Figures and data
Task procedure for Experiment 1 (A) and Experiment 2 (B).
(A) Experiment 1 embedded a perceptual matching task within the delay interval of a working memory (WM) change-detection task. After viewing a sample color and a short delay, participants reported the color of a continuously visible target using a color-wheel (perceptual matching), followed by a second delay and a change-detection, where change magnitude varied at every 60° in both directions (−180°:60°:180°). (B) Experiment 2 combined perceptual matching with continuous WM recall. After the sample and delay, participants performed a perceptual matching report, followed by a second delay and a continuous recall of the memorized color using a color-wheel.
Hierarchical Bayesian parameter estimates for Experiment 1.
Posterior distributions of the Mixture model population-level parameters are shown separately for CW (+40°) and CCW (−40°) conditions for bias (μ), imprecision (σ), and guess rate (g). The lower panels show posterior distributions of condition effects (CCW − CW) for each parameter. A credible condition difference was observed for bias (Δμ), with 95% highest density intervals (HDI95%) beyond zero. Shaded areas represent HDI95%.
Mouse-tracking evidence for WM-induced perceptual bias in Experiment 1.
(A) Average trajectory paths (left) and average horizontal trajectories (right), plotted separately for CW (+40°) and CCW (−40°) conditions. Shaded bands indicate 95% confidence intervals. (B) Area-under-the-curve (AUC) analyses of trajectory curvature. Early and late onset represent trials median split of movement onset latency. Error bars standard errors of the mean.
Destination-vector transformation and decomposition analyses in Experiment 1.
(A) Schematic illustration of the destination-vector (DV) transformation method, which represents each movement segment as an angle (θ) relative to the target based on two consecutive moment points, allowing reconstruction of intended feature values across time (Park & Zhang, 2024). (B) Reconstructed DV dynamics for overall deviation (green), capture (cyan), and shift (blue). Thin gray lines show individual participant DV series and thick lines with shaded bands indicate group means and 95% confidence intervals. (C) Slope differences between capture and shift DV series. Capture showed a steeper early slope, whereas shift dominated later in the movement. Colored horizontal bars denote time windows with significant cluster-based permutation effects (p < .05).
Decomposition of trajectory bias into capture and shift components in Experiment 1.
(A) Schematic illustration of the endpoint-based decomposition. The observed overall trajectory (green) is expressed as the sum of an endpoint-inconsistent capture component (cyan) and an endpoint-consistent shift component (blue). (B) Conceptual examples that capture and shift can manifest as independent attractive (positive) or repulsive (negative) deviations from the source of bias (red markers). White circles and green triangles indicate the true target location and the judgment endpoint, respectively. (C) Average time course of horizontal deviation for the overall, capture, and shift trajectories, time-normalized from movement onset to the final click. Shaded bands indicate 95% confidence intervals (CI95%). (D) Correlations of area-under-the-curve (AUC) measures for capture and shift with final perceptual bias (top row) and movement onset latency (bottom row). Solid lines show the best linear fit and dashed lines the CI95%, with red lines indicating statistically significant effects.
Working memory change-detection performance in Experiment 1.
(A) Proportion of “same” responses as a function of change magnitude, with Gaussian fit curves (solid lines) shown separately for CW (+40°) and CCW (−40°) trials. Markers with error bars indicate group means and 95% confidence intervals. Inverted triangles along the x-axis denote the corresponding locations of the preceded perceptual target features. (B) Posterior distributions of the point of subjective equality (PSE, μGauss) estimated separately for CW and CCW conditions.
Hierarchical Bayesian Mixture modeling of continuous-report errors in Experiment 2.
(Top row) Posterior distributions of population-level parameters for perceptual matching (Perc; red) and working memory recall (WM; green), shown separately for clockwise (CW; solid lines) and counterclockwise (CCW; dotted lines) conditions. (Bottom row) Posterior distributions of condition effects (CW − CCW) for each parameter. Markers and horizontal bars along the x-axis represent posterior means and 95% highest density intervals, respectively.
Reconstructed destination-vector (DV) dynamics and decomposition analyses in Experiment 2.
Mean DV series with 95% confidence intervals over normalized time for perceptual matching (Perception; red) and working memory (WM) recall (green), shown separately for DVOVERALL, DVCAPTURE, and DVSHIFT (left to right). Black horizontal bars indicate time intervals with significant cluster-based permutation effects contrasting perceptual matching and sign-flipped WM recall (p < .05).
