Parker J. Banks, Allison B. Sekuler & Patrick J. Bennett. McMaster University, Department of Psychology, Neuroscience, & Behaviour. Introduction. Radiologists ...
The distinct effects of bias on sensitivity and contrast gain Parker J. Banks, Allison B. Sekuler & Patrick J. Bennett McMaster University, Department of Psychology, Neuroscience, & Behaviour Introduction Radiologists and fingerprint detectives work under extreme payoff structures, where a failure to detect illness, or a conviction based on faulty evidence, can have catastrophic consequences. These payoff structures induce changes in response bias that influence perception and classification, but their effects on perceptual learning are poorly understood [1,4]. Therefore, we investigated how differing monetary rewards and punishments interact with learning during a texture identification task.
Experiment 1
Experiment 2
Methods
• Same-different texture identification task.
A: Thresholds as a function of training day and economic adversity. B: Changes in sensitivity (d’) with training and economic adversity. C: Bias, by training and adversity. D: Bias as a function of image contrast and adversity.
• Contrast varied with method of constant stimuli. • 75% contrast thresholds, sensitivity (d’), and bias (c) were measured. • Experiment 1: 24 subjects; 4000 trials over 5 days. • Experiment 2: 16 subjects; 1920 trials over 2 days. • 3¢ won for each correct response and 3-100¢ lost for false positives and misidentifications.
E: Contrast thresholds for each day & subject, plotted against level of adversity. F: Sensitivity plotted as a function of economic adversity and training day. G: Bias as a function of economic adversity. H: Contour plot demonstrating the effects of contrast and adversity on bias.
Results and Conclusions
References
Experiment 1: Altering payoff structures significantly affected thresholds [F (2, 114) = 18.39, p < .001], bias [F (2, 114) = 357.52, p < .001], and sensitivity [F (2, 114) = 24.71, p < .001]. • Training improved contrast thresholds [F (1, 114) = 7.96, p = .006] and d’ [F (1, 114) = 20.49, p < .001]. • Rate of improvement (learning) in thresholds [F (2, 114) = .154, p = .857] and d’ [F (1, 28) = 1.35, p = .265] unaffected by adversity. Experiment 2: Thresholds degrade continuously as adversity increases [F (1, 28) = 6.58, p = .016]. • Bias [F (1, 28) = 26.17, p < .001] and d’ [F (1, 28) = 5.64, p = .025] increase with adversity. • Training improved d’ [F (1, 28) = 5.40, p = .028].
[1] Garland, H. (1949). On the scientific evaluation of diagnostic procedures. Radiology, 52, 309-328. [2] Hawkins et al. (1990). Visual attention modulates signal detectability. Journal of Experimental Psychology: Human Perception and Performance, 16, 802-811. [3] Macmillan, N., & Creelman, D. (1990). Response bias: Characteristics of detection theory, threshold theory, and "nonparametric indexes". Psychological Bulletin, 107, 401-413. [4] Norman et al. (1986). Expertise in visual diagnosis: a review of the literature. Academic Medicine, 67, 78-83.
• Extreme payoffs degrade thresholds, but increase sensitivity, suggesting increased attention [2]. • Without adversity, bias remains constant, but varies with contrast as adversity increases. • Shifts in bias may highlight distinct detection and identification strategies at low and high contrast levels [3].
Funding This work was supported by McMaster University and the Natural Sciences and Engineering Research Council of Canada.