Using Diffusion Models for Symbolic Numeracy Tasks to Examine Aging Effects.

We present a model-based analysis of aging effects in three symbolic numeracy tasks using three groups of subjects (young adults, 60- to 69-year-olds, and 70- to 90-year-olds). The tasks are number discrimination (is this number greater or less than 50), number memory (was this number in the list of numbers just displayed), and number line (point to where this number is on this number line). The first two tasks were fit by the standard two-choice diffusion model and the last one by the spatially continuous diffusion model (Ratcliff, 2018). Results showed good fits of the models to accuracy (choices) and response time distributions. In the tasks, nondecision time (the time to encode a stimulus and make a response) increased with age, but the amount of evidence needed for a decision (boundary settings) increased in the number discrimination and number memory tasks, but not the number line task. The number discrimination task produced conflicting accuracy and response time results as a function of age, but the model-based analyses resolved these differences. In the number memory task, drift rates (evidence used to drive the decision process) were lower for the older adults than for young adults, but for the other two (easier) tasks, there was no change in drift rate with age. The analyses extracted differences among individuals in model components, some of which were systematic across tasks. In particular, drift rates were correlated across tasks, which shows consistent individual differences across tasks, results that could not have been obtained without model-based analyses.