Fig 1.
Dissociating space-specific from choice-specific reward expectation.
A. Schematic of reward-cued change detection task. Participants were cued (central colored arrowheads) about expected reward contingencies (fixed or variable, see panels D, G) for correct responses on each side, and reported the occurrence of an orientation change on the probed side (yellow arrowhead). Audio-visual feedback indicated the current trial’s reward and the cumulative reward accrued in the trial block (see text for details). B. Stimulus-response contingency table. Rows: event types. Columns: choice types. Shapes: Response types; octagon: hits/H, pentagon: misses/M, heptagon: false alarms/FA, hexagon: correct rejections/CR. C–E. Space-specific reward expectation session. C. Schematic of reward contingencies, in “mini-blocks”, for the space-specific reward expectation session (“gain” blocks only; see text for details). D. Rows: reward contingencies on the fixed (FX; top row, filled polygons) and variable (VR; bottom row, open polygons) reward sides for the space-specific reward expectation session. Columns: reward contingency conditions, “VR > FX” (left) or “VR < FX” (right). Each cell: contingency table, with the same conventions as in panel B. Numbers within polygons indicate the reward (INR) for the respective (correct) response type. Blank represents no reward for the incorrect response types. E. One-dimensional signal detection theory (SDT) models showing hypothesized modulations of sensitivity (d′) for space-specific reward expectation. Rows and columns: same conventions as in panel C. Gaussians: decision variable distributions at the respective location and reward contingency. Gray and orange: noise and signal distributions (FX, filled; VR, open). Horizontal arrows: perceptual sensitivity (d′). F–H. Choice-specific reward expectation session. F. Schematic of reward contingencies, in “mini-blocks”, for the choice-specific reward expectation session (“gain” blocks only). G. Same as in panel D but for the choice-specific reward expectation session. Columns: reward contingency conditions; greater expected reward for correct Yes responses (“Yes > No”/“Liberal”), or vice versa (“Yes < No”/“Conservative”). H. Same as in panel E, but for choice-specific reward expectation. Row and column conventions are as in panel G. Solid vertical lines: choice criterion (C). Arrows: Hypothesized criterion change for each reward contingency. (G–H). Other conventions are the same as in panels D and E, respectively.
Table 1.
Reward expectation effects on psychophysical parameters.
Fig 2.
Space-specific and choice-specific reward expectation independently modulate sensitivity and criterion.
A. Sensitivity (d′) for detecting changes at the FX (filled symbols) and VR (open symbols) sides for the two reward contingency conditions (“VR > FX”, left pair and “VR < FX”, right pair) in the space-specific reward expectation session (gain blocks only). Markers: individual participants (n = 24). Box plots limits denote the first and third quartiles; midline: median; whiskers: the minimum and maximum data points not considered outliers. Violin plots: kernel density estimates. Asterisks: statistical significance levels assessed with a Wilcoxon signed rank test (***p < 0.001, **p < 0.01, *p < 0.05, n.s.: not significant). Dashed line: datum (d′ = 0). B. Sensitivity modulation induced by reward contingency change (Δd′ = d′VR > FX – d′VR < FX) at the FX (x-axis) and VR (y-axis) sides in the space-specific reward expectation session (gain blocks only). p-value in the lower left: statistical significance for differences between Δd′FX and Δd′VR. Violin and box plots along the axes: marginals showing Δd′FX (filled symbols, horizontal) and Δd′VR (open symbols, vertical). Other conventions are the same as in panel A. C. Same as in panel A, but for criterion (c) in the space-specific reward expectation session. Other conventions are the same as in panel A. D. Same as in panel B, but showing criterion modulation induced by reward contingency change (Δc = cVR > FX – cVR < FX) in the space-specific reward expectation session. Other conventions are the same as in panel B. (E–F). Same as in panels (C–D), respectively, but comparing criteria (E) and c modulations (F) between the two reward contingencies (“Yes > No” and “Yes < No”) in the choice-specific reward expectation session. Other conventions are the same as in panels C–D. (G–H). Same as in panels (A–B), respectively, but comparing sensitivities (G) and d′ modulations (F) between the two reward contingencies (“Yes > No” and “Yes < No”) in the choice-specific reward expectation session. Other conventions are the same as in panels A–B. Data are available at https://doi.org/10.6084/m9.figshare.25966015 [34].
Fig 3.
Space-specific, but not choice-specific, reward expectation modulates attention-related ERPs.
A. (Top) ERP waveform from posterior electrodes (see bottom panel) in the space-specific reward expectation session (data averaged across n = 24 participants and both “gain” and “loss” blocks). x-axis: time relative to stimulus onset. Shaded regions: time epochs for quantifying contralateral N2pc (grey shading) and contralateral P300 (purple shading). (Bottom) Scalp topography of (left) N2pc and (right) P300 amplitudes for each electrode. Convention: FX and VR sides are on the left and right visual hemifields, respectively. White circles: occipitoparietal electrodes for ERP quantification. B. ERP waveform modulation induced by reward contingency change (ΔERP = ERPVR > FX – ERPVR < FX) in the space-specific reward expectation session for the FX (solid line) and VR (dashed line) sides. Waveforms were measured from electrodes contralateral to the respective side. Shading on traces (orange and gray): s.e.m. for ΔERP traces (FX and VR sides, respectively). Light gray overlaid trace and colored overbars: ERP waveform (not to scale) and time epochs, respectively, for reference, from panel A. C. (Left) Contralateral N2pc component modulation induced by reward contingency change (ΔN2pc = N2pcVR > FX – N2pcVR < FX) for the FX (filled symbols) and VR (open symbols) sides in the space-specific reward expectation session. (Right) Same as in the left panel, but for contralateral P300 component modulation (ΔP300 = P300VR > FX – P300VR < FX). y-axis in both panels is plotted in symmetric logarithmic scale. Other conventions are the same as in Fig 2A. Asterisks: statistical significance levels assessed with a Wilcoxon signed rank test (***p < 0.001, **p < 0.01, *p < 0.05, n.s.: not significant). D. Same as in panel B, but showing ERP waveform modulation induced by reward contingency change (ΔERP = ERPLib. – ERPCons.) in the choice-specific reward expectation session. Other conventions are the same as in panel B. E. Same as in panel C, but showing contralateral N2pc component (ΔN2pc = N2pcLib. – N2pcCons.; left) and contralateral P300 component (ΔP300 = P300Lib. – P300Cons.; right) modulation in the choice-specific reward expectation session. Other conventions are the same as in panel C. Data are available at https://doi.org/10.6084/m9.figshare.25966015 [34].
Table 2.
Reward expectation effects on neural and motoric metrics.
Fig 4.
Space-specific, but not choice-specific, reward expectation produces lateralized alpha power suppression.
A. Time–frequency spectrogram showing the difference in the reward-induced modulation of contralateral alpha power (Δα = αVR > FX – αVR < FX) between the FX and VR sides (ΔαFX – ΔαVR) in the space-specific reward expectation session (data averaged across all n = 24 participants). x-axis: Time relative to stimulus onset (vertical dashed line); y-axis: frequency relative to the individual alpha frequency (IAF) of each participant (horizontal dashed lines: IAF ± 0.5 Hz). Inset: Alpha-band power spectrum and IAF peak for a representative participant, measured in a 500 ms pre-stimulus time window from occipitoparietal electrodes (see panel B). Gray shading: IAF ± 0.5 Hz. B. Scalp topography of Δα (defined in panel A) for each electrode in the space-specific reward expectation session. White circles: occipitoparietal electrodes used for quantifying contralateral IAF alpha power in panel A, and for subsequent analyses. Other conventions are the same as bottom panel in Fig 3A. C. Reward-induced modulation of contralateral alpha power (Δα, at IAF ± 0.5 Hz) for the FX (filled) and VR (open) sides in the space-specific reward expectation session. Other conventions are the same as in Fig 2A. Asterisks: statistical significance levels assessed with a Wilcoxon signed rank test (***p < 0.001, **p < 0.01, *p < 0.05, n.s.: not significant). D. Same as in panel A, but showing the difference in the reward-induced modulation of contralateral alpha power (Δα = αLib. – αCons.) between the FX and VR sides in the choice-specific reward expectation session. Other conventions are the same as in panel A. E. Same as in panel B, but showing the scalp topography of Δα for the choice-specific reward expectation session. Other conventions are the same as in panel B. F. Same as in panel C, but showing reward-induced modulation of contralateral alpha power for the choice-specific reward expectation sessions. Other conventions are the same as in panel C. Data are available at https://doi.org/10.6084/m9.figshare.25966015 [34].
Fig 5.
Space-specific, but not choice-specific, reward expectation biases microsaccades and reaction times spatially.
A. (Top) Microsaccade rate (MSC) toward the FX (solid line) and VR (dashed line) sides, respectively, for the “VR > FX” reward contingency in the space-specific reward expectation session (data averaged across n = 24 participants). x-axis: time relative to stimulus onset. Shading (orange and gray): s.e.m. for MSC traces (FX and VR sides, respectively). Gray horizontal bar: time epoch for quantifying MSC rates. Black horizontal overline: temporal clusters with significant differences in MSC rates (p < 0.05). (Bottom) Same as in the top panel, but for the “VR < FX” reward contingency. Inset: Microsaccade traces of horizontal (solid lines) and vertical (dashed lines) gaze positions, for an exemplar participant (SI Methods). B. Microsaccade bias induced by reward contingency change (ΔMSC = MSCVR > FX – MSCVR < FX) toward the FX and VR sides in the space-specific reward expectation session. Other conventions are the same as in Fig 2A. Asterisks: statistical significance levels assessed with a Wilcoxon signed rank test (***p < 0.001, **p < 0.01, *p < 0.05, n.s.: not significant). (C–D). Same as in panels (A–B) and inset, but showing MSC rates, sample traces and bias for the “liberal” and “conservative” reward contingencies in the choice-specific reward expectation sessions. E. (Top) Distribution of reaction times (RT) for correct responses on the FX (deep shading fill) and VR (unfilled) sides for the “VR > FX” reward contingency in the space-specific reward expectation session (data averaged across n = 24 participants). Light shading: region of overlap. Solid and dashed vertical lines: RT medians for the FX and VR sides, respectively. (Bottom) Same as in the top panel, but for the “VR < FX” reward contingency. F. Same as in panel B, but showing reward-induced modulation of reaction times (ΔRT = RTVR > FX – RTVR < FX) in the space-specific reward expectation session. Other conventions are the same as in panel B. (G–H). Same as in panels (E–F), but showing RTs and their modulation for the “liberal” and “conservative” reward contingency conditions in the choice-specific reward expectation sessions. Data are available at https://doi.org/10.6084/m9.figshare.25966015 [34].
Fig 6.
Space-specific, but not choice-specific, reward expectation engages a conserved attentional resource.
A. (Top and Middle rows) Average sensitivity dynamics (n = 24 participants) during reward-contingency switch from “VR > FX” to “VR < FX” (top) and vice versa (middle), measured in sliding windows (width: 9 trials, shift: 2 trials), on the FX (filled circles) and VR (open circles) sides in the space-specific reward expectation session. x-axis: trial count relative to the “switch” trial (SI Methods). Lines: cubic polynomial fits for FX (solid) and VR (dashed) sides. Error bars: s.e.m. (Bottom row) d′ on the VR side (x-axis) plotted against d′ on the FX side (y-axis) in individual windows. Light filled circles and dark filled squares: “VR > FX” to “VR < FX” switch, or vice versa, respectively. Dashed line: x + y = 0. (B–D). Same as in panel A but showing normalized post-stimulus alpha power suppression (αpost), microsaccade rates (MSC) and reaction times (RT) dynamics (see text for details), respectively, in the space-specific reward expectation session. Other conventions are the same as in panel A. E. Regression coefficient estimates for a model in which reward-induced sensitivity modulation (Δd′ = d′VR > FX – d′VR < FX) was fit with a linear combination of (left) neural markers – N2pc amplitude, P300 amplitude, pre-stimulus alpha power (αpre), and post-stimulus alpha power (αpost) modulations – and (right) motoric markers – microsaccade rates and reaction time – in the space-specific reward expectation sessions (n = 24 participants). Error bars: jackknife s.e.m. Asterisks: statistical significance levels assessed with a permutation test (***p < 0.001, **p < 0.01, *p < 0.05, n.s.: not significant). F. Observed reward-induced modulation of sensitivity (Δd′) (x-axis) plotted against its predicted value (y-axis) with neural (left) or motoric (right) predictors. Filled and open symbols: Δd′ − s on the FX and VR sides, respectively. r-values: robust correlations. G. Same as in panel A, but showing criterion dynamics in the choice-specific reward expectation sessions. (H–L). Same as in panels B–F but during the choice-specific reward expectation sessions. Other conventions are the same as in corresponding panels (A–F). Data are available at https://doi.org/10.6084/m9.figshare.25966015 [34].