1
Supplementary Materials and Results
Figure S1 depicts the experimental setup used in the study.
Figure S1. (a) Front view of the Siemens Trio 3T scanner used in the experiment with a chess position projected. (b) Twelve-channel head coil with the mirrors and the light source for the eye tracker. (c) The iView X MEyeTrack Long Range (LR) infra-red remote long-range eye-tracking camera. (d) Response buttons used in the study.
Eye Movement Parameters – additional analysis.
Figure S2a-b presents the number of fixations and average duration of fixation among expert and novice chess players in the chess and control tasks. The eye movement parameters were in line with the behavioral results (see Figure 2). Just like they need less time, experts also needed considerably fewer fixations than novices to enumerate the objects of interests in the chess task [ANOVA for expertise in the chess task – F(1, 9) = 33.1, p < .0001]. Normal positions required fewer fixations than random positions mostly because experts needed fewer fixations in normal than in random positions [ANOVA for type of position in the chess task – F(1, 9) = 16.3, p = .003; normal vs random positions among experts, t(4) = 3.5, p = .024]. There were no differences in the duration of fixation between experts and novices, but both groups, in particular experts, made longer fixations in the normal positions [ANOVA for type of position in the chess task – F(1, 9) = 7.7, p = .022; normal vs random positions among experts – t(4) = 3.4, p = .027]. The longer fixation duration on normal positions may imply that players, in particular experts, were trying to capture the structure and relations in the position. Since the common relations and structure were disrupted in random positions, long fixations were probably of less use.
Novices made more fixations than experts in the control task, in particular on the normal positions, the differences were not statistically significant. Although there were no significant differences between experts and novices in the fixation duration in the control task, experts’ fixations were longer, in particular on the normal positions [ANOVA for type of position – F(1, 9) = 8.4, p = .017; ANOVA for expertise × type of position interaction in the chess task – F(1, 9) = 10.6, p = .010; normal vs. random positions among experts, t(4) = 3.6, p = .024]. The longer fixation duration on normal positions among experts may imply interference from automatically elicited task-irrelevant chess knowledge. Due to these longer fixations, experts were also probably slower and made more mistakes (see Figure 2).
Figure S2.(a) Average number of fixations per position experts and novices needed to complete the chess and control task depending on the position type. (b) Average duration of a fixation for expert sand novices in chess and control tasks depending on the position type. Error bars indicate SEM. *p < .05 in a two-tailed t-test for dependent samples.
The summary for the first second parameters are presented in Figure S3. Both experts and novices produced a similar number of fixations on both tasks and on both position types in the first second of trials (Figure S3a). The durations of fixations in the first second did not differ between experts and novices on the chess task and were also not depended on the position type (Figure S3b). In the control positions the only difference was that both groups had longer fixations on normal positions in the first second [ANOVA for type of position in the control task – F(1, 9) = 6.6, p = .030]. Although there was a trend that this difference is primarily driven by experts, the interaction and the actual comparison among experts did not reach the statistical significance.
Even in the very first second of a trial (Figure S3c), the efficiency of experts and novices was different (main effect of expertise): experts focused closer to the objects of interest than novices did [ANOVA for expertise in the chess task – F(1, 9) = 8.5, p= .017]. Similarly, both groups focused closer to the objects of interest on normal than on random positions (main effect of position type), but this trend was mostly driven by the differences among experts [ANOVA for type of position in the chess task – F(1, 9) = 12.9, p = .006; normal vs. random positions among experts, t(4) = 6, p = .004]. In contrast, there were no significant differences on the control task.
When we calculated the distance from the nearest object, irrespective of its relevance (Figure S3d), there were no differences among experts and novices, but both groups tended to focus nearer the objects in normal positions [ANOVA for type of position in the chess task – F(1, 9) = 6.8, p = .028]. In contrast, there were no significant differences in the control task.
Figure S3. (a) Average number of fixations per position experts and novices made in the first second of the chess and control tasks depending on the position type (b) Average duration of a fixation for experts and novices made in the first second of the chess and control tasks depending on the position type. (c) Average distance (in pixel) of fixation from the nearest object of interest in the first second. (d) Average distance (in pixel) of fixation from the nearest object irrespective of the interest in the first second. Error bars indicate SEM. *p < .05 in a two-tailed t-test for dependent samples.
Neuroimaging – additional analyses.
Here we present the results of a 2 x 2 ANOVA on the identified brain areas (see Figures 5 and 6) in the chess task (note that there were no significant effect in the control task). The presented ANOVAs just confirm the whole brain analysis performed on the same task and are mostly presented for control purposes. Some of the contrasts (e.g., main effects of position type and interactions) provide indications we used for more specific claims in the main text.
MAIN EFFECT OF EXPERTISE in the CHESS TASK (whole brain analysis brain areas): F(1, 21) = 30.5, p < .001; F(1, 21) = 26.6, p < .001; F(1, 21) = 24.4, p < .001; F(1, 21) = 22.2, p < .001; F(1, 21) = 30.6, p < .001; and F(1, 21) = 38.2, p < .001, for rCoS, pMTG, OTJ, SMA, M1, and insula, respectively. expertise × type of position interaction in the chess task: F(1, 21) = 7.1, p = .014; F(1, 21) = 3.9, p = .062, and F(1, 21) = 10.7, p = .004 for SMA, M1, and insula, respectively; normal vs random positions among experts – t(7) = 3.2, p = .015; t(7) = 1.3, p = .25, and t(7) = 2.9, p = .024, for pre-SMA, M1, and insula, respectively. ANOVA for expertise × type of position interaction in the chess task – F(1, 21) = 41.2, p < .001, for the right CoS; normal vs random positions among experts – t(7) = 7.1, p < .001, for the right CoS.
INTERACTION EXPERTISE x POSITION TYPE in the CHESS TASK (whole brain analysis brain areas): ANOVAs for expertise × type of position interaction in the chess task – F(1, 21) = 30.1, p = .001, and F(1, 21) = 53.6, p = .001, for left and right CoS, respectively; normal vs. random positions among experts – t(7) = 11.8, p < .001 and t(7) = 4.5, p < .003, for left and right CoS, respectively. ANOVAs for expertise in the chess task – F(1, 21) = 16.2, p = .001, and F(1, 21) = 39.8, p = .001, for left and right CoS, respectively. ANOVAs for type of position in the chess task – F(1, 21) = 24.8, p = .001, and F(1, 21) = 31.1, p = .001, for left and right CoS, respectively.
We showed that there were no substantial behavioural and eye-movement differences in the first second of trials modelled in the fMRI analysis. A remaining question is whether our approach of modelling the first second effectively controls for the different durations of trials. We thus added another factor to our design by splitting the trials into two halves – short and long trials (different cut-offs were used for experts and novices). If there were a systematic influence of trial duration even after only the first second of each trial is included in the analysis, then we would expect to find a difference between longer and shorter trails. Such a difference could eventually also produce patterns of results different from that of the original analysis (see Figures 5 and 6 in the main text). There were, however, no differences in the activation between short and long trials, even with a liberal significance threshold of p < 0.01. Similarly, the patterns of activations when the main effects and interaction were tested using only short or long trials resembled that found in the original analysis (the significance level was of course lower given that the trials were further split into two halves in the new analysis). We further verified these results by using already established ROIs and extracting the activation levels for short and long trials (see Figure S5). There were no significant differences between short and long trials, and the patterns of results were identical to the original analysis.
Figure S4. The influence of long and short trials on the brain regions activated in the original analysis. Orange color represents short trials; green color long trials. Error bars indicate SEM.
Video Legend
Video 1. Expert_chess_task_normal. An example of eye movements of an expert in the chess task on a normal position.
Video 2. Novice_chess_task_normal. An example of eye movements of a novice in the chess task on a normal position.
Video 3. Expert_chess_task_random. An example of eye movements of an expert in the chess task on a random position.
Video 4. Novice_chess_task_random. An example of eye movements of a novice in the chess task on a random position.
Video 5. Expert_control_task_normal. An example of eye movements of an expert in the control task on a normal position.
Video 6. Novice_control_task_normal. An example of eye movements of a novice in the control task on a normal position.
Video 7. Expert_control_task_random. An example of eye movements of an expert in the control task on a random position.
Video 8. Novice_control_task_random. An example of eye movements of a novice in the control task on a random position.
1