Computational Models of Brain and Behavior

Ahmed A. Moustafa

• Publisher: Wiley-Blackwell
• Year: 2017
• ISBN:
  9781119159186
• Edition: 1
• Topic: Psychologie
• Subtopic: Neurosciences cognitives et neuropsychologie

A comprehensive Introduction to the world of brain and behavior computational models

This book provides a broad collection of articles covering different aspects of computational modeling efforts in psychology and neuroscience. Specifically, it discusses models that span different brain regions (hippocampus, amygdala, basal ganglia, visual cortex), different species (humans, rats, fruit flies), and different modeling methods (neural network, Bayesian, reinforcement learning, data fitting, and Hodgkin-Huxley models, among others).

Computational Models of Brain and Behavior is divided into four sections: (a) Models of brain disorders; (b) Neural models of behavioral processes; (c) Models of neural processes, brain regions and neurotransmitters, and (d) Neural modeling approaches. It provides in-depth coverage of models of psychiatric disorders, including depression, posttraumatic stress disorder (PTSD), schizophrenia, and dyslexia; models of neurological disorders, including Alzheimer's disease, Parkinson's disease, and epilepsy; early sensory and perceptual processes; models of olfaction; higher/systems level models and low-level models; Pavlovian and instrumental conditioning; linking information theory to neurobiology; and more.

• Covers computational approximations to intellectual disability in down syndrome
• Discusses computational models of pharmacological and immunological treatment in Alzheimer's disease
• Examines neural circuit models of serotonergic system (from microcircuits to cognition)
• Educates on information theory, memory, prediction, and timing in associative learning

Computational Models of Brain and Behavior is written for advanced undergraduate, Master's and PhD-level students—as well as researchers involved in computational neuroscience modeling research.

Part 1
Models of Brain Disorders

1
A Computational Model of Dyslexics' Perceptual Difficulties as Impaired Inference of Sound Statistics

Sagi Jaffe-Dax, Ofri Raviv, Yonatan Loewenstein, and Merav Ahissar

The Hebrew University of Jerusalem, Jerusalem, Israel

Introduction—Contraction Bias in Simple Discrimination Tasks

Perception is a complex cognitive process in which noisy signals are extracted from the environment and interpreted. It is generally believed that perceptual resolution is limited by internal noise that constrains people's ability to differentiate physically similar stimuli. The magnitude of this internal noise is typically estimated using the two-alternative forced choice (2AFC) paradigm, which was introduced to eliminate participants' perceptual and response biases during experiments (Green & Swets, 1966; Macmillan & Creelman, 2004). In this paradigm, a participant is presented with two temporally separated stimuli that differ along a physical dimension and is instructed to compare them. The common assumption is that the probability of a correct response is determined by the physical difference between the two stimuli, relative to the level of internal noise. Performance is typically characterized by the threshold of discrimination, referred to as the Just Noticeable Difference (JND). Thus, the JND is a measure of the level of internal noise such that the higher the JND, the higher the inferred internal noise.

However, if the stimuli are highly predictable, perceptual resolution may not be limited by the magnitude of the internal noise. In other words, the assumption of a one-to-one correspondence between the JND and the internal noise may ignore this potential benefit that derives from previous experience. If the internal representation of a stimulus is noisy and hence unreliable, prior expectations should bias the participant against unlikely stimuli. The larger the uncertainty of the measurements, the larger the contribution of these prior expectations is likely to be. The Bayesian theory of inference defines computationally how expectations regarding the probability distribution of stimuli should be combined with the noisy representations of these stimuli in order to form an optimal posterior percept (Knill & Richards, 1996).

One limitation of the Bayesian model is that it relies heavily on the assumption that the prior distribution of stimuli is known to the observer. While this assumption may be plausible in very long experiments comprising a large number of trials (e.g., thousands in Körding & Wolpert, 2004) or in experiments utilizing natural tasks (e.g., in reading; Norris, 2006), it is unclear to what extent a rich Bayesian inference is formed when participants have less experience with a task.

Here, we studied participants' patterns of responses on a 2AFC tone discrimination task in relatively short experiments consisting of tens of trials. We found a substantial context effect, whose extent depended on the statistics of the stimuli used in the task and on participants' internal noise level. Participants' pattern of behavior was consistent with an “implicit memory” model in which the representation of previous stimuli is a single scalar that continuously updates with examples. Thus, this model can be viewed as a simple implementation of the Bayesian model that provides a better account of participants' perceptual decision making. We then applied this model to a special population of dyslexic subjects and found that this model captures their difficulties on such tasks.

Contraction Bias—a Simple Experimental Measure of Context Effects

In order to evaluate the impact of the stimulus statistics on perception parametrically, we used a 2AFC frequency discrimination task. On each trial, participants were sequentially presented with two pure tones and instructed to indicate which had a higher pitch (illustrated in Fig. 1.1). The mean frequency of each pair was uniformly selected from a broad range and the frequency difference was chosen either adaptively or according to a pretesting decision. We termed this protocol the No-Reference, since it differs from typical psychophysical assessments where one of the two stimuli on each trial serves as a reference and repeats across trials. Though frequency discrimination tasks are traditionally used as an assessment of low-level sensory bottlenecks, we have shown that performance is highly affected by context, both in the No-reference protocol (Raviv, Ahissar, & Loewenstein, 2012) and in the various reference protocols. In fact the form of integration of previous stimuli explains seemingly inconsistent biases in success rate depending on the position of the reference stimulus within the trial (Raviv, Lieder, Loewenstein, & Ahissar, 2014).

Specifically, expectations, formalized as the prior distribution of the stimuli used in the experiment, have been shown to bias participants' responses in a way that is often (though not always) consistent with the Bayesian framework (reviewed in Körding, 2007). In particular, responses in the 2AFC paradigm have been shown to be biased by prior expectations. Thus, when the magnitude of the two stimuli is small with respect to the mean of the previous stimuli used in the experiment, participants tend to respond that the second stimulus was smaller, whereas when the magnitude of both stimuli is large they tend to respond that the second stimulus was larger (Preuschhof, Schubert, Villringer, & Heekeren, 2010; Woodrow, 1933). We have shown that this bias, known as the “contraction bias,” can be understood within the Bayesian framework. Rather than comparing the two noisy representations of the stimuli, the participant combines the noisy representations of the two stimuli with the prior distribution of the stimuli to form two posterior distributions. The two posteriors are compared to maximize the probability of a correct response. The contribution of the prior distribution to the two posteriors is not equal. The larger the level of noise in the representation of the stimulus, the larger the contribution of the prior distribution to the posterior (Ashourian & Loewenstein, 2011). The level of noise in the representation of the first stimulus is larger than the level of noise in the representation of the second stimulus because of the additional noise associated with the encoding, and maintenance of the first stimulus in memory during the inter-stimulus interval of sequential presentation tasks (Bull & Cuddy, 1972; Wickergren, 1969). As a result, the posterior distribution of the first stimulus is biased more by the prior distribution than the posterior distribution of the second stimulus. Since the posterior of the first stimulus is contracted more than the posterior of the second stimulus, participants' responses are biased toward overestimating the first stimulus when it is small and underestimating it when it is large with respect to the prior (distribution of previous stimuli).

The expected outcome of the contraction bias on performance is that the combination of the mean frequency on the trial with respect to the mean frequency of the experiment, and the relative frequency of the two tones on the trial determines the impact of experiment's statistics in the following manner: Bias+ trials are trials in which the experiment's statistics is expected to improve performance. Specifically, a stronger “pulling” of the first (compared to the second) tone toward the average frequency increases the difference between the representations of the two ...

Table of contents

• Cover
• Title Page
• Table of Contents
• Notes on Contributors
• Acknowledgment
• Introduction
• Part I Models of Brain Disorders
• Part II Neural Models of Behavioral Processes
• Part III Models of Brain Regions and Neurotransmitters
• Part IV Neural Modeling Approaches
• Index
• End User License Agreement