An official website of the United States government
Structures built by animals, such as nests, often can be considered extended phenotypes that facilitate the study of animal behaviour. For rodents, nest building is both an important form of behavioural thermoregulation and a critical component of parental care. Changes in nest structure or the prioritization of nesting behaviour are therefore likely to have consequences for survival and reproduction, and both biotic and abiotic environmental factors are likely to influence the adaptive value of such differences. Here we first develop a novel assay to investigate interspecific variation in the nesting behaviour of deer mice (genus Peromyscus). Using this assay, we find that, while there is some variation in the complexity of the nests built by Peromyscus mice, differences in the latency to begin nest construction are more striking. Four of the seven taxa examined here build nests within an hour of being given nesting material, but this latency to nest is not related to ultimate differences in nest structure, suggesting that the ability to nest is relatively conserved within the genus, but species differ in their prioritization of nesting behaviour. We also find that latency to nest is not correlated with body size, climate or the construction of burrows that create microclimates. However, the four taxa with short nesting latencies all have monogamous mating systems, suggesting that differences in nesting latency may be related to social environment. This detailed characterization of nesting behaviour within the genus provides an important foundation for future studies of the genetic and neurobiological mechanisms that contribute to the evolution of behaviour.
more »
« less
Formalising the role of behaviour in neuroscience
Abstract We develop a mathematical approach to formally proving that certain neural computations and representations exist based on patterns observed in an organism's behaviour. To illustrate, we provide a simple set of conditions under which an ant's ability to determine how far it is from its nest would logically imply neural structures isomorphic to the natural numbers . We generalise these results to arbitrary behaviours and representations and show what mathematical characterisation of neural computation and representation is simplest while being maximally predictive of behaviour. We develop this framework in detail using a path integration example, where an organism's ability to search for its nest in the correct location implies representational structures isomorphic to two‐dimensional coordinates under addition. We also study a system for processing strings common in comparative work. Our approach provides an objective way to determine what theory of a physical system is best, addressing a fundamental challenge in neuroscientific inference. These results motivate considering which neurobiological structures have the requisite formal structure and are otherwise physically plausible given relevant physical considerations such as generalisability, information density, thermodynamic stability and energetic cost.
more »
« less
- Award ID(s):
- 2201843
- PAR ID:
- 10513940
- Publisher / Repository:
- Wiley-Blackwell
- Date Published:
- Journal Name:
- European Journal of Neuroscience
- Volume:
- 60
- Issue:
- 5
- ISSN:
- 0953-816X
- Format(s):
- Medium: X Size: p. 4756-4770
- Size(s):
- p. 4756-4770
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
[This paper is part of the Focused Collection in Investigating and Improving Quantum Education through Research.] The ability to relate physical concepts and phenomena to multiple mathematical representations—and to move fluidly between these representations—is a critical outcome expected of physics instruction. In upper-division quantum mechanics, students must work with multiple symbolic notations, including some that they have not previously encountered. Thus, developing the ability to generate and translate expressions in these notations is of great importance, and the extent to which students can relate these expressions to physical quantities and phenomena is crucial to understand. To investigate student understanding of the expressions used in these notations and the ways they relate, clinical think-aloud interviews were conducted with students enrolled in an upper-division quantum mechanics course. Analysis of these interviews used the symbolic forms framework to determine the ways that participants interpret and reason about these expressions. Multiple symbolic forms—internalized connections between symbolic templates and their conceptual interpretations—were identified in both Dirac and wave function notations, suggesting that students develop an understanding of expressions for probability both in terms of their constituent pieces and as larger composite expressions.more » « less
-
Abstract Behavior is shaped by genes, environment, and evolutionary history in different ways. Nest architecture is an extended phenotype that results from the interaction between the behavior of animals and their environment. Nests built by ants are extended phenotypes that differ in structure among species and among colonies within a species, but the source of these differences remains an open question. To investigate the impact of colony identity (genetics), evolutionary history (species), and the environment on nest architecture, we compared how two species of harvester ants, Pogonomyrmex californicus and Veromessor andrei, construct their nests under different environmental conditions. For each species, we allowed workers from four colonies to excavate nests in environments that differed in temperature and humidity for seven days. We then created casts of each nest to compare nest structures among colonies, between species, and across environmental conditions. We found differences in nest structure among colonies of the same species and between species. Interestingly, however, environmental conditions did not have a strong influence on nest structure in either species. Our results suggest that extended phenotypes are shaped more strongly by internal factors, such as genes and evolutionary history, and are less plastic in response to the abiotic environment, like many physical and physiological phenotypes.more » « less
-
Precise and reliable prediction of soft and structured materials’ behavior under flowing conditions is of great interest to academics and industrial researchers alike. The classical route to achieving this goal is to construct constitutive relations that, through simplifying assumptions, approximate the time- and rate-dependent stress response of a complex fluid to an imposed deformation. The parameters of these simplified models are then identified by suitable rheological testing. The accuracy of each model is limited by the assumptions made in its construction, and, to a lesser extent, the ability to determine numerical values of parameters from the experimental data. In this work, we leverage advances in machine learning methodologies to construct rheology-informed graph neural networks (RhiGNets) that are capable of learning the hidden rheology of a complex fluid through a limited number of experiments. A multifidelity approach is then taken to combine limited additional experimental data with the RhiGNet predictions to develop “digital rheometers” that can be used in place of a physical instrument.more » « less
-
One of the central elements of any causal inference is an object called structural causal model (SCM), which represents a collection of mechanisms and exogenous sources of random variation of the system under investigation (Pearl, 2000). An important property of many kinds of neural networks is universal approximability: the ability to approximate any function to arbitrary precision. Given this property, one may be tempted to surmise that a collection of neural nets is capable of learning any SCM by training on data generated by that SCM. In this paper, we show this is not the case by disentangling the notions of expressivity and learnability. Specifically, we show that the causal hierarchy theorem (Thm. 1, Bareinboim et al., 2020), which describes the limits of what can be learned from data, still holds for neural models. For instance, an arbitrarily complex and expressive neural net is unable to predict the effects of interventions given observational data alone. Given this result, we introduce a special type of SCM called a neural causal model (NCM), and formalize a new type of inductive bias to encode structural constraints necessary for performing causal inferences. Building on this new class of models, we focus on solving two canonical tasks found in the literature known as causal identification and estimation. Leveraging the neural toolbox, we develop an algorithm that is both sufficient and necessary to determine whether a causal effect can be learned from data (i.e., causal identifiability); it then estimates the effect whenever identifiability holds (causal estimation). Simulations corroborate the proposed approach.more » « less
