Mark E. Walton, Joseph T. Devlin, & Matthew F.S. Rushworth

1. INVESTIGATING THE ROLE OF THE ANTERIOR CINGULATE CORTEX IN THE SELECTION OF WILLED ACTIONS AND PERFORMANCE MONITORING FMRIB Mark E. Walton, Joseph T. Devlin, & Matthew F.S. Rushworth Department of Experimental Psychology, Oxford Univeristy, UK & FMRIB Centre, JR Hospital, Headington, Oxford, UK EMAIL: mark.walton@psy.ox.ac.uk TH 169 INTRODUCTION Numerous imaging studies have implicated the anterior cingulate cortex (ACC) in willed response selection, error detection and conflict monitoring (Frith et al., 1991; Botvinick et al., 2001; Holroyd & Coles, 2002). These notions stand in contrast to recent electrophysiological studies in monkeys that have instead emphasized a role for ACC in using reward information to guide responses and in monitoring performance (Shidara and Richmond, 2002; Matsumoto et al., 2003; Ito et al. 2003). We have previously demonstrated selective ACC activation in a task switching paradigm when people have both to decide what response to make and evaluate its consequences (Walton et al., 2002). However, it was not clear from this previous experiment exactly which of the factors was driving ACC activity. Therefore, we used a version of the response switching task to investigate the degree to which this ACC activity could be explained by choosing what response to make (entailing willed action selection and response conflict) or by the combination of selecting a self-generated action and evaluating its consequences. METHODS SCANNING PROTOCOL & IMAGE ANALYSIS EXPERIMENTAL PARADIGM PRE-SCANNING Subjects taught one of 3 visuo-motor response rules outside of the scanner • Functional EPI was acquired on a 3T MRI scanner (TR = 3s, TE = 30ms, Spin Angle = 90º, 64 x 64 resolution, 256 x 160 mm FOV). • 28 x 4mm coronal slices with 2.5 mm in-plane resolution were taken starting from the front of the brain back to approximately the level of the central sulcus. • Subjects performed 2 blocks of 200 trials of each version of the task, making a total of 8 blocks per subject. • Analysis was carried out using using tools from the FMRIB software library (FMRIB, Oxford, UK, www.fmrib.ox.ac.uk/fsl). BOLD signal time-locked to the ‘switch’ cue was compared to that after a ‘stay’ cue in each block. Random effects statistical analysis was carried out using a general linear model approach with local autocorrelation correction. All images were thresholded using cluster detection statistics with z > 2.3 and p = 0.01. The response switching task involved making a response to 1 or 3 shapes according to which of the 3 above response rules was currently in use. After 8-12 trials, a cue appeared either instructing subjects to continue responding with the same set of rules (‘STAY’) or to switch to a different set (‘SWITCH’) SWITCH RESPONSE RULE A middle 1200 ms 70 ms 70 ms 70 ms Stay cue Switch cue correct STAY RESPONSE RULE B middle correct stimuli . . . feedback . . . middle RESPONSE RULE C correct index index correct correct BUTTON PRESS 1250 ms RT + 300 ms 1250 ms ring 4 types of block: INDEX MIDDLE RING CONDITION 1. GUESS 2. GENERATE 3. FIXED 4. CONTROL Stay Stay Switch Stay Switch Stay Switch Switch Cue incorrect GUESS condition: Subjects could choose their response on the 1st trial following a switch cue, but had to monitor its outcome to work out which set of response rules to use. GENERATE condition: Subjects could decide what response to make, but were not required to monitor feedback as they were told that whatever option they chose would always be correct. FIXED condition: Subjects were instructed by the experimenter to make a particular response, but, as it was only correct on 50% of trials, again had to monitor feedback to work out which rule to use. CONTROL condition: Subjects were instructed as to which response to make after a switch and were told that it would always be correct. Stay block Switch block Monitor feedback? Pre-cue response rule A A A A A A A A YES NO INDEX PRESS    ? ? ? ? INDEX PRESS Self-generated response? YES GUESS GENERATE Post-cue response rule A B C A B C A B C A B FREELY-CHOSEN ACTION MONITORING WILLED ACTION SELECTION EXTERNALLY-INSTRUCTED ACTION MONITORING SET-SWITCHING CONTROL NO FIXED CONTROL CHOOSING RESPONSE OUTCOME MONITORING RESULTS BEHAVIOURAL DATA • Large switch cost (difference between RT after ‘switch’ cue compared to after ‘stay’ cue) on trial 1 after cue in GUESS and GENERATE blocks. • No switch cost on 1st trial in FIXED and CONTROL blocks. • Significant residual switch cost on trial 2 in GENERATE, FIXED and CONTROL, but not GUESS. RTs after switch cue Median RT (msec) GUESS GENERATE RTs after stay cue Trials after switch / stay cue FIXED CONTROL IMAGING DATA CONCLUSIONS Is ACC activity being driven by willed action selection, performance monitoring, or combination deciding what to do and monitoring the consequences? ACC ACTIVITY NOT SIMPLY DRIVEN BY: (i) REQUIREMENT TO MAKE WILLED ACTION SELECTION (GUESS > GENERATE) (ii) CONFLICT WHEN CHOOSING BETWEEN TWO POSSIBLE RESPONSES (GUESS > GENERATE) (iii) ERROR DETECTION (GUESS > FIXED) GUESS (Sw – St) – FIXED (Sw – St) GENERATE (Sw – St) – FIXED (Sw – St) X NO AREA OF ACTIVATION WITHIN ACC * p < 0.05 * * * % signal change ACC: 0, 18, 36 ACC – DECIDING WHAT TO DO AND EVALUATING CONSEQUENCES OF THAT CHOICE • Significant main effects of ACTION GENERATION and FEEDBACK MONITORING • Caused by signal change in GUESS being significantly greater than in GENERATE, FIXED or CONTROL. • Identical results (GUESS signal > GENERATE / FIXED / CONTROL) when ACC ROI taken from previous task switching experiment (Walton et al., 2002). Gue Gen F C Acknowledgements: This work was supported by the Wellcome Trust, Royal Society and the MRC. We would like to thank all the analysis group at the FMRIB centre for help with analysing the data along with Peter Hobden, Sarah Hudson and Heidi Johansen-Berg for radiography.