Categorizing Cognition: Toward Conceptual Coherence in the Foundations of Psychology Halford, G.S., Wilson, W.H., Andrews, G., & Phillips, S. (2014). Cambridge, MA: MIT Press http://mitpress.mit.edu/books/categorizing-cognition ISBN: 9780262028073 (hardback) or 9780262320696 (ebook) Background: All sciences need ways to classify phenomena to achieve conceptual coherence (cf. the periodic table and biological taxonomic system). This book proposes a categorization of cognition based on core properties of constituent processes, recognizing correspondences between cognitive processes with similar underlying structure but different surface properties. For copyright reasons, a full-text version of this book may not be posted online. This document provides information about the book, including the Preface. You can find further sample material by searching for "Categorizing Cognition" in Google Books ( books.google.com ). Graeme Halford Specialisation: cognitive complexity, processing capacity, workload, processing load, reasoning. Graeme Halford is Adjunct Professor, School of Applied Psychology, Griffith University, where he continues his work on cognitive processes, with particular reference to complexity of reasoning and decision making.professor Halford's recent research includes a new metric for conceptual complexity applicable to humans and higher animals, with applications to mathematics education and command and control tasks. Bill Wilson Bill Wilson is an Honorary Associate Professor at UNSW Sydney. He is engaged in research on applications of (mathematical) category theory to cognitive modelling, and on modelling relational reasoning. A continuing theme is the use of the concept of relational complexity, to measure the difficulty of mental tasks, and the connections between relational complexity and cognitive coordinate systems.
Glenda Andrews Glenda Andrews is an Associate Professor in the School of Applied Psychology at Griffith University. Her research interests are in higher cognitive processes such as reasoning, problem solving and decision making. She conducts research examining the role of relational complexity in these processes as well as changes related to age and to brain injury due to stroke. Steven Phillips Dr Steven Phillips is Chief Scientist at the National Institute of Advanced Industrial Science and Technology, Japan. His research interests concern cognition, and category theory as it helps to understand why cognitive systems work. Endorsements Categorizing Cognition provides an integrative treatment of complexity from the perspective of cognitive science. The volume lays out theoretical principles for understanding variations in complexity across tasks, highlighting implications for cognitive development, adult human cognition, cognitive neuroscience, comparative psychology, and computational models. For all those interested in understanding complexity (and how a human mind can try to cope with it), this book will be an invaluable resource. Keith Holyoak, Distinguished Professor of Psychology, University of California, Los Angeles; coauthor of Mental Leaps: Analogy in Creative Thought; coeditor of The Oxford Handbook of Thinking and Reasoning Educators urgently need to build on sound, substantial efforts such as this one if they are serious about supporting development of cross-domain intellectual skills. Their long-standing reliance on weakly specified constructs like critical thinking or Bloom's taxonomy can t do the job. Deanna Kuhn, Professor of Psychology and Education, Teachers College, Columbia University Graeme Halford and his colleagues have written an extraordinary book that draws on several decades of first-rate theorizing and research to provide a comprehensive account of the mechanisms underlying higher cognitive functions and their development one that recognizes the fundamental importance of cognitive complexity to an understanding of the human mind. This book is a must-read for any serious student of cognition. Philip David Zelazo, Nancy M. and John E. Lindahl Professor, Institute of Child Development, University of Minnesota
Table of Contents Plan of Chapters ix Preface xi Acknowledgments xv 1 Introduction and Statement of the Problem 1 2 Properties of Cognitive Processes 27 3 Relational Knowledge in Higher Cognition 65 4 Cognitive Complexities and Correspondences 91 5 Representational Rank 137 6 Acquisition of Relational Knowledge and the Origin of Symbols 177 7 Neural Nets as Models of Acquisition Processes 211 8 Human Reasoning and Relational Knowledge 231 9 Applications of Relational Knowledge Theory 253 10 Conclusion 275 Glossary of Terms and Abbreviations 287 References 297 Index 337
Preface Conceptual coherence is of great value in all sciences a fact illustrated by the importance of the periodic table in chemistry and the classification system in biology. Such conceptual coherence has proved elusive in psychology so far, but several important lines of progress in the last few decades have made it more possible. These include relational knowledge, a cognitive complexity metric that has a high degree of generality, new conceptions of working memory, detailed and penetrating conceptions of human reasoning, and identification of the properties of the major categories of cognitive processes. In this book, we utilize these advances to propose a categorization of cognition based on core properties of the constituent processes. Our formulation is based on numerous demonstrated correspondences between cognitive processes with similar underlying structures but different surface properties. We also have shown that there is a mathematical basis for these correspondences that are therefore not coincidental. Coherence depends on finding the right criteria for assigning phenomena to categories, which in turn depends on discovering the fundamental properties of key phenomena. Higher cognition is distinguished by the representation of the structure of tasks and situations. These representations have to be accessible to a wide range of processes, including some that are applicable beyond the situation where the representation was learned, and they must do more than compute outputs. For example, higher cognitive representations enable mappings between representations based on common structure, permitting content-independent transfer and providing a means to adapt to as-yet-unforeseen situations. We proposed a long time ago that the structure of representations could form the basis of a system for categorizing cognitive processes, and experience has shown that the categorizations hold up over a wide range of paradigms and domains. Categories based on common sense or traditional distinctions such as abstract versus concrete are inadequate for the purpose of achieving coherence. While conducting a very extensive reading of the literature covering a wide range of topics, going back several decades, we have thought carefully about the possibilities for achieving coherence, and now we are convinced that at least the beginning of a coherent formulation is feasible. Our reading of the literature has provided multiple examples of contradictions and inconsistencies, and we have found that they can be resolved by our categorization. The relational knowledge literature that has developed over the last three decades has been important for providing the means to a coherent formulation. From this literature, we have identified dynamic structural alignment as the foundational property of higher cognition, as discussed in chapter 2. We have found that bringing areas of advancing knowledge together has led to unexpected solutions to some fundamental issues. For example, by integrating relational knowledge, a contemporary theory of working memory, and longstanding findings in infant cognition, we have been able to devise a new approach to the origin of symbolic processes. As a result, some performances in infancy can be seen to reflect symbolic processes, albeit of a simple kind. In addition, we have seen links between processes observed in infant cognition and fundamental language processes. Thus, the dynamic creation of bindings in working memory between objects and locations can be seen as essentially the same process as the assignment of words to roles in sentence comprehension. This has been proposed in chapters 2 and 6. Many advances have made a coherent categorization of cognition feasible. Working memory now can be measured with greater precision using latent variable
techniques, and it has been shown to be a major factor in reasoning and intelligence. It has been shown to play a major role in reasoning and higher cognition generally and is essential to the establishment of representations used in symbolic processes. In addition, working memory provides a new way of handling the question of individual and developmental differences in capacity. The nature of symbolic processes is also better understood. Attempts to avoid postulating symbolic processes in the past have tended to obscure fundamental differences in cognitive processes. A task performed symbolically can be very different from one performed by subsymbolic processes, and reluctance to recognize this fact can lead to confusion, inconsistency, and proliferation of exceptions. Extensive research has shown that it is not possible to define a cognitive process solely by the behavior that it manifests. Using the classical conditional discrimination task, we show that an instance of the task can be performed by any of three distinct cognitive processes, each with a unique set of properties. These processes are discriminable by established, objective methodologies. Each category of cognition corresponds to processes that are performed in many other tasks and paradigms. Thus, a cognitive process in the conditional discrimination paradigm might relate to a process in another possibly very different paradigm, more than to another process performed in conditional discrimination. Therefore, a single set of observed behaviors does not permit categorization of the relevant cognitive process. There has been resistance to attempts to achieve coherence for a number of reasons. Explanation of all cognition by a single set of processes sometimes has been seen as more desirable, or more parsimonious, than categorization of cognitive processes. However, parsimony is not achieved by theories that produce inconsistencies and exceptions, and we propose that a categorization based on foundational properties yields both coherence and parsimony. There also has been a tendency to avoid foundational issues, and concentration on exciting findings of previously unknown capabilities has discouraged the investigation of species differences and developmental progressions. We do not deny the achievements in animal cognition and cognitive development, but we have shown that their significance is actually increased by defining their place in a coherent conception of cognition. Large accumulated databases on infant and animal cognition entail both the need and the opportunity to identify the core properties that distinguish different levels of cognition. Cognitive complexity and processing capacity have been eschewed in much of the research conducted in recent decades because they are seen as conflicting with efforts to demonstrate cognitive precocity in infants and young children and/or to attribute humanlike cognitions in other animal species. However, this has often led to fragmentation and inconsistency, as when higher cognitions are attributed on the basis of very restricted criteria. Cognitive complexity and its corollary, cognitive capacity, also have been neglected in applied cognition. Complexity is a major issue in 21st-century life as well as in cognitive psychology, and we propose to give it at least some of the attention that it deserves. In chapter 4, we show how the usual objections to these concepts are without validity. In chapter 9, we provide a brief sketch of how our theory can be applied to practical issues in human factors, based on our own experience of consultation in that field. The purpose of this book is to provide at least a beginning of a categorization system that will yield the coherence that often has been advocated for psychology, but never achieved.