Abstract

Signal detection theory (SDT) provides a framework for interpreting psychophysical experiments, separating the putative internal sensory representation and the decision process. SDT was used to analyse ferret behavioural responses in a (yes–no) tone-in-noise detection task. Instead of measuring the receiver-operating characteristic (ROC), we tested SDT by comparing responses collected using two common psychophysical data collection methods. These (Constant Stimuli, Limits) differ in the set of signal levels presented within and across behavioural sessions. The results support the use of SDT as a method of analysis: SDT sensory component was unchanged between the two methods, even though decisions depended on the stimuli presented within a behavioural session. Decision criterion varied trial-by-trial: a ‘yes’ response was more likely after a correct rejection trial than a hit trial. Simulation using an SDT model with several decision components reproduced the experimental observations accurately, leaving only ∼10% of the variance unaccounted for. The model also showed that trial-by-trial dependencies were unlikely to influence measured psychometric functions or thresholds. An additional model component suggested that inattention did not contribute substantially. Further analysis showed that ferrets were changing their decision criteria, almost optimally, to maximise the reward obtained in a session. The data suggest trial-by-trial reward-driven optimization of the decision process. Understanding the factors determining behavioural responses is important for correlating neural activity and behaviour. SDT provides a good account of animal psychoacoustics, and can be validated using standard psychophysical methods and computer simulations, without recourse to ROC measurements.

Highlights

  • Signal detection theory (SDT) is a quantitative framework to analyse subjects’ responses in psychophysical tasks (Green and Swets, 1966)

  • It assumes that a subject’s ability to distinguish between different stimuli is limited by the variability of an internal sensory representation, and that behavioural responses are based on a comparison between the internal representation and a criterion value

  • Consistent with this, psychometric functions, expressed as the criterion-free measure P(c)max, were not significantly different whether obtained with the Methods of Limits or Constant Stimuli (Fig. 2; p > 0.05; one-way MANOVA applied to P(c)max at different signal-to-noise ratio (SNR)), except for one ferret (F3)

Read more

Summary

Introduction

Signal detection theory (SDT) is a quantitative framework to analyse subjects’ responses in psychophysical tasks (Green and Swets, 1966) It assumes that a subject’s ability to distinguish between different stimuli is limited by the variability of an (unspecified) internal sensory representation, and that behavioural responses are based on a comparison between the internal representation and a criterion value. Using certain assumptions sensory acuity (d0) can be derived, filtering out the influence of the cognitive decision process. Those assumptions underlying SDT have been tested in a wide range of psychophysical (and memory) experiments, for human subjects, and generally hold (e.g. classically in hearing, Tanner et al, 1956). Measured d0 can depend on the number of intervals and references in a forced choice task (Creelman and Macmillan, 1979) or differ between detecting change and discriminating the direction of change (Semal and Demany, 2006), and the appropriate sensory model is still debated (Kaernbach, 1991; Pastore et al, 2003; Micheyl et al, 2008)

Methods
Results
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call