Date: Wed, 2 Apr 2003 17:11:55 +0100 From: C A Ankerstein Subject: The Big Book of Concepts
Murphy, Gregory L. (2002) The Big Book of Concepts, MIT Press, A Bradford book.
Carrie Ankerstein, Department of Human Communication Sciences, University of Sheffield, England.
Gregory Murphy's "Big Book of Concepts" is organized around the issues that surround the study of concepts, i.e. it is not organized around the various theories of concepts. After the introductory first chapter, the theories of concepts are discussed in chapters 2 and 3 and the following chapters describes an issue or set of phenomena within concepts, these chapters end with a discussion of how each theory of concepts would handle the issue.
The book is aimed towards an advanced undergraduate or beginning graduate student audience. Murphy assumes that the reader has a general knowledge of experimental psychology and some experience with cognitive psychology, though knowledge about concepts is not assumed.
The chapters are self-contained and may therefore be read out of sequence. Though this is the case, I think it may be useful to read some chapters before others. For example, Murphy discusses "Knowledge Effects" and "Induction" in separate chapters, though these issues come up again in later chapters. In those later chapters, Murphy is careful to give adequate background in these issues, however the reader may find it more fulfilling to have read these chapters beforehand. Also, since many of the issues discussed in the book are discussed with reference to the theories of concepts, it is advisable to at least glance at the theories that Murphy discusses (see Chapters 2-4).
In the first chapter the "Introduction", Murphy outlines some fundamental terminology: a "concept" is a mental representation of classes of things and a "category" refers to the classes themselves. He also states the purpose of his book: "[it] is not so much to tell you all about concepts as to provide some kind of basis to your continuing acquisition of knowledge" (p. 8).
Chapter 2: Typicality and the Classical View of Categories In this chapter, Murphy discusses the Classical or Definitional View of concepts, which states: concepts are composed of necessary and sufficient features. On this view, a concept is either an X or it isn't and all concepts that are X are all on par with each other -- no member is better than the other. It is these predictions, however that proved disastrous for the Classical View as typicality effects were found and the mythical "necessary and sufficient" features weren't.
Chapter 3: Theories After the fall of the Classical View arose the Prototype View which states that each category is represented by a single prototype or best example. Very typical members of the category will share the most overlap of features with the prototype and atypical members will share only a few features and may share other features with another prototype of another category (e.g. tomato as a fruit and/or vegetable). Thus the Prototype View solves many of the problems of the Classical View: there are no defining features and typicality effects are predicted. However, this view is not without problems either.
Another possibility is the Exemplar View: concepts are defined by personal experience with that concept. For example, "dog" is the set of all dogs that the reader of the word "dog" has encountered. The obvious argument against this is that one does not run through all dogs one has seen, instead there is a "general" idea of dogs.
The last view discussed is the Knowledge Approach which argues that concepts are part of our general understanding of the world or our mental theories about the world. This view is often called the Theory-Theory View.
Chapter 4: Exemplar Effects and Theories The issue discussed in this chapter is that of generalization or knowing things in general without thinking about a specific example. Murphy discusses two different approaches to this view:
1. Exemplar Effects: when people encounter a new item, they use a particular remembered exemplar to categorize the new item. 2. Exemplar Models: a formal (usually mathematical) approach to similarity quantification.
In this chapter, Murphy also returns to the "prototype or exemplar" question that arose in the preceding chapter: are representations best examples or generalized feature summaries. The question is not answered, but the discussion is though provoking.
Chapter 5: Miscellaneous Learning Topics The chapter is purposely named. In this chapter very specific issues are discussed as well as broader issues that affect other phenomena discussed in later chapters. These include: base-rate neglect in category learning, feature correlations in concepts, category construction and category use and the implications that follow.
By "base-rate" Murphy is referring to what may be considered a feature that occurs most frequently in a given category, e.g. bird often fly but mammals don't. In category learning, people will generally ignore the base-rate or standard set of features and focus on more defining features, e.g. they won't focus on the fact that birds often fly, but look for something more distinctive.
Feature correlation refers to the fact that an item is not a collection of unrelated features, instead these often go together, e.g. birds have wings and can fly, a knife has an edge and can cut. Categories are composed of clusters of correlated features that differ from other categories that have different clusters.
Category construction refers to a commonly used methodology in which subjects form new categories on their own without being told by the experimenter what forms a category. By category use, Murphy is referring to the use of specific exemplars or use of the category as a whole and how this use can alter the categorization of concepts.
Chapter 6: Knowledge Effects Many studies of concept learning use novel stimuli, e.g. things people have never seen before like dot patterns or chimera, to explore the principles that may be applied to other categories or domains. Though Murphy questions the relevance of such findings -- they may not be applicable to real-life concept acquisition.
In this chapter, Murphy discusses the type of effects found in more realistic stimuli, more specifically, background knowledge and the use of this knowledge in concept learning. By knowledge, Murphy means the post-hoc use of the general world knowledge people have (right or wrong) and use to explain a category when it is being learned. For example, upon learning that an ostrich is a bird, but it doesn't fly, people may use their knowledge about birds, wings and flight to rationalize a reason why ostriches don't fly: their wings are not strong/big enough to lift the massive body. In the case of novel stimuli, people are unable to use their knowledge about that category, because they have never encountered it before.
Chapter 7: Taxonomic Organization and the Basic Level of Concepts This chapter focuses on the hierarchical structure of concepts. A taxonomy is "a sequence of progressively larger categories in which each category includes all previous ones" (p. 199), e.g. animal-mammal-dog-bulldog. Murphy discusses the phenomenon of the basic level or preferred level of categorization, e.g. cat, dog, tool as opposed to Siamese cat, Labrador retriever, and wrench. He addresses the following questions: Is there evidence that people have basic levels? Do people use such hierarchies?
As Murphy acknowledges, this is rather murky water, but claims that there are at least two generalizations that can be made about hierarchies: 1. People are able to learn and use taxonomic relations in order to draw inferences. 2. People are able to reason taxonomically about novel categories.
Chapter 8: Induction If people were unable to make sensible inferences about new things, there were be little advantage in knowing that something belongs to a specific category. Categorization is not useful in and of itself, but it is being able to apply category knowledge that is useful. Murphy discusses "induction" as "the kind of reasoning that one uses when drawing conclusions about the category in general" (p. 243). For example if you were asked to check up on your neighbor's dog, you would from your knowledge of dogs know what to do, e.g. feed and water it, let it out, etc.
The chapter discusses various research tasks that have investigated the kinds of inductions that people make, how people use multiple categories when making inductions about things there are uncertain how to categorize and the different types of methodologies used in these experiments.
Chapter 9: Concepts in Infancy Murphy argues that children's concepts are largely ignored in reviews of the psychology of concepts and he offers two reasons why we should pay attention to concept development: 1. Development speaks to one of the most central questions of cognitive science: how knowledge comes into being. 2. Developmental evidence may place constraints on theories of adult competence and performance.
Chapters 9 and 10 discuss these issues. Chapter 9 covers the development of concepts from birth through the first year of life and Chapter 10 covers the conceptual development of toddlers through the early school years.
In Chapter 9, Murphy discusses many studies, focusing not only on their findings, but also the methods used. Categorization and natural and artificial concept learning are also discussed.
Chapter 10: Conceptual Development In this chapter, the fundamental question is: Are children's concepts radically different to adult concepts? After showing that children perform similarly to adult on many of the tasks generally used in the study of concepts, Murphy discusses the developmental counterparts of the issues that surround adult concepts: typicality, taxonomic organization and knowledge effects.
He concludes that though there are some differences in the content of children's concepts and their performance on experimental tasks, there is no radical difference between adult and child concepts.
Murphy also discusses other aspects of concept acquisition including constraints on conceptual learning: syntactic constraints and fast mapping. Syntactic constraints apply limits to the possible referents of new words, e.g. Look at the quassle! In this case "quassle" can only refer to a thing rather than an action. Fast mapping refers to the ability of children to use a word after minimal exposure, e.g. 2 or 3 instances. Without such constraints, concept learning would be an unbelievable achievement.
Chapter 11: Word Meaning Murphy discusses the relationship between "words" and "concepts" though the distinction has not been made in preceding chapters. In this chapter he discusses the distinction. A "concept" is "a nonlinguistic psychological representation of a class of entities in the world" and "word meaning" is "the aspect of words that gives them significance and relates them to the world" (p. 385).
Murphy's own claim is that "word meanings are psychologically represented by mapping words onto conceptual structures" (388) and there are many possibilities to how this might work, e.g. there may be a one-to-one mapping where each word is mapped onto one concept. This is unlikely though, as soon as one begins to think about polysemy -- the multiple senses of a word such as theater can be used to refer to the institution which puts on plays or the building in which plays are performed. Murphy offers a more complicated, though satisfactory relationship based on three principles which take into account the problem of ambiguity that arises out of polysemy.
Chapter 12: Conceptual Combination This chapter addresses the fact that we rarely deal with concepts in isolation, instead they general come in linguistic chains, e.g. "I found a really great dog book yesterday" where "dog book" refers to a separate concept, combined by the concept of "dog" and "book". The question is: How do people construct complex concepts out of these parts? The chapter is mostly a presentation of the various models that try to explain concept combination. He concludes with the importance of the role that knowledge plays in the combination of concepts.
Chapter 13: Anti-Summary and Conclusions Murphy concludes the book with issues that have not been discussed and why, including: computational models of conceptual processes (e.g. categorization, learning, induction), perception (the role of perceptual information in categorization and judgments), similarity and cross-cultural issues. After acknowledging these left-overs, he quickly sums up the theories discussed, the messiness of the field and future directions in the study of concepts and defends his organization and approach in the writing of the book, e.g. focusing on the issues surrounding the study of concepts rather than the theories of concepts.
Murphy's book covers a lot of the more interesting issues surrounding concepts. He adds some new insights into the field. The style is generally light, witty and clear, though not always concise. His approach of organizing the book according to issues and phenomena is novel and refreshing as most books on concepts focus on theories without integration of issues and things that affect them all.
There are, however a few things that have been surprisingly left out. For instance, nowhere is there a discussion of the view that concepts may not be featural representations, but instead atomistic representations (the Conceptual Atomism View). All the theories and discussions in the book approach concepts as featural representations, though there is an opposing view to this. Murphy assumes a featural approach without pausing to defend it.
Though Murphy covers the development of concepts, claiming its importance to the study of adult concepts, he neglects to mention the deterioration of concepts as seen in semantic dementia. Though the issue of semantic breakdown is complicated, e.g. is access to concepts or the conceptual store itself damaged, it does offer some very interesting insight into the organization and content of conceptual knowledge -- compatible with some views discussed in the book and problematic for others.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER Carrie Ankerstein is a PhD student in the department of Human Communication Sciences at the University of Sheffield, England. She has a Masters in Applied Linguistics from the University of Cambridge, England and a Bachelor's degree in German Linguistics from the University of Wisconsin-Madison, USA / University of Freiburg, Germany. Her research interests include the organization and representation of concepts in semantic memory.