Dissociating task acquisition from expression during learning reveals latent knowledge

Abstract

Performance on cognitive tasks during learning is used to measure knowledge, yet it remains controversial since such testing is susceptible to contextual factors. To what extent does performance during learning depend on the testing context, rather than underlying knowledge? We trained mice, rats and ferrets on a range of tasks to examine how testing context impacts the acquisition of knowledge versus its expression. We interleaved reinforced trials with probe trials in which we omitted reinforcement. Across tasks, each animal species performed remarkably better in probe trials during learning and inter-animal variability was strikingly reduced. Reinforcement feedback is thus critical for learning-related behavioral improvements but, paradoxically masks the expression of underlying knowledge. We capture these results with a network model in which learning occurs during reinforced trials while context modulates only the read-out parameters. Probing learning by omitting reinforcement thus uncovers latent knowledge and identifies context- not “smartness”- as the major source of individual variability.

Prior Experience Modifies Acquisition Trajectories Via Response-Strategy Sampling

“…Studies in mice have provided insight into the role and tuning of the auditory cortex and surrounding cortical (Chen et al, 2019) and striatal (Chen et al, 2019;Xiong et al, 2015;Znamenskiy & Zador, 2013) structures, as well as deeper structures such as the basal forebrain (Hangya et al, 2015) and thalamus (Chen et al, 2019). Such research has yielded useful neurobehavioral computational models largely targeted at steady-state behavior (Hangya et al, 2015;Kuchibhotla et al, 2019).…”

“…Additionally, Kurt and Ehret (2010) showed that although amplitude discriminations are more difficult to acquire than frequency discriminations, amplitude discrimination training facilitates acquisition of frequency discriminations. More recently, Kuchibhotla et al (2019) showed that evidence of acquisition appears sooner in a mass of signaled non-reinforced probe trials than in training trials alone, suggesting either that trial spacing affects acquisition expression (Barnes, 1957;Kalmbach et al, 2019;Sunsay et al, 2004) or competition between the reinforcer and task-relevant stimuli for behavioral control interferes with acquisition expression. More recently, while (Xiong et al, 2015) revealed a role for corticostriatal plasticity in acquisition, (Hangya et al, 2015) demonstrated a role for basal forebrain cholinergic neurons projecting to the cortex in signaling and communicating both reward and punishment.…”

“…In principle, performance in 2AFC tasks can be described by popular associative or reinforcement learning theories (Daniels & Sanabria, 2018;Dayan & Daw, 2008;Kuchibhotla et al, 2019;Mill et al, 2014;Sewell et al, 2019). These models assume that in any 2AFC signal detection task, individuals seek to maximize outcomes by making choices with the greatest value and do so by learning the mapping and values between stimuli (tone plus white noise, white noise) responses (left lever, right lever), and outcomes (reward, no-reward).…”

“…Prior experience could affect acquisition by shortening or altering the composition of the pre-solution period; alternatively, it could affect acquisition by inducing higher learning rates across the pre-solution and solution period. Practically, prior experience induced facilitation of acquisition should cut down on the time it takes to train subjects in a signal detection task, which can take as much as 3000-6000 trials (e.g., Heinemann & Chase, 1970;Kuchibhotla et al, 2019).…”

Touchscreen response technology and the power of stimulus‐based approaches in freely behaving animals

“…Indeed, learning paradigms in freely moving and behaving animals such as those utilizing touchscreen‐based operant systems simulate more naturalistic foraging behavior with the addition of an appreciable level of experimental control, and while animals have the option to engage in the task (or not), as in the real world. Freely moving and body/head‐fixed behavioral methods may reveal the same underlying process, 1 but there are also important and potentially growing examples of divergence 2,3 …”