The Medial Frontal Cortex and the Rapid Processing of Monetary Gains and Losses William J. Gehring and Adrian R. Willoughby

Supplementary Material

The electroencephalogram was recorded using tin electrodes embedded in a nylon mesh cap (Quik-cap, Neuromedical Supplies, Sterling, Virginia, USA), with a left mastoid reference and a forehead ground. An average mastoid reference was derived off-line. EEG data were corrected for ocular movement artifacts using the Gratton algorithm (1). Data were recorded from 0.01 to 70 Hz (half-amplitude cutoffs) and digitized at 400 Hz, using a SynAmps data acquisition system (Neuroscan Labs, Sterling, Virginia, USA). Prior to analysis the data were filtered with a 9-point Chebyshev type II low-pass zero-phase shift digital filter (Matlab 5.3, Mathworks, Inc., Natick, Massachusetts, USA), with a half-amplitude cutoff at 12 Hz.

ERP activity was quantified as the mean amplitude in the 200-300 ms epoch following the onset of the stimulus, relative to a 100 ms pre-stimulus baseline. A 2 (gain/loss) 38 (electrode location) repeated-measures analysis of variance (ANOVA) revealed a significant gain vs. loss effect, F (1, 11) = 9.71, P = 0.0098, MSE = 28.54, and a significant gain/loss location interaction, F (37, 407) = 8.21, P = 0.00022, MSE = 0.56. When necessary, in this and all analyses, P-values were adjusted using the Greenhouse-Geisser correction for violations of the ANOVA assumption of sphericity. All figures show error bars consisting of 2 standard errors of the mean, derived from the corresponding mean squared error (MSE) from the ANOVA.

In the analysis dissociating gain/loss from correct/error effects on the amplitude of the MFN (Figure 3), the 2 (gain/loss) 2 (correct/error) ANOVA showed a significant effect of gain/loss status, F (1, 11) = 21.19, P = 0.00076, MSE = 9.22. The error/correct effect was not significant F (1, 11) = 1.63, P = 0.23, MSE = 4.58.

To analyze the effect of position within the block on the proportion of risky responses, we used a repeated-measures ANOVA with a single four-level factor representing the quarter within a block. The effect of block quarter was significant, F (3, 33) = 4.24, P = .038, MSE = 0.0049, and a separate test contrasting the first quarter to the last quarter was also significant, F (1, 11) = 6.74, P = 0.024, MSE = 0.0089. The means for the quarters were (first to last): 0.58, 0.51, 0.53, 0.48.

The analysis of the effects of previous outcome on the proportion of risky choices and on the MFN effect (Figure 4) distinguished between two kinds of trials: trials where the alternatives were different ([5][25] or [25][5]) and trials where alternatives were equal ([5][5] and [25][25]). We refer to the former case as "choice" trials (where subjects had the opportunity to make a risky or cautious choice), and the latter case as "no-choice" trials. In the analysis of proportion of risky choices, a 2 (gain vs. loss) 2 (25 vs. 5) 2 (choice vs. no-choice) repeated measures ANOVA showed a significant effect of the gain/loss status of the previous trial, F (1, 11) = 23.54, P = 0.00051, MSE = 0.018. The significant gain/loss 25/5 interaction F (1, 11) = 5.59, P = 0.038, MSE = 0.021, showed that the proportion of risky choices was affected by the gain/loss status of the previous trial more when the outcome was a gain or loss of 25 cents than when it was 5 cents. The MFN effect was also affected by the gain/loss status of the previous trial, F (1, 11) = 6.38, P = .028, MSE = 5.17, with a greater MFN effect following losses than following gains. The gain/loss 25/5 interaction was not significant, F (1, 11) = 0.94, P = 0.35, MSE = 5.71. The linear trend apparent in Figure 4 was significant for both the probability of a risky choice and for the MFN effect (F (1, 11) = 17.05, P = 0.0017, MSE = 0.032 and F (1, 11) = 5.04, P = 0.046, MSE = 7.56, respectively). Standard error bars are derived from the gain/loss x 25/5 interaction MSE.

Using the Brain Electrical Source Analysis (BESA) software (2), we derived a best-fit single-dipole model of the MFN, based on the amplitude of the gain-loss difference waveform computed at its peak, 265 ms following the stimulus. Then, with BESA's coordinates for the dipole solution, we located the dipole within a canonical magnetic resonance imaging template of the human head, derived from an average of scans from 152 individual heads (file avg152T1.img, available as part of the SPM99 software at http://www.fil.ion.ucl.ac.uk/spm, Wellcome Department of Cognitive Neurology, London). We aligned the spherical geometry of BESA with the MRI template by fitting a sphere to a convex hull circumscribing the average head defined in the MRI image, using a least-square best-fit criterion. The origin of the sphere within the voxel coordinates of the template was, x = 46.1, y = 55.8, z = 39.8, and the radius of the sphere was 47.5 voxels. The spherical coordinates of the dipole in BESA's coordinate system were theta = 67.0, phi = 83.4, eccentricity = 40.9%. Figure 2 depicts the dipole solution using a sphere with a radius of 5 voxels, centered at those coordinates. The residual (unaccounted-for) variance associated with the best-fit dipole model shown in Figure 2 was less than 5%.

References

1. G. Gratton, M. Coles, E. Donchin, Electroenceph. Clin. Neuro. 55, 468 (1983).

2. M. Scherg in Advances in Audiology: Vol 6. Auditory Evoked Magnetic Fields and Electrical Potentials, F. Grandori, M. Hoke, G. L. Romani, Vol. Eds., M. Hoke, Series Ed. (Karger, Basel, Switzerland, 1990), pp. 40-69.