Mantis shrimp rank the shape of an object over its color during recognition

Mantis shrimp are predatory crustaceans that commonly occupy burrows in shallow, tropical waters worldwide. Most of these animals inhabit structurally complex, benthic environments with an abundance of visual features that are regularly observed, including conspecifics, predators, prey, and landmarks for use in navigation. While these animals are capable of learning and discriminating color and polarization, it is unknown what specific attributes of a visual object are important for its recognition. Here we show that mantis shrimp of the species Neogonodactylus oerstedii can learn the shape of a trained target. Furthermore, when the shape and color of a target which they had been trained to identify were placed in conflict, N. oerstedii significantly chose the target of the trained shape over the target of the trained color. Thus, we conclude that the shape of a target is more important than its color for its recognition by N. oerstedii. Our findings suggest that the shapes of learned structures, such as landmarks or other animals, are important for N. oerstedii during object recognition.


28
Each species of animal living in a given space experiences its own distinct sensory world, 29 known as its "umwelt" (von Uexküll, 1957Uexküll, /1934). The sensory structures responsible for an 30 animal's perception of its environment are metabolically taxing tissues that are often under 31 strong selection pressures to permit the recognition of biologically relevant stimuli, while 32 ignoring much of the available information an environment has to offer. Despite their 33 complexity, the visual systems of stomatopod crustaceans should follow this generalization. 34 Better known as mantis shrimp, these animals are renowned for their visual systems which in 35 most species enable spatial and motion vision, color and multispectral UV vision, and linear and Patel 2 circular polarization receptivity (Cronin et al., 2014a). The compound eyes of many stomatopod 37 species have a relatively high visual acuity; for instance, Gonodactylus chiragra, an animal 38 typically about 8 cm in length, achieves a resolution of 0.8 cycles/degree (Marshall and Land,39 1993). The ability of stomatopods to learn novel visual stimuli has been previously demonstrated 40 with color, linear polarization, and circular polarization cues (Marshall et al., 1996;Marshall et   Since color may be informative in many aspects of a mantis shrimp's life and since these 58 animals use landmarks for navigation when available, this raises the question of what makes an 59 object salient to a mantis shrimp for recognizing it. Considering that mantis shrimp have 60 reasonably acute visual systems and are known to possess color vision, we were interested in 61 determining whether N. oerstedii learns to recognize a visual target using its shape and/or its shape combination (either a red rectangle, red triangle, green rectangle, or green triangle) using a 68 paired food reward in a dichotomous choice y-maze (Fig 1). Since stomatopods in previous 69 behavioral experiments successfully learned to discriminate red and green colored targets 70 (Marshall et al., 1996), targets of these colors were chosen for the current study. The target of the 71 alternate color and shape of the trained target was placed in the other arm of the y-maze and was 72 associated with no reward (for example, a rewarded red triangle was paired with an unrewarded 73 green rectangle). Animals on average responded in this situation (i.e. made a choice) 74 approximately half of the time ( Fig. 2A). From these choices, animals learned to associate food 75 with their respective trained targets over time (P < 0.05; Fig. 2B). Of the 78 stomatopods that Patel 3 were trained, 20 individuals reached the criteria set to progress to the testing procedure (see the 77 Methods section for the criteria).

79
Neogonodactylus oerstedii recognized the trained target by its shape, not its color 80 Once animals reached the performance criteria to enter the testing phase, they were tested 81 in three separate procedures: a shape recognition test, a color recognition test, and a conflicting 82 cue test. During all testing procedures, food was not paired with a target.

83
During the shape recognition test, both arms contained targets of the color an animal had 84 been trained to, but the target in each arm was of a different shape (e.g. a red triangle paired with 85 a red rectangle). In this experiment, individuals significantly chose the arm with the shape to 86 which they had been trained, indicating that they recognized the shape of their trained target (P < 87 0.05, Z = 1.976, n = 19; Fig. 3).

88
During the color recognition test, both arms contained targets of the shape they were 89 trained to but the color of the target differed per arm (e.g. a red triangle paired with a green 90 triangle). During this task, stomatopods more often than not chose the arm with the same color 91 target that they were trained to; however, this relationship was not significantly different from a 92 random choice distribution (P > 0.1, Z = 0.934, n = 21; Fig 3).

93
During the conflicting cue test, one arm contained a target with the same shape but 94 opposite color to the target to which they were trained while the other arm had a target with the 95 same color but alternate shape to the trained target (e.g. a green triangle paired with a red  Our study is the first to demonstrate that mantis shrimp are able to recognize distinct shapes. 103 We found that mantis shrimp ranked the shape of an object higher than its color when 104 recognizing it (Fig. 3). Since mantis shrimp use landmarks during navigation (Patel and Cronin,105 2020c), the findings in our study suggest that the shape of a landmark may be more important 106 than its color when being identified by a mantis shrimp during navigation. Similarly, the shapes 107 of prey, predators, and of body structures used in signaling may be critical for recognition and 108 for generating appropriate behavioral responses independent of their roles in navigation.

109
Identifying an object by its shape might be more effective than recognizing its color when   Reflectance measurements of the colored targets were taken in a dark room using an 195 Ocean Optics USB2000 spectrometer connected to a 3 m long, 400 µm diameter, fiber-optic 196 cable. Reflectances were measured from 300 to 700 nm relative to a "Spectralon" white standard 197 using a PX-2 pulsed xenon light source.  Of the 78 animals that were trained, a total of 20 animals achieved the training criterion 223 and moved on to the testing phase. The procedure of the testing phases was identical to that of 224 the training phase except that no food reward was offered during testing sessions. Trained 225 stomatopods were subjected to three distinct tests: a shape recognition test, a color recognition 226 test, and a conflicting cues test (Fig. 3). Individuals experienced these tests in a randomized 227 order. Two training sessions were administered between tests to facilitate reward seeking 228 between tests. 229

231
In order to test if N. oerstedii could distinguish the shape of the trained target, the cue of   Generalized linear mixed modelling was used to analyze the data for each of the three 265 tests. Our models used animal choices during testing as the variable of interest, specifying a 266 binomial error distribution (link function "logit"). Since individual stomatopods were tested more 267 than once, the models for each test included individual ID as a random term. Since we used both 268 males and females for our study, sex was also included as a random term for our full models; 269 however, since sex did not significantly increase the explanatory power of our models, it was 270 removed from our final models. Individual ID did not significantly increase the explanatory 271 power of our models, but was left in the final models to account for repeated measures.