How the Brain Evolved Language [Hardcover]

If you've ever wondered how humans are able to learn, this book is for you. Even thought Don Loritz's book concentrates on language learning, the same cognitive model can be extended to learning in general. From the inner workings of a small set of neurons to the combined computational power of potentially billions (or more) neural connections, he builds on solid brain science and biology. Unlike other authors who approach neural network research from a purely engineering perspective, Loritz approach is grounded in physiology, and his assertions are largely supported by what we know about the brain's structure and how the emergent properties of mind are formed. I would highly recommend it: it's a nice change from the myriad of books which posit how the brain/mind work without even mentioning the word "neuron."

When I first read this book (or attempted to read it) a few years back, I got bogged down at several points in the mass of detail Loritz presents. I "finished" it, but I can't say I understood it. Loritz's narrative started off engagingly enough, and he sprinkled humor throughout the book in an effort to keep it light, but there was an incredible amount of detail. In retrospect, Loritz prepared the reader for every new concept, but he rarely says anything twice, so again and again I found myself having to go back to understand how I got to the current page. Loritz made a big deal about Stephen Grossberg, so I looked at some of his publications. They were even harder to read, but between the two of them I began to get a clearer notion of how brain and language might fit together, and after much rereading, I think I've got their basic arguments straight.

Then, a few weeks ago, I read Jeff Hawkins' book "On Intelligence". Hawkins presents many of the same arguments that Loritz and Grossberg do, but while Loritz presents them in intricate contexts of language and biology, (and Grossberg presents them in differential difference equations!), Hawkins (or maybe his professional science co-writer, Sandra Blakeslee) presents them in the context of more everyday experiences.

I wish I had read Hawkins first. It would have made Loritz and Grossberg so much easier. Loritz loses a star in my book because I didn't "get it" the first time, but if you harbor a suspicion that the brain has something to do with intelligence, if you read and liked Hawkins, and if you have any special interest in language at all, then you should read this book. There is definitely a "new" model of intelligence out there (is the neocortex really "new"??), and these books are definitely onto it.

In his 1998 book "Consilience: The Unity of Knowledge," biologist E. O. Wilson argues that all of the sciences fit together in a hierarchical structure. Physics forms the foundation of the edifice of science. Chemistry builds on physics-theories in chemistry must be consistent with what is known in physics, although phenomena may arise at the chemical level that are not necessarily predictable from physical laws. Likewise, biology must be consilient with chemistry, and psychology with biology. Wilson goes on to argue that no real progress will be made in the social sciences until the biological underpinnings of human behavior are recognized.

This is a stance that many social scientists would disagree with. Indeed, the so-called Standard Social Science Model argues that human behavior is not limited in any way by biology. Hence, social scientists need not take any of the natural sciences into consideration when building their theories.

Early in the development of generative linguistics, the general expectation was that this approach was consilient with biology and psychology. Chomsky talked of an innate language acquisition device and critical periods, and he argued that universal grammar was hard-wired in the brain. Furthermore, transformational-generative grammar was intended to be a psychological explanation of language processing, and it produced hypotheses that were testable by the methods of the newly developing field of cognitive psychology. However, after successive failures to confirm the predictions of generative linguistics, Chomsky and his followers retreated to the safety of the Standard Social Science Model.

However, minority of linguists has held on to the belief that, if linguistics is to be a legitimate science, its theories must be psychologically and biologically plausible. One such linguist is Donald Loritz, who, in his 2002 book "How the Brain Evolved Language," builds a strong case for the consilience of linguistics with biology.

A major issue for linguists attempting a brain-based explanation of language has to do with the distinction between parallel and serial processors. The brain is a massively parallel computational device. Although its processing speed is slow, the brain can accomplish complex computations quickly because it divides problems into component parts that are each processed simultaneously. Most human-made computers, on the other hand, are serial processors. They solve problems on step at a time, but because their processing speed is so fast, they can far outperform brains on many tasks.

Language is clearly a serial process. Phonemes are produced (roughly) one at a time; word follows word, sentence follows sentence. Yet, if serial-processed language is produced by a parallel-processing brain, then there is a gap that needs to be explained. In principle, a parallel processor can mimic a serial processor (and vice versa), and it is a general assumption among cognitive psychologists and artificial intelligence researchers that this is what the brain is doing when it processes language. However, Loritz challenges the need for positing a virtual serial processor in the brain. Instead, he argues that, even though language is a serial process, it has features that could only arise if it were produced by a parallel processor.

Take metathesis as an example. Metathesis is the switching of two elements in a series, and it is a ubiquitous phenomenon in language. Metathesis occurs at the phonological level in historic sound changes in words, such as "three" and "third," where the "r" and the vowel switch positions, and it also occurs in dialectal variants, such as those who say "ax" for "ask." The bulk of spontaneous speech errors involve metathesis, as for example the legendary spoonerism "our queer old dean" for "our dear old queen." We also find metathesis at the level of syntax, as for example in the inversion of subject and verb in questions and in the repositioning of agents and patients in passive sentences.

Metathesis has long been recognized as a serious problem for a serial language processor. In short, there is no way for a process that produces elements of a string one at a time to metathesize those elements. Generative linguists have proposed various transformation and movement rules to account for metathesis. However, it does seem strange to posit rules for producing errors. Furthermore, Loritz argues, all types of metathesis can be accounted for without the need for extra rules if we assume that language is processed in parallel at the level of the brain. In sum, language has a serial structure because of production limitations, namely that the speech organs can only produce one sound at a time, and not because it is processed that way in the brain.

Loritz's thesis is essentially this: language has exactly the kind of qualities we would expect if it were learned and produced by the cerebral cortex. In the middle chapters of the book, Loritz discusses adaptive resonance theory, a model of how the cerebral cortex stores and processes information. These chapters are laden with mathematical formulae and complex diagrams, and the lay reader will want to skim or skip these chapters. But in the final chapter, Loritz ties this all together with a compelling argument for what he calls adaptive grammar--a linguistic structure that is determined by brain structure.

Another interesting idea that Loritz presents is the role of the cerebellum in language production. The cerebellum is an evolutionarily ancient system in the brain, and it is responsible for coordinating fine motor activity, especially those requiring rhythmic movement, such as walking. Although it has been known since the last half of the nineteenth century that the cerebral cortex is involved in language processing, evidence for a linguistic role for the cerebellum is quite recent.

Spoken language output involves a rhythmic pattern of beats or stresses, although the precise pattern varies somewhat from language to language. The neural organization of the cerebral cortex allows for fast, flexible learning and remains relatively plastic through the entire life span of the organism, but the cerebellum, on the other hand, is a slow learner and becomes set in its ways early in life.

This difference in learning styles is relevant to second language acquisition. It is generally agreed that children are better second language learners than adults, but Loritz points out that the real difference is in the fine points of pronunciation and morphology. Loritz argues that these require input from the cerebellum, which, once set early in childhood, cannot easily adapt to new patterns. Thus, while the flexible cerebral cortex learns the second language's general rules of pronunciation and grammar, the finer points are beyond the grasp of the inflexible adult cerebellum.

"How the Brain Evolved Language" is a tour de force of scientific consilience. Loritz brings to bear evidence from computational science, micro- and macro-neuroanatomy and evolution to provide an explanation for the structure of language that is firmly grounded in what is already known in the natural sciences. The book is thought provoking and a must-read for anyone interested in language issues.