When we make faster decisions, we make more mistakes. When we make slower decisions, we miss more deadlines and we limit the number of decisions we can make. These principles are intuitively obvious and are applicable to decisions in any domain, on any timescale, by any species or automated system. Their resolution defines the speed-accuracy trade-off (SAT). The SAT has long been the subject of experimental and theoretical enquiry. In experiments, subjects make slower, more accurate decisions when motivated to favour accuracy and make faster, less accurate decisions when motivated to favour speed. Computationally, the SAT is well characterized within the framework of bounded integration, where noisy evidence for the alternatives is integrated over time. When the evidence for one of the alternatives reaches a threshold or bound, a choice is made for that alternative. Higher thresholds therefore favour accuracy at the expense of speed. This framework captures a remarkable volume of experimental data, but there is a prominent discrepancy between model fits to behavioural and electrophysiological data. The latter have been taken to suggest that decision thresholds are fixed. If so, then how do we trade speed and accuracy? This question is the focus of intense research interest.
A growing number of studies have investigated the neural mechanisms underlying the SAT within the framework of bounded integration, where a convergence of neuroimaging and electrophysiological methods with mathematical and biophysical modelling has provided new perspectives on the mechanisms by which decision times are determined. For example, with a fixed decision threshold, the time spent integrating evidence may be adjusted by modulating the baseline activity of neural integrators, the rate and onset of integration, or the functional connectivity between integrators and other sources of input to thresholding circuitry. Furthermore, there is increasing evidence that the encoding of elapsed time plays a crucial role in the SAT, sometimes referred to as urgency. These and other hypotheses suggest that the conflicting demands of speed and accuracy may be resolved by the differential encoding, readout and integration of evidence. Striking the optimal balance between speed and accuracy in a given context further requires a means to control these mechanisms. In this Research Topic, we welcome articles that characterize or explain the SAT and its optimization according to any experimental factor or neural mechanism, using any experimental or theoretical methodology. While we take temporal integration as a starting point, we encourage articles expressing disagreement with the premises of the bounded integration framework and reporting evidence in favour of alternative explanations of the SAT. All Frontiers article types are welcome, including original research articles, methods articles, hypothesis and theory articles, opinions, perspectives and reviews.
When we make faster decisions, we make more mistakes. When we make slower decisions, we miss more deadlines and we limit the number of decisions we can make. These principles are intuitively obvious and are applicable to decisions in any domain, on any timescale, by any species or automated system. Their resolution defines the speed-accuracy trade-off (SAT). The SAT has long been the subject of experimental and theoretical enquiry. In experiments, subjects make slower, more accurate decisions when motivated to favour accuracy and make faster, less accurate decisions when motivated to favour speed. Computationally, the SAT is well characterized within the framework of bounded integration, where noisy evidence for the alternatives is integrated over time. When the evidence for one of the alternatives reaches a threshold or bound, a choice is made for that alternative. Higher thresholds therefore favour accuracy at the expense of speed. This framework captures a remarkable volume of experimental data, but there is a prominent discrepancy between model fits to behavioural and electrophysiological data. The latter have been taken to suggest that decision thresholds are fixed. If so, then how do we trade speed and accuracy? This question is the focus of intense research interest.
A growing number of studies have investigated the neural mechanisms underlying the SAT within the framework of bounded integration, where a convergence of neuroimaging and electrophysiological methods with mathematical and biophysical modelling has provided new perspectives on the mechanisms by which decision times are determined. For example, with a fixed decision threshold, the time spent integrating evidence may be adjusted by modulating the baseline activity of neural integrators, the rate and onset of integration, or the functional connectivity between integrators and other sources of input to thresholding circuitry. Furthermore, there is increasing evidence that the encoding of elapsed time plays a crucial role in the SAT, sometimes referred to as urgency. These and other hypotheses suggest that the conflicting demands of speed and accuracy may be resolved by the differential encoding, readout and integration of evidence. Striking the optimal balance between speed and accuracy in a given context further requires a means to control these mechanisms. In this Research Topic, we welcome articles that characterize or explain the SAT and its optimization according to any experimental factor or neural mechanism, using any experimental or theoretical methodology. While we take temporal integration as a starting point, we encourage articles expressing disagreement with the premises of the bounded integration framework and reporting evidence in favour of alternative explanations of the SAT. All Frontiers article types are welcome, including original research articles, methods articles, hypothesis and theory articles, opinions, perspectives and reviews.
