The supplementary data set includes data from 48 upper secondary and university students. Thestudents were matched into two similar groups based on cognitive scores, gender, and mathematicsgrade. Cognitive scores include Raven's Advanced Progressive Matrices [2] and Operation Span [3].Raven's matrices is a standardized test of abstract non-verbal problem-solving while operation Span isa complex working-memory task which measures executive functions involved in coordinating theprocessing and storage of information (see Refs. [2,3] for detailed information). The participantspracticed on mathematics tasks while their eye-movements were recorded by an EyeLink 1000, videobasedeye tracker with which was set to a sampling frequency of 500 Hz, a sampling duration above50 ms and a gaze resolution of approximately one degree across participants (see Ref. [1] for moredetails). One week after the practice session, the students took a mathematics test to evaluate whatthey remembered from the practice session. For a detailed description of the method, see Ref. [1]. Thedata analysis was based on Block distance and Ward's method [4] which through an iterative processmerged data into clusters of increasing dissimilarities (see supplementary data file and [4] for detailedinformation).Two types of mathematics tasks are presented (Tables 1e3). The general task design is indicated byFig. 1, where the five areas of interest are marked. Examples can be seen in Fig. 2. The tasks in Table 1contains a solution template, similar to textbooks, and were given to the first experiment group. Thetasks in Table 2 do not include a solution template, and were given to the second experiment group.Table 3 contains the illustrations included the tasks. The difference between the two task types is theinformation given in the areas marked as ‘formula’ and ‘example’. The purpose of the tasks was to