Introducing Professor Philip Smith: Our CNH February Feature Story
Can you describe your research interests?
"The fundamental computational problem that must be solved by the brain and central nervous system is the translation of perception into action. To solve this problem we (our brains) must be able to answer the questions: What is out there, and what should I do about it? My research focuses on developing whole-of-system cognitive and neural models of how processes of perception, attention, memory, and decision making act in concert to solve this problem. My published work has dealt with mathematical models of attention, visual working memory, and decision making. My research on decision making has focused on diffusion process models, which assume that the brain makes decisions by accumulating successive samples of noisy sensory evidence to a decision criterion. During the last decade, I have been particularly focused on continuous-outcome decisions, in which decisions are expressed on continuous scales, rather than categorically. Many real-world, action-oriented decisions are of this kind, like the decisions we make while walking or driving. Diffusion process models successfully account for the speed and accuracy of these kinds of decisions in great detail and provide a theoretical account of how they are implemented neurally."
What inspired you to pursue this research topic?
"I was inspired to pursue this research by lectures I heard as an undergraduate at the University of Adelaide on signal detection theory. Signal detection theory explains why human decision processes are inherently noisy or variable: We don’t necessarily make the same decisions when presented with the same options on two different occasions and we don’t take the same amounts of time to make them. Signal detection theory attributes this variability to noise or statistical variability in the cognitive representations of the choice options and provides a way to measure and characterise it in behavioural experiments."
What do you like most about your work?
"Diffusion process models provide a theoretical solution to a deep problem in biological computation that I refer to as “von Neumann’s problem,” after a work by the mathematician John von Neumann in 1956, on “the synthesis of reliable organisms from unreliable components.” The components in question are the individual neurons in the brain, which are unreliable because they are noisy, in a statistical sense. Diffusion models assume that the brain’s solution to the reliability problem is the same as the statistician’s solution. The brain achieves reliability by accumulating successive samples of noisy evidence over time, which allows the effects of noise to be averaged out. This idea explains and unifies a vast array of behavioural and neural data. My lab members develop skills in designing, implementing, and running experiments with human participants and in developing and fitting mathematical models to the resulting data. We focus on so-called small-N experimental designs, in which large amounts of data are collected from a small sample of participants, each of whom serve in multiple experimental sessions. The aim of such experiments is to maximise measurement precision to allow us to test between complex models of the underlying cognitive processes."
Do you have any exciting projects or news upcoming?
"The first volume of my two-volume monograph with Roger Ratcliff was published by Cambridge University Press in November 2025: Smith, P. L., & Ratcliff, R., Diffusion Process Models of Decision Making, Volume 1: Fundamental Processes. The second volume, with Gail McKoon, is scheduled for completion in May: Ratcliff, R, Smith, P. L., & McKoon, G., Diffusion Process Models of Decision Making, Volume 2: Theory, Data, and Applications. The two volumes provide a comprehensive treatment of the theory and applications of diffusion process models in cognitive psychology and neuroscience over the last several decades."