Elsevier

Cognition

Volume 82, Issue 2, December 2001, Pages 157-178
Cognition

Error monitoring in the hemispheres: the effect of lateralized feedback on lexical decision

https://doi.org/10.1016/S0010-0277(01)00150-0Get rights and content

Abstract

Does each hemisphere have its own system for monitoring and responding to errors? Three experiments investigate the effect of presenting lateralized accuracy feedback in a bilateral lexical decision task. We presented feedback after each trial in either the left visual field (LVF) or right visual field (RVF). In Experiment 1 the feedback stimuli were faces smiling or frowning, in Experiment 2 we used colored squares, and Experiment 3 tested the effect of verbal feedback. Negative feedback presented in the LVF tended to improve performance on the following trial, while the same negative feedback in the RVF tended to disrupt performance on the following trial. This result was strongest with the faces as feedback, was less pronounced with colored squares, and disappeared with verbal feedback. The results are interpreted as suggesting a right hemisphere superiority for error monitoring that depends on the mode of feedback.

Introduction

As humans, we are prone to making all sorts of errors – slips of the tongue, miscalculations, typographic mistakes, or fumbled footballs. The ability to monitor and compensate for these errors is an essential part of our fallible cognitive systems. However, relatively little is known about the neural mechanisms of error monitoring. Even less is known about how the two hemispheres of the brain differ in their ability to detect and respond to errors.

In 1977, Rabbitt and Rodgers posed the question “What does a man do after he makes an error?” (Rabbitt & Rodgers, 1977). One generality that comes out of Rabbitt's work on serial choice reaction time tasks is that people tend to slow down after they've made an error (Rabbitt, 1966a, Rabbitt, 1966b, Rabbitt and Phillips, 1967, Rabbitt and Rodgers, 1977, Rabbitt and Vyas, 1970). This happens in the absence of performance feedback, indicating that people monitor their own performance in these tasks. They seem to know, at least implicitly, that they have made an error. The cause of this slowdown is not clear. It may be related to ongoing processing of the error, suppression of the tendency to correct errors, or it may even be an artifact of whatever caused the error in the first place (Gehring et al., 1993, Rabbitt and Rodgers, 1977).

Recent ERP data provide evidence that there is actually error-related processing in the brain. In several choice reaction time paradigms, a negative-going deflection in the EEG was observed only on trials that were errors (Gehring, Coles, Meyer, & Donchin, 1995). This potential is known as the error-related negativity (ERN). The ERN begins immediately after the response and peaks at 100–150 ms post-response. Gehring et al. (1995) showed that the ERN is larger when accuracy is made important to participants by means of a reward system. Larger ERNs are associated with weaker error responses as measured by a dynamometer, and a greater probability of a correct response on the following trial. Taken together, these data suggest that the ERN reflects a neural system concerned with the detection and compensation for errors. However, it is not clear what the anatomical substrate for this system is. The poor spatial resolution of EEG does not allow for accurate localization, although the ERN seems to be stronger at frontal and central scalp electrodes. It is hypothesized that the potential may be originating in the anterior cingulate, which is known to play a role in attention-related tasks (Vogt, Finch, & Olson, 1992). Dehaene, Posner, and Tucker (1994) used source dipole localization techniques to estimate the source of the ERN, and confirmed that the anterior cingulate is a likely source.

Carter, Braver, Barch, Botvinick, and Cohen (1998) have an alternate interpretation of the ERN. They propose that anterior cingulate activity is not related to the occurrence of errors, but rather that it reflects conditions of increased response competition. If the cingulate monitors competition between ongoing processes, we would expect it to show error-related activity inasmuch as errors are a result of response competition. Using fMRI they demonstrated that during conditions of high response competition the anterior cingulate was activated for both correct and incorrect trials. However, a recent fMRI study by Kiehl, Liddle, and Hopfinger (2000) claims to show error-related activity in the anterior cingulate.

There are, then, no definite conclusions that can be drawn about the neural mechanisms for monitoring. We approach this issue from the perspective that each hemisphere can function as an independent cognitive unit, complete with its own perceptual, motoric, and linguistic abilities (Zaidel, Clarke, & Suyenobu, 1990). This view leads to questions about the ability of each hemisphere to contribute to monitoring. It is possible that each hemisphere has its own independent executive control, including the ability to monitor errors and adjust performance accordingly. Another possibility is that one hemisphere may be specialized for the monitoring of errors. Although the ERN does not appear asymmetric, if the source is a midline structure like the anterior cingulate, hemispheric differences might not be detectable in the ERP data.

Zaidel (1987) has argued for distinct error processing modules in the two hemispheres based on evidence from the lexical decision task. In lexical decision, participants must decide whether a string of letters is a real English word or not. Lateralized versions of this task usually fit a “direct access” model, meaning that each stimulus is processed by the hemisphere that receives it directly (Zaidel et al., 1990), thus making it a suitable task to investigate each hemisphere's role in error monitoring. Stein and Zaidel (1987) administered a version of this task in which participants were encouraged to correct their errors by a reward system. The results showed that the pattern of error correction responses was markedly different than that of initial responses. Whereas initial responses showed the typical lexical decision pattern of a right visual field (RVF) advantage, and faster responses to words than to non-words, the error correction responses showed no visual field advantage, and faster responses to non-words. The different characteristics of error correction responses were interpreted as evidence that error correction is performed by a distinct error correction module rather than by recomputation in the system that made the initial decision. Furthermore, error correction also appeared to fit the direct access model, which means that each hemisphere was able to independently monitor errors. Error correction was performed equally well by both hemispheres overall, but the right hemisphere (RH) showed an early monitoring advantage that decreased with practice, while the monitoring performance of the left hemisphere (LH) increased with practice. Stein and Zaidel (1987) suggest that this initial RH superiority in error correction may be due to a general advantage in processing feedback about the external environment.

Iacoboni, Rayman, and Zaidel (1997) made a similar suggestion based on their analysis of how the previous trial affects the current trial in a lateralized lexical decision task. In this experiment, accuracy improved on left visual field (LVF) trials following errors, while performance on RVF trials following errors was unaffected. An improvement after error may be interpreted as an appropriate compensatory response, reflecting a shift in strategy, allocation of resources, or some other adjustment towards better performance. Thus, the increase in accuracy in the LVF following errors may reflect a RH error processing advantage.

The present experiments are designed to investigate each hemisphere's ability to respond to external feedback about its performance. If each hemisphere has a different response to feedback information, this would provide further evidence for hemispheric modularity and independence of executive control in the two hemispheres. Notice that by giving feedback about performance we are bypassing the issue of hemispheric differences in detecting errors in favor of investigating differences in compensation for errors. Thus, in conditions where participants are receiving feedback, we are measuring what we call explicit monitoring. The hemispheres are explicitly informed of their performance and must react accordingly. In contrast, in conditions where no feedback is given, we can measure implicit monitoring, the extent to which participants react to errors in the absence of explicit feedback.

Few studies have addressed hemispheric reactions to feedback. The most relevant line of work is that of Derryberry, 1989, Derryberry, 1990 who investigated the effects of feedback in the context of emotional arousal. Since these experiments bear closely on our research, they will be described in detail.

Derryberry (1989) used positive and negative feedback presented both centrally and laterally in a lateralized simple reaction time task in order to examine the effects of different emotional states as feedback for the hemispheres. Positive feedback was provided after fast trials, and negative feedback was provided after trials that were inaccurate or slow. By manipulating the reaction time criteria for positive feedback, Derryberry created some blocks containing mostly negative feedback, and others with mostly positive feedback. Experiments 1 and 2 used feedback presented centrally. The main result was that in mostly negative feedback blocks, the reaction time was faster to RVF targets, while in mostly positive feedback blocks the reaction time was faster to LVF targets. Comparing these conditions to a control in which no feedback was presented led to the interpretation that negative emotion interferes with RH performance, while positive emotion facilitates RH performance. Experiment 3 replicated these results with lateralized feedback stimuli. In this experiment, the feedback stimuli were letter grades, with ‘A’ serving as positive feedback, ‘C’ as neutral feedback, and ‘F’ as negative feedback. Reaction time analysis showed that responses were faster in the LVF after positive feedback, and faster in the RVF after negative feedback, with no difference occurring after neutral feedback. Thus, positive feedback shifted performance in favor of the RH, while negative feedback shifted performance in favor of the LH.

Manipulations of the time between the feedback signal and the target stimulus showed that the feedback effects were greatest at 500 ms SOAs and less at 250 or 740 ms SOAs. Derryberry interpreted these results as indicating a “phasic arousal” mechanism activated by the RH. There is extensive evidence that the RH is sensitive to emotional processing and may control arousal mechanisms (Davidson, 1995, Tucker and Williamson, 1984). Tucker and Williamson (1984) have proposed more specifically that the RH responds in opposite ways to positive and negative emotions. According to Tucker and Williamson, right frontal regions become activated in negative mood states and serve to inhibit posterior perceptual regions of the brain, while positive moods decrease activity in the right frontal lobe leading to less inhibition. Derryberry considers the feedback in his experiments to elicit emotional responses, and suggests that the RH role in emotional control may account for his results.

Derryberry (1990) further explored the mechanism of this feedback effect by manipulating stimulus-response compatibility. The aim was to specify the locus of this emotional interference. The reasoning goes that if the interference is perceptual, then manipulating stimulus-response mappings should not interact with feedback effects. However, if the interference affects the RH at a motor or pre-motor level then this manipulation should interact with the feedback effects. There were three experiments in this paper that all used the same computer-based reaction time task with letters as feedback. Results showed that spatial compatibility did indeed interact with the feedback effects. Derryberry (1990) interprets these findings to “provide additional evidence that feedback-related emotional states modulate information processing within the right hemisphere” (p. 1268). These data are also used to form a more sophisticated interpretation of the feedback mechanism. Derryberry reasons, in line with Tucker and Williamson (1984), that the negative feedback leads to increased right frontal activity. However, in contrast with Tucker and Williamson, he claims that the resulting inhibition from the frontal lobe affects the communication between perceptual and motor systems, thus explaining the interaction with S-R compatibility in his results.

There are, however, alternate interpretations of these results. It is not clear that an “emotional” state was elicited by the feedback stimuli. Derryberry refers to the emotion resulting from negative feedback as frustration, but this state does not fall obviously along the positive-negative mood axis the RH has been associated with. Secondly, the nature of the feedback signals must be considered. The letters “A” and “F” are used to indicate good and poor performance, respectively, and “C” is used as a middle baseline. Feedback with the letter “C” may not be an appropriate baseline. The subjects in these experiments were college students to whom a “C” may represent unsatisfactory performance. Moreover, in this task if subjects are striving to respond as quickly as possible, a “C” may indicate failure to achieve the fastest category. Thus, the axis that has been interpreted as corresponding to positive-negative emotion may instead reflect increasing error awareness. In this view, reaction times following positive feedback may be considered a baseline in which no error has been detected. Then, the slowing of RH reaction times following “C” and “F” feedback may be interpreted as reflecting increased failure-related processing. Our experiments use a content-neutral stimulus as a control to ensure there is no success- or failure-related processing.

Other experiments that have investigated hemispheric responses to feedback have substantial interpretational difficulties. Kostandov (1988) used three different types of feedback in a lateralized task that required participants to distinguish time intervals between stimuli. Unfortunately, participants always used the left hand to respond that the interval was short and the right hand to indicate that it was long, which makes it difficult to draw conclusions about each hemisphere's role in processing the feedback.

The present experiments use a bilateral version of the lateralized lexical decision task to investigate hemispheric response to feedback. In bilateral lexical decision, a distractor stimulus accompanies the target stimulus, but is lateralized to the opposite visual field. Bilateral conditions have been shown to increase hemispheric independence (Iacoboni & Zaidel, 1996). Feedback was lateralized to ensure at least initial processing by the contralateral hemisphere. In Experiment 1, the feedback stimuli consisted of a woman's face either smiling or frowning to indicate the accuracy of the previous trial. Experiment 2 attempts to replicate the results of Experiment 1 using less meaningful stimuli, colored squares, as feedback. Experiment 3 examines the effect of using verbal feedback which may be considered a LH biased stimulus.

Section snippets

Participants

Twelve male and 12 female UCLA undergraduate students participated in this study for partial course credit. All participants learned English as their first language and were strongly right-handed as determined by a modified Oldfield–Edinburgh handedness inventory. All had normal or corrected-to-normal vision.

Materials and apparatus

Two lists of three, four, five, and six letter strings were created, each consisting of 64 English nouns and 64 pronounceable non-words. Words were counterbalanced for spelling-sound

Experiment 2

The purpose of this second experiment was to clarify the results of Experiment 1 by providing a lateralized feedback stimulus that was “hemisphere-neutral”. The ideal feedback stimulus for this experiment is one that both hemispheres can identify with equal ease. We chose to use square patches of different colors to indicate right and wrong answers. In addition, in contrast to Experiment 1, we used a within-subjects design to allow for better cross-conditional analyses.

Participants

For this experiment, 32 different UCLA undergraduate students (16 male, 16 female) participated for partial course credit. All participants learned English as their first language and were strongly right-handed as determined by a modified Oldfield–Edinburgh handedness inventory. All had normal or corrected-to-normal vision.

Materials and apparatus

The chinrest setup, response box, computer, and computer software were identical to those used in Experiments 1 and 2. The word lists were identical to those used in

General discussion

The general pattern observed here is a RH advantage in error processing that became progressively less pronounced when we changed the feedback stimulus from faces to colored squares to words. One of the goals of these experiments was to test whether the hemispheres differ in their sensitivity to different types of feedback stimuli. Comparing error rates across experiments, we find that trials following negative feedback when presented to the RVF have the greatest error rates in Experiment 1

Acknowledgements

This research was supported by NIH Grant NS20187. The authors thank Mr Ian Gizer for his organizational support, as well as Jean Zheng, Joanna Cheng, and Sheila Soleymani for their research assistance. We also thank Jan Rayman for her statistical advice.

References (36)

  • L.I Aftanas et al.

    Event-related desynchronization (ERD) patterns to emotion-related feedback stimuli

    International Journal of Neuroscience

    (1996)
  • E Bisiach et al.

    Anosognosia related to hemiplegia and hemianopia

  • C.S Carter et al.

    Anterior cingulate cortex, error detection, and the online monitoring of performance

    Science

    (1998)
  • R.J Davidson

    Cerebral asymmetry, emotion, and affective style

  • S Dehaene et al.

    Localization of a neural system for error detection and compensation

    Psychological Science

    (1994)
  • W.J Gehring et al.

    A brain potential manifestation of error-related processing

  • W.J Gehring et al.

    A neural system for error detection and compensation

    Psychological Science

    (1993)
  • E Goldberg et al.

    Three possible mechanisms of unawareness of deficit

  • Cited by (19)

    • Post-error adjustments and ADHD symptoms in adults: The effect of laterality and state regulation

      2016, Brain and Cognition
      Citation Excerpt :

      This interpretation is based on performance on lexical decision tasks with visual hemi-field stimulation. Zaidel and colleagues have argued that lateralized lexical decision task can provide a “direct access model”, referring to the fact that stimuli presented to the left visual field (LVF) are initially processed by the right hemisphere, and those presented to the right visual field (RVF) are initially processed by the left hemisphere (Iacoboni, Rayman, & Zaidel, 1997; Kaplan & Zaidel, 2001; Narr, Green, Capetillo-Cunliffe, Toga, & Zaidel, 2003; Zaidel, 1987; Zaidel, Clarke, & Suyenobu, 1990). In these studies, either the effect of an error on a subsequent performance or the effect of feedback on performance has been tested.

    • Hemispheric integration is critical for intact error processing

      2011, Neuropsychologia
      Citation Excerpt :

      A series of behavioral studies using HVFP by Zaidel and his colleagues, showed different laterality patterns for different post error compensatory behaviors. Error feedback presented to the right hemisphere (RH) caused a decrease in error rate (Iacoboni, Rayman, & Zaidel, 1997; Kaplan & Zaidel, 2001) while the same feedback presented to the left hemisphere (LH), increased latency on post error trials (Kaplan & Zaidel, 2001). The diversity of the laterality patterns observed for different error related indices, raises the possibility that the error process is a complex routine where several mechanisms, each allocated to a different hemisphere are orchestrated, shifting back and forth between the hemispheres.

    View all citing articles on Scopus
    View full text