1. Introduction: Focus in description and explanation
In recent decades, the notion of focus has become a central explanatory mechanism for a wide variety of phenomena across linguistic disciplines and across the boundaries of theoretical frameworks. It has been instrumental in the development of the architecture of most current linguistic theories and has been adduced to account for phenomena as diverse as assignment of sentence stress, word order, quantifier scope, and pronoun interpretation, among others. Such recognition of the wide-reaching influence of information-structural factors in the relationship between linguistic form and meaning is undoubtedly to be welcomed.
It does not follow from this that there must be a single, universal linguistic category ‘focus’, but this is how focus is widely treated: diverse structures in different languages – prosodic, syntactic and morphological – are regularly interpreted as being manifestations, or ‘realisations’ of focus. This is often implicit, in unquestioning comparisons of structural phenomena that may share little but the perception that they express focus. But it is not hard to find explicit statements to this effect, too. For example:
[F]ocusing is variously reflected in prosodic phrasing …, constituent ordering, via special focus morphemes, and perhaps in some cases not at all. … languages just choose some aspect of their grammatical structure, prosodic, syntactic, or morphological, to realize focus. (Büring Reference Büring2010: 177)
[T]he heterogeneity of (non-neutral) focus-marking mechanisms attested in natural language – positional, prosodic and morphological – differ only in terms of superficial realization of an identical feature. (Kidwai Reference Kidwai1999: 224)
While some analysts apparently treat the universality of focus as established empirical fact, others overtly prejudge the issue by raising this assumption to the status of a methodological principle. Thus, Erteschik-Shir (Reference Erteschik-Shir2007: 40) states that ‘[a]ny definition of focus must measure up to the requirement that it be universal’. Similarly, Vallduví & Engdahl (Reference Vallduví and Engdahl1996: 459) say that ‘[a] set of information-packaging primitives that are cross-linguistically sufficient and methodologically useful needs to be identified’. It is clear that Vallduví & Engdahl mean that the set of information-packaging primitives should be both minimal and directly involved in determining the structures of the world's languages: they proceed simply to claim that a variety of structural effects in a range of languages are manifestations of these simple primitives.
Such systematic attempts to develop a cross-linguistic notion of focus are among the most obvious examples of work that is based on the assumptions that we identify and criticise in this article. However, we believe that similar reasoning is employed very widely in linguistics: a great deal of work assumes the existence of a primitive called focus and applies it in cross-linguistic comparison. This is despite the fact that the notion of focus is notorious for its vagueness: ‘Terminological profusion and confusion, and the underlying conceptual vagueness, plague the relevant literature to a point where little may be salvageable’ (Levinson Reference Levinson1983: xi). In part, this vagueness is merely an artefact of unsettled terminology (see Gundel Reference Gundel, Bosch and van der Sandt1999, Kruijff-Korbayová & Steedman Reference Kruijff-Korbayová and Steedman2003, Gundel & Fretheim Reference Gundel, Fretheim, Horn and Ward2004, among others). But this is not the whole story. Focus is defined in at least a dozen different, often conflicting ways. The definitions range from simple reference to new/unexpected information to the detailed formal machineries of Alternative Semantics and the Structured Meanings approach. What is more, imprecise and/or divergent definitions inevitably lead to an imprecise extension of the category. Any careful analysis of the relevant literature, like that conducted by Hedberg (Reference Hedberg2006), would reveal that it is unclear which linguistic phenomena count as focus and which do not, so that, apart from some very central examples, the same structural pattern might be treated as ‘focus’ or ‘non-focus’, often without any explanation either way.
We believe that the terminological and notional confusion surrounding the notion of focus is no mere historical accident, but reflects real linguistic diversity, our understanding of which has been obscured by inappropriate conceptions of categorisation and scientific generalisation. That is, focus is an inherently problematic category, which has been used to draw together phenomena in the wrong way: as instances of a single underlying entity, as opposed to potentially independent entities that produce interestingly similar effects.
The main purpose of the paper is to argue against the prevailing essentialist views of focus as a stable, discrete entity that is simply ‘realised’ through the grammars of different languages. We point out how such views are questionable at a fundamental theoretical level, and also how they are related to dubious methodological practices with negative consequences for empirical coverage, which we illustrate in some detail.
At the same time, we argue that the phenomena commonly comprised by the notion of focus are comparable due to the similarity of interpretative effects that they evoke. This is different to positing an underlying identity that may be encapsulated in a cross-linguistic category: as in any domain, the existence of comparable outputs should not be assumed to imply identical inputs, nor identical mechanisms (the relevant mechanisms in this case being grammatical strategies for encoding meaning). We see the ultimate purpose of cross-linguistic research on focus in identifying reasons for the similarity of these effects and modes of convergence. Compared to the strategy of spotting apparent instances of pre-established categories across different languages, we believe this is not only better justified by fundamental theoretical considerations, but also a more interesting exercise and in principle more explanatory.
A number of our arguments below are of a broad conceptual and methodological kind, and many readers may see their force extending well beyond the matter of focus. We would not discourage this line of thought. On the other hand, some of our arguments are quite specific to focus, and we do not mean our point to stand or fall on the reader's opinion of other cases of categorisation. We hope that those who would defend other traditional kinds of universal linguistic category may be persuaded that the particular case of focus is highly problematic and that it might therefore stand at least as a cautionary example for work in other areas.
We begin by considering in the abstract the questions that Vallduví & Engdahl implicitly raise but fail to examine: What is it for a linguistic category to be both cross-linguistically sufficient and methodologically useful? We then turn to an empirical demonstration of the diversity of putative focus phenomena and the inadequacy of common attempts to capture this diversity – including the widespread strategy of splitting focus into ‘ordinary’ and contrastive types, and the popular semanticists' notion that focus is at heart the invocation of alternatives – before returning to more conceptual issues in the conclusion.
Given the nature of these aims, it is important to pre-empt one potential objection at the outset. It has been argued that the accumulation of counterexamples to postulated language universals equates to mere butterfly collecting (originally Chomsky Reference Chomsky1979: 57) or a freak show (Duffield Reference Duffield2010: 2673), with no theoretical relevance whatsoever. Whether or not this charge is justified with regard to some parts of the literature, we wish to emphasise that it cannot be levelled at the present work. We do include much linguistic evidence that the current notions of focus are unable to account for the documented diversity, but the essence of our argument is much more abstract, centring not on the facts of diversity per se, but on specific ways in which common analytical strategies fail to account for such diversity, and do so more or less inevitably. That is, we argue that the very idea of a uniform category of focus is flawed, we trace this to underlying assumptions about linguistic theory and methodology, and we illustrate it through systematic critiques of some prominent work on focus.
In any case, there is perhaps some irony in the accusation that highlighting empirical diversity equates to butterfly collecting. It is arguably the more conventional approach of freely positing universal linguistic categories that leads to mere ‘collecting’ behaviour, such that analysis may consist in placing linguistic structures in the box marked ‘focus’ (or perhaps in a few boxes marked ‘presentational’, ‘contrastive’, ‘corrective’, and so on), on the basis of some relatively superficial observations. In contrast, a concern for the extent of linguistic diversity can encourage investigation of deeper, more explanatory generalisations.
In the following section, we continue to consider relevant issues at a broad theoretical and methodological level, before turning to the more empirically driven parts of the article.
2. Focus and linguistic categorisation
What is a cross-linguistic category? What is it for? This will depend to some extent on the theoretical stance and analytical purposes of the individual linguist, but below we set out some very general criteria for a useful theoretical entity and we argue that any fixed, universal category of focus fails to meet them. First, let us try to separate some possible interpretations of the term ‘category’, to clarify the nature of our claims.
2.1 Interpretations of the category of focus
One interpretation of the claim that focus is not a universal linguistic category might be the following: focus is not a grammatical category, but it may still be considered an interpretive category. Although we are very keen to differentiate interpretive effects from grammatical machinery, this dichotomous labelling of linguistic categories is far too simplistic. For one thing, any generally applicable conception of focus is inherently meaning-based, in one sense or another of ‘meaning’ – whether in terms of discourse newness, interpretive procedures (as in Enric Vallduví's work), truth-conditional semantics (as in Rooth Reference Rooth1992, Reference Rooth and Lappin1996) or some other conception of meaning. Even when focus is treated as a formal syntactic feature (e.g. Rizzi Reference Rizzi and Haegeman1997), it is essentially meaning-based in our terms. The motivation for positing such a feature and all evidence for its existence lie in the apparent effects on interpretation of using certain linguistic structures (see Abusch Reference Abusch2008: 320). It is thus quite normal to have linguistic categories that are in part interpretive and in part grammatical – grammar is, after all, a system that pairs forms and meanings.
Moreover, any simple differentiation of grammatical and interpretive categories loses all clarity in the context of the diverse theoretical proposals that must be considered. For example, focus in the frameworks of Vallduví (Reference Vallduví1992) and Erteschik-Shir (Reference Erteschik-Shir1997) is claimed to inhabit an autonomous, extra-syntactic system of information structure, which is based in information-processing strategies, and so may seem not to be a grammatical category as such. But this idea of information structure effectively treats it as a component of a modular grammar, with fixed and universal features (including [+focus]) participating in the definitively grammatical task of effecting systematic form-meaning relations (albeit independently of other postulated grammatical operations).
The crucial point is not whether we call some category interpretive or grammatical, but what role it is assumed to play in linguistic analysis. To recognise some particular interpretive effect, and perhaps to label it ‘focus’, is one thing. It is another thing to assume that all linguistic phenomena that produce this effect are ‘realisations’ of something called focus, or in any other sense defined by it or dedicated to its expression (whether in formal or typological work; see further Haspelmath Reference Haspelmath2010). In the latter case, the structure of the language is assumed to be sensitive to the presence of that category, and to produce the interpretive effect in question only in this way. This implicitly confers on that category the status of being a linguistic, not merely interpretive, category. For our purposes, then, there is no substantial difference between an explicitly grammatical category – even a core syntactic feature – and an interpretive category that gets ‘linguistically realised’, or one that is taken to be definitive of a class of structures, as in typological work.
We stress this point because of confusions that can be seen in the literature. For example, Zimmermann & Onea (Reference Zimmermann and Onea2011) claim that they treat focus outside the grammar, viewing focus as essentially an interpretive phenomenon, rather than a ‘genuine linguistic category’ (p. 1658). In practice, however, they make the standard assumption that focus must be something that is ‘realised’ in different ways in different languages, and this entails the assumption of a universal linguistic category, in any meaningful sense of the term. For something to be linguistically realised, that thing must exist at some theoretically relevant level. Focus is thus implicitly assumed to be some discrete and stable entity and, moreover, this must be a kind of entity to which linguistic structure is sensitive (see further Section 5 below).
In order to facilitate a coherent discussion of these matters, we need a sharper distinction between the interpretative effects of linguistic utterances, on the one hand, and the linguistic entities or mechanisms that trigger them, on the other. In arguing against the treatment of focus as a universal linguistic category, we certainly do not mean to deny the possibility of addressing relevant interpretive effects systematically. We take it that (i) it is often important for human communicators to indicate the status of different pieces of information within an utterance or a text, (ii) there are at least some discrete and identifiable components of such ‘informational meaning’ which are universally available to human minds, and (iii) people often do convey such meanings through the use of particular linguistic structures. But this does not justify the assumption that one such component of meaning has a privileged status as something that is always the core denotation of the linguistic structures it is associated with, and something that all the languages are sensitive to. The evidence, as we outline below, is that so-called focus constructions in different languages may be underlyingly quite different things, though they show overlapping interpretive effects.
The problem is that this diversity is inevitably suppressed by methodological practices which are common to most forms of linguistic analysis. Put bluntly, but (as we illustrate below) not unfairly, the common analytical strategy is as follows. First, identify linguistic phenomena that, in use, are systemically accompanied by a particular interpretive effect (say, the evocation of alternatives or an indication of how an assertion relates to some implicit or explicit question). Then label each of these phenomena as realising/being dedicated to that interpretive effect (i.e. ‘a focusing strategy’ or ‘focus construction’), thus unifying cross-linguistic phenomena. It should be clear that this practice is logically flawed, equating effects (interpretations, as outputs of linguistic mechanisms) with their own causes or triggers (in the form of linguistic categories that are inputs to linguistic mechanisms, or perhaps elements of the mechanisms themselves).Footnote 2 It should also be obvious that this analytical strategy risks distorting our understanding of the empirical landscape by assuming that the mere presence of one particular interpretive effect suffices to characterize all of the constructions that manifest it, irrespective of any further features of an individual construction.
Importantly, this kind of reasoning is at the root of the most common procedure to identify and define focus. This is via ostension, as summarised in the following quote:
I assume, with much of the literature, that we can identify focus via the pragmatics. Concretely, certain (semantic or pragmatic properties of certain) contexts systematically trigger focus. … there is a common extensional core of contexts that people seem to agree trigger focusing. (Büring Reference Büring2010: 178)
The key focus-triggering context is question–answer pairs; they are the basis of the question–answer (Q–A) test, according to which the material in the answer that corresponds to the question word is identified as focus (see Kasimir Reference Kasimir, Ishihara, Schmitz and Schwarz2005 and Büring Reference Büring, Ramchand and Reiss2007: 448 for more detail). Question–answer congruence has a dual status in current research: it is both a practical heuristic for identifying focus and a formalisation of what has become known as discourse anaphoricity of focus, i.e. the sensitivity of prototypical focus structures to context (Rooth Reference Rooth2008). A congruent question (explicit or implicit) subsumes all types of contexts in which a particular focus structure is felicitous.Footnote 3 Ultimately, this necessarily results in the assumption that, if a certain linguistic structure regularly shows Q–A congruence (i.e. the same type of context sensitivity), then it must be an instance of focus, and, by implication, essentially identical at some level to all other structures in other languages that pass this test. This is how the categorial status of focus receives its empirical justification. In Section 3 below, we adduce a number of clear counterexamples to this effect-based establishment of the category of focus and we discuss the theoretical problems it poses in Sections 4 and 5. This procedure is fallacious in just the ways discussed above: it presupposes that a pragmatic effect – the regular appearance of a structure in certain contexts – necessarily follows from the existence of a primitive category (see Matić Reference Matić2009), disregarding the possibility that this effect can have multiple causes and neglecting the differences that may exist between structures that display the same effect.
To be quite clear, we do not wish to suggest that focus effects are never directly encoded in the grammars of languages. On the contrary, there are clearly many linguistic structures whose primary communicative purpose relates to the relative status of certain bits of information. But we do not assume that these structures must have one unitary underlying meaning. In other words, we do not think that any one definition of focus need be basic or universal, and we see no basis for any such assumption – here the cross-linguistic data strongly support our position, as we indicate below. Nevertheless, it is clear that this assumption is very commonly made. The literature is teeming with attempts to formulate the right ‘theory of focus’, whether in formal semantics (e.g. Alternative Semantics, Structured Meanings) or in less formal approaches (in terms of newness, information processing procedures or other concepts); our contention is that, from a linguistic point of view, the term focus may simply not denote a cohesive phenomenon about which to theorise.
2.2 The notion of category
Having looked at what a linguistic category may be, let us now turn to the purpose of proposing such categories. We take it that practically all kinds of cross-linguistic analysis share one general aim: to establish, and in some sense account for, the extent and nature of the diversity of human languages. To achieve this, the analyst requires categories, or concepts, that perform certain functions:
(i) They must facilitate the initial identification of meaningful points of comparison across languages.
(ii) They must (by definition) unify phenomena in different languages at some appropriate level of abstraction, but they must do this without imposing excessive uniformity at inappropriate levels, in contradiction of the data.
(iii) They must play some kind of explanatory role, in the sense of predicting the range of possible variation across languages. Note how this point is related to the previous one. In effect, the ‘appropriate level of abstraction’ for positing some unifying category, as referred to in (ii), is determined by the degree to which any putative common entity helps to account for the variation that is found in the data. Conversely, attempting to unify phenomena at the wrong level of abstraction will typically obscure or suppress variation, not account for it.
We contend that the category of focus, as it is commonly invoked in studies of many kinds, falls a long way short of these requirements.
For one thing, focus is often poorly defined, meaning that (i) and (ii) above are not properly achieved. Individual studies may propose some particular definition, but too often focus is introduced with a vague reference to some concept such as new information, or is simply assumed to be a familiar and self-explanatory category, despite the notorious terminological instability noted above. Of course, the existence of sloppy analytical practice does not in itself entail the inappropriateness in theory of a linguistic category of focus. However, it may mean that much of the apparent evidence in favour of such a category is built on unstable foundations.
Another major problem is the excessively superficial nature of focus as a putatively primitive notion in cross-linguistic analysis. As we have already emphasised, the observation that two linguistic structures appear to relate to (say) new information or the relevance of alternatives concerns only the effects of interpreting that structure. Such interpretive effects may be partly or wholly inferred on the basis of some other encoded meaning and, indeed, there are good reasons to expect that the particular effects that are commonly associated with the notion of focus may be created in various different ways. Consequently, requirements (i) and (ii) are not satisfied by the notion of universal focus, as it may (and indeed does) turn out that at least some points of comparison are not meaningful, and that the phenomena unified by the category of focus do not belong together. In this context, it is significant that focus effects are by nature closely connected to context (newness, the existence of alternatives) and to ‘speech act’ or interpersonal meaning (assertion, emphasis, contrast). That is, focus has a great deal to do with what is achieved through the use of language. This in itself should make us question whether it is always right to see focus as part of language.
There are also more general reasons to question the assumptions that underlie this conception of focus (bringing us back to point (iii), above). To posit a category like focus is to define one's basic theoretical entities in the same vocabulary as one's descriptions of superficial effects. The methodology that produces such entities is essentially that which Webster & Goodwin (Reference Webster and Goodwin1996: 24) counsel against (in work on the philosophy of biology): one whereby ‘we start with given, concrete particulars and, by a process of comparison, abstract the “common features”, discarding the remainder’. Far from predicting the particular range of variation found in the data, this approach simply suppresses much of that variation. It may have some unifying force, but at the cost of distorting the empirical landscape and potentially obscuring deeper insights.
This point is relatively subtle and decidedly abstract, so let us clarify with a non-linguistic (and intentionally simplistic) example. The leaves of most plants, many kinds of algae, and blooms of cyanobacteria are green. Descriptively, this is an accurate and appropriate generalization, yet it would plainly not be explanatory to appeal to a primitive category of greenness when analysing the nature of these organisms and their interrelations. Indeed, to do so would be an active distortion, suppressing important kinds of variation – not just because ‘green’ is a very broad and obviously superficial characterization of the relevant external effects but because the chemical causes of greenness vary in these cases (involving quite distinct kinds of chlorophyll). To find useful generalizations here, we need to look beyond static components of a superficial description, to consider the processes that produce them and the contexts in which they occur. Greenness in these organisms is a by-product of photosynthesis and is ultimately explicable in terms of the chemical properties of chlorophyll. Once this is understood, our analysis of these cases can be unified with organisms that employ non-green pigments to effect photosynthesis, while unrelated greenness (like that of tree frogs) can be excluded. In contrast, any analysis that took greenness itself to be fundamental would both overgeneralise and undergeneralise wildly.
Crude as this example may be, the general lesson should be clear, along with its particular relevance to notions like focus, which, as noted above, are inherently connected to superficial usage effects. It is for this reason that, as we show below, focus likewise lacks the discriminatory power to delimit a natural class, i.e. it overgeneralises, and lacks explanatory power, i.e. it cannot define and predict the range of attested variation.
Ultimately, if a unifying theoretical entity such as a cross-linguistic category is to be explanatory (or even useful), it should participate in a chain of causal reasoning from the existence of this entity, through any processes it triggers or participates in, and through its interactions with any relevant external factors, to the different effects we see in the data. This is not apparent in the use of a category like focus, which merely abstracts away from diversity, without offering any way back to accounting for it. The combination of the dictum that our notion of focus ‘must be universal’ and the methodology of defining focus in terms of interpretive effects leads inevitably to a theory that is based in identifying the ‘lowest common denominator’, the consequences of which are a distorted and unhelpful picture of the empirical landscape, as we show below. From here, the only possible way to reintroduce the richness of the data we want to account for would be to relax the universality requirement, allowing a potentially empirically adequate but inevitably unexplanatory multiplication of categories, ultimately just re-stating the data.
3. The diversity of ‘focus’ phenomena: Prima facie evidence
In this section, we illustrate the inadequacy of any unified notion of focus to capture the diversity of relevant linguistic phenomena. Part of doing so is simply a demonstration that focus effects do differ across different structures in different languages; that quite distinct functions are performed in the different cases, contrary to the idea that they are simply separate realisations of a single underlying focus primitive. We are aware that some readers may find the mere existence of micro-variation across the interpretations of focus structures to be unremarkable. For reasons outlined above, we would maintain that this is theoretically significant. If two structures demonstrably display focus effects such as Q–A congruence, relevance of alternatives, etc., the effect-based categorisation predicts that it must belong to the same class of entities, ‘focus’. It is this line of thought that our examples refute: there is strong prima facie evidence that the ‘focus’ interpretations in question not only show variation but are arrived at in different ways, from different sources, in spite of any overlaps in superficial effects.
We illustrate our arguments with phenomena that have for the most part been extensively discussed in the focus literature, frequently in explicitly comparative work, and generally with a clear sense that they manifest some universal category. Moreover, all pass common diagnostic tests for focushood, most notably being associated with congruent answers to explicit or implicit questions. It is therefore of some importance if on closer inspection these structures prove to differ from each other, or from other putative focus structures, in significant ways.
One obvious and much discussed way in which focus phenomena may vary is that some appear to involve the expression just of new information (or some related meaning), while others appear to have some kind of inherently contrastive force. Widely cited examples of each are English focal pitch accenting and Hungarian ‘focus movement’, respectively. Both of these must be judged to instantiate focus according to the Q–A criterion. Hence the only felicitous answers to the context question in (1) are (1a) in English (where capital letters indicate focal accenting) and (1b) in Hungarian (the ‘focus position’ being immediately preverbal and triggering the postposing of any ‘verbal modifier’ element, such as the particle meg).Footnote 4
(1) [Who did John invite?]
(a) John invited [MARY]focus.
(b) János [Marit]focus hívta meg.
János Mari called ptl
‘It was Mary (and no other contextually relevant person) who John invited.’
As the translation in (1b) indicates, even as the Hungarian ‘focus position’ is the only unmarked question-answering form, it is widely held also to encode some contrastive element, of a kind that resembles the interpretation of the English it-cleft. More specifically, it is typically said to express ‘exhaustive listing’ of those individuals who bear the property that is expressed by the rest of the sentence (here, having been invited by John). This reading is not necessarily conveyed by the English focal pitch accent. Thus, despite the two structures being used in the same context, there is clearly a significant distinction here, immediately calling into question the idea that there is a thing called focus that is instantiated by prosody in English and syntactically in Hungarian. These observations are by no means new. Still, this has not stopped a wider perception that Hungarian gives us a straightforward case of ‘focus through syntax’, as reflected in textbooks (Saeed Reference Saeed2008) and comparative approaches to information structure (Drubig Reference Drubig2003, Erteschik-Shir Reference Erteschik-Shir2007).
We return in Section 4 to the idea that there is a fundamental division between ‘ordinary focus’ and some contrastive kind. It is in any case not the only kind of variation in focus, nor the most striking example. Consider the Somali morpheme baa. On the basis of standard elicitation tests, this appears to indicate something rather similar to the Hungarian ‘focus position’, with baa obligatorily accompanying a narrow (NP-sized) focus, as indicated by the Q–A test in (2) (Saeed Reference Saeed and Nolan2004: 262), and also being available to indicate what is intuitively some form of contrastive focus, as in (3) (Saeed Reference Saeed1999: 233; see also Saeed Reference Saeed1984 and Tosco Reference Tosco2002, and references therein).
(2) [Where is the girl?]
(a) (Inántii) Qólkáas bày kú jirtaa.
girl.the room.that BAA.she in is
‘(The girl), She's in [that room]focus.’
(b) #Inántíi bàa qólkáas kú jirtá.
girl.the BAA room.that in is
‘[The girl]focus is in that room.’
(3) Tól iyo fardó, tól bàan doortay.
kinfolk and horses kinfolk BAA.I chose
‘(Of) kin and horses (i.e. wealth), I chose [kin]focus.’
The real significance of Somali baa, however, becomes apparent only when texts are examined, as in Tosco (Reference Tosco2002) (see also Biber Reference Biber1984).Footnote 5 Outside of constructed contexts that are designed to elicit focus readings, it becomes clear that the majority of uses of baa are of a quite different nature. The sentence in (4b) is a typical example:
- (4)
(a) Nin baa waa ari badan lahaa.
man BAA time sheep many had
‘There was once a man who had many sheep.’
(b) Arigii baa cudur xumi ka galay.
sheep.the BAA disease bad in entered
‘The sheep fell prey to a bad disease.’ (Tosco Reference Tosco2002: 36)
The baa-marked phrase in (4b) is plainly not the focus of the sentence, in the sense that it corresponds to the wh-word in any implicit context question. If anything, it looks more like a special kind of topic marker, roughly indicating a change of topic – and as such performing a role that primarily relates to textual coherence, rather than marking the newness or contrast status of information in any straightforward sense.Footnote 6
Somali baa is not only notable for its sheer divergence from more famous examples of focusing, but also for its apparent grammatical properties. Saeed (Reference Saeed1999) notes that in its distributional properties baa resembles parts of the Somali mood system and, following suggestions of Ajello (Reference Ajello, Ajello and Sani1995), that it shows mood-like elements to its interpretation as well: it seems to relate to some form of realis meaning, indicating the true (as opposed to hypothetical, conditional, or desired) existence of its complement. It seems reasonable to suggest that the focusing capabilities of baa are built on this underlying realis mood-marking function, and this in turn can be connected to at least one existing definition of focus, as an assertion-creating device (Lambrecht Reference Lambrecht1994). The kind of realis mood in question here is directly related to assertion, the defining feature of which is the speaker's commitment to the truth of the proposition expressed (Jary Reference Jary2010). What is called focus in Somali, then, may well turn out to be based on a type of mood with idiosyncratic properties. In any case, it has a series of distinctive properties that are simply not accounted for if it is analysed, on the basis of standard elicitation tests, as an instantiation of a purported universal focus primitive. Nor is there any obvious reason to assume that these other properties are any less fundamental or essential to the nature of baa than are traditional conceptions of focus.
Quechua provides another instance of an entity with a focus-like face value but also indications of quite distinct underlying semantics. The morpheme -mi/-n has been argued to encode narrow focus when attached to the focused element of the clause (Muysken Reference Muysken and Kiss1995, Sanchez Reference Sanchez2010); this is illustrated by the following Q–A pair:
- (5)
(a) Pi-n wasita ruwarqan?
who-MI house built
(b) Wasita-qa Pidru-m ruwarqan.
house-top Pedro-MI built
‘Who built the house?’ – ‘[Pedro]focus built the house.’ (Sanchez Reference Sanchez2010: 31)
This same morpheme is a part of the Quechua evidentiality system. As argued by Faller (Reference Faller2002: 140), it ‘indicates that the speaker has best possible grounds for making his or her statement’, via direct evidence or otherwise:
(6) Pilar-qa t'antata-n mikhurqan.
Pilar-top bread-MI ate
p = Pilar ate bread & speaker saw that p (Faller Reference Faller2002: 18)
Whether a given use of -mi/-n conveys focus or direct evidentiality seems to depend on pragmatic factors (see Faller Reference Faller2002: 150). Evidentiality is not obligatorily marked in Quechua, and the direct evidentiality (i.e. the best possible evidence for making a statement) is the default interpretation of all sentences which lack an evidential marker. Accordingly, mi/-n is used when direct evidence is at stake for any reason, and this commonly results in a special kind of emphasis. One possible reason to explicitly mark an assertion as based on best possible evidence is to make manifest its truthfulness (see Davis, Potts & Speas Reference Davis, Potts, Speas, Gibson and Friedman2007 on the strengthening effects of evidentials). Since Quechua allows for different scopes of the evidential, the veridicality claim may pertain to the whole proposition expressed by the sentence or to various parts thereof. All this has an effect of explicit marking of different scopes of assertion, much like the traditional notion of different focus scopes.
In other words, focus in Quechua is plausibly a reading – perhaps partially inferential and partially conventionalised – of the direct evidential, in much the same way that Somali focus may be an effect of applying the realis mood to propositions and elements of propositions.Footnote 7 In this connection, it is notable that the ‘predicational’ analysis of the Hungarian focus position (Wedgwood Reference Wedgwood2005, Reference Wedgwood2006; É. Kiss Reference É. Kiss2006) also relates focus-related phenomena to a grammatical mechanism that in effect creates a certain kind of assertion. These three phenomena are strikingly similar conceptually, while having obviously distinct grammatical bases. At the same time, the specific functions that they fulfil seem to be quite different (if overlapping), precluding the possibility that they have simply grammaticalised a pre-existing, universal category of meaning in different ways. Whereas Hungarian preverbal focus is claimed to be truth-conditional and encode exhaustive listing, baa in Somali seems to be predominantly used as a discourse-organising device, and neither of these meanings/functions is attested in Quechua.
Note that we do not intend to promote some assertion-based approach as a general theory of focus. Rather, this discussion illustrates the interrelatedness of many of the concepts that surround the notion of focus (newness, contrast, assertion, and so on), and reminds us that the question of which (if any) is basic and which derived need not be answered the same way for all putative focusing strategies. While the particular linguistic phenomena discussed here are suggestive of some form of relatively abstract unity among themselves, the identification of such deeper and potentially highly explanatory generalisations depends on the abandonment of any simplistic attempt at a general definition of focus. Meanwhile, other examples of putative focusing strategies (including most of those discussed below) give no indication of fitting into the same pattern.
Focus-related meanings can actually be rather more complex. A famous example is the Bantu language Aghem (Hyman Reference Hyman1979), which, in addition to the ‘neutral’, unmarked sentence structure, has at least five distinct structures with distinct semantic and pragmatic properties. According to Watters (Reference Watters1979) (see also Hyman Reference Hyman, Fiedler and Schwarz2010, Hyman & Polinsky Reference Hyman and Polinsky2010), focus structures in Aghem vary along three dimensions: (i) the method of changing the state of knowledge of the interlocutor, (ii) the number of alternatives satisfying the property expressed by the non-focus part of the sentence, and (iii) the type of these alternatives. The state of the interlocutor's knowledge can be changed either by adding a new proposition to his context set (Watters' assertive focus) or by replacing one (counter-assertive); the alternatives may appear in singleton sets or as multiple member sets; they can be entities/propositions, or they can be truth values. Thus, Aghem has specialised (and unrelated) encodings for, say, corrective polarity focus, corrective focus on entities encoded as NPs, and simple assertive exhaustive listing. Focus in Aghem turns out to be distributed along parameters which a unitary English- or Hungarian-style system leaves completely underspecified.
Perhaps an even more interesting case is that of Tura, an Eastern Mande language of Ivory Coast (Bearth Reference Bearth1992, Reference Bearth2005). Alongside a ‘neutral’ structure with no formal marking of focus, Tura has two basic focus types, F1 and F2, which both pass the Q–A test (Bearth Reference Bearth1992: 80). These do not seem to correspond to any current subdivision of focus meanings (even the more elaborate taxonomies, as in Dik et al. Reference Dik, Hoffmann, Jong, Djiang, Stroomer, Vries, Hoekstra, van der Hulst and Moortgat1981). In fact, in many contexts F2 resembles what has been called contrastive topic in German and English (Büring Reference Büring, Bosch and van der Sandt1999), whereas F1 seems to be somehow exclusive. Bearth (Reference Bearth1992, Reference Bearth2005) reduces this difference to different types of alternatives F1 and F2 encode: F1 evokes alternatives of the same type as the denotation of the F1 marked element, whereas the alternatives evoked by F2 are sets to which the denotation of the F2-marked element belongs. The application of F1 thus often results in an implicature of exhaustivity or of the irrelevance of other alternatives and, in a further inferential step, the strengthening of an assertion, whereas the inference F2 triggers is that of alternative propositions possibly being true, i.e. F2 often weakens the assertion. This is nicely illustrated by the following examples (Bearth Reference Bearth2005: 11ff.; the symbol +> reads as ‘implicates’):
- (7)
(a) Wélé ké i g
n (unmarked)wealth pred.be you with
‘You have wealth (=the bride price)’ + > ‘You can visit the bride's parents’
(b) Wélé-' i g
n le (F1)wealth-F1 you with tm
‘You have wealth...’ + > ‘You can therefore be confident.’
(c) Wélé-le ké i g
n (F2)wealth-F2 pred.be you with
‘You do have the wealth, but…’ + > ‘You seem to lack many other qualities.’
While the unmarked sentence is neutral as to the alternatives to possessing wealth, the F1 sentence highlights the difference in importance between wealth and its alternatives in the context of finding a bride. F2, on the other hand, evokes the set of the typical bridegroom qualities to which wealth belongs and, by non-application of the predicate ‘have’ to them, weakens the assertion. F2 and F1 can even be combined (Bearth Reference Bearth2005: 13), and here the divergence from standard notions of contrastive focus and contrastive topic shows up especially clearly:
(8) Wélé-le-' i g
n lewealth-F2-F1 you with tm
‘Well, yes, but it's the bride-price that counts, and that you have.’
+ > ‘You lack other qualities, but only wealth counts, so you may be confident.’
Here, thanks to F2, the set of the desirable bridegroom qualities is evoked, and the non-application of ‘have’ to them implies the bridegroom's lack of these qualities; at the same time, the application of F1 denies their relevance in the context.
Aghem and, even more palpably, Tura, exemplify the extent to which what has been called focus can be subdivided. Moreover, Bearth's analysis of Tura reveals how highly idiosyncratic these subdivisions can be: neutral, F1 and F2 sentences are distributed along (and beyond) the meaning space of the assumed focus category in a way not attested elsewhere.
Thus, focus variability itself comes in many forms – partial overlaps (Hungarian vs. English), different source meanings (Somali, Quechua), or idiosyncratic subdivisions (Aghem, Tura). Many other variants are conceivable, and indeed attested. Due to lack of space, however, we shall refrain from simply expanding the catalogue of foci and turn to the ways the variation has been dealt with in the literature (and indeed our arguments here invoke further notable examples of focus diversity).
4. The splitting strategy
As noted in the previous section, it has become common to assume that languages may feature one or both of two kinds of focus. As ever, details vary between frameworks, but it is widely held that there is some basic split between ‘ordinary focus’ and a specially contrastive kind (e.g. Rizzi Reference Rizzi and Haegeman1997, Vallduví & Vilkuna Reference Vallduví, Vilkuna, Culicover and McNally1998, Drubig Reference Drubig2003, Neelemann et al. Reference Neelemann, Titov, Koot, Vermeulen and Craenenbroek2009). It is assumed that this splitting of the unified category can account for obvious differences in the behaviour of purported focus structures across languages. In this section, we demonstrate that this approach offers little improvement in relation to the data, compared to a monolithic notion of focus, and that it suffers from significant conceptual problems. The failure of such attempts at refining definitions of focus supports our contention that it is not the detail of any such definitions that is at fault, but rather the underlying methodology that reifies focus as an essential category in the first place.
4.1 Ordinary versus contrastive focus: Foundational problems
There are two basic ways of introducing ordinary/contrastive focus distinction into the repertoire of universal categories. These are neatly exemplified by the work of Vallduví & Vilkuna (Reference Vallduví, Vilkuna, Culicover and McNally1998), who introduce contrast into their universal set of information-structural primitives, and É. Kiss (Reference É. Kiss1998), who argues for a split in focus types which she terms ‘identificational’ and ‘information’ focus (based largely on the nature of Hungarian ‘focus movement’). These approaches differ in a number of ways, the most obvious difference being that É. Kiss proposes a simple division of focus into two kinds, while Vallduví & Vilkuna propose that contrastiveness (in the form of a primitive that they call ‘kontrast’) is independent of topic or focus, combining freely with either or both according to language-specific principles.
More significantly for present purposes, the definition of ‘ordinary’ and ‘contrastive’ that each assumes is quite different. The fundamental semantic notion in É. Kiss's analysis of Hungarian ‘identificational focus’ is exhaustivity, i.e. selection of all and only those alternatives for which a given property holds, whereas Vallduví & Vilkuna's ‘kontrast’ relates simply to the existence of alternatives to the focused item. As we shall see, the adoption of either definition of contrast as a universal primitive that is manifested in many languages leads not only to empirical inadequacy but also to internal incoherence.
In Vallduví & Vilkuna's case, this incoherence springs from the idea that the relevance of alternatives constitutes something that can be added to ‘ordinary focus’ to create contrastiveness. That this is problematic is immediately obvious from the fact that interpretation in the context of alternatives is widely taken to be a definition of ‘ordinary focus’ itself, following the likes of Rooth (Reference Rooth1992, Reference Rooth and Lappin1996). It is therefore unclear how this element of interpretation could underpin the distinction between focus (‘rheme’) and contrast that Vallduví & Vilkuna wish to draw. We return to this conceptual issue in Section 5.
The problems inherent to É. Kiss's exhaustivity-based distinction are more relevant to the principal aim of this section, to demonstrate the empirical inadequacy of any ordinary/contrastive focus split. Indeed, there are problems within É. Kiss's own (1998) treatment of different languages. She claims that many languages show a grammatical split between left-peripheral ‘identificational foci’ and in situ, unmarked ‘information foci’ (compare the ‘high’ and ‘low’ focus positions assumed in the so-called cartographical approach to information structure; Rizzi Reference Rizzi and Haegeman1997, Belletti Reference Belletti2009). The former kind are subject to a degree of parametric variation: in a language like Hungarian, identificational foci are said to bear the feature [+exhaustive], while in many other languages (for example, most of the Romance languages) movement to the identificational focus position requires the feature complex [+exhaustive, +contrastive]. Here ‘contrastive’ means drawn from ‘a closed set of individuals known to the participants in the discourse’ (É. Kiss Reference É. Kiss1998: 268). Note that Hungarian is a crucial case, not only because it represents one side of this putative parameterisation, but also because it provides one of the clearest cases of a syntactic position that is seemingly dedicated just to focus of some kind.
The difference between these putative types of identificational focus is said to be manifested in examples like the following:
(9) Hungarian (É. Kiss Reference É. Kiss1998: 230)
[Olaszországban]focus jártam.
Italy.to I.went
‘I went to Italy.’
(‘… and no other place’; no immediate alternatives implied. Felicitous in context of a simple wh-question, such as Where did you go on holiday?)
(10) Italian (É. Kiss 1998: 269)
[Maria]focus ha rotto il vaso.
Maria has broken the vase
‘Maria broke the vase.’
(Restricted set of alternatives implied. Infelicitous in the context of a simple question, Who broke the vase?. Felicitous after Which of you two broke the vase?)
For the sake of argument, let us initially accept this broad picture. There remains a central problem. Hungarian does not in fact show any kind of unmarked, ‘ordinary’ focus in situ. On the contrary, the unmarked position for foci, as determined by the Q–A diagnostic, is just that left-peripheral ‘focus position’ (henceforth FP) that is now said to host the marked identificational foci (recall example (1b)).
In consequence, É. Kiss's analysis of the crucial case of Hungarian is fundamentally different in kind to that of the other languages she discusses, where her terminology has quite different significance. In relation to these other languages, what É. Kiss calls ‘information focus’ equates to what others have called ‘ordinary focus’, and what she calls ‘identificational focus’ stands in contrast to this as the marked focus type. This cannot be true of what she calls ‘identificational focus’ in Hungarian, given that the supposed identificational focus position is the unmarked position of all foci. Conversely, the in situ foci which É. Kiss labels ‘information focus’ in Hungarian are certainly not ‘ordinary focus’. In fact, these Hungarian in situ foci seem to be of two kinds: either they are sub-parts of broader (e.g. VP-)foci, as in the boldface expression in (11), or highly marked non-exhaustive (‘incomplete’) foci, as in (12) (see Wedgwood Reference Wedgwood2005: 125).
(11) Mari meg-evett egyalmá-t.
Mari ptl-ate an apple-acc
(What did Mary do?) ‘She [ate an apple]focus.’
(12) Meg-hívtak (például) János-t.
ptl-they.invited for.example János-acc
(Who did they invite?) ‘They invited János, for example/among others.’
The upshot of all this is that the characteristic inhabitants of the Hungarian FP – any NP-sized foci that are not of the marked kind in (12) – are divided across the two possible positions for focus in languages like Italian. Moreover, it follows that ‘identificational focus’ in these other languages is something quite distinct from any category for which Hungarian syntax gives any evidence.
Therefore, É. Kiss's (Reference É. Kiss1998) analysis does not logically argue for what it has widely been taken to support (and its own terminology suggests), a universal split between ‘ordinary’ and contrastive focus, where the latter is defined in terms of exhaustivity. On the contrary, it reveals a fairly stark cross-linguistic contrast, with Hungarian behaving quite differently to other languages. This flatly contradicts the invocation of Hungarian in support of a universal distinction in focus types/positions by the likes of Rizzi (Reference Rizzi and Haegeman1997) and Vallduví & Vilkuna (Reference Vallduví, Vilkuna, Culicover and McNally1998).
To summarise, the content of the Hungarian FP is both unmarked/‘ordinary’ (by the Q–A heuristic) and exhaustive. This feature configuration obviously cross-cuts the assumed universal distinction between a focus which merely conveys new information (ordinary or information focus) and an operator-like entity with exhaustive or contrastive features (contrastive or identificational focus). Putting aside these implications for a universal ordinary/contrastive focus split, might we at least conclude that Hungarian syntax provides evidence for a basic focus feature that may be supplemented by some further elements of meaning, such as exhaustivity? Things are not so simple. In many ways, what Hungarian FP most resembles is the English it-cleft, as É. Kiss herself argues. Crucially, there are properties of the it-cleft which unequivocally show it not to be intrinsically a focus construction, and Hungarian FP has these same properties.
It has often been observed that the meaning of an it-cleft is essentially presuppositional: the existence of a unique bearer of some property is presupposed and the post-copular ‘clefted’ expression identifies this unique entity (see Zimmermann & Onea Reference Zimmermann and Onea2011: 1666 for a recent summary). Significantly, this presuppositional meaning is orthogonal to focus, as defined by the crucial Q–A diagnostics (Prince Reference Prince1978, Delin Reference Delin1992). Hence, alongside the ‘citation form’ cleft (e.g. It is BEANS that I like), we also find what Prince calls ‘informative presupposition’ clefts, as in (13), which clearly shows that focus, as defined by the Q–A criterion, may appear anywhere in an it-cleft sentence.Footnote 8
(13) [The leaders of the militant homophile movement in America generally have been young people.] It was they who fought back during a violent police raid on a Greenwich Village bar in 1969. (Prince Reference Prince1978: 898)
Hungarian FP behaves like the it-cleft in allowing ‘informative presupposition’ readings unproblematically:
(14) [Translation of immediately preceding sentence: ‘The secondary schools have always been a place where intellectuals were educated in the spirit of the nation, and they preserved this spirit even when there were prosecuted and marginalised, because the spirit of the nation has always been developed by the secondary schools.’]
Nem véletlen, hogy az elnyomó … hatalom mindig aközépiskolákra tette rá kezét.
not coincidence that the repressive power always the secondary.schools.on put ptl hand
‘It is no coincidence that it's the secondary schools that the repressive powers … have always taken control of.’
(Hungarian National Corpus; see Wedgwood, Pethő & Cann Reference Wedgwood, Pethő and Cann2006)
The sentence presupposes that there is a (contextually) unique element of the society that ‘the repressive powers’ strive to take hold of, and this element is identified as secondary schools (a középiskolákra in FP): this is the standard identificational semantics, not focus, on any account of what focus is. The implicit discourse question (in the sense of Roberts Reference Roberts1996) does not inquire about the identity of this entity, but rather about the relationship of secondary schools with the repressive powers. Therefore, to the extent that the Q–A criterion applies to such examples at all, it determines the focus to be ‘have taken control of’ rather then ‘secondary schools’. In other words, the Hungarian FP and ‘focus’ are not necessarily coextensive. We must conclude that É. Kiss's (1998) terminology is half right: FP is identificational, but it does not follow that FP realises a category of ‘identificational focus’, for we know identificational semantics to be dissociable from focus.Footnote 9
Far from supporting a universal category of focus, or of contrastive focus, Hungarian FP seems to call for a quite distinct kind of explanation. It seems that the underlying meaning of FP must be of some relatively underspecified kind, such that it is compatible both with the expression of regular, Q–A style focus (as (1b)) and with cases like (14), in which there is an identificational meaning but no clear relationship to traditional notions of focus.Footnote 10 Thus, FP regularly has the effect of expressing contrast and/or focus, but this is not the same as being dedicated to, or a realisation of, focus (contrastive or otherwise).
4. Variation in putative contrastive focus phenomena
Some supposed exemplars of the ordinary/contrastive focus distinction in fact fail to show clear grammatical sensitivity to it. Experimental work by Skopeteas & Fanselow (Reference Skopeteas and Fanselow2010) demonstrates that superficial evidence of a syntactically expressed ordinary/contrastive focus split in Georgian is illusory: the relevant form–meaning associations are merely statistical. Whether one looks across speakers or at individual speakers, there is no feature whose presence or absence is criterial for either of the syntactic positions in question. Hartmann (Reference Hartmann and Steube2008) argues something similar with regard to Hausa, on the basis of more traditional elicited data. This is incompatible with a hard and fast distinction between primitive varieties (or components) of focus. Instead it suggests a continuum of interpretations, perhaps along some scale such as ‘attention-worthiness’ or ‘unexpectedness’, with more or less arbitrary cut-off points determining the use of one structural position rather than another by different speakers in different contexts (see Zimmermann Reference Zimmermann2008 for a similar point).
Not only the sharpness of the grammatical distinction but also the precise nature of relevant contrastive meanings is potentially subject to significant variation. It has been suggested, for example, that a number of languages have specific structures for ‘corrective’ focus, as opposed to any other kind of contrastive meaning – Gussenhoven (Reference Gussenhoven, Lee, Gordon and Büring2007) mentions Efik and Lekeitio Basque; Aghem seems to belong here, too (see Section 3 above). Molnár (Reference Molnár, Hasselgård, Johansson, Behrens and Fabricius-Hansen2002) claims that Finnish has distinct structural marking for a focus that ‘refers to alternatives in a contextually limited set where the alternatives are known to the participants of the discourse’, and she explicitly relates this to a kind of contrast found in sentence-initial position in Basque, on the basis of a few isolated examples.Footnote 11Vilkuna's (Reference Vilkuna1989) description of the Finnish facts, on the other hand, suggests that the crucial property of the contrastive position in Finnish is the combination of old and new information, meaning that it hosts both confirmations and corrections. To what extent and exactly how any of these phenomena differ from ‘focus with a closed set of alternatives’, as in analyses of languages like Italian, remains unclear; systematic comparisons at sufficient levels of detail are almost entirely absent from the literature.
What is certain is that Finnish offers little support to a universal information-structural primitive of contrastive focus, or pure contrast. An important part of Vallduví & Vilkuna's (Reference Vallduví, Vilkuna, Culicover and McNally1998) argumentation is the claim that Hungarian FP and the Finnish sentence-initial ‘contrast position’ both manifest a single primitive notion of contrast (or, in their terms, ‘kontrast’) – necessarily combined with focus (or ‘rheme’) in Hungarian, but combinable with either focus or topic (‘link’) in the Finnish case. It should already be clear that this is unlikely, given our observations above about Hungarian. And, indeed, Molnár & Järventausta (Reference Molnár, Järventausta, Molnár and Hetland2003) note a series of significant dissimilarities between these two supposed instantiations of contrast. For example, constraints on quantifier phrase distribution in Hungarian, which are typically understood to reflect the exhaustivity of FP (É. Kiss Reference É. Kiss1998: 251ff.), are not found in Finnish. Also, the use of the Finnish ‘contrast position’ is optional (or must be determined by other, as yet unknown factors): both of (15a,b) are acceptable, in what is plainly a ‘corrective’ context. Indeed, Molnár & Järventausta (Reference Molnár, Järventausta, Molnár and Hetland2003: 131, 134) suggest that the less marked of the two is (15b), without the sentence-initial ‘contrast position’:
(15) [‘Peter flew to Stockholm’]
(a) (Eihän, vaan) Reykjavikiin Pekka lensi.
no but Reykjavik.to Pekka flew
‘(That's not true) It was Reykjavik that Pekka flew to.’
(b) (Eihän, vaan) Pekka lensi Reykjavikiin.
no but Pekka flew Reykjavik.to
‘(That's not true) It was Reykjavik that Pekka flew to.’
In addition, the fact that the relevant Finnish position can host contrastive topics as well as contrastive foci is itself a notable difference to the Hungarian case, and, we would add, more problematic than Vallduví & Vilkuna's presentation suggests. In particular, it is unclear how their notion of contrast-plus-link is to be differentiated from the simple notion of link. Vallduví's link is not simple ‘topic’, but is restricted to ‘switch-topics’. These inherently involve some contrast between alternatives (see further Erteschik-Shir Reference Erteschik-Shir2007, Brunetti Reference Brunetti2009), making it unclear what it would mean to add an additional element of contrast. More generally, such a radical difference in overall semantico-pragmatic outcomes only points to the likelihood that the underlying interpretive mechanisms in the Hungarian and Finnish cases are different.
It may appear at this stage that there remains open a relatively simple analysis based on primitive information-structural features (to cover the languages mentioned in this section, if not in the previous one): one in which we recognise a kind of contrast that is restricted to small, contextually salient sets of alternatives and is canonically manifested as corrective focus. This might be posited alongside both focus and a different kind of contrast, which might perhaps be defined in terms of exhaustivity; this is essentially the position of Molnár (Reference Molnár, Hasselgård, Johansson, Behrens and Fabricius-Hansen2002). Alternatively, it might be that this restricted kind of contrast is the main kind of marked focal meaning, with focus itself unmarked (and, presumably, with phenomena like Hungarian FP left to some quite separate form of explanation); this is close to the position of Selkirk (Reference Selkirk2008). Quite apart from the various misgivings mentioned above – both metatheoretical qualms and the paucity of sufficiently detailed comparative data in the literature – there are strong indications even within English that such an approach, in either variant, is unsustainable. English shows clear cases of ‘ordinary focus’, exhaustive/identificational focus and corrective focus, sometimes in different structures, but in such a way that they are demonstrably attributable to a series of complex, interacting, partially inferential interpretive processes, not to the realisation of distinct, primitive features.
First, all of these meanings can be expressed using pitch accents only. Thus, (16b) is appropriate in any of the contexts indicated in (16a):
- (16)
(a) [‘out of the blue’]
[What does Robert want?/What is it that Robert wants?]
[Robert wants coffee, doesn't he?]
(b) Robert wants TEA.
At the same time, there are English syntactic constructions that limit the expression of focus to a specifically identificational kind. As noted above, this is true of the it-cleft, and here the sense of exhaustive identification is notably connected to a presupposition of uniqueness, rather than the assertion of some special sub-type of focus. There is also the wh-cleft, with a similar interpretation, as in (17):
(17) What Robert wants is TEA.
However, the wh-cleft has its own properties (Prince Reference Prince1978), one of which is that pitch accenting cannot be freely moved around within a wh-cleft as in the case with it-clefts. If the most prominent accent falls on a pre-copula expression, the resulting meaning is necessarily corrective, as in (18):
(18) What ROBERT wants is tea.
(… It's not what Ellen wants. Infelicitous as an answer to Who wants tea?)
Crucially, this obligatorily corrective meaning plainly arises as a result of the interaction of several independent facts about English prosody and syntax; it seems quite implausible to suggest that cases like (18) represent a distinct construction that directly and arbitrarily realises a feature [+corrective]. The significance of the pitch accent in (18) is limited to a corrective interpretation simply because something about the wh-cleft independently prevents anything other than the post-copular constituent taking on an ‘ordinary’ focus reading (we take this to be connected to presuppositions triggered by the wh-word and/or the kind of copula verb involved).Footnote 12 This demands an analysis whereby the pitch accent has some relatively underspecified meaning from which the corrective meaning is derived. Hence, from English data alone, we may conclude that a corrective interpretation must be derived, not encoded, in at least some languages. Note that it is not our intention here to suggest that this interpretation is never directly encoded; we consider this a perfectly plausible possibility and it may be the case in some of the languages mentioned above. What we wish to emphasise is that the mere existence of a mapping from some structure to a corrective meaning – even where this is unique and obligatory – is no guarantee that this meaning is directly encoded.
Appealing to a corrective or ‘restricted’ kind of contrastive focus thus fails to provide an empirically or conceptually adequate account of relevant data. This should come as no surprise, since there is plenty of evidence of wider variation in the meanings of ‘contrastive focus’ phenomena. This ranges from subtle but significant variation, even across closely related dialects, to more starkly divergent phenomena. An example of the former is the difference in conditions on ‘focus fronting’ in Yiddish-influenced American English (‘Yinglish’) and other varieties of English, which Prince (Reference Prince1999) identifies as a difference in sensitivity to ‘hearer-newness’. Some examples of the less subtle kind of variation have already been given, in the Somali, Aghem and Tura phenomena discussed above, each of which could be said to contain at least one distinctive type of contrastive focus, but which all have important properties that are not predicted simply by applying such a designation. In effect, in order to capture this kind of diversity, the splitting strategy has no choice but introduce a new subcategory for every different type of ‘contrastive’ focus.Footnote 13
The idea that focus can be salvaged as a universal category by splitting it along the line of contrastive/ordinary is further undermined by the fact that the evocation of alternatives belonging to a closed set can indeed be encoded by a grammatical category, but without being related to information structure. The Tungusic language Even has a suffix, -d(A)mAr, which has a number of the defining characteristics of canonical focus-marking; most notably, when it is used in a Q–A context, it must be used congruently. Furthermore, it seems to roughly correspond to what has been defined as contrastive focus: as (19) shows, it is used in contexts in which there is a highly restricted (usually binary) distinction between two alternatives.Footnote 14
- (19)
(a) Tiːniw hịgịdụk ŋiː-dmer emden?
yesterday from.forest who-d(A)mAr came
(b) Bujuhemŋe-dmer tiːniw emden.
hunter-d(A)mAr yesterday came
‘Who (out of the two people we expected) came from the forest yesterday?’ – ‘The hunter came from the forest yesterday (not the other person).’
On the other hand, -d(A)mAr lacks one of the purportedly universal features of focus (Zimmermann & Onea Reference Zimmermann and Onea2011: 1655), obligatory interaction with restrictive expressions of the only-type (Matić Reference Matić2009). Actually, it in many ways resembles ‘pure contrast’ marking of something like the Finnish kind, since it may also appear with what are clearly contrastive topics:
(20) Amandʒị ŋeːlrin, eńmi-dmer họːč ajawrịn.
her.father feared her.mother-d(A)mAr very loved
[What was her relationship with her parents?]
‘She was afraid of her father, but her mother, she loved her a lot.’
However, one cannot simply classify -d(A)mAr as a Vallduvían contrast-marker that may freely combine with topic or focus. Similarly to Somali baa, it proves to have other uses when we look beyond conventional elicitation tasks. Thus, it is often attached to various discourse connectors (‘and then’, tail–head linkage, etc.). The notion of contrast this usage implies certainly does not involve Roothian alternatives that are defined relative to a particular propositional slot, but rather a loose textual parallelism. Even more importantly, some uses of -d(A)mAr seem to defy description in terms of information structure altogether. For instance, it is the major means for deriving comparatives (Cincius Reference Cincius1947: 237), and it can modify lexical meanings, as is shown in (21), and replace the possessive suffix with relational nouns (22):
- (21)
(a) Heːruku emuli! vs.
bag bring
‘Bring me the bag!’ vs.
(b) Heːruku-dmer emuli!
bag-d(A)mAr bring
‘Bring me the bag in which other bags are carried (the bag for the bags)!’
(22) Ama-dmar ọmọlgọ-j negirin.
father-d(A)mAr son-refl.poss scolded
(text-initial) ‘A father was scolding his son.’
In all these non-focus and non-topic cases, -d(A)mAr arguably functions as an operator-like expression which marks the element in its scope as a member of a restricted set. If contextually relevant alternatives are not present, it produces an inference of the default set to which the denotatum of its host belongs. In the case of comparatives, it is the set of degrees of a property (see e.g. Schwarzschild Reference Schwarzschild2008); in the case of lexical modification, it is the natural kind to which the denotatum belongs (‘a bag among bags, a bag with respect to bags’). With relational nouns, which are otherwise obligatorily accompanied by a referential possessive suffix, -d(A)mAr replaces it, i.e. it somehow introduces the referent for which the predicate ‘be father/mother/friend/etc. to’ holds true. The mechanism is just as in the previous cases: -d(A)mAr produces an inference of the default set to which the denotatum of this noun belongs – in (22), the set {father, child} – from which the referent necessary to interpret ‘be father to’ can be derived.
All of this shows the nature of -d(A)mAr-interpretation to be quite unlike Roothian focus semantics or common conceptions of information-structural contrast, the relevant set of ‘alternatives’ being subject to quite different constraints and potentially drawn from quite different sources. Overall, this kind of phenomenon markedly contradicts the legitimacy of analysing ‘focus’ structures in terms of minimal universal features. Taken as primitive grammatical entities, neither focus nor contrast (as they are proposed for other languages), nor any combination of them, helps us to produce an adequate characterisation of -d(A)mAr, though it is subject to constraints of Q–A congruence and plainly comes to express contrastive focus in some of its uses.
5. The underlying unity strategy
The splitting strategy thus seems doomed to a choice between empirical inadequacy and lack of generality: the only way to avoid an unordered catalogue of potentially unlimited entities is to posit a finite set of Procrustean categories unable to capture, let alone explain, observable linguistic diversity. Yet this is plainly contrary to the original motivations for splitting the category of focus.
An alternative way to preserve an essentialist category of focus might be to take exactly the opposite tack: to maximally increase the generality of the category by reducing the number of necessary attributes. That is, one might seek the very essence of focus phenomena by identifying and explaining them via one single defining feature.
Typically, this means finding a very general denotation with the potential to host the widest possible range of readings that the purported members of the category have. With regard to focus, the most influential applications of this strategy are Rooth's (Reference Rooth1992, 1996) Alternative Semantics and the related Structured Meanings framework (Jacobs Reference Jacobs1983, von Stechow Reference Stechow, Wunderlich and Stechow1991, Krifka Reference Krifka, Féry and Sternefeld2001). Rooth assumes that focus is a cross-linguistically applicable category:
[T]he common core [to a cross-linguistic notion of focus] might turn out to be the weak semantics of the prominence feature in English, with some constructions and morphemes expressing additional semantic content – such as existential presupposition or exhaustive listing – in addition to and in terms of the basic semantics. (Rooth Reference Rooth and Lappin1996: 296)
For Rooth the invocation of alternatives is in effect the meaning of focus itself (unlike the use of alternatives in Vallduví & Vilkuna Reference Vallduví, Vilkuna, Culicover and McNally1998; see Section 4.1 above). Technically, the contribution of focus to the meaning of an utterance is computed at a separate level, the focus semantic value. While the ordinary semantic value, ignoring the contribution of focus, is a standard proposition, the focus semantic value is a set of propositions that differ from each other only in that in each the denotatum of the focused expression is replaced by another object of the same type. To model discourse anaphoricity of focus (as reflected in Q–A pairs) and to constrain the focus semantic value to relevant alternative propositions, a context variable C is introduced, referring to a contextually determined set of alternatives, along with a focus operator which induces the requirement that C be a subset of focus-induced alternatives. Structured Meanings, meanwhile, separates the focus of an utterance from the rest by lambda abstraction. This similarly creates a set denotation as the background to a focus, only in this case not a set of propositions, but rather some form of predicate denotation.
There is no doubt that a sense of the relevance of alternatives is intimately associated with phenomena that are recognised as cases of focusing – perhaps inextricably so. It does not follow that focusing is (at its ‘core’) the evocation of alternatives. That is, the question we should ask is not whether the invocation of sets of alternatives is empirically adequate to characterise focus phenomena, but rather why it is. This relationship between focus and alternatives may be a mere reflection of deeper facts about linguistic communication. Indeed, we contend that the reason alternatives can be used to model focus at all is because the existence of alternatives is inseparable from the notion of assertion, indeed from the very existence of communication.
There is a very basic, essentially pragmatic truth that connects assertion, newness or any other definition of focus to the idea of alternatives. An assertion is worth making just to the extent that it addresses a live issue (or is being conveyed as doing so), which means an issue that could at least conceivably be resolved in more than one way. The existence of alternatives is therefore a fundamental part of what makes any assertion relevant (Stalnaker Reference Stalnaker and Cole1978). Thus, it follows from the act of making an assertion that there are alternatives to what has been asserted – not because each assertion harbours some formal entity called focus, but for essentially communicative reasons.
It is therefore no surprise, and no particular source of insight, if there seems to be consistent empirical support for the strategy of formalising the interpretation of focus as the evocation of set denotations, as in Alternative Semantics or Structured Meanings. Essentially by definition, wherever one finds any of the commonly cited properties of focus – newness, assertion-creation or, most obviously, contrast – one will be able to identify the potential relevance of alternatives, to some (greater or lesser) extent. This is why the Roothian strategy of associating alternatives with different parts of utterances successfully covers all such phenomena. But it is also why this connection to alternatives merely reflects a relationship of family resemblance between them: they all relate to the same broad aspects of communication. It does not follow that Alternative Semantics picks out some natural class of linguistic phenomena, still less one that is attributable to a single underlying grammatical entity.
The alternatives-based approach captures what a range of linguistic phenomena have in common just at the level of their effects on a strictly denotational characterisation of meaning. This of course follows in an esteemed philosophical tradition, but it imposes a doubly limited analytical perspective, especially in the context of a diverse and communicatively grounded field of study like focus. On the one hand, it concentrates exclusively on the effects of interpretive processes, rather than the nature of those processes, a general problem we have already criticised above. In addition, it relies on a single characterisation of these effects and therefore concentrates exclusively on that subset of interpretive effects that this characterisation gives access to. This adoption of a single analytical tool amounts to a self-imposed limitation on the analyst's view. The danger here is not only of obscuring important details but also encouraging the perception of inappropriate kinds of structure in the data. By way of analogy, the inherent limitations of human vision might lead one to account for the night sky as a collection of two-dimensional constellations of homogenous light-emitting heavenly bodies. In the same way, if one restricts oneself to characterising linguistic phenomena in terms of their effects on representations of denotational meaning, then one will inevitably conclude that a wide range of phenomena is unified in its relationship to sets of alternatives. This naturally enough leads to positing a common interpretive process, which is taken to be triggered by a shared grammatical primitive. This mode of analysis simply gives no access to distinctions such as those between the direct marking of new information (in a dynamic, Vallduvían sense), the encoding of a presuppositional relation of identification, a predicative relation, direct evidentiality, realis modality, and plain ‘highlighting’ (Section 3 above). As we have emphasised, such underlying diversity is well supported when one considers the full range of data from languages like Somali, Hungarian, Quechua and Even. Thus, restricting oneself to the tools of denotational analysis, far from being a legitimate idealisation or a strategy that is justified by considerations of practicality, has the consequence of obliterating what are not only fine distinctions in interpretative outcomes, but quite independent interpretive processes. And this is surely the kind of substantial variation that grammatical analysis must take account of.
Importantly, a linguistic phenomenon may directly encode the evocation of alternatives – we certainly should not rule out the possibility that this could be conventionalised in some languages. In fact, this last possibility may well be a conclusion that we have to draw in the case of some uses of the English focal pitch accenting, given evidence presented by Beaver & Clark (Reference Beaver and Clark2008). Our point, however, is that this kind of encoding of alternative sets cannot be assumed to hold in any given case of putative focusing. At the same time, we note that the kinds of alternatives invoked by apparent focus phenomena may differ from the kind suggested by formal approaches like Alternative Semantics – as in the case of the Even particle -d(A)mAr (Section 4.2 above), which encodes a contrast, but with a set of alternatives whose size is restricted even as its nature and source is completely underspecified, unlike Roothian alternatives that share a property denoted by the remainder of the sentence. Thus, the encoded meaning of a given putative focus phenomenon, even if it truly involves some kind of contrast, need not wholly coincide with the communicatively based effect of evocation of alternatives. This is quite contrary to the assumption of Roothian mechanisms as the maximally general semantic basis of a universal category of focus.
Recall that our discussion of Alternative Semantics is presented as an example of the more general strategy of unifying focus through a semantically very general ‘common core’. This ultimately suffers from the same flaws as the splitting strategy. In unquestioningly mapping a formalisation of the relevance of alternatives to a grammatical category, it treats an interpretive effect as a cause of interpretations. Moreover, it does so in a way that actively obscures a variety of possible form–meaning mappings. Because of their shared flaws, the ‘common core’ strategy actually reaches much the same result as the splitting strategy, if seriously applied to a wide range of linguistic structures: to characterise the diversity of the data, this common core has to be supplemented by an unrestricted number of additional features, and thus lose its main attraction, apparent generality.Footnote 15
Note that to reject the category of focus is in no way to deny that the effects of newness, contrast, evocation of alternatives, etc., are omnipresent in human languages, and necessary in communication. In fact, it is precisely this communicative necessity that renders their encoding as a dedicated grammatical category unnecessary (though, of course, possible). This point applies well beyond those approaches that, like Rooth's, encapsulate some general denotation in a syntactic feature or phrase structure configuration. The same kind of fallacy is represented by weaker assumptions, as in approaches that operate with separate information structure modules and/or with discrete denotations which are realised by the grammar in unpredictable ways. A recent paper by Zimmermann & Onea (Reference Zimmermann and Onea2011) exemplifies this looser way of attempting to maintain the universality of focus effects. They claim that focus is outside grammar, being a kind of general-purpose cognitive process – a position ostensibly compatible with ours. Nevertheless, they seek to provide the right definition of focus in terms of denotational semantics and to illustrate in which ways this universal denotation is realised in the grammars of the world's languages. As we have argued throughout this article, it is futile to pursue the ‘right’ definition for focus: since focus-like interpretations are merely a set of interpretive effects (the clustering of which is probably rooted in one or more general cognitive mechanisms), they can stem from highly divergent form–meaning pairs, and can come about in ways that are more or less independent of the grammar. As we have shown, the possible source denotations range from mood and evidentiality to highlighting mechanisms or, indeed, alternative inducers. Simply to discuss by what grammatical means focus is realised in grammar is to preclude all of this variety of explanation – not merely to ignore superficial diversity in the data – and, in so doing, to obscure important insights into the nature of language and its interpretation. Seeking formal common denominators for, say, evidential morphemes and highly idiosyncratic contrast markers is otherwise unmotivated and liable to distort any theory of language. To understand why they might have overlapping effects would be a quite different, potentially much more valuable achievement – but this is just what is obscured by the reification of focus as a linguistic category.
6. Conclusion
There appears to be little left of the single, invariant cross-linguistic category of focus from which this paper set out. The category decomposes into a potentially unlimited number of encoded meanings, which for different reasons happen to be compatible with the effects usually associated with the putative discrete focus meaning. Having said this, we do think that the observed effects can be of high cross-linguistic relevance, but in a sense which sharply deviates from the implications of a unified (or binary, or ternary) category.
Let us first recapitulate our argumentation. Cross-linguistic categories can count as adequate only if they satisfy the following three conditions: they must facilitate identification of meaningful points of comparison, unify phenomena at a certain level of abstraction without violence to facts, and have an explanatory power in the sense of being able to account for the range of possible variation across languages. And to be sure, they can be considered categories in the first place only if they are based on category-defining attributes that establish a sort of natural kind, not on superficial similarities. Focus seems to fare quite poorly when measured against these criteria. As we have shown in Sections 3 to 5, the notion of focus is based on observed pragmatic factors, which surface as changes of cognitive states (newness) or speech-act-related phenomena (Q–A congruence, assertion, correction), or are otherwise representable as abstract constructs (alternatives, compositional restructuring). Despite their pervasiveness in the language, these are merely pragmatic effects, which may be – and, we aim to have shown, routinely are – derived from different, often quite unrelated underlying form–meaning mappings. This makes the status of focus as natural category untenable. The possible escape routes for focus – lumping and splitting – collapse on both empirical and methodological grounds.
Focus thus fails to meet the relevant criteria for a category. All but one, that is: the pragmatic effects with which focus is associated – newness, contrast, use in certain types of contexts, etc. – can facilitate the identification of meaningful points of comparison across languages, if they are applied in their proper ontological capacity, namely as a heuristic tool. In other words, while we have up to this point referred to cross-linguistic categorisation, we would urge the drawing of a distinction similar to that proposed by Haspelmath (Reference Haspelmath2010), between (i) linguistic categories, qua elements of the structural analysis and (ii) comparative concepts, which are essentially practical points of comparison, defined relative to the particular needs of the analyst on a particular occasion and afforded no deeper theoretical role. Focus is then a purely descriptive tool, which should facilitate both language-internal analysis and comparison across languages without constituting that analysis.
It is only in this sense that we claim that the notion of focus is an important notion, albeit a currently misconstrued one. The comparative concept of focus delimits the phenomenological field of contexts and structures that are in one way or another connected with information update and the speech acts based on it. Given the centrality of these in human communication, this comparative tool stands a good chance of being a source of important insights into the nature of human language. But it is not a category – a thing among things, a natural kind, an abstraction over related structures, a prototype, or however one wishes to conceive of categories. It is the metalanguage that is wrong, not the idea of comparability itself, nor its broad field of application.
Why is it important to set the ontology of focus straight? Employing an erroneous metalanguage sooner or later results in erroneous research results. From the epistemological point of view, inappropriate ontologies are an infinite source of false classifications and, ultimately, spurious explanations. At a more specifically linguistic level, the idea of a discrete focus category has been an impediment to both descriptive and comparative work. The belief in the universality of the category has resulted in such misconceptions as the formal equation of English clefts and Hungarian FP's, the introduction of a universal focus position high in the left periphery despite evidence to the contrary, and the frequent suppression of the really distinctive properties of phenomena like Somali baa. An incorrectly established category thus turns into a Procrustean bed for the variety of linguistic categories found in the phenomenological field of focus across languages. In contrast, an open-end comparative tool merely facilitates identification of relevant grammatical categories within languages and leaves the semantic and formal characterisation of these categories open.
It is important to note how theory and methodology are intertwined here. Our discussion of focus has highlighted a number of problematic practices in linguistic analysis, the most important of which are a readiness to precipitately elevate vague similarities to grammatical universals, and, even more importantly, a lack of differentiation between causes and effects and the concomitant reification of the latter. Crucially, these analytical shortcomings are intimately related to underlying theory. Given a theoretical commitment to an essentially ‘realisational’ understanding of linguistic structure, the analyst has little choice but to seek to fashion categories out of relatively superficial effects, both in the analysis of an individual language and in comparative work that leads to positing universals. The promise of subtler and more explanatory analysis arises once one considers that, as in the case of biological species, languages may be best unified not in terms of gross components of either form or function, but rather in deeper and often dynamic terms – in processes of computation at various levels, and via constraints on developmental trajectories. To examine such possibilities would take us well beyond the scope of the present article; our point here is simply to note that the problems of linguistic categorisation that we have raised are not merely the results of sloppy practice but relate to core elements of one's theoretical approach – and that there are coherent alternative approaches.
As noted above, this still leaves a role for focus (now understood as a cover term for certain interpretive effects), as a heuristic tool. The notion of focus is then a means of identifying structural patterns which languages use to generate a certain number of related pragmatic effects, with the ultimate aim of uncovering patterns of semantic and pragmatic variation and establishing higher-order ‘system preferences and functions, with recurrent solutions’ (Evans & Levinson Reference Evans and Levinson2009: 446). The important point is the difference between identifying structural patterns related to certain effects and determining how these effects are arrived at. The open-ended comparative tool we are suggesting is able to do the former, but is fully agnostic as to the latter; any universal linguistic category of focus attempts to account for the latter by fiat and consequently fails.
