Showing papers by "George Lakoff published in 1973"

Hedges: A Study In Meaning Criteria And The Logic Of Fuzzy Concepts

Abstract: Logicians have, by and large, engaged in the convenient fiction that sentences of natural languages (at least declarative sentences) are either true or false or, at worst, lack a truth value, or have a third value often interpreted as ‘nonsense’. And most contemporary linguists who have thought seriously about semantics, especially formal semantics, have largely shared this fiction, primarily for lack of a sensible alternative. Yet students of language, especially psychologists and linguistic philosophers, have long been attuned to the fact that natural language concepts have vague boundaries and fuzzy edges and that, consequently, natural language sentences will very often be neither true, nor false, nor nonsensical, but rather true to a certain extent and false to a certain extent, true in certain respects and false in other respects.

Fuzzy Grammar and the Performance/Competence Terminology Game

Abstract: Fazey §rae§ eae tie §§?f§f§§§§€§§§§§€§$§§€ ?er§en§2egy Gene eeerge Leksff fieeeereizy ef €eE:§e:nie, Berkeieg I; Fuzzy Qreeger Anycne who has taught or taken an intrceuctsry eyntax course can tell you that sneakers £5 not always mas clear er uniform Judgments abeut whether sentences are well—foree er what they mean. Almost every syntactic er semantie phenomenen has n ehaeowy area in which speakers beccme anciear eith resgect to juegmenfie about meaning and vell~fermeeness. For the past few years, this phenomena has been studiefi intensi?el§ by 3033. His general results are as follows: (1) Rulea of grammar ec no? simyly ayply or fail to apply; rather they apply to a degree. (ii).Grammatical elements are not simply members or nonmembers of grammatical categories; rather they are members to a degree. (111) Grammatical cwnstructions are not simply ielanda or nonielands; rather they may be islands to a degree. (iv) Grammatical conatructions are not simply environments or non» envircnmente for rules; rather they may be environments to a degree. (v) Grammatical phenmmena form hierarchies which are largely constant for speaker to speaker, and in many cases, from language to language.‘ (vi) Different speakers (and eifferent languages) will have different acceptability threshcles along these hierarchies. Ross has made these claims in the absence of a theory of fuzzy grammar; no current thesry cf grammar can even begin to accomodate the facts that Kass has observefi. What I voule like to do in this paper is ask some fietailee questions about how current theories of grammar might be changee to accemedate even the most basic of Ross‘ observations“ 3 weuld like ta make ciear at the outset that I do not have a theory of fuzzy grammar ta propese; the beat 1 can do is make extremely tenfiative suggesfiiens for beginning to eeal with a handful of Bees’ §hencmena and ta yreviee a framework for further ingniry.

Some Thoughts on Transderivational Constraints

Abstract: HUB (5) The boys scratched their arms and so did the girls. (scratch their arms) (6) John scratched his arm and so did Mary. (scratch her arm or scratch his arm) (7) I scratched my arm and so did Bill. (scratch his arm or scratch my arm) There are two principal dialects corresponding to these sentences. Dialect I: All of H, 5, 6, and 7 are ambiguous. Dialect II: 4 and S are ambiguous. 6 and 7 have only the scratch his arm and scratch my arm readings, respectively. The difference between the dialects appears to involve the definition of constituent identity. In dialect II, the pronouns must have the same phonological form in order for the two verb phrases to be considered identical. Let us assume that the dialect difference is based on the difference in identity requirements, as I have suggested, and SOME THOUGHTS ON TRANSDERIVATIONAL CONSTRAINTS George Lakoff University of Michigan Recent work by Grinder, Postal, and Perlmutter has shown that the theory of derivational constraints is too weak to ac- count for the facts of such natural languages as English, Span- ish, French, and Slovenian. The basic point of the Grinder- Postal examples is that in certain structural configurations, ambiguities of a certain sort are not permitted, though they are permitted in other structural configurations. Consider (l) and (2). (l) John and Mary entered the room, and he took off his Coat- that dialect II requires the pronouns to have identical phono- (2) *John and Bill entered the room, and he took off his logical forms. Then there exists one (and I think only one) ooat- general hypothesis under which the facts of dialect II can be handled without transderivational constraints, namely, that there is a rule of verb phrase deletion and that the phonologi- cal forms of the pronouns appear in the derivation before the application of that rule. Then, one could define constituent identity in dialect I so that the phonological forms of pro- nouns were irrelevant, but define it in dialect II so that the phonological forms of pronouns had to be identical. No trans- derivational constraint would be required. However, there are two other possible hypotheses which would require transderivational constraints.. Hypothesis A: There is a rule of verb phrase deletion, but the phonological forms of pronouns are not introduced until after it has applied. Hypothesis B: There is no rule of verb phrase deletion. Instead, the meanings of missing verb phrases (and other anaphoric expressions) are to be In (l) John is the antecedent of he. In (2), where there are two possible antecedents, the sentence is ill—formed on both readings. This is not generally true, as (3) shows. (3) John told Bill that he had won the sweepstakes. In (3) the sentence is acceptable on both readings, rather than unacceptable on both readings. Grinder and Postal observed that (2) cannot be blocked within the theory of derivational constraints, since ambiguity is not a property of a single derivation. Since blocking occurs just in the case where ambiguity would arise, the only way a rule of grammar could block either of the two derivations of (2) was if it knew about the other possible derivation. Thus, there must be rules that apply not to individual derivations but to classes of derivations. In short, transderivational constraints are required, since there are cases where the well—formedness of one derivation depends on certain properties of other, related derivations. Perlmutter's cases are even more interesting, determined by a surface structure interpre- since there the application of an otherwise obligatory trans— tation rule, formation Pole is blocked if it would Yield 3“ ambigUitY- Dist I : Under Hypothesis A, deletion would take place before the phono— cussion of such cases is, however, beyond the scope of this ‘ ‘ logical forms of pronouns were introduced, so that the deletion pap€P- rule could not be sensitive to a difference in phonological I would like to discuss a number of cases which I suspect formS_ The only way that dialect [I could be described would will require transderivational constraints, though I have no be to say that IF the verb phrase had not been deleted, them proof of this at present. I am bringing these cases up nerely there WOULD HAVE BEEN a difference in the phonological forms to stimulate further discussion of the matter. Consider the Qf the pronouns in (5) and (7)_ In other words, there is a fOllOWingt corresponding derivation in which the deletion does not take (4) John scratched his arm and so did Bill. (scratch place’ and in which there 15 a difference in the phonological his arm) forms of the pronouns. That is, a transderivational constraint

Notes on what it would Take to Understand how One Adverb Works

Abstract: Sores on what it would take to understand hw one adverb uorks* George Lakoff Department of Linguistics University of'California, Berkeley September 1972 To appear in The Monist *This work was partially supported by grant GS-35119 from the National Science Foundation to to University of California at Berkeley. V'\) 1. A natural language is a unified and integrated system, and the serious study of one part of the system inevitably involves one in the For this reason, the study of small, isolated fragments of a language--however study of many other parts, if not the system as a whole. necessary, valuable and difficult this may be -- will often make us The fact is that you can't really study one phenomenon adequately without studying a think that we understand more than we really do. great many other related phenomena, and the way they fit together in This is the sort of thing Experience in descriptive terms of the linguistic system as a whole. a linguist learns very early in his career. linguistics, even at an elementary level will force a linguist to come to grips with a wide range of complex data in some language, perhaps even English, and the truism soon emerges. But, due to the vagaries of our educational institutions, few philosophers or logicians receive training in linguistic description. Consequently much of the discussion of natural language in the Philosophical and logical literature is based on a very small sampling of data which is skewed in nontrivial ways. True, one has to start somewhere, and a great deal has been learned by ordinary language philosophers who have looked at only a handful of relatively simple examples and by logicians who have studied what by natural language standards are only miniscule fragments (e.g., first- order predicate calculus, the various modal logics, etc.). But now that philosophers and logicians are turning to more detailed studies of natural language phenomena, it is perhaps the right time to suggest that philosophical and logical training be expanded to include the study of natural languages as entire systems. I don't mean to suggest, for example, that logicians should stop their systematic study of small fragents, but rather that a knowledge of the kinds of phenomena out- side of those fragments can enrich the study of fragments and give one a more rezatistic picture of what one does and does not know about natural language. The study of adverbs is a good case in point. Reichenbsch,1 in his analysis of conversational language, made a brave attempt to study a number of natural language phenomena that had previously been ignored

Lexicography and generative grammar i: hedges and meaning criteria *

Abstract: Lexicography is an art-and one where practical considerations outweigh scientific considerations. Linguistics has discovered a great many new facts about words in recent years, but this does not necessarily mean that such facts can or even should be included in any dictionary no matter what its degree of abridgement. In fact, including technical linguistic discoveries in a dictionary may be the last thing a good lexicographer would want to do. On the other hand, technical linguistic discoveries, if used as a very subservient handmaiden to lexicographic craft, could conceivably lead to better dictionaries. What I would like to do is report on some linguistic discoveries, knowing that they cannot and should not be directly applied to the making of dictionaries, but nonetheless hoping that some lexicographers will find a way to make practical use of them. As any lexicographer knows, natural language. concepts are fuzzy; the boundaries are not clear-cut. Zadeh has suggested that fuzziness should be handleable formally in terms of what he calls fuzzy set theory.’ In a fuzzy set, an individual is not simply a member or a non-member, but may be a member to some degree, for example, any real number between 0 and 1. Take the fuzzy concept of tallness. Someone who is 6’5” is clearly tall, and someone who is 5’0” is clearly not tall. But what about someone who is 5’7”, 5’9”, 5‘1 l”? Zadeh suggests that such individuals exhibit intermediate degrees of tallness. Obviously, as one’s height becomes greater, so one’s degree of tallness becomes greater-until a certain threshold is reached. Zadeh has suggested that concepts such as tall are to be represented graphically, pretty much as in FIGURE 1. Interestingly enough, there are words that change fuzziness. Consider, for example, the expression “sort of.” The concept “sort of tall,” if represented graphically, would look approximately as in FIGURE 2. Clearly, if someone is 6’5“, he is not sort of tall-he is just plain tall. But if you are, say, 5’9”, then you are sort of tall. Essentially, the expression “sort of” raises the intermediate values and lowers the high values for “tall.” One might think that tallness is a rather special case. After all, tallness can be described as a continuous function of height plus various contextual factors. But what about more “discrete” concepts? Take the concept of a bird. Heider,* in a psychological experiment, has shown that people do not perceive birdiness as a yes-or-no affair. Rather there seem to be gradations of birdiness. A hierarchy like that given in ( 1 ) emerged from her studies.