The conventional wisdom in psycholinguistics in the 1980s was that thought is like external language in all important respects. Each of us, the argument went, comes genetically equipped with a 'language of thought' that is reflected in the structure and organization of speech. Thought is not remotely similar to perception or imagery, or to the exercise of motor skills. The basic rules governing human thought and language were believed to be largely unique and substantially innate, the result of genetic novelty. Understand language, and — the psycholinguists used to say — you understand thought.

In this line of reasoning, languages relate to the world as follows: names denote, as Henry VIII denotes Henry VIII; type terms, such as planet, refer to the set of all actual planets. Reference, singular or general, is supposedly fixed when a single person first coins a word — for example, planet, while pointing to Jupiter. The proper scope of that term is then said to include all things that 'have the same nature' as Jupiter, where the relevant sameness relation is said to be fixed by physical factors (probably unknown). Were it not so, the story goes, I would not mean what you mean by planet, so communication would founder. Fortunately, says this argument, the ancient Greeks did mean exactly what I mean by planet, owing to one having cleverly dubbed Jupiter a planet. Unfortunately for this theory, the Greeks also called the Sun, but not the Earth, a planet.

This approach to word meaning is about as applicable to real meaning as 'Dungeons and Dragons' is to real life. Aptly ridiculed by critics as 'font-change semantics', the theory still has its disciples. Including Steven Pinker.

Indeed, it is essentially font-change semantics that Pinker defends and deploys in his latest engaging doorstop, The Stuff of Thought. He has revised a few features, but the core ideas — innateness ad libitum, and the quest for the nature of thought in the analysis of language — are intact. Like his earlier books The Language Instinct and The Blank Slate, The Stuff of Thought has very little to do with the stuff with which we think — namely, neurons.

In leaving neurons out of the story, Pinker is not alone. Jerry Fodor elevates ignorance of neuroscience to a methodological virtue, proclaiming, "If you want to understand the mind, study the mind ... not the brain, and certainly not the genes". His metaphor, embraced by some psychologists and philosophers, says that the brain is merely the hardware that happens to implement the cognitive software. Neurons and their connectivity are as irrelevant to understanding the nature of mental function as a computer's transistor configurations are to my using Powerpoint.

Advances in neuroscience and genetics during the past 30 years have put such thinking on the defensive. For one thing, extravagant claims about human uniqueness must deal with the discovery that humans have only about 28,000 genes, and differ from mice in just 300 or so. Additional constraints emerge with the discovery that human brains are stunningly similar to other mammalian brains — in components, connectivity, development, biochemistry and physiology. Topographic maps in the neocortex, cerebellum, spinal cord and subcortical structures are standard for representing and computing with neurons. As such, they suggest constraints relevant to semantics and reasoning.

Maps that represent which parts of the body are receiving what kind of stimuli are probably crucial to the very nature of our self-representation and in what we mean by self. The pathways connecting sensory maps to those representing motor preparation are likely to be important for reasoning what to do next. Meanings, as W. V. O. Quine realized some 50 years ago, fundamentally relate to the world not piecemeal (planet means planet) but through connected representational networks in the brain that, with varying accuracy, map as a whole on to reality. These are the maps that get us around the physical and social worlds.

And getting us around is the basic evolutionary rationale of nervous systems. Unlike plants that must take what comes, animals are movers. More sophisticated behaviour emerged with improved capacities to plan, predict and draw on past experience, which improved chances of surviving and reproducing.

This observation motivated neuroscientist Rodolfo Llinás, in his 2002 book I of the Vortex, to propose that, at bottom, thinking is the evolutionary internalization of movement. He meant that thinking is the generation in the brain of images of a future action, and its consequences. And generating these images depends on flexibility in categorizing the current problem as an instance of one kind of event rather than another, which, in turn, depends on memory for past experience. Fundamentally, thinking is neural activity in the service of behaviour (for example, should I flee or fight? Is this attacker weak or strong?). This almost certainly shapes thinking that seems detached from motor preparation (such as, where did Earth come from?).

As is so often the case in biology, discovering structure is crucial in coming to understand function — as in William Harvey's seventeenth-century revelation that hearts are actually pumps, not biological cauldrons for concocting animal spirits. To figure out how brains actually think and what reasoning really amounts to, we need to focus on understanding their many levels of organization, from neurons to large-scale systems to behaviour. If thinking is rooted in internalized movement, it may be more akin to a skill than to a syllogism. Language may not be the "stuff of thought" after all.