Author: Thompson, Ellen Title: Time in Natural Language Subtitle: Syntactic Interfaces with Semantics and Discourse Series: Interface Explorations 11 Publisher: Mouton de Gruyter Year: 2005
Judith Tonhauser, Department of Linguistics, Stanford University
Ellen Thompson's book _Time in Natural Language_ investigates the syntax-semantics interface in the domain of temporality. The book's central claim is that syntax has an effect on the temporal interpretation of sentences and discourses. Thompson (T, henceforth) proposes a mapping between the semantic representation of temporality and the structure of clauses. This mapping is argued to account for a diverse set of English data in the 7 chapters of the book (which also contains a list of references and an index).
Chapter 1 of _Time in Natural Language_ (pp.1-14) presents the empirical domain of the book and introduces its theoretical framework. Empirically, T is concerned with phenomena which, she argues, illustrate the effect of syntax on temporal interpretation. One such phenomenon is the position of temporal phrases like ''at 2pm'' in examples like (1).
(1a) Mary had left the store at 2pm. (1b) At 2pm, Mary had left the store.
(1a), where ''at 2pm'' is clause-final, can mean either that Mary's leaving occurred at 2pm (event time modification), or that she was already gone by 2pm and her leaving was prior to 2pm (reference time modification). (1b), where ''at 2pm'' is clause-initial, only has the latter interpretation. Thus, the position of the time adverbial correlates with different temporal interpretations. _Time in Natural Language_ argues that this empirical phenomenon (as well as others discussed in the book) should be analyzed as an effect of syntax on temporal interpretation and the book offers a general framework for capturing this effect.
Formally, _Time in Natural Language_ is based on the Minimalist Program (Chomsky 1995) and Reichenbach's (1947) analysis of temporality, which uses three times to represent the temporal meaning of clauses (event time E, reference time R, and speech time S). T proposes a mapping from the semantic representation of temporality to the structure of clauses: according to this mapping, the three times E, R, and S are semantic features associated with the heads of VP, AspP and TP, respectively. The tree in (2) illustrates the structure of clauses assumed by T ((16), p. 7).
English tense morphemes order R with respect to S, which is the speech time in the default (non-embedded) case. In the past tense, R precedes S (written as R_S), in the future tense S precedes R (S_R), and the two times are contemporaneous in the present tense (written as S,R). The aspectual morpheme ''have'' orders E with respect to R: when ''have'' is present, E precedes R (E_R), absence of ''have'' results in E overlapping R (E,R). The common theme of the analyses presented in chapters 2-7 is that the functional position in which an expression is located or to which an expression is adjoined (i.e. TP, AspP, VP) corresponds with the time relative to which the expression is interpreted (i.e., S, R, E, respectively).
Chapter 2 (pp. 15-49) discusses the temporal interpretation of time adverbials like ''at 2pm'' (cf. example (1)). T proposes that such adverbials can adjoin either to AspP (the functional projection associated with the reference time R) or to VP (the projection associated with the event time E). In the former case, the adverbial temporally locates R, resulting in the interpretation of (1a,b) that Mary is gone by 2pm and her leaving occurred at a time prior to 2pm. In the latter case, the adverbial temporally locates the event time E, resulting in the interpretation of (1a) where Mary left at 2pm and is observed to be gone at a time after 2pm.
The formal analysis implements (in the Minimalist Program) Uriagereka's (1988:208) assumption that adverbials can only modify the denotation of the part of the phrase they syntactically govern. T presents data from a wide variety of phenomena, including linear order restrictions, coordination, and preposition stranding, in support of this analysis.
Chapter 3 'Adjunct Clauses and the Structural Representation of Simultaneity' (pp.51-85) develops the idea that there is a correlation between the temporal interpretation of adjunct clauses headed by e.g. ''before'', ''after'' and ''while'', and their adjunction site.
(3a) John slept while/when Mary sang. (3b) John slept after/before Mary sang.
T proposes that temporal adjunct clauses with a simultaneous reading like (3a) are adjoined to VP (which is associated with the event time E) whereas temporal adjunct clause with a nonsimultaneous reading like (3b) are adjoined to TP (which is associated with the speech time S). The relation between the position of a clause and its interpretation is captured by the condition in (4) (T's (23a), p.60):
(4) Condition on Simultaneous Interpretation: In order for an Event time ALPHA to be interpreted as simultaneous with an Event time BETA, ALPHA must be in the modification domain of BETA.
According to (4), the two events in (3a) are interpreted as simultaneous because a clause headed by ''while'' or ''when'' is adjoined to VP. Hence, the two clauses are in the same modification domain (according to Chomsky's (1995) definition of a checking domain). The two events in (3b), on the other hand, are not interpreted as simultaneous because a clause headed by ''before'' or ''after'' is adjoined to TP, and therefore does not end up in the same modification domain as the main clause. Again, T presents support for this analysis with data from a variety of phenomena, including preposition stranding with temporals, direct object/adjunct asymmetries, and ellipsis structures.
The empirical focus of Chapter 4 'The Temporal Syntax of Arguments: Reduced Relatives in Subject Position' (pp.87-117) is the interpretation of gerundive relative clauses, such as ''waiting for flight #307'' in (5).
(5) The passengers waiting for flight #307 left the room.
T argues that such clauses do not possess their own speech time S since they do not have a TP projection. The proposal is that their interpretation is determined by the position in which the noun phrase that embeds the reduced relative is interpreted. There are two possibilities: if the noun phrase is interpreted in Spec of TP, the noun phrase and the reduced relative are temporally interpreted at the speech time S, whereas an interpretation of the noun phrase in Spec of VP results in an interpretation at the event time E. T goes on to show how this analysis accounts for data from coordination, existential constructions, and the scope of quantificational adverbs.
Chapter 5 ''Principles of Time in Discourse: Temporal Syntax beyond the Sentence'' (pp.119-155) turns to the temporal interpretation of discourses. In this chapter, T focuses on the interpretation of discourses with ''then'', as in (6).
(6) Mary will speak to the reporters. (6a) Then Bill will photograph her. (6b) Bill will photograph her then.
The empirical observation (which goes back at least to Schiffrin 1992 and Glasbey 1993) is that clause-initial ''then'' (6a) leads to a sequential interpretation of the two sentences in the discourse, whereas clause-final ''then'' (6b) leads to a contemporaneous interpretation (with simple tenses). T argues that the ''temporal interpretation across sentences in discourse is subject to the same principles as temporal interpretation within sentences'' (p. 150). While in previous chapters the adjunction of clause-internal adjuncts were argued to correlate with different interpretations, T here accounts for the interpretation of discourses in relation to the position of ''then''. She proposes a constraint to the effect that clause-initial ''then'' links the reference times of the two clauses (hence allowing the event times to be ordered) while clause-final ''then'' links the event times of the two clauses (and hence requires a contemporaneous interpretation of the events).
Chapters 6 and 7 extend the analysis of the relation between the structural representation of temporality and temporal interpretation to the subclausal level. In Chapter 6 ''The Structure of Aspect'' (pp.157-181), T suggests that the position in which a verb is interpreted correlates with whether the event denoted by the verb phrase receives a telic or an atelic interpretation. In particular, telic events are interpreted in AspP while atelic events are interpreted lower than AspP. She presents support for this proposal on the basis of the distribution and interpretation of the adverb ''quickly'', and the syntax of durative and time frame adjuncts.
In Chapter 7 ''Syntax and Semantics of Aspectual Verbs'' (pp.183-204), T argues that the meaning of aspectual verbs like ''start'', ''begin'', ''continue'', ''keep'', ''finish'', ''cease'' correlates with the syntactic position in which they are interpreted. Aspectual verbs like ''start'' and ''begin'', which denote the initial phase of an event, are interpreted in V. Verbs like ''continue'' and ''keep'', which denote the middle phase of an event, syntactically raise from V to v. And verbs like ''finish'' and ''cease'', which denote the end of an event, raise from V to v to Asp. T proposes that the ''beginning, middle, and end of an event are represented syntactically by the beginning, middle and end of the verb chain created by verb raising from V to v to Asp(ect)'' (p. 184). Verbs such as ''begin'' and ''start'', for instance, which describe the beginning of an event have only the tail of the verb chain interpreted at LF. T further motivates this proposal with data from the behavior of ''only'', existential and extraposition constructions, and quantifier ambiguities.
_Time in Natural Language_ lives up to its goal of presenting an analysis of the effect of syntax on temporal interpretation, and the book gives a unified analysis of an impressively diverse set of English data. The proposed mapping between the semantic representation of temporality and clause structure is a relatively simple one, and it is intriguing to see how far T can take it in drawing connections between temporality and other domains of grammar, such as word order, scope, presuppositions, and extraction. The book illustrates one particular way in which the syntax-semantics interface in the domain of temporality can be viewed, which makes it interesting to researchers working on temporality or the syntax-semantics interface, more generally.
The book is well-written, and the structure of the argument is clearly presented and easy to follow. Despite its theoretical orientation, the book should be easy to understand even with minimal background in the Minimalist Program or Reichenbach's analysis of temporality. I now raise three concerns with T's proposal.
My first concern relates to the use and presentation of data. The interpretation of many semantic phenomena, including temporality and focus sensitive operators like ''only'', depends on the discourse context. In several cases, T's grammaticality judgements in _Time in Natural Language_ are hard to follow because of the lack of a discourse context. (7), for instance, which T deems ungrammatical, is considered fine by two native English speakers I consulted in a context where the question under discussion is what will happen tomorrow at 3pm, after Mary delivers her speech at 2pm.
(7) *Mary will have spoken to the reporters. Bill will photograph her then. ((50), p.144)
Similarly, on p.187, T claims that (8a) (T's (7a)) does not have the interpretation in (8b) (T's (7c)), which she marks with a *, just like ungrammatical sentences.
(8a) John only continued the book. (8b) *John continued the book, as opposed to doing something else to it.
Again, my native speaker consultants find the example perfectly acceptable in the following context: John has been writing a book for 3 years now and is way behind his deadline. He has had writer's block from January to April 2006 but has finally started writing again in May 2006. His publisher inquires with a mutual friend whether John has managed to finish the book by now (July 2006), but his friend reports ''John only CONTINUED the book''. In this context, (8a) has the interpretation in (8b).
The inclusion of discourse contexts would have benefited the reader's understanding of T's judgements.
In other cases, naturally occurring data and English consultants' judgements do not accord with T's grammaticality judgements. For instance, in chapter 7, T claims that aspectual verbs like ''continue'' do not occur in existential there-constructions, as in (9) (T's (12b), p.190).
(9) *There continued a lecture on anaphora.
A Google search finds numerous examples with ''there continued a'', for instance (10):
(10) Down the Dark and Middle Ages there continued a constant struggle by enlightened men to use their minds without losing their heads. (http://www.butterfliesandwheels.com/articleprint.php?num=41)
The example in (11) (T's (46b), p.107) is claimed to be ''slightly marginal'' (which T indicates with a ?), again to the disagreement of my English consultants.
(11) ?A passenger waiting for the announcement entered the room from California.
In sum, the empirical side of the argumentation in _Time in Natural Language_ would have benefited from the use of naturally occurring data and the presentation of discourse contexts. Given that the empirical focus of the book is English, a language for which there exist many corpora and native speakers are easily accessible, the book's reliance on constructed examples is puzzling.
My second concern is a theoretical one, and relates to the semantic theory presented in ''Time in Natural Language.'' For a start, T assumes a division of E, R and S into a tense relation, which relates S and R, and an aspect relation, which relates E and R. This division is not part of Reichenbach's original proposal, on which T relies in her book, but T neglects to cite authors who have proposed such a division, like Comrie (1985) and Klein (1994). Next, T's introduction to Reichenbach's system is rather informal, and it is misleading with respect to the characterization of the reference time R. On p.4, T introduces E, R and S on the basis of the sentence in (12).
(12) At 2:00, Mary had left the store.
T claims (p.4) that the event time E is ''the time of Mary's leaving'' and the reference time R is ''the time by which Mary leaves (in this sentence, 2:00).'' Typically, however, the reference time (if is given a characterization at all) is described as the time from which the event is presented (cf. Klein's (1994) conceptualization of the reference time as the topic time). Thus, the reference time R in (12) is 2:00, but it is not the time by which Mary leaves!
The semantic analysis of particular phenomena in _Time in Natural Language_ is often imprecise. A particularly striking example of the weakness of the semantic analysis is the proposed interpretation of sentence-medial ''then''. Just like sentence-initial ''then'', sentence-medial ''then'' results in ordered interpretations of the events denoted by the two clauses in a discourse like (13) (T's (24a) p.132).
(13) Mary will speak to the reporters. Bill will then photograph her.
T claims that sentence-medial ''then'' requires the reference time R of its clause to be linked to the reference time R of the previous clause. Thus, as illustrated in (14) (T's (24b) p.132), T's analysis predicts that the interpretation of (13) results in linked speech times S, linked reference times R, and that S precedes R (since the two clauses in (13) are in future tense).
(14) S_R,E (representation of ''Mary will speak to the reporters.'') | | (link between the S and R of the two utterances) S_R,E (representation of ''Bill will then photograph her.'')
Now, T claims that ''[s]ince the Event times of [(14)] are not linked, they are interpreted as noncotemporal, and the ordered reading of [(13)] results'' (p.132). This is not correct: although the two Es are not linked, they are cotemporal with their respective Rs (R,E), since the clauses do not contain the aspectual morpheme ''have'' (which orders R and E). Thus, since the Rs are cotemporal, and the Es are cotemporal with the Rs, it follows from (14) that the Es are cotemporal, too, despite T's assertion to the contrary. T addresses this dilemma in footnote 9 (p.152), where she writes that ''times are interpreted as cotemporal only if they are linked''. This, however, explicitly contradicts her assumption that if ''two times are separated by a comma, they are interpreted as contemporaneous'' (p.4), which is also the way Reichenbach's times E, R and S are typically used.
In sum, while _Time in Natural Language_ certainly strives to integrate a semantic analysis to the syntactic analysis of temporality, its execution is not always satisfying.
The third issue I would like to raise here concerns the conceptualization of the syntax-semantics interface promoted in ''Time in Natural Language.'' As mentioned above, the general tenor of the book is that observed meaning differences between related constructions can be accounted for by different syntactic structures. In chapter 3, for instance, T acknowledges the interpretive differences between connectors like ''before/after'' versus ''when/while'', but the interpretation of clauses featuring these connectors are not directly attributed to the meaning of the connectors. Instead, the meaning differences are only indirectly attributed to the connectors: clauses headed by ''before/after'' adjoin to a different site than clauses headed by ''while'', and the adjunction site determines whether the clause receives a simultaneous or non-simultaneous interpretation. Similarly, the analysis of aspectual verbs in chapter 7 does not directly implement their meaning in the semantic analysis but encodes the differences in the syntactic analysis. Thus, _Time in Natural Language_ analyzes observed meaning differences in the syntactic representation of time rather than in the semantic one, thereby assigning weight to the syntactic component of the syntax-semantics interface in temporality where others have proposed semantic solutions (e.g. Schiffrin (1992) and Glasbey (1993) for the interpretation of ''then'', Beaver and Condoravdi (2003) for the interpretation of ''before'' and ''after'', and Partee (1984) and Kamp and Reyle (1993:ch5) more generally). Previous semantic analyses, if mentioned at all, are not discussed in ''Time in Natural Language.'' While I do not want to argue here against the particular view of the syntax-semantics interface presented in ''Time in Natural Language,'' I believe the book would have benefited from a discussion of the way in which the syntax-semantics interface is conceptualized and the ways in which T's approach is different from and/or superior to semantic analyses.
In conclusion, Thompson's _Time in Natural Language_ is of interest to researchers working on the syntax-semantics interface in temporality as a noteworthy and thought-provoking take on the topic. I find the book less suitable as a reading for a semantics seminar on temporality for the reasons outlined above.
A final note on editing: I have encountered over 60 spelling mistakes in the book (in particular in chapters 1, 5, 6 and the references), which is appalling for a book that costs over $100.
Beaver, David and Cleo Condoravdi (2003): A uniform analysis of ''before'' and ''after'', in Rob Young and Yuping Zhou (eds), Proceedings of SALT XIII, Cornell: CLC Publications, 37-54.
Chomsky, Noam (1995): The Minimalist Program. Cambridge, MA: MIT Press.
Comrie, Bernard (1985): Tense. Cambridge: Cambridge University Press.
Glasbey, Sheila (1993): Distinguishing between events and times: some evidence from the semantics of then. Natural Language Semantics 1, 285-312.
Kamp, Hans and Uwe Reyle (1993): From Discourse to Logic. Dordrecht: Kluwer.
Klein, Wolfgang (1994): Time in Language. London: Routledge.
Partee, Barbara (1984): Nominal and temporal anaphora, Linguistics and Philosophy 7, 243-286.
Reichenbach, Hans (1947): Elements of Symbolic Logic. New York: The Macmillan Company.
Thompson, Ellen (1996): The syntax of tense. Ph.D. dissertation, Department of Linguistics, University of Maryland.
Uriagereka, Juan (1988): On government. Ph.D. dissertation, Department of Linguistics, University of Connecticut at Storrs.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER:
Judith Tonhauser is about to receive her Ph.D. in Linguistics from Stanford University, and will join the faculty of the Linguistics department at the Ohio State University in September 2006. Her research interests lie in crosslinguistic syntax and semantics, with an empirical focus on American indigenous languages. She has conducted fieldwork on Yucatec Maya (Mexico) and Guarani (Paraguay), and has worked on the temporal interpretation of noun phrases and of tenseless languages, the syntax and semantics of questions and focus constructions, and voice and argument realization in Mayan languages.