Beginnings of a storey

The fields of cognitive science can be traced back at least as far as the Hixon symposium on cerebral mechanisms in behaviour held at the California Institute of Technology in 1948. Reaching public prominence in the 1970s, this consortium of psychology, philosophy, linguistics, anthropology, neuroscience and computer science raised high hopes for settling age-old questions: the relation of brain to mind, the nature of consciousness and thought, the acquisition of knowledge and language. Despite impressive successes such as the design of programmes that could beat all but the grandmasters at chess, the achievements fell short. To illustrate this, the editors of this book invoke a Chinese folktale about a man who envied his rich neighbours and their two-storied house. Having hired a contractor to build a house as tall as theirs, he was appalled to find the builder first framing the ground floor of the house. How could he waste money on a ground floor when what was wanted was a second storey?

Cognitive science has been dominated by a computationalist approach that took language and logic as the dominating features of human mentality, and has gone after sophisticated problem-solving, theorem-proving, and grammar-mastering. But these, say Herbert Roitblatt and Jean-Arcady Meyer, are based on humbler functions such as sensory perception, sensory-motor coordination and categorical discrimination. Unlike full-blown linguistic ability, these functions are without doubt shared with other animals and hence are open to comparative approaches. In addition, the biomimetic stance comprises the design of computer simulations or robots informed by observations of how animals cope with their worlds. This "animat" approach, behavioural studies, neuropsychology, and philosophy are represented in this book, which aims to provide the foundation for a new cognitive science.

The 20 chapters are grouped in six sections: introductory issues, intentions and the organisation of behaviour, representation, memory and attention, communication, motivation and emotion. Meyer's introductory review of the history of animats is complemented by Michael Dyer's "Toward the acquisition of language and the evolution of communication". Dyer has succeeded in designing a system that can acquire language of a sort, that can register, describe, and remember content and events in a simple environment called Blobs World. He has also populated an artificial environment, "Bioland", with different species of simple "biots" who interact in ways favouring the evolution of intraspecies signalling. Although crude, their creation may well contribute to filling the "enormous gaps in our theoretical understanding of the neurodevelopment and neuroevolution of both animal and human cognitive systems''.

The other side of the biomimetic initiative envisaged by Roitblatt and Meyer, the comparative ethological approach to cognition, is illustrated in the chapter on "Language and animal communication: parallels and contrasts" by Christopher Evans and Peter Marler. They take three basic features of human language: partitioning of acoustic continua to yield categorical perception of vocal utterance, functional referentiality or the encoding and decoding of information about external events, and the modulation of communication behaviour as a function of social context. Experimental evidence for all three is elegantly laid out for various birds (a cock produces significantly more alarm calls when a hen and not a quail sits next door) and mammals (acoustically distinct vervet monkey alarm calls correspond to distinct classes of predators - mammalian carnivores, raptors, and snakes). While these results imply either phylogenetic antiquity or evolutionary convergence, the authors could find no nonhuman cases of lexical syntax, and only very limited capacities for making reference to past or future.

The comparative work which thus reveals both continuities and discontinuities between human language and animal communication, is further exemplified by Roger K. R. Thompson's account.. He presents evidence that animals categorise their world on the basis of perceptual, relational similarity, discusses the mirror experiments, and mentions the problem of understanding the concept of a mirror in the common interpretation which reserves the concept of self for apes. He arrives at the conclusion that there is more than a grain of truth in the despair Andersen's Ugly Duckling experienced when first inspecting his reflection. Finally, Marc Bekoff in his chapter on "Cognitive ethology and the explanation of nonhuman animal behaviour" cites the ethical issues of animal rights.

Philosophy, the third leg of cognitive science, is represented in the chapters by Margaret Boden, Colin Allen, and Daniel Dennett. All turn on questions of definition. Boden compares several concepts of creativity and settles for the kind of novelty resulting from change in one of the generative rules constituting the relevant conceptual domain. In her view, animals lack the required conceptual complexity, and thus cannot be creative, while geniuses, computers, and robots can. This attribution hinges on her definition, which prevents the invention of tool use by a macaque from being regarded as creative. Reminiscent of the equation of "consciousness" with "use of human language", which precludes anyone not possessing human language from being conscious, this definition appears to aim at the second storey of the Chinese house. Allen discusses the uses of "intentionality", and Dennett focuses on "belief". Belief obtains wherever action is influenced by information. We can thus invoke belief wherever it helps to explain what an animal is up to, but the term is stripped of its discerning power.

Several chapters, notably that of Catherine Thinus-Blanc on "Spatial-information processing in animals", deploy the notion of cognitive maps. In contrast, Thomas Bourbon, in his piece on "Perceptual control theory" uses computer simulations to demonstrate how movement can be precisely guided and coordinated without recourse to a map or control by a command system, but by keeping sensory input within the limits set by some referencing value. He distinguishes between behaviour as seen by an outside observer, and what an animal experiences itself doing.

Some chapters have overlapping themes making for a dialogue. Julie Neiworth's ambitious article on "The integration of content with context: spatiotemporal encoding and episodic memories in people and animals" deals with different kinds of memory, and invites comparison with Jean Delacour's model of the brain and its memory systems. A yet more explicit confrontation is between the back-to-back chapters on motivation by David McFarland and Fred Toates. McFarland mounts a case against assuming goal-directedness in the control of behaviour. Instead of referring to an internal representation of a goal in the guidance of action, decisions about what to do are determined by trade-off considerations which obviate reliance on a represented goal. In this analysis, however, no reason is given as to why goal-directedness could not implement a trade-off decision; after all, trade-off considerations of fuel economy and comfort affect the setting of a thermostat although the device still works by negative feedback. Toates makes a similar point, finding goal-directedness to be essential to much motivational control, and presents a useful comparison of cognitive and stimulus-response learning.

The book shows what comparative approaches to cognitive science have to offer. The field is still a ragbag of new and old. As William James said of the psychology of his day: "At a certain stage in the development of every science a degree of vagueness is what best consists with fertility."

Petra Stoerig is professor of medical psychology, Ludwig-Maximilians-University, Munich. Colin Beer is professor of psychology, Rutgers University, New Jersey.