Consider the complexity of the mental processes involved in reading. Our visual apparatus interprets a random set of squiggles as symbols representing the letters of a word. The arbitrary pairing of each word with its associated concept is decoded in our mental dictionary, and our grammatical processor assembles the words of the sentence so that we can establish who is doing what to whom. Because this process evolves over time, our working memory must retain critical aspects of these representations until the processing of the sentence is complete. Ultimately, the overall meaning of the sentence emerges.

One central theme of Steven Pinker's 1997 book, How the Mind Works [1], is the computational theory of cognition, according to which a massive number of simple processing devices work in concert to accomplish complex tasks. This view is outlined lucidly for several cognitive domains, including visual perception, spatial relations, and logical reasoning. Consider one example related to reading-the task of interpreting the visual squiggles that we call a letter. Remarkable flexibility is required to recognize all of the different fonts in all of the possible sizes in which letters can be written. Commuters appreciate that letters can be recognized upside-down, sideways, and in a mirror (with a little practice), despite the bouncing that newspaper readers endure in the subway. Letters are routinely written in radically different shapes (for example, "A" and "a"), and yet these shapes are mundanely recognized as tokens of the same letter with lightning speed and flawless accuracy. And I haven't even begun to talk about the recognition of your handwriting. Observations such as these suggest that we do not have in our minds a template corresponding to each of the letters of the alphabet. Instead, we may have a legion of highly specialized visual line detectors and orientation detectors that rapidly decode the squiggles, probably with the help of "top-down" contextual support, and the output of this process maps on to abstract mental representations of letters.

Cognitive neuroscientists have pointed out that these kinds of processes are not magical. Many details remain to be worked out, but we have begun to demonstrate some of the ways in which a network of specific regions of the brain works to subserve a complex process, such as letter recognition. The occipital lobe in the back of the brain contains the primary visual cortex, where retinal information is interpreted as a squiggle with a particular shape and orientation. Disease affecting this part of the brain can result in so-called cortical blindness. This area is intimately connected with visual association cortex that interprets the perceptual form of visual input. Injury to the visual association cortex can result in a visual agnosia, or difficulty interpreting a particular kind of visual information. For example, insult to the ventral occipital lobe results in "alexia without agraphia"; patients with this condition cannot recognize letters and words that they themselves wrote a minute earlier. Reciprocal interconnections integrate these areas with a region in the left posterior temporal cortex that is crucial for the mental representation of written word forms.

In The Language Instinct, published in 1994 [2], Pinker describes how we use letters and sounds to produce meaningful messages. This book negotiates the complex technical world of linguistics with exceptional clarity and a good dose of humor. Regardless of age, education, sex, or race, our constant creativity is truly astounding: I have not previously produced the sentence I am now writing, and it is unlikely that will I repeat this sentence again in my lifetime. How can I create so many millions of new sentences? Linguists such as Noam Chomsky have described the hierarchical, tree-like grammatical structures of sentences that allow us to construct and understand countless novel and coherent utterances. By "grammar," I do not mean to invoke your third grade teacher (and her threatening yardstick), drilling you with "i before e except after c." Instead, I refer to the rules that govern the manipulation of the tokens we use when we communicate linguistically. A word in a sentence is a member of a particular grammatical form class (for example, a noun or a verb). These words are grouped into phrases, which in turn can be grouped into larger phrases until a sentence is formed. Because English relies heavily on the order in which words and phrases occur in a sentence, the agent of an action typically maps onto the subject that precedes the verb, and the recipient corresponds to the object after the verb.

In this fashion, we can begin to establish who is doing what to whom: The sentence "Fran put the book on the shelf," for example, describes Fran, the agent, and the book that is being put. The verb at the core of each sentence also demands that certain "arguments" are arrayed around it. The verb put requires a "putter," a "puttee," and the target location of the puttee: Sentences such as "Fran put the book" or "Fran put on the shelf" sound odd because all of the arguments associated with "put" are not included. Our grammar system also features a series of mental rules that allow us to mentally manipulate the phrases of a sentence. Consider "mistakes were made," the all-too-popular political confession that leaves us wondering about the agent of the mistakes. We can move phrases around and use a "trace" as a place-marker for a role that is needed by a verb but is not stated in a sentence. "Trace made mistakes" thus might be the representation of the sentence we hold in mind (possibly until the next election) to remind us that someone really did err.

While our grammatical processor manipulates the tokens of a sentence, we are looking up the meanings of the words. This is no mean feat, given the huge corpus of words that we accumulate during our lives. Experts in language acquisition, such as Pinker, estimate that we learn a new word every 90 minutes that we are awake, yielding a vocabulary of about 60 000 words by the time we graduate from high school (Shakespeare used half as many), and we continue to acquire new words over our entire lives (consider e-mail, fax, and modem). It is likely that the concepts associated with words are not listed alphabetically in our mental dictionary but obey organizational principles of their own related to categorization. This allows us to negotiate a massive amount of information efficiently. Moreover, categorization allows us to make inferences about less familiar words and their concepts. For example, if birds are oviparous and if emus are birds, it is legitimate to infer the previously unknown fact that emus are oviparous.

Despite the potpourri of some 5000 languages used in the human village, important linguistic universals transcend superficial differences between these tongues. All languages have nouns and verbs, subjects and objects, words and phrases that can be manipulated and combined in a way that appears to be sensitive to the hierarchical grammatical structure so elegantly described by Chomsky. From this perspective, it is reasonable to expect that a network of brain regions contributes to these uniquely human linguistic processes. For example, higher-order perceptual processing regions that are located more anterior in the temporal lobe contribute to concepts. It is not clear whether these brain areas are crucial for recognizing the visual perceptual attributes of objects or actually represent categories of knowledge, but insult to these areas results in category-specific language deficits: The world of fruit, animals, and other natural kinds is inaccessible to some patients with such injury; other patients with a slightly different distribution of disease have difficulty naming tools, furniture, and other manufactured artifacts. Verbs seem to be vulnerable to prefrontal disease. Concepts are complex, and our knowledge of object and action concepts is not unimodal: An apple is associated with a crunchy sound and a sweet smell and taste, is eaten, and can be thrown at unruly authors. These bits of information are likely to be distributed in other modality-specific association regions of the brain, and an area in the posterior superior left temporal lobe may organize the distributed representation of information contributing to a concept. As far as grammatical processing is concerned, focal cerebral insult to the left frontal lobe often limits patients to speaking and understanding single, unconnected words, as if grammatical relations between words in a sentence cannot be established.

Sentences are often long and complex, and "working memory" is needed to retain specific aspects of a sentence for a brief period while the sentence is being processed. Daniel Schacter reviews this and other aspects of memory through a series of fascinating vignettes and comprehensive discussions in his 1996 book, Searching for Memory: The Brain, the Mind, and the Past [3]. Intensive work over the past two decades has indicated that memory is a multifaceted cognitive system. We are all familiar with the explicit request to retain three named objects for a minute and then report them back. This task uses a form of anterograde memory, with which we consciously attempt to learn and then recall specific information. Autobiographical memory-the recollection of specific events in our remote personal histories-is another form of explicit memory (remember that third grade teacher?). This can be contrasted with implicit forms of memory, with which we learn without the conscious goal of later recall, so that we may unconsciously recognize a word that we cannot specifically recall from a previously presented list. Implicit memory has been harnessed by advertising executives: We may rate an advertised product positively, even if we cannot recall having been previously exposed to the product or its advertising, because of the effect of advertising on implicit memory.

Distinctions among these forms of memory are supported by the selective forms of amnesia that follow insult to different brain regions. The crucial role in anterograde memory of one brain region-the hippocampus, deep in the temporal lobe-was discovered as an unforeseen side effect of treatment for intractable epilepsy in 1953: Difficulty learning new information after bilateral hippocampal excisions relegated the unfortunate patient to the life of an eternal "child of the fifties." Damage to the hippocampus also seems to explain some of the learning difficulties of patients with Alzheimer disease. Persons with amnesia who cannot recall three words presented to them a minute earlier nevertheless can have impressively preserved implicit memory. Some implicit forms of memory may be compromised in patients with intact hippocampal systems, a pattern of memory difficulty that has been associated with disease in visual association cortex. The prefrontal cortex plays an important role in dating and placing the events that constitute our autobiographical memory.

Pinker and many other cognitive neuroscientists have hypothesized that the computational basis for memory, language, and visual perception is the golden path to understanding human cognition. However, mounting evidence suggests that even exhaustive cataloguing of these cognitive processes will not yield an adequate account of human thought. We are subject to the frailties (and benefits) associated with emotion. Schacter has outlined in his book, for example, the important contribution of emotionally charged material in learning. The amygdala may be the neural substrate for learning about "fight-or-flight" situations. The 1994 book Descartes' Error: Emotion, Reason, and the Human Brain [4], by Antonio Damasio, has further emphasized the crucial role that emotional factors play in human cognition. Damasio describes patients who can voice rational decisions derived from intact memory, language, and visual processes but cannot live normal lives because they are unable to integrate an affective perspective into their decision-making processes. They cannot anticipate that their decisions have consequences, they cast responsibilities to the wind, and they cannot complete plans for the future, no matter how rational. Often, this will be accompanied by a change in personality and a form of disinhibition that makes patients unable to control their impulses. Damasio has related these changes to a portion of the frontal lobe. This frontal brain region appears to act as a higher-order interface with our body as well: That queasy sensation that occurs in the pit of the stomach when a discussion with a difficult coworker is anticipated may originate in this part of our brain.

Of course, I can provide only the briefest glimpse of the captivating observations reviewed in these four volumes. The books are authoritative and well documented for readers wishing to pursue a particular topic in detail, and yet they are extremely entertaining for the casual reader. The scope of each text is manageable, although How the Mind Works may be overly ambitious at times and often speculates about the role of Darwinian natural selection in the evolution of human thinking. The authors provide appropriate caveats about the limitations associated with the processes they describe. Schacter emphasizes that our memory can be surprisingly inaccurate and susceptible to suggestion under some circumstances: For example, given appropriate interrogation conditions, an eyewitness to a crime can be easily convinced to recall an event in a way that differs from the way in which it had apparently occurred. Pinker indicates that we do not exhaustively identify all of the possible interpretations of a sentence, as a computer might. Rather, as we encounter each word during sentence interpretation, we make calculated guesses that are based on our world knowledge and our (flawed) sense of probability. Damasio's text carefully describes the difficult anatomic concepts underlying emotion and reason, and the other three books might have benefited from greater emphasis on the neural substrates for these fascinating syndromes. Recent advances in monitoring the cerebral activity of healthy adults during cognitive challenges could have received greater attention. Perhaps the most important point underemphasized in these books about cognitive neuroscience is the potential for productive interplay between cognition and neuroscience: Theories about memory, language, visual perception, and reasoning can constrain hypotheses about the roles played by various brain regions, and our knowledge of regional brain functioning can be used to constrain the development of realistic cognitive theories.

With these minor shortcomings kept in mind, these four books outline the new frontier of cognitive neuroscience with commendable clarity. Those who forecast the computer-based dominance of the human world by such deadlines as "1984" or "2001" wildly misjudge the complexity of human thought. While being entertained with fascinating stories exploring the inner cosmos, readers of the four books reviewed here will be exposed to intriguing and captivating lessons about the cognitive and neural basis for complex human thought processes.