Thomas Lengyel, "The Role of Arbitrariness in Psycholinguistic Theory" (1972).


In grammar, a rule that applies only in certain specified contexts. Adjective: context-sensitive.

A context-free grammar is one in which the rules apply regardless of context.

See also:

Examples and Observations:

  • "In the use of natural language, the truth value of a sentence can depend upon the context of its utterance: this is most evident in context-sensitive aspects of a language like tense and the use of personal pronouns."
    ("Formal Logic and Modal Logic." The Linguistics Encyclopedia, 2nd edition, ed. by Kirsten Malmkjaer. Routledge, 2004)
  • Phrase-Structure Grammar
    "There are different types of phrase-structure grammar. Context-free grammars contain only rules that are not specified for particular contexts, whereas context-sensitive grammars can have rules that can only be applied in certain circumstances. In a context-free rule, the left-hand symbol can always be rewritten by the right-hand one regardless of the context in which it occurs. For example, the writing of a verb in its singular or plural form depends on the context of the preceding noun phrase."
    (Trevor A. Harley, The Psychology of Language: From Data to Theory, 2nd edition. Psychology Press, 2001)
  • Word Meanings and Context Sensitivity
    "All linguistic morphemes are context-sensitive in the way that their semantic value depends partly on their semantic environment (tender does not have the same meaning in a tender steak and in a tender man) but transcategorial morphemes have a particular property: they are also syntactically context-sensitive. This means that their morphosyntactic status depends on their position inside the utterance and on their syntactic environment: for instance, when English now is used after a verb, it functions as a temporal adverb, while before a clause it functions as a subordinating conjunction."
    (Stéphane Robert, "The Challeneg of Polygrammaticalization for Linguistic Theory." Linguistic Diversity and Language Theories, ed. by Zygmunt Frajzyngier, Adam Hodges, and David S. Rood. John Benjamins, 2005)

    "[A]ll cognitive-functional research accepts axiomatically that '[word] meaning is highly context-sensitive, and thus mutable' (Evans 2005:71). The question for a linguist faced with such an unstable object of study is not how to render sense variation more stable, but how to reveal structure in its variation. Zelinsky-Wibbelt (2000) poses this fundamental question for the study of polysemy: 'Is polysemy a case of lexical representation or rather . . . a case of contextual differentiation?' (Zelinsky-Wibbelt 2000:114). . . . Theoretically in line with Tyler and Evans's (2001) premise, Zelinsky-Wibbelt phrases the problem as a methodological question: in the description of polysemy, 'what should be represented at the level of the lexicon and what should be computed by contextual functions?' (Zelinsky-Wibbelt 2000:145).

    "At play here is the role of real-time processing versus learnt-automated structure. Necessarily, the former is how one deals with entrenched meaning production and the latter with entrenched meaning structure."
    (Dylan Glynn, "Polysemy, Syntax, and Variation: A Usage-Based Method for Cognitive Semantics." New Directions in Cognitive Linguistics, ed. by Vyvyan Evans and Stéphanie Pourcel. John Benjamins, 2009)
  • A Functional Perspective on Language
    "Much of the complexity of NLG [natural language generation] arises out of the fact that producing language is a knowledge-intensive, flexible, and highly context-sensitive process. This context sensitivity reveals itself best when we consider connected texts rather than isolated sentences. Consider the following example. Suppose you were to express the idea: [LEAVE (POPULATION, PLACE)]. It is instructive to watch what happens if one varies systematically the expression of the different concepts leave, population, and place by using either different words (abandon, desert, leave, go away from in the case of the verb, and place or city in the case of the noun), or different grammatical resources: e.g. a definite description ('the + N'), possessives ('yours,' 'its'), etc. Concretely, consider the alternatives given [below].
    'X-town was a blooming city. Yet, when hooligans started to invade the place, [insert one of: (a)-(e) [below]. The place was not livable any more.'

    (a) the place was abandoned by (its/the population)/them.
    (b) the city was abandoned by its/the population.
    (c) it was abandoned by its/the population.
    (d) its/the population abandoned the city.
    (e) its/the population abandoned it.
    The interested reader may perform all the kinds of variations mentioned above and check to see what extent they affect grammaticality (the sentence cannot be uttered as finished), clarity (some pronouns will create ambiguity), cohesion, and rhetorical effect. In particular, while all the candidate sentences we offer in (a)-(e) are basically well-formed, each one has a specific effect, and not all of them are equally felicitous. Some are ruled out by poor textual choices (e.g. in (a) 'the place' is suboptimal, since it immediately repeats a word), others because of highlighting the wrong element, or because of wrong assignment of the informational status (given-new) of some element (e.g. in (d) 'the city' is marked as 'minimal' new information, while actually it is known, i.e. old information). Probably the best option here is (c), since this preserves the given-new distribution appropriately, without introducing potentially ambiguous pronouns. . . .

    "Getting a text 'right' is therefore a major problem."
    (John Bateman and Michael Zock, "Natural Language Generation." The Oxford Handbook of Computational Linguistics, ed. by Ruslan Mitkov. Oxford University Press, 2004)

    Also Known As: context sensitiveness, context-restricted