Information Structuring Begins with the Numeration

 Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 Information Structuring Begins with the Numeration 1 Enoch O. Aboh [email protected] Universiteit van Amsterdam Abstract: This paper argues for the strong minimalist view that information structure starts in the numeration in the form of discourse-related lexical items which drive the derivation. Though sometimes covert, these discourserelated particles encode features, such as, Interrogative force, Topic, or Focus and display very specific structural behavior indicating that they project in syntax. The analysis further demonstrates that these features are comparable to other optional formal features (e.g., Case, -features) that are added arbitrarily when the lexical item enters the numeration. Keywords: Information structure, numeration, lexicon, derivation. Resumen: Este artículo defiende la visión minimalista fuerte de que la estructura informativa comienza en la numeración en forma de piezas léxicas discursivas que dirigen la derivación. Aunque algunas veces no explícitas, estas partículas discursivas codifican rasgos tales como Fuerza interrogativa, Tópico o Foco y muestran un comportamiento estructural muy específico que indica que se proyectan en la sintaxis. El análisis también demuestra que estos rasgos son compatibles con otros rasgos formales opcionales (por ejemplo, Caso, rasgos-) que son añadidos arbitrariamente cuando las piezas léxicas se incorporan entran a la numeración. Palabras clave: Estructura informativa, numeración, lexicón, derivación. Resumo: Este artigo defende a forte visão minimalista de que a estrutura informacional começa com a numeração sob a forma de itens lexicais discursivos que conduzem a derivação. Embora por vezes não explícitas, 1 This paper was written in the context of my NWO-Vidi grant on the interface between information structure and syntax. The paper has been presented in various forms at the MPI conference on Information Structure in Adult and Child Language, Nijmegen, at the Syntax Circle, Meertens Institute, at the Syntax seminar, University of Groningen, at the Syntax Colloquium, University of Potsdam and at the Optimal Communication Colloquium series, Radboud University Nijmegen. Sections of this paper were also discussed during my MIT undergrade course in spring 2008 and my 2010 LOT courses at Radboud University Nijmegen. I thank all the participants of these meetings for their reactions. I’m also grateful to Boban Arsenijević, Anna Cardinaletti, Michel DeGraff, Marina Dyakonova, Irene Heim, Luigi Rizzi, Jenneke van der Wal, and Jan Wouter Zwart for their comments.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 13 Enoch O. Aboh estas partículas discursivas codificam traços, como, por exemplo, Força Interrogativa, Tópico ou Foco, e apresentam um comportamento estrutural muito específico indicando que se projectam na sintaxe. A análise demonstra ainda que tais traços são compatíveis com outros traços formais opcionais (por exemplo, Caso, Traços ) que são adicionados arbitrariamente quando o item lexical entra na numeração. Palabras clave: Estructura informacional, numeração, lexicon, derivação. 1. Introduction In this paper, I argue for what seems to me the most natural approach to information structure, and, which in the end, appears an inescapable conclusion if one adopts the strict minimalist approach to the study of language (Chomsky 1995, 1999, 2005). In current minimalist theory, a linguistic expression is a pair (π, ) consisting of a PF representation (i.e., sound) associated to an LF representation (i.e., a meaning). Under this view, the role of the computational system CHL is to map some array of lexical choices to the pair (π, ). The array of lexical choices is referred to as the numeration (N), which Chomsky (1995: 225) defines as: (1) ‚a set of pairs (LI, i), where LI is an item of the lexicon and i is its index, understood to be the number of times that LI is selected‛. Each time a lexical item is selected from the numeration (N), its index is reduced by one so that the converging derivation (i.e., the one that forms a linguistic expression) is the one for which N is reduced to zero. Given this description, it seems obvious that there should be nothing in the pair (π, ), for instance with regard to its basic (compositional) semantics, that is not already part of the numeration (N). As often assumed in the literature without further demonstration, this would imply that additional discourse information (e.g., topic and focus) that discourse participants assign to a linguistic expression, and which is not part of the numeration, is invisible to CHL. Therefore, adding these features as the derivation proceeds would violate the inclusiveness condition (Chomsky 1995). While one can live a happy life with this conclusion, it is still an empirical fact that, in many languages of the world, notions of information structure such as Topic, Focus, and Interrogative force are determined by lexical choices that are manipulated by CHL in the course of various syntactic operations (e.g., feature matching, displacement). In these languages therefore, the numeration N of a sentence containing a Topic, a Focus or an Interrogative expression (π, ) must include Topic, Focus, or Interrogative lexical choices. This leads me to the formulation in (2).  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 14 Information Structuring begins with the Numeration (2) A numeration N pre-determines the Information Structure of a linguistic expression. 2 If we take languages where lexical items encode information structure non-ambiguously not to be too exotic or dramatically different from English, generative comparative syntax teaches us that (2) should generalize to all languages in a uniformitarian way (see Chomsky 1999: 2 for discussion of the uniformity principle). Languages therefore only vary with regard to the various choices they make in implementing (2). 3 Taking this stance further implies that core syntax embeds properties of Information Structure. In terms of Rizzi (1997) and much related work, this means that core syntax must involve information sensitive functional projections (e.g., TopP, FocP, InterP). While not revolutionary, this controversial position must be motivated both empirically and theoretically. This paper tries to address both issues on minimalist grounds. Section 2 briefly introduces the traditional debate on the interaction between Information Structure and syntax and concludes that there is a certain bias in the literature in considering Topic and Focus as pragmatic notions, while other comparable expressions (e.g., Interrogative) are seen as proper formal features 2 In this paper, I only consider Interrogative, Topic, and Focus, the cornerstone of Information Structure, which involves non-ambiguous morphological expressions correlating with distinct semantic properties. 3 The alert reader may wonder why we should extend the formulation in (2) to all languages given that it is apparently empirically motivated for certain languages only. Therefore in languages where there are no distinct markers for topic, focus, and question (e.g., Germanic, Romance), one may assume that no corresponding functional structure is needed for these languages. Anticipating the discussion in section 5 where I propose that topic and focus features are optional formal features comparable to Case, I would like to stress that decades of comparative syntax should guide our steps here. In his (1977) personal letter to Lasnik and Chomsky, Vergnaud made the crucial point that even languages with virtually no case morphology pattern with case morphology languages with regard to the distributive properties of cased marked NPs. This observation later on inspired Chomsky’s theory of abstract case that is now commonly accepted within generative grammar (but see Marantz 2000 and much related work for an alternative). Yet, if we consider languages like Gungbe, to which this theory extends, they show no case morphology at all on full DPs (i). (i) a. Àjàkà ù gbàdó b. Àsé ù àjàkà mouse eat corn cat eat mouse ‘A/the mouse ate corn’ ‘A/the cat ate the/a mouse’ The only traces of case that we find are confined to a corner of the pronominal system (i.e., 1sg and 2sg as in ùn 1sg-Nom versus mì 1sg-Acc, or à 2sg-Nom versus wè 2sg-Acc, see Aboh 2004, chapter 4.). Jamaican creole and several creoles exhibit no case distinction at all, and the same forms and set of pronouns occur both in nominative and accusative positions. Despite their complete lack of case morphology, the theory of abstract case naturally extends to these languages as well since they manifest similar set of restrictions on case marked NPs. Following this tradition therefore, I assume that it is just as natural to extend the theory of information structure proposed here to even languages with little (or no) morphology for expressing information structure.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 15 Enoch O. Aboh encoded in syntax (i.e., in C, Chomsky 1995: 240). This paves the way for sections 3 and 4 where I present strong empirical evidence for (2) at the CP and DP level, respectively. Section 5 discusses the implications of (2) with regard to the Topic feature and the Focus feature as proper formal features comparable to tense, Case, and -features. Section 6 concludes the paper 2. Information Structure and Syntax It is a well-known fact that Information Structure affects syntax in some particular ways, though the interaction between these two modules is still not well understood. 2.1. The Debate Aside from cleft constructions, which I’m not discussing here, examples that one often finds in the literature concern Left or Right Dislocations (3a-b) or Focus movement constructions (3c). In addition to word order, these sentences have specific intonation patterns that distinguish them from the ‘neutral’ statement in (4). (3) a. John, I like him very much [LD: Topic reading] b. I like him very much, John [RD: Topic reading] c. John I like very much [LD: Focus reading] (4) I like John very much [Basic order: neutral or default new information focus] If we ignore, for the moment, the various movement operations (i.e., the so-called surface effect) that may have happened in (3a) to (3c), it appears that the numeration leading to the derivations in (3a-c) and (4) is virtually the same and we predict it should lead to roughly the same meaning in all contexts, namely: [X likes Y very much]. Under this view, the fact that the DP John may front to check some EPP feature under C follows from general UG principles and is not intrinsic to the lexical entry John (Chomsky 1995: 236). Given this description, the fact that, in actual discourse, the sentences in (3a-c) acquire additional meaning in relation to the displaced categories (e.g., topic, focus) must be explained, but the process producing this discourse-related meaning cannot be directly ascribed to syntax (i.e., the computation that produces the phrase marker). In this regard, Chomsky (1995: 220) suggests that the observed surface effects (e.g., topic-focus articulation) ‚seem to involve some additional level or levels internal to the phonology component, postmorphology but prephonetic, accessed at the interface along with PF and LF‛. Put differently, such effects are extraneous to aspects of the computation that are imposed by the numeration (i.e., the N computation). Similar approaches are found in recent work by several researchers (e.g., Zubizaretta 1998, Szendroi 2001, Fanselow 2006, 2007)  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 16 Information Structuring begins with the Numeration who assume that constituent rearrangements as in (3) and (4) are prosodicallydriven and therefore differ from displacement rule triggered by checking operations. Under such a view, information structure and syntax interact, but indirectly (via PF and presumably LF). 2.2. On the inclusiveness condition Indeed, when we consider the examples under (3), there appears to be no clear (context-invariant) property of the alleged Topic and Focus features driving the fronting operations which in turn correlate with specific meanings (e.g., topic vs. focus) acquired by the displaced phrase. As often suggested in the literature, such displacements correlate with stress assignment in Germanic and Romance rather than the assignment of topic and focus features per se (Zubizarreta 1998, Szendroi 2001, Samek-Lodovici 2006). The inconclusive status of examples such as those in (3) in Germanic and Romance languages reinforces the general view that crucial notions of Information Structure, such as, Topic and Focus are not part of syntax but must be added to the linguistic expression once computed by CHL. The wisdom is to claim that introducing these notions in syntax violates Chomsky’s (1995: 228) inclusiveness condition: (5) Given the numeration N, CHL computes until it forms a derivation that converges at PF and LF *…+ A ‚perfect language‛ should meet the condition of inclusiveness: any structure formed by the computation *…+ is constituted of elements already present in the lexical items selected for N; no new objects are added in the course of computation apart from rearrangements of lexical properties. 2.2.1. On question-answer pairs As a way of evaluating the inclusiveness condition, let us consider question-answer pairs, which typically involve focus constructions. Adopting (5) would suggest that question (7) derives from the numeration in (6). (6) C, T, who, did, love, John (7) [CP who [C did [TP John [T did [VP John love who]]]]] The idea here is that interrogative force, the so-called wh-feature, is a property of C, which in English, forces wh-movement of the interrogative whphrase to [spec CP] and I-to-C movement, as a requirement of clause-typing (Cheng 1991). It is important to observe at this stage already that the lexicon of English contains the functional category C endowed with the feature interrogative (i.e., Cwh). Given this state of affairs, it is admissible (and even widely accepted in the literature) that the lexicon contains certain discourserelated (or informationstructure-sensitive) functional items that project in syntax. This is the first crack in the common argumentation rejecting the idea of  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 17 Enoch O. Aboh any direct interaction between information structure and syntax. 4 As the above description shows, the correct characterization is that certain information sensitive features (e.g., interrogative force) are directly encoded in syntax. Given this state of affairs, the next question to ask is what is so unique to interrogative feature that necessitates a different treatment as opposed to topic and focus features? Suppose that the question in (7) ‘naturally’ requires the answer in (8b), with new information focus on Mary, though the numeration (8a) contains no functional focus category that may combine with Mary. (8) a. C, T, ed, love, John, Mary b. John loved Mary[New Information Focus] Some questions that immediately come to one’s mind are: (i) why is (7) a product of syntax to the exclusion of the focus structure (8b) that it requires? (ii) Why must a natural answer to a (wh)question always contain a focused expression? (iii) What triggers the matching between question and focus as observed in (6-7)? (iv) Where does the (extra) focus feature in (8b) come from if it is not present in the numeration? This paper tries to answer these questions by considering questions and answers in parallel. The interaction between (7) and (8) suggests that even though speakers have the freedom of choice with regard to which linguistic expression to use in a particular context or discourse, the form of this expression is a product of syntax that directly relates to the Numeration (see also Zwart 1998). Assuming that discourse participants are collaborative, we can make the reasonable hypothesis that the trigger of the focus feature in a felicitous reply represents an attempt of the speaker to provide a matching expression to the question. This recalls Zubizaretta’s (1998: 4) ordered Assertion Structure (AS) where, ‚in the case of question-answer pairs, the presupposition provided by the context question is part of the AS of the answer statement *…+ The focuspresupposition structure of such statements is represented in terms of two ordered assertions. The first assertion (A1) is the existential presupposition provided by the context question. The second assertion (A2) is an equative relation between a definite variable (the restriction of which is the presupposition provided by the context question) and a value.‛ 4 Another feature that is commonly assumed to project in syntax though it primarily relates to discourse (and therefore information structure) is definiteness/specificity as encoded by (in)definite articles (Abney 1987, Szabolcsi 1987, 1994, Longobardi 1994). Given the close relation between definiteness/specificity and topicality (see following discussion and section 4.1.) it is not clear to me why one is seen as a legible formal feature expressed by the nominal periphery D, while the other is banned from the clausal periphery.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 18 Information Structuring begins with the Numeration In looking at question-answer pairs this way, we realize that the structure of the answer (i.e., with regard to clause-typing and focus) is not completely orthogonal to that of the context question. That the matching answer contains a focused expression seems to be a requirement of the question operator in the question. In terms of Chomsky (1971), this means that questionanswer pairs are structured such that the question introduces a variable to which focus assigns a value. Accordingly, focus assignment (which in some languages, e.g., Gungbe, require constituent displacement) satisfies a syntactic requirement generated in the question through clause-typing and wh-licensing. Treating questions and answers this way, appears compatible with the format of question-answer pairs in Mandarin Chinese, where we observe a more transparent correlation between the form of a question and that of the answer. In their discussion of word order in Mandarin Chinese, Li and Thompson (1975) gave the contrast in (9) to show that the sentence in (9b), though a perfect syntactic choice, ‚is extremely unnatural as an answer to (9a) *MC, Li and Thompson 1975: 173+.‛ (9) a. Yàoshi ne ? key INTER ‘What about the key(s)?’ b.??W wàng le yàoshi le I forget ASP key ASP ‘I forgot the key(s)’ According to Li and Thompson, (9b) is strange because in the particular context of question (9a), the noun phrase yàoshi ‘the key(s)’ is forced to be interpreted as discourse anaphoric, and therefore as a topic. The authors further give a number of alternatives that satisfy this requirement and represent felicitous answers for (9a). These are given under (10). (10) a. Yàoshi, wo wàng le [Topicalization] Key(s) I forget ASP ‘As for the key(s) I forgot it/them’ b. Wo bă yàoshi wàng le [The bă-marked DP is discourse anaphoric] I ba key forget ASP ‘I forgot the keys’ c. Wo wàng le [Topic drop] I forget ASP ‘I forgot them’ The different strategies in (10) indicate that the felicitous answer to (9a) requires the noun phrase yàoshi ‘the key(s)’ to be syntactically marked as a topic, hence its sentence-initial position in (10a). In (10b) we have a bă-construction about which Li and Thompson (1981) report that a noun so marked denotes ‚something about which the speaker believes the hearer knows‛. As I suggest  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 19 Enoch O. Aboh later, this recalls Prince’s (1981) notion of familiarity or aboutness which relates to the notion of topicality. In this regard, it is interesting to note that (10b) is an SOV construction, a fact that would suggest that ‘object shift’ here interacts with definiteness and topicality. Finally, example (10c) is an instance of topic drop, which means that the DP yàoshi is discourse-anaphoric. The correct description therefore seems to be that question (9a) does not open the floor for a flexible situation where any choice between the grammatical sentences under (9b) and (10) could fit in. Instead, (9a) calls for the choice of a linguistic expression in (10) excluding (9b). It is clear from this description that the choice of a linguistic expression is sensitive to discourse considerations external to syntax. But what matters for the present discussion is that whatever linguistic expression is chosen by the speaker has very specific syntactic properties. In the case of (9a), the felicitous expressions have a syntactic topic that formally distinguishes then from the infelicitous answer in (9b), which itself has a specific syntactic format (i.e., simple SVO declarative). There seems to be no principled way of approaching these facts, if we interpret (as often implied in the literature) the inclusiveness condition as an exclusion of topic and focus features from core syntax. Indeed, it cannot be said that the sentences under (10) acquired the topic reading only in context. If this were the case, they would be ambiguous and therefore undistinguishable from (9b). Instead, what happens here is that there is a request for a topic in (9a) which is fulfilled by the answers in (10) but not (9b). Based on (2), I argue in what follows that strict application of the inclusiveness condition requires that core notions of information structure (interrogative force, topic, focus) be part of the numeration, and project in syntax. 2.2.2. wh-words and interrogative force Let us first step back and consider the wh-question in (7). In discussing this example, we suggested that the numeration contained an interrogative Cwh that requires wh-movement in English. In this view, the wh-phrase moves to [spec CP] to clause-type the question (Cheng 1991, Cheng & Rooryck 2000). This analysis cannot extend to example (9) though: there is no wh-phrase in this example, all we have is a noun phrase followed by a question marker: ne. In terms of Li and Thompson (1981: 305–306), this declarative typing particle can also be used to type various questions, including truncated questions that include a noun phrase only (9a). The authors report that though ne retains its declarative meaning (e.g., signifying: ‘this is what I say in connection with your previous claim’) its meaning in questions can be interpreted as: ‘In connection with your claim or expectation, let me find out…’ or ‘with respect to what you have just said, let me ask you…’ This description suggests that (9a) should not  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 20 Information Structuring begins with the Numeration be regarded as an isolated DP including the particle, but rather as a fragment question. Adopting Merchant’s (2004) view of fragment answers, this would mean that (9a) involves ellipsis of the complement of ne. More precisely, (9a) presumably represents a structure where ne, the ellipsis site, heads a functional projection within C, whose specifier contains the attracted DP yàoshi, and whose complement (i.e., the presupposition described by Li and Thompson) is silenced as described in (9’a), see section 3.2.2. on Gungbe. This question therefore recalls English what-about questions where the relative clause is elided (e.g., as in what about the keys <that I asked you to bring>?). The only difference between English and Mandarin Chinese therefore seems to be that in the latter, the DP constituent that moves to [spec CP] does not contain a wh-word. (9’) CP 3 spec C’ [yàoshi]F 3 C[E] <XP> ne 6 [yàoshi that I asked you to bring] What appears from this description is that the particle ne has typical properties of functional elements responsible for ellipsis (as described in Merchant 2004): it instructs non-pronunciation of it complement, which must be given or presupposed. In addition, ne appears to interact with interrogative force in Mandarin Chinese. As such it is endowed with an EPP-feature that forces attraction of the DP-topic in its specifier. These properties of neconstructions, clearly distinguish them from Mandarin Chinese wh-questions, which involve genuine wh-phrases (e.g., shéi: who; shénme: what; duō: how). Given that the fragment question (9a), as described in (9’), is properly interpreted as wh-question though it lacks a wh-word, we can reasonably assume that wh-words do not express the feature interrogative. In the context of the English example (7b), this leads us to hypothesize that: (i) the expression of interrogative force does not depend on the wh-phrase who, but on some silent question particle within C (e.g., as in Mandarin Chinese), and (ii) wh-movement satisfies other requirements (e.g., focus, EPP) than clause typing. Contrary to what is often discussed in the literature therefore (e.g., Cheng 1991, and much related work) this description suggests that we must tease wh-movement and clause-typing apart. The distribution of wh-phrases in English provides us with a first set of empirical facts that support this view. Indeed, English wh-phrases are not restricted to interrogatives only but can freely occur in declarative clauses that also involve wh-movement. These examples indicate that English wh-movement is not contingent on interrogative clause typing.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 27 Enoch O. Aboh A priori, one could think of the Gungbe low tone as a simple prosodic requirement on yes-no questions similar to rising or falling intonation in some languages. This would follow because yes-no questions require a falling intonation in Gungbe and other Gbe languages, contrary to the rising intonation often found in Indo-European languages. A look at the closely related language Fongbe, however, indicates that yes-no questions include a full question morpheme: à (26). This suggests to me that the low tone in (25b) is the Gungbe equivalent of the Fongbe question morpheme. 9 (26) É ù nú à? [Fongbe] 3SG eat thing INTER ‘Did he eat? Granting this, it appears that, in Gungbe as well as in Fongbe, topic and focus markers, which mainly occur in the left periphery and the right peripheral marker (i.e., the yes-no question marker) may co-occur. When this happens, we may obtain the sequences in (27) for Gungbe, and (28) for Fongbe (see Aboh 2004a, chapter 8 for discussion). (27) a. Ùn kànbí  Kòfí ù lsì w? [Gungbe] 1SG ask that Kofi eat rice FOC-INTER ‘I asked whether KOFI ATE RICE *e.g. he shouldn’t do so because he is taking medicine+?’ b. Ùn kànbí  Kòfí ù lsì ya? 1SG ask that Kofi eat rice TOP-INTER ‘I ask whether Kofi ate rice *as planned/mentioned+?’ c. Ùn kànbí  Kòfí ní x mótò w ya? 1SG ask that Kofi MOOD buy car FOC TOP-INTER ‘I asked whether KOFI SHOULD BUY A CAR *as planned/mentioned+’? d. *Ùn kànbí  Kòfí ní x mótò yà w? 1SG ask that Kofi MOOD buy car TOP FOC-INTER (28) É ù nú a w à? [Fongbe] 3Sg eat thing NEG FOC INTER ‘Is it that s/he has not eaten? [Lefebvre & Brousseau 2002: 135, 485] These examples are particularly illustrative of various facts: (i) some of the left peripheral markers described in (22-24) (e.g., topic and focus) can also occur to the right. (ii) When they co-occur to the right, they appear in the mirror image of that in (24). In (27c), for instance, we observe the sequence (w  >yà) as opposed to (yà>w  ) in (24). Notice from example (27d) that preserving the order in (24) in sentence-final position leads to ungrammaticality. 9 See Aboh 2004a and references cited there for discussion on the Gbe languages. It is proposed there that such a floating tone could be seen as a leftover of a full morpheme that the language has had at earlier stages.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 28 Information Structuring begins with the Numeration I have proposed elsewhere that this mirror image effect is the result of snowballing movement which rolls the clause up to the specifier of the interrogative phrase, as described in (29), (Aboh 2004a, b, d). (29) [ForceP [Force° [InterP [Inter°  [TopP [Top° yà [FocP [Foc° w [FinP Kòfí ní x mótò]]]]]]]]]    On the basis of these examples, we can say that the peripheral markers (i.e., Topic, Focus, Interrogative) form a paradigm where they mark discourse properties. Since Chomsky (1977), and more recently Cheng (1991), it is unanimously accepted that question particles head a functional projection (ForceP/InterP) in syntax. If we look at the Gungbe facts from this perspective, it is reasonable to conclude that the topic and focus markers are similar to the question particle in that they project in syntax. 10 Put another way, nothing in Gungbe suggests that interrogative force is a particular information-structure- sensitive feature that requires a different treatment from topic and focus. If this is so for Gungbe (and generally Gbe), then the burden of proof is on the adherents of the view that Romance and Germanic are different and that interrogative force in these languages requires a special treatment. In continuing our discussion, I now turn to fragment answers in Gungbe, which I show support the view advocated in this paper. 3.2.2. Ellipsis A particularly insightful conclusion reached in Merchant (2004: 664) is that subsentential answers can function as propositions (i.e., correspond to assertion) because (i) they are syntactically generated as parts of full sentences, (ii) they are subject to a movement operation that evacuates them from the non pronounced part of the sentence. Under this view therefore, sentence (30a) can be assigned the representation in (30b) (Merchant 2004: 670). (30) a. Abby was reading something, but I don’t know what <Abby was reading t> b. CP c. syntax of E: E [uwh, uQ] 3 spec C’ whatwh 3 Cwh[E] < TP> 6 Abby was reading twh 10 See Aboh (2004a) where it is shown that both C and I involve markers (i.e., free morphemes) that are the expressions of distinct functional projections.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 29 Enoch O. Aboh In the representation (30b) the feature E, which marks the ellipsis site, has the syntax in (30c), forcing it to attach to a C head bearing the relevant whfeatures, which are checked by the fronted wh-phrase. Merchant (2004) further argues that while the feature E may attach to different heads cross-linguistically, its phonological and semantic properties are uniform across languages: it instructs non-pronunciation of its complement and the unpronounced complement must be given. What emerges from Merchant’s analysis is that ellipsis derives from a formal syntactic feature, which in English and some languages, attaches to a head within C, whose complement remains unpronounced. Given the inclination of Gungbe to spell out all the heads within the left periphery (24), we may hope to find evidence in this language where the covert stranded Cwh[E] in English will get pronounced. Finding such cases will not only provide strong empirical support to Merchant’s analysis of ellipsis, but will also support the point made here that the functional elements represented in (24) are real syntactic objects endowed with formal syntactic features (see also Baltin 2006). Once again, Gungbe makes life easy for us. Though ellipsis is very restricted in this language, the following examples are possible: (31) a. Súrù x nŭ é, àmn má ny nŭ ě *(w) Suru buy thing INDEF but 1SG.NEG know thing that FOC ‘Suru bought something, but I don’t what’ b. M é wá àmn má ny m ě *(w) person INDEF come but 1SG.NEG know person that FOC ‘Someone came, but I don’t know who’ Compared to their English translations, these Gungbe examples show that: (i) the locus of the feature E in Gungbe is the focus head (i.e., Foc realized here by the focus marker), (ii) ellipsis is not possible in the absence of the focus marker. Adapting Merchant’s analysis to Gungbe, I partially represent (31a) as in (32a), where the feature E has the syntax in (32b). (32) a. FocP b. syntax of E: E [uFoc, uQ] 3 spec Foc’ *nŭ ě+F 3 Foc[E] < FinP> w 6 Súrù x t*nŭ ě+ This analysis sheds light on the fact that the element that fronts in Gungbe ellipsis is not a simple wh-phrase of the English type, but a complex phrase involving a dummy noun phrase nú ‘thing’ and a Q-morpheme (see section 5). This clearly relates to the point made in this paper that the feature at  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 30 Information Structuring begins with the Numeration work in these Gungbe constructions is a focus feature that is anchored under Foc, the focus head. Under this view, the only difference between English and Gungbe is that the latter, but not the former, spells out the focus head whose specifier is occupied by the fronted wh-phrase. Given our minimalist standpoint that the role of the computational system CHL is to map the numeration onto a converging linguistic expression, I conclude from the discussed facts that Gungbe involves discourse particles endowed with formal features that are introduced lexically in the derivation as part of the numeration. Accordingly, Gungbe is a discourse-configurational language because the syntax of discourse particles determines the surface structure of the sentence. We therefore have a clear case where information structure, encoded by discourse particles, has direct access to syntax. The discussed examples do not only support the viewpoint defended here, but they also indicate that any analysis departing from this view will violate the inclusiveness condition in (5). 11 If information structure is indeed a property of the lexicon, we expect to find its effects in other domains besides the clausal ones. Using data form the Gungbe noun phrase, the following section shows that the DP is another field where we find information-structure-sensitive functional items. 4. Some parallels between the sentence and the noun phrase Previous discussion shows that Gungbe (and Gbe languages in general) include discourse particles (e.g., Topic, Focus, Inter) that are introduced in the derivation as part of the numeration. In this regard, a remarkable property of these languages is that we find similar particles within the DP (Aboh 2004c). 4.1. The activation of Topic within D Gbe languages distinguish between specific and non-specific referents. Specific referents are defined as follows: 12 (i) A specific definite noun phrase is strongly D(iscourse)-anaphoric and represents a unique referent assumed to be known to both speaker and hearer, and which the speaker intends to refer to. 11 Indeed, assuming that the markers are introduced at PF (as one could imagine) is equivalent to saying that lexical items are added to the derivation once completed by CHL. A seemingly possible way out could be to propose that these discourse markers are prosodic cues comparable to stress-assignment in Indo-European. Though a possibility, this view makes these languages even more exotic than they already are because they will be the only cases known to me where prosody translates into words! Therefore the burden of proof is on those who embrace this view. 12 See also Ionin (2006) for discussion on specificity.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 31 Enoch O. Aboh (ii) A specific indefinite noun phrase need not be D-linked. It represents an existing referent that the hearer may not know about, but which the speaker has in mind and intends to refer to (Aboh 2006b: 224). In Gungbe, a noun phrase without a determiner (referred to here as a bare DP) may be interpreted as (in)definite or generic depending on the context. 13 The sequence távò títán in (33a) is understood as definite (because it is unique in the discourse setting) but non-specific because it was not previously mentioned in discourse. On the other hand, a noun phrase including a determiner (referred to as a full DP) must be interpreted as specific. When these DPs have been explicitly mentioned in previous discourse or represent entities, referents or events which both the speaker and the hearer know about or have some shared experience of, they are interpreted as definite specific. Accordingly, the sequence távò l  (33a) is interpreted as definite specific (because it refers back to távò títán ‘table first’). On the other hand, indefinite specifics (e.g., távò  é in 33b) refer to a referent that is known to the speaker but not necessarily to the hearer. Observe that this specific reading is absent in the sequence távò cè (33b) which is only definite. In addition, example (33c) shows that l  and  é are mutually exclusive, though they can co-occur with other DP-internal markers (e.g., demonstrative, possessive) (33d). This is taken as a piece of evidence that l  and  é target the same position, and should be distinguished from other nominal modifiers such as demonstratives (see Aboh 2004a, chapter 4, 2004c). (33) a. Kkú mn távò títán bò  émì ná x távò l Koku see table first and say 3SG FUT buy table DET[+spec, +def] ‘Koku saw the first table and said that he would buy that specific table’ b. Kkú mn távò cè bò  émì ná x távò é Koku see table 1SG.GEN and say 3SG FUT buy table DET[+spec -def] ‘Koku saw my table and said that he would buy a certain table’ c. *Kkú mn távò l é Koku see table DET[+spec, +def] DET[+spec, -def] d. Kkú mn távò cè titan éhè l bò  …. Koku see table 1SG.GEN first DEM DET and say ‘Koku saw this first table of mine and said….’ It appears from this brief description of the DP that a Gungbe nonspecific noun need not be marked while a specific noun phrase must necessarily take the marker l  or  é (see Aboh 2004a, chapters 3, 4 for discussion ). The Gungbe definite specific DPs recall Li and Thompson’s (1975) description of baconstructions illustrated in (10b), or Prince’s (1981) assumed familiarity. A common aspect to these DPs is that they interact with discourse topic. Thinking of the parallels between CP and DP (Aboh 2004c), a hypothesis that we may 13 See Aboh (2004, chapters 3, 4) for arguing that such bare noun phrases are DPs.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 32 Information Structuring begins with the Numeration formulate is that specificity marking within DP matches topicality at the clausal (i.e., CP) level. That this is the right characterization is suggested by an interesting contrast between focalization and topicalization in Gungbe. Both bare DPs and full DPs can be freely focused and fronted to the left of the focus marker. In addition, the process is not sensitive to definiteness, as indicated by the various options in (34). (34) a. Távò w Kófí sà bò mn kw table FOC Kofi sell and find money ‘Kofi sold TABLE(S) and found money *i.e., got rich+’ b. Távò cè w Kófí sà ….. table 1SG.GEN FOC Kofi sell ‘Kofi sold MY TABLE …..’ c. Távò l w Kófí sà …… table DET FOC Kofi sell ‘Kofi sold THAT (TYPE OF) TABLE…..’ d. Távò é w Kófí sà ….. table DET FOC Kofi sell ‘Kofi sold SOME/A (TYPE OF) TABLE…..’ With regard to topicalization, however, only definites and/or specific constituents can undergo topic movement to the left of the marker yà. 14 Consider the ungrammatical examples (35a-b) and contrast them with the grammatical ones under (35c-d): (35) a. *Távò yà Kófí sà  bò mn kw table TOP Kofi sell 3SG and find money ‘As for table(s) Kofi sold (it/them) and found money *i.e., got rich+’ b. *Távò é yà Kófí sà ….. table DET TOP Kofi sell 3SG ‘As for some/a certain table(s) Kofi sold it…..’ c. Távò cè yà Kófí sà ….. table 1SG.GEN FOC Kofi sell 3SG ‘As for my table, Kofi sold it …..’ d. Távò l yà Kófí sà ..… table DET FOC Kofi sell 3SG ‘As for that specific table, Kofi sold it….’ Similarly, adjuncts that have a unique reference (e.g., time adjuncts) can freely topicalize, but other adjuncts (e.g., place adjuncts) can only front if they are made definite or specific. Contrast again example (36a) with the time 14 Topics, unlike focus constructions require a resumptive pronoun in Gungbe except in the case of referential time adjuncts (e.g., today, tomorrow). Contrast the examples (34c-d) to (35a) below. See Aboh (2004a, chapters 7, 8) for discussion.  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 33 Enoch O. Aboh adjunct to examples (36b-c) where only the specific-definite adjunct x  l  kpá ‘beside that specific room’ is allowed as topic. (36) a. S yà, Kòfí ná zà x l kpá] gbáú ! tomorrow TOP Kofi FUT sweep room DET POST[beside] necessarily ‘Tomorrow, Kofi has to sweep beside the room!’ b. [X l kpá] yà, Kòfí ná zà fln gbáú ! room DET POST[beside] TOP Kofi FUT sweep there necessarily ‘As for beside the aforementioned room, Kofi will sweep there!’ c. *[X kpá] yà, Kòfí ná zà fln gbáú ! room POST[beside] TOP Kofi FUT sweep there necessarily ‘As for beside room(s), Kofi will sweep there!’ Given that both bare DPs and full DPs can be focused (34), we cannot attribute the contrast between (34) versus (35) and (36) to some structural deficiency making bare DPs unsuitable for movement operations. Instead, it seems as if definiteness (however encoded in this language) or specificity marking with l  is a pre-requisite for clausal topicalization. This is not surprising since topics are discourse-anaphoric (or D-linked) by definition. The interesting question instead is this: Why would we have this matching between the topic head and the DP it attracts if the interaction at stake were just a pragmatic one? Keeping to minimalist assumptions, the observed matching appears a classical instance of the probe-goal interaction whereby the Gungbe topic head probes over a definite or specific (i.e., discourse-anaphoric) constituent, which is attracted to its specifier position and interpreted as topic. Needless to say that CHL properly computes this operation because the topic marker yà and the specificity marker l  are introduced in the derivation as part of the numeration. These elements are then selected and merged with other elements (i.e., D with N, and Top with IP) to form new syntactic objects: DP and TopP. 4.2. The activation of focus within D A similar conclusion can be reached on the basis of wh-questions in Gungbe. This language does not have English-type wh-words. Instead, question words consist of a noun phrase and a question particle, as in the forms nú-t   ‘what’ and fí-t   ‘where’ in (37a-b). (37) a. [Nú-t] w Kòfí x? Thing-Q FOC Kofi buy ‘What did Kofi buy?’ b. [Fí-t] w Kòfí yì? Place- Q FOC Kofi go ‘Where did Kofi go?’ As example (38) shows, the noun, in these complex wh-words, can be separated from the question particle by intervening modifiers. This is partial  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 34 Information Structuring begins with the Numeration evidence that Gungbe wh-phrases are complex phrases where a phrase is associated with a Q-morpheme. (38) [Fí j-fí t] w Kòfí trn bò  kú íó mn? Place well-known Q FOC Kofi come and has die dirt so ‘What possible important place could he come from to be so dirty?’ Let us compare these wh-phrases to the format of focused constructions and wh-questions. (39) a. [[é-t] w] Kòfí x? 3SG- Q FOC Kofi buy ‘What did Kofi buy?’ b. *é w Kòfí x ? 3SG FOC Kofi buy c. [[àkwékwè] w] Kòfí x banana FOC Kofi buy ‘Kofi bought BANANA(S)’ Here, we observe a striking parallel between wh-questions and focus expressions. The sentences under (39) indicate that, in wh-questions and focus constructions, the attracted DP is immediately adjacent to the focus marker w   (39a). That Gungbe resorts to the same focus marker clearly shows that fronting in both cases involves the same functional head. Looking closely at whquestions, we further observe that the fronted wh-phrase must contain the Q- morpheme t  , hence the ungrammatical example (39b). 15 The proper nature of this morpheme is not clear to me and it could be considered a pure question particle or a general operator sensitive to both focus and question operations, which I tentatively labeled here as Q. Given this description, the pattern in (39) is reminiscent of the data discussed in the context of topicalization, where we have shown that a marker inside the DP allows the latter to be attracted at the clausal level by a topic marker. Therefore, just as the topic head probes over a DP marked as definite or specific, the focus head, in relation to interrogative, appears to probe over question phrases that are marked by t  . Accordingly, we have here another instance of a matching between a probe and its goal (see footnote 16 for the discussion on long distance agreement). What these Gungbe facts tell us is that DPs (or else the relevant constituents) do not end up being topicalized or focused by chance, but because they embed the relevant Topic and Focus features. These are spelled out on the probe by lexical material in the form of the discussed markers. 16 The next 15 On a purely methodological side, the sentences in (39) show that it would be highly undesirable to assume that wh-question in (39a) is a pure product of syntax, while the focus example in (39b) is leftover to pragmatics even though the two derivations seem to be dragged by the focus marker. 16 Assuming the view developed here, another relevant empirical domain to explore  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 35 Enoch O. Aboh question is what type of features are Topic and Focus compared to other (traditional) formal features (e.g., tense, aspect, mood, case)? 5. Topic and focus features It seems clear from the Gungbe facts that nothing in principle distinguishes between Topic and Focus features and other formal features such as Tense, Case or -features. 17 We have already shown in previous sections that with regard to interrogative, topic, and focus features is that of long distance Agree. Indeed, it has been shown in the literature that formal features can be checked under long distance Agree between the probe and the goal. The analysis put forward here suggests that the same should hold of information-structure-sensitive features. Though further work on this issue is in order, certain facts from Standard Arabic (SA), as discussed by Ouhalla (1993), suggest that focus may be checked under long distance Agree. Ouhalla (1993: 279) reports that SA allows both in-situ or fronted focused expressions as indicated below: (i) a. allaf -at Zaynab-u qasiidat-an Write.PERF-3SG.F Zaynab-NOM poem-ACC ‘Zaynab has written a POEM.’ b. qasiidat-an allaf -at Zaynab-u poem- ACC write.PERF-3SG.F Zaynab- NOM ‘It is a poem Zaynab has written.’ According to Ouhalla, (ia) encodes new information focus while (ib) expresses contrastive focus (Kiss 1998). In addition, it is possible to express contrastive focus with an insitu constituent thanks to the focus markers  inna and la. The former occurs in sentence-initial position while the latter attaches to the focused in-situ constituent. (ii) inna Zaynab-a (la)-shaa’irat-un (laa riwaaiyyat-un) FOC Zaynab- ACC FM poet- NOM not novelist- NOM Zaynab is a POET (not a novelist).’ In terms of Ouhalla  inna heads a focus phrase, while la Ouhalla is comparable to ‚a tonic accent (focus stress) in that it serves the function of identifying the constituent which is being contrastively focused. This element can therefore be considered a constituent focus marker (p. 281).‛ This characterization recalls my own description of DP-internal topic and focus in Gungbe. In current Minimalism framework, Ouhalla’s (1993) proposal can be seen as an instance of probe goal relation, where the feature focus under the probe is checked by long distance Agree between  inna and la. If this characterization turns out to be the right one, then we may have to change our perspective on languages where there seems to be no segmental marking at all for topic and focus. Indeed, the fact that assignment of in situ focus often correlates with main accent/stress in many languages (e.g., English) can be seriously taken as an instance of long distance agreement between the focus probe or the topic probe within the left periphery and the relevant goal within IP. This means that English is just a less obvious realization of the Arabic situation. This description would also suggest that ex situ versus in situ focus distinction boils down to the choice between Move and long distance Agree in these languages. 17 This need not mean that these features form a uniform class. While Case and - features relate to nouns only, topic and focus apply to all categories (see Aboh 2004). Similarly, Rizzi (2004) proposes that topic, focus, (and interrogative) are criterial features, but this does not  Iberia: An International Journal of Theoretical Linguistics vol 2.1, 2010, 12-42 pp. http://www.siff.us.es/iberia/index.php/ij/index ISSN 1989-8525 36 Information Structuring begins with the Numeration the licensing of the Topic or Focus head implies a matching operation that is necessitated by Probe-Goal relations. In this section, I take the comparison a step further. An interesting fact about Gungbe (and Kwa in general) is that, in addition to discourse-related markers (e.g., Topic, Focus, Inter), they also involve INFL-related markers that encode tense, mood, aspect, negation, etc. (Aboh 2004a). Let us consider the category Tense in Gungbe. This language has a future tense marker ná only, illustrated in (40): here we see that the only difference between (40a) and (40b) is precisely the presence of this marker in the latter but not in the former. Thus, (40b) is interpreted with future reading while (40a) is read as perfective. The ungrammatical examples (40c-d) indicate that the subject cannot remain to the right of this marker (e.g., in [spec VP]), recall that Gungbe is SVO. (40) a. Kòfí x xwé l Kofi buy house DET ‘Kofi (has) bought the house’ b. Kòfí ná x xwé l Kofi FUT buy house DET ‘Kofi will buy the house’ c. *ná Kòfí x xwé l FUT Kofi buy house DET d. *xwé l ná Kòfí x house det FUT Kofi buy Aboh (2004a, chapter 5) demonstrates that the marker ná encodes T in Gungbe. In the context of the present discussion, and following Chomsky (1995: 237, ff), this amounts to saying that the Numeration that produced sentence (40b) contained ná as expression of T. Now, let us contrast the sentence in (40b) to the topic and focus sentences in (41a-b). Examples (41c-d) show that these markers must necessarily be associated to some constituent immediately to the left. (41) a. Kòfí w x xwé l Kofi FOC buy house DET ‘KOFI bought the house’ b. Kòfí yà é x xwé l Kofi top 3SG buy house DET ‘As for Kofi, he bought the house’ apply to Case and -features. Further work is needed for understanding the intrinsic properties of these features and how they relate to the observed differences. What matters for the present discussion though, is that the CHL manipulates these two classes of features the same way (I thank Anna Cardinaletti for commenting on this issue).