Abstract
Perceptual sensitivity is usually measured either as a contrast threshold at a particular level of stimulus discriminability (e.g., Gabors of 45°±15°) or as a threshold feature difference for discrimination at a fixed stimulus contrast. The existing observer models assume that any single stimulus activates only one perceptual template. For example, in the perceptual template model (PTM), the observer is characterized by four parameters: a single gain to the signal stimulus β, exponent of the non-linear transducer function g, internal additive noise Nadd, and coefficient of the multiplicative internal noise Nmul (Lu & Dosher, 1999). These models are only applicable to contrast threshold tasks. Here, we attempted to elaborate the PTM to model both contrast and feature thresholds. Using the method of constant stimuli, full contrast psychometric functions of three observers were measured in an orientation identification task at fovea with four orientation differences (±3°, ±6°, ±15°, and ±45° from vertical) across a wide range of external noise levels. Threshold versus external noise contrast functions at three performance criterion levels (65, 75, and 85% correct) were estimated in each orientation condition. In such two-alternative identification tasks, the performance of an observer can be modeled by gains of two templates; the gain of the better-matching template β and the gain of the less-well-matching template β′. The original PTM was developed for the cases where β′ = 0. The simplest elaboration of the PTM model, with the same β, γ, Nadd, and Nmul but varying β′0 across all the orientation conditions, provided the best fit to all the data, accounting for 96–99% of the variance. The elaborated PTM provides a theoretical framework to characterize joint feature and contrast sensitivity of human observers.