Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
LACUS FORUM XXV © 2009 The Linguistic Association of Canada and the United States (lacus). The content of this article is from lacus Forum 25 (published 1999). This article and others from this volume may be found on the Internet at http://www.lacus.org/volumes/25. YOUR RIGHTS This electronic copy is provided free of charge with no implied warranty. It is made available to you under the terms of the Creative Commons Attribution-NonCommercial license version 3.0 (http://creativecommons.org/licenses/by-nc/3.0/) Under this license you are free: • • to Share — to copy, distribute and transmit the work to Remix — to adapt the work Under the following conditions: • • Attribution — You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). Noncommercial — You may not use this work for commercial purposes. With the understanding that: • • Waiver — Any of the above conditions can be waived if you get permission from the copyright holder. Other Rights — In no way are any of the following rights affected by the license: • Your fair dealing or fair use rights; • The author's moral rights; • Rights other persons may have either in the work itself or in how the work is used, such as publicity or privacy rights. Notice: For any reuse or distribution, you must make clear to others the license terms of this work. The best way to do this is with a link to the web page cited above. For inquiries concerning commercial use of this work, please visit http://www.lacus.org/volumes/republication Cover: The front cover of this document is licensed under the Creative Commons Attribution-No Derivative Works 3.0 license (http://creativecommons.org/licenses/bynd/3.0/) and may not be altered in any fashion. The lacus “lakes” logo and Claremont Graduate University logo on the cover are trademarks of lacus and Claremont Graduate University respectively. The Claremont Graduate University logo is used here with permission from the trademark holder. No license for use of these trademarks outside of redistribution of this exact file is granted. These trademarks may not be included in any adaptation of this work. ALMOST, AND COUNTERFACTUAL IMPLICATURES Debra Ziegeler Monash University and National Tsing Hua University An area of infrequent analysis in past pragmatic research is the role of approximating or proximative adverbs such as almost and nearly in producing meanings of counterfactuality, and the place of such adverbs on an implicature scale of the nature of Horn (1972). From first appearances it would seem that there is no reason why proximative adverbs could not appear on such a scale, alongside quantifiers such as some, most, few and more precise indicators of quantity such as cardinal numerals. The function of such items appears to be comparable to that of the quantifiers in restricting a universal proposition, or in grading degree in some adjectives and other quantifiers. However, there emerges a clear difference: quantifiers like some conversationally imply the negative of a stronger proposition, by the observance of the Gricean Maxim of Quantity 1, while almost (and those akin to it) has been believed to contain negative entailments, for the reason that the relevant inferences are non-defeasible, making them unlikely candidates for implicature scales (see, e.g., Atlas 1984; Hitzeman 1992). The paradox which arises is that, as noted by Sadock (1981:257), the negation of the stronger proposition that almost restricts is not part of its semantic content, and yet this counterfactual proposition, in the majority of cases, cannot be cancelled. The present paper aims to resolve the paradox, analysing the meaning of almost as a strong counterfactual implicature, and offering an account by which this and many other counterfactual meanings may be explained.¹ The argument for an implicature analysis for almost will be supported by revealing a predictive function for proximatives, similar to that of certain modal verbs, and comparison will be made with proximatives in other languages, which grammaticalize from the same crosslinguistic sources as predictive meanings. A factor to be considered is whether co-occurrence conditions will affect the interpretation of almost, and a small number of native-speaker intuitions will assess the grammatical distribution of almost in order to investigate whether contradictions, indicating entailments, can be found in all environments. It will be shown that the negative inferences assumed to be part of the meaning of approximators are actually strong counterfactual implicatures, their strength determined by means of a given Principle. The Principle is hypothesized to apply to a wide range of counterfactual meanings. 224 Debra Ziegeler 1. Arguments for an implicature analysis 1.1 Prediction. According to Sadock, it was possible to consider the denial of the truth value of the predicate of almost as a generalized conversational implicature based partly on its meaning, rather than an entailment (1981:261). Sadock had admitted that the cancellation of such an implicature out of context would be difficult, and not parallel to the kinds of cancellation applicable to some/all examples on the implicature scale of Horn (1972),² but that this was mainly because of the strength of the implicature. In order to provide evidence for the presence of an implicature and not an entailment, Sadock argued for a number of cases. One was the use of the conjunctor but in examples such as: (1) He almost, but didn’t quite, win the race. This use is compared with the use of truth-functional equivalents such as and: (1)' ?He almost, and didn’t quite, win the race. (1)' is considered inappropriate since, unlike but in (1), the use of and adds nothing to the meaning of almost. Sadock (1981:264) therefore assumed that almost implies the denial of his winning the race, and but, used in (1), is said to reinforce the implicature, adding more information. However, the conventionally implied meanings in but suggest that the conjunctor typically introduces a contrastive statement, so in (1), the possibility of his winning must also be contained in the meaning of almost for such statements to be appropriate. Sadock does not discuss this possibility, but but-clauses offer firm evidence for the implication of predictive senses in the proximative adverb. All past time reference uses of almost can be appended with a conjoining clause introduced by but, and, as well as adding more information, the conjunctor can be used in most cases to suggest a contrary turn of events; e.g.: (2) She had almost sung Strauss’s ‘Four Last Songs’, but she broke into a sneezing fit before she had finished. Although Sadock refers to the use of but to reinforce the implicature of negation, it is questionable whether an implicature that is strong enough to be considered by some accounts to be an entailment actually still needs reinforcing. The use of but in (1) and (2) to conventionally implicate contrast, adversity, or unexpectedness is unlikely if the inferences surrounding the use of almost are that the predicated event did not occur anyway—just where does the contrast lie? For this reason, it is likely that the relevant inferences are related to the Gricean Maxim of Quantity 2. The Gricean Quantity 2 Maxim (termed q2 by Levinson 1995, or R-based by Horn 1984), in contradistinction to the Quantity 1 Maxim, is infor- almost, and counterfactual implicatures 225 mationally upper-bounded, and may license the implicature that the speaker was inferring more information than was expressed. q2 implicatures have been shown to influence the grammaticalization of predictive meanings from modal verbs historically (Traugott 1989), by the implicatures of future action resulting from present intentions, abilities, or permission. The use of a but-clause in examples such as (1) and (2) reveals the presence of similar inferences associated with almost. This offers strong evidence against the entailment hypothesis, which, if it is sustained, will require an explanation for such cases. The predictive nature of almost in past contexts may therefore contribute to the irrealis meanings of counterfactuality considered to be negative entailments in some accounts. In the following section it will be seen that the proximative functions of almost are grammaticalized diachronically in other languages from verbs of volition. It is such parallels with volition verbs that enable a modal function to be suggested as attributable to the use of almost. 1.2 Evidence from the grammaticalization of proximatives. Although verbs of volition often form the sources for the grammaticalization of predictive future meanings, similar lexical sources have been found to grammaticalize into proximative meanings. Kouteva and Heine (1995) and Heine (1994) were able to isolate in their data from African languages and Bulgarian a volition verb with a corresponding grammatical function of marking proximity, that is, expressing meanings of ‘nearly’ or ‘almost’. In a study of the grammaticalization of negative marking in Chamus, a Maa dialect of the Eastern Nilotic family, Heine (1992) illustrates a gradual chain of concepts: the lexeme -yyéú moves from being a main verb of volition with a noun phrase object, to an auxiliary-like verb with a main verb complement, and then to a grammaticalized, uninflected aspect marker with an inanimate agent as subject, conveying the meaning ‘almost’, and finally grammaticalizing to imply negation. Part of this chain is illustrated in (3): (3) a. k-á-yyéú n-aâ k-1sg F-food.abs ‘want food’ b. k-é-yyéú l-cáni n-é-uróri k-3sg- M-tree.nom nar-3sg-fall ‘the tree almost fell’ (Heine 1992:338–40)³ In (3)b the grammaticalized senses of proximity are well established, as is demonstrated by the presence of an inanimate subject not capable of expressing volition. The use of yyéú is now as an aspect marker, described by Heine (1992:339) as the almost-aspect, but later renamed the proximative aspect in Heine (1994). Other languages in which volitional verbs have provided the lexical sources for the proximative include Swahili, Mvua, and Ewe (ibid); and Heine (1992:342) notes that in Southern Soto, a Bantu language, the verb batla meaning ‘seek’, ‘desire’, or ‘want’ has developed into an auxiliary with a proximative aspect. Matisoff 226 Debra Ziegeler (1991:394) notes that for Chrau (a Mon-Khmer language spoken in Vietnam), the verb cornh (‘want to’) can also function as a non-negatable, pre-verbal particle meaning ‘almost’. It is of interest to note that not all of these attested examples have progressed to become negative markers, but that when the temporal reference is past, negative inferences also begin to appear as a meaning shift in a limited number of environments. This suggests the influence of the Maxim of Quantity 1, which implies the negation of the predicate proposition. The arguments for the negative entailments rather than counterfactual implicatures resulting from the use of almost are further weakened by this cross-linguistic data, illustrating the gradual development of proximative inferences from volition verb sources via the strengthening of q1 implicatures. Given such evidence, it is suggested therefore, that the senses of counterfactuality associated with almost in English may be similarly the result of the gradual strengthening of Quantity implicatures, and that both q1 and q2 implicatures may interact to determine proximative meanings. 2. Co-occurrence conditions. One of the arguments Hitzeman (1992) uses to justify an entailment analysis of almost is the contradiction suggested by almost P and P occurring co-temporally. However, Atlas (1984:348) argues that a sentence such as: (4) Moore almost understood ‘material object’ and he understood it is acceptable. The acceptability of (4) can hold if the proximity to the event and the event itself are considered co-temporal. Leech (1987:25) describes verbs such as understand as ‘verbs of inert cognition’, or ‘unrestricted present’, referring to mental states. As such, they should be easier to construe as situations which are homogeneously distributed over an indefinite time-span, and such temporallyunbounded environments cannot individuate the implied separate events required to create the contradiction necessary to ascertain an entailment account. Perhaps, then, in some environments almost produces a greater likelihood of contradiction than in others, suggesting that the negative inferences may in fact be graded in a continuum-like fashion according to varying strengths. It is possible to see that (4) does not appear so sharply contradictory as, for example, Sadock’s (1981) example would appear: (5) Bill almost swam the English Channel, and he swam it when referring to the same event in both clauses. The reason for this is that almost usually refers to a point close to the boundary of the predicate it quantifies (as suggested by Heine (1994:36) referring to the proximative in African languages), and this point is more vaguely construed in imperfective situations such as (4). almost, and counterfactual implicatures imperfectives Figure 1. 227 perfective events Figure 2 Langacker (1987:79) described the use of imperfective processes to refer to an extension of a stable situation through time, one with no inherent boundaries. Thus, the state of full understanding may not be conceptually distinct from any approach stages or states of approximate understanding which may overlap with it, and may not entirely contradict such states. (4) may be schematized then as in Figure 1, the centre shading marking the area of approximation which overlaps with the darkest area marking the (predicted) fully represented state. Grammatical imperfectivity, using the progressive aspect, can be similarly schematized, using an activity verb (see Vendler 1967), which is also inherently unbounded, as in Bill was almost swimming (see Figure 1). The center shading does suggest for expressions such as almost swimming that some action is on-going at reference time, which may overlap with and indicate a less representative instance of the state of swimming. Punctual or telic achievement verbs, however, do not contain the possibility of grading indicating a part-whole relationship, since they are bounded in time, and the time span over which the event occurs is too brief to be considered as co-occurring with an approach phase. For this reason, the grading provided by almost is pragmatically determined by a time-scale, not semantically determined by the internal composition of the entity, and the approach stage is implied as occurring independently and outside of the bounds of the event. The representation of almost with a punctual achievement event; e.g., Mike almost fell, can be illustrated as follows, the cross marking the point of proximity, and the dotted figure marking a predicted event: Because of the conceptualisation of an external approach point in Figure 2, accounts such as Hitzeman (1992) cannot consider the use of almost with such verbs unless the adverb refers to a time point outside the time at which the action did occur (if at all). The punctuality of many actions makes it impossible to perceive, for example, an event schematized in Figure 3, where the act of falling includes or is fused with the approach point. In reporting the conceptualization of a punctual event, the approach point is often not viewed as discrete from the result, and, furthermore, it would be conversationally uncooperative to report only the approach point without reporting the complete event, if known. It is for such reasons that proximative adverbs can be given an implicature account, rather than an entailment account. Langacker (1987) also ranked perfective and imperfective verb types as direct notional analogues of count and mass nouns, count nouns being comparable to 228 Debra Ziegeler |–—fell—–| Figure 3. perfective processes, and mass nouns to imperfective processes. If the use of proximative adverbs invokes a point close to a conceived boundary of a predicate, then a spatial boundary demarcated in the notion of a count noun should be amenable to approximation by almost in the same way as the temporal boundary demarcated by a perfective verb type. It is possible to use almost to refer to the count nouns Langacker terms ‘interconnected entities’, e.g., team, constellation and alphabet, since these refer to bounded entities which are compositionally distributed over a number of members. The grading of size then provides the internal scalarity needed for the use of almost; and the boundary relevant to approximation is the terminal one; e.g., in We have almost a team here; in which the quantifying expression may be diagrammatically represented as Figure 4. Noun phrases with adjectival quantification can be similiarly represented, some gradable adjectives (e.g., almost bald—see (3)), as well as one reading of accomplishment predicates (the other is plotted as in Figure 2), in which the boundary to which the adverb approximates is the terminal boundary of the object of accomplishment (e.g., the English Channel in Bill almost swam the English Channel). (Again, the dotted boundary indicates a predicted, not an actual, demarcation.) Almost may also occur with some uses of mass nouns, as in This soup is almost water, where it is implied that too much water has been added in the making of it, producing a substance which is partly water, partly soup. In such contexts, the mass noun water has been construed as having an approach point, in which the entity described by the mass noun is seen to partially overlap into the entity described by another mass term, soup, and is not discontinuous with it. This situation may be schematised, as in Figure 1, in which the approach marker almost may be used to indicate the direction in which the composition of the soup is heading, but the interconnected entities Figure 4. almost, and counterfactual implicatures 229 approach phase and the resulting state are not totally discrete, since some of the resulting state may be construed as already present in the approach phase. The same situation can also be seen with some cases of non-gradable adjectives; e.g., almost crazy, where almost crazy is considered as partially overlapping with the absolute state, crazy, at the same time being a less representative instance of it. Such adjectives can also be diagrammed according to Figure 1. Some entities cannot be schematised at all, such as a count noun referring to a non-gradable, physical object (e.g., beagle, in Hitzeman’s 1992 example, Martha has almost one beagle). Such nouns cannot be given any kind of interpretation with almost unless they refer to created entities (e.g., almost a cake, which could be represented as in Figure 4). The reason why certain count nouns and bounded verbs are not direct analogues where approximation is concerned is that time provides the quantifiable scalarity necessary for the expression of an approach point with bounded verbs, and if the count noun has no inherent grading, there is no equivalent scalarity progressing towards a spatial goal without invoking temporal measurement as well. If temporal progression cannot be conceptualised, the expression is unacceptable. The potential for the negative inferences to be cancelled then, appears to vary according to the environment in which almost appears, and with stative verbs, mass nouns, and other time-stable predicates, an approximate representation of their state provides a valid instance of that state without total contradiction of the full representation. The strength of the negative inferences associated with the use of almost are therefore hypothesized to be a factor of their co-occurrence conditions, varying according to the presence of conceived boundaries in the predicates over which they have scope. In order to test this hypothesis, a brief survey was conducted, investigating the extent of such variation perceived by individual speakers. 3. Testing the co-occurrence conditions. The survey consisted of a small number of sentences which were constructed on the basis of the points discussed above, and in which the variables were bounded and unbounded predicates. Unbounded predicates included a quantified noun phrase (sentence 1), a gradable adjective (sentence 2), and a lexically and grammatically imperfective verb (sentence 3). Bounded predicates included a physical object count noun (sentence 4), a past achievement verb (sentence 5), and a past accomplishment verb (sentence 6), both of the last two being grammatically and lexically perfective verb types. The respondents were presented with a group of sentences of the form almost P, in fact P, and asked to list those they thought were contradictory.⁴ The number of example sentences was kept to a minimum to avoid the problems of weakening of judgement due to repetitiveness of the task. It was hypothesised that the first three sentences (1–3), in which the item quantified by almost was felt to be conceptually unbounded and internally-scaled would invoke weaker implicatures (which would be more easily defeasible) than the last three (4–6). If this were the case, the sense 230 Debra Ziegeler Sentence No./20 % 1. 2. 3. 4. 5. 6. 11 10 12 19 15 18 55 50 60 95 75 90 Table 1. Distribution of the assessment of contradictory sentences. (The number of respondents listing all sentences as contradictory was 9 out of 20, or 45%.) of contradiction in sentences 1–3 would be predicted to be less obvious to the respondents than for sentences 4–6. The survey and results are summarized below. (6) The Survey Sentences 1. Almost all, in fact, all of the wheat crop was saved. 2. Sam is almost bald, in fact he is bald. 3. When I saw Sue she was almost running, in fact she was running. 4. The animal in the cellar was almost a mouse, indeed it was a mouse. 5. John almost fell off the bridge, in fact, he did fall. 6. The dog almost crossed the road, in fact, it crossed the road. The results are listed in Table 1. The limited data clearly suggests that opinions vary as to whether or not a sentence with co-occurring almost P and P are contradictory, and that there is a general tendency for externally, temporally-scaled, bounded perfective verbs (sentences 5–6) to have higher evaluations of contradiction (75–90%) than internally-scaled count nouns, imperfective processes, or stative qualities (50–60% for sentences 1–3). (Sentence 4 showed neither internal scaling nor external, temporal scaling, and thus scored the highest.) The differences in the evaluations of contradiction indicate that an entailment reading of almost is not always present, suggesting instead the presence of counterfactual implicatures varying in strength according to environment and speaker intuitions. Given that temporally-bounded situations are more likely to be listed as contradictory according to the survey, it may be hypothesised that implicature strength increases in such environments; i.e. those of more specific conceptual distinction. The implications of such results are discussed in section 4. 4. Determining implicature strength. Lewis (1973:69) claimed that a counterfactual could not be asserted without a factual premise to back it. Thus, in past contexts, the counterfactuality of an expression is dependent on the speaker’s almost, and counterfactual implicatures 231 awareness of certain known facts to the contrary, otherwise, the meaning is considered merely hypothetical. A counterfactual implicature, therefore, is an inverse construal derived from accessible factual evidence. The expression of a bounded, past perfective event is taken as having terminated and is therefore unalterable in the subjective viewpoint of the speaker. As a result, it becomes more highly factualized, and when co-occurring with proximatives, licenses stronger implicatures than a time-stable, stative verb phrase, to the point at which the implicature is interpreted as either conventionalized or an entailment. When almost, then, is used to quantify time-stable entities such as quantifiable nouns and stative verbs, it is more easily defeasible. When used to quantify realized past perfective events, it is not so defeasible since the realisation of a perfective event profiles the termination of the event, and almost P, which indicates the approach stage of an event, cannot be conceived as co-occurring simultaneously with the termination of the event P in the short time period in which such punctual events take place. The termination of the event, in turn, indicates a greater potential for factual recall by the speaker, since it has already passed at the moment of speaking. The enforcement of the implicature, then, depends only on the potential for the speaker to utter a statement about which true facts are known. As this potential increases, so the likelihood of defeasibility decreases. The likelihood of the occurrence of an implicature in past contexts is therefore minimal, since we normally have facts about the past, and we do not need to derive pragmatic inferences for further information. An implicature, then, is simply not anticipated in relation to past contexts, or in similar information-rich situations. This situation can be summarized in the form of a single principle, which will be labelled in the present study the Counterfactual Implicature (ci-) Principle: (7) The ci-Principle: ‘The strength of an implicature is inversely proportional to the probability of its occurrence.’ In linguistic terms, this means that by increasing the information density of an utterance, an implicature will not be likely to form; therefore, in a situation in which the specificity and/or factuality of an utterance is expected to be greater, such as in past perfective contexts, an implicature will be improbable. However, the strongest counterfactual implicatures, such as are derived from predicates with almost, apply to past, perfective contexts, so when in an improbable context a counterfactual implicature does occur, it will be much more powerful than in a more probable context; in this way the pragmatic function is best preserved. Environments with negation or with a first person subject were found in Ziegeler (1994 and 1996) also to produce the strongest implicatures of counterfactuality relating to conditionals and hypothetical complements, negation because of its marked nature, implying a relevant, corresponding positive expression (see 232 Debra Ziegeler Horn 1989), and first person subjects as they are identical with the speaker and imply the highest level of evidential support for an utterance. Sadock (1981:265) observed that the cancellation of implicatures such as some X by in fact, all X, could not be so easily transferred to instances in which the expressed quantity was more specific: ?A fraction of, and possibly all, poodles have spots. In such cases, the difficulty is due to the increased information density which lends the utterance a higher factual evaluation by the hearer. Givón (1994:310) questions the common appearance in subjunctive clauses of past and perfect forms, essentially realis morphology. The reason is that past and perfect forms express factually more specific concepts which are more amenable to abstraction by counterfactuality. The licensing conditions for a counterfactual implicature are therefore not the actual provision of factual information, but the hearer’s assessment of the potential for one given situation to provide more explicit and more detailed information than another. 5. Conclusions. The present paper analyses the meaning of almost in terms of a strong conversational implicature (as suggested in Sadock 1981) of counterfactuality, and argues against the claims by Hitzeman (1992) and others that the proximative adverb is an entailment. Evidence for the implicature analysis comes from: (i) the fact that a negative entailment from almost would suggest a prediction, which is an expression of only possible truth, not actual truth, and which, furthermore, is derived by means of q2 implicatures; (ii) the evidence of proximatives in other languages in which the meaning is grammaticalized via pragmatic implicatures from the same crosslinguistic sources as derive predictive meanings; and (iii) the variation with which contradictions are assessed by native speakers when almost is used in different grammatical environments. The negative inferences associated with almost can be seen to be non-defeasible in many instances, but can be more easily cancelled when quantifying unbounded, internally-gradable entities and states. The phenomenon of increasing and decreasing implicature strength is a factor of the potentiality provided by the context for generating an implicature, and this is summarised in the ci-Principle. The effects of the Principle are clearly seen in the behaviour of almost in different contexts, as shown in the data from a small survey. While the Principle has been shown to be most effectively applied to counterfactual situations in the use of almost, it is not to suggest that it may be limited in the extent of its application to such contexts. The obvious next task is to test the Principle in other counterfactual domains such as conditionals and complement clauses. Further investigations may also look to finding more precise and principled means of predicting pragmatic inferences generally. Such tasks present a challenge for future endeavours in the field. almost, and counterfactual implicatures 233 ¹ The definition of counterfactuality is to be understood here as determined only by speaker knowledge: a counterfactual utterance can be made only in cases where the speaker has full knowledge of facts contrary to the hypothesised proposition; in all other cases the meaning is merely hypothetical. ² Such scales are termed ‘Horn-scales’ by Atlas and Levinson (1981). ³ Heine notes that in these examples the function of the prefix k- is unclear. ⁴ The form almost P, in fact P could be understood by the reader as a self-correction of an utterance. However, this form has been used because it also serves typically to cancel implicatures in scalar expressions, e.g., John ate some, in fact all of the cake. REFERENCES Atlas, Jay D. 1984. Comparative adjectives and adverbials of degree: An introduction to radically radical pragmatics. Linguistics and philosophy 7:347–77. ——— & Stephen C. Levinson. 1981. It-clefts, informativeness, analogical form: Radical pragmatics. In Cole 1981,1–61. Cole, Peter (ed). 1981. Radical pragmatics. New York: Academic Press. Givón, Talmy. 1994. Irrealis and the subjunctive. Studies in language 18:265–337. Heine, Bernd. 1992. Grammaticalization chains. Studies in language 16:335–68. ———. 1994. On the genesis of aspect in African languages: The proximative. Proceedings of the Berkeley Linguistics Society 20:35–46. Berkeley, ca: Berkeley Linguistics Society. Hitzeman, Janet. 1992. The selectional properties and entailments of ‘almost’. cls: Proceedings of the annual meeting of the Chicago Linguistics Society 28 vol 1, ed. by Costas P. Canakis, 225–38. Chicago: Chicago Linguistics Society. Horn, Laurence R. 1972. On the semantic properties of logical operators in English. Los Angeles: University of California dissertation. ———. 1984. Towards a new taxonomy for pragmatic inference: Q-based and R-based implicature. Meaning, form and use in context: Linguistic applications/ gurt ’84, ed. by Deborah Schiffrin, 11–42. Washington, dc: Georgetown University Press. ———. 1989. A natural history of negation. Chicago: University of Chicago Press. Kouteva, Tania & Bernd Heine. 1995. The proximative. International Cognitive Linguistics Conference, 24-28 July, Albuquerque, nm. Langacker, Ronald W. 1987. Nouns and verbs. Language 63:53–94. Leech, Geoffrey. 1987. Meaning and the English verb. Harlow: Longman. Levinson, Stephen C. 1995. Three levels of meaning. Grammar and meaning: Essays in honor of Sir John Lyons, ed. by Frank R. Palmer, 90-115. Cambridge: Cambridge University Press. Lewis, David. 1973. Counterfactuals. Cambridge, ma: Harvard University Press. 234 Debra Ziegeler Matisoff, James. 1991. Real and universal dimensions of grammaticalization in Lahu. Approaches to grammaticalization II, ed. by Elizabeth C. Traugott & Bernd Heine, 383–453. Amsterdam/ Philadelphia: John Benjamins. Sadock, Jerrold M. 1981. Almost. In Cole 1981,257–71. Traugott, Elizabeth C. 1989. On the rise of epistemic meanings in English. Language 65:31–55. Vendler, Zeno. 1967. Linguistics and philosophy. New York: Cornell University Press. Ziegeler, Debra P. 1994. Conditionals and counterfactuality in Singaporean English. Journal of intercultural studies 15:29–49. ———. 1996. A synchronic perspective on the grammaticalization of will in hypothetical predicates. Studies in language 20:411–12.