This book "supplies a vocabulary of English words and idiomatic phrases 'arranged … according to the ideas which they express'. The thesaurus, continually expanded and updated, has always remained in print, but this reissued first edition shows the impressive breadth of Roget's own knowledge and interests."
Peter Ludlow (1999), Semantics, Tense, and Time: An Essay in the Metaphysics of Natural Language. [A Bradford Book] Cambridge, Massachusetts: The MIT Press. xxi + 252 pp. $45.00/�29.95
Reviewed by: Gorel Sandstrom and Rognvaldur Ingthorsson, Department of Philosophy and Linguistics, Umea university
The book under review argues for a version of the so-called "A-series" conception of the nature of time, and against the opposing "B-series" conception. The two series are defined as follows (cf. the Introduction, p. 2) A-series: - The B-series is reducible to the A-series. - Temporal becoming is intrinsic to all events. - There are important ontological differences between past and future. - Change is analyzable solely in terms of A-series relations (past, present, future) B-series: - The A-series is reducible to the B-series - Temporal becoming is psychological. - The B-series is objective. All events are equally real. - Change is analyzable solely in terms of B-series relations (earlier-than, later-than). According to the preface, however, the particular argument outlined is not the main concern of the author, which is rather to "illustrate an approach to metaphysics in which semantical theory and the philosophy of language are central" (p. xvii). Thus, a large part of the text is concerned with examining the extent and kind of "metaphysical commitment" of different semantic theories of tense.
SYNOPSIS The book is divided into ten chapters of varying length, preceded by an Introduction and followed by two "philosophical" and five "technical" Appendices, notes and references. The philosophical appendices contain discussions of I-language as the language of thought (touched on in Chapter 1) and of language-world isomorphism (a possibility hinted at in Chapter 4); the technical appendices present T-theories for fragments of English using different semantic theories.
After the Introduction, which presents the argument in outline and gives an overview of the remainder of the book, Chapters 1-3 lay the groundwork for what is to follow by presenting some basic assumptions and theoretical choices. Chapter 1, "The nature of language", argues for language as I-language in the sense of Chomsky (1986), and against the idea that I-language is "for" communication ("for" is taken here in the rather strict sense of "evolved for the purpose of".) Ludlow also argues against a "third medium" between I-language and the world, on the grounds that such a medium would be redundant, an argument which is expounded on in Appendix P1. A central assumption is that all thoughts we have about the world ("genuine representational thoughts") are propositional ? which here means that they are stored in the form of interpreted logical forms (ILFs), i.e. phrase markers whose nodes are paired with semantic values. In Chapter 2, "The form of the semantic theory", Ludlow argues for a particular form of semantic theory to characterize the first-order semantic knowledge that an agent has, namely, an absolute truth-conditional semantics. To exemplify the "T-theory" approach, he gives a simple fragment from Larson & Segal (1995); more complex fragments are in Appendices T1 and T2. Besides a context-free syntax, the theory contains two kinds of axioms, one assigning semantic values to lexical items, the other stating how the semantic value of a mother node can be calculated from those of the daughter nodes. The question of whether T-theory axioms can display senses receives some discussion, in the course of which Ludlow draws a distinction between first-order semantic knowledge (i.e., knowledge that 'snow' refers to snow) and second-order semantic knowledge, i.e. knowledge of the representation of the knowledge. A correct T-theory must correctly characterize also the latter (which amounts to "displaying senses"). Chapter 3, "Attitudes and Indexicals", starts addressing issues pertaining to the semantics of tense. In order to ground the idea that tense morphemes are indexical predicates wich take interpreted logical forms (ILFs), as arguments, Ludlow discusses first ILFs as used in another context, that of propositional attitude verbs, and then indexicals in general. Propositional-attitude verbs are analyzed as expressing relations between agents and ILFs; Ludlow also proposes to handle other intentional contexts, like intensional transitive constructions ('John wants a donkey'), in the same way: these sentences contain a hidden clause in their LF representation (something like 'John wants [PRO (to have) a donkey]') and are thus amenable to ILF analysis. As for indexicals, Ludlow argues that the "role" or "character" of the indexical need not be kept out of the semantics of the indexical expression, in effect, that indexical expressions like "I" can be disquotationally entered into the right-hand side of a T-theory axiom. In chapter 4, "Drawing Metaphysical Consequences from a T-theory", Ludlow sets out to show where metaphysical commitment arises in an "absolute" theory such as he has adopted. The basically Quinian position he takes is that we are committed to the existence of whatever semantic values we metalinguistically quantify over. An axiom like (i) above, which is shorthand for (i'), thus commits us to the existence of snow. (i) Val(x, _snow_) iff x = snow (i') For all x, Val(x, _snow_) iff x = snow >From this point of departure, he contrasts the different metaphysical commitments which arises depending on what semantic theory we adopt. The cases he explores concern properties, names, and events. Ludlow contends that no T-theory can avoid metaphysical commitment altogether. It is in Chapters 5 through 8 that the central argument of the book is developed, concerning the relationship between tense semantics and metaphysics of time. In Chapter 5 he expounds "what is essentially the received view: the Reichenbachian theory of tense" (p. 77), which he calls the B-theory semantics. The Reichenbachian semantics is a B-theory semantics in that it involves tenseless truth conditions, with temporal relations between events stated in terms of the earlier-than/later-than relations, and where future and past events are equally "real" as present ones. Axioms for tense have the basic form of (ii). (ii) Val(e, PAST, t) iff e is temporally before t He gives axioms for seven tense morphemes, using the triple index S,R,E; and for temporal adverbs like 'yesterday, 'today', and 'tomorrow' as well as temporal connectives like 'when', 'before', and 'after', all exploiting the temporal relations "earlier-than" and "overlap". In Chapter 6, "Problems with the B-Theory Semantics", Ludlow argues (like many others have) that the B-theory semantics is inadequate in that it cannot solve the problem of indexicality. The most promising attempt thus far, the token-reflexive account, is itself dependent on using an indexical to fix the reference to the token referred to. That is, according to the token-reflexive analhysis, the sentence 'my fifth anniversary is _today_' is true iff my anniversary is the day of S -- the day of _this_ very utterance. Instead of picking out the day of the utterance by the indexical expression 'today, the utterance is picked out by the indexical expression 'this'. Even if this problem could be solved, Ludlow continues, the token-reflexive analysis will have as a consequence that sentences like 'there are no linguistic tokens' are not just simply false when tokened or uttered, but necessarily false. Ludlow argues that it must of course be true that there are no linguistic tokens when in fact there are no linguistic tokens, and that an adequate semantic theory should be able to give the conditions under which this would be true. Chapter 7 develops the A-theory semantics, a basically Priorean semantics in which tense is preserved in the metalanguage, so that the indexical sense is "displayed" in the axiom, as in (iii): (iii) Val(x, PAST) iff x was true Ludlow notes the "modest" character of such axioms, defending his approach by arguing that when we evaluate propositions in the past or future tenses we do not imagine some past or future time and determine whether the corresponding present-tensed proposition is true or not. Rather, we examine the evidence available to us right now. This, he claims, is an epistemological advantage to the A-theory. In general, the axioms proposed for the A-theory semantics retain all indexical and temporal relational morphemes of the object language in the metalanguage. Tense morphemes are given axioms like (iii) above, with complex tenses resulting from nesting of simple tenses. Temporal connectives and temporal adverbs also receive a disquotational treatment, as in (iv) and (v). (iv) Val(T, [S1 _when_ S2]) iff Val(T, S1) when Val(T, S2) (v) Val(x, _yesterday_) iff x was true yesterday Also in this chapter, Ludlow notes some linguistic objections that have been raised against an A-theory semantics, notably its inability to account for temporal reference (to be dealt with in Chapter 8), and the major philosophical objections, chiefly the so-called McTaggart paradox. He recapitulates Prior's defense of the A-theory, leading to the version of the A-theory known as presentism. Chapter 8 is perhaps the chapter of most interest to the linguist reader of STT. Here, Ludlow proposes that an A-theory can solve the problem of temporal anaphora by treating them as E-type anaphors in the sense of Evans (1977), or rather, of Cooper (1979) -- expressions that stand proxy for Russellian definite descriptions. According to this approach, in a discourse like (vi), the anaphoric pronoun stands proxy for a definite description, made explicit in (vii). (vi) A man came in. He tripped over a chair. (vii) A man came in. The man who came in tripped over a chair. This, on the theory, frees us from the assumption that there is some object that is the referent of the pronoun; the second sentence makes a general claim about the world, namely that "the world contains exactly one man who came into the room, and he tripped over the chair" (from the Introduction, p. 10). The proposal is, then, that the anaphoric element so frequently observed in temporal discourse is an E-type anaphor, often implicit, sometimes surfacing as "then". In either case, it stands proxy for a when-clause (or other temporal clause). Thus, a sentence like the second one in (viii) could be spelled out along the lines of (ix) (p 11). (viii) Sam adressed Bill. Bill didn't respond (then). (ix) Sam adressed Bill. Bill didn't respond when Sam adressed him. In a case such as this, the previous discourse provides the spelling-out of the anaphoric element, just as with "ordinary" E-type anaphors. In the proposal as developed in chapter 8 and appendix T5, however, there is no context provided and hence no spelling out of the implicit temporal clause. The work which the implicit temporal clause does for Ludlow is similar to that of the R point in Reichenbach's analysis. Each tense morpheme is the manifestation simultaneuously of Absolute Tense and Relative Tense. The former takes the form of a Priorean tense operator, the latter is a temporal connective (possibly implicit, but present in the LF representation). T-axioms are of the form in (x)-(xi). (x) Val(x, PAST) iff x was true (Absolute Past) (xi) Val(T, [IP1 _when_ IP2]) iff Val(T, IP1) when Val(T, IP2) (Relative Present) For a simple past sentence like "Smith swam", applying the axioms to the proposed LF will give us truth conditions like (xii) (ILFs appear within closed square brackets): (xii) "Smith swam" is true iff  there is an e, e is a swimming, Smith is the agent of e and e culminates  was true when  [...]  was true He notes that "when" can no longer , as in the B-theory, mean "at the same time as" -- instead, it must be "a kind of primitive". Ludlow assumes that implicit when-clauses have the same structure as explicit when-clauses, specifically (1) they are tensed, never infinitival and (2) they are "coordinated with the tense of the matrix clause". This "coordination" means that the when-clause has the same (absolute) tense operator as the matrix, whether PRES, PAST, or FUT. Finally in this chapter, he gives a novel solution to the McTaggart paradox, by invoking the implicit when-clauses which on his theory are present in every tensed sentence. A proposition that was future and is now past will be coupled with different when-clauses for the different cases. Ludlow's example is the proposition  there is a dying of Queen Anne , which was future "when Queen Anne was born" and is past "as I write these words". There can be no contradiction, since the content of the when-clauses is different. Chapter 9, "Broadening the investigation", brings in psycholinguistic data from langauge acquisition and language disorders to bear upon the A- versus B-theory controversy. Ludlow argues that studies of the order of acquisition render support to the A-theory semantics as the real one. Ludlow also claims support from acquisition for "when" as a "primitive" notion, distinct from the notion "at the same time". He calls in Merleau-Ponty to support the idea that presentism is also in accord with our temporal intuitions, fairly suggesting that intuitions of time passing are induced by philosophy classes. "If we concentrate on _our_ experience of the world, we have to reject the idea that time is a process or that it involves reference to independent future and past events. Rather, we have to think of the future and past being, as it were, in the present." (p. 143). At this point, Ludlow's position starts having very far-reaching consequences, which he draws up in a rather speculative manner in the final chapter. For philosophy, there is the apparent loss of truth-value links, and consequences for the nature of memory. For linguistics, the consequence is that the grammatical category of tense must be eliminated, along with any talk of specifically temporal anaphora, adverbs, etc. What we are accustomed to calling tense is really " a mixture of modality and evidentiality" (p. 163) The indexical character lies not in the time dimension but in evidentiality, with aspect a "more abstract" form of evidentiality. Languages that have been analyzed as having tense morphemes, aspect morphemes and evidential morphemes "simply have three kinds of evidentials" (p 161 ff)
EVALUATION A book that tries to balance, on less than 300 pages, the interests of the linguist working in tense semantics with that of the philosopher working in the metaphysics of time stands a fair chance of satisfying neither. Both kinds of intended reader will feel that important issues in his or her field have been passed over, with little hint as to how they should be addressed in the proposed framework. The author signals his awareness of this risk on p xvi ff, but he also seems to fear that his linguistic readership will bored by too much philosophy, and vice versa. This fear is certainly unfounded. The linguist or philosopher who opens a book such as this one is in all probability quite willing to listen to "the other side". But the brevity, even sketchiness, of much of the argument is likely to leave the linguist without a true appreciation of the philosophical import of the discussion, and the philosopher without means of evaluating the linguistic side of the argument. Even quite brief reviews of the issues raised, aimed at readers from the other discipline, would have been most helpful. Not all of us are lucky enough to be placed in joint philosophy and linguistics departments, where there are suitably qualified colleagues to pester. In general, if the book places quite high demands on the patience of the reader, it is because so many possible lines of argument are picked up, shown to have far-reaching consequences for all kinds of issues, and then dropped again, often with a comment to the effect that "a lot of interesting work remains to be done".
[A comment from the linguist reviewer, largely on the analysis of temporal anaphora:] Those of his readers who come from philosophy will have to judge whether Ludlow has succeeded in showing the relevance of linguistic analysis to the metaphysical pursuit (see below for one such assessment). As to the relevance for linguistics of the metaphysical argument remains unclear, it is not even obvious that Ludlow meant it to have such relevance. The A-theory "needs to avoid temporal reference" (p. 130), and this concern guides the analysis, rather than linguistic facts and intuitions. Let me expound this a bit. Ludlow rightly points out what are the big hurdles for the respective theories from a linguistic point of view, namely indexical expressions for the B-theory, and temporal anaphora for the A-theory. Most linguists would probably assume that both exist in language, and that consequently, both theories are wrong and right to about the same extent. Neither can be reduced to the other; we need resources from both series to account for the full range of temporal phenomena in language. The pure A- or B-theorist cannot take this way out, but must in some way "explain away" the offending phenomenon from language, at least if he takes the position that language in some sense mirrors reality (a fortiori if he endorses language-world isomorphism). The success of Ludlow's proposal then hinges on whether his explaining away of temporal anaphora is better than the B-theorist's explaining away of indexicality. Ludlow claims the latter fails because it cannot at the same time account for the different meanings of "today" (uttered on March 12) and "March 12" (uttered on March 12) AND for the, according to Ludlow, contingent falsity of propositions like "There is no language". The intuitions that have to be appealed to to substantiate this claim appear less than clear (especially if we restrict ourselves to philosophically "untutored" intuitions, as Ludlow proposes on p. 142). Let's see how Ludlow's own proposal for explaining away temporal anaphora fares in this respect. As we saw, the proposal is basically that a sentence which seems to be "about" some past time, like "Smith swam", contains an implicit when-clause, which could be made explicit, say "when I threw him in the water" (assuming I did so only once; recall that the when-clause is a Russellian description). We should get truth conditions as in (xiii). (xiii) "Smith swam when I threw him in the water" is true iff  there is a swimming by Smith  was true when  there is a throwing-in-the-water of Smith by me  was true. But how are we to interpret these truth conditions? The sort of situation we seem to interpret the natural language sentence as describing is one of an episode of two consequentially connected events: I threw Smith in the water, and (his way of dealing with this was that) he swam. In temporal terms, his swimming follows "just after" my throwing him (cf Partee 1986), but as Moens (1987) and others have shown, felicitous linking of two events by "when" requires that there be a more-than-temporal relation of consequentiality between them (for a critique of Moens and a more detailed proposal, see Sandstrom 1993). Ludlow is certainly correct when he says that "when" cannot mean "at the same time as" (nor can it mean "just after", as Partee suggested). When-linking of two events always envokes an episodic interpretation. With states, however, it is different. Sentences of the form STATE1 when STATE2 mean that STATE1 holds throughout the time picked out by STATE2. Thus given the fact that the Social Democrats held office in Sweden for an unbroken period of more than 40 years, which included my childhood, (xiv) is fine (while the inverse version in (xv) is not). (xiv) The Social Democrats held office when I was a child. (xv) I was a child when the Social Democrats held office. Also note that with states there is no longer any consequentiality requirement; the situations have nothing to do with each other save that the one serves to "date" the other. Now return to the truth conditions in (xiii) above. Here, what appears on either side of the (metalinguistic) "when" are two states, " ...  was true". Interpreting this using our ordinary language intuitions will land us somewhere very different from where we started, with the sentence "Smith swam when I threw him in the water". (And use our natural language intuitions we apparently must, else the T-theory will be no better than "Semantic Markerese", which Ludlow rejects in Chapter 2.) We are now dating the hold-time of a particular state (the being true of "there is a swimming by Smith") by means of the hold-time of another state (the being true of "there is a throwing-in-the-water of Smith by me"). These are simply the wrong truth-conditions. Thus, it is not clear that Ludlow's attempt at explaining away temporal anaphora is successful. (In fact there are more problems with it, which it would take us too far afield to discuss here.) Linguists will probably not be induced by it to change their predilection for an "impure" approach that utilizes both A-theory and B-theory resources.
[A comment from the philosopher reviewer on Ludlow's approach to metaphysics and his claim that his thesis solves McTaggart's paradox:]
It has been argued, e.g. by Craig (1998), that a presentist ontology "adroitly avoids McTaggart's Paradox because the only intrinsic properties there are are present tensed and therefore compatible". It would appear to follow that a presentist semantics, like Ludlow's, would also avoid the contradiction, i.e. if it shows that sentences that _appear_ to refer to entities that exist in the future and past, really do not, and therefore do not involve metaphysical commitment to the existence of such entities. It would not be enough, though, to avoid commitment to the existence of future and past entities. A presentist semantics must be able to tell us what the difference is between future and past tensed sentences, and account for temporal anaphora. Treating temporal anaphora as E-type does the trick, according to Ludlow. Well, it seems to me that Ludlow's approach to metaphysics makes it difficult to decide whether or not he succeeds in solving McTaggart's paradox, or rather, in arguing successfully that McTaggart was wrong about reality. It is often overlooked that McTaggart's argument is not about language, and its conten, but about the reality it describes (see e.g. Ingthorsson 1998). Ludlow's thesis is that one can study metaphysical questions by studying the semantics of natural languages, because "there is a close connection between language and reality" (p. xiv). To oversimplify: doing metaphysics is just doing empirical linguistics; find out how the world is by finding out how language mirrors the world. Ludlow's own criteria for success seem then to be that in order to show that the world is presentistic, one must show that, as a matter of empirical fact, natural language does not contain any such thing as tense, or "at least not the sort of tense system that is compatible with currently favored philosophical theories of time" (p. xiv). As pointed out above, there is reason to doubt that he is entirely successful on this point, although he has certainly made a serious attempt to challenge the thesis that tense is anything like currently favoured theories of time claim it is. The truth of the matter may of course be that natural languages contain traces of 'philosophical' tense, traces of E-type temporal anaphora, as well as traces of all the other conflicting and more or less confused conceptions of time that are around. If this is the case then Ludlow does not have a good case against McTaggart. McTaggart interpreted tensed language according to his fully developed metaphysical system, and found it to contain a contradiction (Ingthorsson 1998). A presentist metaphysics, as Craig (1998) points out, gives an interpretation of tensed language that does not contain a contradiction. Both McTaggart's view and the presentist view are not dependent on how natural language actually is, they can just claim that natural language just describes how the world appears to us, not how it really is. Ludlow makes his argument dependent on how natural language actually is, because he claims to be discovering how reality is by a study of language. That makes his argument weaker, from a metaphysician's point of view. I think however that Ludlow presents enough persuading arguments to support the thesis that natural language does not just contain the 'philosophical' tense, but also a more presentistic tense, and that therefore presentism is not just a philosophical view but one of our untutored intuitions about how the world is. This is important in motivating the shift of focus from a more traditional A-theory of time to a presentist A-theory of time.
[Conclusion:] For the reader who sticks it out until the end, there is a lot of good to be found in this book, not least in the form of almost parenthetical suggestions made along the way. Also, Ludlow's conclusions are radical enough to force anyone not willing to follow him into that particular wilderness to carefully think through precisely where the argument "went wrong". This in turn leads one to explore a lot of little paths of enquiry that one had hitherto neglected. Given the sort of intellectual temperament Ludlow seems to be gifted with, this would probably be just what he wanted for his readers. A rough and often quite unhelpful guide will in all likelyhood make you better acquainted with a place than one who leads you by the hand.
There seem to be few mistakes for a book with a reasonable amount of formal notation. We noted the following: p. 71, l. 5 from bottom: reference to a section 3.5 that does not exist p. 100, example (6): the FUT rule is repeated and there is no PRES rule p. 119, LF for Future in Past should have "after" instead of "when" p. 134, ll. 15 and 18: reference to example (63) seems mistaken p. 222, fn. 6, example (7'.a), second conjunct of the right-hand side should have CONJ' instead of CONJP.
REFERENCES Chomsky, N. 1986. Knowledge of Language. New York: Praeger. Cooper, R. 1979. The interpretation of pronouns. In Syntax and Semantics 10, ed. E. Heny and H. Schnelle. New York: Academic Press. Craig, W.L. 1998. McTaggart's Paradox and the problem of temporary intrinsics. Analysis 58: 122-127 Evans, G. 1977. Pronouns, quantifiers, and relative clauses. Canadian Journal of Philosophy 7:467-536. Ingthorsson, R. 1998. McTaggart and the unreality of time. Axiomathes 9: 287-306. Larson, R & G. Segal. 1995. Knowledge of Meaning. Cambridge: MIT Press. Moens, M. 1987. Tense, aspect and temporal reference. Ph.D. thesis, Centre for Cognitive Science, University of Edinburgh. Partee, B. 1984. Nominal and temporal anaphora. Linguistics and Philosophy 7:243-286. Sandstrom, G. 1993. When-clauses and the temporal interpretation of narrative discourse. Ph.D. thesis, Department of Linguistics, Umea university.
Gorel Sandstrom is Lecturer in General Linguistics at the Department of Philosophy and Linguistics, Umea university. Her research areas include the syntax of nominals and the semantics of tense and aspect.
Rognvaldur Ingthorsson, also of the Department of Philosophy and Linguistics, Umea university, is working on a Ph.D. thesis in the philosophy of time, notably on the debate between the A- and B-views.