Rhesus Macaques (Macaca mulatta) Monitor Uncertainty During Numerosity Judgments Michael J. Beran Georgia State University J. David Smith and Joshua S. Redford University at Buffalo, State University of New York David A. Washburn Georgia State University Two rhesus macaques (Macaca mulatta) judged arrays of dots on a computer screen as having more or fewer dots than a center value that was never presented in trials. After learning a center value, monkeys were given an uncertainty response that let them decline to make the numerosity judgment on that trial. Across center values (3–7), errors occurred most often for sets adjacent in numerosity to the center value. The monkeys also used the uncertainty response most frequently on these difficult trials. A 2nd experiment showed that monkeys’ responses reflected numerical magnitude and not the surface-area illumination of the displays. This research shows that monkeys’ uncertainty-monitoring capacity extends to the domain of numerical cognition. It also shows monkeys’ use of the purest uncertainty response possible, uncontaminated by any secondary motivator. Keywords: uncertainty monitoring, metacognition, numerosity, monkeys, Macaca mulatta The question of whether nonhuman animals experience uncer- tainty in the judgments they make has been investigated with diverse experimental procedures. Like humans (e.g., Carr, Alex- ander, & Folds-Bennett, 1994; Cawley & Miller, 1986; Flavell, 1979; Koriat, 1993; Metcalfe & Shimamura, 1994; Nelson, 1992; Schwartz, 1994; Terrace & Metcalfe, 2004), some nonhuman animal species apparently monitor and respond to uncertainty states (Hampton, 2001; Inman & Shettleworth, 1999; Shields, Smith, Guttmannova, & Washburn, 2005; Shields, Smith, & Washburn, 1997; Smith et al., 1995; Smith, Shields, Allendoerfer, & Washburn, 1998; Smith, Shields, Schull, & Washburn, 1997; Smith, Shields, & Washburn, 2003; Smith & Washburn, 2005; Son & Kornell, 2005). The general approach in the comparative studies has been to use perceptual and cognitive tasks that contain a mix of easy and difficult trials while also giving animals a third uncertainty response with which they can choose to decline mak- ing a primary response. Given this paradigm and this response, animals might choose to complete the trials they think they know while declining the difficult trials on which they would most likely err. This is what they do, and humans and animals have produced uncertainty-monitoring data patterns that are among the strongest human and animal performance similarities in the comparative literature. For example, Smith et al. (1997) combined a dense–sparse discrimination with an uncertainty response. On each trial, partic- ipants could make a dense or sparse response, respectively, after judging whether a box had the critical pixel-density level to be called dense (48 white pixels per cm 2 ) or any density less than this. Or they could use the uncertainty response to decline the present trial and move into a new guaranteed-win one. Trial difficulty was titrated so as to maintain difficulty at a high level and to map carefully monkeys’ response patterns in the region surrounding their threshold for distinguishing dense from sparse. Dense and sparse responses naturally dominated for the clearest cases of the two trial types. However, the discrimination was performed at chance near the monkeys’ thresholds, and the uncertainty response was used most in this region of maximum uncertainty. Monkeys assessed accurately when they were liable to err in the discrimi- nation and declined those trials selectively and adaptively. Humans did so, too, and attributed their uncertainty responses to uncer- tainty monitoring. Here we extend uncertainty-monitoring research for the first time to monkeys’ numerosity judgments. This extension seemed important for several reasons. First, the judgment of numerosity is a basic cognitive capacity of evolutionary significance (Gallistel, 1990; Gallistel & Gelman, 2000) and a continuing focus of com- parative cognition research (e.g., Beran, 2001, 2004; Beran & Beran, 2004; Biro & Matsuzawa, 2001; Boysen & Berntson, 1989; Brannon & Terrace, 2000; Capaldi & Miller, 1988; Hauser & Carey, 2003; Hauser, Carey, & Hauser, 2000; Matsuzawa, 1985; Washburn & Rumbaugh, 1991). Michael J. Beran, Language Research Center, Georgia State University; J. David Smith, Department of Psychology and Center for Cognitive Science, University at Buffalo, State University of New York; Joshua S. Redford, Department of Psychology, University at Buffalo, State Univer- sity of New York; David A. Washburn, Department of Psychology and Language Research Center, Georgia State University. This research project was supported by Grant HD-38051 from the National Institute of Child Health and Human Development and by the Rumbaugh Fellowship of Georgia State University. This article honors the research careers of Donald R. Griffin and Thomas O. Nelson, who pioneered the study of animal awareness and human metacognition. Correspondence concerning this article should be addressed to Michael J. Beran, Language Research Center, Georgia State University, P.O. Box 5010, Atlanta, GA 30302-5010. E-mail: mjberan@yahoo.com Journal of Experimental Psychology: Copyright 2006 by the American Psychological Association Animal Behavior Processes 2006, Vol. 32, No. 2, 111–119 0097-7403/06/$12.00 DOI: 10.1037/0097-7403.32.2.111 111