Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society 317 The British Psychological British Journal of Psychology (2008), 99, 317–340 Society q 2008 The British Psychological Society www.bpsjournals.co.uk Cognitive Ethology: A new approach for studying human cognition Alan Kingstone1*, Daniel Smilek2 and John D. Eastwood3 1Department of Psychology, University of British Columbia, British Columbia, Canada 2Department of Psychology, University of Waterloo, Ontario, Canada 3Department of Psychology, York University, Ontario, Canada We all share a desire to understand and predict human cognition and behaviour as it occurs within complex real-world situations. This target article seeks to open a dialogue with our colleagues regarding this common goal. We begin by identifying the principles of most lab-based investigations and conclude that adhering to them will fail to generate valid theories of human cognition and behaviour in natural settings. We then present an alternative set of principles within a novel research framework called ‘Cognitive Ethology’. We discuss how Cognitive Ethology can complement lab-based investigations, and we show how its levels of description and explanation are distinct from what is typically employed in lab-based research. The study of human cognition has been punctuated by three historical stages of advance (Van Kleeck & Kosslyn, 1991). The first stage, beginning in the late 1950s to early 1960s was marked by a rapid progression propelled by the methods of traditional psychophysics and experimental psychology. The second stage, beginning by the mid-1970s, was fuelled by computational analysis that signalled the arrival of cognitive science. The third phase, which began in the mid-1980s, incorporated evidence from neuropsychology and animal neurophysiology, and most recently an ever increasing array of techniques for scanning the brain of alert participants. In the present article, we take as our starting-point a critical problem that continues to bedevil the study of human cognition that arose precisely from the original and remarkably successful methods of experimental psychology. Those methods, which involved simplifying the issue of investigation by making the experimental context both impoverished and controlled, sought to discover causal relationships between one factor and another. The intention was that by minimizing the complexity of the environment and maximizing the experimental control, investigators could create theories that would be universally valid. However, by the mid-1970s it had become very * Correspondence should be addressed to Dr Alan Kingstone, Department of Psychology, University of British Columbia, Vancouver, British Columbia, Canada (e-mail: [email protected]). DOI:10.1348/000712607X251243 Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society 318 Alan Kingstone et al. clear that most statements were true if, and only if, particular laboratory conditions were met. In other words, the relationship between factor A and factor B was predictable if, and only if, specific conditions were established within the lab; the relationship between factors became unpredictable when these laboratory situations were not met. Thus, for example, memory experiments found that what people remembered depended on factors such as (a) what processing they performed on the stimulus materials; (b) what stimulus materials they expected to receive; (c) what materials were actually presented; (d) what people were doing before their memory was measured; (e) how their memory was measured, and so on and so forth. The take home message was that cognitive processes vary and are affected by what is happening elsewhere within the cognitive system, and therefore cognitive processes depend critically on the specific situational context in which a subject is embedded. The field’s response to the above fact has generally taken one of the two forms. One reaction is to deny that there is a problem. This ‘response’ enables one to maintain the initial assumption that cognitive processes are invariant and unaffected by what is happening elsewhere, and thus allows one to continue to create and study laboratory- specific phenomenon like ‘nonword repetition memory’ or ‘inhibition of return’. The other reaction is to acknowledge that there is a problem, but then continue to conduct research predicated on the assumption that cognitive processes are invariant. Both responses are what Broadbent (1991) has called ‘pathological’. Occasionally, investigators like Donald Broadbent and Ulric Neisser have tried a third response. They acknowledged that cognitive processes change with situational changes and worked hard to bring the implications of this fact to the awareness of others. Perhaps their only mistake was to trust that the next generation of researchers would take their words to heart and try to find a solution to the issue. In hindsight, this faith has proven to be grossly misplaced, as the next generation of researchers have adopted one of the pathological responses of the past and grounded their neuroimaging investigations on the false assumption that cognitive processes are invariant across situations. It is precisely this false assumption that allows researchers to make the remarkable claim that the cognitive processes that they engage and measure in a simple, artificial brain neuroimaging situation captures the same fundamental cognitive processes and associated neural systems that are engaged in a complex natural situation. The aim of the present paper is modest but against this historical backdrop, we believe it is vital. We aim to initiate a dialogue among researchers regarding the fact that cognitive processes vary substantially with changes in context. We also hope to stimulate researchers to find a response to this issue that is not ‘pathological’. By putting forward a possible solution of our own, a novel research approach that we call ‘Cognitive Ethology’, our intention is to encourage other researchers to develop and advance their own positive responses. While what follows for the remainder of this paper focuses primarily on instances of cognition as it pertains to the investigation of human attention, we think that the issues we raise here can be readily extended to other research domains of human cognition. Laboratory research Laboratory research in the field of human cognition is founded on the critical assumption that human cognition is subserved by processes that are invariant and regular across situations. This invariance assumption enables one to conduct a study in Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society Cognitive Ethology 319 the laboratory and then to propose that the process being measured is expressed in everyday life. Importantly, there is a second assumption that falls out of the first. Given that processes are assumed to be invariant across situations, it follows that one can reduce situational variability without compromising the nature of the process one is measuring. Indeed, a basic objective of the experimental environment in the laboratory is to gain as much control over a situation as is possible so that any change can be attributed to the variable that is being manipulated. Together, these assumptions provide a powerful one–two punch. The assumption of process stability enables the scientist to be concerned with real-life situations without ever having to leave the laboratory. In addition, the assumption of control drives the scientist increasingly away from complex real-life situations to paradigms that are simple, contrived, and artificial. These assumptions are not, however, without their risks. For instance, the assumption of invariance eliminates any need or even obligation for the scientist to confirm that the process being manipulated and measured in the laboratory actually expresses itself in the real world. Investigators do, of course, through the process of replication, check that their lab-based effects are regular within the laboratory environment. Unfortunately, a result that is invariant within the strict confines of the laboratory does not mean that it is reproducible outside the lab. Indeed, even a cursory examination of the literature reveals that there are many instances where even the most minor change within a laboratory situation will compromise the replicability of an effect (e.g. Atchley & Kramer, 2001; Berry & Klein, 1993; Bindemann, Burton, & Langton, 2008; Soto-Faraco, Morein-Zamir, & Kingstone, 2005; Wolfe & Pokorny, 1990). In addition, as any researcher knows all too well, failed replications that are published represent just the smallest tip of a very large iceberg of failed replications that are obtained in the laboratory and never published. Upon closer consideration, there is a good reason why lab-based effects should be so remarkably fragile. After all there is a large, well established, and growing body of literature indicating that process stability is tied intimately to the situation used to create it, with participants’ strategies and associated brain configurations changing from one situation to the next (see for instance Duncan & Owen, 2000 for a review). Neisser (1976) referred to these dynamic configurations as ‘schemata’, Monsell (1996) has spoken