The impact of learning on perceptual decisions and its implication for speed-accuracy tradeoffs.

In standard models of perceptual decision-making, noisy sensory evidence is considered to be the primary source of choice errors and the accumulation of evidence needed to overcome this noise gives rise to speed-accuracy tradeoffs. Here, we investigated how the history of recent choices and their outcomes interact with these processes using a combination of theory and experiment. We found that the speed and accuracy of performance of rats on olfactory decision tasks could be best explained by a Bayesian model that combines reinforcement-based learning with accumulation of uncertain sensory evidence. This model predicted the specific pattern of trial history effects that were found in the data. The results suggest that learning is a critical factor contributing to speed-accuracy tradeoffs in decision-making, and that task history effects are not simply biases but rather the signatures of an optimal learning strategy.

Learning optimal decisions with confidence.

Diffusion decision models (DDMs) are immensely successful models for decision making under uncertainty and time pressure. In the context of perceptual decision making, these models typically start with two input units, organized in a neuron-antineuron pair. In contrast, in the brain, sensory inputs are encoded through the activity of large neuronal populations. Moreover, while DDMs are wired by hand, the nervous system must learn the weights of the network through trial and error. There is currently no normative theory of learning in DDMs and therefore no theory of how decision makers could learn to make optimal decisions in this context. Here, we derive such a rule for learning a near-optimal linear combination of DDM inputs based on trial-by-trial feedback. The rule is Bayesian in the sense that it learns not only the mean of the weights but also the uncertainty around this mean in the form of a covariance matrix. In this rule, the rate of learning is proportional (respectively, inversely proportional) to confidence for incorrect (respectively, correct) decisions. Furthermore, we show that, in volatile environments, the rule predicts a bias toward repeating the same choice after correct decisions, with a bias strength that is modulated by the previous choice's difficulty. Finally, we extend our learning rule to cases for which one of the choices is more likely a priori, which provides insights into how such biases modulate the mechanisms leading to optimal decisions in diffusion models.

The mechanistic foundation of Weber's law.

Although Weber's law is the most firmly established regularity in sensation, no principled way has been identified to choose between its many proposed explanations. We investigated Weber's law by training rats to discriminate the relative intensity of sounds at the two ears at various absolute levels. These experiments revealed the existence of a psychophysical regularity, which we term time-intensity equivalence in discrimination (TIED), describing how reaction times change as a function of absolute level. The TIED enables the mathematical specification of the computational basis of Weber's law, placing strict requirements on how stimulus intensity is encoded in the stochastic activity of sensory neurons and revealing that discriminative choices must be based on bounded exact accumulation of evidence. We further demonstrate that this mechanism is not only necessary for the TIED to hold but is also sufficient to provide a virtually complete quantitative description of the behavior of the rats.

Loss of genetic redundancy in reductive genome evolution.

Biological systems evolved to be functionally robust in uncertain environments, but also highly adaptable. Such robustness is partly achieved by genetic redundancy, where the failure of a specific component through mutation or environmental challenge can be compensated by duplicate components capable of performing, to a limited extent, the same function. Highly variable environments require very robust systems. Conversely, predictable environments should not place a high selective value on robustness. Here we test this hypothesis by investigating the evolutionary dynamics of genetic redundancy in extremely reduced genomes, found mostly in intracellular parasites and endosymbionts. By combining data analysis with simulations of genome evolution we show that in the extensive gene loss suffered by reduced genomes there is a selective drive to keep the diversity of protein families while sacrificing paralogy. We show that this is not a by-product of the known drivers of genome reduction and that there is very limited convergence to a common core of families, indicating that the repertoire of protein families in reduced genomes is the result of historical contingency and niche-specific adaptations. We propose that our observations reflect a loss of genetic redundancy due to a decreased selection for robustness in a predictable environment.