On the correct side of performance: Processing of internal and external signals in response speed evaluation

Highlights

• Electrophysiological correlates of cognitive processes sensitive to response speed

• Positive arousal modulates the correct-related negativity (CRN).

• Feedback processing is not restricted to the feedback-related negativity (FRN) but affects earlier ERPs like the P2a.

Abstract

Response appropriateness is not exclusively limited to accuracy. Nevertheless, the processing of parameters other than accuracy for response monitoring has been mostly neglected. The present experiment explored how the cognitive system processes response speed based on internal and external signals. Participants performed a response-choice task where correct responses were classified as fast, average, or slow. External signals informative about performance quality were presented after the response in most of the trials; in some trials, instead, participants had to judge their own performance. Event-related brain potentials (ERPs) of response and feedback processing were analysed to investigate how the cognitive system monitors correct responses. Response quality affected the processing of internal signals. That is, both the response-related negativity (correct-related negativity, CRN) and positivity (correct positivity, Pc) showed modulations related to speed: with the largest and the smallest amplitudes associated with fast and slow responses, respectively. We ascribe these modulations to positive arousal associated with the optimal nature of correct fast responses. Response quality, also affected feedback processing. Here, response speed significantly modulated the feedback-related negativity (FRN) and had an effect on the latency and the magnitude of the preceding positive peak. These effects in feedback processing seem related to feedback expectation in a context where awareness of feedback quality is vague and average performance, although more difficult to detected, is generally expected.

Introduction

Knowing whether an action was adequate to achieve a specific purpose is essential for learning. Performance monitoring is, therefore, a fundamental cognitive process for adaptive behaviour. Since the discovery of the error-related negativity (Ne/ERN) in the early 90s (Falkenstein et al., 1991, Gehring et al., 1993), research in performance monitoring has mostly focused on the brain processes related to error detection and elaboration. However, in everyday life, errors are less common than correct responses and the cognitive system should be sensitive for analysing parameters other than accuracy in order to achieve a comprehensive evaluation of performance quality. The present experiment aims to explore processes related to the evaluation of correct responses, with a specific focus on the monitoring of internal and external signals. To this mean, we used response speed as the additional criterion for appropriate behaviour and categorised correct responses as “very good” when fast, “good” when average, or “ok” when slow. Throughout the experiment, participants performed a choice-response task and informative feedback signalled response quality in most of the trials; in other trials, external signals were not displayed and participants had to subjectively judge their performance. Brain processes related to the evaluation of internal and external signals for speed monitoring were investigated by means of electrophysiological recording. The processing of self-generated internal signals was explored by assessing modulations of response-related potentials; potentials evoked by feedback processing were analysed to gain insight into the processing of external signals.

The electrophysiological investigation of performance monitoring has mostly focused on activity occurring after a response as an index of internal processes for performance monitoring. Internal monitoring has its clearest electrophysiological manifestation in a large medial fronto-central negative potential generally elicited by errors immediately after overt responses (0–100 ms). This potential was called error negativity (Ne; Falkenstein et al., 1991) or error-related negativity (ERN; Gehring et al., 1993) and was originally considered to be an index of error detection. However, the presence of a smaller negative potential, similar to the Ne/ERN in temporal and topographical properties, elicited by correct responses – the so-called correct-related negativity (CRN) – challenged this first interpretation (Vidal et al., 2000). Vidal et al. (2003) showed that the CRN is not associated with any error or residual stimulus processing (Coles et al., 2001) and suggested, instead, that the response-locked negativities (Ne/ERN and CRN) describe processes not exclusively related to error detection (see Allain et al., 2004). Previous research on the electrophysiological signal associated with the evaluation of response speed offered two different interpretations of CRN activity. Luu et al. (2000) proposed that this response-related negative potential reflects the affective reaction elicited by the mismatch between the actual response and an action plan. They observed a larger CRN for slow compared to fast correct responses and this result was interpreted as the emotional reaction to the mismatch between the planned optimal response and the actual correct but slow response. On the other hand, two studies (Heldmann et al., 2008, Stahl, 2010) used an adaptation of Luu et al. (2000) design but with feedback displayed after the response and concluded that CRN modulations describe flexible monitoring shifts from internal to external signals. In these two studies, the CRN was robust when response speed was clearly discernible but no CRN was observed when the cognitive system was unable to correctly evaluate small speed differences (Heldmann et al., 2008, Stahl, 2010). When internal signals were not sufficient to correctly evaluate performance, the negativity was elicited by the feedback (Stahl, 2010). However, CRN enhancements were also observed when internal signals fail an adequate performance evaluation, because of low input discriminability (Pailing and Segalowitz, 2004), increases in perceptual difficulty (Endrass et al., 2012), larger response-set size (Maier et al., 2010), or response conflict (Bartholow et al., 2005, Eppinger et al., 2007).

The response-related negativity is not the only potential evoked by the processing of internal signals for performance monitoring. In addition to the Ne/ERN errors elicit two positive deflections (Arbel and Donchin, 2009): an early fronto-central positivity peaking around 200 ms after response (early-Pe) and a late centro-parietal positivity, peaking 200–400 ms after response (late-Pe). The early- and the late-Pe have been associated with error awareness (Murphy et al., 2012, Nieuwenhuis et al., 2001), evidence accumulation (Steinhauser and Yeung, 2012), additional stimulus processing (Shalgi et al., 2009) and confidence evaluation (Boldt and Yeung, 2015). However, it is still unclear whether these potentials are only linked to errors or apply to the processing of correct responses as well. In a recent investigation, Boldt and Yeung (2015) reported a clear association between the response-related positivity for correct responses (Pc) and response confidence, with larger centro-parietal positivity when participants were unsure about response correctness.

To summarise, the meaning of the cognitive processes inducing the CRN is still unclear whereas the association between later stages of internal signal processing, indexed by the response-locked positivities, and the processing of correct responses has been mostly neglected.

The described response-related negativities and positivities index the monitoring of internal signals for performance monitoring, but, in many situations, such monitoring is based on external signals. Internal and external monitoring seem to be linked and the potentials elicited by these two types of processing show similarities (Koban and Pourtois, 2014, Ullsperger et al., 2014). The most prominent ERP component related to the processing of external signals for performance monitoring is the feedback-related negativity (FRN, Holroyd and Coles, 2002, Miltner et al., 1997). The FRN describes a negative potential over medial fronto-central electrodes peaking approximately 250 ms after feedback onset. Stahl (2010) found FRN modulations when correct evaluation of response quality required monitoring response speed and the cognitive system failed in processing this parameter with the necessary details. The researcher observed that the FRN was elicited mostly by responses given close to a deadline, but the FRN was absent when responses were clearly faster or slower than the deadline (see also Heldmann et al., 2008). These opposite trends of CRN and FRN modulations were interpreted as shifts in monitoring focus from internal to external processing, with the consequent interpretation that response-related and feedback-related negativities describe attentional processes. However, the reported dynamics in response-related and feedback-related activity might describe the emotional effect suggested by Luu et al. (2000), with emotional reactions elicited by the mismatch between ideal and actual response delayed to feedback processing when internal signals are not sufficient to evaluate response quality. In fact, FRN modulations were reported in tasks in which participants do not have any outcome control (Holroyd et al., 2003). For example, in gambling tasks, where an algorithm selects the outcomes based on preset probabilities, the FRN is generally larger for unexpected and unfavourable outcomes, pointing to a potential emotional connotation of the processes described by the FRN.

Although studies on the monitoring of external signals have mostly focused on the FRN, some results point to feedback processing occurring before the FRN. The positive deflection preceding the FRN, the so-called P2a, is affected by outcome expectancy and valence, with a larger positivity for unexpected than expected outcomes and for gains than losses (Potts et al., 2006). In addition, reward magnitude affects the P2a. This component is enhanced when the expected outcome is large, irrespective of its valence, that is, large positive and negative outcomes result in similar P2a enhancements (Goyer et al., 2008). This early modulation is also evoked when the feedback is informative about the performance of another participant (Bellebaum et al., 2010). It follows that feedback processing is not only restricted to processes reflected by the FRN but also takes place at earlier stages of feedback processing.

To summarise, previous studies reported an association between CRN and FRN but further research is required to explore which cognitive process links these two potentials. In addition, potentials evoked before the FRN seem also important for a comprehensive understanding of the processes involved in the analysis of external signals for performance monitoring.

The focus of the present experiment is how internal and external signals are processed to monitor response speed. We investigated the electrophysiological signals associated with internal and external monitoring to shed clear light, and draw some conclusions, about the evaluation of correct responses and the brain processes involved in response monitoring. Participants had to respond as fast and accurate as possible to a target letter in 3 × 3 letters arrays that incorporated four different degrees of response conflict. Response conflict, in the form of Simon conflict (Simon, 1969) and Eriksen flanker conflict (Eriksen and Eriksen, 1974), was employed in the present design to increase the variability in response times (RTs) and make the discrimination of speed differences easier; however, response conflict was not the main focus of the present investigation. Three categories of correct responses were obtained based on two deadlines: fast, average, and slow. In most of the trials, an informative external signal was displayed after each response but, in some trials, this feedback signal was not presented and participants had to judge their performance. These judgements were obtained in order to control for the accuracy of internal monitoring. Response speed was selected as the discriminative factor for performance quality because of its robust effect on CRN and FRN amplitudes and its strong influence on performance monitoring dynamics.

We expect modulations of the CRN and the FRN as previously reported in speed evaluation (Heldmann et al., 2008, Luu et al., 2000, Stahl, 2010). A trade-off between CRN and FRN amplitude would support an interpretation of dynamic shifts from internal to external signals imposed by the low sensitivity of the monitoring system to small response speed differences. In contrast, parametric modulations of the CRN, not linked to any corresponding FRN modulation, would suggest independent processing of internal and external signals. CRN modulations could be related to response speed discriminability, negative valance, or arousal. RTs in response-choice tasks follow a Gaussian distribution, slightly skewed for slow responses (Spieler et al., 2000). Consequently, according to speed discriminability, fast and slow responses should be associated with large CRNs, because of the higher discriminability due to wider-spread RTs along the tails of the distribution. A smaller CRN should be instead observed for average responses because of the narrower time-window in this condition, and consequently the harder response quality evaluation. Harder evaluation of internal signals for average compared to fast or slow responses should also affect the judgements about response quality. On the other hand, largest CRN for slow responses and smallest CRN for fast responses might index an emotional connotation of the response-related negativity. Slow responses are sub-optimal responses and could, therefore, be associated with negative feelings, like the ones elicited by errors. Largest CRN for fast responses and smallest CRN for slow responses might instead suggest emotional arousal: The optimal nature of fast correct responses might generate positive arousal, as errors generate negative arousal.

Previous research has shown that the CRN is also sensitive to task difficulty (Endrass et al., 2012, Pailing and Segalowitz, 2004) and response conflict (Bartholow et al., 2005). Although task difficulty was not directly manipulated in the present experiment, the use of stimuli with different levels of response conflict allowed controlling for the influence of response conflict on internal monitoring. Responses made to stimuli incorporating a conflict are generally slow; therefore, trials in the three speed categories might differ not only in response quality but also in response conflict. Previous research has shown that the CRN is enhanced for stimuli incorporating response conflict due to flanker incongruency (Bartholow et al., 2005) and this might confound the present investigation of speed monitoring based on internal signals. Therefore, the CRN amplitude was further analysed according to response conflict; for this purpose, a subset of trials was selected in order to equalise response speed across the four conflict conditions.

No hypothesis can be formulated on the Pc and P2a components because of the absence of extensive previous research. The analysis of these components is, therefore, mostly exploratory. Nevertheless, regarding the judgement about response quality, we expect a direct link between these behavioural data and the Pc. One previous study (Boldt and Yeung, 2015) showed that response certainty modulated the response-locked positive potential for incorrect and correct responses; therefore, we expect the smallest positivity for the category with the highest judgement accuracy.