Article Figures & Data
Figures
- Figure 1.
Trial structure. The experimental stimuli were in Dutch; for illustrative purposes, the English translation is shown here.
- Figure 2.
Behavioral performance. A, Accuracy (rate of correct responses), separately for pre- versus retro-cue trials and for exemplars (purple) versus features (green). B, Reaction times. C, Speed–accuracy trade-off: rate of correct responses for fast to slow reaction time bins. D, Logarithm base 10 of the word frequency of responses. E, Intersubject variability (how many different responses were given over all participants). F, Results from a generalized linear mixed-effects model predicting reaction times from word frequency with subject as random-effects predictor. For plotting purposes, the raincloud plot shows coefficients of a GLM per subject. ***p < 0.001, **p < 0.010.
- Figure 3.
Time-frequency analysis comparing pre- versus retro-cue conditions. A, TFRs of pre- versus retro-cue trials, showing t values averaged over significant sensors denoted in the topography (right), masked by significance (opacity mask). B, Power averaged over significant sensors and time points (time window 4.9–11.5 s), separately for exemplar (purple) and feature (green) and for pre- and retro-cue conditions. Power values were converted to decibel by calculating the 10 times base 10 logarithm of the ratio versus baseline power (averaged from −0.45 to −0.25 s). C, Left, Time courses of alpha power (8–13 Hz) for pre- (dark gray) and retro-cue conditions (light gray). Horizontal lines indicate significant time windows. Right, Source reconstructions for pre- versus retro-cue, showing t values exceeding the 95th percentile. D, Same as C for the beta band (14–25 Hz). Error bars represent ±1 SEM. ***p < 0.001.
- Figure 4.
Time-frequency analysis comparing exemplar versus feature conditions. A, TFRs of exemplar versus feature in pre-cue trials, showing t values averaged over significant sensors denoted in the topographies (bottom), separately for the alpha (left), and the beta band cluster (right), masked by significance. B, Time courses of alpha (10–12 Hz) (left) and beta power (13–22 Hz) (right) for exemplar (purple) and feature (green). Horizontal lines indicate significant time windows. C, Source reconstructions for exemplar versus feature in pre-cue trials, showing t values exceeding the 95th-percentile (left, alpha; right, beta). Bar plots show power averaged over significant sensors and time points (time window 6–8 s), separately for each condition. Error bars represent ±1 SEM. ***p < 0.001.
- Figure 5.
Relationship between neural measures and reaction times. A, Scatterplot showing a positive relationship between IAF power in left frontal channels (denoted in the topography) and reaction times, as evidenced by a generalized linear mixed-effects model. B, Same as A for the beta band.
Tables
- Table 1.
Model fit statistics and log likelihood comparisons between mixed-effects regression models predicting reaction times
Model n Comparison Predictors df χ2 p Behavioral GLMMs 1 - WF 4 −411.31 ΝΑ 2 2 vs 1 WF + cue 5 −211.88 <0.001 3 3 vs 1 WF + task 5 −396.16 0.001 4a 4 vs 2a WF + cueXtaska 7a −185.58a <0.001a 5 5 vs 4 WFXcueXtask 10 −184.56 0.564 Neural GLMMs 1 - IAF 4 −485.49 ΝΑ 2 2 vs 1 IAF + cue 5 −260.17 <0.001 3 3 vs 1 IAF + task 5 −460.33 <0.001 4a 4 vs 2a IAF + cueXtaska 7a −224.57a <0.001a 5 5 vs 4 IAFXcueXtask 10 −222.74 0.301 Model n: model number. Comparison: contrasted models. Predictors: predictors included in the model. LRT, log likelihood ratio test; WF, word frequency; Task, task rule; IAF, individual alpha peak frequency power at left frontal sensors; χ2, chi-squared test. X denotes main effect and interaction.
↵a Best model based on likelihood ratio tests.
In this issue
