This book is aimed at all scientists interested in semantic cognition, even those with limited experience of computational modeling. It benefits from concise and logically presented text and the effective use of figures and illustrations, with details of simulations tucked away in the Appendix. The length of the book defines it not only as a useful addition to a reference collection, but also an excellent primer for young researchers, postgraduates, and even advanced undergraduates.

This timely text provides an account of the parallel distributed processing (PDP) approach to semantic cognition. Building on the work of Geoffrey Hinton (1981, 1986) and David Rumelhart (1990; Rumelhart et al., 1986) and the observations of Frank Keil (1979, 1989, 1991) and Jean Mandler (1988, 1990, 1992, 1997, 2000a, 2000b, 2002), the authors propose that distributed connectionist networks provide a promising mechanism for implementing and studying semantic cognition. This approach integrates the strengths and overcomes many weaknesses apparent in more “classical” approaches to this topic.

In the PDP framework, all cognitive performance arises through the propagation of graded signals in a system of interconnected processing units. The representations used in performing these tasks are patterns of activation across units, governed by weighted connections among them. Semantic knowledge is acquired through the gradual adjustment of the strengths of these connections in the course of day-to-day experience. In line with Rumelhart's earlier work, the model remains in a simplified form to maintain clarity and tractability. However, the authors extend Rumelhart's model to address the progressive differentiation of conceptual knowledge in development and the progressive deterioration of conceptual knowledge in some forms of dementia.

As a whole, the book is well structured with a pleasant conversational style. Each chapter begins with a brief introduction outlining its aims, tying in the new topic with those discussed in earlier chapters, and ends with an overall summary. Key points are elucidated in each chapter and tangible examples are given to explain complex concepts. Relevant subheadings throughout the book help guide the reader through the text. The authors have a talent to predict the reader's questions and answers are often found in subsequent paragraphs.

The book begins with a short preface outlining the authors' perspective and introducing the key elements of their approach. The first chapter reviews findings and theories central to the field of semantic cognition (e.g., Carey, 1985; Collins & Quillian, 1969; Keil, 1989, 1991). It also introduces the intriguing observation that conceptual knowledge undergoes a progressive differentiation in development, resulting in the gradual elaboration of a conceptual hierarchy, and that this process is reversed in some forms of dementia, resulting in progressive loss of first the finest and later the coarser conceptual distinctions. Chapter 2 provides an overview of PDP theory, with a focus on the seminal work of Rumelhart and Hinton. In the third chapter, the authors describe preliminary simulations addressing the reasons for the observed patterns of development and disintegration of semantic knowledge. The subsequent chapters (4 through 7) in turn describe more targeted simulations to address specific issues including, preverbal conceptual development, naming ability, category coherence, and inductive projection. Each of these chapters conjoins experimental findings with results from the simu\.lations. Readers with minimal knowledge of computational modeling will appreciate the simplicity with which the simulations are described. Chapter 8 reviews the role of causal knowledge in semantic task performance, with a return to the discursive approach used in the opening chapter. In the final chapter, the authors discuss the key principles of their model in relation to the work presented in earlier chapters, and speculate on remaining questions, such as “how is PDP theory implemented in the brain?”

In describing their contribution to understanding human semantic cognition as “a small step in the right direction” (p. 380), I believe that the authors provide a modest description of their work. Indeed, they provide a much-needed mechanistic model of semantic cognition, filling an essential gap often overlooked in most theories of semantic cognition.

For example, although a number of theories of conceptual development have been put forward (Carey, 1985; 2000; Mandler, 2002; Quinn, 2002; Quinn & Eimas, 1997, 2000; Rakison, 2003), the authors of this book are the first to provide a mechanism for an infant's ability to associate different objects as belonging to the same category (Chapter 4). Here, they describe coherent covariation (Chapter 3); a process whereby changes are made to the connection weights in the network capturing shared properties among related items (e.g., can move is shared by all animals, and covaries with properties such as has wings and has feet). Using this process, the network rapidly learns shared properties, and thus forms general categories (e.g., animals). With time, and therefore development, categories become fine-tuned forming specific-level properties, such as, a canary is yellow. This progressive differentiation results from the gradual accumulation of weak error signals associated with unshared properties.

Building on previous theories, coherent covariation can account for the shift in development from general categories to specific-level categories, the acquisition of property salience (see Rakison, 2003), and provides a mechanism for the extraction of conceptual representations from perceptual experience (see Mandler, 1990, 2000a, 2000b, 2002). Although the authors suggest that there are no innate “core” properties on which category formation is based (e.g., Carey, 1985, 2000), they do suggest that the influence of coherent covariation on concept development depends on the infant's initial (and therefore, possibly innate) ability to detect similarities and differences among particular events.

This book will provoke a great deal of discussion and debate in the field of semantic cognition, and hopefully will generate new research ideas. With this in mind, I have no hesitation in recommending that you read this book.