Spatial Learning by Using Non-Visual Geometry and a Visual 3D Landmark in Zebrafish (Danio rerio)

Abstract

Fish conjoin environmental geometry with conspicuous landmarks to reorient towards foraging sites and social stimuli. Zebrafish (Danio rerio) can merge a rectangular opaque arena with a 2D landmark (a blue-colored wall) but cannot merge a rectangular transparent arena with a 3D landmark (a blue cylinder) without training to "feel" the environment thanks to other-than-sight pathways. Thus, their success is linked to tasks differences (spontaneous vs. rewarded). This study explored the reorientation behavior of zebrafish within a rectangular transparent arena, with a blue cylinder outside, proximal to/distal from a target corner position, on the short/long side of the arena. Adult males were extensively trained to distinguish the correct corner from the rotational one, sharing an equivalent metric-sense relationship (short surface left, long surface right), to access food and companions. Results showed that zebrafish's reorientation behavior was driven by both the non-visual geometry and the visual landmark, partially depending on the landmark's proximity and surface length. Better accuracy was attained when the landmark was proximal to the target corner. When long-term experience was allowed, zebrafish handled non-visual and visual sensory stimulations over time for reorienting. We advance the possibility that multisensory processes affect fish's reorientation behavior and spatial learning, providing a link through which to investigate animals' exploratory strategies to face situations of visual deprivation or impairments.

Keywords: 3D landmark; environmental geometry; geometric module; multisensory; reorientation; sensory channels; spatial cognition; transparency; zebrafish.

Figure 1

Experimental apparatus. (a) Setup from the outside. (b) Rectangular transparent arena with upright corridors at the corners. The transparent cylinder hosting the experimental fish at the beginning of each training trial is also shown (credit by Sara Boffelli).

Figure 2

Experimental conditions. Proximal–Short (PS), top left; Proximal–Long (PL), top right; Distal–Short (DS), bottom left; Distal–Long (DL), bottom right (3D virtual graphic reconstruction by Sara Boffelli).

Figure 3

Individual learning curves in the four experimental conditions (PS, PL, DS, DL): total choices. In PS (a), 4/4 fish met the 70% accuracy criterion. In PL (b), 3/4 fish met the criterion. In DS (c), 1/4 fish met the criterion. In DL (d), 0/4 met the criterion. The dotted line indicates the 70% threshold.

Figure 4

Total choices for the corners (geometry and landmark) in PS, PL, and DS collapsed. Mean ± SEM are shown. Significant p values are indicated with asterisks (*** p < 0.001).

Figure 5

Total choices for the diagonals (geometry alone) in the 8 fish achieving the learning criterion ≥ 70% for C (a) and in the 8 fish not achieving it (b). Mean ± SEM are shown. Significant p values are indicated with asterisks (** p = 0.001).

Figure 6

Latency times before leaving the arena through a correct choice, in the four experimental conditions (PS, PL, DS, DL) separately.

Figure 7

Motion patterns by collapsed conditions (PS, PL, DS, DL). (a) Motion strategy: wall-following vs. center-to-corner. (b) Motion direction: left vs. right. Mean ± SEM are shown. Significant p values are indicated with asterisks (*** p < 0.001).