Semantics-Syntax Interface (Fall 2014)

Semantics‐Syntax Interface Volume 1, Number 2 Fall 2014 E‐ISSN 2383‐2010 Semantics‐Syntax Interface Volume 1, Number 2 Fall 2014 Editor‐in‐Chief Negin Ilkhanipour University of Tehran Editorial Board Laura Bortolotto Ca’foscari University of Venice Dustin Alfonso Chacon University of Maryland Natalia Cichosz University College London Despina Ikonomou MIT Moreno Mitrović University of Cambridge Ethan J. Poole University of Massachusetts at Amherst Ayaka Sugawara MIT Anqi Zhang University of Chicago Associate Editorial Board Octav Eugen DeLazero North‐Eastern Federal University Ilaria Frana University of Massachusetts at Amherst Thomas Grano University of Maryland Valentine Hacquard University of Maryland Tim Hunter University of Minnesota Jungmin Kang University of Connecticut Jonathan E. MacDonald University of Illinois at Urbana Ramin Rahmany University of Tehran The editorial board is sincerely honoured to be supported by Rajesh Bhatt University of Massachusetts at Amherst Guglielmo Cinque Ca’foscari University of Venice Kai von Fintel MIT Irene Heim MIT Sabine Iatridou MIT Angelika Kratzer University of Massachusetts at Amherst Howard Lasnik University of Maryland Barbara H. Partee University of Massachusetts at Amherst Ian Roberts University of Cambridge Editorial Assistant Parinaz Dadras University of Tehran CONTENT Original Articles Diego Gabriel Krivochen (Anti‐)Causativity and the Morpho‐ Phonology‐Semantics Tension 82 Short Contributions Sumiyo Nishiguchi Mirative Past in Japanese 118 Daniel Altshuler and Susanna Melkonian In Defense of the Reference Time 133 Corrigendum 150 Publisher University of Tehran Director‐in‐Chief Gholamhosein Karimi‐Doostan University of Tehran Cover Image Jan Koster E‐ISSN 2383‐2010 Website http://semantics‐syntax.ut.ac.ir Email semsyn@ut.ac.ir (Anti‐)Causativity and the Morpho‐ Phonology‐Semantics Tension Diego Gabriel Krivochen* School of Psychology and Clinical Language Sciences, University of Reading, Reading, UK Abstract: In this paper I will analyze the causative‐anticausative opposition from the point of view of semantic construal, and how syntax builds structures following semantic instructions that convey that information, without adding or deleting information. I will use causativity to analyze the tension that arises when a putatively universal semantic construal, (narrow‐)syntactically instantiated, is to be materialized using limited, language‐specific resources. This will touch on the subject of language typology, and its importance to describe the observable effects of this tension between semantics and morpho‐phonology, already noticed by Tesniére (1959). Our theoretical proposal will take mutually consistent elements from Conceptual Semantics, Relational Semantics, Lexical Decomposition, and Minimalism, in the search for the simplest (yet, empirically adequate) theory of the syntax‐ semantics interface. Consequences for comparative linguistics will be suggested, with particular emphasis on Slavic, Germanic, and Romance languages. Keywords: causativity, conceptual primitives, syntax‐semantics interface, cycles * Email: d.g.krivochen@pgr.reading.ac.uk I would like to express my deep gratitude to Peter Kosta, whose views on causativity and the relation between lexicon, syntax, and semantics inspired a good part of this paper, even in those aspects I do not agree. His input and feedback have been invaluable for me while writing this paper. I would also like to acknowledge the helpfulness of an anonymous reviewer, who carefully commented on the paper and proposed challenges to my theory I have done my best to address. This study was supported by the project Linguistic and lexicostatistic analysis in cooperation of linguistics, mathematics, biology and psychology, grant no. CZ.1.07/2.3.00/20.0161, which is financed by the European Social Fund and the National Budget of the Czech Republic. I also thank Monika Pitnerová, L’udmila Lacková, and Alicja Kobylecka for their kind help with grammaticality judgments for Czech and Polish examples. The usual disclaimers apply. Semantics‐Syntax Interface, Volume 1, Number 2, Fall 2014 82–117 © 2014 by University of Tehran http://semantics‐syntax.ut.ac.ir (Anti‐)causativity and the morpho‐phonology‐semantics tension 83 1. Introduction The phenomenon of causativity has been analyzed from various perspectives: purely syntactic, as the projection of certain functional heads (little v and Voice), purely semantic, following a logical approach to natural language semantics (within neo‐ Davidsonian semantics and Montague grammar), purely morpho‐phonological (focusing on the affixal expressions of causative meanings and alternations), and mixed, interface explanations. Interface explanations involve more than one of the aforementioned components, and, unlike single‐system accounts, they tend to resort to less intra‐theoretical stipulations – like feature valuation operations, and approach the problem from the convergence of two or more systems, thus more in tone with Minimalist desiderata (see, e.g., Marantz 1995:380381). For instance, Shibatani and Pardeshi (2001) focus on the syntax‐morphology interface, refining a previous classification of causative constructions in (a) lexical or synthetic kind (in which the root is causative per se), (b) morphological kind (in which there is a morpheme that determines causative alternancies, as in Japanese affix –sase), and (c) syntactic or periphrastic kind (e.g., make/cause + V), all three configuring stages of a single functional continuum, following Givón (1980). This account takes into consideration the Lexicon‐Morpho‐phonology road and functional‐typological variables, but tends to neglect the Logical Form (LF) of causatives, as well as semantic considerations in a broader sense (for instance, the role roots play in causative constructions, and whether causativity is part of the meaning of the root or of the whole construal, as I will try to problematize below). Schäfer (2008), to whom I will return below, takes a different stance, taking into account semantic issues from a syntactic perspective. His focus is set on the causative‐anticausative alternation (with which I will deal here) from a syntactic point of view, using data from a number of languages. The same perspective, although focused mostly on applicatives (Appl), is adopted by Pylkkänen (2002, 2008), partly based on previous neo‐Davidsonian LFs proposed by Kratzer (1996): while distinguishing neo‐Davidsonian semantics from neo‐Davidsonian syntax (such that a neo‐Davidsonian approach to LF does not entail a neo‐Davidsonian approach to narrow syntax), Kratzer introduces a new head, Voice, which Pylkkänen takes instead of Chomsky’s (1995) v (“litle v”), the later being also present in Marantz (1984)  a foundational work of the VP‐internal subject hypothesis in which v comprises broader values, related to any verbal suffixation including derivational suffixes  as a head comprising the LF‐interpretable dimension causativity and introducing an external argument, besides hosting φ‐features and signaling phase boundaries in more recent work (for example, Chomsky 2008). 84 D.G. Krivochen The distinction between VP and a higher functional node introducing causativity and an external argument, an initiator of the event denoted by VP, has also been advocated for from a Generative Semantics perspective, which is quite close in spirit to the one I argue for in the present paper. Culicover and Jackendoff (2005:96) mention works by McCawley, Ross, and Lakoff, arguing that causative constructions involve an underlying structure containing a CAUSE primitive which, after successive instances of a transformation dubbed Predicate Raising, CAUSE, and event‐related primitives (e.g., BECOME), are lexicalized and surface as a single item. The derivation Generative Semantics proposed for a sentence like [John opened the door] goes along the lines of (1), taken from Culicover and Jackendoff (2005:97), in turn borrowing it from Shibatani (1976): (1) a. S CAUSE John S BECOME S NOT S CLOSED door (Predicate Raising) b. S CAUSE John S BECOME S NOT CLOSED door (Predicate Raising) c. S CAUSE John S BECOME NOT CLOSED door (Predicate Raising) d. S CAUSE BECOME NOT CLOSED John door (Lexicalization) open (Anti‐)causativity and the morpho‐phonology‐semantics tension 85 Notice that in Generative Semantics there were no constraints as to the kind of complex structures that could be lexicalized: we have here a causative primitive, an eventive, change‐of‐state primitive, and a result predicate plus a negative operator (CAUSE + BECOME + NOT + CLOSED) all materialized by a single lexical item. As argued by Lakoff and Ross (1973), conditions apply to a set of derivational steps, not to a single representation level, which means that conditions apply all throughout the lexical derivation, not to the final product. Generative Semantics was deeply derivational, and assumed a syntactic structure for semantic construal, an idea I will develop in my own way in the rest of the present work. At this respect, it is useful to compare this perspective with that of Culicover and Jackendoff (2005:20, fn. 8): Algebraic combinatorial systems are commonly said to “have a syntax”. In this sense, music has a syntax, computer languages have a syntax, phonology has a syntax, and so does Conceptual Structure. However, within linguistics, “syntax” is also used to denote the organization of sentences in terms of categories such as NP, VP, and the like. These categories are not present in any of the above combinatorial systems, so they are not “syntax” in this narrower sense. (Culicover and Jackendoff 2005:20, fn. 8) In this paper, and in general within my theory, “syntax” is used in the wider sense, for two main reasons: to begin with, there is no compelling evidence that the “syntactic mechanisms” taken alone (without considering the elements involved, just the combinatory algorithm) vary from one system to another, except that the properties of the units affect the algorithm, in case that actually happens; and also, an adequately wide formalization of syntactic mechanisms could reveal deep facts about the structure of more than a single cognitive system. Moreover, there is recent evidence that syntax, in the sense in which I take it here and in previous publications, is not only evolutionarily previous to, but also simpler than, other forms of combination (Collier, Bickel, van Schaik, Manser and Townsend 2014). Therefore, semantic construal is syntactic in nature (as is a musical phrase; or an equation), insofar as it is structured, but syntax is to be understood independently of both sounds and concepts. From a stance compatible with (and in fact related to) Generative Semantics, Mateu Fontanals (2000, 2002, 2005) claims that VP is a purely transitional node between a causative node that licenses an external argument (which he calls R) and a locative node (which he calls r), and that VP takes no arguments. All arguments are selected by R and r, roughly equivalent to Mainstream Generative Grammar’s (MGG) vP and PP (notice the similarity between Mateu Fontanals’ proposal and Jackendoff’s 1987 Conceptual Semantics in which the dynamic primitive GO, corresponding to VP, takes a locative structure [PATH [PLACE]] as its complement and is in turn the complement of a causative primitive CAUSE). The importance of distinguishing an eventive projection from a causative projection at the syntax‐semantics interface will be obvious 86 D.G. Krivochen throughout the paper, and has also been stressed by the aforementioned authors. Separating cause from event is essential in order to account for alternations in which causative meaning is involved (e.g., alternating ergatives) with or without a materialized agent. R takes an initiator as its specifier and an event (VP, in Mateu Fontanals’ terminology, a transitional node T) as its complement, and r takes two arguments as well, in localist terms, a figure and a ground. Those roles, as in Hale and Keyser’s (2002) strongly componential theory (adopted by Chomsky 2004), are read off the syntactic configuration at LF, which is partly a function of syntactic structure. Causativity, in this paper, will be seen as a phenomenon to be analyzed from the syntax‐semantics interface. 2. Towards a Classification of Events (and Roots) My point of departure will be the theory outlined in considerable depth by Kosta (2011:287, 2014), who follows Schäfer (2008:142) – in turn, using terminology of Levin and Rappaport Hovav (1995)  and Alexiadou, Anagnostopulou and Schäfer (2006) in distinguishing four kinds of causative events (taking into account causative alternations of non‐causative roots, like ergative roots):  Agentive (murder, assassinate, cut…)1  Externally caused (destroy, kill, slay…)  Internally caused (blossom, wilt, grow…)  Cause unspecified (break, open, melt…) Schäfer (2008:142) assumes that “the encyclopedic information associated with Roots is now the factor which decides whether a verb undergoes the causative alternation or not”, and Kosta (2014) also proposes that the content of the root determines the availability of a causative alternation for unaccusatives (thus refining the typology of uncaused verbs to include the potentiality of allowing a causative construal) without the need to resort to featural specifications on lexical items (as HPSG does, among other frameworks). It is to be noticed that this stance implies distancing from traditional Distributed Morphology (DM) approaches (Halle and Marantz 1993; Embick and Noyer 2005), in which the encyclopedia (called “C‐List”) is distinguished from the set of roots (“A‐List”) insofar as roots are made up of formal U(niversal) G(rammar)‐given features (Embick and Noyer 2005; Panagiotidis 2013). 1 A reviewer has pointed out, correctly in my opinion, that cut might not be causative in the same sense as murder, given different specifications for their respective subjects. I share that objection, which applies to Schäfer’s examples (which I have reproduced here). (Anti‐)causativity and the morpho‐phonology‐semantics tension 87 Like Schäfer (2008), Borer (2013), Panagiotidis (2011), and Kosta (2011, 2014), among others, I will assume that roots are indeed interpretatively underspecified semantic substance, but, contrarily to Schäfer (2008:142), I will not claim that this semantic substance determines the availability of alternations (which would mean constraining the lexicon a priori without principled reasons), but that alternations are, in principle, always available for all roots2 (cf. Panagiotidis 2005), and coinage processes (including factors of socio‐historical nature) are not to be confused with system‐internal conditions: a root’s categorial and further semantic properties in a specific sentence are determined by their local relation with distributionally specified semantically interpretable functional‐procedural heads3 (see Borer 2013 for a related perspective; and Krivochen 2012:90 ff. for a development of this idea within Radical Minimalism). My proposal, as will be obvious below, is closer to Kosta’s (2011) full development, more componential and interface‐oriented than Schäfer’s (2008): I would like to propose an alternative semantic classification of cause functors, which only partially overlaps with Kosta/Schäfer/Alexiadou’s:  External cause (CAUSEEXT): particularly noticeable in [let]/[make] light V structures, includes an initiator and an affected object (prototypical transitivity).  Internal cause (CAUSEINT): there is no affectedness. It corresponds roughly to the ‘Activity’ Aktionsart (atelic, durative, intransitive).  Environmental cause (CAUSEENV): there is affectedness but no initiator in the sense of ‘specific, definite sortal entity’ (cf. Van Valin and Wilkins 1996:341315). The event is licensed by stative conditions of the environment4. 2 The idea was already present in earlier work by Beth Levin and Malka Rappaport Hovav (e.g. 2003, 2008), but there are major differences, one of which is the existence of a level of event structure, as well as a more specified root semantic template, closer to Schäfer’s (2008) account than to my own. Their lexical semantic representations are also closer to neo‐Davidsonian semantics than to Conceptual Semantics/Relational Semantics, the latter of which I follow in the present paper. 3 In other words, as Boeckx (2010:26) puts it, “[t]he complex whole is the output of simple processes and interactions, rather than the result of complex specifications.” As should be obvious, compositional semantics is precisely an example of this tendency: simple, atomic semantically interpretable elements combine in such a way that complex meaning is an emergent result (see Goldberg, forthcoming, for a very recent overview), a ‘function’ (Partee 1984:153) of elements – which can be as simple as one wants  and relations, optimally kept to a minimum. Boeckx’s attempt, and my own, is to eliminate ad hoc elements from both syntax and semantics, including, for instance, specific features to trigger particular syntactic operations with no reference to the interface systems of sound and meaning. 4 It is important to distinguish environmental cause from natural force (Schäfer 2012), which is limited to NPs denoting natural events (e.g., storms, earthquakes). In the sense I work with, those NPs would correspond to external causes, since volition is, in this account, not a necessary condition to be an external initiator, quite in the line of Holinsky (1987). I will come back to this below. 88 D.G. Krivochen Crucially, this does not mean I propose three v projections, rather, as the mentions of overall construal in each of the types suggest, I argue for inferential enrichments of the interpretation of the v node at LF, when building a full propositional form. In other words, there are no diacritics or multiplication of syntactic projections, but rather inferential processes in the construction of an explicature (Sperber and Wilson 1995; Wilson and Sperber 2004) which take into account the construal in which v appears and the semantic contribution of co‐appearing nodes, in terms that will be clearer below after introducing Mateu Fontanals’ (2000, 2002, 2005) semantic primitives. I will propose that the semantics of causativity is not directly mapped to the morpho‐syntax of transitivity, in a related claim to Kratzer (1996), insofar as there is no uniform mapping between syntax and semantics, or semantics and morpho‐phonology (a situation I have referred to in past works as “opaque interfaces”). Consider, for example: (2) a. The telephone moved. (because it has an internal device that allows vibration) b. The rock moved down the mountain. (because there is an environmental condition that licenses this movement, such as an inclined plane) Both examples in (2) display causativity but no transitivity (as we are dealing with one‐place predicates), and they differ in the source of the cause: in (2a), the cause of the event one can informally represent in predicate‐argument terms as move (telephone) is an internal device, inherent to the argument (therefore, corresponding to what I have called internal cause). (2b), on the other hand, displays an event move (rock) licensed by an external condition, but not an agent/initiator (a sortal entity that triggers the event). The syntactic structures corresponding to both unergative Vs are the same, displaying a vP‐VP dynamics5, but the thematic interpretation of the element in Spec,vP is not always that of an agent, as is the case of (2a–b). The determination of causativity types, then, goes beyond what syntactic nodes can tell us assuming bi‐ univocal mapping syntax‐semantics; otherwise, I should be forced to assume a rich inventory of functional projections and features, as in Nanosyntax (e.g., Fábregas and Putnam, forthcoming, on the syntax of middle constructions) or cartographic approaches, which in our opinion go against basic considerations of economy in 5 That is, [Agent [[CAUSE] [[DO] √MOVEMENT]]], with conflation of the root onto both DO and CAUSE. See, for instance, Jackendoff (1983, 1987) and Hale and Keyser (2002:15), who use some notational variants (V1 and V2 instead of vP and VP). See also Harley (2003), for some differences: she just assumes conflation of the root onto v. Considering that v conveys cause, I will argue that whenever there is v, there must be V (which conveys event), which justifies the extra layer. Empirical evidence is provided by causativized ergatives: they are events (i.e., VPs) that might or might not appear with a vP layer. (Anti‐)causativity and the morpho‐phonology‐semantics tension 89 syntactic representations, a basic methodological Minimalist desideratum. As should be obvious by now, I am already distinguishing different levels of analysis: on the one hand, causativity is a semantic notion, read off at LF from a syntactic‐semantic construal (i.e., a structure built by syntactic means, optimally just concatenation, manipulating only a minimal number of semantically interpretable elements), whereas transitivity is a morphological phenomenon (consider the accusative vs. ergative case paradigms, see Laka 2002 for an introductory overview), oten related to affectedness. In concrete terms, an external initiator in a syntactic‐semantic construal affects an internal argument via an event denoted by the verbal root. In this construal, the affected object (A relation which, according to Gropen, Pinker, Hollander and Goldberg 1991:158–159, includes possession, change of state/location, etc. identifies the moved theme in a locative construal with the affected object, a claim I share and expand. Configurationally, it corresponds to the figure in a PP, from a localist perspective; see, e.g., Anderson 1971) receives an accusative morphological exponent. Needless to say, causativity and affectedness do not always coincide, since Unergative construals do not license the position for an internal argument in the syntax unless those internal arguments are further specifications of a cognate nominal root (assuming a Hale and Keyser 2002 style for the derivation of denominal verbs; see also Mateu Fontanals 2000, 2002), as will be seen below. Causativity is taken here to mean only the presence of internal, external, or environmental action triggers in the syntactic‐ semantic construal without implying any of the other notions. These distinctions, which might seem anti‐economical, prove useful when considering concrete data, and provide the bases for a highly componential interface‐based system. Causativity, as an interface phenomenon, does not escape the general economy tendency to minimize Spell‐Out when possible, that is, reduce the number of overt elements liking form and meaning (see Krivochen and Kosta 2013:178 for a comparison with Nunes’ 2004:50 maximization of Spell‐Out links in a chain; also Krivochen, forthcoming for an application of this principle to the derivation of Wh‐interrogatives, Parasitic Gaps, and multiple gap constructions)6. This does not mean relying on a rich 6 Far from being a shortcoming, I explicitly claim that not only every linguistic phenomenon is an interface phenomenon, since natural language itself is a system that displays not only structural relations, but substantive elements as well, sound and meaning, but also that an explanation that rests only on syntactic ‘evidence’ (e.g., Agree, being the most pervasive in the literature) is stipulative as it rests on non‐falsifiable assumptions (e.g., the existence of features given by UG as something more than metaphors) and elements that, if only usable by the so‐called ‘narrow syntax’ (like Edge Features, or Agreement Projections) should be eliminated. If there is a change of meaning involved in a construal, I see no obstacle in considering there is interplay of syntax and semantics there. Notice that, for instance, the linking rules used by Levin and Rappaport Hovav (1995) are not narrowly syntactic (just like Jackendoff’s linking rules in the Parallel 90 D.G. Krivochen skeleton of phonologically null functional projections (as cartographical approaches do), but predicts a preference of synthetic forms (Shibatani and Pardeshi’s (a) and (b) groups) over analytical forms (Shibatani and Pardeshi’s (c) group) in language use to convey a certain meaning, as can be seen in the following examples. (3a) and (4a) are periphrastic, analytic causative, (3b) is a resultative construction of synthetic form, and (4b) with overt se clitic is causative in Masullo’s (1992) terms.7 (3) a. (?)John caused the metal to go flat by hammering it. b. John hammered the metal flat. (4) a. (?)Juan hace que el doctor García lo atienda. (Spanish) Juan make.PRES.3SG that the doctor García CL.ACC examine.PRES.SUBJ.3SG ‘John makes Dr. García examine him.’ b. Juan se Juan CL atiende con el doctor García. examine.PRES.3SG with the (Spanish) doctor García ‘Juan goes to Dr. García’s.’ This phenomenon has received the label of “lazy‐Spell‐Out” in Krivochen and Kosta (2013), and is a form of a general principle of economy, which Chomsky (2013:38, fn. 12) has formulated, informally, as “less is better than more”, as a sort of aproristic approach to economy. If this minimization of overt material is to be taken seriously, then deviations from the synthetic case must be justified in interface terms, particularly, in semantic terms: an analytic syntactic construal conveys some extra meaning, absent in the synthetic version (e.g., use of the se clitic in Spanish, resultatives in Germanic languages, among many other options); in other words, more materialized elements mean more nodes in the syntactic structure. In turn, if all nodes are to be justified from a semantic point of view, more nodes mean more elements in the representation to be read, which have to be justified by the cognitive benefits one takes from computing extra material (taking into account the implementational level of the Architecture), as they involve, at least, two levels: Lexical Semantic Representation and Argument Structure (see also Williams 2003 for a related view). 7 List of abbreviations in the glosses: accusative; GEN = genitive; SG CL = clitic; = singular; PL ERG = ergativizer; = plural; subjunctive; AUX = auxiliary; PART = participle. PRES PASS = present; = passive; PST = past; = reflexive; ACC = = perfective; SUBJ = REFL PERF (Anti‐)causativity and the morpho‐phonology‐semantics tension 91 proposal, contrarily to most narrowly‐syntactic proposals)8. This has as a consequence that periphrastic causative constructions are, when confronted with a non‐periphrastic version, rendered awkward by most speakers (when they are not directly ungrammatical, examples and judgments are taken from Kosta 2011:240), as if there was extra information (‘positive cognitive effects’, in Relevance Theoretic terms; see Sperber and Wilson 1995; Wilson and Sperber 2004), as in (5a) and (5b): (5) a. Karel Karel upustil dopis. (Czech) drop.PST.3SG the‐letter ‘Karel dropped the letter.’ b. *Karel Karel dal upustit dopis. give.PST.3SG fall (Czech) the‐letter ‘Karel made the letter fall.’ In this case, it is not the presence of an overt external sortal initiator for the event fall (letter) that influences grammaticality, but, I argue, the violation of a local morphological rule allowing conflation of a root onto a light node, following Hale and Keyser (2002). The structure of a causative construction I assume is Hale and Keyser’s (as well as Mateu Fontanals’ 2002, see also Pylkkänen 2008 for a related view but assuming a different, richer clausal skeleton), distinguishing three layers within the verbal domain (cf. Kosta 2011:251, who claims, following Larson 1988 and related work, that internal arguments are arguments of V, not of P, as I, following Mateu Fontanals 2000 et seq., do): (6) vP Causative layer Initiator v’ [CAUSE] VP [EVENT] Eventive layer PP Locative layer Figure P’ P 8 Ground This does not mean adhering to the Derivational Theory of Complexity, since there is not a relation of direct function between number of nodes/operations and interpretative complexity: it is relativized by the context‐sensitive notion of cost‐benefit relations. 92 D.G. Krivochen According to Mateu Fontanals (2000:5, 2002:33), the nodes cause, event, and P (the latter corresponding to the semantic primitive Location) can adopt two values each (not intended to be taken in the Minimalist sense of features to be valued, but as interpretative possibilities in a Relational Semantic structure, to be linguistically instantiated)9:  [+ cause] → external / internal cause  [ cause] → possession (morphologically transitive, but uncaused)  [+ event] → dynamic V  [ event] → stative V  [+ location] → terminal coincidence relation (see Hale 1986; Hale and Keyser 2002:218, ff.)  [ location] → central coincidence relation Causativity, I propose, is not an isolated, narrowly syntactic phenomenon whose explanation could be exhausted by means of feature valuation operations or purely formal mechanisms of the sort but the result of compositionality within local cycles at the Conceptual‐Intentional (C‐I) interface, once the whole construal has been transferred by cycles assuming that a causative domain, comprising an initiator, a theme, and a location, configure a fully interpretable object (a ‐domain, in Grohmann’s 2003 terms, as all and every possible theta‐position is represented in (6)) and can thus be taken by the interfaces to be assigned an interpretation and phonological matrices (see Krivochen 2012; Krivochen and Kosta 2013:94, ff. for discussion and examples). Assuming the clause structure in (6), a periphrastic causative construction would be the result of spelling out both [cause] and [event] (v and V) as different terminal nodes, via separate vocabulary items. Since most Romance, Germanic, and Slavic languages allow the possibility of V‐to‐v conflation (or incorporation, depending on the framework one assumes), the periphrastic causative is a suboptimal candidate for evaluation, as the non‐periphrastic construction conveys (in most cases) the same propositional information using less material. This happens, of course, when the cause node (syntactically represented as a v head) spells out as a light verb taking a clause 9 Notice that the primitives, and the values they adopt, are strikingly similar to those predicted by Generative Semantics during the 60s, and also used by Jackendoff within his Conceptual Semantics framework (Jackendoff 1983, 1987, 2002). The use of CS, GS, and Relational Semantics is thus justified in terms of the close relations I can find among these theories. As such, ‘features’ are just ways to represent, for instance, the two possible syntactic‐semantic readings of a locative relation, namely, central and terminal coincidence. (Anti‐)causativity and the morpho‐phonology‐semantics tension 93 (Exceptionally Case Marked) or a spatial relation (locative Vs) as its complement, for instance:  make + ECM/small clause, let + ECM/small clause, cause + ECM/small clause, take + PP, give + PP (and, by hypothesis, their equivalent forms in any language). Two key features are to be taken into account here: first, that the competition between candidates is won by the non‐periphrastic forms if and only if it exists as a convergent (optimal) form in a language L. For instance, in the following competition, the interfaces, peering into the syntactic workspace (see Boeckx 2007; Krivochen 2012; Krivochen and Kosta 2013 for details) selects the periphrastic CC as a convergent unit and takes it away for semantic interpretation, as the non‐periphrastic CC is non‐ convergent when attempting to convey the same meaning10 (examples and judgments have been checked with native speakers): (7) a. John made the baby cry. b. Jan Jan rozplakal dítě. (Czech) make‐cry.PST.3SG the‐baby’ ‘Jan made the baby cry.’ c. Juan John hizo llorar al bebé. make.PST.3SG cry (Spanish) to‐the baby ‘John made the baby cry’ d. *John cried the baby. e. *Jan Jan plakal dítě. (Czech) cry.PST.3SG the‐baby ‘*Jan cried the baby.’ f. *Juan John lloró al bebé. (Spanish) cry.PST.3SG to‐the baby ‘*John cried the baby.’ Arguably, the structure [the baby cry] is an event in itself (comprising a ‘manner of emission’ Unergative verb, [cry], and an external argument, the one who ‘emits cry’, [the baby]), and cannot be tampered with once completed and, possibly, transferred to 10 Hale and Keyser’s (2002) account of this impossibility also resorts to a mixed explanation, involving both syntax and semantics, mentioning as a core factor the nature of the involved root (e.g., 2002:3, 15). Also, Uriagereka’s (1998: Chapter 6.2) discussion of some ‘impossible words’ takes into account interface factors and opacity of domains, contra narrowly syntactic accounts. His proposal also draws on elements from Hale and Keyser (1993). 94 D.G. Krivochen C‐I interface (see Stroik and Putnam 2013:135, ff. for discussion in an alternative framework), if the interface is somehow sensitive to propositional forms (which is not incompatible with Chomsky’s early conception about ‘propositional phases’, see Chomsky 1998:20), as seems to be the case if one considers the processes of construction of propositional forms tackled from a Relevance Theoretic perspective (Sperber and Wilson 1995; Wilson and Sperber 2004). That is why all additional material is added on top (namely, the light V [make] and the external initiator [John]). The full representation of (7a) would then go along the lines of (8): (8) [John [CAUSEEXT [the baby [CAUSEINT [DO [√CRY]]]]]] Notice that I am assuming (following Hale and Keyser 2002:47; Mateu Fontanals 2002; among others) that unergative verbs, of the kind of “cry” are lexically derived through conflation of a root onto V, and the complex {V, √} onto v (which can Spell‐Out separately from the V root, as in English, or as part of the V, as in Czech and Slovak). These structures do not admit the configuration (7df) insofar as we are dealing with two events, both of which are caused (see Levin and Rappaport Hovav 1995:83, 108 for a related, two‐event proposal, but at the lexical‐argument structure levels, with no reference to syntactic‐semantic cycles): (9) a. Main Event (cycle 1): [John [CAUSEEXT [cycle 2]]] b. Cycle 2: [the baby [CAUSEINT [DO [√CRY]]]] Arguably, each sub‐derivation is derived in parallel syntactic workspaces (following the “derivational cascades” theory of Uriagereka 2002), transferred when completed, and then unified at the interfaces: semantically, a global complex causative construction is read off the construal; notice that [John] is an initiator of the event [the baby cry], whereas [the baby] is an initiator of the syntactically embedded event of crying.11 The impossibility of unifying both causative phases into a single “John cried the baby” follows from the framework outlined so far if cycles are taken to be impenetrable to external operations once completed and taken by the interfaces as they 11 I use ‘cycle’, ‘cascade’, or ‘sub‐derivation’ following the proposal of Multiple Spell Out of Uriagereka (2002, 2012) and in order to avoid the stipulations related to the static phase framework of Chomsky (1998 et seq.). Uriagereka’s model, phonologically based (command units are monotonically‐assembled Markovian objects which can thus be linearized at PF), has the advantage of not stipulating the size of transferrable units a priori, contrarily to Phase Theory. I adopt a complementary model to that of Uriagereka, in which semantically interpretable units are independent of phonologically interpretable ones, and thus, LF and PF cycles need not coincide (see Krivochen and Kosta 2013 for details and implementation). (Anti‐)causativity and the morpho‐phonology‐semantics tension 95 are no longer in the syntactic workspace (Uriagereka 2002, 2012; Krivochen, forthcoming). Phonologically, attention is paid to the impossibility of tampering with cycles once transferred, thus making conflation (a process that is both syntactic and phonological, as there is p(honological)‐signature copying, but the restrictions over strict complementation for conflation are established in narrowly syntactic configurations; see Hale and Keyser 2002:59) of {CAUSE, {DO, √CRY}} onto {CAUSE} in the main event impossible, once again, because they are no longer in the active workspace. It is to be noticed, although I will come back to this below, that there is a tension between semantics and morpho‐phonology that generates these interpretations as interface results: semantics operates globally (as argued in Krivochen 2013) at the level of determining the propositional information to be conveyed, plus implicatures derived from specific Topic/Focus projections, whereas phonology, as has been argued since Bresnan (1971) operates in a cyclic, local fashion, recently revived within the phase framework: from syllable stress assignment rules to word‐level or phrase‐level materialization. ungrammaticality Taking of into the consideration Czech examples the representation follows in straightforwardly (8), the if one acknowledges the fact that there is no structural position where to Merge the applicative clitic [mu] (in terms of Kosta 2011:253), and it is therefore a superfluous element in the representation at LF which is banned by a strict interpretation of the Full Interpretation Principle (Chomsky 1995:27). The second issue to take into account, in a related note, is the composition of the phonological lexicon, that is, how verbs “get their names” (Harley 2003). In the present framework, verbal Spell‐Out (i.e., the morpho‐phonologically visible “verb”) is the result of either: (a) Conflated p‐signatures: the phonological form of a {D} structure is copied onto the closest higher empty/phonologically defective node (e.g., a clitic or affix). There is no movement, but, if the reader prefers, percolation of phonological features, following Hale and Keyser’s (2002:59) requirement of strict complementation for conflation. This occurs with “heavy” roots, that is, roots that instantiate substantial conceptual information. In a word, only these verbs involve roots, and they constitute the vast majority of verbs in Spanish and English (unergatives, location, locatum, ergatives, and even some unaccusatives), or (b) Default Spell‐Out: this occurs only with very primitive and semantically light verbs. A p‐signature is inserted in an empty [event] node when no conflation process has occurred to prevent PF crash. The piece inserted depends entirely on the syntactic configuration in which {V/event Ø} appears, giving us a strongly compositional system. A list of such verbs include, to our understanding, GO, BE, PUT, TAKE, 96 D.G. Krivochen ARRIVE (see above), LEAVE, DO (and their equivalents in other languages12). These verbs can be exhaustively decomposed in features of v, V, and P (in the sense of Mateu Fontanals 2002, 2005) with no need to resort to a root either to provide content or phonological signature: in GB traditional terms, they would not impose s‐selectional restrictions to co‐occurring arguments (as the pair [John came]/[Winter came] would prove). My claim is that there are no roots involved here, but purely contextual interpretative issues, in the sense that the verbs in the list above are materializations of the combinations of semantic primitives proposed by Mateu Fontanals in different configurations, for instance: GO = dynamic V, positive locative node (terminal coincidence) BE = stative V, negative locative node (central coincidence) It is to be noticed that some of these verbs (e.g., BE, HAVE, DO) also work as auxiliaries marking Time and/or Aspect. Coming back to the “lazy Spell‐Out” claim (“prefer synthetic candidates ceteris paribus”), consider for instance the following paradigm (taken from Kosta 2011:246): (10) a. Petr Peter zmĕnil strategii. change.PST.3SG the‐strategy (Czech) ‘Peter changed the strategy.’ b. Strategie the‐strategy *(se) CL[ERG] zmĕnila. (Czech) change.PST.3SG ‘The strategy changed.’ The presence of an overt external causator of the event change (strategy), namely [Petr], in (10a) is explained by the fact that it increases the informational load in a relevant way: the identity of the external initiator is revealed, and, moreover, it is thematic (i.e., letmost within the TP). (10b), on the other hand, presents an anticausative (ergative, as the event is a change of state) alternation on the light of the absence of a materialized agent. This does not mean that, in the phenomenological world, the event is actually uncaused, but that the speaker has chosen to present it that way, possibly, because the identity of the initiator is not pertinent (a pragmatic reason). Truth conditions vary from (10a) to (10b), as a person could hold (10b) true, but deem 12 Ancient Greek verbs GO and BE, for example, differed only in word stress: εἰμὶ ([be], with both individual and stage level predicates as complements) and εἶμ ([go]), which is evidence in favor of our position that they are really Spell‐Out of the same eventive node (V/T, depending on the terminology one takes) in different syntactic configurations (e.g., central/terminal coincidence, distribution of cognitive Figure and Ground), provided that stress assignment takes place, as DM claims, post‐syntactically. (Anti‐)causativity and the morpho‐phonology‐semantics tension 97 (10a) false if, for instance, the initiator was not Peter. Semantically, the root requires causation in this particular construal (V + NP [strategy]), as a strategy does not contain an internal, inherent capability for change; change must always come from the outside: crucially, this does not mean that there is no semantic cause, just that there is no overt external initiator. A comparison of both structures might help clarifying the scenario I propose (following Mateu Fontanals 2002; also Jackendoff 2002): (11) a. b. VP [GO] vP PP The strategy [TO] Peter P’ √CHANGE v’ [CAUSE] VP [GO] PP The strategy [TO] P’ √CHANGE Both construals depict a dynamic (event primitive [GO]) change of state (positive location primitive [TO]), but (11b) includes an external sortal initiator, licensed in turn by the causative v head on top of the eventive node. In both cases, the root [√CHANGE] conflates to v, following the Head Movement Constraint, through all intermediate heads, which contribute to the final semantic interpretation of the root in its final syntactic destination. Thematic roles, within the construal, are read off at LF from the configuration, following the strongly configurational theta theory of Hale and Keyser (2002), Chomsky (2004) – where it is subjected to the duality of semantics, a proof that it is not narrowly syntactic even for some orthodox generativists, and Krivochen (2012): as I have argued elsewhere, thematic roles have no entity at all in a purely generative syntactic component which consists only on a generative algorithm, but are read off from a configuration in which DPs establish local relations with the procedural nodes cause and P.13 Considerations of theoretical economy, and logical consistency within 13 Arguments in favor of the hypothesis of considering theta‐roles outside the scope of narrow syntax as conceived in MGG (also backed up by extensive bibliography) are both architectural and empirical: on the one hand, if the syntactic component is purely generative, then it cannot read theta‐roles, or else it should have to be both generative and interpretative, thus rendering the sound and meaning interfaces partially redundant. If such a redundancy arises, sound and meaning being necessary, it must be purged from the syntax. On the other hand, as well demonstrated in GB, an element that fails to comply with the (purely 98 D.G. Krivochen the theory prevent me from accepting the thesis that thematic relations have any relevance for the generative engine, unless enriched with interpretative protocols (which belong to the interfaces, by definition). Therefore, the present proposal does not fail to overcome the criticism made by Jackendoff (1987), among others, to the orthodox Theta‐Criterion exposed in Chomsky (1981), in two relevant senses:  An argument can have more than one theta‐role (e.g., in [John sent a letter to Mary], [John] is both Agent and Source).  An argument’s thematic role is not directly mapped from an a priori lexical entry, fixed in the Lexicon, but dynamically read off a syntactic‐semantic construal, as in (8). Therefore, Uniformity Theta Assignment Hypotesis (UTAH)‐based approaches to thematic roles (and also, Montague Grammar and Categorial Grammar along the same lines; see Jackendoff 1997:33) are rejected within this theory because of their static and aprioristic character. I explicitly argue in favor of a componential theory of semantic relations between arguments, to be established at the semantic interface (not at the lexicon, either), which should be the object of thorough research even within syntactico‐centric theories (if one takes Relevance Theory to be a valid context‐sensitive, internalist pragmatic‐semantic theory, it partly has been): theta roles are taken here to be the result of local relations between sortal entities and v or P (for the latter, as either Figure or Ground) at the syntax‐semantics interface, where these relations are read off as theta‐roles. In this vein, I argue that, if thematic roles are not relevant at the syntactic component, but at the semantic component, then there is no difference between Agent, Cause, and so on, as they are all inferential specifications of a more general “thematic sphere”, Agent/Initiator (Krivochen 2012: Chapter 2; Krivochen and Kosta 2013:93),14 a reasoning that is not alien to proposals like Holinsky semantic, by the way) requirements to bear a theta role in a construal does not by itself render it ungrammatical but deviant, whereas, for instance, an element that does not receive Case does: (i) ??The guitar was writing a paper. (ii) *John worked a car. Another argument comes from the very same theories that assume theta‐roles as having entity in the narrow syntax, as features. If movement can be triggered by the need to check a theta‐feature present in a functional head: (a) how come the semantic interface can distinguish between theta roles (say, distinguish Agent from Theme)? Are there diacritics in theta‐features? and (b) exactly at which level is the assignment of theta‐features to functional heads determined? Is it not stipulative? What prevents assigning a certain head theta‐features just to motivate movement (which is, in my view, flagrantly stipulative)? 14 The concept is similar, but not identical to, Van Valin and Wilkins’ (1996) effector (‘the dynamic part doing something in an event’). On the one hand, I consider volition to be inferential, thus, post‐ propositional semantically relevant. On the other, the present proposal does not require of an NP to be (Anti‐)causativity and the morpho‐phonology‐semantics tension 99 (1987). This being so, the ungrammaticality of a certain example cannot be blamed on theta‐roles (an independency statement that can be traced back to Chomsky’s first writings), as they are post‐syntactic: a free, blind‐Merge system (see, e.g., Chomsky 2004; Boeckx 2010; Krivochen 2011 et seq.) is not sensitive to thematic relations, which are not relevant until LF‐Transfer, where configurations are read (see also Stroik and Putnam 2013: Chapter 6 for discussion about the locus of relevance of semantic relations). These interface‐oriented considerations are essential if one is to distinguish between internal and external causation, as it has to be done at two levels: syntax and semantics, independently of their morpho‐phonological manifestation. 3. Dissociating Internal/External Cause Now, after the basic assumptions have been outlined, I can expand on whether the difference between internal and external cause is narrowly semantic, narrowly syntactic, or a combination of both, and how it is materialized. Moreover, is causativity a property of a root (as Schäfer 2008 suggests), or, more componentially, of the construal the root appears in? Distributed Morphology (Halle and Marantz 1993; Embick and Noyer 2005; see also Panagiotidis 2013 for recent references) claims, as I pointed out in section 1, that “all roots are meaningless in isolation” (Panagiotidis 2013:5; Borer 2009), insofar as they can only have meaning within a specific construal, in local relations with affixes, categorizers, etc. I disagree with this claim, as my proposal does not divorce the roots from the Encyclopedia (the reservoir of content): a root has meaning, albeit a highly underspecified one, which allows for a limited range of variation, as in genotype/phenotype dynamics (which derives from a theory with LF‐underspecified roots; see Panagiotidis 2011). What is more, this claim does not invalid any of the empirical evidence Panagiotidis offers in favor of meaningless roots, it only relativizes the theoretical claim. Meaningless roots, if taken in a strict sense, could lead to a situation of overgeneration, as there is no reason why a particular meaningless element should be linked to an encyclopedic entry, or a Spell‐Out pattern: if a root is meaningless, then one would have to resort to PF and LF diacritics for the root to be interpreted at all (or B‐list and C‐list diacritics, using DM terminology) and avoid roots randomly assigned PF and LF properties, which would result in interface crash for definite and animate to be a better candidate for the effector role (Van Valin and Wilkins 1996:314): an eventive entity denoted by a CP, or a nominalization (categorially an NP, but denoting an event) can also belong to the thematic sphere Agent/Initiator. 100 D.G. Krivochen most cases. The content of a root, underspecified though it might be, weakly constrains the possible interpretations of that root, the contexts it can appear in, and the PF/LF representations that are associated to those contexts, all of which I consider advantages over contentless roots. Moreover, if they are really “unexceptional syntactic objects” (Panagiotidis 2013:7), then they cannot be radically empty, as every element is to have an interface motivation, or else it should be eliminated because of Full Interpretation. Functional material on top can specify the reference of the root (either to an entity or to an event), as a subset of the potential extension of the root, but functional nodes (procedural nodes, in Relevance Theory) do not provide semantic substance. Part of this precision or specification is the determination of the source of causation, when the causative alternation is possible. A crucial objection I will make to the radical ‘empty roots’ theory is that, if roots are in fact meaningless, how come they take arguments? Consider the following representation, taken from Embick and Noyer (2005:24): (12) vP AdvP v’ quickly v v √PLAYi T[past] ti v √P DP the trumpet Notice that the argument [the trumpet] originates within the √P, which implies that there are some features in the root that percolate to a label, and, more importantly, that the root has selectional features, which allow the merger of an internal argument. The DP is directly governed by the root, not by the v head. What is more, there is no lexical/eventive projection (i.e., VP) here, just a root Phrase and the vP, in charge of categorization via incorporation of the root to v. Assuming a bottom‐up approach to syntactic derivations, the DP must have entered the syntactic workspace before the root has been categorized by v: how come the root takes a complement if it is semantically empty (see also Harley 2013 for a similar representation assuming a √P)? And how to prevent for the root to take any XP as complement, if there is no content to constrain combination possibilities? A similar analysis to that of Embick and Noyer (2005) can be seen in Kosta’s (2011) bracketing structures, following Alexiadou et al. (2006), where (13a) is assumed to have the structure (13b): (Anti‐)causativity and the morpho‐phonology‐semantics tension (13) a. Petr otevřel Peter open.PST.3SG dveře. 101 (Czech) the‐door ‘Peter opened the door.’ b. [Petr VOICE [CAUSE [dveře √OTEVŘEL]]] Here, the DP [dveře] appears within the domain of the root, not of an eventive variable, as there is no VP (or [EVENT] head) in that representation. However, there is a Voice node which, apparently, “expresses the relation between the element in the specifier and the event in the complement position (CAUS)” (Kosta 2011:285), following and extending the proposal of Kratzer (1996). My take on the matter is quite different. In my opinion, the presence of a Voice head as distinct from v is not justified, as it is the semantically interpretable causative head that requires an initiator, and the roles of the two functional nodes seem to overlap: insofar as v conveys the semantically interpretable dimension causativity, it is v that licenses subject oriented adverbs and [by‐phrases] in passives, licensing (but not always requiring the overt realization of) an external argument which receives a thematic interpretation initator, agent, (natural) force, etc. being inferential (therefore, post‐syntactic) specifications. As I have claimed above, in the present proposal, v has a clear semantic value, which is causativity. It is essential to bear in mind that the initiator argument is not always spelled‐out, whereas this does not take causativity away from the semantic representation of the construal (as I saw above). How to solve this apparent problem? A semantically based solution, focused on economy of representation and derivation would be to abandon the X‐bar intra‐theoretical stipulation that all X project a specifier (see, e.g., Chomsky 1995). If the semantic primitive CAUSE syntactically represented by v does not necessarily project a specifier position because there is nothing in the lexical array to merge in Spec‐v, then one gets something like (14): (14) Dveře the‐door se otevřely. CL[ERG/PASS] open.PST.3SG (Czech) ‘The door opened.’ In this case, I agree with Kosta’s representation (including cause, but without an external argument), which goes along the lines of (15): (15) [CAUSE [the door [√OPEN]] but I would add an eventive variable (bound by the features of T, which anchor the reference of the event, and incidentally generates a V category reading for the root 102 D.G. Krivochen at the semantic interface), which in this case would be a positive value for the V node (i.e., a dynamic event), and a cause primitive that has not been morphologically materialized and can therefore be taken not to be relevant (in the technical sense of Sperber and Wilson 1995) for the construction of a full propositional form and posterior interpretation, thus geting (16): (16) [CAUSEENV [the door [[GO] √OPEN]]] If there is no element in the array to be merged as a specifier to CAUSE, then the position is not projected, in real derivational time, and there is no need to resort to a different Voice projection to account for internal/external causation alternations, thus simplifying the array of functional material. Harley (2013) proposes to keep VoiceP as divorced from both VP and vP, which introduces an initiator (vP being ‘specifier‐less’) based on morphological evidence from Hiaki interpreted through the Mirror Principle (Baker 1985:375, ‘Morphological derivations must directly reflect syntactic derivations (and vice‐versa)’), and assuming each morpheme corresponds to a projection in the syntax and materialization is unambiguous. Harley assumes Pylkkänen’s (2002, 2008) analysis of Appl(icative)P, which is in itself a problematic projection. Pylkkänen (2008) employs the term applicative as a way to denote indirect object arguments added to a V. It is claimed that semantic properties of applicative (i.e., ditransitive) constructions are more or less uniform among languages, but their syntactic properties differ (Pylkkänen 2008:11). Thus, for instance, whereas English does not allow applicative arguments (i.e., benefactive/goal datives) with unergative Vs, Bantu languages do. Moreover, the relative stability of applicative semantics is due to the scope they have, thus defining two kinds of applicatives: high applicatives, which have scope over the VP; and low applicatives, within the VP. The representations are as in (17): (17) Low Applicative High Applicative VoiceP VoiceP DPsubject DPsubject Voice VP V Voice ApplP ApplP DP DP Appl Appl DP VP V DP (Anti‐)causativity and the morpho‐phonology‐semantics tension 103 However, it is to be noticed that not all authors accept this distinction. For instance, Marvin (2011) claims that Slovenian displays ambiguous readings with some ditransitive transmission constructions: (18) Binetu Bine.DAT sem poslal pismo. AUX send.PST.1SG the‐letter.ACC (Slovenian) ‘I sent Bine the letter’/‘I sent the letter to Bine.’ The readings, according to Marvin (2011:100–101), are the following: (a) High Applicative: I sent X the letter instead of Bine (Bine was the recipient of my event of sending the letter). (b) Low Applicative: I sent Bine the letter (Bine is the intended recipient of the letter). While one would expect a logical form analysis of the sentence, so that scope relations are made explicit, Marvin (2011) provides none. However, she does make a point about the ambiguous scope of ApplP, as she finds the same effect with non‐ transmission Vs (like skuhal ‘cook’) and possessor rising constructions, in which the dative (DAT) argument introduced by the Appl can either denote the possessor of the accusative (ACC) object or not. What is more, she extends her claim to other Slavic languages, like Serbo‐Croatian and Macedonian. I will not discuss the empirical appropriateness of Marvin’s (2011) analysis, but rather try to simplify the theoretical apparatus. In Hale and Keyser’s (1993 et seq.) framework, the P node can convey either a central or a terminal coincidence relation, which determines the difference between location and locatum Vs: (19) a. John shelved the book = [John [CAUSEEXT [GO [[the book] [TO [√SHELF]]]]]]  Location V b. John buttered the toast = [John [CAUSEEXT [GO [[the toast] [WITH [√BUTTER]]]]]]  Locatum V Notice that (19b) denotes a possession relation, in which John causes the event of the toast having butter. This possession relation is interpreted because of the nature of the locative node P, a central coincidence relation: possession is a kind of location, in which the possessed thing is merged as the complement of P and the possessor, as the specifier. As Bleam (2003) has pointed out, contrarily to the judgments of Pylkkänen (2008:15), it is ungrammatical to cancel the presupposition with an adversative clause if the main sentence is not negated, as in (20): 104 D.G. Krivochen (20) *Peter sent Lina the book, but she never got it. This is arguably the most salient semantic difference between prepositional indirect object constructions (PIOC), in which the DAT argument is introduced via a preposition, and double object constructions (DOC), in which the DAT‐ACC dynamics is only read off the configuration, there being an adjacency relation between internal arguments, materialized as DAT‐ACC. Taking this into account, (21) is perfectly acceptable, since PIOCs do not generate the presupposition that at the end of the process the locations of Figure and Ground coincided (i.e., there is no possession presupposition): (21) Peter sent a book to Lina, but she never got it. The structures proposed by Bleam (2003; as well as Harley 2002) involves a caused event (in Harley’s terms, a vCAUSE) having scope over a central coincidence relation (in Harley’s terms, a PHAVE). If P takes two arguments, as Hale and Keyser (2002) propose, then the semantic contribution of P is precisely to provide instructions to the semantic system as to how to interpret the relation between those sortal arguments. The question arises whether one needs an Appl node to account for the introduction of an oblique argument in the construal (as Pylkkänen claims) or not (as in Kosta’s 2011, 2014 account). Without even trying to exhaust the topic, my very provisional answer, guided by Minimalist desiderata of economy in representations, will be that, if transitive constructions always involve a P node (in monotransitives, the complement of P conflates onto P0), then there is no need to resort to an extra Appl projection to account for DAT arguments. Moreover, the observations with respect to the ungrammaticality of applicative arguments in Unergative construals (Pylkkänen 2008:11) would be derived from the syntactic‐semantic construal underlying unergativity, which in Hale and Keyser’s (1993) terms, involve root conflation to a defective V head, roughly equivalent to DO or EMIT. With this in mind, if [t]he key argument for the separation of Voice and v is a minor variation on the argument presented by Pylkkänen (2002: 122–125). The key point is that the behavior of the applicative shows that the causative v° head does not introduce the overt external argument in Hiaki causatives. (Harley 2013:3536) Then an argument against Appl as an independent head undermines Harley’s argument indirectly, and forces us to reanalyze the empirical data. Arguments against the Mirror Principle (the other main support of Harley’s discussion of Hiaki data) have risen from outside MGG, from theories that either do not subscribe to Structural Uniformity (e.g., Culicover and Jackendoff’s 2005 Simpler Syntax) or have (Anti‐)causativity and the morpho‐phonology‐semantics tension 105 problematized morpho‐phonological linearization such that each phrase marker can receive two logically possible and equally theoretically valid materializations (Uriagereka’s 2012:56 Mirror Linear Correspondence Axiom). The problem clearly exceeds the limits of the present work, but I think the problems proposed here suffice to consider the validity of alternatives to Harley’s tripartite split‐VP proposal on both conceptual and empirical grounds. 4. A Frustrating Approach to Causativity and Alternations In a semantically based syntax, global tendencies favor information conservation (see Lasnik, Uriagereka and Boeckx 2005; Krivochen 2011, 2012) throughout the derivation, whereas local tendencies favor cyclic transfer of local derivational chunks to the phonological component (as first noticed by Bresnan 1971), a tension I have (using Binder’s 2008 terminology and following Uriagereka 2012) described as a dynamical frustration in cognitive design, particularly pervasive in language under a Multiple Spell‐Out model. This frustration, defined as the tension between opposing forces, or global and local tendencies, is very much present in language (see Tesniére 1959: Chapters 6–7), and interlinguistic variation at certain respects can be looked at from a “frustrated” perspective. For instance, Schäfer (2008) and Alexiadou et al. (2006) justify the need for independent VOICE projections on the light of contrasts like the following (taken from Kosta 2011:284): (22) a. John/the explosion/Will’s banging broke the window. b. Okno bylo rozbito Johnem /explozí/ ránou Willa. (Czech) the‐window AUX break.PST.PART by‐John/the‐explosion /banging Will.GEN ‘The window was broken by John/the explosion/Will’s banging.’ (23) a. *The window broke by John/by the explosion/by Will’s banging. b. *Okno se the‐window CL[REFL/ERG] ránou rozbilo Johnem / explozí / break.PST.3SG by‐John / the‐explosion / Willa. (Czech) banging Will.GEN ‘The window broke (itself) by John/the explosion/Will’s banging.’ However, Spanish (a Romance language) allows constructions of the type of (23), as shown in (24): 106 D.G. Krivochen (24) La ventana the window se rompió por la explosion. CL[ERG] break.PST.3SG by the explosión (Spanish) ‘The window broke because of the explosion.’ Kosta notices this fact, and makes the caveat that an agent cannot be introduced by means of a PP: (25) *The window broke from Mary. However, the mistake here, inducing ungrammaticality, seems to be the preposition choice, as (26) is grammatical: (26) The window broke because of John. How is (26) to be interpreted? Possibly, it is to be interpreted along the lines of (22a), being paraphraseable as “the window broke because of what John did”. In that case, we would not be dealing with an entity as the cause of the relevant event of [breaking] but with an event itself as the cause (whatever John did), which is correctly predicted to be grammatical by Kosta.15 In spite of this, the motivation for Voice is far from clear, since nothing in the nature or semantic contribution of the v head determines that a CP (a full proposition) or a nominalization with a full thematic structure (agent, event, theme) cannot occupy that position. In more concrete terms, once I have distinguished several semantic types of cause, and clarified which arguments are licensed by which heads, the necessity of an independent Voice head fades out, until further evidence forces the revision of these claims. 15 Peter Kosta (pers. comm.) objects to this analysis that I propose [because of] as introducing a [cause] argument, which is admittedly clearly false, but that is not what I actually say. He points out that [because of] introduces a partial ellipsis, along the lines of [the boat sank because of the storm [that has caused that the boat sank]]. However, the objection does not apply: what I am saying is precisely that I am not dealing with a sortal cause here, but with an eventive cause, an observation captured in the paraphrasis provided for (21). To spell it out, [because of] introduces a semantically eventive argument, materialized as a DP. Thus, being an event, it does not behave as a [by‐phrase] agent, as Kosta correctly points out. Moreover, there is no evidence or compelling conceptual argument to claim that eventive and sortal causatives require different projections, thus licensing the presence of VoiceP. In a free‐Merge syntax, there is simply no reason to assume that the generative engine deriving representations is sensitive to such distinctions. If the syntactic component is claimed to be indeed sensitive to the eventive/sortal opposition, then one must concede it is not only generative but also partly interpretative, which would in turn make the interfaces (at least LF) partly superfluous. Arguments in favor of a more economic approach including only vP are thus both conceptual and empirical, deep‐rooted in a cognitive stance. (Anti‐)causativity and the morpho‐phonology‐semantics tension 107 The caveat I have made above provides some bases for a classification of causatives according to the kind of PP causator they allow: [by‐phrases] are licensed by sortal causatives, in which the causator is a sortal entity, denoted by a DP. On the contrary, [because of]/[due to] phrases are licensed by eventive causatives, in which the cause of the event denoted by V is an event itself, regardless its grammatical realization (i.e., it can be grammaticalized as a PP followed by a CP, as in [because of what X did], or as a nominal). In the case of (26), for instance, the eventive causative has a conceptual subject [what John did], grammatically realized as a PP [because of John]. Interestingly, eventive causatives do not have the same semantic restrictions as sortal causatives, that is, the requirement of animacy and volition for sortal DP agents. In Spanish, some apparent ergative constructions can be thought to be in fact eventive causatives, for instance: (27) La puerta se abrió. (Spanish) the door CL[ERG/PASS] open.PST.3SG ‘The door opened/was opened.’ (28) La puerta se abrió por el viento. (Spanish) the door CL[*ERG/PASS] open.PST.3SG because the wind ‘The door was opened by the wind.’ Notice that the eventive causator, roughly [because of what the wind was doing], is introduced by a different preposition than that used by Instrumental arguments (roughly paraphraseable as ‘by means of’): (29) La puerta se the door CL[PASS] abrió con la llave. (Spanish) open.PST.3SG with the key ‘The door was opened with the key.’ This paradigm, admittedly reduced, already introduces interesting conditions for the semantic classification of causatives, and, in my opinion, motivates a revisiting of transparent syntax‐semantics‐morpho‐phonology interface proposals, based on uniform mapping between the aforementioned systems. The concept of tension is thus essential in an interface proposal that seeks to achieve explanatory adequacy while accounting for both sound‐morphology and meaning requirements, from my point of view. 108 D.G. Krivochen 5. Some Problems with Root Classification I would now like to address an apparent counterexample proposed to me by Kosta (pers. comm.; see also Kosta 2014) regarding the classification of verbs and the possibilities of alternations: his evidence consists on distinguishing roots characterizing them via [± cause], [± voice] features, which project from the Lexicon (a strong lexicalist stance, insofar as syntactic structure is not a function of semantic requirements, but of intra‐lexical specifications). This double classification results in four kinds of verbal roots:  √agentive (murder, assassinate, cut) projects only a VoiceP: [+voice], [‐cause].  √internally caused (blossom, wilt, grow) projects only a CausP: [‐voice], [+cause].  √externally caused (destroy, kill, slay) projects a VoiceP and a CausP: [+voice], [+cause].  √cause unspecified (break, open, melt) projects a light vP: [‐voice], [‐cause]. While assuming a uniform system of projection and v as a mere categorizer head (e.g., Marantz 1997; see also Panagiotidis 2011), Kosta’s objection would be indeed correct, and I would need an extra head to account for dissociation; if one treats Specifiers as Adjuncts (see, for instance, Uriagereka 2002; Chomsky 1995), that is, elements that are absent of the core projection system, then the extra Voice head is not necessary, particularly with the added assumption (made above) that v can either project a Specifier position or not, depending on the pre‐linguistic semantic construal to be linguistically instantiated: in our theory, syntactic structure depends on semantic requirements, not on X‐bar theoretical axioms. It is also important to notice that the conception Kosta has of v is, apparently, completely different from my own: in this theory, ergatives have no v layer, insofar as the semantic contribution I assign to the v procedural element is, precisely, causativity; diathesis having no independent projection (i.e., no VoiceP) as it is not narrowly syntactic in nature but more related to theme‐rheme dynamics (as in passivization, taken to be the thematization of the grammatical object), as noticed, among many others, by Frías Conde (2010; pers. comm.); if this is the case, then diathesis is at best an interface phenomenon, but does not belong to the narrow syntax (i.e., the generative algorithm alone). A further point to be considered is that I invert the logics: transitivity and passivization are not sine qua non conditions for causativity, as I have dissociated all three, but this does not mean I have proposed three independent syntactic projections. It is to be noticed that some cases are not as clear‐cut as they seem, something predicted in a blind, free‐ Merge system: the generative system is not constrained, and alternations are always logically possible; constraints on alternations are given by use and historical accidents, and can be reversed (as opposed to a rigid conception of the lexicon in which (Anti‐)causativity and the morpho‐phonology‐semantics tension 109 alternations are a priori limited by the featural composition of the root involved, see also the discussion in Krivochen 2012: Chapter 2). Consider, for example, the case of “read”: (30) a. John read the book. (transitive causative) b. The book was read by John. (personal passive, licensed by v) c. “The text is a faithful rendering of the original in style and register, and it reads easily and smoothly.” (http://www.linguee.com/english‐spanish/translation/reads+easily.html) (middle/impersonal construction?)16 It has been noticed that (30c) presents some properties of anti‐causative constructions, in spite of its apparent middle meaning, and while (31a) could be a suitable paraphrasis of (30), (31b) could not: (31) a. Anyone reads the text easily and smoothly. b. *The text is read easily and smoothly by anyone. The unacceptability of the [by‐phrase] leads to the idea that the syntax of this construction is closer to that of anti‐causatives, in the event that middles are not considered a kind of anti‐causative. In any case, the point is clear: alternations are available in the system, a subset of which constitutes the core data for linguists (roughly, Coseriu’s norm). If generative linguistics analyzes the system that generates structural descriptions of sentences, then a full account of the system is to be given: constraints on alternations must be revisited paying attention to the use. Historically, language development might drive apart from uniform interface mapping, which is problematic for a systematic approach to the causative/anti‐causative distinction unless it is dynamically interpreted.17 The non‐uniform mapping between (syntax and) semantics and phonology (including the claim that PF‐cycles – possibly, tone units ‐ and FL‐cycles – propositional objects ‐ might not coincide) is exactly the kind of problem addressed in Krivochen (2013) from the perspective of dynamical frustrations (that is, mutually opposing forces at global and local levels; see Binder 2008), and which can be useful 16 The exact expression “reads easily” got more than 280.000 hits in Google, which is not a minor number given the specificity of the search. 17 It is possible that the contrast can be semantically analyzed along the lines of a non‐valid LF movement of the universal quantifier, but this possibility will not be analyzed within the context of this paper. 110 D.G. Krivochen here.18 Notice that, be it a CP or a DP, the restriction on the materialization of the causator is not categorial (as such a restriction would imply re‐introducing subcategorization frames from the GB model), but semantic: in structures like (21), the causator, if materialized, corresponds to an event, roughly, [what John did], and not to a sortal entity (of the kind denoted by proper names). When cyclically transferred to PF, the structure collapses if an anti‐causative construal, with no Spec,vP position, contains a proper name introduced by a PP, but not if the PP introduces an event, even if categorized as an NP (e.g., a nominalization). Since the restriction is by nature (broadly) semantic, I find no justification to the addition of a node in the syntactic structure (as Voice), as the effects can be accounted for via interface conditions and the tension between global semantic requirements and local phonological cycles. This perspective can also prove useful when considering the availability of causative/anti‐causative alternations for certain roots, a restriction Schäfer (2008), among others, attributes to the semantic (encyclopedic) content of the root. In my opinion, such a proposal has a crucial disadvantage; it needs a syntactic component that, apart from building structure, is sensitive to the inner characteristics of the elements it manipulates, in order to identify the roots that allow alternations and those which do not, and construe a phrase marker in consequence. This entails directly that the so‐called “Narrow Syntax” (Hauser, Chomsky and Fitch 2002) cannot be just Merge (or Merge + Move/Agree…), but Merge + some interface reading/evaluating function applying locally, still unspecified in the mainstream literature (but see Putnam 2010; Stroik and Putnam 2013 for some proposals). I will take two sets of examples, in order to support our argument that things are not that simple, as “root content” is easy to write, but hard to define; in a strongly componential, semantically‐motivated syntax, construal is always to be taken into account, as in (32)(33): (32) a. The flowers blossomed. b. Kvĕtiny the‐flowers rozkvetly (Czech) blossom.PST.3PL ‘The flowers blossomed.’ c. Kwiatek zakwitł. (Polish) the‐flower blossom.PST.3PL ‘The flower blossomed.’ 18 The existence of a tension between syntax‐semantics and morphology has been acknowledged since at least Tesniére (1959:1922). In Krivochen (2013), I attempted to tackle the problem of its implementation in the mind‐brain, following Uriagereka (2012). (Anti‐)causativity and the morpho‐phonology‐semantics tension d. Las flores the flowers florecieron. 111 (Spanish) blossom.PST.3PL ‘The flower blossomed.’ (33) a. *The gardener blossomed the flowers. b. *Zahradník rozkvetl the‐gardener blossom.PST.3PL kvĕtiny (Czech) the‐flowers ‘*The gardener blossomed the flowers.’ c. *Ogrodnik the‐gardener zakwitł kwiat blossom.PST.3PL the‐flower (Polish) ‘*The gardener blossomed the flower.’ d. *El jardinero floreció las flores. (Spanish) the gardener blossom.PST.3PL the flowers ‘*The gardener blossomed the flowers.’ The paradigm above (including Romance, Slavic, and Germanic languages) shows that an overt CAUSEEXT initiator is incompatible with ergative construals, but it says nothing about whether it is the construal itself that does not allow the presence of an overt external initiator or it is the content of the root, specified enough to make c‐/s‐ selection choices, that bans the external argument. Schäfer (2008), like Embick and Noyer (2005) (see also Alexiadou 2003:19; Harley 2013), assumes that the root not only projects, but also establishes constraints as to its “arguments”. The perspective presented here is different: ergative/unaccusative, unergative, and (di)transitive are primarily conceptual templates (Relational Semantic Structures, in Mateu Fontanals’ 2002 terms; see Jackendoff 1987, 2002 for a Conceptual Semantics proposal) a subject can organize phenomenological information with, and might or might not be instantiated linguistically. When they are, the meaning is to be read off the whole structure, not as a mere projection of the root, which is too semantically underspecified to be interpretable on its own and thus cannot take arguments or impose restrictions to co‐occurring XPs. As Hale and Keyser (1997:40) eloquently put it: I maintain that certain crucial aspects of meaning are dependent on the very structural features whose identification is at issue. If I ‘knew the meaning’ I would know the structure, perforce, because I know the meaning from the structure. [emphasis in the original] (Hale and Keyser 1997:40) What I would like to problematize here, without attempting to reach a decisive answer, is the fact that, in principle, all roots should be equally “causitivisable”, from a strictly syntactic point of view (i.e., just the combinatory engine), as already noticed by 112 D.G. Krivochen Levin and Rappaport Hovav (2003, 2008). From a strictly semantic point of view (unstructured meaning, as structured meaning would involve syntax), the possibility of causative alternations is banned from square one, due to the selectional properties of the root (however they turn out to be encoded). From an interface point of view, like the one I defend here following Hale and Keyser (1993 et seq.) and Mateu Fontanals (2000, 2002) among others, there are at least two factors to be taken into account: (a) The construal in which the root appears, and the structural positions licensed by such construal (Mateu Fontanals 2002:24), and (b) Extra‐linguistic processes of coinage and fossilization of certain structures. The immediate consequence of this proposal is that linguistic diachrony cannot be overlooked when analyzing argumental alternations in a particular language, otherwise, any account would require ad hoc stipulations to determine the availability of alternations for certain Vs and not others within a single typology (e.g., unlike [blossom], some ergatives allow pure transitive alternations, like [grow]: John growsEXT tomatoes/Tomatoes growINT; however, both [grow] and [blossom] are customarily classified as ergatives). Another consequence is that syntactic structure is driven by global semantic requirements, therefore, optimality considerations about the positioning of arguments and competition among candidates must take into account not only Spell‐Out possibilities but also which candidate turns out to be less entropic with respect to the global semantic content to be linguistically instantiated: in this way, I aim to integrate conceptual structures (Krivochen, forthcoming) with conceptual semantics (Jackendoff 2002; Mateu Fontanals 2002 et seq.) and lexical semantics (Kosta 2011, 2014; Hale and Keyser 1993, 1997, 2002), all of which take into account structured meaning (i.e., syntax + semantics), and thus provide valuable insight for interface accounts. Coming back to the ergative paradigm above, it is to be noticed that the causative meaning is not banned in the construal (which could be interpreted as an argument in favor of compositional approaches rather than root‐oriented constraints on construal), since (34) is well formed: (34) The gardener made the flower blossom. (for instance, by using a special fertilizer on it) What is banned is the particular materialization of causative meaning in a synthetic manner, that is, Shibatani and Pardeshi’s (2001) (a) group comprising lexical causatives. The tension mentioned before is also visible here: a global meaning is available (in principle, if thought systems are independent of language, and (Anti‐)causativity and the morpho‐phonology‐semantics tension 113 computational themselves, why would it not be?), but not all materialization possibilities are equally evaluated, as already mentioned, synthetic materializations having the upper hand ceteris paribus.19 In this case, Transfer‐PF by local cycles, as in (9), sets the local limits for the application of PF‐related operations, including conflation. As one can see, broader interface and architectural issues are to be taken into account when considering constraints over alternations, as well as extra‐linguistic issues related to coinage.20 6. Conclusion In this paper, I proposed an interface approach to the phenomenon of causativity, based on the lexical decomposition approach taken by Kosta (2011, 2014) among others, complementing it with notions from Conceptual Semantics and Relational Semantics, configuring a consistent system. In doing so, I highlighted that no linguistic phenomenon can be fully understood or accounted for from a single perspective (be it purely syntactic, semantic, or morpho‐phonological), but from the interaction and mutual conditioning between those components, what is called “interfaces”, and, moreover, I adopted a particular approach to the interaction between the interfaces and the purely generative syntactic component in terms of a tension or frustration, the solution of which in a particular language L (given certain possibilities for morpho‐ phonological expression) results in a materialized form. I revisited the concept of causativity exposed, among others, by Schäfer (2008) and Kosta (2011, 2014), and proposed a different, though related, classification of ‘flavors of causativity’, which in turn helped us eliminate elements in the syntactic representation (e.g., VoiceP) in favor of less in number but more explicit semantically interpretable syntactic heads, thus simplifying the theoretical apparatus while maintaining empirical coverage in the examples analyzed with respect to the cited sources. A more elegant theory of causativity can be further developed along the lines hereby proposed, although there is much pending investigation. Among the problems to tackle, the availability of 19 The ceteris paribus clause refers to the existence of extra positive cognitive effects (or, in Grohmann’s 2003 terms, drastic interface effects). 20 This is related to the following problem (pointed out to me by Phoevos Panagiotidis, pers. comm.): Is it really the case that a root like √CAT is somehow more inclined to be merged with a D procedural node because of some syntactic requirement or is it that, as the N is “more widely used” (because of socio‐ historical factors, once again), and the phonological matrix is perceived in certain environments, the neurological connection is rutinized and the syntactic configuration reflects that statistical asymmetry by merging {D, √CAT} “by default”, as the most accessible (but not the only one) option for the inferential component to work with? See Krivochen (2012:123, ff.) for a take on the mater. 114 D.G. Krivochen alternations, and tests to identify verb typology are particularly salient, and I hope to address them in future research. References Alexiadou, Artemis. 2003. Functional structure in nominals: Nominalization and ergativity. Amsterdam: John Benjamins. Alexiadou, Artemis, Elena Anagnostopulou, and Florian Schäfer. 2006. The properties of anticausatives crosslinguistically. In Phases of interpretation, ed. by Mara Frascarelli, 187–211. Berlin: Mouton de Gruyter. Anderson, John. 1971. The grammar of case: Towards a localistic theory. Cambridge: Cambridge University Press. Baker, Mark. 1985. The Mirror Principle and morphosyntactic explanation. Linguistic Inquiry 16(3):373–415. Binder, Phillip. 2008. Frustration in complexity. Science 320: 322–323. Bleam, Tonia. 2003. Properties of the double object construction in Spanish. In A Romance perspective on language knowledge and use, ed. by Rafael Núñez‐Cedeño, Luis López and Richard Cameron, 233– 354. Amsterdam: John Benjamins. Boeckx, Cedric. 2007. Eliminating Spell‐Out. Linguistic Analysis 33(3–4):414–425. Boeckx, Cedric. 2010. Defeating lexicocentrism. Ms., Outline of Elementary Syntactic Structures. ICREA/Universidad Autónoma de Barcelona. Borer, Hagit. 2009. Roots and categories. Presented at 19th Colloquium on Generative Grammar, April 1–3, Vitoria‐Gasteiz, Faculty of Letters (Gradu Aretoa), University of the Basque Country. Borer, Hagit. 2013. The category of roots. Ms., University of London. Bresnan, Joan. 1971. Sentence stress and syntactic transformations. Language 47(2):257–281. Chomsky, Noam.1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass: MIT Press. Chomsky, Noam. 1998. Minimalist inquiries: The framework. MIT Working Papers in Linguistics 15. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Structures and beyond: The cartography of syntactic structures Vol. 3, ed. by Adriana Belleti, 104–131. Oxford: Oxford University Press. Chomsky, Noam. 2008. On phases. In Foundational issues in linguistic theory. Essays in honor of Jean‐Roger Vergnaud, ed. by Robert Freidin, Carlos Otero and Maria Luisa Zubizarreta, 291–321. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2013. Problems of projection. Lingua 130:33–49. Collier Katie, Balthasar Bickel, Carel van Schaik, Marta Manser, and Simon Townsend. 2014. Language evolution: Syntax before phonology?. Proceedings of the Royal Society B: Biological Sciences (281). doi: 10.1098/rspb.2014.0263. Culicover, Peter, and Ray Jackendoff. 2005. Simpler syntax. Oxford: Oxford University Press. Embick, David, and Rolf Noyer. 2005. Distributed Morphology and the syntax/morphology interface. htp://www.fosssil.in/Embick%20and%20Noyer%202005.pdf. [Printed in Oxford handbook of linguistic interfaces, ed. by Gillian Ramchand and Charles Reiss, 289–324. Oxford: Oxford University Press.] (Anti‐)causativity and the morpho‐phonology‐semantics tension 115 Fábregas, Antonio, and Michael Putnam. forthcoming. Syntactic variation through lexical exponents: Middle formation in Norwegian and Swedish. In Agent demotion in Mainland Scandinavian, ed. by Antonio Fábregas and Michael T. Putnam. Berlin: Mouton. Frías Conde, Xavier. 2010. An introduction to Functional Categorial Grammar. Madrid: Bubok. Givón, Talmy. 1980. The binding hierarchy and the typology of complements. Studies in Language 4(3):333– 377. Goldberg, Adele. Forthcoming. Compositionality. In Routledge handbook of semantics, ed. by Nicholas Riemer. London: Routledge. Grohmann, Kleanthes. 2003. Prolific domains. On the anti‐locality of movement dependencies. Amsterdam: John Benjamins. Gropen, Jess, Steven Pinker, Michelle Hollander, and Richard Goldberg. 1991. Affectedness and direct objects: The role of lexical semantics in the acquisition of verb argument structure. Cognition 41:153–195. Hale, Kenneth. 1986. Notes on world view and semantic categories: Some Warlpiri examples. In Features and projections, ed. by Pieter Muysken and Henk van Riemsdijk, 233–254. Dordrecht: Foris. Hale, Kenneth, and Samuel J. Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In The view from building 20: Essays in honor of Sylvain Bromberger, ed. by Kenneth Hale and Samuel Jay Keyser, 53–109. Cambridge, Mass.: MIT Press. Hale, Kenneth, and Samuel J. Keyser. 1997. The basic elements of argument structure. Ms. MIT. [Reprinted with changes as Chapter 1 of Hale and Keyser, 2002]. Hale, Kenneth, and Samuel J. Keyser. 2002. Prolegomenon to a theory of argument structure. Cambridge, Mass.: MIT Press. Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the pieces of inflection. In The view from building 20: Essays in honor of Sylvain Bromberger, ed. by Kenneth Hale and Samuel Jay Keyser, 111–176. Cambridge, Mass.: MIT Press. Harley, Heidi. 2002. Possession and the double object construction. Yearbook of Linguistic Variation 2:29–68. Harley, Heidi. 2003. How do verbs get their names?. Ms., University of Arizona. htp://dingo.sbs.arizona.edu/~hharley/courses/Oxford/Harley2.pdf. Harley, Heidi. 2013. External arguments and the Mirror Principle: On the distinctness of Voice and v. Lingua 125:34–57. Hauser, Marc D, Noam Chomsky, and W. Tecumseh Fitch. 2002. The faculty of language: What is it, who has it, and how did it evolve?. Science 198:1569–1579. Holinsky, Dee Ann. 1987. The case of the intransitive subject in Tsova‐Tush (Batsbi). Lingua 71:103–32. Jackendoff, Ray. 1983. Semantics and cognition. Cambridge, Mass.: MIT Press. Jackendoff, Ray. 1987. The status of thematic relations in linguistic theory. Linguistic Inquiry 18(3):369–411. Jackendoff, Ray. 1997. The architecture of the language faculty. Cambridge, Mass.: MIT Press. Jackendoff, Ray. 2002. Foundations of language. Oxford: Oxford University Press. Kosta, Peter. 2011. Causatives and anti‐causatives, unaccusatives and unergatives: Or how big is the contribution of the lexicon to syntax. In Formalization of grammar in Slavic Languages, ed. by Peter Kosta and Lilia Schürks, 235–296. Frankfurt am Main: Peter Lang. Kosta, Peter. 2014. On the causative/anti‐causative alternation as principle of affix ordering in the light of the Mirror Principle, the Lexical Integrity Principle and the Distributed Morphology. Ms. forthcoming in Zeitschrift für Slawistik. htp://ling.auf.net/lingbuzz/002055. Krivochen, Diego. 2011. An introduction to radical Minimalism I: On Merge and Agree (and related issues). In IBERIA 3(2):20–62. Krivochen, Diego. 2012. The syntax and semantics of the nominal construction. Frankfurt am Main: Peter Lang. Krivochen, Diego. 2013. A Frustrated mind. Ms., University of Reading. htp://ling.auf.net/lingbuzz/001932. 116 D.G. Krivochen Krivochen, Diego. forthcoming. Tokens vs. copies: Displacement revisited. Studia Linguistica. Krivochen, Diego, and Peter Kosta. 2013. Eliminating empty categories. Frankfurt am Main: Peter Lang. Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase structure and the lexicon, ed. by Johan Rooryck and Laurie Zaring, 109–137. Dordrecht: Kluwer. Laka, Itziar. 2002. Delimitando la gramática universal. In El lenguaje y la mente humana, ed. by Natalia Catalá, Itziar Laka and Carlos Piera, 77–119. Barcelona: Ariel. Lakoff, George, and John Ross. 1973. Is Deep Structure necessary?. Ms., Bloomington: Indiana University Linguistics Club. Lasnik, Howard, Juan Uriagereka, and Cedric Boeckx. 2005. A course in Minimalist syntax. Oxford: Blackwell. Larson, Richard. 1988. On the double object construction. Linguistic Inquiry 19:335–391. Levin, Beth, and Malka Rappaport Hovav. 1995. Unaccusativity. At the syntax‐lexical semantics interface. Cambridge, Mass.: MIT Press. Levin, Beth, and Malka Rappaport Hovav. 2003. Root and template in the representation of verb meaning. Talk delivered in Stanford University, May 15, 2003. htp://web.stanford.edu/~bclevin/su03.pdf. Levin, Beth, and Malka Rappaport Hovav. 2008. The English dative alternation: The case for verb sensitivity. Journal of Linguistics 44:129–167. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge, Mass.: MIT Press. Marantz, Alec. 1995. The Minimalist Program. In Government and Binding Theory and the Minimalist Program, ed. by Gerd Webelhuth, 349–382. Oxford: Blackwell. Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own Lexicon. In Proceedings of the 21st Annual Penn Linguistics Colloquium, ed. by Alexis Dimitriadis, Laura Siegel, Clarissa Surek‐Clark and Alexander Williams, 201–225. Philadelphia: University of Pennsylvania. Marvin, Tatjana. 2011. High and low applicatives in Slovenian and cross‐linguistically. Slovenski jezik – Slovene Linguistic Studies 8:95–113. Masullo, Pascual. 1992. Incorporation and Case Theory in Spanish: A crosslinguistic perspective. Doctoral dissertation, University of Washington, Washington. Mateu Fontanals, Jaume. 2000. Universals of semantic construal for lexical syntactic relations. Volume 99(4) of GGT Reports. ICERA‐Universidad Autónoma de Barcelona. Mateu Fontanals, Jaume. 2002. Argument structure. Relational construal at the syntax‐semantics interface. Doctoral dissertation, Universitat Autónoma de Barcelona. http://www.tesisenxarxa.net/TDX‐ 1021103–173806/. Mateu Fontanals, Jaume. 2005. Impossible primitives. In The composition of meaning and content: Foundational issues, ed. by Markus Werning, 213–230. Berlin: De Gruyter. Nunes, Jairo. 2004. Linearization of chains and sideward movement. Cambridge, Mass.: MIT Press. Panagiotidis, Phoevos. 2005. Against category‐less roots in syntax and word learning: Objections to Barner and Bale (2002). Lingua 115:1181–1194. Panagiotidis, Phoevos. 2011. Categorial features and categorizers. The Linguistic Review 28:325–346. Panagiotidis, Phoevos. 2013. Roots and categories. Lectures delivered at Stony Brook University, October 28th–30th. Partee, Barbara. 1984. Compositionality. In Varieties of formal semantics, ed. by Frank Landman and Frank Veltman, 281–312. Dordrecht: Foris. Putnam, Michael. 2010. Exploring crash‐proof grammars: An introduction. In Exploring crash proof grammars, ed. by Michael T. Putnam, 1–14. Amsterdam: John Benjamins. Pylkkänen, Lina. 2002. Introducing arguments. Doctoral dissertation, MIT. Pylkkänen, Lina. 2008. Introducing arguments. Cambridge, Mass.: MIT Press (Anti‐)causativity and the morpho‐phonology‐semantics tension 117 Schäfer, Florian. 2008. The syntax of (anti)causatives: External arguments in change‐of‐state contexts. Amsterdam: John Benjamins. Schäfer, Florian. 2012. Two types of external argument licensing  the case of causers. Studia Linguistica 66(2):128–180. Shibatani, Masayoshi. 1976. The grammar of causative constructions: A conspectus. In The grammar of causative constructions, ed. by Masayoshi Shibatani, 1–40. New York: Academic Press. Shibatani, Masayoshi, and Prashant Pardeshi. 2001. The causative continuum. Kobe Papers in Linguistics 3:136–177. Sperber, Dan, and Deirdre Wilson. 1995. Relevance: Communication and cognition. Oxford: Blackwell. Stroik, Thomas, and Michael Putnam. 2013. The structural design of language. Cambridge: Cambridge University Press. Tesniére, Lucien. 1959. Elements de syntaxe structurale. Paris: Klincksieck. Uriagereka, Juan. 1998. Rhyme and reason. Cambridge, Mass.: MIT Press. Uriagereka, Juan. 2002. Multiple Spell‐Out. In Derivations: Exploring the dynamics of syntax, ed. by Juan Uriagereka, 45–65. London: Routledge. Uriagereka, Juan. 2012. Spell‐Out and the Minimalist Program. Oxford: Oxford University Press. Van Valin, Robert, and David Wilkins. 1996. The case for effector: Case roles, agents, and agency revisited. In Grammatical constructions: Their form and meaning, ed. by Masayoshi Shibitani and Sandra A. Thompson, 289–322. Oxford: Oxford University Press. Williams, Edwin. 2003. Representation Theory. Cambridge, Mass.: MIT Press. Wilson, Deirdre, and Dan Sperber. 2004. Relevance Theory. In The handbook of pragmatics, ed. by Lawrence Horn and Gregory Ward, 607–632. Oxford: Blackwell. To cite this article: Krivochen, Diego G. 2014. (Anti‐)causativity and the morpho‐phonology‐semantics tension. Semantics‐Syntax Interface 1(2):82–117. SHORT CONTRIBUTION Mirative Past in Japanese Sumiyo Nishiguchi* Tokyo University of Science, Kuki‐city, Saitama, Japan 1. Non‐Past Past The past tense morpheme in simple Japanese sentences can receive non‐past interpretations in contexts of the speaker expressing surprise at discovery, recalling something or at seeing the fulfillment of expectations (Mikami 1953; Kunihiro 1967; Teramura 1984; Machida 1989; Mo 1992; Inoue and Ogoshi 1997; Inoue 2001). For example, (1a) can be utered when the speaker was looking for a book and has now found it. While the present tense may be used, the past tense is more natural, even without past reference. The bus in (1b) is still on its way. The uterance in (1c) is a shout at someone who is blocking the speaker’s way at the time of speech, and the street vendor yells (1d) to a passersby who has not yet purchased anything. I consider the examples in (1c) and (1d) to be imperative uses of the past tense morpheme. The present paper primarily discusses non‐past past1 sentences such as those in (1a) and (1b), and then in section 4 expands the analysis to capture the imperative use shown in (1c) and (1d).2 (1) a. A, koko‐ni oh here‐LOC ‘Oh, it was here.’ * at‐ta/#a‐ru. be‐PST/be‐PRES Email: nishiguchi@rs.tus.ac.jp I am grateful to extensive comments provided by anonymous reviewers. Needless to say, all remaining errors are mine. In this paper, the non‐past past particle ‐ta in Japanese is also referred to as the mirative past, mirative ‐ta, or past tense of discovery interchangeably. 1 2 List of abbreviations in the glosses: COND: locative, PAR: ACC: accusative, BPG: best possible grounds, COMP: complementizer, conditional, DECL: declarative, EVID: evidential, GEN: genitive, IMPF: imperfective, INF: infinitive, LOC: MIR: mirative, NEG: negative, NOM: nominative, NONPST: nonpast, NX.PST: non‐experienced past, particle, PRES: present, PROG: progressive, PST: past, SG: singular, SUBJ: subject, SUR: surprise, TOP: topic. Semantics‐Syntax Interface, Volume 1, Number 2, Fall 2014 118–132 © 2014 by University of Tehran http://semantics‐syntax.ut.ac.ir Mirative past in Japanese 119 b. A, basu‐ga ki‐ta. oh bus‐NOM come‐PST ‘Oh, the bus is coming.’ (Lit. ‘The bus has come.’) c. Doi‐ta, doi‐ta. step.aside‐PST step.aside‐PST ‘Step aside, step aside!’ d. Kat‐ta, kat‐ta. buy‐PST buy‐PST ‘Buy one, buy one!’ The structure of this paper is as follows: this section explains three major characteristics of the past tense of discovery in Japanese. Section 1.1 illustrates the temporal mismatch between matrix tense and future adverbs in English and Japanese. Section 1.2 categorizes the non‐past past as a modal mirative marker. Section 1.3 identifies how mirative past sentences convey old information. Section 2 demonstrates that Ippolito’s (2003) analysis on temporally mismatched matrix sentences in English applies to the non‐past past in Japanese. Section 2.1 introduces Ippolito’s approach and section 2.2 explains how the past tense is interpreted outside the proposition. Section 3 identifies different implicatures raised by mismatched matrix sentences in English and the past tense of discovery in Japanese. Section 4 outlines the analysis of the imperative use of the past tense morpheme. And section 5 concludes the paper. 1.1. Temporal Mismatch In English, the past tense in I was to have an exam tomorrow refers to the past and indicates that, according to her agenda in the past, the speaker had an exam scheduled for the next day, and the exam might have been cancelled later on. On the contrary, in Japanese, Tomorrow, it was that I take an exam, as in (2), indicates that the speaker had forgotten about the exam before her utterance and she now remembers that the exam is scheduled for the next day. The exam is still on schedule at the time of the utterance. (2) Shimat‐ta. Ashita‐wa close‐PST tomorrow‐TOP shiken‐o exam‐ACC uke‐ru‐no‐dat‐ta. take‐INF‐COMP‐DECL‐PST ‘Oh, no! I forgot that I’d take an exam tomorrow.’ (Lit. ‘Tomorrow, it was that I take an exam’). The future adverb tomorrow interferes with the regular past tense, as in (3). (3) (#Ashita/kino)(‐wa), tomorrow/yesterday‐TOP shiken‐ga at‐ta. exam‐NOM be‐PST ‘There was an exam tomorrow/yesterday.’ 120 S. Nishiguchi Such temporally mismatched future adverbs are not compatible with ‘real’ past tense with past reference but can co‐occur with non‐past past tense. Temporal mismatch between the past tense morphology and a future adverbial tomorrow has been discussed in past studies (Ogihara 2000; Ippolito 2003, 2004; Higginbotham 2007). Ippolito (2003) calls such a temporal mismatch in Originally, Charlie left tomorrow as temporally mismatched matrix sentences. Higginbotham (2007) calls it indexical mismatch. The future temporal adverb tomorrow has been known to be infelicitous with the regular past tense with past reference, as in (4b). However, (4a) is felicitous because it refers to Charlie’s cancelled plan for the future. (4) a. Originally, Charlie left tomorrow. b. Charlie played with Lucy (yesterday/#tomorrow). 1.2. Mirativity In this paper, I argue that the past tense morpheme without a past reference in Japanese is a mirative modal marker. Mirativity (admirativity) refers to linguistic encoding of the speaker’s surprise and is related to unprepared mind, new information, and speaker’s unexpectedness (DeLancey 2001; Aikhenvald 2004). Mirativity marks whether the information represents knowledge which is new to the speaker or knowledge which is already integrated into the speaker’s picture of the world (DeLancey 2001). I take mirativity to be a part of evidentiality. It has been observed that evidential markers are fused or composed with tense in Tariana (Aikhenvald 2004). Faller (2004) analyzes the past tense marker ‐sqa in Cuzco Quechua as an indirect evidential marker. The mirative use of ‐sqa is termed as sudden discovery tense, which may have present time reference. Example (5) is utered by a speaker upon unexpectedly seeing Marya, who was not supposed to be there.3 (5) Kay‐pi‐(má) this‐LOC‐SUR ka‐sha‐sqa Marya‐qa. be‐PROG‐NX.PST Marya‐TOP ‘Marya is here!’ (Faller 2004:53) ‐sqa is used for expressing a compliment to the cooking, as in example (6). 3 The word order of Cuzco Quechua, which belongs to Quechuan language family, is SOV. Mirative past in Japanese (6) Lawa‐yki‐qa soup‐2‐TOP 121 smak‐mi qa‐sqa! nice‐BPG be‐NX.PST ‘Your soup is very tasty! ’ (Faller 2004:53) According to DeLancey (2001), Hare (Athapaskan) lō is used at the sudden (direct) perception of an unexpected fact and “has the sense of surprise at an unanticipated situation,” as seen in (7). (7) e‐we’ its‐hide ghálayeda lõ work.2SG.SUBJ.IMPF MIR ‘I see youʹre working on hides! ’ (DeLancey 2001:376) Furthermore, the Turkish past tense morpheme mIş has a mirative use and expresses surprise (Slobin and Aksu 1982). When the speaker has just become aware of something which s/he was not conscious about, mIş can be used for both eventive and stative predicates, as in (8) and (9), respectively. Example (8) can be an expression of surprise upon encountering concrete physical evidence that Kemal came. (8) Kemal Kemal gel‐mIş. come‐MIR ‘Kemal came.’ (Slobin and Aksu 1982:196) (9) Selma Selma bura‐da‐ymIş. here‐LOC‐MIR ‘Selma is here.’ (Slobin and Aksu 1982:193) In view of similar interpretations between the Japanese non‐past past marker ‐ta and the mirative markers in other languages, it is plausible to consider the surprise past tense marker ‐ta to be a mirative (Sadanobu and Malchukov 2006). Direct visual perception of an unexpected or expected fact triggers –ta of discovery or that of realization of the fulfilment of expectation. (10) (Finding a snake in a zoo,) Wa, koret‐te hebi‐dat‐ta. Iwa‐ja‐nakat‐ta. Oh this‐PAR snake‐be‐PST rock‐be‐NEG‐PST ‘Oh, this is a snake. It is not a rock.’ 122 S. Nishiguchi On the other hand, first‐hand information triggers what is called –ta of recalling. (11) (Looking at a calendar,) Ashita‐wa saijitsu‐dat‐ta. tomorrow‐TOP holiday‐be‐PST ‘Tomorrow is holiday.’ Therefore, I propose an addition of the mirative marker –ta in (12d) to the classification of the Japanese evidential markers by Aoki (1986) in (12ac). (12) a. visible, tangible or audible evidence collected through his own senses to make inferences: yo (i) Soto‐wa ame‐no‐yo‐da. rain‐GEN‐EVID‐DECL outside‐TOP ‘It seems to be raining outside. ’ b. circumstantial: rashi (ii) Chizu‐ga machigat‐tei‐ta‐rashii. map‐NOM wrong‐PROG‐PST‐EVID ‘It seems the map was wrong. ’ c. hearsay or inferential about what occurred in the past: so (iii) Chizu‐ga machigat‐tei‐ta‐so‐da. map‐NOM wrong‐PROG‐PST‐EVID‐be ‘I heard that the map was wrong. ’ d. mirative, surprise from direct evidence: ‐ta (iv) A, hebi‐ga oh i‐ta.4 snake‐NOM be‐EVID ‘Oh, there is a snake here. ’ 1.3. Presupposition Although non‐past ‐ta sentences are usually simple sentences, they can be embedded under factive predicates such as realize:5 4 As a reviewer pointed out, the topic marker ‐wa raises the mirative reading of ‐ta in general. However, this example shows that it is not always the case. 5 A reviewer brought my attention to the fact that the past tense of discovery behaves differently from Mirative past in Japanese (13) Asu‐ga 123 shiken‐dat‐ta‐to tomorrow‐NOM waka‐reba, Ken‐wa odoroku‐daro.6 exam‐be‐PST‐COMP realize‐COND Ken‐TOP surprise‐will Lit. ‘When he realizes that he had an exam tomorrow, Ken will be surprised.’ A predicate, such as regret, is a “hole” to presupposition that preserves the presupposition of the embedded sentences (Kartunen 1973). As shown in (14), negation does not cancel the presupposed content: that is, John insulted her. (14) a. John regrets he insulted her. b. John does not regret he insulted her. According to Quirk, Greenbaum, Leech and Svartvik (1985), certain adverbs, called “content disjuncts,” comment on the degree of truth and “present a comment on the truth value of what is said.” Adverbs such as clearly and definitely express conviction while another group of disjuncts, such as perhaps and possibly, express some degree of doubt on the truth value of what is said. In non‐presuppositional constructions, such as in (15a), clearly/definitely can be used to comment or negotiate on what is asserted  (15a) would mean that it was strongly true that John insulted her. On the other hand, these adverbs are incompatible with presuppositional predicates such as regret in (15b) because the truth value of the content is already presupposed. other evidential markers in (10) with respect to the embeddability. Except for the mirative ‐ta, the only marker that allows direct attachment of the complementizer is rashi in (ii) below whose semantics may change in the embedded clause. Accounting for such different behavior falls outside of the scope of this paper. (i) *Soto‐wa outside‐TOP ame‐no‐yo(‐da)‐to wakat‐ta. rain‐GEN‐EVID‐DECL‐COMP realize‐PST ‘I realized that it seemed raining outside.’ (ii) Chizu‐ga machiga‐tei‐ta‐rashii‐to map‐NOM wrong‐PROG‐PST‐EVID‐COMP wakat‐ta. realize‐PST ‘I realized that the map appeared to be wrong.’ (iii) Chizu‐ga map‐NOM machiga‐tei‐ta‐so*(‐da)‐to wakat‐ta. wrong‐PROG‐PST‐EVID‐DECL‐COMP realize‐PST ‘I realized that the map appeared to be wrong.’ (iv) Kumo‐ga spider‐NOM itan‐dat‐ta‐to wakat‐ta. be‐be‐PST‐COMP realize‐PST ‘I realized that there was a spider.’ 6 I would like to thank an anonymous reviewer for providing me with this example. 124 S. Nishiguchi Delin (1992), in her discussion on it‐clefts, demonstrates the non‐negotiability of presupposed content by using the following examples. In presuppositional environments such as the complement of the verb regret, the adverbs that express some degree of doubt are infelicitous, as in (15b). (15) a. John clearly/definitely insulted her. b. ??John regrets that he clearly/definitely insulted her.7 (Delin 1992:301) As per Delin’s test, the propositions in mirative ‐ta utterances are presupposed and non‐negotiable since adverbs such as tabun “probably,” tashikani “certainly,” and akirakani “clearly” interfere with non‐past ‐ta sentences. These adverbs can only be felicitously used when the truth value of the proposition it modifies is not presupposed. (16) a. #A, ashita‐wa tabun/tashika‐ni/akirakani Hanako‐no oh, tomorrow‐TOP probably/certainly/clearly tanjobi‐dat‐ta. Hanako‐GEN birthday‐be‐PST ‘Oh, it is probably/certainly/clearly Hanako’s birthday tomorrow.’ b. A, #tabun/?tashikani/#akirakani hon‐ga oh perhaps/certainly/clearly at‐ta. book‐NOM be‐PST ‘Oh, the book is perhaps/certainly/clearly here.’ On the other hand, yappari ‘as expected’ can be added freely to non‐past past sentences, suggesting that prior expectations are not in conflict with the non‐past past tense and that the proposition is presupposed to be true: (17) a. A, yappari Oh as expected ki‐ta. come‐PST ‘Oh, (s/he) is coming as I thought.’ 7 A reviewer brought the following seeming counterexample to Quirk et al.ʹs (1985) assertion, in which the presupposed content can be modified by clearly. (i) John does not regret that he clearly insulted her. I assume that the negation associates with and “licenses” clearly, which is emphatic and introduces scalar endpoint of the degree of insult in this example. The relation between negation and clearly is comparable to the one between not and much in John did not regret he spoke much, in which the negation licenses the NPI much. Mirative past in Japanese b. Yappari as expected 125 koko‐ni at‐ta. here‐LOC be‐PST ‘(The book) was here as I expected.’ Thus, the embeddability with the fact predicates, non‐negotiability and compatibility with prior expectation suggest the presupposed content of the non‐past past sentences. 2. Modal Use of Past Now, this paper will compare the past tense of discovery in Japanese with mismatched matrix sentences in English. Referring to Ippolito (2003), this section argues that the non‐past past tense in Japanese is a modal past that restricts the accessibility relation of the planning modal. This explains the temporal mismatch with future adverbs. 2.1. A Summary of Ippolito 2003 Ippolito (2003) discusses temporal mismatches in counterfactual conditionals and claims that the past tense is interpreted outside of the proposition in which it occurs. The past tense in the antecedent is interpreted inside the accessibility relation of the modal operator so that the future temporal adverb tomorrow does not interfere with it in (18). (18) If Charlie had left for Paris tomorrow, he would have met my cousin there. In order to strengthen her proposal, Ippolito extends her analysis to the tense mismatch in matrix sentences. In her analysis on temporally mismatched matrix sentences as in (19a), all sentences with a temporal mismatch are interpreted modally. The temporal mismatch between the past tense and the future adverbial in (19a) indicates that the past tense scopes out of the proposition. In (19b), the past tense locates the event in the past so that tomorrow cannot occur. (19) a. Originally, Charlie left tomorrow. (Ippolito 2003:179) b. Charlie played with Lucy (yesterday/#tomorrow). (Ippolito 2003:149) 126 S. Nishiguchi In (19b), the past tense is interpreted clause‐internally so that the future adverbial tomorrow is ungrammatical. On the other hand, in (19a), the past tense is interpreted outside of the clause so that tomorrow is admitted. The past tense is a temporal argument of the accessibility relation between the actual world and the possible worlds in which Charlie’s departure is scheduled for the day after the day of utterance. Charlie’s departure was planned at some point g(2) in the past. (20) [ t2[past] ]g,c = defined only if g(2) < tc; if defined, then [ t2[past] ]g,c = g(2) (Ippolito 2003:157) In planning modal sentences, as in (21a), the worlds quantified over by the modal operator are worlds compatible with what was necessarily scheduled in the actual world at a past time. The past tense restricts the accessibility relation of the modal operator as shown in (21b) and (21c), so that the matrix clause is tenseless. The past tense operator outscopes modal. Therefore, there is no problem with tomorrow. (21) a. Charlie left tomorrow. b. [CP [covert modal<st<st,t>> [R<s<i<st>>>(w1)(t2[past])]<st>]<st,t> [λw2[ Charlie leaves tomorrow in w2]]<st>] c. [[ CP ]]g, c = 1 iff w  W [w is compatible with the plans in wc at g(2)  Charlie leaves in w] defined only if g(2) < tc (tc: utterance time, wc: actual world) On the other hand, in epistemic modal sentences, as in (22a), the past tense is interpreted clause‐internally. The interpretation of (22a) is that “in view of the circumstances at the time of utterance time, Charlie could possibly have left in the past,” as given in (22b) and (22c). The clause‐internal past tense cannot be used with tomorrow. (22) a. Charlie could have left (*tomorrow). b. [CP [modal<st<st,t>>[R<s<i<st>>>(w1)(t2)]<st>]<st,t>[λw2[Charlie left in w2]]<st>] c. [[ CP ]]g, c = 1 iff w  W [w is compatible with what the speaker knows in wc at tc  Charlie left in w] Such scopal interactions between the past tense and modal operator explain (un)grammaticality of future adverbs. The past tense is a temporal argument of the planning modal so that the proposition itself is interpreted tenseless and is compatible Mirative past in Japanese 127 with the future adverb tomorrow. On the contrary, the time of evaluation of the epistemic modal is the utterance time. The past tense is interpreted clause‐internally, which results in contradiction with the future adverb. 2.2. Meaning of Mirative Past If we apply Ippolito’s (2003) analysis to the Japanese example (23c), the past tense is a temporal argument of the accessibility relation between the actual world and the possible worlds in which an exam is scheduled for the day after the day of utterance. The exam was scheduled at some point in the past. (23) a. (#Ashita/kino) shiken‐ga at‐ta. tomorrow/yesterday exam‐NOM be‐PST ‘There was an exam (tomorrow/yesterday).’ b. I was to have an exam tomorrow. c. Shimatta. closed Ashita‐wa shiken‐dat‐ta. tomorrow‐TOP exam‐be‐PST ‘Oh, no! I had an exam tomorrow.’ If Ippolito’s analysis on mismatched matrix sentences applies to the Japanese matrix mirative, the past tense does have temporal contributions, that is, the time argument switches the evaluation time of the accessibility relations in the restrictor of the covert modal backward, as illustrated in (24). The exam was scheduled according to the past perspective. The speaker has found that the exam was scheduled in the past, for the next day. (24) t <st,t> planning modal<st,<st,t>> <st> <st> t2[past] <i, st> R<s,<i,st>> w1 (25) [covert modal<st<st,t>>[R<s<i<st>>>(w1)(t2[past])]<st>]<st,t>[λw2[I have an exam tomorrow in w2]]<st> (26) [ (23c) ]g,c = 1 iff w  W[w is compatible with what the speaker plans in wc at g(2)  there is an exam tomorrow in w] defined only if g(2) < tc 128 S. Nishiguchi In (24) and (25), the accessibility relation of the modal combines with the actual world and the contextually salient past time, which is the restrictor of the modal. What is in the nuclear scope is a tenseless clause, a function from possible worlds to the truth‐values. In all the worlds compatible with the speakerʹs plans at the actual world at the past salient time g(t2), the speaker has an exam scheduled tomorrow. Moreover, as pointed out by a reviewer, the covert modal operates at the CP level, since the past tense of discovery is interpreted as non‐past in topicalized sentences, as seen in (23c). The past tense is interpreted outside of the proposition so that it does not interfere with the future temporal adverb tomorrow. The past tense is interpreted in the domain of the modal operator, i.e. inside of the accessibility relation. (27) a. A, Oh hon‐ga at‐ta. book‐NOM be‐PST ‘Oh, the book was here.’ b. covert modal<st<st,t>> [R<s<i<st>>>(w1)(t2[past])]<st>[λw2[the book is on the table in w2]]<st> In the accessible worlds from the actual world in the past (g(2)), the book was supposed to be on the table. This fact was forgotten or doubted until immediately before the utterance time. Ippolito’s analysis provides a good explanation to the temporal mismatch in the mirative past sentences. The past tense is interpreted as a temporal argument of the accessibility relation of the modal so that it does not contribute to the interpretation of the proposition. However, the non‐past past in Japanese differs from the mismatched matrix sentences in English in a significant way that is not predicted by Ippolito’s analysis. 3. Different Conversational Implicatures Even though the interpretation of the modal past is similar between temporally mismatched matrix sentences in English and sentences with the past tense of discovery in Japanese, Ippolito’s analysis does not explain the different conversational implicatures between English mismatched matrix sentences and Japanese temporally mismatched past sentences. English mismatched matrix sentences such as (Originally,) there was an exam tomorrow give rise to conversational implicature, such that if the speaker used the past tense instead of the non‐past tense, she would probably not in a position to use either the present or future tense morpheme, and that is why she used Mirative past in Japanese 129 the non‐past past tense. That is, the schedule would not hold anymore at the time of utterance: the exam would have been cancelled or rescheduled, as in (28). (28) There was an exam tomorrow but it was postponed. Thank goodness! (29) Temporally mismatched matrix sentences in English: Conversational implicature: The impossibility of the proposition at speech time (30) a. P ∩ ct2<tc   b. P ∩ ctc =  (P: presuppositions; the set of all worlds w such that the conjunction of all the presuppositions of p is true in w) (31) context set ct = {w  W: w  p for all p that the speaker presupposes to be true at t} Nevertheless, the Japanese mirative past does not have such implicature, as the content of the proposition is presupposed and factive at the utterance time, as previously shown in section 1.3. The Japanese mirative past asserts that “the exam is scheduled for tomorrow or “the book is on the table,” both at the time of utterance. (32) Temporally mismatched matrix mirative past sentences in Japanese: Implicature: p was not known in common ground at t2 and t2 < tc (33) a. P ∩ ct2<tc =  b. P ∩ ctc   (34) The book was here. Implicature: The book’s presence was not in common ground. Japanese matrix mirative past sentences do not imply the falsity of the proposition at speech time but it does give rise to another implicature in accordance with the Gricean Maxim of Quantity. (35) Maxim of Quantity (Grice 1975): Make your contribution as informative as is required for the current purposes of exchange. Do not make your contribution more informative than required. (36) a. Statement on past belief on p: non‐informative if he had believed p. 130 S. Nishiguchi b. Statement on past belief on p: informative if he had believed p or was not sure of p. As stated in (37), the mirative past statements are not informative if both the speaker and the hearer had already believed p. Based on the assumption that the speaker should be informative enough, the utterance conveys the conversational implicature that the speaker had not believed p or was not sure of p. (37) a. English: The speaker’s utterance of [[ must[R(w1)(t2[past])]][ λw2 [there is a meeting tomorrow in w2]]]g,c is less informative than the utterance [[ must[R(w1)(t2)]][ λw2 [there is a meeting tomorrow in w2]]]g,c, which gives rise to the implicature [[ must[R(w1)(t2)]][ λw2 [ [there is a meeting tomorrow in w2]]]]g,c b. Japanese: The speaker’s utterance of [[ must[R(w1)(t2[past])]][ λw2 [there is a meeting tomorrow in w2]]]g,c gives rise to the implicature that the speaker believed [[ must[R(w1)(t2[past])]][ λw2 [there is a meeting tomorrow in w2]]]g,c Assertion in mismatched matrix sentences imply that until the speech time, the speaker had not perfectly believed or had forgotten that the proposition held in the past. The assertion in mismatched matrix sentences is thus informative, even though the information is old. The implicature of the modal past is cancellable, as the following examples show: (38) a. There was an exam tomorrow but it was postponed. Thank goodness! b. There was an exam tomorrow and there will be one as scheduled. The uterance in (38a) suggests that the implicature that arises from the modal past is real, in that the exam was in fact postponed. However, the implicature is cancellable, rendering example (38b) felicitous. The speaker of (39a) was not ready when the bus came, while the speaker of (39b) expected the arrival of the bus. The implicature of uncertainty was cancelled. (39) a. A, basu‐ga ki‐ta. Jitsuwa kuru‐to oh bus‐NOM come‐PST in.fact come‐COMP omot‐te nak‐at‐ta. think‐INF NEG‐be‐PST ‘Oh, the bus is coming. In fact, I didn’t expect it to come.’ Mirative past in Japanese b. A, oh 131 basu‐ga ki‐ta. Jitsuwa kuru‐to omot‐te‐ta. bus‐NOM come‐PST in.fact come‐COMP think‐be‐PST ‘Oh, the bus is coming. In fact, I expected it to come.’ 4. Remarks on the Imperative Use of Past The given analysis can be expanded to the imperative use of the past morpheme which does not have the past reference either, as mentioned in section 2. (40) a. Hashit‐ta, run‐PST hashi‐ta. run‐PST ‘Run!’ b. Saa, let’s hatara‐i‐ta, hatara‐i‐ta. work‐be‐PST work‐be‐PST ‘Go back to work!’ Parallel to other past tense morphemes without the past reference, the covert modal, which is deontic, is present; this accounts for why the past tense is not interpreted inside of the proposition. The accessibility relation of the deontic modality takes the past tense as its argument. In view of the rules, the addressee was supposed to run or work. An extensive discussion on the imperative use of the past tense morpheme is beyond the scope of this paper, and remains a topic for future research. 5. Conclusion The past tense of discovery without past reference restricts the accessibility relation of the modal operator so that the past tense is interpreted outside of the clause and a temporal mismatch does not occur. This past is a modal past to which Ippolito’s (2003) analysis on temporally mismatched matrix sentences in English directly applies. Nevertheless, the non‐past past tense in Japanese differs from English in conversational implicature. The non‐past, mirative past implicates that the speaker was uncertain or was not sure of p out of a Gricean maxim of quantity, while temporally mismatched matrix sentences in English give rise to another implicature that the proposition does not hold at the utterance time. 132 S. Nishiguchi References Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford: Oxford University Press. Aoki, Haruo. 1986. Evidentials in Japanese. In Evidentiality: The linguistic coding of epistemology, ed. by Wallace Chafe and Johanna Nichols, 223238. Norwood, NJ: Ablex. DeLancey, Scot. 2001. The mirative and evidentiality. Journal of Pragmatics 33:369382. Delin, Judy. 1992. Properties of it‐cleft presupposition. Journal of Semantics 9:289306. Faller, Martina. 2004. The deictic core of ‘non‐experienced past’ in Cuzco Quechua. Journal of Semantics 21:4585. Grice, Paul. 1975. Logic and conversation. In Syntax and semantics 3: Speech acts, ed. by Peter Cole and Jerry L. Morgan, 4158. NY: Academic Press. Higginbotham, James. 2007. The anaphoric theory of tense. In Proceedings of SALT 16. Ithaca, NY: Cornell University. Inoue, Masaru. 2001. Gendai nihongo no ta [ta in contemporary Japanese]. In Ta no Gengogaku [Linguistics of ta], ed. by Tsukuba Gengo Bunka Forum, 97163. Tokyo: Hituzi Syobo. Inoue, Masaru, and Naoki Ogoshi. 1997. Kakokei no shiyo ni kakawaru goyoronteki yoin: Nihongo to chosengo no baai [Pragmatic factors for the use of past tense: In Japanese and Korean]. Nihongo Kagaku 1:3752. Ippolito, Michela. 2003. Presuppositions and implicatures in counterfactuals. Natural Language Semantics 11:145186. Ippolito, Michela. 2004. Imperfect modality. In The syntax of time, ed. by Jacqueline Guéron and Jacqueline Lecarme, 359388. Cambridge, MA: MIT Press. Kartunen, Lauri. 1973. Presupposition of compound sentences. Linguistic Inquiry 4:169193. Kunihiro, Tetsuya. 1967. Kozoteki imiron [Structural semantics]. Tokyo: Sanseido. Machida, Ken. 1989. Nihongo no Jisei to asupekuto [Tense and aspect in Japanese]. Tokyo: Arc. Mikami, Akira. 1953/1972. Gendai goho josetsu [Introduction to contemporary grammar]. Tokyo: Kuroshio. Mo, Sejong. 1992. Hakken, omoidashi‐ni‐okeru ru‐kei‐to ta‐kei [Ru‐form and ta‐form at discovery and recalling]. Nihongogaku 12:8897. Ogihara, Toshiyuki. 2000. Counterfactuals, temporal adverbs, and association with focus. In Proceedings of SALT 10, 115131. Ithaca, NY: CLS Publications, Cornell University. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik. 1985. A comprehensive grammar of the English language. London: Longman. Sadanobu, Toshiyuki, and Andrej Malchukov. 2006. Modal extensions of aspecto‐temporal forms in Japanese. Paper presented at CHRONOS 7, University of Antwerp. Slobin, Dan, and Ayhan Aksu. 1982. Tense, aspect, modality, and more in Turkish evidentials. In Tense‐ aspect: Between semantics and pragmatics, ed. by Paul J. Hopper, 185200. Amsterdam: John Benjamins. Teramura, Hideo. 1984. Nihongo no shintakusu to imi II [Japanese syntax and semantics II]. Tokyo: Kuroshio. To cite this article: Nishiguchi, Sumiyo. 2014. Mirative past in Japanese. Semantics‐Syntax Interface 1(2):118–132. SHORT CONTRIBUTION In Defense of the Reference Time Daniel Altshulera* and Susanna Melkonianb a,b Heinrich Heine University, Düsseldorf, Germany 1. Introduction Intuitions about past‐tensed narratives were instrumental in the development of a theory of tense and aspect in the 1980s, giving rise to Discourse Representation Theory (DRT; Kamp 1981). The key observation was that eventive sentences in past‐tensed narratives ‘move the story forward’, whereas stative sentences do not. Consider the two discourses below, modified from Kamp, Genabith and Reyle 2011: (1) a. Josef turned around. b. The man pulled his gun from his holster. c. Josef took a step back. (2) a. Josef turned around. b. The man had a gun in his holster. c. Josef took a step back. The only difference between (1) and (2) is the b‐sentence. Whereas the former discourse contains the eventive VP pulled a gun, the latter discourse contains the stative VP had a gun. This impacts how we understand the ordering of the described eventualities. The discourse in (1) exemplifies narrative progression: the events are understood to occur in the order in which they are described. The discourse in (2), * Email: daltshuler@gmail.com smelkonian@phil.uni‐duesseldorf.de We would like to thank two anonymous reviewers for their questions, comments and suggestions on a previous version of this squib. The first author is supported by the Strategic Research Grant (Strategisches Forschungsfond) from Heinrich Heine University for the project Temporal Constraints on Discourse Structure. The second author is supported by the DFG‐Sonderforschungsbereich 991: The Structure of Representations in Language, Cognition and Science, Subproject C04: Conceptual Shifts – Their Role in Historical Semantics. Semantics‐Syntax Interface, Volume 1, Number 2, Fall 2014 133–149 © 2014 by University of Tehran http://semantics‐syntax.ut.ac.ir 134 D. Altshuler and S. Melkonian however, is more complicated. Although we infer narrative progression in (2a) and (2c), the state described in (2b) is understood to hold at the time that Josef turned around. Such intuitions led to a theory of temporal anaphora that builds on Partee’s (1973) idea of treating the past tense like a pronoun. In particular, the past tense seeks a salient time antecedent, called a reference time, prior to the speech time. This is analogous to she seeking a salient individual antecedent that is female. In turn, the reference time is related to the time of the eventuality (event or state) by other mechanisms (e.g. the aspect). In other words, the reference time mediates between the speech time and the event time. This idea, first proposed by Reichenbach (1947), has often been referred to as the “two‐dimensional” analysis of tense (Kamp 1999/2013). Relating this analysis to the discourses above, the crux is that sentences with eventive VPs (or “eventive sentences”) provide a reference time of the right kind, that is, one that satisfies the presupposition of the past tense, whereas sentences with stative VPs do not provide a reference time (Partee 1984; Hinrichs 1986). For example, in (1c), the reference time is provided by the eventive sentence in (1b), namely some time in the past that is ‘just after’ the man pulled his gun from his holster.1 As a result, we understand the gun pulling to take place after the turning around. In (2c), however, the antecedent is some time in the past ‘just after’ Josef turned around because (2a) is the only other eventive sentence in the discourse and hence the only other sentence that could provide a reference time. As a result, we understand the taking a step back to take place after the turning around, but not necessarily after the man had a gun in his holster. Research in AI in the 1980s pioneered by Jerry Hobbs also analyzed narrative discourse, but from a different point of view. Instead of explaining our intuitions about event ordering in terms of semantic rules, Hobbs (1985, 1990) appealed to commonsense reasoning with non‐linguistic information. In particular, Hobbs extended David Hume’s principles of idea association to model how we interpret a given discourse; he assumed that there is a primitive class of rhetorical relations according to which a discourse is organized, that is, a discourse is coherent because it is organized by particular rhetorical relations.2 1 The ‘time just after’ is purposely vague. It is “determined by the hearer’s understanding of the nature of events being described in a narrative, the overall degree of detail in which events are being described, and common knowledge about the usual temporal relationships among events...each successive sentence presents the very next event that transpires that is important enough to merit the speaker’s describing it to the hearer, given the purpose of the narration” (Dowty 1986:47). 2 Hume famously wrote: “To me there appear to be only three principles of connection among ideas, namely Resemblance, Contiguity in time or place, and Cause or Effect” (Hume 1748/2008:16). In defense of the reference time 135 For example, the Humean notion of contiguity is often characterized by the so‐ called “OCCASION relation” (Kehler 2006:250), which is defined below, in (3). It has played a vital role in a variety of research projects within the cognitive sciences3 because it is seen as the backbone to any narrative (Hobbs 1990: Ch. 5).4 (3) OCCASION(S0, S1): a. A change of state can be inferred from the assertion of S0, whose final state can be inferred from S1 b. A change of state can be inferred from the assertion of S1, whose initial state can be inferred from S0 (Hobbs 1990:87). To see the impact of the definition above, let us relate it to the discourse in (1). In particular, let’s suppose that (1a) corresponds to discourse segment S0 and (1b) corresponds to discourse segment S1. In this case, we can say, following (3a), that a change of state could be inferred from (1a) due to the telic predicate turned around, whose final state, that is, Josef having change his physical location, can, in some sense, be ‘inferred’ from the man pulling his gun from his holster. It’s natural to understand the man’s pulling his gun from his holster to be contingent on (and hence inferable from) Josef having turned around.5 In contrast, it is unlikely that (2b) is linked to (2a) via OCCASION because, given world knowledge reasoning, it’s hard to see how a man having a gun in his holster could be contingent on Josef’s turning around. Instead, (2b) is linked to (2a) by BACKGROUND (Lascarides and Asher 1993), whose temporal import is ‘overlap’ between two described eventualities.6 In sum, coherence‐based theories like Hobbs’ derive the temporal ordering of events in narrative discourse by appealing to particular kinds of world knowledge reasoning, such as the kind encoded by OCCASION and BACKGROUND. This, according to Kehler (2002), allows for a much simpler analysis of the past tense: it constitutes a precedence relation between the speech time and the event time (Kehler 2002:191); there is no need for an intermediary notion of a reference time because narrative 3 See, e.g., Kehler 2002, Wolf and Gibson 2005, Altshuler 2012, Rhode and Horton 2014 and references therein. 4 A satisfactory analysis of narrative discourse is much more complex, however, involving a multitude of rhetorical relations (in addition to OCCASION). We consider some of these relations in the next section. 5 Note that the clause in (3a) is sufficient to establish OCCASION between (1a) and (1b). See Hobbs 1990, for examples which motivate the clause in (3b) and narrative sequences which blur the distinction between (3a) and (3b), i.e. where OCCASION can be established by either clause. 6 Lascarides and Asher 1993 define BACKGROUND(S0, S1) as: The state described by S0 is the ‘backdrop’ or circumstance under which an event described by S1 occurred. 136 D. Altshuler and S. Melkonian progression (or lack thereof) is an epiphenomenon of speakers seeking coherence in a given discourse.7 A similar conclusion was also reached by Lascarides and Asher (1993:43738) who were the first to point out, based on the discourses below, that grammatically‐ based theories of narrative progression do not explain the backward movement of time in (4). Moreover, “they are unable to explain why the natural interpretations of [(5a) and (5b)] are different” (Lascarides and Asher 1993:43738). That is, it is surprising on Partee’s analysis outlined above that two eventive sentences in (4) don’t invoke narrative progression. Similarly, it is surprising that the stative sentence The room was pitch dark in (5a) triggers narrative progression, but the same sentence in (5b) does not. (4) Max fell. John pushed him. (5) a. Max opened the door. The room was pitch dark. b. Max switched off the light. The room was pitch dark. An appeal to commonsense reasoning could, however, explain the data above according to Asher and Lascarides. For example, the second sentence in (4) is naturally understood to be an explanation of the first sentence (invoking the so‐called EXPLANATION relation8), while (5a) and (5b) differ in that the first sentence of (5a) naturally invokes BACKGROUND, while the first sentence of (5b) invokes RESULT9, a relation that could be thought of in terms of Humean cause‐effect.10 More recently, Bitner (2008, 2014) has attempted to defend the grammatical view of narrative progression. Based on a wide array of genetically unrelated language with rich morphological marking, she argued that aspectual distinctions are, in fact, responsible for the understood temporal location of an eventuality. Moreover, she suggests that English verbs, which exemplify impoverished morphology, are underspecified for aspectual type, thereby making it extremely difficult to come up with meaningful generalizations about narratives in English. While Bittner does not 7 Kehler (2002) does, ultimately, conclude that the perfect requires the notion of a reference time. This, according to Kehler, is what makes the perfect different from all other tenses. 8 Lascarides and Asher 1993 define EXPLANATION(S0, S1) as: The event described by S1 explains why the event described by S0 happened (perhaps by causing it). 9 Lascarides and Asher 1993 define RESULT(S0, S1) as: The event described by S0 caused the event or state described by S1. 10 (5) could also be argued to exemplify Hobbsian OCCASION. However, Lascarides and Asher (1993) don’t consider this relation to be primitive. More recently, Asher and Lascarides (2003) treated trigger for NARRATION, which is discussed in the next section. OCCASION as a In defense of the reference time 137 characterize the alleged underspecification in examples like (4)(5)11, her position is supported by Caenepeel’s and Moens’ (1994:10) observation that if we assume that EXPLANATION (or commonsense reasoning more generally) triggers backward movement of time for simple past sequences, then it is unclear why (6a) and (7a) are infelicitous, or at least worse than their b‐counterparts: (6) a. ?Everyone laughed. Fred told a joke. b. Fred told a joke. Everyone laughed. (7) a. ?The committee applauded. Niegel announced his promotion. b. Niegel announced his promotion. The committee applauded. Clearly, a salient causal link between someone telling a joke and people laughing and between someone’s promotion being announced and people applauding is evident. Nevertheless, that does not make (6a) and (7a) felicitous. The data above suggests that perhaps Bittner is right and that generalizations about narrative progression should not be based solely on English data.12 To that end, we choose to consider narrative discourses from Russian and French, languages with explicit aspectual marking. Adopting the coherence‐based approach to discourse interpretation advocated by Asher, Hobbs, Kehler, Lascarides and others, the aim of this squib is to provide new evidence in favor of a Partee‐type meaning of the past tense, namely one that encodes a reference time. This aim is achieved as follows. In the next section, we argue that on the one hand, OCCASION is a necessary ingredient to model narrative discourse, but on the other hand, we also need a relation which is defined in purely temporal terms; something like Asher and Lascarides’ (2003) NARRATION. Subsequently, in section 3, we show that the choice between these two relations is governed by various factors, such as the use of aspect and the presence (or lack thereof) of temporal locating adverbials. In section 4, we suggest that this observation warrants a semantic explanation that involves the notion of a reference time, and sketch out some ideas for what the explanation could be like, while remaining neutral about the particular formal implementation. In sum, the methodology assumed in this squib is as follows. We look at narrative discourses of various kinds and consider which rhetorical relations adequately model the inferences that are invoked. Subsequently, we ask what in the grammar is responsible for triggering the rhetorical relations that we observe. After 11 See Kamp, Genabith and Reyle 2011 for a proposal. 12 See Toosarvandani 2014 for more discussion of this point. 138 D. Altshuler and S. Melkonian reaching a natural hypothesis, we conclude that only a particular kind of semantic analysis, namely one that involves the notion of a reference time, could explain why these grammatical forms should trigger the rhetorical relations that we observe. So, in essence, the methodology employed here starts from the discourse level and concludes with questions about form and meaning. We believe that this methodology is essential for explaining intuitions about the temporal location of eventualities. Moreover, by looking at Russian and French, we hope to advance Bittner’s motto: in order to better understand narrative discourse, study languages with rich morphology that encodes aspectual distinctions. 2. OCCASION, NARRATION or Both? In the previous section, we saw how OCCASION plays a vital role in accounting for the contrasting inferences in (1)(2). However, we also noted that narrative discourse is more complex than what is suggested by (1)(2). One complication is that there appear to be discourses that exemplify narrative progression, yet the described events are not understood as being related via a contingency relationship involving changes of states. For example, consider the discourses below, in (8)(10): (8) A man murdered Mitys. Shortly ater, a statue of Mitys fell on the murderer, killing him instantly (Aristotleʹs Poetics, cited in Cumming 2010). (9) “That will be 72 cents”, the grocer said. “AAAAAAAAAAAA!” Jack screamed, looking down to see a busy bee stinging him on the little finger (after Richard Brautigan, Revenge of the Lawn, Sam Cumming, pers. comm.). (10) A wave crashed down on Bob, ruining his new suit. “Iʹll sue you for that,” said Bob. Another wave crashed down on him shortly thereafter, carrying off his bowler hat (Sam Cumming, pers. comm.). In (8), a man is described as killing another man and then being killed by a statue. Since statues are not the sorts of things that react to murders, the second killing cannot be contingent on the first. In (9), we understand that two speech acts—the grocer asking Jack to pay and Jack screaming—happened in temporal succession, but there is no understood contingency between them, e.g. we don’t infer that Jack was screaming at the grocer. Finally, in (10), we infer that a wave crashed after Bob playfully In defense of the reference time 139 threatened the wave. Since waves are not the kind of things that react to threats, once again, there is no understood contingency between the two events. If we conclude that there is no understood contingency between the events described in (8)(10), then it seems far‐fetched to account for the narrative progression in these discourses by appealing to OCCASION. But if not OCCASION, then what?13 One possibility is to adopt Asher and Lascarides’ (2003) NARRATION: (11) NARRATION(S1, S2): the event described by S1 follows an event described by S2 In addition, following Cumming (2013), we will assume that a discourse is coherent if it addresses some question under discussion (QUD).14 For example, the QUD in (8) is something like Who killed who? In (9), the QUD is something like What happened in the store? And in (10), the QUD is something like What happened between Bob and the wave? What all these QUDs have in common is that they are compatible with the temporal order imposed by NARRATION. In contrast, there is no reasonable QUD in (12) below that would also satisfy the temporal order imposed by NARRATION. (12) At five o’clock, my car started and the rain stopped. (Moens and Steedman 1988:22) A reasonable QUD such as What happened at five? would trigger PARALLEL (or perhaps CONTRAST)15, but crucially not NARRATION. As a result, there is no understood narrative progression. 13 According to Jerry Hobbs (pers. comm.), (8) could be analyzed as a kind of parallel since both sentences describe a similar event, namely killing, and the agent and the patient of the two events is the same individual. Perhaps (10) could also be analyzed along these lines. (9), however, does not seem to involve any parallel. 14 This assumption is a departure from Asher and Lascarides’ (2003) proposal that discourses exemplifying NARRATION 15 must also satisfy the so‐called common topic constraint. While there is no consensus as to how PARALLEL and CONTRAST should be defined, most theories of discourse coherence agree that these relations are fundamental. Below, I provide Hobbs’s (1990) definitions. See, e.g. Kehler 2002 and Asher and Lascarides 2003 for different renditions. (i) PARALLEL(S0, S1): Infer p(a1, a2…) from the assertion of S0, and p(b1, b2…) from the assertion of S1, where ai and bi are similar for all i (Hobbs 1990:93). (ii) CONTRAST(S0, S1): a. Infer p(a) from the assertion of S0, and ¬p(b) from the assertion of S1, where a and b are similar. b. Infer p(a) from the assertion of S0, and ¬p(b) from the assertion of S1, where there is some property q such that q(a) and ¬q(b) (Hobbs 1990:99). 140 D. Altshuler and S. Melkonian Similarly, there is no possible QUD in (13) below that would satisfy the temporal order imposed by NARRATION. Unlike (12), however, (13) is incoherent because it arguably does not satisfy any rhetorical relation. (13) #Christ was born no later than 4 B.C. and today Fortuna won the match. In sum, there are narrative discourses that do not exemplify OCCASION. NARRATION seems to be helpful in analyzing these discourses. A question that arises, however, is how exactly QUD is linked to NARRATION and other rhetorical relations. Could it be the case that a sufficient answer to this question would eliminate the need for OCCASION altogether? While no concrete answer to this question has been provided to the best of our knowledge, and none will be provided here, the next section will look at data that is directly relevant. We will first look at discourses suggesting that the Russian imperfective is incompatible with OCCASION. Then, we will see narratives with the Russian imperfective where NARRATION is exemplified. We will take these data as evidence that both relations are linguistically relevant and this will set the stage for us showing that semantic factors contribute to NARRATION, rather than OCCASION, being exemplified by a given discourse. 3. Aspect in Past‐Tensed Discourses In this section we would like to show that both OCCASION and NARRATION are linguistically relevant. To do so, we would like to first consider discourses that describe two events that could, in principle, be related to one another in multiple ways. We show that using a particular aspect in Russian (i.e. perfective vs. imperfective) constrains the possible ways in which the event ordering could actually be understood. Let’s begin by considering a discourse that describes the following two events: An event of entering the castle and an event of reading a brochure about this castle. In principle, we could imagine the following ways in which these events could be related to form a coherent discourse: (a) reading a brochure about a castle sets up the occasion for visiting it, (b) the castle was entered with the knowledge gained from having read the brochure, (c) the castle was entered while reading the brochure, and (d) the entering of the castle set up the occasion to read a brochure about it. (14) Dudkin za‐še‐l v zamok. On čita‐l Dudkin PFV‐go‐PST.3SG into castle brošjuru ob ètom zamke. he read.IPF‐PST.3SG brochure about this castle ‘Dudkin entered the castle. He {had read/was reading} a brochure about this castle.’ In defense of the reference time 141 Interestingly, the use of the imperfective (‘IPF’) čital (‘read’) in (14) is compatible with the situations described in (a)(c), but not in (d). To describe (d), the perfective (ʹPFVʹ) počital is preferred. Let us now turn to the discourse in (15), cited by Altshuler (2012: 86): (15) Gruzinskaja Georgian storona ešče s utra čto zajavila, even from morning announce.PFV‐PST.3SG‐FEM that side peredača uže sostoja‐l‐a‐s’, exchange already take.place.PFV‐PST.3SG‐FEM‐RFL ‘The Georgians said in the morning that the exchange had already taken place,’ rossijskaja storona èto Russian oprovergla‐l‐a. this deny.IPF‐PST.3SG‐FEM side ‘the Russians had denied this…’ (Izvestija, 2002.10.04) Here, we understand that the Russians denied that an exchange had taken place. The question relevant for the purposes here is: When is this denial understood to have taken place relative to the announcement made by the Georgians? World knowledge tells us that denying must take place after the claim that is being denied. However, with the imperfective verb oproverglala (‘deny’), the denial is actually understood to be the general position that had been held by the Russians regardless of what the Georgians claimed. Put differently, the imperfective verb describes a habitual state that is not contingent on the Georgian claim. To describe a denial that is contingent on the Georgian claim, the perfective counterpart, oprovergla, would have to be used. In sum, (14) is a case in which the described events could, in principle, be ordered in every possible way, and (15) is a case in which world knowledge reasoning biases narrative progression. However, in both discourses, narrative progression is not inferred when the imperfective is used. Let us now consider a pair of discourses that force narrative progression. We start with (16), where the use of imperfective would lead to an infelicitous discourse: (16) Musčina vo‐še‐l man v <<White Hart>>. On byl PFV‐go‐PST to White Hart he be.IPF‐PST ‘A man entered the White Hart. He wore a black jacket.’ Bill {OKda‐l/ Bill give.PFV‐PST #dava‐l} give.IPF‐PST ‘Bill gave him a mug of beer.’ (after Kamp and Reyle 1993) emu him bakal pivo. mug beer v černom pidžake in black jacket 142 D. Altshuler and S. Melkonian This discourse describes a man who entered a bar while wearing a black jacket. Subsequently, the discourse describes this man receiving a mug of beer. Given that people receive beer after they enter a bar (rather than before), narrative progression is forced (or heavily biased). Consequently, the imperfective daval (‘gave’) is not possible; the perfective dal must be used instead. An analogous point could be made with respect to the discourse in (17) below: (17) Dva goda spustja, v 1977 gody, amerikanskij geometr two years removed in 1977 year American Robert Konnelli geometrician Robert Connelli po‐stroi‐l pervye primery izgibaemyx, mnogogrannikov PFV‐build‐PST.3SG first examples flexible i tem samym {OKoproverg/ and that same polyhedron #oproverga‐l} disprove.PFV.PST.3SG gipotezu Ejlera. disprove.IPF.PST.3SG hypothesis Euler ‘Two years ago, in 1977, the American geometrician Robert Connelli built the first flexible polyhedron, and thereby disproved Euler’s hypothesis.’ (N.P. Dolbilin, Žemčužiny teorii mnogogrannikov) The phrase i tem samym (‘and thereby’) forces a contingency relationship between the polyhedron having been built by Connelli and the disproval of Euler’s hypothesis. Consequently, the imperfective verb oproverglal (‘disproved’) cannot be used to assert the disproval of Euler’s hypothesis; its perfective counterpart must be used instead. Based on the discourses above, Altshuler (2012) proposes that the Russian imperfective is incompatible with OCCASION. While we agree with this hypothesis, it is important to see why a more general hypothesis fails, namely that the Russian imperfective is incompatible with NARRATION and is only therefore incompatible with OCCASION. Consider (18): (18) a. Osen’ju 1888 godu Ul’janovu bylo fall 1888 year Ul’janov razrešeno vernut’sja v was.IPF allowed Zdesʹ on vposledstvii vstupil v odin iz here he subsequently joined.PFV in one return Kazanʹ. to Kazan marksistskix kružkov... from Marxist circles ‘In the fall of 1888, Ul’yanov was allowed to return to Kazan. Here he subsequently joined one of the Marxist circles...’ b. V 1924 godu N.K. Krupskaja pisala in 1924 year N.K. Krupskaja wrote.IPF in “Pravda” “Plexanova, Vladimir Il’ič Plexanov v “Pravde”: ljubil strastno...” Vladimir Ilich loved.IPF passionatly ‘In 1924, N.K. Krupskaja wrote in “Pravda”: “Vladimir Ilich loved Plexanov passionately.’ In defense of the reference time 143 The discourse above is a snippet from a long narrative about Lenin’s life. (18a) describes two subsequent events starting in 1888: Lenin’s permission to return to Kazan and his subsequent joining one of the Marxist circles. The narrative then goes on to describe several other events in Lenin’s life, leading to the end of a paragraph. The new paragraph opens with (18b) above, which describes an event that took place 36 years later, in 1924, when Lenin’s wife, Krupskaja, wrote about Lenin in the newspaper Pravda. What’s crucial for our purposes is the fact that the imperfective past tense verb pisala (‘wrote’) is used in a context where there is clear narrative progression. Given the temporal locating adverbial that begins (18b), Krupskaja’s writing about Lenin must be understood as occurring after the events described in (18a). Several comments are in order. To begin with, note that the writing event described in (18b) is not understood to be contingent on the events described in (18a). Therefore, saying that OCCASION relates Krupskaja’s writing to a previously mentioned event would be misleading. Rather, something like NARRATION is exemplified because the imperfective clause, along with the adverbial that modifies it, is simply used to assert that an event (i.e. writing) took place in 1924, ater the other events. The QUD is something like What happened in Lenin’s life?16 As noted in the previous section, much more needs to be said about how QUD is linked to the choice of a retorical relation, and NARRATION in particular.17 For our purposes, what is crucial to take away is three‐fold: (a) past imperfectives are used in narrative discourse18, but (b) only when the narrative involves something like NARRATION rather than OCCASION and (c) it also involves a temporal locating adverbial. Based on (a) and (b), we conclude that the Russian imperfective is allergic to OCCASION, but welcomes NARRATION. This conclusion is supported by the fact that (17) and (18) are infelicitous with the imperfective precisely because contingency (which is the hallmark trigger of OCCASION) is forced. We now turn to the observation in (c) and its relation to (a) and (b). In the next section, we argue that the notion of a reference time allows for a natural explanation for why the imperfective is allergic to OCCASION but not NARRATION. 16 Based on this QUD, it’s tempting to argue that a parallel is being established. This, however, seems unlikely given the definition of PARALLEL in fn. 15. In particular, note that the agent of the writing event described in (18b) is not Lenin but his wife. 17 With respect to (18), one relevant observation is that the writing event described in (18b) is not at‐issue; the writing by Krupskaja serves as evidence for what Lenin’s feelings towards Plexanov are like and these feelings are at issue. See Simons 2007 for an analysis of propositional atitude verbs as evidentials. 18 This observation was noted by Chvany (1985), though her discussion centers around the imperfective present, which is the form used in Russian for the historical present. This squib has nothing to say about this usage of the imperfective. 144 D. Altshuler and S. Melkonian 4. In Defense of the Reference Time In what follows, we would like to advance the hypothesis that temporal locating adverbials license the use of the Russian imperfective in particular narrative progression contexts, that is, those discourses that exemplify NARRATION. This idea is, in part, motivated by the observation that it would be odd to utter (18b) above as a continuation of (18a) without the temporal locating adverbial V 1924 godu ‘In 1924’. In addition, there is evidence for the hypothesis from French. It has often been observed that the imparfait is odd in narrative contexts (see, e.g., Kamp and Rohrer 1983). For example, the imparfait pénétrait ‘entered’ in (19) below is infelicitous due to the narrative progression. (19) #Maigret reprit la petite auto noire et il pénétrait dans la brasserie de la place de la République. ‘Maigret once more took the little black car and he entered the café at the Republic square.’ Interestingly, inserting a temporal locating adverbial, quelques minutes plus tard ‘a few minutes later’, in the imparfait clause of (19) renders the discourse perfectly natural: (20) Maigret reprit la petite auto noire et, quelques minutes plus tard, il pénétrait dans la brasserie de la place de la République. ‘Maigret once more took the little black car and a few minutes later he entered the café at the Republic square’ (Grønn 2008). The use of the imparfait in (20) has been called “the imparfait narrative” (or “imparfait de rupture”) and it has often been noted that it requires a temporal locating adverbial (often in fronted position). For example, Tasmowski (1985:6) claims: “the development of the IR (= “imparfait de rupture”) crucially depends on a temporal adverb at the beginning of the sentence, this adverb being in charge of introducing a new temporal moment in the text”. Tasmowski cites Klum (1961), who compared how many times in a corpus the temporal locating adverbials appeared with the imparfait narrative. Klum suggested that in presence of temporal locating adverbials, the imparfait is often preferred to the passé simple. Kamp and Rohrer (1983:258) provide the beginnings of an explanation for this observation when they write: “[…] Indeed when an imparfait sentence is to refer to a time other than the reference point it must contain a temporal adverbial (which may take the form of a single adverbial, a prepositional phrase or a subordinate clause) to indicate that time.” One way to In defense of the reference time 145 interpret this idea is that the imparfait naturally occurs without the presence of temporal adverbials, for example, when it’s embedded in a discourse and gets its reference time from surrounding context. In the presence of temporal adverbials, however, it is that adverbial that provides the temporal anchor, that is, the reference time required by the imparfait. This invites a compositional semantic analysis of the interaction between tense, aspect and temporal adverbials. Crucially, such an analysis must explain how this interaction constrains the structure of a discourse, that is, whether OCCASION or NARRATION is inferred. In what follows, we would like to offer a suggestion for what this analysis could be like, while remaining neutral about the formal implementation. Crucial in this suggestion will be the notion of a reference time. In this way, we will come full circle. Recall that we started with Partee’s analysis, which appealed to reference times to explain narrative progression. Subsequently, we saw that coherence‐ based analyses can explain the same data without positing reference times. And now, the suggestion will be that past oriented narratives involving the imperfective aspect show that we need reference times to explain the interaction between tense, aspect and temporal adverbials and, in particular, how this interaction impacts the discourse structure. So, in essence, Partee’s analysis was on the right track, but for reasons that were not considered until now. Here is the crux of our suggestion about how to model temporal meanings. Aspectual meaning is responsible for constraining the temporal location of a described eventuality relative to a time argument that is saturated by the tense. If there are no temporal locating adverbials, the saturated time is left as a free variable; it functions like the aforementioned reference time used to define the past tense by Partee (1984; see discussion in §1). It could lead us to infer OCCASION if the particular relation between the reference time and eventuality time encoded by the aspect is compatible with OCCASION. With respect to the Russian imperfective, what we would like to say, then, is that its semantics imposes a relation between the reference time and eventuality time that is in direct conflict with OCCASION (as defined in §1).19 19 Altshuler (2012) shows how this can be formally implemented. He proposes that the Russian imperfective denotes a function from a set of events denoted by the VP that it combines with to a set of VP‐event parts. The consequent states of these event parts are related to a temporal coordinate that is specified by the discourse context; it functions similar to Partee’s reference time discussed at the outset with one key difference: the reference time corresponds to the run time of salient consequent state previously introduced into the discourse context (cf. Webber 1988; Bitner 2008, 2014). Now, recall from §1 that OCCASION is defined in terms of salient (final and initial) states. Altshuler’s idea is that the semantics of the Russian imperfective rules out OCCASION because the particular relation that it imposes (between the consequent state that it describes and the salient consequent state introduced into the discourse context) is in direct conflict with the state ordering in the definition of OCCASION. 146 D. Altshuler and S. Melkonian Now, if there is a temporal locating adverbial present, then the time left free by the tense—which can then function as a reference time—is existentially bound by the adverbial. In other words, temporal locating adverbials are devices of anaphoric closure because they prevent the otherwise free time variable from getting its value from the context. One question that comes up for this toy analysis is why the presence of temporal adverbials often leads us to infer NARRATION rather than OCCASION in narrative discourse. One possible hypothesis, pursued in recent work (Altshuler and Melkonian 2014) is that temporal adverbials introduce a reference time which establishes a temporal gap between two events e and e’ that is inconsistent with e being contingent on e’ (see Dowty 1986 for a related idea). Therefore, OCCASION is often not exemplified in narrative segments that contain temporal adverbials. This idea is nicely illustrated by the aforementioned discourse in (18). Recall that the temporal adverbial in (18b) is used to open the beginning of a paragraph. We believe that the temporal adverbial is especially well suited here because it allows Krupskaja’s writing in “Pravda” to stand after a large enough temporal gap such that we could further infer that there were intermediary events not mentioned in the discourse that Krupskaja’s writing in “Pravda” was contingent upon. As noted above, we suggest to model this idea as anaphoric closure by the adverbial. Without this anaphoric closure, the hearer would attempt to establish a contingency relationship directly, by picking out reference times established in the local context. However, such a relationship is not warranted given the facts described in the discourse, which is why (18b) is odd without an adverbial. We end this section by noting that there are temporal adverbials which are not devices of anaphoric closure. That is, they do not provide a reference time because they are “discourse transparent”. As illustrated by Altshuler (2014), that same day is an adverbial of this kind. Altshuler observed that adding this adverbial to a narrative discourse does not alter the narrative progression—the adverbial is, as it were, ‘transparent to the progress’. Consider, for example, the discourse in (21) below: (21) a. On May 12, 1984, Sue gave Fido a bath and cleaned our house. b. That same day, my wife hired her and gave her a check for one month in advance. Specifying that the hiring took place on the same day as the house cleaning does not block the additional inference that the hiring took place after the house cleaning. Crucially, (21) makes no claim about the exact temporal distance there is between the In defense of the reference time 147 house cleaning and the hiring. The only claim made is that the temporal distance is relatively short (see fn. 1). The discourse transparency that we see in (21) can be replicated with related temporal adverbials and in non‐narrative discourse contexts. Consider (22b) below, which differs from (21b) in containing the pluperfect (rather than the simple past). As a result, we understand that the hiring took place prior to the house cleaning. Crucially, this inference is present regardless of whether that same month and that very month are used. That is, these adverbials are transparent to the narrative regression. (22) a. Some months ago, Sue gave Fido a bath and cleaned our house. b. (That same month/That very month) my wife had hired her and had given her a check for one month in advance. In (23) below, there is no order that the events in (23b,c) are understood to have occurred in (though they are understood to precede the event described in (23a)). Such is the case whether or not that same week or that very week are present in (23c). That is, specifying that the events described in (23b,c) all took place during the same week does not provide any new information that is not already inferred without that same week or that very week. (23) a. Bill will move next week. b. Last week, his house burned down. c. (That same week/That very week) he divorced Sue and he was fired. In sum, what distinguishes an adverbial like that same day from an adverbial like in 1924 is that the time described by the former always preserves the temporal structure of a discourse, that is, it is transparent to the independent rules that account for the temporal structure of a given discourse. Following Altshuler 2014, this can be modeled on the toy analysis by saying that a distinguishing property between temporal adverbials is whether or not they introduce a reference time. Temporal adverbials like in 1924 clearly do, and this impacts the structure of a discourse, mediated by tense and aspect. Adverbials like that same day, however, do not introduce a reference time, thereby preserving the discourse structure that has been constrained by the tense and aspect. A prediction of this analysis is that in contingency contexts that trigger OCCASION, the presence of an adverb like that same day should not license the use of the imperfective (unlike the presence of an adverb like in 1924). This prediction is borne 148 D. Altshuler and S. Melkonian out in the Lenin discourse in (18). If we were to replace v 1924 godu in (18b) with v tot že samyj den’ (‘that same day’), the discourse would be odd; the perfective form of pisala ‘write’, namely napisala, would be preferred. 5. Conclusion We hope that the toy analysis developed in the previous section provides a glimpse of what a formal implementation of the data considered in this squib could be like. The contribution should be thought of as providing some constraints on whatever formal implementation is chosen. We acknowledge that we have not showed how, exactly, grammatical components such as tense, aspect and temporal adverbials interact with pragmatic principles such as OCCASION or NARRATION. Rather, we have provided an argument, involving empirical evidence, that there must be such an interaction and argued that the notion of reference time is necessary to make sense of it. References Altshuler, Daniel. 2012. Aspectual meaning meets discourse coherence: A look at the Russian imperfective. Journal of Semantics 29:39108. Altshuler, Daniel. 2014. A typology of partitive aspectual operators. Natural Language and Linguistic Theory 32(3):735775. Altshuler, Daniel, and Susanna Melkonian. 2014. Taxis, contingency and imperfectivity. Presented at the International Conference “The Linguistic Heritage and Development of Yu.S. Maslovʹs Ideas”, St. Petersburg, Russia. Asher, Nicholas, and Alex Lascarides. 2003. Logics of conversation. Cambridge: Cambridge University Press. Bitner, Maria. 2008. Aspectual universals of temporal anaphora. In Theoretical and crosslinguistic approaches to the semantics of aspect, ed. by Susan Rothstein, 34985. Amsterdam: John Benjamins. Bitner, Maria. 2014. Temporality: Universals and variation. Malden, MA: Wiley‐Blackwell. Caenepeel, Mimo, and Marc Moens. 1994. Temporal structure and discourse structure. In Tense and aspect in discourse, ed. by Co Vet and Carl Vetters, 520. Berlin: Mouton de Gruyter. Chvany, Catherine. 1985. Backgrounded perfectives and plot‐line imperfectives. In Issues in Russian morphosyntax, ed. by Michael S. Flier and Richard D. Brecht, 247–73. Columbus: Slavica. Cumming, Sam. 2010. Narrative and point of view. Ms., University of California, Los Angeles. Cumming, Sam. 2013. The atentional foundations of coherence. Ms., University of California, Los Angeles. Accessed on December 14, 2014 at: htps://dl.dropboxusercontent.com/u/31660624/atention‐coherence.pdf Dowty, David. 1986. The effects of aspectual class on the temporal structure of discourse: Semantics or pragmatics?. Linguistics and Philosophy 9:3761. Grønn, Atle. 2008. Imperfectivity and complete events. In Interdependence of diachronic and synchronic analyses, ed. by Folke Josephson and Ingmar Söhrman, 149165. Amsterdam: John Benjamins. In defense of the reference time 149 Hinrichs, Erhard. 1986. Temporal anaphora in discourses of English. Linguistics and Philosophy 9:63–82. Hobbs, Jerry. 1985. On the coherence and structure of discourse. Report No. CSLI‐85‐37, Center for the Study of Language and Information, Stanford University. Available at: http://www.isi.edu/~hobbs/discourse‐reference/discourse‐references.html. Hobbs, Jerry. 1990. Literature and cognition. Stanford: CSLI Publications. Hume, David. 1748/2008. An enquiry concerning human understanding: And selections from a treatise of human nature. New York: Cosimo Classics. Kamp, Hans. 1981. A theory of truth and semantic representation. In Formal methods in the study of language, ed. by Jeroen Groenendijk, Theo Janssen and Martin Stokhof, 277–322. Amsterdam: Mathematical Centrum Tracts. Kamp, Hans. 1999/2013. Deixis in discourse: Reichenbach on temporal reference. 1999 Reichenbach Lecture at UCLA. Published 2013 in Meaning and the dynamics of interpretation: Selected papers of Hans Kamp, ed. by Klaus von Heusiger and Alice ter Meulen, 105–159. Leiden: Brill. Kamp, Hans, and Christian Rohrer. 1983. Tense in texts. In Meaning, use and interpretation of language, ed. by Rainer Bäuerle, Cristoph Schwarze and Arnim von Stechow, 250269. Berlin: De Gruyter. Kamp, Hans, Josef van Genabith, and Uwe Reyle. 2011. Discourse Representation Theory. In Handbook of philosophical logic, ed. by Dov Gabbay and Franz Guenthner, 125–394. Dordrecht: Kluwer. Klum, Arne. 1961. Verbe et adverbe: Étude sur le système verbal indicatif et sur le système de certain adverbes de temps à la lumière des relations verbo‐adverbiales dans la prose du français contemporain. Stockholm: Almqvist and Wiksell. Kehler, Andrew. 2002. Coherence, reference and the theory of grammar. Stanford: CSLI Publications. Kehler, Andrew. 2006. Discourse coherence. In The handbook of pragmatics, ed. by Laurence R. Horn and Gregory Ward, 241266. Oxford: Blackwell. Lascarides, Alex, and Nicholas Asher. 1993. Temporal interpretation, discourse relations and commonsense entailment. Linguistics and Philosophy 16:437493. Moens, Marc, and Mark Steedman. 1988. Temporal ontology and temporal reference. Computational Linguistics 14(2):15–28. Partee, Barbara H. 1973. Some structural analogies between tenses and pronouns in English. Journal of Philosophy 18:601–609. Partee, Barbara H. 1984. Nominal and temporal anaphora. Linguistics and Philosophy 7:243–286. Reichenbach, Hans. 1947. Elements of symbolic logic. New York: Macmillan. Rohde, Hannah, and William Horton. 2014. Anticipatory looks reveal expectations about discourse relations. Cognition 132:667691. Simons, Mandy. 2007. Observations on embedding verbs, evidentiality, and presupposition. Lingua 117:10341056. Tasmowski‐De Ryck, Liliane. 1985. L’imparfait avec et sans rupture. Langue française 67:5977. Toosarvandani, Maziar. 2014. The temporal interpretation of clause chaining in Northern Paiute. Ms., University of California, Santa Cruz. Accessed on December 14, 2014 at: http://people.ucsc.edu/~mtoosarv/papers/temporal‐interpretation‐clause‐chaining.pdf Webber, Bonnie. 1988. Tense as discourse anaphor. Computational Linguistics 14:61–73. Wolf, Florian, and Edward Gibson. 2005. Representing discourse coherence: A corpus‐based study. Computational Linguistics 30:249287. To cite this article: Altshuler, Daniel, and Susanna Melkonian. 2014. In defense of the reference time. Semantics‐ Syntax Interface 1(2):133–149. CORRIGENDUM Semantics‐Syntax Interface, Volume1, Number 1, Spring 2014  A Brief History of the Syntax‐Semantics Interface in Western Formal Linguistics Barbara H. Partee Page 12, section 8, lines 8−9, the correct form is: ... that it was partly a reaction to a perceived attack on the autonomy of syntax, even though syntax is descriptively autonomous in Montague grammar. Semantics‐Syntax Interface, Volume1, Number 1, Spring 2014  Persian Complex Predicates: How Compositional Are They? Pollet Samvelian and Pegah Faghiri Page 63, example (34b), the correct form is: (34) b. divār‐hā wall‐PL pārsāl rang šo‐d‐and last‐year paint become‐PST‐3PL ‘The walls were painted last year.’ Semantics‐Syntax Interface, Volume 1, Number 2, Fall 2014 150 © 2014 by University of Tehran http://semantics‐syntax.ut.ac.ir Semantics‐Syntax Interface Volume 1, Number 2 Fall 2014 CONTENT Original Articles Diego Gabriel Krivochen (Anti‐)Causativity and the Morpho‐Phonology‐Semantics Tension 82 Short Contributions Sumiyo Nishiguchi Mirative Past in Japanese 118 Daniel Altshuler and Susanna Melkonian In Defense of the Reference Time 133 Corrigendum 150 ‫ﺷﻨﺎﺳﻨﺎﻣﺔ ﻧﺸﺮﻳﻪ ‪Semantics‐Syntax Interface‬‬ ‫ﺻﺎﺣﺐ اﻣﺘﻴﺎز‪ :‬داﻧﺸﻜﺪة ادﺑﻴ‪‬ﺎت و ﻋﻠﻮم اﻧﺴﺎﻧﻲ داﻧﺸﮕﺎه ﺗﻬﺮان‬ ‫ﻣﺪﻳﺮ ﻣﺴﺌﻮل‪ :‬ﻏﻼﻣﺤﺴﻴﻦ ﻛﺮﻳﻤﻲدوﺳﺘﺎن )اﺳﺘﺎد زﺑﺎنﺷﻨﺎﺳﻲ‪ ،‬داﻧﺸﮕﺎه ﺗﻬﺮان(‬ ‫ﺳﺮدﺑﻴﺮ‪ :‬ﻧﮕﻴﻦ اﻳﻠﺨﺎﻧﻲﭘﻮر )داﻧﺸﺠﻮي دﻛﺘﺮي ﺗﺨﺼ‪‬ﺼﻲ زﺑﺎنﺷﻨﺎﺳﻲ‪ ،‬داﻧﺸﮕﺎه ﺗﻬﺮان(‬ ‫داﻧﺸﮕﺎه ﺻﺎدرﻛﻨﻨﺪة ﻣﺠﻮ‪‬ز‪ :‬داﻧﺸﮕﺎه ﺗﻬﺮان‬ ‫زﻣﻴﻨﺔ اﻧﺘﺸﺎر‪ :‬ﻋﻠﻤﻲ‬ ‫ﺗﺮﺗﻴﺐ اﻧﺘﺸﺎر‪ :‬دوﻓﺼﻠﻨﺎﻣﻪ‬ ‫ﮔﺴﺘﺮة ﺗﻮزﻳﻊ‪ :‬داﻧﺸﮕﺎه ﺗﻬﺮان‬ ‫ﺗﺎرﻳﺦ اﻧﺘﺸﺎر‪ :‬ﭘﺎﻳﻴﺰ ‪1393‬‬