Automated measurement of long-term bower behaviors in Lake Malawi cichlids using depth sensing and action recognition

Abstract

20Measuring naturalistic behaviors in laboratory settings is difficult, and this hinders progress 21in understanding decision-making in response to ecologically-relevant stimuli. In the wild, 22many animals manipulate their environment to create architectural constructions, which 23 represent a type of extended phenotype affecting survival and/or reproduction, and these 24behaviors are excellent models of goal-directed decision-making. Here, we describe an 25 automated system for measuring bower construction in Lake Malawi cichlid fishes, whereby 26 males construct sand structures to attract mates through the accumulated actions of 27 thousands of individual sand manipulation decisions over the course of many days. The 28 system integrates two orthogonal methods, depth sensing and action recognition, to 29 simultaneously measure the developing bower structure and classify the sand manipulation 30 decisions through which it is constructed. We show that action recognition accurately (>85%) 31 classifies ten sand manipulation behaviors across three different species and distinguishes 32 between scooping and spitting events that occur during bower construction versus 33feeding. Registration of depth and video data streams enables topographical mapping of 34 these behaviors onto a dynamic 3D sand surface. The hardware required for this setup is 35inexpensive (<$250 per setup), allowing for the simultaneous recording from many 36 independent aquariums. We further show that bower construction behaviors are non-uniform 37in time, non-uniform in space, and spatially repeatable across trials. We also quantify a unique 38 behavioral phenotype in interspecies hybrids, wherein males sequentially express both 39 phenotypes of behaviorally-divergent parental species. Our work demonstrates that 40 simultaneously tracking both structure and behavior provides an integrated picture of long-41 term goal-directed decision-making in a naturalistic, dynamic, and social environment. 42 45 many hours to weeks and are critical for survival and reproduction in a wide range of invertebrate 46and vertebrate species (Tucker 1981, Feng, Fergus et al. 2015, Russell, Morrison et al. 2017 Mouritsen 2018). These behaviors may be expressed inflexibly according to fixed sets of rules, 48or plastically in response to changing environmental and social stimuli. Understanding the 49 underlying logic of long-term behaviors and how they are encoded in the genome and the nervous 50 system will require accurately measuring them as they unfold over extended periods of time in 51 complex, naturalistic, dynamic, and often social environments. 52 53Long-term natural behaviors are also often goal-directed, in which animals integrate external 54 stimuli, internal physiology, and previous experience to coordinate decisions and actions towards 55 a specific goal. For example, many species exhibit construction behaviors in which they 56 manipulate the environment to build extended phenotype structures such as burrows, dens, 57 tunnels, webs, nests, or bowers; and th...