Related Experiment Videos
Comparison of two weighted integration models for the cueing task: linear and likelihood
Steven S Shimozaki1, Miguel P Eckstein, Craig K Abbey
1Department of Psychology, University of California, Santa Barbara, CA, USA. mshimozak@psych.ucsb.edu
Journal of Vision
|May 2, 2003
Summary
Human observers integrate visual cues to improve signal detection. A sum of weighted likelihoods model, not a linear combination, best explains performance in cued discrimination tasks across varying signal-to-noise ratios.
Area of Science:
- Cognitive psychology
- Computational neuroscience
- Visual perception
Background:
- Cued discrimination tasks demonstrate improved performance with valid precues.
- Existing models explain cue validity effects via limited capacity attention or unlimited capacity weighted integration.
- Distinguishing between these models requires quantitative analysis across signal-to-noise ratios (SNR).
Purpose of the Study:
- To quantitatively compare two weighted integration models (linear vs. sum of weighted likelihoods) for explaining cue validity effects.
- To test whether human observers employ a weighted linear combination or a weighted combination of likelihoods.
- To determine the best model for describing psychophysical data in a cued discrimination task.
Main Methods:
- Three observers performed a cued discrimination task with Gaussian targets and an 80% valid precue.
- Data were collected across a wide range of signal-to-noise ratios (SNR).
- Psychophysical results were analyzed to compare model predictions against empirical data, including rejection of a limited capacity attentional switching model.
Main Results:
- Both linear and sum of weighted likelihoods models qualitatively predicted cue validity effects.
- Quantitative differences emerged at higher signal-to-noise ratios (SNR).
- The sum of weighted likelihoods model provided a significantly better fit to the observed psychophysical data.
Conclusions:
- Human observers appear to approximate a weighted combination of likelihoods rather than a simple weighted linear combination.
- This finding supports more complex Bayesian integration mechanisms in visual cueing.
- The results challenge explanations solely based on limited attentional capacity.