Akan is a Kwa language of the Niger-Congo phylum that is spoken mainly in the central and southern parts of Ghana. Altogether, it has over 8.3 million speakers (Lewis 2009). The three literary dialects of Akan are Asante Twi, Fante and Akuapem Twi. The present paper is concerned solely with the Asante Twi dialect. Akan is a syllable-timed tone language, with high and low tones used to distinguish lexical or grammatical meaning (Dolphyne 1988; Aboh and Essegbey 2010).1 The unmarked word order of Akan is SVO (Boadi 1974; Saah 1994), as in (1).

The sentence in (1) is information-structurally neutral. It is compatible with a context that either licenses new information focus (henceforth NIF) on the entire IP, or on the VP, or on the object DP, allowing for a so-called focus projection (Reinhart 1995; 2006). In other words, the sentence in (1) can occur in contexts of questions of the type ‘What’s new?’, or ‘What does Adwoa do?’, or ‘Who does Adwoa take care of?’. Narrow focus on the object DP can additionally be represented via a marked strategy. That is, a question that licenses NIF on the object, as in (2Q), can be answered not only by a simple SVO structure in (2A1) but also by a so-called focus construction in (2A2).

The focus construction in (2A2) is characterised by the focused constituent occurring leftmost in the sentence. In (2A2), it is the focused object that surfaces in this position. In (3A2), it is the focused subject. As can be seen from (2A2) and (3A2), the leftmost focused constituent is immediately followed by the morphological marker na, which is referred to in the linguistic literature on Akan as a focus marker. Moreover, the focused constituent is obligatorily doubled by the so-called resumptive pronoun in the thematic position that agrees with the focused constituent in number and person (see (2A2) and (3A2)).2

As can be seen from (2A1) and (3A1), Akan exhibits a subject/non-subject asymmetry in that a narrowly focused object can remain in its thematic position (see (2A1)), whereas a narrowly focused subject cannot (see (3A1)) (Boadi 1974; Saah 1994; Fiedler et al. 2010). In other words, a narrowly focused subject is obligatorily represented via a marked operation.3

The focus marker na found in Akan focus constructions has been analysed in the linguistic literature on Akan as not the only focus marker found in the language. The morphological marker de(ɛ) that occurs in the same type of construction (see (4)) has been referred to as a focus marker as well (Boadi 1974; Saah 1988; Boadi 1990; Saah 1994; Marfo and Bodomo 2005).

In what follows I question the view of de(ɛ) as a focus marker. From the theoretical point of view, economy considerations should rule out employment of two morphologically distinct markers na and de(ɛ) for exactly the same information-structural (henceforth IS) interpretation. Moreover, as can be seen from the translations in (4), the interpretation of a de(ɛ)-sentence is different from that of a na-sentence. In Section 2, I maintain that only na is a true focus marker in Akan, whereas de(ɛ) is a marker of contrastive topic (CT). This conclusion is based on the interpretive and formal properties of de(ɛ)-clauses. After I have established that na is the only focus marker, I will proceed with an analysis of the interpretive properties of focused arguments in na-constructions. In Section 3, I argue that an ex-situ object obligatorily maps onto an interpretation that involves quantification over a discourse-salient set of alternatives,4 resulting in either contrastive or/and exhaustive or/and emphatic construal. I demonstrate that an SVO sentence can map onto exactly the same IS interpretation, but for an in-situ object, quantification over a discourse-salient set is merely available, not obligatory. I account for this observation by adopting the interface-based approach to IS encoding that assumes mapping of syntactic representations onto IS templates that is regulated by the principles of economy (Neeleman and van de Koot 2008; Titov 2017a; b). Section 4 proceeds with an analysis of focused subjects and the question of why these obligatorily occur in focus constructions. It will be argued that a marked strategy is cross-linguistically used for the encoding of narrow focus on a subject due to subjects not belonging to the default focus set (Reinhart 1995; 2006). Section 5 offers a syntactic analysis of the na-construction and proposes that it is an inverse pseudocleft. This conclusion is empirically based on the lack of reconstruction and theoretically on the Theta Criterion ruling out a mono-clausal analysis. The proposed syntactic analysis is further supported by focus constructions having properties of restrictive relative clauses (Schachter 1973; Boadi 1974; Ofori 2011). Section 6 concludes the paper.

In the previous section, we looked at the interpretation of Akan focused objects. This section is concerned with Akan focused subjects. As mentioned in the introduction, Akan exhibits a subject/non-subject asymmetry in the distribution of focused arguments. We have seen that a focused object can remain in situ, whereas Akan narrowly focused subjects can occur only in a na-construction. A recent trend in the linguistic literature on Akan is to argue that non-exhaustive subjects do not obligatorily occur in focus constructions (Duah 2015; Pfeil et al. 2015; Grubic et al. to appear). However, it is not entirely clear to what extent this is the case. For instance, Pfeil et al. (2015) argue on the basis of the data in (42Q), (42A1) and (42A2) that Akan non-exhaustively focused subjects cannot occur in focus constructions.

Since the context in (42Q) (which is given by the authors only in English) presupposes that Ama is not the only member of the set of individuals that provides a value for x in ‘x arrived’ (i.e., Kofi does that too), the subject in the reply to this question must be non-exhaustive. The fact that (42A2) is pragmatically infelicitous is taken by Pfeil et al. (2015) to suggest that this is due to the non-exhaustive construal of the subject, which is incompatible with focus constructions. However, if this were true, the felicitous reply in (42A3), would also be ruled out in this context, as it contains a non-exhaustively focused subject that occurs in a focus construction. The only difference between the infelicitous (42A2) and the felicitous (42A3) is that the latter contains the additive discourse particle nso ‘also/too’, which makes the sentence compatible with the context by signalling that a proposition (i.e., ‘Ama arrived’) is added to the discourse that also applies to at least one alternative (i.e., Kofi). Plausibly, the additive discourse particle cannot be omitted in an answer to (42Q) because it expresses VF, which must be realised in an answer to (42Q) because this question asks for a truth-value of the proposition ‘x arrive’ as applied to an alternative to Kofi. If so, the data in (42) do not demonstrate that non-exhaustive subjects cannot occur in focus constructions. At best, they show that such occurrence is optional. However, (42A1) may be available due to the fact that the sentence contains a VF. To be precise, the data in (42) demonstrate that an answer to (42Q), can either realise only the VF, as in (42A1), or the VF and the subject focus, as in (42A3), but not only the subject focus, as in (42A2). In other words, all that (42) shows is that VF realisation cannot be omitted in the context of (42Q). Conversely, when the subject is the only focus of the sentence the subject/non-subject asymmetry holds. Moreover, this asymmetry extends to wh-questions, where it is absolutely uncontroversial (see (43) and (44)). Unlike a wh-object, as in (43b), a wh-subject cannot occur in an SVO sentence, as in (44b). Instead it always occurs in a focus construction, as in (44a).

A further difference between Akan focused objects and subjects is that a subject in a na-construction does not have to involve quantification over a discourse-salient set of alternatives. That is, all of our reasoning about the difference in the interpretation of leftmost and in-situ objects does not apply to Akan narrowly focused subjects, with any of the discussed interpretations available for (45A), including the interpretation of a non-contrastive, non-exhaustive NIF. Similarly, wh-subjects, as in (45Q), do not have to be [+D-linked] in Akan. This observation is expected, given that there is no simpler SVO alternative to these structures that also involves wh-subjects or expresses narrow focus on the subject. Recall that interface economy chooses a simpler syntactic representation over a costly one if and only if the former achieves an interpretive effect that the latter fails to express. (In the case of focused objects this effect had to do with interpretive disambiguation.) However, in the absence of a simpler alternative for the given truth-conditional and IS interpretation, interface economy has no choice but to consistently select the marked representation, without any additional interpretive effect being achieved. In other words, if a na-construction with a focused subject could only capture a contrastive interpretation, there would be no structure to capture the interpretation of a non-contrastive narrow NIF on the subject. It is therefore unsurprising that the only available structure for narrow focus on the subject is interpretively ambiguous and is used in both types of context – contrastive and non-contrastive.

A logical question that follows from the above observation is why the unmarked SVO representation cannot capture the interpretation of narrow focus on the subject and the marked na-construction has to be used instead. The hypothesis adopted here is that narrow focus on a subject must be encoded via a marked strategy because subjects are not in the default focus set illustrated in (46) (Reinhart 1995; 2006).

We have seen that for Akan wh-objects and narrowly focused objects the in-situ complement to V position is unmarked. It is from this position that focus projection takes place, allowing for VP- and IP-wide focus. And it is in this position that a focused object is interpretively ambiguous as regards the set it operates on. A focused object that surfaces in a na-construction, conversely, is interpretively restricted in that it MUST operate on a discourse-salient set of alternatives, and no focus projection is available from the leftmost position. Similarly, in intonation languages, the unmarked position of a focused object is also the thematic complement to V position. Depending on whether the language is head-initial or head-final in the verbal domain, the object occurs either to the left or to the right of the verb. In this position the object can be interpreted as a narrow NIF or a narrow CF. And it is from this position that focus projection takes place resulting in VP- or IP-wide focus (Reinhart 1995; 2006). An object that A’-scrambles from this position is interpretively restricted in that it must quantify over a discourse-salient set of alternatives (Neeleman and van de Koot 2008; Neeleman and Titov 2009; Titov 2012a; b; 2013a), and, unsurprisingly, no focus projection can take place from an A’-position. In other words, the behaviour of focused objects in intonation languages is similar to that of focused objects in Akan.

Encoding of the interpretation of narrow focus on the subject, conversely, cross-linguistically requires a marked operation because narrowly focused subjects are not in the default focus set, i.e., a subject that belongs to the default focus set can only occur in IP-wide focus sentences (Reinhart 2006), as in (47A1) where the entire IP including the subject is in focus.

As expected, the IP-wide focus interpretation is captured by the unmarked SVO structure in (47A1) that we have already seen to allow for focus projection. Conversely, a focus construction where the subject is followed by the focus marker na, as in (47A2), is incapable of capturing the interpretation of a IP-wide focus. Nor is it compatible with the interpretation of a VP-wide focus (see (48A2)), strongly suggesting that no focus projection is possible for it.

In fact, it appears that no other IS interpretation but the one of narrow focus on the subject, as in (49A1), is available for a focus construction where the subject is followed by the focus marker na. As soon as the focus is wider than the subject DP, this interpretation has to be captured by the unmarked SVO structure.

In other words, narrow focus on the subject interpretation can be captured only by the focus construction and it is the only interpretation that is available for this construction. Hence, we can argue that the unmarked SVO representation is chosen by the interface system for the encoding of any focus that belongs to the default focus set but it fails to encode a focus that does not belong to this set, in which case the marked focus construction is chosen over the simpler SVO structure by the interface economy, as this is the only structure that can express this interpretation.21

Crucially, the syntactically and prosodically unmarked SVO construction fails to capture the narrow focus on the subject interpretation cross-linguistically (see also Ameka 2010 for the claim that subject focus constructions in Kwa languages involve an obligatory marking in the clause). However, languages differ in the choice of the linguistic tool (prosodic, syntactic or morphosyntactic) used to achieve a marked representation. In English, a marked operation of stress shift takes place at PF in order to mark a subject as the narrow focus of the sentence (see (50)).22 In Russian, a marked operation of A-scrambling takes place in syntax in order for the subject to be interpreted as the narrow focus of the sentence (see (51)) (Titov 2012a; 2013b).23 Both operations occur so as to place the subject in the position of stress.

By analogy, marking of narrow focus on Akan subjects also requires a marked operation. It follows, then, that Akan employs the morphosyntactically marked focus construction for the encoding of narrow focus on the subject in the same way as English uses a prosodically marked construction with a stress shift and Russian uses a syntactically marked A-scrambled OVS construction. The only difference is that English and Russian aim to place the subject in the position of stress in order for it to be interpreted as the narrow focus of the sentence, whereas Akan places the subject in the position immediately preceding the focus marker na, as this position is reserved for narrowly focused constituents.

The above analysis relies on the idea that there is a certain parallelism in the encoding of IS interpretations between syllable-timed tone languages, such as Akan, and intonation languages, such as English and Russian. We have seen that both types of languages optionally encode contrastive interpretation on the object via a marked representation. We have also seen that both types of languages employ a marked operation for the encoding of narrow focus on the subject, as subjects are not in the default focus set. The resulting representation can be marked either prosodically (e.g. stress shift in English), or syntactically (e.g. A-scrambling in Russian) or morpho-syntactically (Akan focus construction). The fact that Akan has a position in which only narrow focus surfaces, namely the position immediately preceding the focus marker na, allows it to make use of exactly the same surface position for the encoding of narrow focus on the subject and narrow contrastive focus on the object.24 Languages that do not have such a designated narrow focus position, on the other hand, have to resort do distinct operations (e.g. A-scrambling vs. A’-scrambling in Russian, and stress-shift vs. A’-scrambling in English). It is worth mentioning, however, that in intonation languages, a contrastively focused object either scrambles to an A’-position or stays in its thematic A-position, whereas a narrowly focused subject typically remains in an A-position irrespective of its interpretation as regards contrast. This means that an A’-position is available only for foci that quantify over a discourse-salient set of alternatives, whereas an A-position is compatible with both contrastive and non-contrastive readings. If non-contrastive construal is cross-linguistically available exclusively for arguments in A-positions, we expect the position that precedes na to be an A-position because this positon is used for both contrastive and non-contrastive foci. The next section addresses the syntactic properties of the Akan focus construction and provides support for the A-nature of the focus position.

The first question that we must address in this section is whether the na-construction involves A’-movement25 (Boadi 1974; Korsah and Murphey 2017) or base-generation (Saah 1994;26 Ofori 2011) of the focused XP. This question is easily resolved using reconstruction tests as A’-movement is known to reconstruct to the position of the trace (see Vanden Wyngaerd 1989; Zwart 1993; Jacobs 1997; Haider and Rosengren 1998; Mahajan 1990; Neeleman 1994; and Neeleman and Van de Koot 2008 for a discussion of diagnostics for A- vs. A’-position). An XP that is base-generated in an A-position, conversely, has no position to reconstruct to. Hence, if focused objects in na-constructions undergo A’-movement across the subject, they are expected to reconstruct to their thematic position below the subject, in which case the subject must be able to bind into the A’-moved object and take scope over it.27 Conversely, if focused objects in na-constructions are base-generated in an A-position above the subject, it should be impossible for the subject to bind into the object or take scope over it. As (52) and (53) demonstrate, Akan data support a base-generation analysis.

That is, the sentence in (52A) is impossible on the co-referential reading, suggesting that the variable embedded in the focused object cannot be bound by the subject.28 Similarly, the sentence in (53A) is incompatible with a distributive reading, suggesting that only surface scope is available for this sentence.29 The fact that no reconstruction for binding or scope is available for the focused object strongly suggests that this object is base-generated in its surface position.30

Admittedly, there appears to be variation among native speakers of Akan Asanti Twi as regards judgments related to reconstruction tests. Thus, while Korsah and Murphey (2017) give the sentence in (54) two question marks due to an alleged WCO violation, all of my language consultants judge it as absolutely grammatical on the coreferential reading. In fact, while the coreferential interpretation is readily conceivable in (54), the alternative construal is less plausible and therefore harder to reach.

Similarly, the sentences in (55) and (56) are judged by Korsah and Murphey (2017) as capable of having a pair-list answer, whereas all of my language consultants find such an answer incompatible with either (55) or (56), with only an individual answer possible.31

The following analysis is based on the judgements provided by my language consultants. These judgements indicate that at least in the grammar of these speakers no syntactic reconstruction takes place in Akan na-constructions, strongly suggesting that the surface position of the focused object is not derived by movement.

The conclusion that the focused object in na-constructions is base-generated in its surface position poses the question of the co-occurrence of focused XPs with corresponding resumptive pronouns. In particular, the assumption that the focused XP is generated within the same clause as the resumptive pronoun in the corresponding thematic position immediately runs into a problem with the Theta-Criterion (Chomsky 1981), according to which it is impossible for a predicate to assign the same q-role to more than one argument. Given that the focused XP does not move, we cannot assume that the relevant q-role is assigned to a chain formed by the focused XP and its pronominalized trace – the resumptive pronoun (Boadi 1974; Korsah and Murphey 2017).32 Hence, the only option we are left with is to assume that the focused XP and the resumptive pronoun occur in separate clauses and receive their q-roles from distinct predicates, i.e., the Akan na-construction has a biclausal structure.

It has been noticed by a variety of scholars that Akan focus constructions and constructions involving restrictive relative clauses show striking formal similarities, a fact which suggests that there is some deep relationship between these constructions (Schachter’s 1973; Boadi 1974; Bearth 1999; Dakubu 2005; Ofori and Osam 2006; Schwarz and Fiedler 2007; Ofori 2011; Korsah and Murphey 2017). To begin with, both types of constructions contain resumptive pronouns (see (57c) and (57d)). Yet, the most fascinating resemblance between the two constructions concerns tone. That is, in the formation of both restrictive relative clauses and focus constructions, certain underlying low tones must be replaced by high tones. Strikingly, exactly the same replacements must occur in both construction types (see (57)). (In (57), acute and grave accents represent high and low tone, respectively).

The restrictive relative clause and the focus construction in (57c) and (57d), respectively, differ only with respect to the marker (i.e., relative or focus), while everything else is identical. Crucially, when they are compared to the SVO sentences in (57a) and (57b), it becomes apparent that focus constructions and restrictive relative clauses involve changes in tone that are absent in simple clauses.33

The above observations suggest that a successful analysis of the Akan focus construction must capture its relatedness to constructions involving restrictive relative clauses. A popular trend in the linguistic literature on Akan is to assume that the Akan focus construction, as in (58), is diachronically derived from a cleft construction.34 One such analysis proposes that (58) is derived from the biclausal clefted focused sentence in (59) via deletion of the introductory ɛyɛ ‘it is’ clause (Boadi 1974).

Boadi’s (1974) analysis is based on the observation that the sentences in (58) and (59) are semantically related to extraposed and non-extraposed pseudoclefted sentences, containing restrictive relative clauses, as in (60) and (61). He analyses the copula ne that is used in extraposed and non-extraposed pseudoclefted sentences as the allomorphic variant of na and assumes that ne results from na-yɛ fusion.

Although Boadi’s (1974) analysis successfully captures the semantic and formal relatedness of Akan focus constructions, as in (58), and Akan pseudoclefts, as in (60), it runs into a problem with the Theta Criterion as it assumes a process that results in a monoclausal structure. To be precise, Boadi (1974) assumes that the focused constituent in the Akan focus construction undergoes movement from its thematic position, with the resulting sentence undergoing an obligatory Pronominalization Rule that turns the lower copy into a pronoun. Under this analysis, the corresponding q-role is assigned by the predicate to the chain formed by the moved focus and its pronominalized trace.35 However, we have seen that reconstruction tests do not support a movement analysis. Yet, if the focused constituent is base-generated (Saah 1994; Ofori 2011), the Theta Criterion dictates that it must be generated in a clause that is separate to the one hosting the pronoun. Consequently, the structure of the Akan focus construction must be biclausal.

The hypothesis that I would like to put forward is that the biclausal Akan focus construction, as in (58), is not derived from a cleft construction, but is itself a pseudocleft construction that exists independently in the language. To be precise, the Akan focus construction is an inverse pseudocleft.36 This conclusion is based on the observation that constructions of the type given in (58) can be found in other languages, such as English (see (62)), where they are not taken to be derived from any other construction involving a restrictive relative clause.

The English inverse pseudocleft, as in (62), is not only syntactically parallel to the Akan focus construction, it also has the same IS partitioning. That is, just like the Akan focus construction, the English inverse pseudocleft construction can only be used when the sentence-initial XP is the narrow focus of the sentence, as in (63), but not when it is part of a wider focus, as in (64), or (part of) a background, as in (65). Plausibly, the inverse pseudocleft construction is employed in both languages to signal narrow focus on the sentence-initial XP.37

The analysis of the na-construction as a biclausal inverse pseudocleft solves the problem with the Theta-Criterion, as the focused argument in SpecIP gets its q-role from the CP predicate it c-commands in (66) (Den Dikken et al. 2000).38

Furthermore, the analysis in (66) successfully captures the anti-reconstruction effects in Akan focus constructions. Given the position of the focused DP in (66), it is unsurprising that it fails to reconstruct for binding into the CP containing the resumptive pronoun. After all, this DP is generated in an A-position in the higher IP (see also den Dikken et al. 2000 who show that subjects of inverse pseudoclefts occupy A-positions). As for the surface quantifier scope facts, they also follow from the analysis in (66) under the assumption that quantifiers can take scope over non-clausemate quantifiers which they c-command from an A-position (see (67) and (68)).

Following Ofori (2011), the analysis in (66) takes the focus marker na to be a result of a fusion of the copula ne and the relative marker a.39 Since the copula and the relative marker are productively used in contemporary Akan (see (60)), we can hypothesize that na is merely a result of a contraction of ne [nI] and a [a] that takes place at the PF level in Akan inverse pseudoclefts, as no phonological material ever intervenes between the copula and the relative marker in such constructions. In extraposed pseudoclefts, in contrast, intervening phonological material prevents the relevant contraction from taking place (see (60)). According to Ofori (2011), the lack of morphophonological intervention creates /nI/ and /a/ contiguity, as in (69-i). A consequence of this contiguity is the deletion of /I/ by the morphophonological rule, /I/ Ø /n_+a/ ([a] being the relativizer), and the fusion of [n] and [a] to derive na, which results in the na-focus sentence alternation, as in (69-ii). In line with the above analysis, Ofori (2011) proposes that na is a ‘portmanteau morpheme’ (Hockett 1947; Newman 2000) consisting of ne –copula – and a ‘the relativizer.’40

The above analysis assumes that the focused constituent and the focus marker na do not form a constituent (see (66)). Such a conclusion is supported by the observation that the string containing the focused constituent and the marker does not pass the answer ellipsis or coordination constituency tests (see (70) and (71), respectively).

To summarize, in this section we have established that a focused constituent in Akan focus constructions is generated in an A-position in the higher clause of a biclausal inverse pseudocleft. The resulting structure successfully captures the reconstruction facts, the co-occurrence of focused constituents with resumptive pronouns and the formal and semantic similarities of Akan focus constructions with other types of pseudocleft constructions. Moreover, the Akan inverse pseudocleft has been demonstrated to have the same IS properties as the corresponding construction in English. The focus marker na has been argued to result from a phonological contraction of the copula and the relative marker. This conclusion is further supported by the fact that the focus marker does not form a constituent with the focused XP.

In this paper, we have established that the Akan morphological markers na and de(ɛ) perform the same interpretive function as the prosodic markers of focus and CT in intonation languages. We have determined that the morphological marker de(ɛ) is a marker of CT and that the morphological marker na is the only true morphological marker of focus in Akan. This marker occurs in inverse pseudocleft constructions and is plausibly a result of a phonological contraction of the copula ne and the relative marker a found in Akan extraposed and non-extraposed pseudoclefted sentences, where contraction cannot take place due to phonological interference (Ofori 2011). The inverse pseudocleft construction is used in Akan to mark narrow focus on the constituent in SpecIP of the higher clause. When the relevant constituent is an object, it is additionally associated with the interpretation of quantification over a discourse-salient set of alternatives, which can be achieved via various strategies, including contrastive, emphatic and exhaustive construal. Although this interpretation is available for in-situ objects as well, the SVO structure is largely ambiguous as to the available IS interpretations, suggesting that the pseudocleft is used in order to disambiguate the IS interpretation of the sentence. A parallel has been drawn with intonation languages in that they also employ a marked representation for the aforementioned interpretive disambiguation.

The Akan subject/non-subject asymmetry observed by a variety of linguists has been accounted for by making a reference to the cross-linguistic tendency for a marked encoding of narrow focus on constituents that are not in the default focus set (Reinhart 2006). The observation that Akan uses the same inverse pseudocleft construction for the encoding of narrow focus on the subject and narrow contrastive focus on the object has been argued here to be due to the fact that this construction contains a position associated with narrow focus (i.e., SpecIP of the higher clause), which is an A-position located outside of the clause containing the main verb and its arguments. The resulting structure is capable of encoding not only the thematic prominence of arguments by keeping them in their thematic positions but also the narrow focus on one of the arguments by placing it in a separate clause where it receives a q-role from the CP predicate it c-commands.