Figure 1. Direction-imagination task.

(a) Twelve evenly spaced directions were sampled using 18 buildings distributed regularly across Donderstown. We sampled each direction (indicated by black arrows) from different start locations (yellow circles), which dissociated the directions from visual features of imagined views (Figure 1—figure supplement 2), and employed a counterbalancing regime ensuring equal sampling of directions and start locations throughout the experiment (see Materials and methods). Buildings marked with a green circle served as target locations only. Importantly, the regular arrangement of building locations did not correspond to the street layout and was not revealed to participants, who experienced Donderstown only from a first-person perspective (see also Figure 1—figure supplement 1d). (b) Trials began with a cue indicating start (top building name) and target (bottom building name) location and thereby defining the relevant direction (black arrow). During an imagination period the screen was black and participants were instructed to imagine the view they would encounter when standing in front of the start building facing the direction of the target building. An auditory signal terminated the imagination period and participants indicated the imagined direction (red arrow) in a sparse VR environment, followed by a confidence judgment. Performance was measured as the absolute angular difference between the correct and the indicated direction (red arc). Note that only the bottom row of images was presented to participants, top row for illustration only. (c) Circular histogram of average absolute angular difference between correct and indicated directions across participants (mean error 33.68° ± 19.09° SD).

DOI: http://dx.doi.org/10.7554/eLife.17089.003

Figure 1—figure supplement 1. Overview of behavioral training.

(a) To familiarize themselves with the controls of the computer game and the layout of the city, participants explored Donderstown for 10 min and searched for a set of landmarks irrelevant for the direction imagination task. This exploration phase was omitted in the second training session. The black circle and arrow on the map in the top panel indicate the participants’ position and orientation when first encountering Donderstown. (b) Subsequently, participants learned the names of 18 task-relevant buildings (top) to criterion. Knowledge of the building names was assessed in test blocks during which participants had to select the building belonging to the presented name from a display of three buildings by pressing one of three buttons (bottom). (c) For the remainder of the session, participants were trained on the building locations in Donderstown. Bottom row shows the trial structure as presented to the participants, top row for illustration only. Participants were instructed to navigate to the building whose name was presented on the screen. Once the building was located, participants encoded the position and were then asked to estimate the direction to the following target building. Performance was measured as the number of buildings located during the training session and the absolute angular difference between the estimated direction and the correct direction as defined by the current location and the new target building. (d) Overview of Donderstown highlighting the task-relevant buildings, which largely differed in features salient from the first-person perspective such as size, shape and rotation with respect to the hexagonal building layout (red arrows), which makes an influence of the regular arrangement of their entrances on participants’ cognitive representation of the city unlikely.

DOI: http://dx.doi.org/10.7554/eLife.17089.005

Figure 1—figure supplement 2. Sampling of directions in the imagination task.

(a–f) From each of the six start locations (yellow circles) ten directions were sampled. Directions (black arrows) were defined based on the angle of the vector connecting the start and the target locations (green circles). Screenshots show view from Donderstown corresponding to direction indicated by dashed arrow. Note that start locations could also be goal locations. The building combinations used in the direction imagination task were carefully counterbalanced so that in a task block of 24 trials each direction was sampled twice, each start building served as a start location four times and each building combination did not occur more than twice throughout the experiment (see Materials and methods). Trials sampling directions using buildings located on the same street (purple in b and c) were subject to an additional control analysis (see Materials and methods).

DOI: http://dx.doi.org/10.7554/eLife.17089.006

Figure 1—figure supplement 3. Accuracy of direction judgments during imagination task is related to behavioral performance during training and a post-scan map test.

(a) Across subjects, the accuracy of direction judgments in the direction imagination task (π minus the mean angular difference between the correct and indicated directions in radians) correlated significantly with navigation success indexed by the number of buildings found during training. (b) Accuracy of direction judgments during the training sessions also correlated highly with performance during the imagination task. (c) Additionally, performance during the imagination task was correlated with z-scored accuracy in the post-scan map test. All correlations remained significant when controlling for variability in the time spent navigating the VR city during training using partial correlations (partial correlations coefficients r>0.84, all p<0.001). Correlations were also significant when calculated between the accuracy of direction judgments and training performance measures separately for the two training sessions (all correlations coefficients r>0.80, all p<0.001).

DOI: http://dx.doi.org/10.7554/eLife.17089.007