(A) General Structure of a sample trial of the experimental task. In this task, participants are presented with two images, sequentially, such that only one image is present on the screen at each time. After the second image is presented, participants are asked to determine the numeric value corresponding to the pair of images and to provide their answer via a button box. Feedback is provided after the answer. Crucially, the numeric value depends on the task condition (see B and C). (B) Left: Value matrix in the hierarchical task (HIER). Here, the color of the first object determines the ‘context’ set of possible values ([1, 3] if red and [2, 4] if green). Then the shape category of the second ‘item’ (in relation to the first one) determines the final value of the pair. In this rule system ‘same’ shape does not mean exact visual similarity but rather membership to the same category set (see methods and Figure S3for details). Right: Value matrix in the iterative task (ITER). Each object category is associated with a specific value which does not depend on the context. The value of each pair is the sum of the objects’ individual values. (C) Regions of Interest (ROI) for hypothesis testing ([L]eft and [R]ight). IFG pars opercularis and fronto-medial cortex masks extracted from Harvard-Oxford probabilistic map (http://neuro.debian.net/pkgs/fsl-harvard-oxford-atlases.html) with 50% and 10% threshold, respectively; Hippocampus mask taken from (Tian et al., 2020) and pSTS mask from (Schaefer et al., 2018). These ROIs are based on regions known to yield activity during hierarchical processing either in human language or animal spatial navigation (see literature review in the text).

(A) Whole brain-analysis. Main effect of TASK (HIER-ITER, red) and interaction TASK x POSITION ([H2-H1]-[I2-I1], green; [H1-H2]-[I1-I2], blue). Statistical t-maps were thresholded and binarized at the cluster level using an FWE-corrected p-threshold of 0.05 and projected on a brain mesh using Mango. (B, C, D) ROI analyses. An asterisk indicates a significant effect of TASK (*task) or of the interaction TASK x POSITION (*int; both p<.05). (B) Hippocampus (HIP) and medial Prefrontal Cortex (mPFC). (C) Inferior Frontal Gyrus pars opercularis (IFG op) and posterior Superior Temporal Sulcus (pSTS). (D) hippocampus sub-divisions along its long axis. (E) Schematic POSITION x TASK effect within the hippocampus. In all subplots, the y-axis describes the average height of beta values for a given condition. For ROI-plots with subject-level datapoints see Supplementary Figure S3.

Summary of whole brain results for the main effect of TASK and interaction TASK x POSITION.

All activations are significant at cluster-level FWE-corrected p <.05, and voxel-level at p<.001. Cluster and main peaks were labeled using the AAL3 toolbox. In this summary, we only include activations that make up > 5% of a cluster and of the labeled area. Abbreviations: cn – cluster number, k – number of voxels in cluster %c – percentage of the cluster belonging to the area, nv – number of voxels of that cluster within the respective area, %r percentage of region occupied by voxels from the cluster.

(A) Whole brain analysis interaction TASK x EXPERIENCE. Contrast [Hier Late - Hier Early] – [Iter Late - Iter Early] (red). Statistical t-maps were thresholded and binarized at the cluster level using a FWE-corrected p-threshold of 0.05 and projected on a brain mesh using Mango. (B, C) ROI analyses. Hippocampus (HIP), right and left Inferior Frontal Gyri (IFG), right and left posterior Superior Temporal Sulcus (pSTS) and medial Prefrontal Cortex (mPFC). An asterisk indicates a significant effect of TASK (*task), of EXPERIENCE (*exp) or their interaction (*int; all p<.05).