Contents lists available at ScienceDirect Intelligence journal homepage: www.elsevier.com/locate/intell Curvilinear dependency of response accuracy on response time in cognitive tests Haiqin Chen a, ⁎ , Paul De Boeck b , Matthew Grady a , Chien-Lin Yang a , David Waldschmidt a a American Dental Association, United States b KU Leuven and Ohio State University, United States ABSTRACT The relationship between response time and accuracy in cognitive tasks is an important topic in experimental cognitive psychology as well as in the domain of cognitive testing, but the relationship is much more difficult to capture for the latter. Using data involving five cognitive tests: three academic achievement tests (knowledge tests) and two reasoning tests (perceptual and quantitative reasoning), the relationship between response time and response accuracy is explored after controlling for possible confounds associated with individual and item differences. The tests are different in terms of contents and type of test (achievement or ability test), but it was nonetheless found for all tests that response accuracy shows the same kind of curvilinear dependency on re- sponse time. Accuracy rates first increase rather rapidly and then decrease more slowly as a function of response time. The turning point came earlier for the three knowledge tests than for the two ability tests. The results are not easily reconcilable with simple principles that may apply to tasks used in cognitive experimental psychology. Possible explanations refer to discontinuities in the cognitive processes such as switching strategies, or a decline of cognitive efficiency and increasing cognitive depletion with passing time while working on problems that take much more time than the common tasks in experimental psychology. 1. Introduction The relationship between response time and response accuracy re- presents an important area of study within the cognitive performance literature. This relationship has been investigated primarily in experi- mental cognitive psychology, with rather simple cognitive tasks (Heitz, 2014; Meyer, Irwin, Osman, & Kounios, 1988; Ratcliff, Smith, Brown, & McKoon, 2016; Wickelgren, 1977). Two major principles appear to play a role in cognitive experimental studies: (a) cognitive capacity or the efficiency of information processing (Wenger & Gibson, 2004; Wenger & Townsend, 2000), and (b) the speed-accuracy tradeoff (SAT) (Donkin, Little, & Houpt, 2014; Ratcliff et al., 2016). Following the first prin- ciple, as the efficiency of information processing increases, the response times will tend to decrease, and the responses will tend to be more accurate. The available capacity can be manipulated (reduced) by in- creasing cognitive load through extra tasks or more demanding tasks (e.g., Donkin et al., 2014). Following the second principle, caring more about accuracy increases the accuracy rate at the cost of a larger re- sponse time, and caring more about speed makes one respond faster at the cost of the accuracy rate (Heitz, 2014). In short, the SAT can be manipulated through incentives or response time limits. The diffusion model is a successful model that captures the two principles for binary decision tasks with very short response times and homogeneous processing (Ratcliff et al., 2016). The efficiency or speed of information accumulation is captured through the drift rate para- meter that reflects the steepness of the drift toward a decision threshold A vs. B, whereas the SAT is captured through the boundary separation parameter, which is the distance between the two thresholds, one for the correct response and the other for the incorrect response. The larger the boundary separation is, the larger the time to reach one of the two boundaries and the higher the probability that the response is correct if the drift rate is positive. More generally and not limited to the diffusion model, the two principles affect the relationship between response time and response accuracy differently. A higher level of cognitive efficiency leads to shorter response times and higher accuracy (i.e., higher speed and ac- curacy), and more emphasis on accuracy as opposed to speed leads to longer response times and higher accuracy rates (i.e., lower speed and higher accuracy). The two principles apply to homogeneous types of cognitive processing during task performance (Ratcliff et al., 2016; Wickelgren, 1977). Sequential sampling and incremental evidence ac- cumulation are popular examples of such homogeneous and continuous https://doi.org/10.1016/j.intell.2018.04.001 Received 10 October 2017; Received in revised form 4 April 2018; Accepted 8 April 2018 ⁎ Corresponding author. E-mail address: Chenh@ada.org (H. Chen). Intelligence 69 (2018) 16–23 0160-2896/ © 2018 Elsevier Inc. All rights reserved. T