Fig 1.
(A) Examples of hierarchical Navon stimuli used in this study, illustrating both congruent (e.g., a large ‘H’ made of smaller ‘H’s) and incongruent (e.g., a large ‘H’ composed of smaller ‘S’s) conditions. (B) A schematic of a single misaligned composite face trial, showing the sequence of events. Study faces were consistently presented in the aligned condition, while test faces were displayed in either aligned or misaligned configurations. (C) Illustration of the complete composite face paradigm, including congruency and alignment manipulations, designed to assess holistic processing through congruent-aligned, congruent-misaligned, incongruent-aligned, and incongruent-misaligned conditions. (D) Overview of the experimental timeline. Each participant completed three sessions involving cTBS: left PPC stimulation, right PPC stimulation, and sham stimulation. The sessions were counterbalanced across participants following a Latin square design, ensuring at least a 24-hour gap between sessions to prevent carryover effects. Note. The face images shown are standardized stimuli from the face-categorization-lab dataset and do not depict any participant from the current study.
Fig 2.
TMS coil positioning and stimulation sites.
(A) Configuration of the figure-of-eight TMS coil during stimulation over the P3 region. The upper image shows the positioning of the coil over the identified “hot spot” (blue circle) on the P3 region, while the subsequent images depict the stimulation sites targeted for cTBS application over the P3 region. (B) Configuration of the figure-of-eight TMS coil during stimulation over the P4 region. The upper image illustrates the positioning of the coil over the identified “hot spot” (blue circle) on the P4 region, and the subsequent images represent the stimulation sites targeted for cTBS application over the P3 region. The locations of the left and right PPC were determined using the 10–20 EEG electrode placement system, ensuring precise targeting of cortical regions involved in attentional and perceptual processing.
Fig 3.
Simulated diffusion model trajectories for congruent and incongruent conditions.
The mean state of the diffusion process for congruent (top) and incongruent (bottom) conditions, modeled for a sample of two tasks: the Navon task in the global condition, where participants identify the global letter of the stimulus, and the composite face task in the aligned condition, where participants decide whether the top half of the face has changed. The shapes of the diffusion trajectories are determined by the parameter values: (amplitude of the irrelevant process),
(time scale of the irrelevant process),
(shape parameter of the Gamma distribution),
(drift rate of the relevant process),
(starting point of the decision variable), and
(decision boundaries). The time to maximum irrelevant interference, mathematically defined as
, simplifies to
because
is fixed at 2. The figure captures the temporal evolution of the decision variable under each condition, highlighting the transient peak of irrelevant information and its subsequent decline, which align with the theoretical dynamics proposed by the DMC. Note. The face images shown are standardized stimuli from the face-categorization-lab dataset and do not depict any participant from the current study.
Fig 4.
Model fit: predicted vs. actual behavioral data.
Comparison of predicted and actual values of behavioral data across experimental conditions, where the horizontal axis represents the actual values of behavioral data and the vertical axis represents the predicted values. Each data point represents the predicted versus actual values for error rates (left), correct response time quantiles (middle), and error response time quantiles (right). Panel (a) corresponds to the Navon task, with conditions of time, stimulation, and level stacked within the plots. Panel (b) corresponds to the composite face task, with conditions of stimulation and alignment stacked. Rows indicate congruency conditions (congruent vs. incongruent), highlighting the model’s ability to accurately capture behavioral patterns across tasks and conditions.
Fig 5.
Behavioral performance in the Navon task.
Violin plots with overlaid data points depict accuracy (top) and response times (RTs, bottom) across conditions. Error bars indicate one standard error of the mean, while shaded boxes represent the 95% confidence interval, assuming a hypothetical normal distribution of the mean. For the Navon task, conditions of congruency (congruent vs. incongruent) and level (global vs. local) are compared.
Fig 6.
Behavioral performance in the composite face task.
Violin plots with overlaid data points depict accuracy (top) and response times (RTs, bottom) across conditions. Error bars indicate one standard error of the mean, while shaded boxes represent the 95% confidence interval, assuming a hypothetical normal distribution of the mean. In the composite face task, results are organized by congruency and alignment (aligned vs. misaligned), reflecting the holistic interference effects.
Fig 7.
Parameter values of the DMC model for the Navon task.
Violin plots illustrate the distributions of key parameters—drc (drift rate), amp (amplitude),tau (latency), bnds (decision boundary), and res (residual time)—across distinct stimulation (LPPC, RPPC, CZ), Level (global, local), and Time (pre-test, post-test) conditions. Overlaid data points reflect individual participant values, while error bars indicate one standard error around the mean. Shaded boxes represent the 95% confidence intervals assuming a normal distribution for the mean. These visualizations highlight variations in cognitive processing dynamics under different experimental conditions.
Fig 8.
Parameter values of the DMC model for the composite face task.
Violin plots illustrate the distributions of key parameters—drc (drift rate), amp (amplitude), tau (latency), bnds (decision boundary), and res (residual time)—across distinct stimulation (LPPC, RPPC, CZ) and alignment (aligned, misaligned) conditions. Overlaid data points reflect individual participant values, while error bars indicate one standard error around the mean. Shaded boxes represent the 95% confidence intervals assuming a normal distribution for the mean. These visualizations highlight variations in cognitive processing dynamics under different experimental conditions.