[ CogSci Summaries home | UP | email ]
http://www.jimdavies.org/summaries/

Luger, G. F. (1994) Chapter 13: Language representation and processing. In Cognitive Science: The Science Of Intelligent Systems. Academic Press, San Diego, CA


Author of the summary: Debajyoti Pati, 2000, gte811q@prism.gatech.edu

The study of language is being addressed in numerous discipline including anthropology, education, philosophy, neurolinguistics, AI etc. The most influential studies, however, originate from linguistics and psycholinguistics. In linguistics, once again, the direction of inquiry differ between the theoretician and the empiricists. In general, linguistic studies concern themselves in providing explanations for the working of the language as a whole (as against individual performance), where the theoreticians focus on discovering sets of rules that could generate all possible (meaningful and grammatical) utterances in a language. The empiricists, on the other hand, make use of data collected from field settings along with various analytical and statistical tools to discover properties common to many languages (universals). While the linguistic studies concentrate on describing and accounting for regularities in tacit knowledge of native speakers (competence), psycholinguistics concern themselves in describing how the knowledge is represented and used during comprehension and production (performance). Earlier studies in psycholinguistics were based on Chompsk"s linguistics theory of transformational grammar, and focused on comprehension and language development of adults as well as children. The later periods witnessed the development of theories within the discipline. Typically, psycholinguistic studies focus on four levels of language; sound, meaning, syntax and pragmatics.

The origin of human language: While scientists are unsure about the actual origin of human language, a number of speculative hypothesis have been propounded in the past. One theory by Philip Lieberman (1984) proposed that language developed owing to the lengthening of the pharynx, while others point at a possible gestural origin. Derek Bickerton"s (1990) proposal of that language appeared abruptly after the development of protolanguage has been a controversial hypothesis. There is common agreement, however, that human language is unique owing to four major characteristics; semanticity, arbitrariness, displacement, and productivity. The first important characteristic is that language convey meaning. The arbitrariness associated with the relation between the signifier and the signified (the form-referent relationship) means that humans must possess a large and complex symbol system in order to use language. While the presence of iconic words including onomatopoeia and phonetic symbolism may seem contradictory to the idea of arbitrariness, study of the American sign language as well as language learning in children prove otherwise. Moreover, the presence of abstract words and concepts necessitates arbitrariness. Displacement, the ability to convey information across time is another unique feature, which is one of the crucial differences between the human language as compared to language in other species. Finally, human language has the capability of productivity; to produce novel utterances, that have never been said before. The possibility of infinite productions was one of the main arguments against Skinner"s (1957) operant learning theory. Language comprehension and processing, owing to these unique features, is an enormously complicated tasks.

The first major focus area in psycholinguistic studies is "sound" (phonetics and phonology). The stimulus is analyzed from two viewpoints; production (articulatory phonetics) and perception (acoustic phonetics). In general phonology is concerned with describing rules used to combine sounds into permissible sequences. One"s knowledge of phonological rules help in segmenting spoken output. The basic unit of phonology is the phoneme, which could have variants (allophones). One important difference in sound is caused by the voice onset time (VOC), and small differences in sound could result in different meanings. Differences that are not phonemic in one language could cause meaning difference in another. Range of possible sounds is assigned to different meanings in different languages. Speech perception deals with how humans decode and understand sound waves. While phoneme and VOC have been shown to play certain role in perception, most of the research have focused on finding whether speech perception is different from other acoustical input. A major finding termed categorical perception suggest that humans impose certain stages of processing of speech sound which aids on ignoring irrelevant variations and focus on the meaningful aspects of the signal. Comprehension of words in connected speech presents similar problem in another level. Logically, listeners should be able to identify individual words in speech to be able to refer to the mental lexicon. An additional problem is caused by the fact that the same intended phoneme could be produced in more than one way. Morton & Long (1976) and Warren (1985) proposed the notion of top-down approach. The top-down approach is contextual and knowledge driven, as compared to the bottom-up (Liberman & Mattingly, 1985), which is stimulus and data driven. The author"s views an interactive approach, where perception rely on both the processes.

Meaning (morphology and semantics) constitute the second important focus area in research. Most semantic theories are concerned with the relationship between meanings in the lexicon (lexical semantics). Another area of study is at the sentence level, where the study focus on describing meaning relations expressed in entire sentences, by studying the relationship between its words. Morphology, on the other hand, deals with the study of the structure of words. The smallest unit is word structure are the morphemes. The next larger units include plural morphemes (allomorphs). Other kinds of morphemes include the free morpheme, and bound morpheme. Some of the important bound morphemes include inflections and derivational morphemes. Scientists argue that derivational processes ( and its reverse) are used during perception and production. This leads to the argument that humans must possess a mental lexicon where stems and sets of bound morphemes are stored together, which are then combined in various manner as needed. An alternative view suggests the storage of high frequency words in complete form, and low frequency ones in derivative form. There are several theories accounting for meaning. The more important ones include the referential theory, feature representation, network representation, and prototype theory. A major area of interest in the study of meaning deals with lexical ambiguity, particularly how people represent words with multiple meanings. There are two basic theoretical positions concerning the processing of ambiguities. One view suggests that the meanings of ambiguous words are stored separately, while the other view suggests that all meanings of such words are accessed together. The role of context in meaning has been a topic of debate, and some views argue that the meanings are stored together and accessed automatically, where appropriate types of relation between the word and sentence context may result in selective access of meaning.

The study of syntax (the third major focus area) deals with explaining the structure of grammatically acceptable sentences. Linguists believe that native speakers possess tacit knowledge of competence, although they may not be aware of fit. In attempts to describe phrase and phrase structure, researchers have developed rule systems involving noun phrases, verb phrases, prepositional phrases, etc. (phrase structure rules), which along with lexical insertion rules provide some understanding of sentence structures. The power of recursion provides the way to create infinite number of sentences based on a small set of rules.


Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)
Last modified: Wed Apr 26 11:14:25 EDT 2000