Limits...
Detecting and correcting partial errors: Evidence for efficient control without conscious access.

Rochet N, Spieser L, Casini L, Hasbroucq T, Burle B - Cogn Affect Behav Neurosci (2014)

Bottom Line: Two parameters of the partial errors were found to predict detection: the surface of the incorrect EMG burst (larger for detected) and the correction time (between the incorrect and correct EMG onsets; longer for detected).The correct(ive) responses associated with detected partial errors were larger than the "pure-correct" ones, and this increase was likely a consequence, rather than a cause, of the detection.The respective impacts of the two parameters predicting detection (incorrect surface and correction time), along with the underlying physiological processes subtending partial-error detection, are discussed.

View Article: PubMed Central - PubMed

Affiliation: Laboratoire de Neurosciences Cognitives, UMR 7291, Fédération de Recherche 3C, Aix-Marseille Université and CNRS, Case C, 3, Place Victor Hugo, 13331, Marseille, France.

ABSTRACT
Appropriate reactions to erroneous actions are essential to keeping behavior adaptive. Erring, however, is not an all-or-none process: electromyographic (EMG) recordings of the responding muscles have revealed that covert incorrect response activations (termed "partial errors") occur on a proportion of overtly correct trials. The occurrence of such "partial errors" shows that incorrect response activations could be corrected online, before turning into overt errors. In the present study, we showed that, unlike overt errors, such "partial errors" are poorly consciously detected by participants, who could report only one third of their partial errors. Two parameters of the partial errors were found to predict detection: the surface of the incorrect EMG burst (larger for detected) and the correction time (between the incorrect and correct EMG onsets; longer for detected). These two parameters provided independent information. The correct(ive) responses associated with detected partial errors were larger than the "pure-correct" ones, and this increase was likely a consequence, rather than a cause, of the detection. The respective impacts of the two parameters predicting detection (incorrect surface and correction time), along with the underlying physiological processes subtending partial-error detection, are discussed.

Show MeSH
a Scatterplot of IncSurf as a function of CT for all participants (after z-score computation). The overlap between undetected and detected partial errors is large, and it is clear that neither of the two parameters in itself allows for a clear prediction of partial-error detection. It can be noted, however, that above a virtual decreasing diagonal, most of the points belong to the detected class, confirming that the combination of the two parameters is necessary for classifying the trials. b Receiver operating characteristic curves for EMG surface (solid line) and CT (dashed line). The area under the curve (AUC) is larger for EMG surface than for CT
© Copyright Policy - OpenAccess
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4125819&req=5

Fig3: a Scatterplot of IncSurf as a function of CT for all participants (after z-score computation). The overlap between undetected and detected partial errors is large, and it is clear that neither of the two parameters in itself allows for a clear prediction of partial-error detection. It can be noted, however, that above a virtual decreasing diagonal, most of the points belong to the detected class, confirming that the combination of the two parameters is necessary for classifying the trials. b Receiver operating characteristic curves for EMG surface (solid line) and CT (dashed line). The area under the curve (AUC) is larger for EMG surface than for CT

Mentions: Although the two selected parameters differed significantly between detected and undetected partial errors, this difference was true on average, but the dissociation was far less clear at the individual-trial level. Indeed, Fig. 2 (panels c and d) shows the cumulative density functions (CDFs) of the EMG surface (panel c) for undetected and detected partial errors, along with the CDF for the overt-error surface, as an inset. Panel d shows the CDFs of CT for undetected and detected partial errors. As can clearly be seen, both parameters have large overlaps in their distributions of detected and undetected partial errors. The overlap is also evident on Fig. 3a, which presents the scatterplot of the EMG surface as a function of CT for all of the participants (z-score transform) for detected and undetected partial errors. To better characterize the predictive values of these two parameters, we first computed the receiver operating characteristic (ROC) curves and the corresponding areas under the curve (AUCs) for both parameters using the pROC library (Robin et al., 2011) under the R environment (R Development Core Team, 2012), on the pooled data across participants (Fig. 3b). Second, we computed the corresponding regression coefficients on the z-score values for the EMG surface and CT. The two AUCs were significantly different from 0.5 [as assessed by one-tailed t tests: 0.863 for EMG surface, t(15) = 15.0, p < .001, and 0.742 for CT, t(15) = 7.7, p < .001], indicating that the two parameters have clear discriminative power. However, this discriminative power also differs between the two parameters.Fig. 3


Detecting and correcting partial errors: Evidence for efficient control without conscious access.

Rochet N, Spieser L, Casini L, Hasbroucq T, Burle B - Cogn Affect Behav Neurosci (2014)

a Scatterplot of IncSurf as a function of CT for all participants (after z-score computation). The overlap between undetected and detected partial errors is large, and it is clear that neither of the two parameters in itself allows for a clear prediction of partial-error detection. It can be noted, however, that above a virtual decreasing diagonal, most of the points belong to the detected class, confirming that the combination of the two parameters is necessary for classifying the trials. b Receiver operating characteristic curves for EMG surface (solid line) and CT (dashed line). The area under the curve (AUC) is larger for EMG surface than for CT
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4125819&req=5

Fig3: a Scatterplot of IncSurf as a function of CT for all participants (after z-score computation). The overlap between undetected and detected partial errors is large, and it is clear that neither of the two parameters in itself allows for a clear prediction of partial-error detection. It can be noted, however, that above a virtual decreasing diagonal, most of the points belong to the detected class, confirming that the combination of the two parameters is necessary for classifying the trials. b Receiver operating characteristic curves for EMG surface (solid line) and CT (dashed line). The area under the curve (AUC) is larger for EMG surface than for CT
Mentions: Although the two selected parameters differed significantly between detected and undetected partial errors, this difference was true on average, but the dissociation was far less clear at the individual-trial level. Indeed, Fig. 2 (panels c and d) shows the cumulative density functions (CDFs) of the EMG surface (panel c) for undetected and detected partial errors, along with the CDF for the overt-error surface, as an inset. Panel d shows the CDFs of CT for undetected and detected partial errors. As can clearly be seen, both parameters have large overlaps in their distributions of detected and undetected partial errors. The overlap is also evident on Fig. 3a, which presents the scatterplot of the EMG surface as a function of CT for all of the participants (z-score transform) for detected and undetected partial errors. To better characterize the predictive values of these two parameters, we first computed the receiver operating characteristic (ROC) curves and the corresponding areas under the curve (AUCs) for both parameters using the pROC library (Robin et al., 2011) under the R environment (R Development Core Team, 2012), on the pooled data across participants (Fig. 3b). Second, we computed the corresponding regression coefficients on the z-score values for the EMG surface and CT. The two AUCs were significantly different from 0.5 [as assessed by one-tailed t tests: 0.863 for EMG surface, t(15) = 15.0, p < .001, and 0.742 for CT, t(15) = 7.7, p < .001], indicating that the two parameters have clear discriminative power. However, this discriminative power also differs between the two parameters.Fig. 3

Bottom Line: Two parameters of the partial errors were found to predict detection: the surface of the incorrect EMG burst (larger for detected) and the correction time (between the incorrect and correct EMG onsets; longer for detected).The correct(ive) responses associated with detected partial errors were larger than the "pure-correct" ones, and this increase was likely a consequence, rather than a cause, of the detection.The respective impacts of the two parameters predicting detection (incorrect surface and correction time), along with the underlying physiological processes subtending partial-error detection, are discussed.

View Article: PubMed Central - PubMed

Affiliation: Laboratoire de Neurosciences Cognitives, UMR 7291, Fédération de Recherche 3C, Aix-Marseille Université and CNRS, Case C, 3, Place Victor Hugo, 13331, Marseille, France.

ABSTRACT
Appropriate reactions to erroneous actions are essential to keeping behavior adaptive. Erring, however, is not an all-or-none process: electromyographic (EMG) recordings of the responding muscles have revealed that covert incorrect response activations (termed "partial errors") occur on a proportion of overtly correct trials. The occurrence of such "partial errors" shows that incorrect response activations could be corrected online, before turning into overt errors. In the present study, we showed that, unlike overt errors, such "partial errors" are poorly consciously detected by participants, who could report only one third of their partial errors. Two parameters of the partial errors were found to predict detection: the surface of the incorrect EMG burst (larger for detected) and the correction time (between the incorrect and correct EMG onsets; longer for detected). These two parameters provided independent information. The correct(ive) responses associated with detected partial errors were larger than the "pure-correct" ones, and this increase was likely a consequence, rather than a cause, of the detection. The respective impacts of the two parameters predicting detection (incorrect surface and correction time), along with the underlying physiological processes subtending partial-error detection, are discussed.

Show MeSH