors. We have collected data from 16 participants using this procedure and we are currently in the process of analyzing the data. Both the overall error rates as well as specific kinds of errors might be informative with respect to a potential role of the right cerebellum in speech production monitoring. Additionally, we will also analyze response times and voice onset times in search for differences as a function of laterality. Though less specific in nature, such effects would also indicate a causal involvement of a specific part of the cerebellum in the process of speech production.
Part C
This part aimed at answering the same broad question as parts A and B, namely whether mechanisms used for supervision of other domains of cognition may be engaged during language processing. The approach to this question was to disrupt domain general executive control through the ingestion of alcohol, and examine the impact of alcohol intake on language monitoring. Alcohol is known to disrupt domain general executive control functioning (e.g. Riderinkhof et al., 2003; Bombeke et al., 2013), so the question is whether it will also disrupt speech monitoring. Three groups of participants (alcohol, placebo and control, 54 participants in total) were tested on speech production tasks targeting speech monitoring at different levels (discourse level, morphosyntactic and lexico-phonological). As a control, we also included a task measuring non-linguistic executive control performance that is already known to be sensitive to alcohol consumption. We are currently analyzing the data searching for main effects of alcohol on all or some of the tasks, as well as correlations between the performance on the speech production and the executive control tasks. This study has the potential of shedding light on the relationship between the supervision of language production and that of other cognitive skills.

Final Report Summary - SMAC (Speech monitoring as action control)

Diese Seite teilen

Herunterladen