1.1 Purpose and organization of the paper

The purpose of this paper is neither to defend the cartographic project nor to gather more arguments against it. Rather, it mainly aims to counterbalance a number of definitive points made by Neeleman et al. (2009) with respect to focus in Russian as part of their more general anti-cartographic stance. At the same time, staying critical of cartography, the paper opts for an alternative to both the left periphery focus and the focus-final hypothesis. Developing such an alternative in detail is not the primary aim of the present paper, though – this would require more attention and space. Rather, I will present a brief preview of an alternative approach at the end of the paper, which is meant merely as an invitation for further study.

The remainder of Section 1 recalls the key notions of the grammar of focus relevant to the subsequent discussion. Section 2 presents the three types of evidence produced by Neeleman et al. (2009) in favor of their hypothesis that focus is uniformly clause-final. Section 3 discusses two conceptual problems arising from this stance and Section 4 provides empirical counterevidence to their proposal. Section 5 raises a few speculative points and implications for a general theory of focus arising from the previous discussion and Section 6 offers concluding thoughts.

1.2 Focus in cartography

Under a cartographic view, focus is linked to a dedicated position in the left periphery, whose classic formulation, resulting from the splitting of CP, is Rizzi’s (1997) template, as in (1) below:

Thus, a phrase designated for a focus interpretation (somehow) finds its way to the position of the specifier of the Focus Phrase in which it enters a checking relationship with a focus head with respect to the feature [+F], as schematized in (2) below:

The placement of a phrase in Spec. FocP triggers specific interpretations at the interfaces: special focus semantics on the meaning side and the marked intonation on the sound side.

Rizzi (2014) characterizes the cartographic approach as attempts to “‘syntacticise’ as much as possible aspects of scope-discourse semantics, in that fundamental scope-discourse interpretive properties are traced back to basic syntactic configuration in a transparent and straightforward manner”, so that “[…] syntax wears interface properties on its sleeves” Rizzi (2014: 527).

1.3 Focus type, focus fronting, and cartography

A debatable issue is how a position of focus relates to the dichotomy between identification focus and (new) information focus, originally postulated by É. Kiss (1998). ^[1] Under one view, focus in the left periphery (ex situ focus) always marks (exhaustive) identification or contrast (e. g., Drubig 2003; López 2009). The instances of overt focus fronting from Hungarian, Polish and Catalan respectively, given in (3)–(5) below, all involve exhaustive identification of the capitalized constituent.

Thus, contrastive (exhaustive) ex situ focus is distinguishable from (new) information focus, which is always in situ. The latter is associated with the position of main sentence stress, predictable by some variant of the Nuclear Stress Rule (Chomsky and Halle 1968; Halle and Vergnaud 1987; Cinque 1993), as illustrated by (6):

On a different view, contrastive focus can also be found in a low in situ position (e. g., Rizzi 1997, 2006; Samek-Lodovici 2005), as illustrated by the following Italian example: ^[3]

A way to reconcile an apparently low position of contrastive focus, as in (7), with cartography is suggested by Belletti (2004), who argues that objects like la ragazza in (7) are anyhow raised to Spec. FocP in the left periphery with the rest of the sentence remnant-moved to a higher Spec. TopP. In this way, a clause-final position of focus need not, in principle, be incompatible with a cartographic, left-peripheral view of focus. ^[4]

Another question is whether new information focus can also be found ex situ. In a non-cartographic framework, López (2009) argues against this option for Romance languages, but Benincá and Poletto (2004) point to Italian examples like (8) below, to show this possibility exists: ^[5]

Also, as discussed later (Section 4.5.), Fanselow and Lenertová (2011) argue for the option of ex situ information focus for German.

1.4 Focus projection

A distinguishing property of new information focus is its ability to project. Contrastive and exhaustive foci cannot project beyond their immediate phrasal projection – they must be narrow; new information focus, on the other hand, can project: it may be narrow or broad (Chomsky 1971; Gussenhoven 1983; Selkirk 1984; Reinhart 1995; Gussenhoven 2007; Selkirk 2007; Büring 2006). Thus, under a standard view of VP focus projection, formulated by Selkirk (1995), focus can be projected to a VP from its accented V head (a case of vertical focus projection) or from its accented complement (horizontal focus projection). ^[6] Such a projection results in the formation of broad VP focus, which answers wh-questions of the form: What did x do? or What happened?^[7]

As illustrated by (9) and (10), if focal stress falls on football, the new information part may be the nominal phrase (DP) itself, the verb phrase, or the whole sentence, i. e., the information focus may be extended to VP or TP. However, the same extension is not possible for contrastive focus – the element in focus is just the constituent carrying focal stress, as shown by (11) and (12). Also, if the accented constituent is the head verb, e. g., danced, then a new information focus can be extended from V (through VP) to TP, but the same extension is not possible if the accented verb is used contrastively, as illustrated by (13)–(16):

Büring (2006) argues that Selkirk’s (1995) standard formulation of focus projection should be supplemented by the formulation of the rules of Default Prominence to account properly for, inter alia, transitive structures. For such cases, Default Prominence should guarantee that horizontal focus projection prevails over the vertical one, so that in any V-complement structure, a complement, not a V head, projects its focus to the VP. This prevents cases like We PLAYED football from being interpreted as instances of VP (broad) focus allowing only their narrow focus interpretation (e. g., answering the question Did you WATCH football?, but not the question What did you do?).

1.5 Critique of cartography

Despite its descriptive appeal, a cartographic view of sentence left-periphery has been subjected to criticism harsh enough to have Bailyn (2011) declare that the “overall picture that emerges for cartography is fairly bleak” (Bailyn 2011: 667). Furthermore, alternative, non-cartographic accounts of how to encode focus and topic in formal grammar have been advanced (see papers in van Craenenbroeck 2009). One of the toughest points raised against focus cartography is its questionable empirical coverage across languages. Overt “focus movement” to the left of a clause is not universal, and, even when it occurs, the landing site may differ from language to language (Bouchard 2009). ^[8] Next, there are languages, like Finnish, in which there is room for just one left-dislocated XP and not for complex templates like (1) (López 2009). It can also be argued that, in principle, focus can be associated with any clause-internal position and that the strict ordering of topic and focus, predicted by the cartographic approach, is questionable (Wagner 2009). Moreover, the alleged focus movements can be viewed as triggered by independent factors, such as quantification or contrast (e. g., Horvath 2007; López 2009; Neeleman et al. 2009; Horvath 2010), or as instances of movement for checking a semantically neutral edge feature of C (Fanselow and Lenertová 2011).

As noted above, a contentious point for the cartographic view is the versatile position of focus in a sentence. As acknowledged by Rizzi (2014), focus can be realized in syntax in two ways: (i) it may be “superimposed to the hierarchical structures created by Merge […], an element can be focalized in situ, without being displaced from its argument position”, or (ii) the element in focus “is displaced from its thematic position to another position, typically in the initial periphery of the sentence, where it receives its appropriate scope-discourse property” (Rizzi 2014: 4). Of these two ways, the latter has been widely explored within the cartographic project. However, the facts from Russian (and Polish) appear to show that focus in situ (both for new information and contrast) appears to be even more widespread than “focus fronting” (limited to contrastive focus). One way to respond to this observation is to advance a proposal such as Neeleman et al.’s (2009), which associates movement with contrast and links focus exclusively with a right periphery instead. A plausible alternative, as it seems, would be to develop a non-positional view of focus freely available in different sentence positions, fronted, rear and intermediate, with its specific interpretation as new information or contrast determined only at the C-I interface (see footnote 21 below).

2.1 The anti-cartographic stance of Neeleman et al. (2009)

Neeleman et al. (2009) present their empirical evidence and arguments against the cartographic view in three major parts. The first deals with the instances of A’-scrambling in Dutch targeting phrases with contrastive interpretation, i. e., contrastive topics and contrastive foci. The second is concerned with the cases of Japanese topic fronting and their uniform linkage to a clause-initial position. The third tackles the conditions of focus in Russian. Jointly, these three parts are meant to refute two major claims of cartography by showing that (i) there is no one-to-one relation between a syntactic position and interpretive effect, and (ii) there is no rigid and cross-linguistically fixed “clausal skeleton”. In what follows, I will not assess the general anti-cartographic stance of Neeleman et al.’s (2009) proposal as a whole. In particular, I will not offer an assessment of, or counter-argumentation to, the facts from Dutch and Japanese. Instead, the discussion is confined to the evaluation of their arguments with regard to the position of focus in Russian.

2.2 Unitary clause-final focus and privative features

The major premises of Neeleman et al.’s (2009) proposal with respect to focus in Russian may be summarized as follows:

1. There is no dedicated functional Foc head in the left periphery.

2. Focus is unitary; there is no grammatical distinction between new information focus and identification (contrastive, exhaustive) focus.

3. The position of unitary focus is clause-final.

4. There is no principled distinction between in situ and ex situ focus (all focus is in situ).

5. The observed ex situ focus effect is the result of (A’-) fronting for contrast; there is no fronting for focus.

6. [+focus] and [+contrast] are privative syntactic features; there are rules applying to categories carrying [+contrast] and other rules applying to categories carrying [+focus], but no rules applying specifically to categories carrying [+contrast, +focus].

The authors’ argumentation in favor of a canonical, clause-final position of focus in Russian is based on three observations: (i) there is complementary distribution of contrastive and new information focus, (ii) contrastive foci obligatorily take the lowest scope position, and (iii) split scrambling is allowed only from the clause-final position. Below is a brief exposition of these three sorts of argument.

3.1 On being clause-final

Neeleman et al. (2009) do not provide any structural account of the clause-final position of focus. Thus, it is not evident how [focus] is accommodated within the CP in (22):

In other words, “clause-final” position is taken to be syntactically relevant, but the position is not related to any structural property of the “right periphery”. ^[12] The question of why all foci should appear in a specified “clause-final” position remains unresolved since this position is not structurally dedicated for focus in any way. It would therefore be problematic if dismissing left-periphery focus should not lead to formulating a structural account of the right-periphery focus.

An anonymous reviewer points to two structural options for the focus-final hypothesis to be reconcilable with cartography. One relies on remnant movement and the other on the idea of the right specifier. Under the former, the effect of stranding focus in a final position is achievable by a left-bound movement of the focused phrase to the Spec. of FocP and by (remnant) moving the rest of the structure to a higher Spec. TopP position. This type of analysis has been adopted by Belletti (2004) in her cartographic analysis of in situ contrastive focus in Italian. However, given the encompassing nature of the anti-cartographic argumentation in Neeleman et al. (2009), it is difficult to consider remnant movement as a structural option for their hypothesis. Besides, as noted by Brunetti (2003) and López (2009), it raises the question why contrastive focus should ever surface in the left-periphery if the movement of TP to Spec. TopP were obligatory for its interpretation as topic. Furthermore, one should seriously regard Bouchard’s (2009) criticism of remnant Move (in the context of the analysis of the scope of adjectives in French, and more generally) as a tool so powerful that “it is just as easy to derive all the wrong orders” and “to relate any surface order to any underlying structure”, so that “anything is possible” (Bouchard 2009: 258). He also argues that such tools cannot “make correct empirical generalizations” and the cartographic system which uses them is neither precise nor informative.

Under the right specifier option, a focus phrase (XP) would appear in a specifier of a dedicated FocP located in the right periphery, as schematically given in (23):

The configuration in (23) would call for the examination of the linearization effects, especially in view of Kayne’s (1994) Linear Correspondence Axiom (LCA) – it is unclear how a focused phrase in (23) could fail to c-command the material in TP in order to be linearized as final. Besides, there is no indication in Neeleman et al. (2009) that the authors could opt for a right-periphery variant of cartography.

3.2 On unitary focus

The second general point relates to Neeleman et al.’s (2009) unitary treatment of focus. They define it, seemingly broadly, as “the information highlighted in a proposition”, and provide the following exemplification:

They further assume that focus is the constituent that corresponds to the wh-expression. The two conditions may be summarized as in (38):

Thus, in (37b) a gun is the focus element because it corresponds to what, and it is highlighted with respect to the rest of the sentence, which functions as the background containing a variable, i. e., Rutger bought x (see Neeleman et al. 2009: 17). ^[22]

Neeleman et al’s (2009) characterization of focus in (38) corresponds to what, in less-unitary approaches to focus (e. g., Drubig 2003), is referred to as completive focus, or a completive function of focus. This type has typically been juxtaposed with contrastive focus, exemplified by (39) below:

It should be noted that in (39b) a gun does not directly correspond to the wh-expression since there is no wh-expression in the immediate context of (39b). ^[23] The criterion in (38a) is therefore not met and a gun in (39a) could not qualify as a focus element. Further, the only sense in which a gun in (39b) is highlighted is by providing contrast (by correction) to a proposition expressed by (39a). Unlike in (37b), then, it is not highlighted by completing the gap in the background proposition, i. e., Peter bought x.

In sum, it appears that either the case illustrated by (39) does not qualify as focus under formulation (38), or (39b) is a genuine instance of focus, but the formulation (38) is too narrow. Neeleman et al. (2009: 4) classify cases like (39b) as instances of contrastive focus, derived from the combination of two privative features: [focus] and [contrast]. The problem seems to be, however, that if such examples are deprived of their contrastive aspect, nothing appears to be left to classify them as instances of focus under formulation (38). As noted, a gun in (39b) does not correspond to a wh-expression, neither is it highlighted with respect to a background proposition containing a variable. It thus transpires that the apparently broad and encompassing definition of focus adopted by Neeleman et al. (2009) is fairly restrictive, as it excludes cases of contrastive focus like (39b).

4.1 Focus and scope of quantifiers revisited

Let us start from the proposal that the observations on scope in examples like (19), repeated below, need not be interpreted as supporting a uniform clause-final position of focus.

First, for a contrastive phrase to take a lower scope position with respect to a quantifier is not logically equivalent to taking the lowest (i. e., sentence-final) position. One way to account for the high LF-position of the quantifier is to adopt a Quantifier Raising (QR) analysis. ^[24] Alternatively, a fronting contrastive QP could be reconstructed for interpretation in some position c-commanded by the existential QP (i. e., odin mal’čik) in the course of a derivation. ^[25] A further option to explain an unexpectedly low scope position of the focused QP could be that contrastive foci systematically scope below quantifiers, as proposed by É. Kiss (2003) in her cartographic template for Hungarian. ^[26]

If we wish to use any of the above alternatives as a principled account of the low scope position of the focused QP in (19), it has to be shown that the focused phrase KAŽDUJU DEVOČKU systematically differs from the other quantifying phrase odin mal’čik and hence has to take a lower scope position. To this end, it may be worthwhile to consider the semantics of KAŽDUJU DEVOČKU more closely. First, it should be noted that, contrary to the typographic convention used by Neeleman et al. (2009), the stressed part of the QP in (19) is only the nominal complement of the quantifier, and not the whole phrase, as indicated in (40). This restriction follows from the application of the Nuclear Stress Rule at a phrase-level (e. g., Cinque 1993; Zubizarreta and Vergnaud 2005):

Next, the type of contexts in which (41a) can be used are, for example, the answers in (41b) and (41c), but not the answer in (41d):

By asking the question in (41a), in which contrastive stress is placed on the noun girl, the speaker wants to learn if the hearer knows every girl or rather every boy in the class. A suitable answer is (41b) in which the interlocutor explains he knows every boy, not every girl, marking the contrast by placing focal stress on the noun boy. Another possible reply is (41c) in which the hearer explains (again by using contrastive stress) that it is boys, not girls he knows, incidentally adding that it is only some boys, not all. Finally, answer (41d) is infelicitous because the hearer does not respond relevantly to question (41a) (which is about the sex of pupils) and explains instead he knows only some girls, not all of them, placing contrastive stress on some.

The contextual facts in (41) show that the semantics of the phrase in (40) is such that the alternatives are evoked and dismissed in the position of the nominal complement of the quantifier within the QP, but not in the position of the quantifier head itself (assuming the structure of the QP to be: [_QP Q [_NP N]]). However, the QP of (41c) differs from the QP of (41a) not only in the content of the complement, but also in the choice of the type of quantifier: the universal quantifier každogo is replaced with the existential one nekotoryx. Despite this replacement, the answer (41c) is fully felicitous – it looks like what matters in the discourse is the opposition devočki vs. mal’čiki, which is marked by focus. Notice, that this sort of substitution would be impossible within an exchange like (42) below, in which the underlined quantifier phrases are not in focus:

A way to account for this difference between (41) and (42) is that the presence of focal stress within the phrase nekotoryx MAL’ČIKOV in (41c) renders the semantics of the quantifier inert, so that even a change in the (type of) quantifier in the reply, which should result in a crucial change of the truth conditions, becomes undetectable. The same does not occur in (42b), where contrastive focus falls outside of the underlined QP and the change of the type of quantifier results in the infelicity of (42b) as a response to (42a). It looks as though the whole focused phrase was no longer interpreted as a quantifier phrase; the semantics of focus appears to have suppressed the semantics of the quantifier.

If the interpretation of the facts in (41) and (42) above is correct, then it sheds new light on example (19). On this interpretation, (19) can no longer be regarded as showing the relative scope of two quantifiers; rather, it involves one quantifier and one phrase whose semantics is effectively determined by contrastive focus. There is no interaction between two quantifiers in (19), hence the question of their relative scope does not arise. The only expression interpreted as a quantifier in (19) would be odin mal’čik and it would take a scope position at LF. ^[27] The fact that a contrastively focused phrase is reconstructed for interpretation below the QP would follow from a systematically lower position of focus with respect to quantifiers at LF. Therefore, examples like (19) do not provide any argument from scope facts for the reconstruction of the focused phrase in clause-final position.

It should be noted that, assuming the Copy Theory of movement (Chomsky 1993), the silent copy of KAŽDUJU DEVOCKU in (19) does occupy the clause-final (most deeply embedded) position. This, however, is not to be linked with a clause-final site of focus, but rather with the fact that the accusative object is the element which merges with the verb as first. An interesting case is (43), another of Neeleman et al.’s (2009) cases, allegedly showing the ∃ > ∀, *∀ > ∃ scope relation, in which the fronted focused phrase is a nominative subject. Here, the in situ position of the silent copy of the subject in the inner Spec.vP can still be argued to be below the surface position of the quantifier which has overtly moved to the outer Spec.vP, as schematized in (44) below. ^[28]

Furthermore, it is worth confronting Neeleman et al.’s (2009) hypothesis that focused quantified phrases are obligatorily reconstructed in clause-final position with a Polish example (45) below, in which the quantifier itself, and not its complement, carries focal stress:

The preferred scope relation in (45) is ∀ > ∃, *∃ > ∀, which imitates the surface order of quantifiers in this sentence. A contrastively focused QP takes scope over a non-focused QP, which contradicts the prediction that focus always has to be reconstructed in the lowest position. Thus, (45) aligns with the pattern of the Russian example (20) above, in which none of the elements carries contrastive focus.

The reason for the difference between cases like (19) and (45) may be related to the possible suppression of the semantics of quantifiers by the semantics of focus; this suppression apparently does not occur if the quantifier itself is focused. In cases when a QP with a focused quantifier co-occurs with another QP, as in (45) above, the relative scope of quantifiers falls back on the standard preference determined by their surface linear order. The rationale for such an outcome may be derived from the facts of focus projection discussed in Section 1.3. If Büring’s (2006) rule of Default Prominence is extended to apply to narrow focus, then focus can be projected to a QP only horizontally, i. e., from the complement of Q, but not vertically, from the Q head of QP (see footnote 6). Accordingly, a phrase like (46) becomes as a whole a contrastive “focus-phrase”, while a phrase like (47) does not, and its contrastive focus is confined to the quantifier itself:

Next, the difference in the extension of focus in (46) vs. (47) will have a bearing on their interpretation. Although both phrases are syntactically QPs, the former has the semantics of a contrastively focused phrase, whilst the latter retains the semantics of a quantifier. That this is actually the case may be corroborated by the following simple question–answer tests in (48) and (49), and a simple cleft-test in (50): ^[29]

In (48), A1 is a felicitous answer but not A2. This may be interpreted as follows: the question opens a set of alternatives (posits a variable) in the position of the complement of talked to; the answer is expected to provide a value for that variable. This is accomplished in A1, in which the QP is a “focus phrase” (it is marked [+F] in Selkirk’s [1995: 555] terms). In A2, however, focus is not projected from Q to QP, so that the QP is not marked [+F] and cannot provide a value for the variable posited by the question. Hence, answer A2 is infelicitous. The case in (49) is complementary to that in (48). Here, the question opens a set of alternatives for the head of QP, and not the whole QP, and the answer should provide a value for the variable in the position of the head of QP. This is accomplished in A1 but not in A2.

The pattern in (50) illustrates contrastive focus instantiated by cleft-sentences. In (50a), the non-quantifying DP the boys is a felicitous alternative for the QP with a contrastively focused complement of Q. But the same DP is not an appropriate alternative for the “cleft” QP in (50b) in which the quantifier carries focal stress. Example (50c) shows that the appropriate alternative in this case would be a similar QP with the quantifier stressed. Additionally, the complement in the alternative cannot be altered.

The facts in (48), (49), and (50) show that there is a difference in the semantics of the two types of QPs in (46) and (47); the former displays the semantics of a contrastively focused phrase, while the latter the semantics of a quantifier. It appears that for LF purposes a phrase must function either as a contrastive focus element or as a quantifier, but cannot take on both functions at one time. A result like this is expected if both quantifiers and foci are quantificational elements (see, e. g., Rizzi 1997) with conflicting requirements – they link different variables, and have different scope properties (quantifiers c-command foci at LF). In sum, the fact that focused QPs systematically scope below non-focused QPs, regardless of their mutual surface position, need not be linked with a unitary clause-final position of focus.

4.2 Split scrambling revisited

As discussed in Section 2.2.3, Neeleman et al. (2009) argue that focusing split scrambling applies in Russian only from a clause-final position. This claim should be confronted with a larger set of data. First, as shown below, the neutral position of the adjectival modifier of a noun is to its left: ^[30]

If the surface position of the adjective correlates with its position in the base structure, then the left-bound extractions observed in examples in (52) are from the left of the noun, not from a clause-final position: ^[31]

The above generalizations concerning word order and scrambling naturally extend from Russian to Polish. Thus, as seen in (53)–(55), Polish quite naturally allows split scrambling from a variety of non-clause-final positions:

Example (53) shows left branch extraction of a focused adjective, (54) features leftward extraction of the focused genitive phrase occurring before the adverbial adjunct, and (55) shows extraction of a pre-nominal numeral. ^[32] In sum, a larger set of data from Russian, supported by similar facts from Polish, shows that split scrambling of a focused element does not appear to be restricted to a clause-final launching site.

4.3 Heads of phrases in focus

For focus to be clause-final, it has to be phrase-final first, as schematically presented in (56):

In other words, in head first languages, for any clause-final phrase it holds that it is the complement of this phrase, not its head, which carries sentence focal stress. This is generally predicted by the aforementioned Nuclear Stress Rule (Chomsky and Halle 1968; Selkirk 1984; Halle and Vergnaud 1987; Cinque 1993). The situation becomes complicated, though, for cases of contrastive focus in which focal stress may be shifted to a largely unpredictable position (as widely discussed in the literature; see, e. g., Zubizarreta 1998; Reinhart 2006), as schematized in (57):

The difference between cases (56) and (57) is difficult to handle in a framework like Neeleman et al.’s (2009), in which information focus and identification (exhaustive, contrastive) focus are not distinguished, and in which any instance of focus exposed by nuclear stress should be linked to a clause-final position. Thus, it is left unexplained why the head of the fronted prepositional phrase and not its complement is focused in cases such as (58) below:

If the only position licensing focus was clause-final, then the only option for a contrastive stress on the fronted phrases should be (59), which is contrary to fact:

4.4 Complex Focus and Multiple Focus

The postulation of a single focus position in a sentence should also be confronted with cases in which two instances of focus co-occur in a single clause. As noted earlier, each instance of focus has to be appropriately licensed, which amounts to the restriction that each instance of focus in a clause needs to be linked with an operator (Krifka 1992; Rooth 1997; Krifka 2007; Krifka and Musan 2012; Hedberg 2013). The first case to consider is an instance of Complex Focus, exemplified by (60) and its Polish equivalent in (61):

The semantics of (60) is: “The only pair <x. y> such that John introduced x to y is <BILL, SUE>” (Krifka 2007: 35). This means that (60) is a case in which one operator binds a combination of foci.

The second case is Multiple Focus, which Krifka (2007: 35) defines as a case “in which in one and the same sentence, one expression introduces alternatives that are exploited in one way, and another expression introduces alternatives that are exploited in a different way”. Unlike in the case of Complex Focus, here each instance of focus is linked with a separate operator. A relevant example is (62), with its Polish counterpart in (63):

Leaving aside the problem of the semantics of examples involving Complex and Multiple Focus (see Rooth 1997), the very existence of Complex and Multiple Focus appears to refute the hypothesis of a single focus position per sentence. Although, the instances of Complex and Multiple Focus are not specifically dealt with by Neeleman et al. (2009), presumably the ban they propose on the co-occurrence of contrastive and new information foci should apply to them as well. If focus is unconditionally linked to a single (clause-final) position, instances of Complex and Multiple foci should not be licensed, which is contrary to fact.

4.5 Ex situ new information focus without contrast

In Neeleman et al. (2009), it is predicted that the only instances of ex situ focus in Russian are cases involving contrast. This follows from the assumption that the only factor responsible for A’-fronting of a focused element is the presence of the feature [+contrast]. Numerous examples, however, can be observed in which a non-contrastive phrase with the interpretation typical of new information focus is fronted, as illustrated below:

Cases like (64)–(66), discussed at length by Fanselow and Lenertová (2011), are referred to as instances of Subpart Focus Fronting, in which the fronted phrase is only a subpart of a larger focus projection extending to the whole verb phrase. ^[33]

Another category of counterexamples are cases of ex situ information focus in which the focused element appears to the left of the verb, but does not reach as high as the front of the clause, as exemplified in (67):

4.6 In situ, non-clause-final information focus

A further complication for Neeleman et al.’s (2009) descriptive statement that the position of focus is uniformly clause-final arises from the felicity of exchanges like the one in (68):

The type of focus in (68b) can only be identified as (new) information focus because it is linked to a variable expressed by čto ‘what’. The sentence in (68b) is perfectly well-formed, and the focus extends to the whole VP, as represented in structure (69): ^[34]

The situation described in (68b) can be readily explained in structural, not linear terms. The accusative object in (68b) is deepest embedded, but not clause-final, as illustrated in (70):

The fact that nuclear stress falls on the object and not on the final PP in (68b) follows straightforwardly if one adopts a derivational approach such as Reinhart’s (2006), in which the addition of stress grids (to a metrical tree) is mapped on the formation of syntactic trees. Roughly, nuclear stress is automatically placed on a word which carries most stress grids at the completion of a derivation. In (69), the DP jazz-pianista stays longest in the derivation and therefore collects most stress grids.

4.7 Non-fronted contrastive focus

Finally, the cases of contrastive fronting are regarded as dislocations to the left-most position, presumably Spec. ContrastP (see Neeleman et al. 2009: 42). ^[35] This, however, gives rise to the question why cases of in situ, as well as ex situ but non-left-most, contrastive focus should ever be allowed. ^[36] That this possibility exists is illustrated by the following examples:

Example (71) is a case of short split scrambling of a focused phrase to a position below the subject (topic), and (72) is an instance of in situ contrastive focus. In fact, it is common for contrastively focused phrases to occupy positions lower than the fronted one. The only syntactic type which systematically requires fronting in Russian and Polish are clefts, as illustrated by (73) and (74) below: ^[37]

Leaving aside the issue of the syntax of clefts, it may be finally observed that examples (73) and (74) are truth-conditionally equivalent with their non-cleft equivalents given in (75) and (76):