Peter Fonagy and Mary Target present a broad and provocative

Transcription

1 jap a Philip Robbins / Jeffrey M. Zacks 55/2 ATTACHMENT THEORY AND COGNITIVE SCIENCE: COMMENTARY ON FONAGY AND TARGET Peter Fonagy and Mary Target present a broad and provocative argument for the potential contributions of attachment theory to clinical psychoanalysis, together with an analysis of why these contributions have thus far not been realized. Central to the authors discussion is the claim that the movement within contemporary cognitive science known as embodied cognition provides the theoretical machinery to bring attachment theory and psychoanalysis together. This conclusion depends on two claims about attachment theory s relationship to cognitive science. First, Fonagy and Target argue that attachment theory as originally developed by Bowlby was based on a concept of mental representation derived from classical, disembodied cognitive science. Second, they argue that what is new about contemporary cognitive science is its turn to embodied cognition. These claims support their conclusion that attachment theory should be revised by bringing it more into line with contemporary cognitive science. Our perspective is that of working cognitive scientists one of us is a philosopher of mind and the other a cognitive neuroscientist. From our perspective, the conceptual map presented by Fonagy and Target has recognizable landmarks, but deviates from our experience of the terrain. To begin with, attachment theory is not nearly so well aligned with classical cognitive science as they suggest; on the contrary, it shares at least as much ground with embodied cognition, if not Philip Robbins, Assistant Professor of Philosophy, Washington University, St. Louis. Jeffrey M. Zacks, Associate Professor of Psychology, Washington University, St. Louis. This commentary was a fully collaborative effort. The order of the authors names is arbitrary.

2 Philip Robbins / Jeffrey M. Zacks more. Likewise, embodied cognition lies at a considerable remove from mainstream contemporary cognitive science. So we see the geography somewhat differently. We agree with Fonagy and Target that attachment theory can be updated so as to better integrate it into contemporary cognitive science, but because we view contemporary cognitive science differently, we see that updating as taking a different form. In what follows, we address these three points in turn. The first two sections lay out our understanding of the relevant features of attachment theory and cognitive science; the third section addresses the way forward. ATTACHMENT THEORY AND CLASSICAL COGNITIVE SCIENCE 458 As noted above, Fonagy and Target s argument rests on the claim that attachment theory is deeply rooted in classical cognitivism and the associated computer model of the mind. But even in its original formulation, attachment theory does not fit neatly into this paradigm. For example, Fonagy and Target describe Bowlby s internal working model mechanism a kind of social-cognitive mapping device as prototypical of a now discredited disembodied information-processing approach. But judging from Bowlby s classic writings (e.g., Bowlby 1982), the roots of the theory lie mostly elsewhere: in control theory, evolutionary biology, and ethology. Against this background, Fonagy and Target s claim that Bowlby s theory requires thorough revision to bring it into line with alternative developments in cognitive science, embodied approaches in particular, seems questionable. A brief sketch of the background will suggest why we are skeptical on this point. At the heart of attachment theory is the concept of attachment behavior, a form of instinctive social behavior the main goal of which is to minimize the distance between infant and mother (or other primary caregiver), thereby fostering interaction between them. Such behavior occurs in a wide range of animal species, from mice to humans. In Bowlby s account, attachment theory forms a chapter of the general theory of instinctive behavior a theory that naturally lends itself to treatment within the theory of control systems, or cybernetics (Bowlby 1982, pp ). In this framework, attachment behavior is seen as the product of a special kind of homeostatic control system: the attachment system. In much the same way that a thermostat regulates the temperature of a room by checking the temperature, comparing it with

3 COMMENTARY ON FONAGY AND TARGET a set-point, and engaging heating or cooling systems as necessary, the infant s attachment system computes the distance to mother, compares it with a set-point, and engages locomotor systems as necessary. In the course of elaborating this idea, Bowlby cites work on applications of control theory both to general biology (Bayliss 1966; Grodins 1963) and to psychology and neuroscience (Young 1964; McFarland 1971). This work a direct predecessor of recent dynamical-systems-theoretic approaches in cognitive science (Thelen and Smith 1994; Port and van Gelder 1995) looms much larger in the conceptual landscape of attachment theory than anything in the classical rules-and-representations tradition emphasized by Fonagy and Target. When Bowlby does allude to this tradition, he cites work that draws heavily on control theory (Miller, Galanter, and Pribram 1960). The biological flavor of his discussion, which extends to a conjecture about the neural basis of behavioral control systems, is evident throughout. A second salient aspect of attachment theory is the influence of evolutionary biology. This influence is clear from Bowlby s emphasis on attachment as an instinctive social-behavioral capacity that contributes no less to an animal s adaptive fitness than does the capacity to find food, warmth, or a mate. In developing his general account of behavioral systems (of which the attachment system is an example), Bowlby makes extensive use of ideas from evolutionary theory, as when he notes that the only relevant criterion by which to consider the natural adaptedness of any particular part of present-day man s behavioural equipment is the degree to which and the way in which it might contribute to population survival in man s primeval environment (Bowlby 1982, p. 59). Like basic ideas from control theory, these evolutionary ideas are not incompatible with classical cognitivism. Nonetheless, they inform an approach to the study of mind that diverges in important ways from the mind-as-computer analogy at the core of early cognitive science. Indeed, it diverges from that analogy in much the same way that the embodied approach does: as Fonagy and Target put it, by making the evolutionary adaptive function of cognition the main focus of study. A third, and related, way in which Bowlby s work on attachment theory differs from classical cognitivist research lies in its ethological orientation: specifically, in its attention to studies of cognition and behavior in the wild. For example, in his discussion of attachment behavior in nonhuman species, Bowlby emphasizes the results of 459

4 Philip Robbins / Jeffrey M. Zacks 460 observation of animals in their natural habitats (or facsimiles thereof), as opposed to the results of tightly controlled studies in the laboratory (Bowlby 1982, pp ). Indeed, much of the basis of his discussion of attachment behavior in humans comes from observation of infants and mothers interacting in more or less ordinary settings. This concern with ecological validity, and the way that environmental context shapes mind and behavior, puts him at odds with the classical computationalist tradition. It also provides a precedent for a central feature of embodied cognition as Fonagy and Target describe it, namely, a move away from reified laboratory studies and an interest in ecological approaches to cognition (for further discussion, see the following section). In short, Bowlby s original framing of attachment theory seems to us to be less classical in spirit, and more embodied, than Fonagy and Target make out. A further problem with their story here is the way it represents, or rather misrepresents, the history of cognitive science. As they note, cognitive psychology, unlike behaviorism, posited unobservable mental representations to account for observable behavior. However, cognitive psychology is slightly older and more continuous with previous psychological theorizing than Fonagy and Target suggest. The term cognitive psychology was not (as the authors report) first coined by Neisser in the 1967 textbook of that name (Neisser 1967); it was in use at least by 1939, when another textbook with that title was published (Moore 1939; see also Surprenant and Neath 1997; Knapp 1985). The notions of information processing and internal representation were central to a generation of psychologists that preceded Neisser s cohort (e.g., Miller 1956; Piaget and Inhelder 1956). To go back yet further, internal representations played a central role in Gestalt psychology (Köhler 1929). These developments within psychology, and others, share a core notion of internal mental representations but not the specific commitment to a view that mental life is similar to the calculations performed by a digital computer. Thus, while we agree that the original framing of attachment theory, with its focus on internal working models of the social environment, bore a certain affinity with 1960s cognitive psychology, we doubt that this affinity binds attachment theory to the specifics of the computer metaphor. There is, in any case, little or no textual evidence of this further commitment to be found in the locus classicus of attachment theory (Bowlby 1982).

5 COMMENTARY ON FONAGY AND TARGET EMBODIED COGNITION AND CONTEMPORARY COGNITIVE SCIENCE Fonagy and Target provide a creative synthesis of several important threads within contemporary cognitive science. From their perspective these empirical and theoretical trends cohere into a broad movement that they refer to collectively as embodied cognition. However, this cyclopean view misses some empirical and conceptual distinctions that may be important for their argument. (For a comprehensive and readable overview of embodied cognition, see Anderson 2003.) Fonagy and Target summarize current developments in cognitive science as follows: The model of mind assumed by attachment theorists is consistent with the important discoveries of the first generation of cognitive psychology, but this approach has been supplanted by a number of recent developments collectively called embodied cognition or enactive mind. Features of this new approach include (1) the increasing use of introspection as a research method; (2) a keen interest in the understanding of emotion as organizer as well as motivator of behavior; (3) rapidly advancing brain-imaging technology that has made cognitive neuropsychology into a brain as well as a mental science and led to increasingly functional cognitive accounts; and (4) a move away from reified laboratory studies and an interest in ecological approaches to cognition. 461 We will address each of these points, but we will take them out of order, addressing points 1 and 4, and then 2 and 3. The increasing use of introspection as a research method. We find little evidence for this. Introspection reached its zenith in psychology with the structuralist psychology of Titchener and his colleagues at the beginning of the twentieth century, and depended on the observations of small numbers of highly trained observers (Boring 1950). Recent proponents have argued for broader use of introspection, particularly by larger numbers of more representative participants but have lamented the fact that these techniques are rarely used (Jack and Roepstorff 2002; Hurlburt and Heavey 2001). A move away from reified laboratory studies and an interest in ecological approaches to cognition. We also find little evidence that cognitive scientists are abandoning the laboratory. The oversimplified and unnatural environment of the laboratory has been a concern of experimental psychology s critics for most of its history (see, e.g., James

6 Philip Robbins / Jeffrey M. Zacks , p. 192). Movements away from the laboratory under the rubric ecological psychology had their strongest influence in the 1960s and 1970s (e.g., Barker 1968; Gibson 1979). In fact, Neisser s Cognitive Psychology (1967), which Fonagy and Target describe (rightly) as a pioneering work in cognitive science, has a strongly ecological bent. Thus, if ecological approaches are more friendly to the bridging of attachment theory and psychoanalysis, it makes more sense to look backward for missed connections than to look forward for new ones. A keen interest in the understanding of emotion as organizer as well as motivator of behavior. We agree with Fonagy and Target that there is increasing interest in the role of emotion in cognition, and in its neural basis. These developments go by the labels affective neuroscience and social neuroscience. As Fonagy and Target note, some researchers have developed theories of the emotions that ground the psychological experience of emotion in bodily experiences; examples include Damasio s somatic marker hypothesis (1994) and Ekman s theory of the facial expression of emotion (1982). However, much research on the neural basis of emotion places no particular emphasis on bodily states in understanding the relationship between brain states and the psychological experience of emotion (see, e.g., Davidson and Irwin 1999; LeDoux 1996). Thus, there are two problems with the thesis that these developments support the grounding of the mind in the body in contemporary cognitive science. First, much of social and affective neuroscience could fairly be called disembodied. Second, social and affective neuroscience are not characteristic of most of contemporary cognitive science; rather, they are newly emerging fields whose influence on cognitive science as a whole has yet to be determined. Rapidly advancing brain-imaging technology that has made cognitive neuropsychology into a brain as well as a mental science and led to increasingly functional cognitive accounts. We also agree with Fonagy and Target that cognitive science has increasingly become a brain science as well as a mental science. (They use the term cognitive neuropsychology here, but we think the comment applies better to cognitive science as a whole. Cognitive neuropsychology is a subdiscipline that historically has studied the effects of brain lesions on cognition, and so has been a brain science all along.) The part of cognitive science that addresses mind-brain relationships empirically is known as cognitive neuroscience. As Fonagy and Target point out, a major driver of cognitive neuroscience has been the availability of noninvasive

7 COMMENTARY ON FONAGY AND TARGET brain-imaging technologies as well as the use of invasive electrical recording and stimulation techniques in nonhuman animals and neurosurgical patients. The article suggests that cognitive neuroscience has helped undercut the 1960s computer metaphor of the mind. We agree. However, we think this has been due less to neuroimaging than to the development of computational models of psychological processes based on the information-processing properties of the nervous system. This type of modeling is referred to as neural network modeling or connectionism. It traces its roots in part to the pioneering work of Hebb (1949) and McCulloch and Pitts (1943) in the 1940s, but the modern field of connectionism really was launched with a two-volume work by Rumelhart et al. (1986). Connectionist models incorporate several principles of neural systems that differentiate them from classical computer-style models. These include analog-valued computing and distributed representations. Neither of these is compatible with classical information-processing theories. Connectionist modeling is by no means the dominant paradigm in cognitive science, but its core ideas have had broad influence among theorists who may not adopt its formal methods. As a result primarily of the influence of connectionism, most theories in cognitive neuroscience are embodied in the sense that they claim that the form of mental representations depends importantly on the nature of the physical devices nervous systems that implement them. However, this is not the sense of embodiment that embodied cognition theorists usually have in mind when they use the term. Cognitive neuroscience is not the same thing as embodied cognition, because the claim that mental representations are grounded in the brain is not the same as the claim that they are grounded in perceptual-motor experiences. So if embodied cognition isn t cognitive neuroscience, what is it? As Fonagy and Target indicate, a central claim of embodied cognition theorists is this: Mental representations are not abstractions that stand in an arbitrary relationship to the things they represent (as in a computer); rather, they are grounded in specific perceptual and motor experiences, the structure of which depends on the structure of one s body, and they retain the marks of that grounding even when one is thinking about abstract concepts (Barsalou 1999; Barsalou et al. 2003). One implication of this view is that the meanings of external symbols such as words include not just definitional criteria and associations to other symbols, but also perceptual images (see, e.g., Zwaan, Stanfield, and Yaxley 2002) and motor programs (see, e.g., Glenberg 463

8 Philip Robbins / Jeffrey M. Zacks 464 and Kaschak 2002). Another implication is that when people think about abstract concepts such as truth or jealousy, they ground these concepts in physical experiences (Lakoff and Johnson 1999). To some extent, then, Fonagy and Target s description of embodied cognition is accurate. The main point of confusion, we think, is that they fail to differentiate between several quite different trends in recent cognitive science: embodied cognition, cognitive neuroscience, and social/affective neuroscience. Thus, we believe that points 1 and 4 are incorrect, and that 2 and 3, while reasonably accurate summaries of recent developments, fit better under the headings of social-affective neuroscience and cognitive neuroscience, respectively, than under the heading of embodied cognition. This is not merely a terminological point the important thing is that not all social-affective neuroscience theories assign a significant role to the body, and relatively few cognitive neuroscience theories do. These conclusions lead us to question whether contemporary cognitive science is any closer in spirit to psychoanalysis than was classical cognitive science. We find Fonagy and Target s argument that embodiment is a natural concept for psychoanalysis intriguing, but we doubt that any implications of this point carry over from the specific theoretical orientation of embodied cognition to cognitive science more broadly. EXPANDING THE TOOLBOX OF ATTACHMENT THEORY Our analysis leads us to some suggestions for what contemporary cognitive science has to offer attachment theory, which differ somewhat from those offered by Fonagy and Target. First, we think that social-affective neuroscience offers a set of techniques for investigating the mechanisms by which the regulatory mechanisms proposed by Bowlby are implemented (see Panksepp 1998). As Fonagy and Target suggest, brain imaging is one important set of tools; invasive lesion and electrical stimulation and recording studies in animals are others. The increasingly widespread use of these techniques is surely something that Bowlby would have applauded, given his biological bent. Second, we think that cognitive neuroscience particularly connectionist modeling provides a set of tools for thinking about the social-cognitive maps posited by attachment theory. On our analysis,

9 COMMENTARY ON FONAGY AND TARGET the problem with attachment theory vis-à-vis internal representation wasn t so much that it used discrete representations that were rigid or disembodied; rather, it was that it used control-theoretic formulations that lacked representational power. Modern connectionist representations are much better suited to control theory and its intellectual successor, dynamical systems theory (Eliasmith and Anderson 2002). Because representations in connectionist models are distributed and continuous-valued, they provide a natural means of implementing homeostatic control mechanisms. At the same time, because they have the ability to implement complex structured representations, maintain them over time, and learn, connectionist models provide a natural mechanism to characterize the development of internal models of the social world that can guide behavior. In conclusion, like Fonagy and Target we see the future promise of attachment theory as resting upon its integration with contemporary cognitive science. But where they see embodied cognition as the basis of that integration, we see the various neurosciences cognitive, affective, social, and computational as playing that role. As to which perspective will prevail, we will have to wait and see. In the meantime, we hope that our view of cognitive science provides a useful counterpoint to the view taken by Fonagy and Target. Examining the relationship between cognitive science and attachment theory from this perspective leads, we think, to some new conclusions. We note, however, that our analysis by itself says little about the relationship between attachment theory and psychoanalysis. 465 REFERENCES ANDERSON, M.L. (2003). Embodied cognition: a field guide. Artificial Intelligence 14: BARKER, R.G. (1968). Ecological Psychology: Concepts and Methods for Studying the Environment of Human Behavior. Stanford: Stanford University Press. BARSALOU, L.W. (1999). Perceptual symbol systems. Behavioral & Brain Sciences 22: SIMMONS, W.K., BARBEY, A.K., & WILSON, C.D. (2003). Grounding conceptual knowledge in modality-specific systems. Trends in Cognitive Science 7: BAYLISS, L.E. (1966). Living Control Systems. San Francisco: W.H. Freeman. BORING, E.G. (1950). A History of Experimental Psychology. 2nd ed. New York: Appleton-Century-Crofts.

10 Philip Robbins / Jeffrey M. Zacks 466 BOWLBY, J. (1982). Attachment and Loss: Vol. 1. Attachment. 2nd ed. New York: Basic Books. DAMASIO, A.R. (1994). Descartes Error: Emotion, Reason, and the Human Brain. New York: Putnam. DAVIDSON, R.J., & IRWIN, W. (1999). The functional neuroanatomy of emotion and affective style. Trends in Cognitive Science 3: EKMAN, P. (1982). Emotion in the Human Face: Emotion and Social Interaction. 2nd ed. Cambridge: Cambridge University Press. ELIASMITH, C., & ANDERSON, C.H. (2002). Neural Engineering: Computation, Representation, and Dynamics in Neurobiological Systems. Cambridge: MIT Press. GIBSON, J.J. (1979). The Ecological Approach to Visual Perception. Boston: Houghton Mifflin. GLENBERG, A.M., & KASCHAK, M.P. (2002). Grounding language in action. Psychonomic Bulletin & Review 9: GRODINS, F.S. (1963). Control Theory and Biological Systems. New York: Columbia University Press. HEBB, D.O. (1949). The Organization of Behavior: A Neuropsychological Theory. New York: Wiley. HURLBURT, R.T., & HEAVEY, C.L. (2001). Telling what we know: Describing inner experience. Trends in Cognitive Science 5: JACK, A.I., & ROEPSTORFF, A. (2002). Introspection and cognitive brain mapping: From stimulus-response to script-report. Trends in Cognitive Science 6: JAMES, W. (1890). The Principles of Psychology. Vol.1. New York: Dover. KNAPP, T.J. (1985). Contributions to the history of psychology: T.V. Moore and his Cognitive Psychology of Psychological Reports 57(3, pt. 2): KÖHLER, W. (1929). Gestalt Psychology. New York: Liveright. LAKOFF, G., & JOHNSON, M. (1999). Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. New York: Basic Books. LEDOUX, J.E. (1996). The Emotional Brain: The Mysterious Underpinnings of Emotional Life. New York: Simon & Schuster. MCCULLOCH, W.S., & PITTS, W.A. (1943). A logical calculus of the ideas immanent in nervous activity. Bulletin of Mathematical Biophysics 5: MCFARLAND, D.J. (1971). Feedback Mechanisms in Animal Behavior. New York: Academic Press. MILLER, G.A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review 63: GALANTER, E., & PRIBRAM., K.H. (1960). Plans and the Structure of Behavior. New York: Holt. MOORE, T.V. (1939). Cognitive Psychology. Philadelphia: Lippincott.

11 COMMENTARY ON FONAGY AND TARGET NEISSER, U. (1967). Cognitive Psychology. New York: Appleton-Century- Crofts. PANKSEPP, J. (1998). Affective Neuroscience. Oxford: Oxford University Press. PIAGET, J., & INHELDER, B. (1956). The Child s Conception of Space. London: Routledge & Kegan Paul. PORT, R.F., & VAN GELDER, T., EDS. (1995). Mind as Motion: Explorations in the Dynamics of Cognition. Cambridge: MIT Press. RUMELHART, D.E., MCCLELLAND, J.L., & THE PDP RESEARCH GROUP (1986). Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Cambridge: MIT Press. SURPRENANT, A., & NEATH, I. (1997). T.V. Moore s (1939) Cognitive Psychology. Psychonomic Bulletin & Review 4: THELEN, E., & SMITH, L.B. (1994). A Dynamic Systems Approach to the Development of Cognition and Action. Cambridge: MIT Press. YOUNG, J.Z. (1964). A Model of the Brain. Oxford: Oxford University Press. ZWAAN, R.A., STANFIELD, R.A., & YAXLEY, R.H. (2002). Language comprehenders mentally represent the shape of objects. Psychological Science 13: Philip Robbins Washington University Department of Philosophy Campus Box 1073 St. Louis, MO Fax: probbins@artsci.wustl.edu 467 Jeffrey M. Zacks Washington University Department of Psychology Campus Box 1125 St. Louis, MO Fax: jzacks@artsci.wustl.edu