Neural dynamics of error processing in medial frontal cortex
Introduction
To survive in changing environments, an organism must be able to adapt its behavior to the situation at hand. This flexibility can be achieved by evaluating response outcomes and adjusting behavior accordingly (Dickinson, 1985). In this regard, error signals provide important evaluative information, since they indicate that a behavior was inadequate given the current context and that, in future, a different response needs to be selected (Holroyd and Coles, 2002).
Existing data on the neural substrates of action selection indicate that the medial frontal cortex plays a crucial role in selecting actions on the basis of their outcomes (Matsumoto and Tanaka, 2004) and subsequent monitoring of response outcomes (Holroyd et al., 2004a, Ridderinkhof et al., 2004). Rather than attributing a single role to this vast cortical expanse, recent studies have started to associate different functions to the different anatomical structures that lay within the medial frontal cortex (Picard and Strick, 2001, Rushworth et al., 2004). In this context, an anterior portion of the cingulate cortex, the rostral cingulate zone anterior (RCZa), has been specifically associated with processing of error information and selecting appropriate behavioral adjustments (Holroyd and Coles, 2002, Rushworth et al., 2004, Fiehler et al., 2004).
These inferences on the neural bases of error processing have been obtained in the context of a “static” experimental environment, in which the organism knows the behavior that is appropriate for the current situation. Thus, a given response can be evaluated immediately against an internal representation of the correct stimulus–response relationship. Should the response be incorrect, error information is available from an internal error-detection process at the time of the response (Gehring et al., 1993, Holroyd et al., 2005). However, in a novel environment, with as yet unknown stimulus–response associations, error information is not available until the delivery of external performance feedback. This implies that, during the learning of stimulus–response associations by trial and error, the time at which error information is available will change. Prior to learning, error information will not be available until external performance feedback is delivered, but after learning, error information will be available earlier from internal sources at the time of the response itself. Thus, a neural structure that adjusts behavior as a function of the evaluation of response outcomes should dynamically shift its responsivity as a function of learning, from external sources provided by error feedback to internal sources associated with the error response itself. We predicted that, following error feedback, activity in the anterior cingulate cortex would decrease as learning proceeds; conversely, following an erroneous response, activity in the anterior cingulate would increase as learning proceeds. These predictions can be derived from a neuro-computational model (Holroyd and Coles, 2002) that formally describes the relationship between neural systems involved in outcome evaluation with those involved in action selection.
To test these predictions, we asked human subjects to learn arbitrary visuomotor mappings (Wise and Murray, 2000, Toni et al., 2001), using performance feedback, while measuring their cerebral activity using functional magnetic resonance imaging (fMRI). Participants were presented with line drawings, each of which was associated with pressing one of four response buttons (Fig. 1). We manipulated the degree of learning achieved during the scanning session by varying the number of times a given visuomotor mapping was presented. For one condition (High Learning, HL), four distinct visuomotor mappings were presented 36 times each over the course of the scanning session, enabling the subject to fully learn the visuomotor associations. For a control condition (Low Learning, LL), 24 different mappings were presented 6 times each. A reaction time (RT) deadline ensured that participants made errors, even during learned performance. Crucially, by varying the delay between response and feedback, and by introducing neutral feedback on some of the trials, we were able to dissociate the hemodynamic responses elicited by response and feedback (see Experimental timing).
Section snippets
Subjects
We studied eight right-handed male volunteers (mean age = 30.4 years, SD = 13.4) with normal or corrected-to-normal vision after obtaining informed consent according to institutional guidelines of the local ethics committee (CMO region Arnhem-Nijmegen, Netherlands). They were paid €10 per hour for their participation. Imaging data from 5 additional subjects were discarded, since these subjects either failed to learn the appropriate stimulus–response mappings adequately (2 subjects, less than
Behavioral data
Behavioral data indicated that our design was successful in manipulating the degree of learning achieved by the participants during the scanning session. Participants learned the stimulus–response mappings at a faster rate in the High Learning condition than in the Low Learning condition (Condition × Time interaction on Error Rate: F(7,49) = 3.2, P = 0.035, Fig. 3a). Although participants never reached error-free performance during the scanning session in either condition (because of the RT
Discussion
The present data indicate that, over the course of learning a set of arbitrary visuomotor mappings, a region along the cingulate sulcus (RCZa) shifts its responsiveness to different sources of error information as a function of learning. Error feedback-related activation decreases as learning proceeds, while error response-related activation increases, and these effects are reciprocal (Fig. 4). These results show not only that the anterior cingulate cortex responds to both internal (Carter et
Acknowledgments
The authors would like to thank Paul Gaalman for excellent technical assistance and Peter Hagoort, Karl Magnus Petersson, and Guillen Fernandez for helpful comments on an earlier draft of the manuscript. C.B.H. was supported by National Institute of Mental Health (NIMH) postdoctoral fellowship MH63550. S.N. was supported by the Netherlands Organization for Scientific Research (NWO).
References (39)
- et al.
Cognitive and emotional influences in anterior cingulate cortex
Trends Cogn. Sci.
(2000) - et al.
Event-related fMRI: characterizing differential responses
NeuroImage
(1998) - et al.
Stochastic designs in event-related fMRI
NeuroImage
(1999) - et al.
Dissociable executive functions in the dynamic control of behavior: inhibition, error detection, and correction
NeuroImage
(2002) - et al.
Thresholding of statistical maps in functional neuroimaging using the false discovery rate
NeuroImage
(2002) - et al.
Generalisability, random effects and population inference
NeuroImage
(1998) - et al.
The role of the medial prefrontal cortex in achieving goals
Curr. Opin. Neurobiol.
(2004) - et al.
Volition and conflict in human medial frontal cortex
Curr. Biol.
(2005) - et al.
Imaging the premotor areas
Curr. Opin. Neurobiol.
(2001) - et al.
Action sets and decisions in the medial frontal cortex
Trends Cogn. Sci.
(2004)