Enriching Psychological Research by Exploring the Source and Nature of Noise

Abstract

In psychological research, noise is often considered a nuisance that obscures rather than contributes information. This simplification overlooks that noise can be informative and that by exploring the nature of the noise one can often draw additional conclusions concerning the underlying psychological processes. It is arguably only in recent years that the mainstream of researchers has taken this idea to heart and demonstrated that it can lead to breakthroughs in the understanding of human behavior. The aim of this special section is to showcase some of the ways in which systematic exploration of noise can be achieved and how it can enrich psychological research. In this introductory article, we introduce the idea of treating noise as endogenous as opposed to exogenous to the theoretical and statistical models of psychological phenomena. We then contribute a historical review of the role of noise in psychological research, including discussions of previous endogenous treatments of noise in the literature. As an illustration, we describe our own research on the precise/not precise model and show how noise distributions can be used to delineate analytic and intuitive modes of reasoning. Finally, we briefly introduce the other contributions to this special section.

Keywords: exogenous versus endogenous; methodology; noise; statistical modeling.

Fig. 1.

Two ways to treat the noise in the scientific inquiry. (Left) The origin of noise is treated as exogenous to the psychological model, a nuisance to be controlled for in the identification of the process. In this case, there is little scientific interest in understanding the origins of the noise. (Right) The noise is treated as endogenous to this psychological model, part of the explanandum, and informative about the nature of the process.

Fig. 2.

Examples of the stimuli used in Experiment 1 in Sundh et al. (2021). (a) The condition with a conceptual triangle, (b) the condition with a visual triangle, and (c) the condition with a blob shape. In all conditions, the participants were asked to estimate the area covered by the object in square centimeters (cm²). Below each stimulus is the distribution of deviations from the criterion value in the corresponding condition, with a distinctly leptokurtic distribution for the conceptual triangle, a (roughly) Gaussian distribution for the blob shape, and an intermediate result for the perceptual triangle. The figures are reproduced from Sundh, J., Collsiöö, A., Millroth, P., & Juslin, P. (2021). Precise/not precise (PNP): A Brunswikian model that uses judgment error distributions to identify cognitive processes. Psychonomic Bulletin & Review, 28(2), 351–373.

Fig. 3.

Examples of results in Experiment 2 in Sundh et al. (2021). The responses by two participants in a willingness-to-pay task, with estimated parameter values from the precise/not precise (PNP) model and a conventional regression model. A line representing the predictions of the regression model (small-dashed line) and a reference line x = y (large-dashed line) is included. In the upper panel (ID = 24), the predictions by the PNP model coincide with the reference line, and in the lower panel (ID = 84), the predictions of the PNP model coincide with the predictions of the regression model. Note that some data points overlap. The figures are reproduced from Sundh, J., Collsiöö, A., Millroth, P., & Juslin, P. (2021). Precise/not precise (PNP): A Brunswikian model that uses judgment error distributions to identify cognitive processes. Psychonomic Bulletin & Review, 28(2), 351–373.

Fig. 4.

The distribution of responses to the bat-and-ball problem from Experiment 1 in (left) Barr et al. (2015) and (right) Pennycook et al. (2015). The data from Pennycook et al. contain two outlier responses of 210, which were excluded to improve readability.