Signal detection model vs. Guided search model

In the last decade, models of visual search have moved away from the original parallel/serial dichotomy (Treisman & Gelade, 1980).  Two prominent models are the Signal Detection Theory (SDT) model (Swensson & Judy, 1981; Shaw, 1982; Palmer, 1994; Eckstein, 1995) and Wolfe’s Guided Search model.  In the SDT model, set-sizes are caused by the variability in the internal response to the stimuli (e.g. noise in the firing of cells). Noise will degrade performance as set-size increases owing to the increasing probability that the response to any distractor exceed the target response.  In the Guided Search (GS) model a parallel stage guides a subsequent serial stage. In the GS model, set-size effects are caused by the serial allocation of visual attention. The two models are fundamentally different, however, progress towards achieving a consensus in the field has been hampered by the difficulty associated with directly comparing these two models.  Firstly, the two models have been developed for different experimental paradigms: SDT models for search accuracy and the GS model for response time studies. Secondly, current implementations of the GS model generate predictions with the use of computationally time-consuming Monte Carlo simulations, a method that precludes fitting the model quantitatively to human data.  Recently, we have developed a mathematical implementation of Wolfe’s Guided Search model to predict visual search accuracy experiments that will allow comparisons between the SDT model and the GS model (ECVP, 1998; NASA Tech. Memoranda, 1999). 
For the localization of a target in one of M locations we have shown that both the SDT model and GS model can be used to accurately predict visual search performance across set-size conditions and target saliency conditions. 
Sixteen data points were fit with 4 fitting parameters for the GS model and 1 fitting parameter for the SDT model. We are currently investigating other model predictions to enable us to distinguish between both models. 

Publications:

Eckstein, M.P., Beutter B.R., Stone L.S., Analytic Guided-Search Model of human performance accuracy in target localization search tasks, NASA/TM 209-594 (2000)

Eckstein, M. P., Beutter B.R., Bartroff, L., Stone, L.S., Guided search vs. signal detection theory in target localization tasks, Investigative Ophthalmology & Visual Science (Suppl.), 40, 1836, (1999)

Eckstein, M.P., Beutter, B.R., Analytic implementation of the guided search model allows comparisons to signal detection models for accuracy in visual localization tasks, Perception (Suppl.), 27, (1998)