3-Dimensional Scene Perception during Active Electrolocation in a Weakly Electric Pulse Fish

Abstract

Weakly electric fish use active electrolocation for object detection and orientation in their environment even in complete darkness. The African mormyrid Gnathonemus petersii can detect object parameters, such as material, size, shape, and distance. Here, we tested whether individuals of this species can learn to identify 3-dimensional objects independently of the training conditions and independently of the object's position in space (rotation-invariance; size-constancy). Individual G. petersii were trained in a two-alternative forced-choice procedure to electrically discriminate between a 3-dimensional object (S+) and several alternative objects (S-). Fish were then tested whether they could identify the S+ among novel objects and whether single components of S+ were sufficient for recognition. Size-constancy was investigated by presenting the S+ together with a larger version at different distances. Rotation-invariance was tested by rotating S+ and/or S- in 3D. Our results show that electrolocating G. petersii could (1) recognize an object independently of the S- used during training. When only single components of a complex S+ were offered, recognition of S+ was more or less affected depending on which part was used. (2) Object-size was detected independently of object distance, i.e. fish showed size-constancy. (3) The majority of the fishes tested recognized their S+ even if it was rotated in space, i.e. these fishes showed rotation-invariance. (4) Object recognition was restricted to the near field around the fish and failed when objects were moved more than about 4 cm away from the animals. Our results indicate that even in complete darkness our G. petersii were capable of complex 3-dimensional scene perception using active electrolocation.

Keywords: environmental imaging distance; object feature; object recognition; rotational invariance; size constancy.

Figure 1

(A) Experimental setup: PE, position of the experimenter; EA, experimental area; LA, living area; P, mesh partition; O, objects; Gr, grids; G, gates; S, shelter; W, water plants. (B) Schematic list of tasks. Different fish were trained either with constant S+ and variable S− (left column) or constant S+ and constant S− (right columns). In the tests, S+ and/or S− were rotated, changed in their distances from the gates, or new S− were presented.

Figure 2

Choice frequency for S+ of a G. petersii trained to discriminate between a fixed S+ (‘little man’) and six differently shaped objects. Light grey bars represent choice frequency when known S− objects were used, while dark grey bars depict the choice frequency when previously unknown objects were paired with the S+. The symbols below each bar indicate the object combinations (Table 1) used (n > 264 for each combination, asterisks indicate significant choice on the 5% level as evaluated using the Chi-square test). Here and in the subsequent figures, choice frequency for S+ was determined in non-rewarded and non-punished test trials.

Figure 3

(A) Choice frequency for S+ of a G. petersii trained to discriminate between a fixed S+ (‘little man’) and six differently shaped objects, symbols of which are shown below each bar along the abscissa. The total number of choices recorded for each column was above 132. (B) Choice frequency of the same fish as in (A) when components of the former S+ were presented with other previously unseen objects. Symbols below each column show the object combination used, with the object above depicting the S+ and the object below the alternative objects used in the tests. N > 44 test trials per column.

Figure 4

Results of experiments testing for size constancy, during which fish 2 had to discriminate between a small (S+) and a large cube. (A) Choice frequency for the small cube is plotted for different distance combinations of the objects. Below each graph, symbols show the object pair, with the object on the left depicting the S+. Numbers give the distances of S+ (first number) and S− from the gates. Distances are measured from the gate towards the edge of the object closest to the gate. Dots within each graph give mean choice latencies with standard deviation (right ordinate). N > 27 test trials per column. (B) Results of control experiments done under dim light conditions (light grey column) and in complete darkness (dark grey column). Both objects were placed 3 cm from the gates. N > 50 test trials per column.

Figure 5

Tests for rotational invariance. (A) Mean choice frequencies for S+ by fish 1, which was trained and tested to discriminate between a S+ (‘little man’) and one of six different S−. Each column shows the mean results with the S+ rotated in the vertical plane parallel to the dividing wall by a certain angle (numbers give the rotation angles). N > 195 (B) Mean choice frequencies of fish 3, which was trained to discriminate between a non-rotated (0°) pyramid (S+) and a non-rotated (0°) cube, symbols of which showing the rotation angle are given below each bar along the abscissa. The symbols show the view one sees of the S+ and S− if one were to look through the gates from the living area. Both objects were rotated by various angles (numbers above symbols). Dots within each column give mean choice latencies with standard deviation (right ordinate). N > 35. (C) Results of control experiments with fish 3 done under dim light conditions (light grey columns) and in complete darkness (dark grey columns). The two columns on the left show results during which both objects were not rotated (0°), while on the right both objects were rotated by 45°. N > 50.

Figure 6

(A) Mean choice frequencies of fish 4, which was trained to discriminate between a 3-dimensional A (S+) and a mushroom (S−), symbols of which showing the rotation angle are given below each bar along the abscissa. Both objects were rotated by various angles (numbers above symbols). (B) Mean choice frequencies with rotated objects of fish 5, which was trained to discriminate between a cone (S+) and a pyramid (S−). Dots within each column give mean choice latencies (right ordinate). N ≥ 50.

Figure 7

(A) Mean choice frequencies of fish 2, which was trained to discriminate between a small (S+) and a large cube. Objects were placed at different distances from their gates (abscissa). (B) Mean choice frequencies at different distances of objects for fish 3, which was trained to discriminate between a pyramid (S+) and a cube (S−). In both graphs, sigmoid curves were fitted to the data. Dots within each column give mean choice latencies with standard deviation (right ordinate). N > 60.

Figure 8

(A) Mean choice frequencies of fish 3, which was trained to discriminate between a pyramid (S+) and a cube. Both objects were made from clay and were painted with varnish, which sealed them to water. Objects were placed at different distance from their gates (abscissa). Fish were tested either with visible lights on (light grey columns), or in complete darkness (infrared illumination; dark grey columns). Sigmoid curves were fi tted to the two data sets (solid line = dark conditions, dashed line = lights on) N > 45. (B) Same as in (A) but with clay objects, which were not painted and therefore were soaked with water. This hampered their detection through active electrolocation. Dots within each column give mean choice latencies with standard deviation (right ordinate). N > 33.