Copyright © 1999 Massachusetts Institute of Technology
A linguistic unit is said to be ambiguous when it is associated with more than one MEANING. The term is normally reserved for cases where the same linguistic form has clearly differentiated meanings that can be associated with distinct linguistic representations. Ambiguity is thus distinguished from general indeterminacy or lack of specificity.
Ambiguity has played an important role in developing theories of syntactic and semantic structure, and it has been the primary empirical testbed for developing and evaluating models of real-time language processing. Within artificial intelligence and COMPUTATIONAL LINGUISTICS, ambiguity is considered one of the central problems to be solved in developing language understanding systems (Allen 1995).
Lexical ambiguity occurs when a word has multiple independent meanings. "Bank" in the sentence "Jeremy went to the bank" could denote a riverbank or a financial institution. Ambiguous words may differ in syntactic category as well as meaning (e.g., "rose," "watch," and "patient"). True lexical ambiguity is typically distinguished from polysemy (e.g., "the N.Y. Times" as in this morning's edition of the newspaper versus the company that publishes the newspaper) or from vagueness (e.g., "cut" as in "cut the lawn" or "cut the cloth"), though the boundaries can be fuzzy.
Syntactic ambiguities arise when a sequence of unambiguous words reflects more than one possible syntactic relationship underlying the words in the sentence, as in:
In (1a), the ambiguity lies in whether the adjective smart modifies (provides information about) both women and men resulting in a practice not to hire unintelligent people of either sex, or whether smart modifies women only. In (1b), the phrase with the knife could be used to describe the manner in which the burglar threatened the student, or to indicate which student was threatened. Chomsky (1957) used similar ambiguities to argue for the necessity of abstract syntactic structure.
The different underlying relationships in ambiguous sentences can frequently be observed directly by manipulating the form of the ambiguous sentence. When the order of the string smart women and men is reversed to men and smart women, the sentence can only be understood as involving modification of women but not men. When the adverb wildly is inserted before with the knife, only the reading in which the burglar is using the knife remains possible.
In its most technical sense, the term ambiguity is used to describe only those situations in which a surface linguistic form corresponds to more than one linguistic representation. In lexical ambiguities, one surface phonetic form has multiple independent lexical representations. For syntactic ambiguities, one surface string has different underlying syntactic structures. A subtler and more controversial example is the phenomenon of scope ambiguity, exemplified in:
In (2a), the sentence can be understood as referring to a single woman who tolerates each and every man, or alternatively, it can mean that every man is tolerated by at least one woman (not necessarily the same one). Sentence (2b) can mean either that John believes that the King of France is not bald, or that John does not hold the particular belief that the King of France is bald. It is difficult to find clear syntactic tests for scope ambiguities that would demonstrate different underlying structures. For instance, reversing the order of some woman and every man does not eliminate the ambiguity (although it may affect the bias towards one reading):
May (1977) argues that sentences such as (2) reflect different underlying structures at a level of linguistic representation corresponding to the LOGICAL FORM of a sentence. Subsequently, much of the linguistic literature has considered scope ambiguities as genuine ambiguities.
A broader notion of ambiguity includes a pervasive ambiguity type that involves not multiple possible structures, but rather, multiple associations between linguistic expressions and specific entities in the world. The sentence in (4) is an example of referential ambiguity:
The ambiguity resides in the understanding of the pronoun he, which could refer to either Mark, Christopher, or some other salient entity under discussion.
Language processing necessarily involves ambiguity resolution because even unambiguous words and sentences are briefly ambiguous as linguistic input is presented to the processing system. Local ambiguities arise in spoken language because speech unfolds over time and in written language because text is processed in successive eye fixations (cf. Tanenhaus and Trueswell 1995).
The sentence in (5) illustrates how globally unambiguous sentences may contain local ambiguities.
The pupil is the object of a relative clause. However, the underlined sequence is also consistent with the pupil being the subject of a main clause, as in "The pupil spotted the proctor." The ambiguity arises because the morphological form "-ed" is used for both the simple past and for the passive participle, illustrating the interdependence of ambiguity at multiple levels.
Laboratory studies have established that multiple senses of words typically become activated in memory with rapid resolution based on frequency and context. For example, when "pupil" is heard or read, both the "eye part" and the "student" senses become briefly active (Simpson 1984). Similarly, "elevator," "elegant," and "eloquent" are briefly activated as the unambiguous word "elephant" is heard because they are consistent with the initial phonemic sequence "eluh" (Marslen-Wilson 1987).
Syntactic ambiguities exhibit consistent preferences. Readers and listeners experience processing difficulty and sometimes a conscious feeling of confusion when the sentence becomes inconsistent with the preferred structure. The example in (6), from Bever (1970), is a classic example of a so-called garden-path, illustrating that the main clause is the preferred structure for the main clause/relative clause ambiguity.
In (5), resolution in favor of the relative clause does not cause conscious confusion. Nonetheless, processing difficulty at by the proctor can be observed using sensitive measures, for instance the duration of eye-fixations in reading. Theoretical explanations for syntactic preferences can be roughly divided into structural and constraint-based approaches. In structural theories, principles defined over syntactic configurations determine an initial structure, which is then evaluated, and if necessary, revised. Different principles may apply for different classes of ambiguities (e.g., Frazier 1987). In constraint-based theories, preferences arise because a conspiracy of probabilistic constraints, many of them lexically based, temporarily make the ultimately incorrect interpretation the more likely one (MacDonald, Pearlmutter, and Seidenberg 1994; Tanenhaus and Trueswell 1995). The example in (7) illustrates a sentence with the same structure as (6) in which the probabilistic constraints initially favor the relative clause because "buried" is typically used as a passive participle without an agent, and land mines are more typically themes than agents of burying events. Whether or not constraint-based systems can provide a unified account of ambiguity resolution in language using general principles that hold across other perceptual domains remains a central unresolved issue.
Allen, J. (1995). Natural Language Understanding. Redwood City, CA: Benjamin/Cummings.
Bever, T. (1970). The cognitive basis for linguistic structures. In J. R. Hayes, Ed., Cognition and the Development of Language. New York: Wiley.
Chomsky, N. (1957). Syntactic Structures. The Hague: Mouton.
Frazier, L. (1987). Sentence processing: a tutorial review. In M. Coltheart, Ed., Attention and Performance XII: The Psychology of Reading. London: Erlbaum.
MacDonald, M., N. Pearlmutter, and M. Seidenberg. (1994). Lexical nature of syntactic ambiguity resolution. Psychological Review 4:676-703.
Marlsen-Wilson, W. D. (1987). Functional parallelism in spoken word recognition. Cognition 25:71-102.
May, R. (1977). The Grammar of Quantification. Ph.D. diss., MIT. Distributed by the Indiana University Linguistics Club, Bloomington.
Pritchett, B. (1992). Grammatical Competence and Parsing Performance. Chicago: University of Chicago Press.
Simpson, G. (1984). Lexical ambiguity and its role in models of word recognition. Psychological Bulletin 96:316-340.
Small, S., G. Cottrell, and M. Tanenhaus, Eds. (1988). Lexical Ambiguity Resolution: Perspectives from Psycholinguistics, Neuropsychology, and Artificial Intelligence. San Mateo, CA: Morgan Kaufmann Publishers.
Tanenhaus, M., and J. Trueswell. (1995). Sentence comprehension. In J. Miller and P. Eimas, Eds., Handbook of Cognition and Perception. New York: Academic Press.
Zwicky, A., and J. Sadock. (1975). Ambiguity tests and how to fail them. In J. Kimball, Ed., Syntax and Semantics, vol. 4. New York: Academic Press.
Copyright © 1999 Massachusetts Institute of Technology