Paradigm and results of reanalysis of Kerrén et al.

a The stimuli space for all experiments followed the same rationale. Images varied along a semantic (Animal / Object) and a perceptual (Photograph / Drawing) dimension. b The paradigm used in 37, where participants studied between 1 and 4 items and were later tested on one of the items using either a perceptual or semantic probe. c Using drift-diffusion modelling, we found reliable evidence for non-decision time (NDT) and drift rate differences between semantic and perceptual questions for load 3 (highlighted with grey bar) d Zooming in on NDT in load 3, semantic probes had lower NDT as compared to perceptual ones. e The effect was flipped for drift rate, where we observed a higher drift rate for semantic probes. f When instead organising the data according to lag between study and test, we observed a continuous build-up of increasing difference in drift rate between semantic and perceptual probes, with reliable evidence in lag 3 and 4. NDT was significant for all lag levels g In lag 3, we found very similar results to load 3, with a shorter NDT for semantic questions. h There was a smaller, yet reliable, difference in drift rate between semantic and perceptual questions.

Paradigm and results of Experiment 1.

a Behavioural paradigm. During study, participants saw three items sequentially in pseudorandom positions around the fixation cross. After encoding, they either saw a single circle (valid condition), indicating which item was going to be tested later, or three circles (neutral condition), indicating the three positions in which items were presented during study. Following a 1000ms blank screen, they were tested for the cued item (in the valid condition), or a random item (neutral condition). b Average accuracy for semantic trials was higher than perceptual trials across conditions. c Average reaction times (RTs) for semantic trials were faster than perceptual trials. d We observed faster semantic RTs in both valid and neutral conditions. e We observed higher accuracy for the semantic questions in the neutral condition, but not in the valid one. f No difference between conditions for valid trials. g And not for neutral trials either. h A significant negative slope was found for the valid condition, indicating that semantic prioritisation increased with lag. i This was not the case in neutral trials.

Balanced integration score.

a BIS was significantly higher in neutral trials (brown) as compared to valid trials (orange). b When organising the data in the valid condition according to lag between study and test, we observed a significant positive slope across lags, reflecting more semantic prioritisation for larger lags c This was not the case for the neutral condition, where a similar semantic prioritisation was present in all lag levels.

Drift-diffusion modelling of Experiment 1.

a (top) We observed similar non-decision time for perceptual (blue) and semantic (red) in the valid condition (solid lines), whereas for the neutral condition, perceptual trials had significantly longer non-decision time. (bottom) Same as top but subtracted perceptual from semantic for valid (solid line) and neutral (dashed line) conditions, separately. b (top) In the valid condition, we observed significantly lower drift rate for perceptual trials as compared to semantic trials. This was not the case for the neutral condition. (bottom) Same as top, but subtracted perceptual from semantic, for valid and neutral trials, separately.

Paradigm and results of Experiment 2.

a The study phase was identical to Exp. 1. After encoding, participants saw three different words (“Now”, “Wait”, or “Later”), indicating which condition they were in (Immediate, Delayed, or Prospective). In the Immediate condition, they were randomly tested on one of the presented items from the study phase, immediately following a brief 1000ms blank screen. In the Delayed condition, participants were exposed to a symbol consisting of a triangle, square and a circle, stacked on top of each other for 2000ms. After this, they were randomly questioned about one of the studied items. In the Prospective condition, participants saw a triangle, square or a circle with the question “Triangle?”, “Square?”, or “Circle?”, and were asked to respond yes if the word matched the symbol, and no otherwise. After 2000ms, the trial continued in the same manner as the two other conditions. In the Immediate condition, we asked participants the same symbol matching question after they had responded to the questions about the item from encoding (not shown in figure). b Average accuracy for semantic trials was higher than perceptual trials. c Average reaction times (RTs) for semantic trials were faster than perceptual trials. d We observed faster semantic RTs in all conditions, but no difference between conditions. e The difference between semantic and perceptual accuracy was lowest for the Immediate condition, followed by Delayed and Prospective, with a significant difference between Immediate and Prospective conditions.

Drift diffusion modelling of Experiment 2.

a We observed a reliable difference in non-decision time between semantic and perceptual questions between Prospective and Immediate and Prospective and Delayed conditions. b No such difference was seen in drift rate.

Comparison of the 4 drift-diffusion models used.

a The most complex model showed the best fit to the data followed by a model with non-decision time and drift as free parameters when analysing load 3 (left) and lag 3 (right). b The same was true in Experiment 1. c and Experiment 2.