T. Winograd, Language as a Cognitive Process, Volume I: Syntax

1985

"Does this butcher knife handle frozen foods, too?" There was a time when children in grade school were required to parse, orally and on paper. Today, we study how to program computers to parse sentences in natural language and write books to introduce beginning graduate students to the complexities of computational linguistics. This change cannot be ignored when reviewing Terry Winograd's solid tome. He wanted it to be several things: textbook (including "detailed technical material and exercises designed to help the student master a body of concepts and techniques"); practical guide ("which should be useful in both the design and testing of systems" that deal with natural language); and reference source ("with many pointers to the literature of both linguistics and computer science"). Common sense requires one to consider a book in respect to the aims of its author. I doubt, however, that saying (and proving) that this is an excellent textbook and/or a successful practical guide and/or a comprehensive reference source is of any significance. Winograd is far from being a novice in the field. He is rightly credited for many innovative ideas (since his graduate studies at MIT in the AI laboratory) and deserves a challenging discussion of his work. Mentioning that parsing used to be a classroom exercise, I intended to put the subject of the book in a cultural perspective. And I do so because the book lacks such a perspective. The two sections, "What every language user knows" and "The evolution of linguistic science" are but minimal reference alibis. The cultural perspective is necessary in order to understand why we ask questions about the syntactic level of language, and how such questions have changed through the years. The change of paradigms, which Winograd tends to embrace (although aware of Kuhn's rather sketchy model of The Structure of Scientific Revolutions), explains changes in the conceptual framework that scientists Artificial Intelligence 27 (1985) 353-361 0004-3702/85/$3.30 O 1985, Elsevier Science Publishers B.V. (North-Holland)

Cognitive linguistics began as an approach to the study of language, but it now has implications and applications far beyond language in any traditional sense of the word. It has its origins in the 1980s as a conscious reaction to Chomskyan linguistics, with its emphasis on formalistic syntactic analysis and its underlying assumption that language is independent from other forms of cognition. Increasingly, evidence was beginning to show that language is learned and processed much in the same way as other types of information about the world, and that the same cognitive processes are involved in language as are involved in other forms of thinking. For example, in our everyday lives, we look at things from different angles, we get up close to them or further away and see them from different vantage points and with different levels of granularity; we assess the relative features of our environment and decide which are important and need to be attended to and which are less important and need to be backgrounded; we lump information together, perceive and create patterns in our environment, and look for these patterns in new environments when we encounter them. As we will see in this volume, all of these processes are at work in language too.

Journal of Linguistics, 2003

At a recent conference a colleague described a conversation that he had had with a high school student. The student proclaimed, with an obvious sense of satisfaction and excitement, that he had recently completed Mark Baker's Atoms of language; he then asked somewhat more sheepishly, as though he might be probing more deeply than should be done in polite company, 'Do YOU believe in parameters?' The student's reaction nicely captures three notable qualities of Baker's book. First, Baker has authored a monograph that is eminently readable, even for the novice in linguistics. Second, the content of the book is sufficiently intricate and well discussed that the book is a satisfying intellectual experience (and this even for the professional linguist). Finally, while Baker succeeds in raising some of the Big Questions that lie beneath structural similarities and variations in language, he falls short of providing an apologetic strong enough to sway skeptics to his linguistic worldview. Undoubtedly, he will also cause some offense to readers who already believe very differently about how linguists should begin to confront the Big Questions. Baker identifies a core task of modern linguistics (perhaps the core task?) to be solving the 'Code Talker Paradox'. The paradox, simply stated, is: How is it possible that a language such as Navajo can be similar enough to English to translate and transmit complicated messages, yet be so dissimilar from Japanese (or English for that matter) that the use of Navajo by American forces in World War II baffled Japanese cryptographers ? Baker is optimistic that a decipherment of the paradox is made possible by applying the concept of a parameter, i.e. a structural variable which is found in all languages and which can have one of two values in a particular language. For example, the NULL SUBJECT PARAMETER governs whether overt subjects are required in a finite clause. Some languages have one value for the parameter such that subjects are required (e.g. English). All other languages will carry the opposite value for this parameter such that subjects need not be expressed overtly (e.g. Italian). With a full complement of parameters, rather significant morpho-syntactic differences, even on the order of what distinguishes Navajo and Japanese, can arise between languages. On the other hand, since the parameters are constant across languages, and are presumably coupled together with other universal principles of language structure, there is also an account of the commensurability of languages. Paradox resolved. Of course, this is not a new idea in linguistics; rather, it is a well-articulated version of Principles & Parameters Theory, which can be traced back two decades (Chomsky 1981). What is novel about The atoms of language is that the concept of parameters is given more thorough treatment here than in any other previous work that comes to mind. Not only does Baker gather together putative parameters that have been individually proposed in the literature (from both formalist and functionalist sources), he also explores the relationships which obtain among them. Baker makes five crucial assumptions about parameters and how they interact to determine language structure: 1) the overall number of parameters is a small proportion of observable differences in languages because 2) a series of consequences for language structure can be traced to a single parameter; thus, 3) knowing the value for a particular parameter allows the linguist to make a set of predictions about what the morpho-syntax of a language will look like. 4) Parameters are not independent properties of language, but form a hierarchical system with some parameters being logically prior to others. Therefore, the settings for some parameters will determine whether other parameters are relevant or irrelevant. 5) Logically prior parameters will be learned first.