Tonal reflexes of movement in Asante Twi


We argue that Asante Twi has a process of tonal overwriting on verbs that are crossed by an A’-dependency. It is shown that this view captures the distribution of the process across ex-situ focus constructions, relative clauses and adverbial clauses, which are all contexts involving operator movement. Furthermore, we illustrate that this process is unbounded and applies to each verb in a long-distance dependency. We therefore conclude that this is a reflex of successive-cyclic movement through vP. Additionally, we provide a detailed study of resumption in Asante Twi, showing that despite island-insensitivity, resumption is still derived by movement. Finally, the morpho-phonological side of the phenomenon is investigated. It is shown that overwriting affects only those affixes below v and not those above, which follows from cyclic Vocabulary Insertion. This provides support for Kandybowicz’ (2017) assumption that aspect and negation are lower than vP in Asante Twi.


There is a well-known, yet still poorly understood, tonal alternation in Asante Twi originally noticed by Schachter and Fromkin (1968). As (1a) shows, both syllables of the verb kita are low-toned in an ordinary declarative clause. However, when the object is wh-moved, both of these tones surface as high (1b).

  1. (1)

The productivity and robustness of this low-high alternation has been independently corroborated at various points in the literature (Schachter 1973; Marfo and Bodomo 2005; Marfo 2005a,b; Fiedler and Schwarz 2005; Schwarz and Fiedler 2007; Ameka 2010; Genzel 2013). Nevertheless, it has generally been assumed that it is a specific quirk of the na-focus construction (e.g. Marfo 2005a,b; Genzel 2013). In this paper, we will argue that the high tone insertion in (1b) is actually a reflex of successive-cyclic A̅-movement through Spec-vP. There are a number of arguments that point to this conclusion, in particular the fact that high-tone insertion is found in a wider range of A̅-constructions, and also that the process is unbounded.

In standard phase theory (Chomsky 2000, 2001), long-distance extraction must proceed ‘successive-cyclically’ (cf. Chomsky 1973, 1977, 1986), making a stop at the edge of each intermediate phase (vP and CP), as shown in (2).

  1. (2)

A considerable body of empirical evidence has amassed in support of the claim that phasal domains constitute (at least) both vP and CP (for overviews, see Boeckx 2008; Georgi 2014, 2017; Citko 2014; van Urk 2015). These varied arguments come from PF interface phenomena (Legate 2003; Adger 2007; Kratzer and Selkirk 2007; Kahnemuyipour and Megerdoomian 2011; Sato 2012), the presence of overt material at intermediate positions (du Plessis 1977; McCloskey 2000; Felser 2004; Wiland 2010; Manetta 2010), interpretational effects along the movement path (Barss 1986; Fox 1999; Nissenbaum 2000; Legate 2003; Sauerland 2003) and the opaque licensing properties of moved items at intermediate positions, e.g. with inversion (Kayne and Pollock 1978; Torrego 1984; Henry 1995), gaps (Thiersch 1978; van Urk and Richards 2015), and discourse particles (Bayer et al. 2016). However, some of the most direct and compelling evidence for successive-cyclic movement comes from languages that exhibit dedicated morphological reflexes of movement (Lahne 2008; Georgi 2014, 2017). For example, Bennett et al. (2012) show that A̅-movement in Defaka triggers the morpheme -ke on the verb (3b). Furthermore, cases of long-distance movement show this marking on verbs in both clauses (3c).

  1. (3)

    ke-marking in Defaka (Bennett et al. 2012:296f.):

Tonal overwriting in Asante Twi has a very similar profile, as it also affects all verbs crossed by A̅-movement in a long-distance dependency. In (4), the low-toned syllables in both the matrix and the embedded verb alternate to high in the wh-movement construction in (4b).

  1. (4)

We will therefore argue that tonal overwriting on the verb tracks whether successive-cyclic movement has taken place in that clause. In particular, we analyze overwriting as being triggered by a floating high-tone that is the realization of a phase head v bearing a checked edge feature. This adds to the already significant body of literature on reflexes of successive-cyclic movement, but also provides a good example of a reflex of movement at the vP level, which is still a somewhat contentious issue (e.g. Keine 2016, 2017; Dayal 2017). Furthermore, tonal overwriting provides a clear case of a purely phonological reflex of syntactic movement. While similar phenomena have been reported, e.g. downstep deletion in Kikuyu (e.g. Clements et al. 1983; Zaenen 1983; Clements 1984a,b), downstepping in Igbo (Amaechi and Georgi 2019), and soft mutation in Welsh (Willis 2000; Borsley et al. 2007), they have either not yet been sufficiently explored or come with additional complications and caveats.

In addition, we provide evidence that, despite showing island-insensitive resumption, A̅-constructions in Asante Twi are derived by genuine syntactic movement, tracked by tonal overwriting. This conclusion is supported by the fact that island effects correlate with the availability of a resumptive pronoun and by the divergent properties of genuine base-generated topic constructions. Finally, we also provide a systematic investigation of the scope of the overwriting process, showing that only affixes originating lower than vP are affected by high-tone insertion. It is shown that this affix generalization follows from independently-motivated assumptions about clause structure in Asante Twi put forward by Kandybowicz (2015).

This paper is organized as follows: Sect. 2 outlines the phenomenon of tonal overwriting, reviews previous approaches and provides additional data motivating a movement-based analysis. The following Sect. 3 justifies the movement analysis in light of island-insensitive resumption, demonstrating that a movement analysis is still well-motivated. Section 4 discusses the morpho-phonological side of the phenomenon. It is shown that only certain types of affixes are subject to overwriting and that this follows from the syntax of the verbal domain. Finally, Sect. 6 concludes.

High tone overwriting in Asante Twi

Asante Twi is a dialect of the Kwa language Akan, spoken in Ghana (Dolphyne and Kropp Dakubu 1988; Kropp Dakubu 2009). It is an SVO, terraced-level tone language with a distinction between high and low tones, as well as downstep. The syllable is typically assumed to be the tone bearing unit (see Dolphyne 1988:53; Kügler 2016b:92f., cf. Abakah 2005:110ff.). An example of a simple declarative sentence in Asante Twi is given in (5).Footnote 1

  1. (5)

Since Schachter and Fromkin (1968), it has been reported that Asante Twi exhibits a tonal overwriting processes where underlying low-toned verbs become high in particular contexts. The most widely-discussed context involves ex situ focus constructions (see Schachter 1973; Boadi 1974; Marfo and Bodomo 2005; Marfo 2005b; Fiedler and Schwarz 2005; Genzel 2013). This focus strategy involves displacement of a constituent to the left of the focus particle na (see, e.g. Boadi 1974; Saah 1988; Ameka 1992; Ermisch 2006, 2007; Amfo 2010; Genzel and Kügler 2010; Ofori 2011; Pfeil et al. 2015). In this construction, it has been noted that verb roots with underlying low tones surface as high (sometimes referred to as the ‘link tone’; Fiedler and Schwarz 2005:115). In (6a), the verb (‘be’) bears a low tone, however in the corresponding example (6b) with subject focus, the tone of the verb changes to high.

  1. (6)

The same effect can be seen in the following example. The verb ba (‘come’) is low-toned in a discourse-neutral declarative clause (7a). If the subject is focused, the tone on the verb surfaces as high (7b). It is particularly interesting to note that the low-toned past tense marker -a and the person marker me- are not affected. We return to this in Sect. 4.1.

  1. (7)

High tone overwriting also applies if an object is fronted. Furthermore, the process extends to fronting of wh-phrases, since ex situ wh-questions are also a sub-type of the na-focus construction. As (8) shows, the verb becomes high-toned when wh-object is moved.

  1. (8)

In the corresponding in situ variant, the low tone on remains unaffected.

  1. (9)

An important point, to which we return in detail in Sect. 3, is that extraction of an animate DP triggers obligatory resumption in the base position (10b). Even with resumption, we still observe high tone overwriting.

  1. (10)

In bisyllabic verbal roots, the process of tonal overwriting also singles out low tones. For example, the LL root kita in (11a) becomes HH in the presence of an extracted object (11b).Footnote 2

  1. (11)

The same can be seen with the HL stems such as nóm (‘drink’) which also bear the HH tonal sequence in the presence of wh-movement (12b).

  1. (12)

Although the existence of this process has been rather widely reported and experimentally corroborated (e.g. Genzel 2013:208; Korsah and Murphy 2016:229ff.), the nature of it is still poorly understood. Most of the previous literature assumes that high-tone overwriting is a construction-specific quirk of the na-focus construction (e.g. Boadi 1974:19; Marfo 2005b:79; Genzel 2013:207f.). This becomes particularly clear in Marfo (2005b), which is to our knowledge the only explicit analysis of this phenomenon to date. We will illustrate his analysis on the basis of the example in (13b).

  1. (13)

Marfo (2005b) proposes the following derivation for the na-focus construction in (13b). First, the syntactic structure in (14) is mapped onto phonological phrases, roughly corresponding to each XP (14a) (cf. Truckenbrodt 1999; Selkirk 2011). Subsequently, a rule of ‘focus restructuring’ moves the boundary of the initial ϕ-phrase to include the Spec-TP position, i.e. the subject in (14b). The next process is ‘prosodic raising’ that maps the expanded ϕ domain created in the previous step onto the next highest domain in the Prosodic Hierarchy (Selkirk 1986; Nespor and Vogel 1986), i.e. the intonation phrase (I) (14c). Now, it is important to note that Marfo (2005b) assumes that there is an ‘inserted’ floating H-tone which is lexically associated with the focus marker na. For reasons that are relatively unclear, ‘this inserted H prefers to dock on a constituent at the left-edge of a succeeding I, irrespective of its syllable structure. This left-edge constituent happens to be the verb’ (Marfo 2005b:108ff.) (14d). After the high tone has been repositioned, it triggers spreading throughout the verbal stem, as in (14e), a process Marfo (2005b:109) refers to as the ‘inserted H spread rule.’

  1. (14)

    High tone overwriting by prosodic restructuring (Marfo 2005b:108f.):

      [FocP Kofí [Foc′ na [TP [VP re-boá Á!bénáá ]]]]  
    a. (Kofí naH-)ϕ (ɔ-)ϕ(re-boá)ϕ!bénáá)ϕ (Mapping XPs toϕ)
    b. (Kofí naH- (ɔ-)ϕ) (re-boá)ϕ!bénáá)ϕ (Focus Restructuring)
    c. (Kofí naH- (ɔ-)ϕ)I ((re-boá)ϕ!bénáá)ϕ)I (Prosodic Raising)
    d. (Kofí na (ɔ-)ϕ)I (H-(re-boá)ϕ!bénáá)ϕ)I (H-tone shift)
    e. (Kofí na (ɔ-)ϕ)I (H-(ré-bóá)ϕ!bénáá)ϕ)I   (Inserted H spread rule)

Many aspects of Marfo’s (2005b) analysis are problematic, however. First, there does not appear to be any independent motivation for the prosodic ‘restructuring’ processes he proposes, beyond the data they are designed to capture. Furthermore, the core assumption of his analysis is that the high tone that triggers spreading in the verb stem is associated with the focus marker na. The main motivation for this comes from the observation that the superficially similar déέ-construction does not trigger tonal overwriting, as shown by (15b).

  1. (15)

However, it will be argued that this assumption cannot be maintained in the light of new data from long-distance extraction. Furthermore, closer inspection reveals that déέ-constructions do not share the same movement derivation as their na counterparts, despite their putative semblance (see Sect. 3.4). Instead, we argue that the tonal alternation in Asante Twi is a reflex of successive-cyclic movement. Descriptively, whenever A̅-movement takes place in a clause, then the low tones of the verb in that clause are replaced by high tones. More technically, it will be shown that this can be modelled as the allomorphic realization of a v head bearing an edge feature as a floating high tone. The following section provides arguments against the characterization of overwriting by high tones as an idiosyncratic property of the morpheme na, and in favour of an analysis in terms of a phonological reflex of A̅-movement.

Tonal overwriting as a reflex of Ā-movement

This section presents evidence in support of the claim that high tone overwriting in Asante Twi verbs is triggered by the presence of an A̅-dependency, and is not specific to the na-focus construction. The first piece of evidence involves new data from long-distance dependencies where it is shown that the process affects all verbs along the extraction path. The second argument comes from the fact that the tonal overwriting is found in contexts other than focus constructions, namely relative and adverbial clauses. These contexts can all be unified as involving an A̅-dependency of some kind.

Long-distance dependencies

Recall that for examples such as (16), Marfo (2005b) proposed that the high tone that triggers overwriting on the verb is lexically associated to the na focus particle. Consequently, the tonal overwriting process is a construction-specific property of na-constructions.

  1. (16)

What previous literature on this phenomenon neglected to consider was long-distance dependencies. These reveal a very important characteristic of the low/high-alternation, namely that it is an unbounded process. If extraction takes place from an embedded clause to a focus position in the matrix clause, then both the matrix and embedded verbs exhibit high-tone overwriting. In (17b), the matrix verb kaé (‘remember’) and the embedded verb kita (‘hold’) both surface with high tones.

  1. (17)

The same effect can be seen with different verbs in (18).

  1. (18)

    (cf. Boadi 1990:81)

The unbounded nature of this process is further illustrated by example (19), which contains two levels of embedding. In the case of extraction, the low tones on both the embedded verbs and the matrix verb alternate to high (19b).

  1. (19)

Consider this now in the light of Marfo’s (2005b) analysis discussed in the previous section. The intuition is that a floating H tone originates on na and then docks onto the closest verb. Such an analysis cannot account for the fact that all verbs along the extraction path are raised to high, since there is only ever a single high tone associated with the focus marker. Thus, it seems impossible to maintain the idea that the high tone originates on na in the light of these data. If one wished to maintain the construction-specific view of tonal overwriting, one could formulate a rule stating that all verbs in a na-focus construction are raised to high.

However, this would be problematic when we consider (20). As (20b,d) show, extraction of a matrix argument does not affect the embedded verb. Thus, it is not only the construction-type, but also the origin of the displaced constituent that must be taken into account when defining where tonal overwriting takes place.

  1. (20)

On the other hand, these data make perfect sense under the view that high tone overwriting on verbs tracks the presence of A̅-movement in that clause. Thus, this kind of tonal overwriting patterns exactly like other established reflexes of successive-cyclicity in that they are only found in domains crossed by movement, for example in Kikuyu (Clements 1984a:47), Defaka (Bennett et al. 2012:297), Kitharaka (Muriungi 2005:48) and Indonesian (Saddy 1991:194).Footnote 3

Relative clauses

Another argument against the idea that tonal overwriting is specific to the na-construction comes from the fact that the low/high alternation is also found in relative clauses. Relative clauses in Asante Twi are not obviously related to na-focus constructions. Instead, they involve the relative complementizer áa preceding the pivot constituent of the relative clause (see Saah 2010; McCracken 2013). As (21) shows, the underlyingly low-toned verb hu (‘see’) bears a high-tone when it occurs inside a relative clause (21b).

  1. (21)

Although much previous work on this tonal alternation has focused exclusively on focus constructions, its presence in relative clauses can be found in earlier literature, i.e. since at least Schachter (1973:23). For example, while Saah (2010) does not discuss the tonal alternation on the verb, it can be seen in many of his examples, such as (22b).

  1. (22)

Fiedler and Schwarz (2005:122), on the other hand, explicitly mention the tonal alternation, reporting that in (23b) ‘the verb in the relative clause changes its tone pattern in adopting a H tone.’

  1. (23)

Similar to the na-construction discussed in the previous section, this tonal overwriting is also an unbounded process, triggered on all verbs crossed by long-distance relativization (24).

  1. (24)

The existence of the low/high-alternation inside relative clauses also follows from our hypothesis that it is a reflex of successive-cyclic movement, since both head-internal and head-external approaches to relative clauses posit some kind of A̅-movement originating inside the relative CP (for overviews, see e.g. Alexiadou et al. 2000; de Vries 2002; Bhatt 2015; Salzmann 2017b). At this point, it is important to mention the fact that we find resumption inside relative clauses (and also in na-constructions) if the tail of the dependency corresponds to an animate DP. Nevertheless, Sect. 3 argues that these resumptive dependencies still involve genuine movement.

Adverbial clauses

A further context beyond the na-focus construction where the tonal alternation can be found is adverbial clauses. Dolphyne (1988:69) points out that there is a distinction between the tonal patterns of verbs in what she calls a ‘subordinate clause.’ What Dolphyne (1988) is actually referring to is tone raising inside an adverbial clause, since this effect is not found in clausal complements. An example of this is given in (25).

  1. (25)

Although we observe the low/high-alternation here, this is clearly not an instance of a focus construction (note that high-toned here is a subordinating conjunction). The fact that there does not seem to be any overt movement inside the adverbial clause could be viewed as a challenge to our claim that tonal overwriting is a reflex of A̅-movement. However, it has already been argued on independent grounds that there is movement of a covert operator in adverbial clauses (see e.g. Geis 1970, 1975; Larson 1987, 1990; Johnson 1988; Thompson 1995; Demirdache and Uribe-Etxebarria 2004; Haegeman 2007, 2010; Zentz 2014; Kusumoto 2017). The main empirical motivation for this comes from ambiguities found in temporal adverbial clauses such as those in (26), originally noticed by Geis (1970).

  1. (26)
    1. a.

      Joan left before Harry told her to.   (Geis 1970:127)

    2. b.

      I saw Mary in New York after John said that she left. (Larson 1987:261)

    3. c.

      The professor wrote the letter after the student said he needed it. (Zentz 2014:375)

In each of these cases, there are two possible readings: one in which the adverbial refers to the time of the matrix event (e.g. saying), and another where it refers to the time of the embedded event (leaving or needing). One influential analysis of this ambiguity involves positing movement of a null operator inside the adverbial clause (see e.g. Geis 1970:127; Larson 1987:261, 1990:178; Demirdache and Uribe-Etxebarria 2004:171; Haegeman 2007:293). If the temporal operator is merged in the matrix clause of the adverbial, then we derive the high construal when it moves to Spec-CP (27a). However, the operator can also be merged in the embedded clause and then moved successive-cyclically to the matrix clause to derive the low construal (27b).

  1. (27)

    I saw Mary in New York after John said that she left.

    1. a.

      [PP after [CP Op1 [TP John said [CP that she left ] t1 ]]]

      High construal: ‘The speaker saw Mary after hearing from John that she had left.’

    2. b.

      [PP after [CP Op1 [TP John said [CP t1 that she left t1 ]]]]

      Low construal: ‘The speaker saw Mary after the point of her reported departure.’

Assuming that there is also A̅-movement of a null operator in Asante Twi temporal adverbial clauses, such as (25b), then the presence of tonal overwriting is unsurprising. Cross-linguistically, we also find movement reflexes in adverbial clauses in other languages such as Irish (McCloskey 2001) and Akɔɔse (Zentz 2014). In (28), the movement-related complementizer aL that triggers lenition is found inside a temporal adverbial (see Sect. 2.2 for more discussion of Irish).

  1. (28)

If tonal overwriting in Asante Twi adverbial clauses results from movement of a null operator, then we might expect the presence or absence of the movement reflex to correlate with different interpretations. While Geis ambiguities of the kind found in English prove rather difficult to reconstruct faithfully in Asante Twi, the following example is highly suggestive of such an effect:

  1. (29)

In (29), the adverbial clause clearly refers to the time of the saying event, rather than the state expressed by the verb (‘like’). As such, we have an obligatory high construal of the temporal operator, i.e. no movement takes place in the embedded clause. This is reflected by the fact that tonal overwriting is found on the matrix verb (on the prefix re), but not on the embedded verb , suggesting that the null operator originates in the matrix clause in (29).

A movement-based analysis

This section will try to formalize the assumption that tonal overwriting is a reflex of movement. The standard approach to locality in Minimalism comes from phase theory (Chomsky 2000, 2001, 2008). Central to phase theory is the Phase Impenetrability Condition (30).

  1. (30)

    Phase Impenetrability Condition (Chomsky 2000:108):

    In phase α with head H, the domain of H is not accessible to operations outside α, only H and its edge are accessible to such operations.

While the exact inventory of phase heads is still a matter of debate, there is a general consensus that it includes at least vP and CP, as proposed by Chomsky (2001:13).Footnote 4 The consequence of this is that a wh-phrase is not accessible to a probe on C once the complement of v has undergone Spell-Out (31a). In order to allow for wh-movement of objects, it was proposed that ‘edge features’ can be added to phase heads in order to allow moving operators to escape transfer (Chomsky 2000:109, 2001:34; Chomsky 2008:149; also see Müller 2011:2f. and Kandybowicz 2009:321f.). Thus, in order for a wh-object to move to Spec-CP, an edge feature (that will simply be represented as an EPP feature, following Chomsky 2000) is added to the v head, triggering movement to its specifier (31b). Subsequently, the wh-phrase is accessible to the movement probe on C and the wh-phrase will move its criterial position in Spec-CP (31c).

  1. (31)

The result of this is that long distance extraction must involve movement through each intermediate Spec-vP and Spec-CP position, as shown in (32).

  1. (32)

    Long distance successive-cyclic movement:

In (32), each of the heads through which successive-cyclic movement passes bears an edge feature. A straightforward way of capturing morphological reflexes of successive-cyclicity is to make the realization of a particular head sensitive to whether or not it bears such a feature. In other words, we can posit a contextual allomorph of certain phase heads that tracks whether or not that head bears an edge feature, and thus indirectly whether or not successive-cyclic movement took place via the specifier of that head.Footnote 5 Consider the classic case of Irish complementizer allomorphy in (33). While the ordinary declarative complementizer takes the form go as in (33a), the presence of an A̅-dependency in a clause leads to the other complementizer form a (33b, c).

  1. (33)

McCloskey (2002) analyzes this reflex of movement by specifying distinct Vocabulary Items that track whether a C head has triggered intermediate movement steps, i.e. whether a head bears an edge feature (however see Noonan 1999, 2002 and Bošković 2008:209, fn. 19 for some criticism of this view).Footnote 6 In clauses in which no movement has taken place, the form go simply realizes the category feature of C (34b). However, if the C head also bears an EPP feature (i.e. edge feature) that was inserted to facilitate successive-cyclic movement, then this complementizer will be realized by the aL (34a) rather than go, given standard assumptions about the ordering of VIs relative to specificity (e.g. Halle and Marantz 1993:124; Halle 1997:428; Embick 2015:95).

  1. (34)

    Vocabulary Items for Irish (cf. McCloskey 2002:203):

    a. [c, epp] aL
    b. [c] go

We can then adopt an entirely analogous approach for tonal overwriting in Asante Twi. For long distance wh-movement such as (35b), there will be successive-cyclic movement through each vP and CP projection triggered by an edge feature on the corresponding phase head, as in (32).Footnote 7

  1. (35)

The fact that tonal overwriting occurs on verbs points to the conclusion that this reflex is on v rather than C. As a first approximation, overwriting can be analyzed as a floating tone that spreads through the verb. This is captured by the following Vocabulary Items for v in Asante Twi, where the more specific exponent of v with an edge feature is realized as a floating H tone:

  1. (36)

    Vocabulary Items forv(preliminary):

    a. [v, epp] H-
    b. [v] Ø

The core intuition will be that the floating H tone realized on v then triggers ‘tonal overwriting’ throughout the v + V complex, as shown in (37). The finer details of how tonal overwriting is derived, as well as discussion of which affixes are affected, are provided in Sect. 4.

  1. (37)

    Tonal overwriting ofv+V :

For cases such as (20b), where movement only takes place in the matrix clause, an edge feature will only be inserted on matrix v. Thus, the floating H tone will only trigger overwriting on the matrix verb.

Movement and resumption in Asante Twi

So far, we have not yet addressed the fact that we often find resumptive pronouns in the constructions which we are assuming to involve movement. In many languages, resumptive A̅-dependencies do not show the typical properties of movement (i.e. island-sensitivity, reconstruction) and have often been analyzed as base-generation. If this were true for Asante Twi, this would undermine the analysis of tonal overwriting as a reflex of movement, since it is also found with resumption. In this section, we will argue that, despite the lack of island-sensitivity, resumption is still best treated as the result of a movement derivation.

Descriptively, the pattern we observe is that resumptive pronouns obligatorily appear in the base position of animate, but not inanimate DPs (e.g. Saah 1988:23f.; 1994:101f.; Saah and Goodluck 1995:383). Consider the following example involving a double object construction. If the animate indirect object is extracted, then the resumptive pronoun must appear in its base position (38b). However, if the inanimate direct object is extracted (38c), then no resumptive pronoun may appear, and we instead find a gap.

  1. (38)

Taking this pattern at face value, it is conceivable that dependencies with gaps and resumptives could have radically different derivations. For example, it has been proposed that whereas displaced constituents leaving a gap are generated by movement (39a), resumption involves base-generation and binding (39b) (e.g. Chomsky 1977:80f.; Bayer and Salzmann 2013:305).

  1. (39)

These two different types of structures have been argued to be discernible on the basis of movement diagnostics such as island-sensitivity. For example in Hebrew, an object resumptive can occur inside a Complex NP Island (40a), but an object gap may not (40b) (see McCloskey 1979:32ff. for similar facts in Irish).

  1. (40)

This would then seem to follow from the structures in (39); a movement derivation is postulated only when we find island-sensitivity, i.e. with gaps. A somewhat surprising fact about Asante Twi is that we find island-sensitivity neither with resumptives (41) nor with gaps (42) (Saah 1994; Saah and Goodluck 1995; Goodluck et al. 1995):

  1. (41)
  1. (42)

This led Saah (1994) to propose that both constructions involve base-generation of a constituent in the left-periphery, which binds a resumptive pronoun in situ. A̅-constructions with gaps (42) are assumed to involve binding of a null resumptive pronoun (Saah 1994:172f.). Such a derivation then contradicts the present proposal for tonal overwriting since it would mean that long-distance dependencies would not involve movement. In fact, it seems difficult to see what property of A̅-constructions high tone overwriting on verbs would be tracking. One possibility would be successive-cyclic binding. In Irish, there is a dedicated complementizer form for resumptive dependencies, namely aN (McCloskey 1979, 2001, 2002). Long-distance dependencies involving resumptives, and which involve the complementizer aN are insensitive to islands (McCloskey 1979:34). This leads to the conclusion that such constructions do not involve movement. For resumptive dependencies, McCloskey (2002:199) reports that it is possible for each complementizer between the binder and resumptive to be marked with the complementizer aN, which reflects base generation of a proform in its specifier. This can be seen in (43a), which is analyzed as a chain of successive binding dependencies reaching down to the resumptive pronoun (RP) (43b).

  1. (43)

In this analysis, we observe a particular reflex on all heads between the RP and its binder. However, there is no necessity to have binding be successive-cyclic. Since base-generation is mainly motivated by island-insensitivity, then binding by pro must be able to cross clause boundaries. Indeed, this means that the intermediate positions need not be filled by other binders, giving rise to ‘mixed’ chains such as (44a). Here, the intermediate C head does not select a pro binder in its specifier and therefore surfaces in its default form go (44b) (McCloskey 2002:190).

  1. (44)

The major difficulty in developing an analogous proposal for Asante Twi is that it does not have mixed chains of the kind found in Irish. Furthermore, we find a wide range of movement diagnostics (e.g. crossover and reconstruction effects) that are difficult to capture in the absence of movement (see Sect. 3.3) (see Adger and Ramchand 2005; Rouveret 2008; Pan 2016 for Agree-based analyses for resumptives without reconstruction effects).

Instead, we will argue that a movement-based account of tonal overwriting can be maintained under the assumption that resumption amnesties island violations. This effect of so-called ‘intrusive’ resumption is even found in languages without grammatical resumption (45) where it functions as a kind of Last Resort (cf. Shlonsky 1992).

  1. (45)

    Resumptives inside islands in English (Ross 1986:260f.):

    1. a.

      I just saw that girl who1 [DP Long John’s claim [CP that she1 was a Venusian ]] made all the headlines.

    2. b.

      King Kong is a movie which1 you’ll laugh yourself sick [CP if you see it1 ].

The following sections will show that once we take the full range of facts into account, a much clearer picture emerges: extraction of a DP always triggers resumption in its base-position. However, there are independent constraints on the realization of inanimate pronouns that also apply to resumptive pronouns. Following Korsah (2017), it will be shown that movement of inanimate DPs does, in fact, show overt resumption in a well-defined set of contexts. This means that pseudo-gaps such as those in (42) are actually phonologically null resumptive pronouns. If it is a general property of Asante Twi resumptives that they circumvent island violations, then the lack of island-sensitivity with nominal extraction can be straightforwardly accounted for. This conclusion will be bolstered by the fact that island effects are found with extraction of categories that lack resumptives, namely VPs and PPs. Since resumptive pronouns also show a wide range of reconstruction effects, we conclude that resumptives are the phonological realization of lower copies generated by movement. The final piece of the puzzle is provided by the déέ-construction, which lacks tonal overwriting. While this was previously taken as evidence for the construction-specific nature of high tone insertion, we show that this construction systematically lacks reconstruction effects, and therefore involves base-generation rather than movement. This lends further support to the conclusion that A̅-movement is the trigger of high tone overwriting, even in the presence of resumption.

On nominal resumption in Asante Twi

Recall that examples such as (38) initially suggest that extracted animate arguments require resumptive pronouns (46a), while inanimate object arguments seem to leave gaps in their base position (46b) (e.g. Saah 1992). However, this section shows that the ‘gap’ in (46b) is actually a phonologically null resumptive pronoun, subject to a general process of pro-drop.

  1. (46)

The fact that third person inanimate object pronouns are obligatorily phonetically null has long been observed in the literature on Akan grammar (cf. Riis 1854:60; Christaller 1875/1964:85). As the following examples from Osam (1996:160) show, a pronoun with an inanimate referent is obligatorily null (47b).

  1. (47)

Although this is the general rule, there are at least three contexts in which inanimate pronouns are obligatorily pronounced. These include clause-final adverbs (48a) (Saah 1994), change-of-state verbs (48b) (Boadi 1971; Osam 1996), and those with secondary predicates (48c) (Korsah 2017).

  1. (48)

Putting aside what unifies these contexts for a moment, it is clear that the expected application of the inanimate pro-drop rule in Asante Twi is blocked in the contexts in (48). Furthermore, we see the same effect with resumptive pronouns corresponding to inanimates. While inanimate resumptives must normally be dropped (49a), as we saw in (42), the presence of a clause-final VP-adverb forces it to appear overtly (49b).Footnote 8

  1. (49)

Again, while we do not find resumptives corresponding to inanimate objects with most verbs (50a), change-of-state predicates such as bu (‘break’) trigger overt resumption (50b).

  1. (50)

Finally, when the antecedent of a resumptive is an argument of a secondary predicate, it cannot be omitted, as shown in (51) (also see Korsah 2017:25).

  1. (51)

From the preceding discussion, it is clear that the absence of inanimate resumptives follows conditions on the realization of object pronouns more generally. Showing that this effect is also found in other Kwa languages, Korsah (2017) argues that inanimate object pronouns regularly undergo PF deletion and shows that all contexts where the object pronoun is obligatorily pronounced constitute situations where it has escaped the deletion operation, for principled reasons. We refer the interested reader to Korsah (2017) for details of the analysis. As far as the present paper is concerned, we will assume that the absence of an inanimate resumptive pronoun in Asante Twi actually involves phonological non-realization of a syntactically-present resumptive pronoun.

Islands appear with true gaps

The previous section established that the ‘gaps’ found with extracted inanimate DPs are actually phonologically unrealized resumptive pronouns. Thus, they behave on a par with overt resumptives in circumventing island violations. A clear prediction of this would be that, if an extracted category lacks a resumptive pronoun, then island effects should re-emerge. This is exactly what we find with extracted PPs and VPs. In (52a), we observe that Asante Twi has PPs headed by the postposition (‘in’). While it is possible to extract this constituent in focus constructions, such as (52b), it does not leave a resumptive pronoun in its base position. We can verify that this pronoun is not just phonologically deleted, as was the case with inanimate DPs, by adding a clause final adverb. We observe that there is still an obligatory gap (52c).

  1. (52)

Furthermore, PPs can also undergo long-distance movement across a finite clause boundary (53).

  1. (53)

However, PP extraction differs from DP extraction in that it is sensitive to Complex NP islands (54) and wh-islands (55).

  1. (54)
  1. (55)

The difference in island obviation between PP and DP extraction follows naturally if it is linked to the availability of a resumptive pronoun.

Another extractable constituent that lacks resumptives is the verb phrase. As Hein (2017) shows, VPs can undergo long-distance focalization (56).

  1. (56)

    (Hein 2017:9)

Since Asante Twi lacks VP proforms and thereby resumptive pronouns, the clear prediction is that extraction of VPs should also show island effects. As Hein (2017) reports, this is precisely what we find. Movement of a VP out of a Complex NP or wh-island is ungrammatical (57).

  1. (57)

These facts strongly support the idea that the lack of island-sensitivity with nominal extraction is linked to obligatory resumption. As soon as a resumptive strategy becomes impossible, as with PPs and VPs, then island effects re-emerge. Of course, this could still mean that the derivation of PP and VP focalization involves movement, whereas nominals are exclusively base-generated. Aside from the conceptual objection that a unified approach to focalization (via movement) would be most desirable, we actually find a number of properties of resumptive dependencies that are typically associated with movement, e.g. reconstruction and crossover effects. These are presented in the following section.

Movement diagnostics

Binding reconstruction

The first diagnostic we will consider involves reconstruction of a moved phrase to satisfy the conditions on binding (Chomsky 1981). As Saah (1989:18) shows, the reflexive in Asante Twi is subject to Condition A, in that it must be bound be a local c-commanding antecedent (58a). In order to test whether this is truly reconstruction in the presence of resumption, we create an analogous embedded context to (58a) and include a clause-final adverb, which will force an inanimate resumptive to appear. As (58b) shows, coreference between the reflexive and the embedded subject is possible, thereby indicating reconstruction to the position of the resumptive.

  1. (58)

Further evidence for reconstruction to the position of the resumptive pronoun comes from variable binding (Aoun et al. 2001; Sichel 2014). This effect has been shown for Arabic by Aoun and Benmamoun (1998). A fronted expression containing a variable bound by a quantifier reconstructs at the resumption site (59a). If the resumptive is higher than the binder (59b), the result is ungrammatical due to lack of c-command.

  1. (59)

In Asante Twi, we see a similar effect. A variable bound by a quantified expression is still grammatical even if this DP is fronted with a resumptive in its base position (60b).

  1. (60)

However, it is important to establish that reconstruction for variable binding is also found across clausal boundaries. Schneider-Zioga (2009) has argued for Kinande that, while reconstruction for variable binding is found in monoclausal environments (61a), it becomes impossible under non-local extraction (61b).

  1. (61)

This leads her to propose a non-movement based account of long-distance dependencies in Kinande, involving successive operator-binding configurations, similar to the analysis of the Irish example in (43) (iterative prolepsis; also see Finer 1997; Davies 2003; Boeckx 2008:98). Importantly, reconstruction effects in Asante Twi also pertain in bi-clausal contexts, as (62) shows for variable binding (see Sect. 3.6.1 for islands).

  1. (62)

This would be unexpected on an iterative prolepsis account of long A̅-dependencies, but is entirely consistent with the approach here.

Weak crossover

Another potentially revealing diagnostic comes from ‘weak crossover’ (WCO) effects, which are found when an extracted operator crosses a coreferent element in a non-commanding position (63a) (e.g. Postal 1971; Higginbotham 1980; Koopman and Sportiche 1982; Safir 1984; Lasnik and Stowell 1991; Ruys 2000).

  1. (63)

In Asante Twi, resumptive A̅-dependencies give rise to WCO effects both with pronouns (64b, c) and lexical epithets (64d).Footnote 9

  1. (64)

    Weak Crossover in Asante Twi:

The existence of WCO with resumption in Asante Twi provides evidence that the resumptive pronoun behaves like a trace of the A̅-movement in (63). In fact, this is what is reported for Vata by Koopman and Sportiche (1982), where resumptives behave like traces in triggering WCO effects (65).

  1. (65)

    Weak Crossover with resumptives in Vata (Koopman and Sportiche 1982:143):

One could try to argue that base-generation and A̅-binding of the resumptive also creates the same configuration for WCO. However, in languages like Irish where such an analysis has been proposed, we do not find WCO effects with resumption (66a), unlike with gaps (66b) (also see Safir 1996 on English).

  1. (66)

Thus, it seems that resumptives resulting from base-generation and binding are often not sufficient to induce WCO in other languages. Consequently, the fact that Asante Twi shows weak crossover effects suggests that the resumptive is generated by movement, as with island-sensitive resumption in Vata for example.

Scope reconstruction

More evidence for reconstruction to the position of a resumptive involves quantifier scope ambiguities. In particular, the availability of pair-list readings in wh-questions with a universal quantifier is often assumed to involve reconstruction of the wh-phrase to a position below the quantifier (e.g. Agüero-Bautista 2001; Panitz 2014). Consider the following Spanish example from Agüero-Bautista (2001:172), where a pair-list answer is reported to be possible even in the presence of a resumptive pronoun in the base position of the extracted wh-phrase:

  1. (67)

In Asante Twi, we also find that a resumptive dependency still permits wide-scope of the universal quantifier, i.e. a pair-list reading (68).

  1. (68)

This is important since it rules out the possibility of deriving this scopal relation via Quantifier Raising over the wh-phrase (É. Kiss 1993), since QR is typically assumed to be a clause-bound process (e.g. May 1985; Larson and May 1990). Thus, the only way to derive the availability of a pair-list reading in (69) is by reconstruction to the position of the resumptive pronoun.

Idiom reconstruction

A final diagnostic for reconstruction of displaced material to the position of the resumptive comes from the interpretation of idiom chunks (Brame 1968; Schachter 1973; Vergnaud 1974). Following Chomsky (1993:38f.), a standard assumption is that the idiomatic interpretation of phrasal idioms requires adjacency at LF. Thus, if part of an idiom chunk is A̅-moved (69), then it must reconstruct to its base position in order to receive a non-literal, idiomatic interpretation.

  1. (69)

In Asante Twi, as in many other Kwa languages, there are so-called inherent complement verbs (ICVs), which are fixed VP chunks that have a non-literal, idiom-like interpretation (Nwachukwu 1985; Essegbey 1999, 2010; Korsah 2016b). For example, the VP to ndwom (‘throw song’) has the non-compositional interpretation ‘to sing’ (70b) (Kandybowicz 2015:266). We find that, even when the complement of an ICV is extracted, the idiomatic interpretation is maintained. In (70b), resumption is obligatory with extraction of ndwom due to the clause-final adverb, nevertheless the idiomatic interpretation is still available. From this, we can conclude that the NP reconstructs to the position of the resumptive at LF.

  1. (70)

Similar facts hold in other languages. For example, in Hebrew relative clauses with resumption we also find idiom reconstruction (71).

  1. (71)

Base generation in déέ-constructions

Further evidence for the fact that resumptive dependencies involve movement comes from the déέ-construction. Recall it was contrasts such as those in (15), repeated below as (72), that led Marfo (2005b) and others to conclude that high tone overwriting was an idiosyncractic property of the na-construction. As (72b) shows, there is a superficially similar construction involving the left-peripheral particle déέ that lacks the tonal reflex entirely.

  1. (72)

However, the déέ-construction turns out to be the exception that proves the rule. Under the present analysis, the lack of tonal overwriting in (72b) must indicate that these constructions are not derived by movement and, in fact, there is evidence that the déέ-construction involves a base-generated left-peripheral topic.

The déέ-construction has sometimes been described as a variant of the focus construction, but this seems to be a mischaracterization. Boadi (1974) and Saah (1994:142) assume that déέ is a marker of (non-exhaustive) focus, however it actually displays the properties of a topic construction (see e.g. Saah 1992:236f.; Ermisch 2006:58f.). For example, the pivot of the déέ-construction cannot contain new information, e.g. in the question-answer pair in (73) (cf. Marfo 2005b:93).

  1. (73)

Furthermore, wh-phrases, which constitute new (non-given) information, are not possible in the déέ-construction (as originally noted by Boadi 1974:53), further supporting the conclusion that déέ is a marker of topic, rather than focus:

  1. (74)

Having established that déέ-constructions are topic constructions, it is interesting to note that there is a cross-linguistic tendency for topic constructions to involve base-generation. If this were also the case for Asante Twi, then the absence of movement reflexes in these constructions would not be surprising. For example, Collins (1993, 1994) shows that Ewe has a reflex of successive-cyclicity involving optionality in the form of the pronoun in embedded clauses. In the presence of local movement dependency, the embedded 3sg.obj pronoun can optionally take the form (75b). Interestingly, Collins (1993) notes that this effect is absent in topic constructions (75c), which leads him to the conclusion that they involve base-generation.

  1. (75)

In fact, we find supporting evidence for the absence of movement in -constructions from the lack of reconstruction effects found with other resumptive A̅-dependencies.Footnote 10 For example, we do not find reconstruction for idiomatic interpretation for some speakers (see Sect. 3.3.4). Consider the verb phrase gya nán (‘leave leg’) that has the idiomatic reading ‘to defecate’ (76).

  1. (76)

We observe that, under displacement, the idiomatic reading is still available in na-focus constructions (77a), but disappears with déέ (77b). This follows if (77b) does not involve movement.

  1. (77)

We also find the same pattern with a different (subject) idiom ne-hó dáné, which can have both the literal meaning of ‘turn one’s self (body/skin)’, but also the idiomatic meaning of ‘become pregnant.’ Importantly, A̅-extraction of the idiom chunk in (78a) preserves both readings, whereas the idiomatic reading is lost in the base-generated déέ-construction (78b).Footnote 11

  1. (78)

Another asymmetry between na- and déέ- constructions pertains to another putative reflex of movement noted by Korsah (2016a, 2017). As we have seen, subject extraction triggers obligatory resumption on the verb. There are two types of 3sg.obj subject pronouns in Asante Twi; ɔ- for animate referents (79a), and Ɛ- for inanimate referents (79b) (see Korsah 2017:106).Footnote 12

  1. (79)

While resumptive pronouns generally agree with their antecedents in φ-features, Korsah (2016a, 2017) shows that Ā-movement of an animate subject allows for the unagreeing resumptive Ɛ- in both focus constructions (80a) and relative clauses (80b). This can be understood in terms of (optional) anti-agreement, which is an established reflex of subject extraction (see Ouhalla 1993; Campos 1997; Schneider-Zioga 2007; Henderson 2013; Baier 2017, 2018). If this is correct, then it is particularly revealing that anti-agreement is not an option in the déέ-construction (80c).

  1. (80)

Again, this follows naturally if no movement is involved in the derivation of the déέ-construction.


A reviewer points out the following data, which they suggest may counterexemplify the generalization that verbs crossed by (A̅-)movement lead to tonal overwriting. In (81b), it looks like the subject has raised into the matrix clause leaving behind a resumptive. However, we do not see high-tone raising on the verb .

  1. (81)

We suggest that (81b) does not actually involve movement and should instead be analyzed as resumptive prolepsis, where Ám! is base-generated as a ‘proleptic’ object of the matrix verb and binds a pronoun in the embedded subject position (see e.g. Salzmann 2017a,b):

  1. (82)

This would then be analogous to cases of prolepsis in English where the proleptic object is introduced inside a PP:

  1. (83)
    1. a.

      They say [PP of Maryi ] that shei is smart.

    2. b.

      We know [PP of Johni and Maryj ] that theyi+j visited Joseph. (Branigan and MacKenzie 2002:390)

If the prolepsis analysis is correct, then this explains why there is no tonal overwriting in the embedded clause, since there is no movement. This analysis also accounts for another fact pointed out by the same reviewer, namely that this particular construction does not allow the unagreeing subject resumptive Ɛ- (84).

  1. (84)

Recall from (80) that the option of unagreeing Ɛ- was interpreted as an instance of anti-agreement licensed by A̅-movement. As such, both the lack of tonal overwriting and anti-agreement in this construction converge on the conclusion that this construction does not involve movement.

Finally, what we are analyzing as prolepsis is subject to arbitrary lexical restrictions. While some predicates allow for a proleptic object, others such as ka (‘say’) and dwene (‘think’) do not:

  1. (85)

Since these are bridge predicates, it is unclear why (85) should be ungrammatical if movement were involved. The prolepsis account is therefore better equipped to handle this idiosyncrasy.

Reconstruction and resumption

The evidence for reconstruction effects presented in this section strongly points to the conclusion that resumptives in Asante Twi behave like traces of movement (Zaenen et al. 1981; Koopman and Sportiche 1982, 1986; Alexopoulou 2006). Given the positive results of various reconstruction diagnostics, we can therefore conclude that, despite the general lack of island-sensitivity, resumption in Asante Twi is derived by movement. Nevertheless, there are potential alternative analyses that have been proposed, which could also account for this pattern. These are addressed in the following sections.

Reconstruction into islands

Aoun et al. (2001) propose that there are actually two strategies for resumption in Lebanese Arabic. Movement-generated resumptives are the default option and give rise to reconstruction effects. Inside islands, however, no reconstruction effects are found. This leads Aoun et al. (2001) to propose that a base-generation strategy is used only when the resumptive is situated inside a strong island. For this reason, it is important to establish that the effects described above also pertain in island configurations. In Asante Twi, this is the case. As (86) shows for sentential subject and adjunct islands respectively, resumption inside the island still co-occurs with tonal overwriting on the verb.

  1. (86)

    Reflexes of succesive-cyclicity with extraction from island:

Furthermore, evidence for movement is still found with island-internal resumption. This is illustrated for Complex NP islands in (87), where we observe both reconstruction for variable binding (87a) and WCO effects (87b).

  1. (87)

We take this to be evidence against assuming a distinction between ‘true’ and ‘apparent’ resumption (in the sense of Aoun et al. 2001) for Asante Twi.

Reconstruction without movement?

Nevertheless, Guilliot and Malkawi (2006, 2011) argue that, in Jordanian Arabic, certain reconstruction effects with island-internal resumptives can be derived by base-generating two identical phrases and applying NP ellipsis to derive the resumptive pronoun in the lower DP (either NP ellipsis; Postal 1966; Elbourne 2001).

  1. (88)

While it is unclear whether such an approach can actually account for all reconstruction effects without movement (see Salzmann 2017b:223ff. for discussion), we can rule out such an analysis for Asante Twi resumption on independent, language-internal grounds. Recall that resumptives and definite determiners in Asante Twi are homophonous (). Initially, this may seem to legitimize an analysis such as the one in (88), as Arkoh and Matthewson (2013:27) suggest. Despite the appeal of such a unification, there are good reasons to believe that resumptive pronouns in Asante Twi are not derived by NP ellipsis. The evidence for this comes from a haplology effect reported by Saah (1994). Saah shows that a sequence of homophonous elements is tolerated when one is a resumptive pronoun and the other is the clausal determiner, as in (89a). However, if the two homophonous items both correspond to determiners, then one of them must be deleted (89b). This anti-haplology rule is therefore sensitive to syntactic category and projection type, i.e. it rules out adjacent, identical D0 elements (also see Kramer 2010:231ff.).

  1. (89)

This is problematic for the view that resumptive pronouns are derived by NP ellipsis, since the ‘resumptive pronoun’ in (89a) would actually correspond to the determiner of the elided NP as in (90). Under this analysis, we would expect it to be subject to the same haplology effect as in (89b), contrary to fact.

  1. (90)

Instead, behaves like a genuine pronoun in such cases, which is presumably a phrasal projection DP and therefore immune from the haplological dissimilation rule. In fact, Sect. 3.1 also showed that inanimate resumptive pronouns share the same distribution as anaphoric pronouns, namely that they are obligatorily null in all but a few clearly-defined contexts. This last observation in particular suggests that we are dealing with genuine pronouns, which is somewhat puzzling since they show properties of movement gaps.

Pronoun Conversion

A tentative proposal at this point would be to suggest that resumptive pronouns are indeed the Spell-Out of a lower copy in a movement chain (see e.g. Kandybowicz 2008; Baier 2014; van Urk 2018). Rather than deriving resumptives by somehow reducing the lower copy (e.g. Kandybowicz 2008:135; van Urk 2018), we suggest that there is a PF process of Pronoun Conversion in which a copy is transformed into a pronoun.Footnote 13 As schematized in (91), the basic idea is that the lowest copy of a movement chain is replaced by a corresponding pronoun.

  1. (91)

This process can circumvent island violations as follows. Let us assume that there is a distinction between derivational constraints, which constraint operations within the syntactic derivation, and representational constraints, which specify an illicit linguistic representation. We adopt the widely-held view that syntactic islands are representational constraints at PF (see Merchant 2001; Lasnik 2001; Hornstein et al. 2007; Boeckx 2012; Griffiths and Lipták 2014). Thus, violations of such constraints can be circumvented by altering the offending representation. For traditional Subjacency islands, this can be achieved either by means of ellipsis (e.g. Ross 1969; Merchant 2001) or by resumption (e.g. Sells 1984; Shlonsky 1992; Pesetsky 1998). This is also what we find for other representational constraints such as COMP-trace effects repaired by ellipsis (Merchant 2001:185) and resumption (Kandybowicz 2006; Sato and Dobashi 2016), but crucially not with derivational constraints such as P-stranding. Of course, there remains the question of how to deal with those languages with island-sensitive resumption that still show reconstruction effects. Here, it is possible that another mechanism is responsible for the generation of resumptive pronouns (e.g. copy-deletion; van Urk 2018), which does not sufficiently alter the island-violating representation. As a reviewer points out, Bošković (2002:377f.) argues that lower copy Spell-Out in Romanian is also sufficient to obviate an island violation. Alternatively, this could be an effect of ordering on the PF branch. If the point at which island constraints are checked at PF precedes Pronoun Conversion, then island violations will still be incurred despite resumption.Footnote 14 Under this view, resumption in Asante Twi differs from intrusive resumption in English in being a case of island-obviating grammatical resumption, rather than an extra-grammatical repair strategy (e.g. Heestand et al. 2011; Beltrama and Xiang 2016; Morgan and Wagers 2018; however, see Kroch 1981; Ackerman et al. 2018).

Whatever its broader implications may be, this view seems to make the right cut when it comes to resumption in Asante Twi. There is a movement derivation in the syntax, which explains why we find reflexes of movement inside islands, as well as reconstruction effects at LF, and island obviation comes relatively late in the grammar at the PF interface. A direct empirical advantage is that we can explain why resumptive pronouns derived by movement share the same overt distribution as regular pronouns, as shown in Sect. 3.1. Furthermore, the discussion in Sect. 3.2 showed that it seems correct to link the obviation of island violations to resumption since island effects resurface for extraction of categories that lack proforms independently.

Morpho-phonological aspects of tonal overwriting

The scope of the process

While the previous sections dealt with the syntactic aspects of movement and tonal overwriting in Asante Twi, we now turn to the morpho-phonological side of the phenomenon. So far, we have said that low tones on the verb are replaced with high tones in the presence of an A̅-dependency. While all tones on the verbal root are affected, it is not always the case that affixes undergo the alternation. For example, (92) shows that the low-toned progressive prefix re- surfaces as high in extraction contexts, whereas the low-toned resumptive marker does not.

  1. (92)

While progressive aspectual prefixes are affected, the past suffix -a in (93b) remains low, despite the verb root alternating to high (Marfo 2005b:109, fn.31; Genzel 2013:208).

  1. (93)

Previous literature only mentions these facts in passing and the full scope of the tonal overwriting process has not yet been systematically investigated. As shown by Paster (2010), Asante Twi has a relatively rich inventory of verbal inflection (also see Boadi 1965:41ff.; Dolphyne 1988:87ff.; Ofori 2006:7ff.; Boadi 2008:13; Stump 2016:136). Inflection for tense, aspect and negation is primarily expressed by the affixes in (94).

  1. (94)

    Verbal affixes in Asante Twi (cf. Paster 2010:107):

    (stative) bisá ‘asks’
    -V (past) bisá-a ‘asked’
    bƐ- (future) bέ-bísá ‘will ask’
    re- (progressive) re-bisá ‘is asking’
    a- (perfective) á-bísá ‘has asked’
    kɔ- (egressive) kɔ-bísá ‘go and ask’
    be- (ingressive) be-bísá ‘come and ask’
    N- (negation) ḿ-bísá ‘not ask’

The examples in (92) and (93) showed that the low-toned progressive prefix re- is subject to high-tone raising, whereas the low-toned past suffix is not. This poses the question of how the other affixes behave. Testing this is not entirely straightforward, since some affixes (such as future bέ-) are invariably high and thus do not allow us to check whether they have been affected by tonal overwriting. However, Paster (2010) shows that the tone of other prefixes is conditioned by the preceding subject. For example, the perfective marker á- is typically reported as bearing a high tone (95a), however Paster (2010) shows that if the preceding subject ends in a low tone (e.g. Yaw), then the prefix surfaces as low (95b).

  1. (95)

We will leave aside the question of whether this involves tonal spreading or phonologically-conditioned allomorphy. What is clear is that, in contexts such as (95b), we are now able to check whether we find the low/high-alternation in A̅-contexts. In the extraction equivalent of (95b), the low-toned a- prefix is affected by tonal overwriting (96).

  1. (96)

There are also the low-toned motion prefixes and be- expressing egressive and ingressive aspect, respectively (Osam 2008; Kusmer 2011; Paster 2010).Footnote 15

Such prefixes are also affected by the tonal overwriting process (97b).

  1. (97)

Furthermore, negation, expressed as a homorganic nasal N-, behaves like the aspectual prefix a- in that its tone is conditioned by the preceding subject. When preceded by a low-toned subject such as Saka, it surfaces as low (98a). As (98b) shows, it also changes to high in A̅-contexts.

  1. (98)

So far, we have seen that aspectual and negation prefixes are affected by tonal overwriting, and the past tense suffix is not. While this may invite the conclusion that we are dealing with a prefix/suffix-distinction, we will argue that this is not correct. First, there is another suffix -yƐ that appears in combination with the past tense. While the exact nature of this marker will be discussed further in Sect. 4.3, Kandybowicz (2015) analyzes this as the default realization of aspect in a prosodically-vacuous domain. A prosodically-vacuous VP can be created by moving the object as in (99b). When we do so, we see that the ordinarily low-toned morpheme surfaces with a high tone.

  1. (99)

Although Kandybowicz (2015) does not treat as an affix, there is good reason to believe it is. First, non-verbal material can never intervene between and the verb (e.g. Kandybowicz 2015:261). Second, it undergoes ATR-harmony with the verb, suggesting its integration in the verbal complex. As discussed by Dolphyne (1988:94) and Ofori (2006:42ff.), the suffix -yƐ in Asante Twi surfaces either as invariant [(j)Ɛ] or as [i]/[I], depending on the ATR specification of the preceding vowel (100).

  1. (100)

In terms of autosegmental representation, we can assume apocope of the final , with the glide becoming moraic and harmonizing with the verbal stem in ATR values (101). This supports the integration of -yƐ into the word.Footnote 16

  1. (101)

A similar argument can be made for the 3rd singular subject marker ɔ-. As (102) makes clear, this prefix also participates in ATR-harmony (Saah 1994:54, fn.7; Osam 1994:150ff.).

  1. (102)

As examples such as (97b) and (99b) showed, despite being part of the verbal complex, the subject marker/resumptive is not affected by high-tone insertion.

So far, we have seen that aspectual affixes such as progressive, perfect, motion and -yƐ, as well as negation, are all affected by tonal overwriting, whereas the past-tense and the subject markers are not. We therefore propose the following preliminary generalization:

  1. (103)

    Affix generalization (to be revised):

    Tense and agreement affixes are not affected by high tone insertion in A̅-constructions.

It will ultimately be shown that this generalization follows from independently-motivated assumptions about the clause structure of Asante Twi. Before showing this, we will first briefly consider how the tonal overwriting works.

The nature of overwriting

In Sect. 2.2, we proposed that tonal overwriting involves the realization of v[epp] with a floating H tone that subsequently spreads overwriting the tones in the verb. In light of the preceding discussion, the challenge for an analysis of overwriting is how to limit spreading to affixes of a particular kind. Interestingly, tonal overwriting is also found elsewhere in Asante Twi. Paster (2010) shows that a floating L tone overwrites any H tones on the verb in the imperative:

  1. (104)

    L-tone overwriting in the imperative (Paster 2010:115):

      Habitual   Imperative  
    H tɔ́ ‘buy’ tɔ pέn ‘Buy a pen!’
    HL nóm ‘drink’ nom insyu ‘Drink water!’
    LH kaé ‘remember’ kae kofí ‘Remember Kofi!’
      bisá ‘ask’ bisa asέ ‘Ask something!’

Even within Asante Twi, there is evidence for overwriting with both H and L tones. We therefore need a general theory of tonal overwriting. We will follow Trommer (2011) in proposing that tonal overwriting involves a floating tonal circumfix (H- -H) (for alternative approaches to tonal overwriting, see McPherson 2014; McPherson and Heath 2016; Kim 2016). Trommer (2011:126) proposes that overwriting is driven by a constraint Contτ (105) requiring that tones belonging to the same morpheme are adjacent (see Landman 2003; Zimmermann 2017 on morpheme contiguity).

  1. (105)


    Tautomorphemic tones form a contiguous melody.

Additionally, we assume the following standard markedness/faithfulness constraints for tonal processes in (106) (see Yip 2002:79ff.). Note that the constraint Float(h) refers to high tones specifically to allow for floating L tones that trigger downstep.

  1. (106)
    1. a.


      No floating high tones.

    2. b.


      Do not delete tones.

    3. c.


      Every tone-bearing unit must be linked to a tone.

In Trommer’s (2011) analysis, the combined effect of a discontinuous morpheme and a high-ranked contiguity constraint leads to inward spreading to achieve adjacency between the parts of the circumfix. This is illustrated in the following OT analysis of overwriting with the verb re-somá in (92). The faithful candidate in (107a) violates contiguity, since the H tones belonging to the circumfix are not adjacent. These tones can delete the underlyingly linked tones, as in (107b), to avoid *Float(h) violations, but contiguity is still violated by the intermediate L.Footnote 17 An alternative candidate with a floating L (realized as a downstep) equally violates contiguity (107c). Removing this intermediate L tone satisfies contiguity, but introduces a fatal Specify violation (107d). Thus, the optimal candidate is (107e), where one of the H tones spreads to derive full overwriting.

  1. (107)

    Tonal overwriting in Asante Twi (rè-sòmá ⇒ ré-sómá):

The attachment of a floating circumfix therefore deterministically results in overwriting due to the importance of contiguity. Thus, morphemes that obligatorily trigger overwriting such as H- -H on v and the imperative marker L- -L (104) can be analyzed in this fashion.

It is important to note that there are a number of advantages to assuming that overwriting is due to a tonal circumfix rather than spreading of a single tone. First, it allows us to make a distinction between floating tones which associate to a single tone and those which intrinsically overwrite in a given domain. The former type can be seen in associative constructions, for example, where the floating high tone docks to the nearest syllable, but does not overwrite (Abakah 2010:62). The second advantage is that the tonal circumfix delimits the domain of spreading, since it spreads inwards. In an alternative approach in which spreading of the floating tone is driven by Align (McCarthy and Prince 1993; Hyde 2012) or Share constraints (McCarthy 2010), something additional must be stipulated to prevent spreading to affixes to the past tense marker, for example, at later cycles. Finally, the constraint in (105) means that overwritten L tones must be deleted to ensure contiguity. A consequence is that we do not find outputs such as (107c) that would lead to downstep. This is different when L tones are left floating by spreading of a single H, which often results in downstep (e.g. Paster 2010; Abakah 2010).

Consequently, we can refine the Vocabulary Items for the v proposed in (36) to involve a floating tonal circumfix (108).

  1. (108)

    Vocabulary Items forv(final):

    a. [v, epp] H- -H
    b. [v] Ø

The following section will now demonstrate how this view of tonal overwriting can explain its selective nature with regard to the affixes affected.

Deriving the affix generalization

Recall the preliminary version of the affix generalization from (103), repeated below.

  1. (109)

    Affix generalization:

    Tense and agreement affixes are not affected by high tone insertion in A̅-constructions.

This section will demonstrate that this generalization can be rephrased in structural terms based on the independently-motivated assumption that aspect and negation are actually lower than v (110), as proposed by Kandybowicz (2015:257).

  1. (110)

    Clause structure of Asante Twi:

Kandybowicz (2015) primarily motivates this structure on the basis of the distribution of the morpheme . He argues that is the realization of Asp in a prosodically vacuous Spell-Out domain, i.e. the complement of v. This morpheme is only found in past tense contexts with the suffix -V (111a). In other words, is in complementary distribution with aspect (111b).Footnote 18

  1. (111)

This distinction is explained by assuming that the verb moves to T, where the -V ending is hosted, unless ‘blocked by an overt/contentful head’ such as aspect (Kandybowicz 2015:249).Footnote 19 With an intransitive verb (111a), V-to-T movement will render the Spell-Out domain of v prosodically vacuous, leading to insertion of . The fact that aspectual affixes obviate -insertion (111b) suggests that they must remain in the Spell-Out domain of v, i.e occupying a position lower than v, as in (110). The assumption that -insertion has to do with prosodic vacuity is further supported by the fact that even when the verb moves, is blocked if there is a direct object of a transitive verb (112a) or a low VP-adverb (112b).

  1. (112)

We can also see that negation is lower than aspect given the affix order in (113).

  1. (113)

Given the structure in (110), Kandybowicz (2015) argues that na occupies T and the verb does not move out of the vP domain. Since we see that the aspectual prefix re- must precede negation (113), mirror principle reasoning would therefore dictate that negation is located lower than aspect as in (110).Footnote 20

Given this clause structure, then the generalization in (109) translates into a structural one—affixes that originate in a position lower than v (e.g. aspect and negation) are affected by tonal overwriting, whereas those in a higher position (e.g. tense, subject agreement/resumption) are not. This is summarized in (114).

  1. (114)

    Affix generalization (final):

    Only affixes lower than v are affected by high tone insertion in A̅-constructions.

The generalization in (114) will now follow both from the previously established analysis of tonal overwriting as a floating tonal circumfix and standard assumptions about word formation in Distributed Morphology (DM). In standard DM approaches (e.g. Halle and Marantz 1993; Harley and Noyer 2003; Embick and Noyer 2007; Embick 2015), Vocabulary Insertion and concatenation of affixes applies cyclically from the root outwards inside a complex head (Bobaljik 2000; Embick 2010). Consider (115) where both the tones of the root and the progressive prefix re- have been affected by tonal overwriting (re-boá ⇒ ré-bóá).

  1. (115)

Assuming that aspect blocks verb movement to T (Kandybowicz 2015), then the complex head in (116) is formed either by head movement or post-syntactic Lowering (Embick and Noyer 2001) (or a combination thereof). Since AspP is structurally lower than v, the low-toned prefix re- combines with the root first (117)[a]. Subsequently, v combines with the Asp constituent. Since it is realized as a floating tonal circumfix, this now attaches outside both the aspectual prefix and the root (117)[b]. This then triggers overwriting of both the prefix and the root (117)[c], as shown in Sect. 4.2.

  1. (116)
  1. (117)
    1. a.

      Asp\(+\sqrt{\text{bo{\'{a}}}}\) ⇒ rè-bòá

    2. b.

      v+[rè-bòà] ⇒ H-rè-bòá-H

    3. c.

      H-rè-bòá-H ⇒ ré-bóá

Given the Mirror Principle (e.g. Baker 1985; Harley 2011), prefixes located lower than vP in the clause (such as aspect and negation) will always combine with the root before v and therefore always intervene between the floating tonal circumfix. Affixes originating higher than vP, on the other hand, will be attached after v and therefore be immune from the effects of overwriting. To see this, let us consider the more complex example from (97), repeated as (118). Recall that this example involves both aspect and tense morphemes, and we observe that, while the root and egressive prefix kɔ- become high, the past tense -a and subject prefix ɔ- do not.

  1. (118)

In this example, the verb moves to T and therefore receives overt past tense inflection expressed by the -V suffix (following Kandybowicz 2015). Subsequently, the complex head T in (119) formed by head movement is subject to Vocabulary Insertion, as above.

  1. (119)

First, the low-toned aspectual affix kɔ- is attached to the root. At the next cycle, the exponent of v (the floating H-tone circumfix) attaches to this complex and triggers overwriting of all low tones. Subsequently, the subject resumptive marker is attached. We assume this to be a realization of an Agr projection, since subject resumptives differ from object resumptives in being strictly bound morphemes that also show optional anti-agreement in certain contexts (see Sect. 3.4). Since ɔ- attaches after v, it is not subject to tonal overwriting and remains low-toned. The same holds for the past tense exponent in T, which is realized as a floating moraic suffix that triggers lengthening of the final segment of the verb (e.g. Ofori 2006:29; Zimmermann 2017:188). Since this affix is also structurally higher than v, it is immune from the effects of overwriting.

This section has shown that the selective nature of the tonal overwriting process follows from the cyclic nature of Vocabulary Insertion, the assumption of a floating H-tone circumfix and the independently-motivated assumption that both aspect and negation affixes orginate lower than v. The fact that both Kandybowicz’s (2015) analysis of -insertion and the affix generalization in (109) require the same assumptions about the clause structure of Asante Twi constitutes a striking convergence across empirical domains and thereby lends further support to the assumption of a low AspP and NegP projection in the language.

Further issues

Subject extraction

One issue we have not yet addressed regards tonal overwriting with subject extraction. Recall from Sect. 2.2 that we are proposing that the floating high-tone exponent that triggers overwriting is the realization of v bearing an edge feature ([epp]) inserted to facilitate successive-cyclic movement. With local extraction of a wh-object, this feature is inserted to trigger movement to the edge of the vP phase (120).

  1. (120)

However, wh-subjects already originate at the edge of vP and can therefore move to their local Spec-CP position without any intermediate step (121).

  1. (121)

All else being equal, we would not expect to find movement reflexes with local subject extraction since there is no trigger for edge-feature insertion in (121). This has been reported to be the case in Duala (Epée 1976), Indonesian (Saddy 1991) and Defaka (Bennett et al. 2012). However, like many other languages (Clements et al. 1983), local subject extraction in Asante Twi does show reflexes of movement, i.e. tonal overwriting on the verb (122b).

  1. (122)

In order to reconcile this with the analysis proposed so far, we could either propose that wh-subjects must undergo string-vacuous intermediate movement for independent reasons (e.g. Müller 2007:86) or we can dissociate the phase edge and the base position of the subject.Footnote 21

We will adopt the latter view in line with others (e.g. Richards 2010; Baltin 2012; Harley 2013) and propose the existence of vP shells as in (123). While the outer vP is often given the name Voice, we label it vcP following Richards (2010:14) and also adopt his proposal that it is this head, and not the lower one, that is the phase head.

  1. (123)

With this structure, subject extraction now requires insertion of an edge feature on vc (124), which is realized as the familiar overwriting circumfix.

  1. (124)

The postulation of vP shells in Asante Twi is not only motivated by a theory-internal quandary regarding the base position of the subjects, there is also a potential empirical advantage. Hein (2017) discusses the fact that topicalization of a VP in Asante Twi leads to a process of -support, analogous to do-support in English (125a). While it may be tempting to treat this as a variant of the morpheme discussed by Kandybowicz (2015), Hein (2017) points out that this is not possible since the two morphemes co-occur (125b).Footnote 22

  1. (125)

For (125b), we can place in v and in Asp (following Kandybowicz 2015) (126).

  1. (126)

The general idea would be that v is realized as zero in the context of the root (i.e. inside a complex head). If v is not local to the root, for example because the VP has been fronted, then it receives a default realization as . This is then very similar to Thoms’ (2011) and Baltin’s (2012:417) approach to do-support/British-do in English. Locating in v is also supported by the fact that it is subject to tonal overwriting (127b), in line with the affix generalization (109) stating that material structurally lower than the floating circumfix (now in vc) is affected by overwriting.

  1. (127)

Covert movement and wh-in-situ

Another issue pertains to wh-in-situ: wh-words can stay in-situ in Asante Twi and do not trigger tonal overwriting when they do (128a).

  1. (128)

This is a cross-linguistically familiar picture where in-situ wh-phrases do not trigger putative reflexes of A̅-movement (but cf. Reintges et al. 2006). A prevalent view of wide-scope wh-in-situ constructions would be that they involve wh-movement in the syntax, but pronunciation of the lowest copy at PF (e.g. Bobaljik 2002). Under this view, we would expect to find the same reflexes that we do with overt movement.

While not being able to resolve this issue completely, we will point to two pieces of evidence which suggest that wh-in-situ and wh-ex-situ in Asante Twi do not share the same syntactic derivation. The first argument comes from the observation by Kobele and Torrence (2006) that in-situ wh-phrases in non-echo questions are sensitive to intervention effects by negation, for example (129) (Pesetsky 2000; Beck 2006). Ex-situ wh-phrases, on the other hand, are not subject to intervention effects (129c).

  1. (129)

The second asymmetry pertains to wh-in-situ in embedded contexts. As originally noticed by Kobele and Torrence (2006) (and corroborated by Kandybowicz 2017:117), an in-situ wh-phrase cannot take wide scope out of an embedded clause (130a). By comparison, overt movement out of an embedded CP is unproblematic (130b).

  1. (130)

Both of these asymmetries serve to show that the syntax of wh-in-situ cannot simply involve the same syntactic derivation as wh-ex-situ, with differences in the pronunciation site. Instead, we can find alternative ways to compute wide-scope of wh-phrases without movement (e.g. Pesetsky 1987; Reinhart 1998; Cole and Hermon 1998) or a different kind of movement, genuine LF-movement (Huang 1982; Nissenbaum 2000; Richards 2001) or feature movement (Bošković 1998; Pesetsky 2000; Soh 2005).Footnote 23 Under these kinds of approaches to wh-in-situ, we would not necessarily expect to find the same reflexes as overt phrasal movement.


This paper has argued that, contrary to previous claims, high-tone overwriting on verbs in Asante Twi is not a property of only focus constructions. Instead, we have shown that it has a much wider distribution across other A̅-constructions, which leads to the conclusion that it is best analyzed as a reflex of successive-cyclic operator movement through vP. The most compelling evidence for this view comes from the observation that tonal overwriting affects all verbs crossed by an A̅-dependency, which is a hallmark of such movement reflexes (see Georgi 2014). We also addressed a challenge for the view that movement is involved in A̅-dependencies in Asante Twi, namely the presence of obligatory, island-insensitive resumption. Nevertheless, we showed that obviation of islands is linked directly to resumption, since island effects re-emerge with extraction of categories that lack resumptives independently (i.e. PPs and VPs). Furthermore, resumptive pronouns show an array of reconstruction effects which would be unusual for genuine pronouns. What is more, it was shown that déέ topic constructions lack both movement reflexes and reconstruction effects, and are therefore genuine base-generated structures. All of this constitutes evidence against the view that resumptive constructions involve base-generation (Saah 1994).

This paper also provided the first detailed investigation of the morpho-phonological aspects of overwriting. In particular, we showed that there is an asymmetry regarding the affixes which undergo the alternation. While the verb root and aspect and negation affixes are subject to overwriting, tense and agreement affixes are not affected. We showed that this generalization corresponds to the position of the affix in the structure of the clause. Kandybowicz (2015) proposed, on independent grounds, that both AspP and NegP are lower than vP in Asante Twi. Furthermore, there are the affixes which undergo the alternation together with the verb. Given that the structure of complex heads is assumed to mirror their hierarchical order in the clause, any affix that is below v will be in the scope of the overwriting tonal circumfix and therefore affected. Affixes generated higher attach later and are immune from its effects.

Aside from adding to the evidence for successive-cyclic movement through vP, the preceding discussion also shows that even formal features such as the EPP can be realized as floating autosegmental material (cf. Akinlabi 1996). Finally, we also see the importance of the syntax/phonology interface in informing syntactic theory. Boeckx (2008:23) points out that ‘Kikuyu is the only language I am aware of that offers phonological reflexes of successive-cyclic movement. But this may be due to the fact that the relation between tone and syntactic movement hasn’t been studied as much as it ought to be.’ We concur with Boeckx and hope to have demonstrated on the basis of Asante Twi that tone can potentially tell us a lot about movement if we start looking in the right places.


  1. 1.

    All non-cited data come from the first co-author of the paper, who is a native speaker of Akan and has native intuitions about the Asante Twi and Fante dialects. Crucial judgments have been cross-checked with other native speakers of the Asante Twi dialect. High tones are marked with an acute accent (e.g. á) and low tones are generally unmarked (but sometimes with a grave accent à). Downstepped high tones are marked with a superscript exclamation mark (!). Note that we focus on Asante Twi in particular here since other dialects of Akan (such as Fante) exhibit tonal polarity where the corresponding low-toned habitual verb stems in Asante Twi actually surface with high tones even in discourse-neutral contexts (see Abakah 2005:123ff.). While Akan has a rich array of segmental phonological processes such as assimilation (Schachter 1969) and ATR harmony (Casali 2012; Kügler 2015), we follow standard Akan orthography and do not represent such processes graphemically, unless immediately relevant.

  2. 2.

    Note that this process is not limited to extraction of arguments. Tonal overwriting is also found with extraction of adjuncts such as dabέn (‘when’) (11b) and Ɛhéńfá (‘where’) (11d).

    1. (i)
  3. 3.

    It is worth noting that there are other kinds of tonal overwriting in Asante Twi. For example, Dolphyne (1988) discusses high-tone spreading in associative constructions, and Paster (2010) discusses low tone overwriting in the imperative (see Sect. 4.1). Certain kinds of prefixes also trigger tonal alternations on verb roots (Schachter and Fromkin 1968; Paster 2010) (see Sect. 4.2). As far as we can tell, these processes are fully independent of the particular kind of high-tone overwriting described here.

  4. 4.

    It was originally suggested that only transitive vP (i.e. v*P) was a phase. However, subsequent work argued that unaccusative and passive vP should also constitute phases (see e.g. Legate 2003; Sauerland 2003).

  5. 5.

    It is important to note however that this is only straightforward for transparent reflexes which occur uniformly along the extraction path (what Georgi 2014 calls Pattern I reflexes). As Georgi (2014, 2017) discusses, there are also patterns which mark exclusively final or non-final movement steps, which can motivate treating the reflex as a realization of Spec-Head wh-agreement.

  6. 6.

    There is an additional complementizer aN (triggering nasalization of the following segment) that occurs with resumptive dependencies (34).

    1. (i)

    McCloskey argues that this results from a base-generated pro in Spec-CP that binds the resumptive pronoun í directly. Thus, no movement is involved in an example such as (33). Thus, we would have to refine the Vocabulary Items in (34) to involve an OP(erator) feature, as McCloskey (2002) suggests. Such cases are not relevant for the Asante Twi data. Although Asante Twi has overt resumptives with animates, they behave exactly like gaps do, and also show movement effects (see Sect. 3).

  7. 7.

    Titov (2019:29f.) follows Boadi (1974) and Ofori (2011) in assuming that na-focus constructions are actually bi-clausal structures, where the focus particle na is the result of phonological fusion of the copula ne and the relative marker áa. In particular, Titov (2019) treats this as an underlying inverse pseudocleft with fusion of ne and áa (35).

    1. (i)

    This is not a plausible synchronic analysis for the na-construction, however, since free/headless relative clauses in Asante Twi do not take the áa relative marker, but rather the topic marker déέ (35) (Boadi 2005:155ff.).

    1. (ii)

    Furthermore, despite some similarities between focus constructions and relative clauses (e.g. regarding tone), there are also surface differences such as the distribution of the clausal determiner (cd). While the CD is obligatory with relative clauses (Saah 2010:100), it is typically optional in na-focus constructions. This difference would be unexpected if they shared the same underlying source.

  8. 8.

    It is important to note that this is not simply an effect of ‘clause-finality,’ since structurally higher clause-final material, such as the question particle anaa, does not license an inanimate pronoun (49b) (also see Korsah 2017:27f.).

    1. (i)

    Furthermore, adding a pause before the adverb in examples such as (48a) and (49b) leads to the disappearance of the pronoun. As a reviewer suggests, this may highlight the role of prosodic phrasing, i.e. pro-drop applies when the pronoun is final in its intonation phrase (see Reglero 2007 on with wh-in-situ in Spanish). Furthermore, the special intonation of the yes-no question in (48b) may also have a similar effect (as also noted by the reviewer).

  9. 9.

    Titov (2019) disputes the existence of WCO in Akan. This is not too surprising, since WCO is known to be a somewhat fragile diagnostic, subject to inter-speaker variation (see e.g. Salzmann 2017b:197). Regarding WCO with resumptives in Swedish, Asudeh (2012:245) notes that ‘as is common with weak crossover judgments, there is some speaker uncertainty and variation here. For some speakers, the judgments are quite robust, though.’ (also see Baker and Kramer 2018:1045 on WCO variation in Amharic). As far we can tell, the situation is similar in Asante Twi, with some speakers having robust sensitivity to WCO. Speakers without WCO might be employing a logophoric strategy, or treating the possessor as the resumptive (i.e. Left-Branch Extraction). Examples such as (64b) become relevant in this regard because epithets do not function as resumptives in Asante Twi and are generally assumed to be anti-logophoric (e.g. Dubinsky and Hamilton 1998).

  10. 10.

    A reviewer points out an interesting example in which a -construction is embedded under a focus construction. Here, we still observe tonal overwriting on the verb (76).

    1. (i)

    While we cannot investigate this construction in detail, there are a number of potential analyses that would be consistent with our proposal. One such analysis is that the subject simply moves across a base-generated topic pronoun. The alternative would be to assume that a moving phrase can exceptionally stop at Spec-Top (and trigger resumption) en route to its final landing site (however, we might expect to find a D-linked interpretation). For now, we leave such examples for further investigation in future research.

  11. 11.

    An anonymous reviewer provides the following example in which the idiomatic reading of the idiom ka ne mpaboa so (‘to polish one’s shoe’), i.e. ‘to be drunk,’ is also possible in the déέ-construction (78b).

    1. (i)

    If there are certain idiom chunks that license idiomatic readings even under base-generation, this is not necessarily fatal for our analysis, since idioms are known to vary with regard to their ‘transparency,’ i.e. their ability to participate in syntactic dependencies (see e.g. Fraser 1970; Nunberg et al. 1994). Thus, some idioms, such as the one in (77), could be able to control the (null) pronoun in a base-generation structure, whereas others cannot and would therefore require reconstruction to derive an idiomatic interpretation.

  12. 12.

    There are a couple of, still poorly understood, exceptions to this basic picture. As discussed by Korsah (2017:111), speakers allow for optional Ɛ-marking on the verb with animate subjects in some restricted contexts. These include serial verb constructions and with subjects of intransitive verbs (79a). Importantly, this option requires an overt DP subject and is not possible under pro drop (79b). We leave further investigation of these exceptions to future research.

    1. (i)
  13. 13.

    This can be thought of as analogous to the LF process of Trace Conversion (Fox 1999, 2002). It is also conceivable that Pronoun Conversion is the PF equivalent of vehicle change (Fiengo and May 1994), where an R-expression can be construed as a coreferent pronoun at LF. Importantly, vehicle change has also been claimed to apply to traces in movement chains (Giannakidou and Merchant 1998; Safir 1999; Hunter and Yoshida 2016). It could also be viewed as the same process as with island-sensitive resumptives, e.g. deletion (van Urk 2018) or m-Merger (Harizanov 2014), but with variation in the timing of when island constraints are evaluated. Crucially, Pronoun Conversion must apply before, and therefore feed satisfaction of, island constraints. In island-sensitive languages, the Pronoun Conversion process applies too late (counter-feeding).

  14. 14.

    It seems that extrinsic ordering of Pronoun Conversion must be assumed in Asante Twi with regard to pro drop. Since island violations are still obviated by pro-dropped resumptives, we must assume that PF-islandhood is checked after Pronoun Conversion, but before pro drop applies (thanks to Ivy Sichel for pointing this out).

  15. 15.

    Although these affixes are cognates of the verbs for ‘come’ and ‘go,’ it is not possible to analyze them (synchronically) as serial verb constructions. In serial verb constructions, both verbs typically inflect for tense/aspect/agreement (see Hellan et al. 2003). In some aspects, the second verb in an SVC bears ‘infinitive’ or ‘consecutive’ marking as with in (97), which makes it distinct from its grammaticalized use as an aspectual prefix.

    1. (i)
  16. 16.

    Note that, in (101), although past tense is a floating mora (Ofori 2006; Paster 2010), Akan does not allow heavy syllables (*CVC, *CVV). Long vowels and final nasals are associated with separate syllables (Dolphyne 1988:52ff.). We assume that the additional syllable node is inserted as a repair to a constraint militating against bi-moraic syllables. Also, Kügler (2015) shows that there are some limited cases of ATR-harmony across words in Akan, however they only include regressive harmony between a sequence of [−ATR]-[+ATR] and only within a phonological phrase (φ). We therefore do not believe that such exceptions fundamentally undermine the applicability of this diagnostic.

  17. 17.

    We assume that the H tones belonging to the circumfix cannot be deleted due to a high-ranked constraint on morpheme realization (e.g. RealizeMorpheme; Kurisu 2001).

  18. 18.

    We diverge from Kandybowicz (2015) by continuing to represent the morpheme as , following standard Akan orthography.

  19. 19.

    This presupposes that the Head Movement Constraint cannot be absolute, but rather relativized for certain heads, see Baker and Collins (2006:313) for a similar assumption. Kandybowicz (2015) provides further evidence for V-to-T movement from the periphrastic expression of past tense with na, which will not be discussed here.

  20. 20.

    Furthermore, motion aspect seems to occur even lower than negation (and in conjunction with other aspects) (114) (also see Boadi 2008:14), suggesting it occupies another aspect head even lower than negation (e.g. Mot0; Kusmer 2011:18).

    1. (i)

    This aspectual marker does not block movement to T, since it is compatible with the -V past marker (97).

  21. 21.

    An anonymous reviewer points out that this problem might speak in favour of a different mechanism of successive-cyclic movement. For example, Bošković (2014, 2016) assumes that the entire phase (including the edge) is sent to Spell-Out (also see Harwood 2015). Bošković (2016) argues that in such an approach, successive-cyclic movement targets the phrase immediately dominating the phase. This would mean that subjects would also have to move from their base-position inside the phase. A potential drawback of this kind of non-feature-driven approach to successive-cyclic movement comes in accounting for the morphological reflexes at hand. Although Bošković (2008) suggests that some of these data could be reanalyzed as intervention effects, feature-driven approaches offer the most straightforward way of accounting for them, as they can tie allomorphic alternations to the presence of a movement-related feature. Furthermore, Georgi (2014, 2017) shows that they can also handle ‘opaque’ patterns, which seem more problematic for featureless, ‘Greed’-based analyses.

  22. 22.

    Also, (125a) shows that this morpheme occurs with progressive aspect re-, unlike (111b).

  23. 23.

    Indeed, a reviewer points out that the restrictions on wh-in-situ in Asante Twi seems to mirror those described for French (see Bošković 1998 and Cheng and Rooryck 2000 for arguments for a feature movement approach).


  1. Abakah, Emmanuel N. 2005. Tone rules in Akan. Journal of West African Languages 32 (1–2): 109–134.

    Google Scholar 

  2. Abakah, Emmanuel N. 2010. Tone and the associative construction in Akan. Journal of West African Languages 37 (2): 57–79.

    Google Scholar 

  3. Ackerman, Lauren, Michael Frazier, and Masaya Yoshida. 2018. Resumptive pronouns can ameliorate illicit island extractions. Linguistic Inquiry 49 (4): 847–859.

    Google Scholar 

  4. Adger, David. 2007. Stress and phasal syntax. Linguistic Analysis 33 (3–4): 238–266.

    Google Scholar 

  5. Adger, David, and Gillian Ramchand. 2005. Merge and move: Wh-dependencies revisited. Linguistic Inquiry 36 (2): 161–193.

    Google Scholar 

  6. Agüero-Bautista, Calixto. 2001. Cyclicity and the scope of wh-phrases. PhD diss., MIT.

  7. Akinlabi, Akinbiyi. 1996. Featural affixation. Journal of Linguistics 32 (2): 239–289.

    Google Scholar 

  8. Alexiadou, Artemis, Paul Law, André Meinunger, and Chris Wilder. 2000. Introduction. In The syntax of relative clauses, eds. Artemis Alexiadou, Paul Law, André Meinunger, and Chris Wilder, 1–51. Amsterdam: John Benjamins.

    Google Scholar 

  9. Alexopoulou, Theodora. 2006. Resumption in relative clauses. Natural Language and Linguistic Theory 24 (1): 57–111.

    Google Scholar 

  10. Amaechi, Mary, and Doreen Georgi. 2019. Quirks of subject (non-)extraction in Igbo. Glossa 4 (1): 69–136.

    Google Scholar 

  11. Ameka, Felix K. 1992. Focus constructions in Ewe and Akan: A comparative perspective. In Kwa comparative syntax workshop, eds. Chris Collins and Victor Manfredi, 1–25. Cambridge: MITWPIL.

    Google Scholar 

  12. Ameka, Felix K. 2010. Information packaging constructions in Kwa: Microvariation and typology. In Topics in Kwa syntax, eds. Enoch O. Aboh and James Essegbey, 141–176. Dordrecht: Springer.

    Google Scholar 

  13. Amfo, Nana Aba Appiah. 2010. Lexical signaling of information structure in Akan. Linguistics 48 (1): 195–225.

    Google Scholar 

  14. Aoun, Joseph, and Elabbas Benmamoun. 1998. Minimality, reconstruction and PF movement. Linguistic Inquiry 29 (4): 569–597.

    Google Scholar 

  15. Aoun, Joseph, Lina Choueiri, and Norbert Hornstein. 2001. Resumption, movement and derivational economy. Linguistic Inquiry 32 (3): 371–403.

    Google Scholar 

  16. Arkoh, Ruby, and Lisa Matthewson. 2013. A familiar definite article in Akan. Lingua 123: 1–30.

    Google Scholar 

  17. Asudeh, Ash. 2012. The logic of pronominal resumption. Oxford: Oxford University Press.

    Google Scholar 

  18. Baier, Nico. 2014. Spell-out, chains, and long distance Wh-movement in Seereer. Ms. University of California, Berkeley.

  19. Baier, Nico. 2017. Antilocality and antiagreement. Linguistic Inquiry 48 (2): 367–377.

    Google Scholar 

  20. Baier, Nico. 2018. Anti-agreement. PhD diss., University of California, Berkeley.

  21. Baker, Mark, and Ruth Kramer. 2018. Doubled clitics are pronouns: Amharic objects (and beyond). Natural Language and Linguistic Theory 36 (4): 1035–1088.

    Google Scholar 

  22. Baker, Mark C. 1985. The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16 (3): 373–415.

    Google Scholar 

  23. Baker, Mark C., and Chris Collins. 2006. Linkers and the internal structure of vP. Natural Language and Linguistic Theory 24 (2): 307–354.

    Google Scholar 

  24. Baltin, Mark. 2012. Deletion versus pro-forms: An overly simple dichotomy? Natural Language and Linguistic Theory 30 (2): 381–423.

    Google Scholar 

  25. Barss, Andrew. 1986. Chains and anaphoric dependence: On reconstruction and its implications. PhD diss., MIT.

  26. Bayer, Josef, and Martin Salzmann. 2013. That-trace effects and resumption—how improper movement can be repaired. In Repairs: The added value of being wrong, eds. Patrick Brandt and Eric Fuß, 275–333. Berlin: de Gruyter.

    Google Scholar 

  27. Bayer, Josef, Jana Häussler, and Markus Bader. 2016. A new diagnostic for cyclic wh-movement: Discourse particles in German questions. Linguistic Inquiry 47 (4): 591–629.

    Google Scholar 

  28. Beck, Sigrid. 2006. Intervention effects follow from focus interpretation. Natural Language Semantics 14 (1): 1–56.

    Google Scholar 

  29. Beltrama, Andrea, and Ming Xiang. 2016. Unacceptable but comprehensible: The facilitation effect of resumptive pronouns. Glossa 1 (1): 29–124.

    Google Scholar 

  30. Bennett, William G., Akinbiyi Akinlabi, and Bruce Connell. 2012. Two subject asymmetries in Defaka focus constructions. In West Coast Conference on Formal Linguistics (WCCFL) 29, eds. Jaehoon Choi et al., 294–302. Somerville: Cascadilla Proceedings Project.

    Google Scholar 

  31. Bhatt, Rajesh. 2015. Relative clauses and correlatives. In Syntax – Theory and analysis: An international handbook, eds. Artemis Alexiadou and Tibor Kiss, 708–749. Berlin: de Gruyter.

    Google Scholar 

  32. Boadi, Lawrence A. 1965. Some Twi phrase structure rules. Journal of West African Languages 2 (1): 37–46.

    Google Scholar 

  33. Boadi, Lawrence A. 1971. The passive in Akan. Journal of African Languages 10 (3): 34–41.

    Google Scholar 

  34. Boadi, Lawrence A. 1974. Focus marking in Akan. Linguistics 140: 5–57.

    Google Scholar 

  35. Boadi, Lawrence A. 1990. Questions in Akan. Frankfurter Afrikanistische Blätter 2: 70–92.

    Google Scholar 

  36. Boadi, Lawrence A. 2005. Three major syntactic structures in Akan: Interrogatives, complementation and relativisation. Accra: Black Mask.

    Google Scholar 

  37. Boadi, Lawrence A. 2008. Tense, aspect and mood in Akan. In Aspect and modality in Kwa languages, eds. Felix K. Ameka and Mary Esther Kropp Dakubu, 9–68. Philadelphia: John Benjamins.

    Google Scholar 

  38. Bobaljik, Jonathan. 2000. The ins and outs of contextual allomorphy. In University of Maryland Working Papers in Linguistics 10, eds. Kleanthes K. Grohmann and Caro Struijke, 35–71. College Park: University of Maryland.

    Google Scholar 

  39. Bobaljik, Jonathan D. 2002. A-chains at the PF-interface: Copies and “covert” movement. Natural Language and Linguistic Theory 20 (2): 197–267.

    Google Scholar 

  40. Boeckx, Cedric. 2008. Understanding Minimalist syntax: Lessons from locality in long-distance dependencies. Oxford: Blackwell.

    Google Scholar 

  41. Boeckx, Cedric. 2012. Syntactic islands. Cambridge: Cambridge University Press.

    Google Scholar 

  42. Borer, Hagit. 1984. Restrictive relatives in Modern Hebrew. Natural Language and Linguistic Theory 2 (2): 219–260.

    Google Scholar 

  43. Borsley, Robert D., Maggie Tallerman, and David Willis. 2007. The syntax of Welsh. Cambridge: Cambridge University Press.

    Google Scholar 

  44. Bošković, Željko. 1998. LF movement and the Minimalist Program. In Northeast Linguistic Society (NELS) 28, eds. Pius N. Tamanji and Kiyomi Kusumoto, 43–57. Amherst: GLSA.

    Google Scholar 

  45. Bošković, Željko. 2002. On multiple wh-fronting. Linguistic Inquiry 33 (3): 351–384.

    Google Scholar 

  46. Bošković, Željko. 2008. On successive-cyclic movement and the freezing effect of feature checking. In Sounds of silence: Empty elements in syntax and phonology, eds. Jutta Hartmann, Veronika Hegedus, and Henk van Riemsdijk, 195–233. Amsterdam: Elsevier.

    Google Scholar 

  47. Bošković, Željko. 2014. Now I’m a phase, now I’m not a phase: On the variability of phases with extraction and ellipsis. Linguistic Inquiry 45 (1): 27–89.

    Google Scholar 

  48. Bošković, Željko. 2016. What is sent to spell-out is phases, not phase complements. Linguistica 56 (1): 25–66.

    Google Scholar 

  49. Brame, Michael. 1968. A new analysis of the relative clause: Evidence for an interpretive theory. Ms. MIT.

  50. Branigan, Phil, and Marguerite MacKenzie. 2002. Altruism, Ā-movement and object agreement in Innu-aimûn. Linguistic Inquiry 33 (3): 385–407.

    Google Scholar 

  51. Campos, Héctor. 1997. On subject extraction and the antiagreement effect in Romance. Linguistic Inquiry 28 (1): 92–119.

    Google Scholar 

  52. Casali, Roderic F. 2012. [+ATR] dominance in Akan. Journal of West African Languages 39 (1): 33–59.

    Google Scholar 

  53. Cheng, Lisa L. S., and Johan Rooryck. 2000. Licensing wh-in-situ. Syntax 3 (1): 1–19.

    Google Scholar 

  54. Chomsky, Noam. 1973. Conditions on transformations. In A festschrift for Morris Halle, eds. Stephen R. Anderson and Paul Kiparsky, 232–286. New York: Holt, Rinehart and Winston.

    Google Scholar 

  55. Chomsky, Noam. 1977. On wh-movement. In Formal syntax, eds. Peter Culicover, Thomas Wasow, and Adrian Akmajian, 71–132. New York: Academic.

    Google Scholar 

  56. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.

    Google Scholar 

  57. Chomsky, Noam. 1986. Barriers. Cambridge: MIT Press.

    Google Scholar 

  58. Chomsky, Noam. 1993. A Minimalist Program for linguistic theory. In The view from the building 20: Essays in linguistics in honor of Sylvain Bromberger, eds. Ken Hale and Samuel J. Keyser, 1–52. Cambridge: MIT Press.

    Google Scholar 

  59. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essays on minimalist syntax in honor of Howard Lasnik, eds. Roger Martin, David Michaels, Juan Uriagereka, and Samuel J. Keyser, 89–155. Cambridge: MIT Press.

    Google Scholar 

  60. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A life in language, ed. Michael Kenstowicz, 1–52. Cambridge: MIT Press.

    Google Scholar 

  61. Chomsky, Noam. 2008. On phases. In Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud, eds. Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta, 133–165. Cambridge: MIT Press.

    Google Scholar 

  62. Christaller, Johann G. 1875/1964. A grammar of the Asante and Fante language called Tshi [Chwee, Twi]. New Jersey: Gregg Press.

    Google Scholar 

  63. Citko, Barbara. 2014. Phase theory: An introduction. Cambridge: Cambridge University Press.

    Google Scholar 

  64. Clements, George. 1984a. Binding domains in Kikuyu. Studies in the Linguistic Sciences 14: 37–56.

    Google Scholar 

  65. Clements, George. 1984b. Principles of tone assignment in Kikuyu. In Autosegmental studies in Bantu tone, eds. George Clements and John Goldsmith, 281–339. Dordrecht: Foris.

    Google Scholar 

  66. Clements, George, James McCloskey, Joan Maling, and Annie Zaenen. 1983. String-vacuous rule application. Linguistic Inquiry 14 (1): 1–17.

    Google Scholar 

  67. Cole, Peter, and Gabriella Hermon. 1998. The typology of wh-movement: Wh-questions in Malay. Syntax 1 (3): 221–258.

    Google Scholar 

  68. Collins, Chris. 1993. Topics in Ewe syntax. PhD diss., MIT.

  69. Collins, Chris. 1994. Economy of derivation and the Generalized Proper Binding Condition. Linguistic Inquiry 25 (1): 45–61.

    Google Scholar 

  70. Davies, William D. 2003. Extreme locality in Madurese wh-questions. Syntax 6 (3): 237–259.

    Google Scholar 

  71. Dayal, Veneeta. 2017. Does Hindi-Urdu have feature driven wh-movement to Spec vP? Linguistic Inquiry 48 (1): 159–172.

    Google Scholar 

  72. Demirdache, Hamida, and Myriam Uribe-Etxebarria. 2004. The syntax of time adverbs. In The syntax of time, eds. Jacqueline Guéron and Jacqueline Lecarme, 143–180. Cambridge: MIT Press.

    Google Scholar 

  73. de Vries, Mark. 2002. The syntax of relativization. PhD diss., University of Amsterdam.

  74. Dolphyne, Florence A. 1988. The Akan (Twi-Fante) language: Its sound systems and tonal structure. Accra: Ghana University Press.

    Google Scholar 

  75. Dolphyne, Florence A., and Mary E. Kropp Dakubu. 1988. The Volta-Comoé languages. In The languages of Ghana, ed. Mary Esther Kropp Dakubu, 50–90. New York: Kegan Paul.

    Google Scholar 

  76. du Plessis, Hans. 1977. Wh-movement in Afrikaans. Linguistic Inquiry 8 (4): 723–726.

    Google Scholar 

  77. Dubinsky, Stanley, and Robert Hamilton. 1998. Epithets as antilogophoric pronouns. Linguistic Inquiry 29 (4): 685–693.

    Google Scholar 

  78. É. Kiss, Katalin. 1993. Wh-movement and specificity. Natural Language and Linguistic Theory 11 (1): 85–120.

    Google Scholar 

  79. Elbourne, Paul. 2001. E-type anaphora as NP-deletion. Natural Language Semantics 9 (3): 241–288.

    Google Scholar 

  80. Embick, David. 2010. Localism versus globalism in morphology and phonology. Cambridge: MIT Press.

    Google Scholar 

  81. Embick, David. 2015. The morpheme: A theoretical introduction. Boston: de Gruyter.

    Google Scholar 

  82. Embick, David, and Ralf Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32 (4): 555–595.

    Google Scholar 

  83. Embick, David, and Ralf Noyer. 2007. Distributed morphology and the syntax/morphology interface. In Oxford handbook of linguistic interfaces, eds. Gillian Ramchand and Charles Reiss, 289–324. Oxford: Oxford University Press.

    Google Scholar 

  84. Epée, Roger. 1976. On some rules that are not successive cyclic in Duala. Linguistic Inquiry 7 (1): 193–198.

    Google Scholar 

  85. Ermisch, Sonja. 2006. Focus and topic constructions in Akan. Frankfurt Afrikanistische Blätter 18: 51–68.

    Google Scholar 

  86. Ermisch, Sonja. 2007. Issues in the left periphery: A typological approach to topic and focus constructions. Frankfurt a.M: Peter Lang.

    Google Scholar 

  87. Essegbey, James. 1999. Inherent complement verbs revisited: Towards an understanding of argument structure in Ewe. PhD diss., Radboud University Nijmegen.

  88. Essegbey, James. 2010. Inherent complement verbs and the basic double object construction in Gbe. In Topics in Kwa syntax, eds. Enoch O. Aboh and James Essegbey, 177–193. Dordrecht: Springer.

    Google Scholar 

  89. Felser, Claudia. 2004. Wh-copying, phases, and successive cyclicity. Lingua 114: 543–574.

    Google Scholar 

  90. Fiedler, Ines, and Anne Schwarz. 2005. Out-of-focus encoding in Gur and Kwa. In Interdisciplinary studies on information structure 03, eds. Shinichiro Ishihara, Michaela Schmitz, and Anne Schwarz, 111–142. Potsdam: Universität Potsdam.

    Google Scholar 

  91. Fiengo, Robert, and Robert May. 1994. Indices and identity. Cambridge: MIT Press.

    Google Scholar 

  92. Finer, Daniel L. 1997. Contrasting ā-dependencies in Selayarese. Natural Language and Linguistic Theory 15 (4): 677–728.

    Google Scholar 

  93. Fox, Danny. 1999. Reconstruction, Binding Theory and the interpretation of chains. Linguistic Inquiry 30 (2): 157–196.

    Google Scholar 

  94. Fox, Danny. 2002. Antecedent-contained deletion and the Copy Theory of Movement. Linguistic Inquiry 33 (1): 63–96.

    Google Scholar 

  95. Fraser, Bruce. 1970. Idioms within a transformational grammar. Foundations of Language 6 (1): 22–42.

    Google Scholar 

  96. Gazdar, Gerald, Ewan Klein, Geoffrey Pullum, and Ivan Sag. 1985. Generalized Phrase Structure Grammar. Cambridge, MA: Harvard University Press.

    Google Scholar 

  97. Geis, Michael L. 1970. Adverbial subordinate clauses in English. PhD diss., MIT.

  98. Geis, Michael L. 1975. English time and place adverbials. In Working papers in linguistics no. 18, eds. Michael L. Geis, Sheila G. Geoghegan, Jeanette K. Gundel, Ronald L. Neeld, Geoffrey K. Pullum, and Arnold M. Zwicky, 1–24. Columbus: Ohio State University.

    Google Scholar 

  99. Genzel, Susanne. 2013. Lexical and post-lexical tones in Akan. PhD diss., Universität Potsdam.

  100. Genzel, Susanne, and Frank Kügler. 2010. How to elicit semi-spontaneous focus realizations with specific tonal patterns. In Linguistic fieldnotes i: Information structure in different African languages, eds. Mira Grubic, Susanne Genzel, and Frank Kügler, 77–102. Potsdam: Universitätsverlag Potsdam.

    Google Scholar 

  101. Georgi, Doreen. 2014. Opaque interactions of Merge and Agree: On the nature and order of elementary operations. PhD diss., Universität Leipzig.

  102. Georgi, Doreen. 2017. Patterns of movement reflexes as the result of the order of Merge and Agree. Linguistic Inquiry 48 (4): 585–626.

    Google Scholar 

  103. Giannakidou, Anastasia, and Jason Merchant. 1998. Reverse sluicing in English and Greek. The Linguistic Review 15 (2–3): 233–256.

    Google Scholar 

  104. Goodluck, Helen, Kofi K. Saah, and Danijela Stojanović. 1995. On the default mechanism for interrogative binding. Canadian Journal of Linguistics 40 (4): 377–404.

    Google Scholar 

  105. Griffiths, James, and Anikó Lipták. 2014. Contrast and island sensitivity. Syntax 17 (3): 189–234.

    Google Scholar 

  106. Guilliot, Nicolas, and Nouman Malkawi. 2006. When resumption determines reconstruction. In West Coast Conference on Formal Linguistics (WCCFL) 25, eds. Donald Baumer, David Montero, and Michael Scanlon, 168–176. Cambridge: Cascadilla Press.

    Google Scholar 

  107. Guilliot, Nicolas, and Nouman Malkawi. 2011. Weak versus strong resumption: Covarying differently. In Resumptive pronouns at the interfaces, ed. Alain Rouveret, 395–423. Amsterdam: John Benjamins.

    Google Scholar 

  108. Haegeman, Liliane. 2007. The movement analysis of temporal adverbial clauses. Folia Linguistica 41 (3–4): 279–325.

    Google Scholar 

  109. Haegeman, Liliane. 2010. The internal syntax of adverbial clauses. Lingua 120: 628–648.

    Google Scholar 

  110. Halle, Morris. 1997. Distributed Morphology: Impoverishment and fission. In PF: Papers at the interface, eds. Benjamin Bruening, Yoonjung Kang, and Martha McGinnis, 425–449. Cambridge: MITWPIL.

    Google Scholar 

  111. Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. In The View from Building 20, eds. Ken Hale and Samuel J. Keyser, 111–176. Cambridge: MIT Press.

    Google Scholar 

  112. Harizanov, Boris. 2014. Clitic doubling and the syntax-morphophonology interface: A-movement and morphological merger in Bulgarian. Natural Language and Linguistic Theory 32 (4): 1033–1088.

    Google Scholar 

  113. Harley, Heidi. 2011. Affixation and the Mirror Principle. In Interfaces in linguistics: New research perspectives, eds. Raffaella Folli and Christiane Ullbricht, 166–186. Oxford: Oxford University Press.

    Google Scholar 

  114. Harley, Heidi. 2013. External arguments and the Mirror Principle: On the distinctness of Voice and v. Lingua 125: 34–57.

    Google Scholar 

  115. Harley, Heidi, and Rolf Noyer. 2003. Distributed Morphology. In The second Glot international state-of-the-article book, eds. Lisa Cheng and Rint Sybesma, 463–496. Berlin: de Gruyter.

    Google Scholar 

  116. Harwood, William. 2015. Being progressive is just a phase: Celebrating the uniqueness of progressive aspect under a phase-based analysis. Natural Language and Linguistic Theory 33 (2): 523–573.

    Google Scholar 

  117. Heestand, Dustin, Ming Xiang, and Maria Polinsky. 2011. Resumption still does not rescue islands. Linguistic Inquiry 42 (1): 138–152.

    Google Scholar 

  118. Hein, Johannes. 2017. Doubling and do-support in verbal fronting: Towards a typology of repair operations. Glossa 2 (1): 67–136.

    Google Scholar 

  119. Hellan, Lars, Dorothée Beermann, and Eli S. Andenes. 2003. Towards a typology of serial verb constructions in Akan. In Studies in the languages of the Volta Basin, eds. Mary Esther Kropp Dakubu and E. Kwekuk Osam, Vol. 1, 61–86. Legon: University of Ghana.

    Google Scholar 

  120. Henderson, Brent. 2013. Agreement and person in anti-agreement. Natural Language and Linguistic Theory 31 (2): 453–481.

    Google Scholar 

  121. Henry, Alison. 1995. Belfast English and Standard English: Dialect variation and parameter setting. Oxford: Oxford University Press.

    Google Scholar 

  122. Higginbotham, James. 1980. Pronouns and bound variables. Linguistic Inquiry 11 (4): 679–708.

    Google Scholar 

  123. Hornstein, Norbert, Howard Lasnik, and Juan Uriagereka. 2007. The dynamics of islands: Speculations on the locality of movement. Linguistic Analysis 33 (1–2): 149–175.

    Google Scholar 

  124. Huang, C. T. James. 1982. Move WH in a language without wh-movement. The Linguistic Review 1 (4): 369–416.

    Google Scholar 

  125. Hunter, Tim, and Masaya Yoshida. 2016. A restriction on Vehicle Change and its interaction with movement. Linguistic Inquiry 47 (3): 561–571.

    Google Scholar 

  126. Hyde, Brett. 2012. Alignment constraints. Natural Language and Linguistic Theory 30 (3): 789–836.

    Google Scholar 

  127. Johnson, Kyle. 1988. Clausal gerunds, the ECP and Government. Linguistic Inquiry 19 (4): 585–609.

    Google Scholar 

  128. Kahnemuyipour, Arsalan, and Karine Megerdoomian. 2011. Second-position clitics in the vP phase: The case of the Armenian auxiliary. Linguistic Inquiry 42 (1): 152–162.

    Google Scholar 

  129. Kandybowicz, Jason. 2006. Comp-trace effects explained away. In West Coast Conference on Formal Linguistics (WCCFL) 25, eds. Donald Baumer, David Montero, and Michael Scanlon, 220–228. Somerville: Cascadilla Proceedings Project.

    Google Scholar 

  130. Kandybowicz, Jason. 2008. The grammar of repetition: Nupe grammar at the syntax-phonology interface. Amsterdam: John Benjamins.

    Google Scholar 

  131. Kandybowicz, Jason. 2009. Embracing edges: Syntactic and phono-syntactic edge sensitivity in Nupe. Natural Language and Linguistic Theory 27 (2): 305–344.

    Google Scholar 

  132. Kandybowicz, Jason. 2015. On prosodic vacuity and verbal resumption in Asante Twi. Linguistic Inquiry 46 (2): 243–272.

    Google Scholar 

  133. Kandybowicz, Jason. 2017. On prosodic variation and the distribution of wh-in-situ. Linguistic Variation 17 (1): 111–148.

    Google Scholar 

  134. Kayne, Richard, and Jean-Yves Pollock. 1978. Stylistic inversion, successive-cyclicity and move NP in French. Linguistic Inquiry 9 (4): 595–621.

    Google Scholar 

  135. Keine, Stefan. 2016. Probes and their horizons. PhD diss., University of Massachusetts Amherst.

  136. Keine, Stefan. 2017. Agreement and vP phases. In A schrift to fest Kyle Johnson, eds. Nicholas LaCara, Keir Moulton, and Anne-Michelle Tessier, 177–185. Amherst: Linguistics Open Access Publications.

    Google Scholar 

  137. Kim, Yuni. 2016. Tonal overwriting and inflectional exponence in Amuzgo. In Tone and inflection: New facts and new perspectives, eds. Enrique L. Palancar and Jean L. Léonard, 199–224. Berlin: de Gruyter.

    Google Scholar 

  138. Kobele, Gregory M., and Harold Torrence. 2006. Intervention and focus in Asante Twi. ZAS Papers in Linguistics 46: 161–184.

    Google Scholar 

  139. Koopman, Hilda, and Dominique Sportiche. 1982. Variables and the Bijection Principle. The Linguistic Review 2 (2): 136–160.

    Google Scholar 

  140. Koopman, Hilda, and Dominique Sportiche. 1986. A note on long extraction in Vata and the ECP. Natural Language and Linguistic Theory 4 (3): 357–374.

    Google Scholar 

  141. Korsah, Sampson. 2016a. Beyond resumptives and expletives in Akan. In Diversity in African languages: Selected papers from the 46th Annual Conference on African Linguistics, eds. Doris L. Payne, Sara Pacchiarotti, and Mokaya Bosire, 103–117. Berlin: Language Science Press.

    Google Scholar 

  142. Korsah, Sampson. 2016b. On deriving inherent complement verb constructions. In ConSOLE 23, eds. Kate Bellamy, Elena Karvovskaya, Martin Kohlberger, and George Saad, 52–69. Leiden: Leiden University Linguistics Centre.

    Google Scholar 

  143. Korsah, Sampson. 2017. Issues in Kwa syntax: Pronouns and clausal determiners. PhD diss., Universität Leipzig.

  144. Korsah, Sampson, and Andrew Murphy. 2016. What can tone tell us about successive-cyclic movement? evidence from Asante Twi. In Proceedings of NELS 46, eds. Christopher Hammerly and Brandon Prickett, 227–240. Amherst: GLSA.

    Google Scholar 

  145. Kramer, Ruth. 2010. The Amharic definite marker and the syntax-morphology interface. Syntax 13 (3): 196–240.

    Google Scholar 

  146. Kratzer, Angelika, and Elisabeth Selkirk. 2007. Phase theory and prosodic spellout: The case of verbs. The Linguistic Review 24 (2–3): 93–135.

    Google Scholar 

  147. Kroch, Anthony. 1981. On the role of resumptive pronouns in amnestying island constraint violations. In Papers from the 17th regional meeting of the Chicago Linguistic Society, eds. Roberta A. Hendrick, Carrie S. Masek, and Mary F. Miller, 125–135. Chicago: Chicago Linguistic Society.

    Google Scholar 

  148. Kropp Dakubu, Mary Esther. 2009. Akan. In Concise encyclopedia of languages of the world, eds. Keith Brown and Sarah Ogilvie, 17–20. Amsterdam: Elsevier.

    Google Scholar 

  149. Kügler, Frank. 2015. Phonological phrasing and ATR vowel harmony in Akan. Phonology 32 (1): 177–204.

    Google Scholar 

  150. Kügler, Frank. 2016a. Embedded clauses and recursive prosodic phrasing in Akan. In Tonal Aspects of Languages 2016, ed. Christian DiCanio et al., 38–42. Baixas: ISCA.

    Google Scholar 

  151. Kügler, Frank. 2016b. Tone and intonation in Akan. In Intonation in African tone languages, eds. Laura J. Downing and Annie Rialland, 89–130. Berlin: de Gruyter.

    Google Scholar 

  152. Kurisu, Kazutaka. 2001. The phonology of morpheme realization. PhD diss., University of California Santa Cruz.

  153. Kusmer, Leland. 2011. There is no future: Diachronic verbal morphology in Fante Twi. Ms. Swarthmore College.

  154. Kusumoto, Kiyomi. 2017. Geis-ambiguity and tense harmony. In A schrift to fest Kyle Johnson, eds. Nicholas LaCara, Keir Moulton, and Anne-Michelle Tessier, 203–205. Amherst: Linguistics Open Access Publications.

    Google Scholar 

  155. Lahne, Antje. 2008. Where there is fire there is smoke. Local modelling of successive-cyclic movement. PhD diss., Universität Leipzig.

  156. Landman, Meredith. 2003. Morphological contiguity. In Papers in Optimality Theory II: University of Massachusetts-Amherst Occasional Papers in Linguistics, eds. Angela Carpenter, Andries Coetzee, and Paul de Lacy, Vol. 26. Amherst: GLSA.

    Google Scholar 

  157. Larson, Richard K. 1987. ‘Missing prepositions’ and the analysis of English free relative constructions. Linguistic Inquiry 18 (2): 239–266.

    Google Scholar 

  158. Larson, Richard K. 1990. Extraction and multiple selection in PP. The Linguistic Review 7 (2): 169–182.

    Google Scholar 

  159. Larson, Richard K., and Robert May. 1990. Antecedent containment or vacuous movement: Reply to Baltin. Linguistic Inquiry 21 (1): 103–122.

    Google Scholar 

  160. Lasnik, Howard. 2001. When can you save a structure by destroying it? In Northeast Linguistic Society (NELS) 31, eds. Minjoo Kim and Uri Strauss, 301–320. Amherst: GLSA.

    Google Scholar 

  161. Lasnik, Howard, and Tim Stowell. 1991. Weakest crossover. Linguistic Inquiry 22 (4): 687–720.

    Google Scholar 

  162. Legate, Julie Anne. 2003. Some interface properties of the phase. Linguistic Inquiry 34 (3): 506–516.

    Google Scholar 

  163. Manetta, Emily. 2010. Wh-expletives in Hindi-Urdu: The vP phase. Linguistic Inquiry 41 (1): 1–34.

    Google Scholar 

  164. Marfo, Charles. 2005a. Akan focus and topic constructions and the prosody-syntax interface. Journal of West African Languages 32 (1–2): 45–59.

    Google Scholar 

  165. Marfo, Charles. 2005b. Aspects of Akan grammar and the phonology-syntax interface. PhD diss., University of Hong Kong.

  166. Marfo, Charles, and Adams Bodomo. 2005. Information structuring in Akan question-word fronting and focus constructions. Studies in African Linguistics 32 (4): 179–208.

    Google Scholar 

  167. May, Robert C. 1985. Logical Form: Its structure and derivation. Cambridge: MIT Press.

    Google Scholar 

  168. McCarthy, John J. 2010. Autosegmental spreading in Optimality Theory. In Tones and features: Phonetic and phonological perspectives, eds. John A. Goldsmith, Elizabeth Hume, and Leo Wetzels, 195–222. Berlin: de Gruyter.

    Google Scholar 

  169. McCarthy, John J., and Alan Prince. 1993. Generalized alignment. In Yearbook of morphology 1993, eds. Geert Booij and Jaap van Marle, 79–153. Dordrecht: Kluwer.

    Google Scholar 

  170. McCloskey, James. 1979. Transformational syntax and model theoretic semantics: A case study in Modern Irish. Riedel: Dordrecht.

    Google Scholar 

  171. McCloskey, James. 2000. Quantifier float and wh-movement in an Irish English. Linguistic Inquiry 31 (1): 57–84.

    Google Scholar 

  172. McCloskey, James. 2001. The morphosyntax of WH-extraction in Irish. Journal of Linguistics 37 (1): 67–100.

    Google Scholar 

  173. McCloskey, James. 2002. Resumption, successive cyclicity, and the locality of operations. In Derivation and explanation in the Minimalist Program, eds. Samuel Epstein and T. David Seely, 184–226. Oxford: Blackwell.

    Google Scholar 

  174. McCloskey, James. 2011. Resumptive pronouns, Ā-binding and levels of representation in Irish. In Resumptive pronouns at the interfaces, ed. Alain Rouveret, 65–119. Amsterdam: John Benjamins.

    Google Scholar 

  175. McCracken, Chelsea. 2013. Relative clauses in Asante Twi. Rice Working Papers in Linguistics 4: 1–28.

    Google Scholar 

  176. McPherson, Laura. 2014. Replacive grammatical tone in the Dogon languages. PhD diss., UCLA.

  177. McPherson, Laura, and Jeffrey Heath. 2016. Phrasal grammatical tone in the Dogon languages: The role of constraint interaction. Natural Language and Linguistic Theory 34 (2): 593–639.

    Google Scholar 

  178. Merchant, Jason. 2001. The syntax of silence: Sluicing, islands and the theory of ellipsis. Oxford: Oxford University Press.

    Google Scholar 

  179. Morgan, Adam M., and Matthew W. Wagers. 2018. English resumptive pronouns are more common where gaps are less acceptable. Linguistic Inquiry 49 (4): 861–876.

    Google Scholar 

  180. Müller, Gereon. 2007. Towards a relativized concept of Cyclic Linearization. In Interfaces + recursion = language?, eds. Uli Sauerland and Hans-Martin Gärtner, 61–114. Berlin: de Gruyter.

    Google Scholar 

  181. Müller, Gereon. 2011. Constraints on displacement: A phase-based approach. Amsterdam: John Benjamins.

    Google Scholar 

  182. Muriungi, Peter. 2005. Wh-questions in Kitharaka. Studies in African Linguistics 34 (1): 43–104.

    Google Scholar 

  183. Nespor, Marina, and Irene Vogel. 1986. Prosodic phonology. Dordrecht: Foris.

    Google Scholar 

  184. Nissenbaum, Jon. 2000. Investigations of covert phrase movement. PhD diss., MIT.

  185. Noonan, Máire. 1999. Successive cyclicity phenomena in wh-movement constructions. In West Coast Conference on Formal Linguistics (WCCFL) 18, eds. Sonya Bird, Andrew Carnie, Jason Haugen, and Peter Norquest, 406–419. Somerville: Cascadilla Press.

    Google Scholar 

  186. Noonan, Máire. 2002. CP-pied-piping and remnant IP movement in long distance wh-movement. In Dimensions of movement: From features to remnants, eds. Artemis Alexiadou, Elena Anagnostopoulou, Sjef Barbiers, and Hans-Martin Gärtner, 269–295. Amsterdam: John Benjamins.

    Google Scholar 

  187. Nunberg, Geoffrey, Ivan A. Sag, and Thomas Wasow. 1994. Idioms. Language 70 (3): 491–538.

    Google Scholar 

  188. Nwachukwu, P. Akujuoobi. 1985. Inherent complement verbs in Igbo. Journal of the Linguistics Association of Nigeria (JOLAN) 3: 61–74.

    Google Scholar 

  189. Ofori, Seth Antwi. 2006. Topics in Akan grammar. PhD diss., Indiana University.

  190. Ofori, Seth Antwi. 2011. On the basic focus marker, and the basic focus sentence, in Akan (Twi). Nordic Journal of African Studies 20 (3): 241–262.

    Google Scholar 

  191. Osam, Emmanuel K. 1994. Aspects of Akan grammar: A functional perspective. PhD diss., University of Oregon.

  192. Osam, Emmanuel K. 1996. Animacy distinctions in Akan grammar. Studies in the Linguistic Sciences 23 (2): 153–164.

    Google Scholar 

  193. Osam, Emmanuel K. 2008. Akan as an aspectual language. In Aspect and modality in Kwa languages, eds. Felix K. Ameka and Mary Esther Kropp Dakubu, 69–89. Amsterdam: John Benjamins.

    Google Scholar 

  194. Ouhalla, Jamal. 1993. Subject-extraction, negation and the anti-agreement effect. Natural Language and Linguistic Theory 11 (3): 477–518.

    Google Scholar 

  195. Pan, Victor J. 2016. Resumptivity in Mandarin Chinese. Berlin: de Gruyter.

    Google Scholar 

  196. Panitz, Ezekiel J. 2014. Strong islands, resumption and reconstruction. UCL Working Papers in Linguistics 26: 13–42.

    Google Scholar 

  197. Paster, Mary. 2010. The verbal morphology and phonology of Asante Twi. Studies in African Linguistics 39 (1): 77–120.

    Google Scholar 

  198. Pesetsky, David. 1987. Wh-in-situ: Movement and Unselective Binding. In Representation of (in)definiteness, eds. Eric Reuland and Alice Ter Meulen, 98–129. Cambridge: MIT Press.

    Google Scholar 

  199. Pesetsky, David. 1998. Some Optimality principles of sentence pronunciation. In Is the best good enough? optimality and competition in syntax, eds. Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky, 337–383. Cambridge: MIT Press.

    Google Scholar 

  200. Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge: MIT Press.

    Google Scholar 

  201. Pfeil, Simone, Susanne Genzel, and Frank Kügler. 2015. Empirical investigation of focus and exhaustivity in Akan. In Mood, exhaustivity and focus marking in non-european languages, eds. Felix Bildhauer and Mira Grubic, 87–109. Potsdam: Universitätsverlag Potsdam.

    Google Scholar 

  202. Postal, Paul M. 1966. On so-called ‘pronouns’ in English. In Report of the 17th annual round table meeting on language and linguistics, ed. F. P. Dinneen, 177–206. Washington, D.C: Georgetown University Press.

    Google Scholar 

  203. Postal, Paul M. 1971. Cross-over phenomena. New York: Holt, Rinehart and Winston.

    Google Scholar 

  204. Reglero, Lara. 2007. Wh-in-situ interrogatives in Spanish. Probus 19 (1): 267–297.

    Google Scholar 

  205. Reinhart, Tanya. 1998. Wh-in-situ in the framework of the Minimalist Program. Natural Language Semantics 6 (1): 29–56.

    Google Scholar 

  206. Reintges, Chris H., Philip LeSourd, and Sandra Chung. 2006. Movement, wh-agreement and apparent wh-in-situ. In Wh-movement: Moving on, eds. Lisa Cheng and Norbert Corver, 165–194. Cambridge: MIT Press.

    Google Scholar 

  207. Richards, Norvin. 2001. Movement in language: Interactions and architectures. Oxford: Oxford University Press.

    Google Scholar 

  208. Richards, Norvin. 2010. Uttering trees. Cambridge: MIT Press.

    Google Scholar 

  209. Riis, Hans N. 1854. Grammatical outline and vocabulary of the Oji-language with especial reference to the Akuapim-dialect. Basel: G. Detloff.

    Google Scholar 

  210. Ross, John R. 1969. Guess who? In Papers from the 5th regional meeting of Chicago Linguistic Society (CLS), eds. Robert Binnick, Alice Davidson, Georgia Green, and Jerry Morgan, 252–286. Chicago: Chicago Linguistic Society.

    Google Scholar 

  211. Ross, John R. 1986. Infinite syntax! Norwood: Ablex Publishing.

    Google Scholar 

  212. Rouveret, Alain. 2008. Phasal agreement and reconstruction. In Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vegnaud, eds. Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta, 167–195. Cambridge: MIT Press.

    Google Scholar 

  213. Ruys, Eddy G. 2000. Weak Crossover as a scope phenomenon. Linguistic Inquiry 31 (3): 513–539.

    Google Scholar 

  214. Ruys, Eddy G. 2004. A note on weakest Crossover. Linguistic Inquiry 35 (1): 124–140.

    Google Scholar 

  215. Saah, Kofi K. 1988. Wh-questions in Akan. Journal of West African Languages 18 (1): 17–28.

    Google Scholar 

  216. Saah, Kofi K. 1989. Reflexivization in Akan. Journal of West African Languages 19 (2): 15–28.

    Google Scholar 

  217. Saah, Kofi K. 1992. Null object constructions in Akan. In Proceedings of the Kwa comparative syntax workshop, eds. Chris Collins and Victor Manfredi, 219–244. Cambridge: MITWPL.

    Google Scholar 

  218. Saah, Kofi K. 1994. Studies in Akan syntax, acquisition and sentence processing. PhD diss., University of Ottawa.

  219. Saah, Kofi K. 2010. Relative clauses in Akan. In Topics in Kwa syntax, eds. Enoch O. Aboh and James Essegbey, 91–109. Dordrecht: Springer.

    Google Scholar 

  220. Saah, Kofi K., and Helen Goodluck. 1995. Island effects in parsing and grammar: Evidence from Akan. The Linguistic Review 12 (4): 381–409.

    Google Scholar 

  221. Saddy, Douglas. 1991. Wh-scope mechanisms in Bahasa Indonesia. In More papers on wh-movement, eds. Lisa L.-S. Cheng and Hamida Demirdache, 183–218. Cambridge: MITWPIL.

    Google Scholar 

  222. Safir, Ken. 1984. Multiple variable binding. Linguistic Inquiry 15 (4): 603–638.

    Google Scholar 

  223. Safir, Ken. 1996. Derivation, representation and resumption: The domain of Weak Crossover. Linguistic Inquiry 27 (2): 313–339.

    Google Scholar 

  224. Safir, Ken. 1999. Vehicle Change and reconstruction in Ā-chains. Linguistic Inquiry 30 (4): 587–620.

    Google Scholar 

  225. Salzmann, Martin. 2017a. Prolepsis. In The Wiley-Blackwell companion to syntax, eds. Martin Everaert and Henk van Riemsdijk, Vol. 5, 3203–3245. Somerset: Wiley-Blackwell.

    Google Scholar 

  226. Salzmann, Martin. 2017b. Reconstruction and resumption in indirect A-dependencies: On the syntax of prolepsis and relativization in (Swiss) German and beyond. Berlin: de Gruyter.

    Google Scholar 

  227. Sato, Yosuke. 2012. Multiple spell-out and contraction at the syntax-phonology interface. Syntax 15 (3): 287–314.

    Google Scholar 

  228. Sato, Yosuke, and Yoshihito Dobashi. 2016. Prosodic phrasing and the That-trace effect. Linguistic Inquiry 47 (2): 333–349.

    Google Scholar 

  229. Sauerland, Uli. 2003. Intermediate adjunction with A-movement. Linguistic Inquiry 34 (2): 308–314.

    Google Scholar 

  230. Schachter, Paul. 1969. Natural assimilation rules in Akan. International Journal of American Linguistics 35 (4): 342–355.

    Google Scholar 

  231. Schachter, Paul. 1973. Focus and relativization. Language 49 (1): 19–46.

    Google Scholar 

  232. Schachter, Paul, and Victoria Fromkin. 1968. A phonology of Akan: Akuapem, Asante, Fante. Los Angeles: University of California, Los Angeles.

    Google Scholar 

  233. Schneider-Zioga, Patricia. 2007. Anti-agreement, anti-locality and minimality: The syntax of dislocated subjects. Natural Language and Linguistic Theory 25 (2): 403–446.

    Google Scholar 

  234. Schneider-Zioga, Patricia. 2009. Wh-agreement and bounded unbounded movement. In Merging features: Computation, interpretation and acquisition, eds. José M. Brucart, Anna Gavarró, and Jaume Solà, 46–59. Oxford: Oxford University Press.

    Google Scholar 

  235. Schwarz, Anne, and Ines Fiedler. 2007. Narrative focus strategies in Gur and Kwa. In Focus strategies in African languages: The interaction of focus and grammar in Niger-Congo and Afro-Asiatic, eds. Enoch O. Aboh, Katharina Hartmann, and Malte Zimmermann, 267–286. Berlin: de Gruyter.

    Google Scholar 

  236. Selkirk, Elisabeth. 1986. On derived domains in sentence phonology. Phonology Yearbook 3: 371–405.

    Google Scholar 

  237. Selkirk, Elisabeth. 2011. The syntax-phonology interface. In The handbook of phonological theory, eds. John Goldsmith, Jason Riggle, and Alan Yu, 435–485. Oxford: Blackwell.

    Google Scholar 

  238. Sells, Peter. 1984. Syntax and semantics of resumptive pronouns. PhD diss., University of Massachusetts Amherst.

  239. Shlonsky, Ur. 1992. Resumptive pronouns as a Last Resort. Linguistic Inquiry 23 (3): 443–468.

    Google Scholar 

  240. Sichel, Ivy. 2014. Resumptives pronouns and competition. Linguistic Inquiry 45 (4): 655–693.

    Google Scholar 

  241. Soh, Hooi Ling. 2005. Wh-in-situ in Mandarin Chinese. Linguistic Inquiry 36 (1): 143–155.

    Google Scholar 

  242. Stump, Gregory. 2016. Inflectional paradigms: Content and form at the syntax-morphology interface. Cambridge: Cambridge University Press.

    Google Scholar 

  243. Thiersch, Craig. 1978. Topics in German syntax. PhD diss., MIT.

  244. Thompson, Ellen. 1995. Temporal ambiguity of clausal adverbs and the syntax of simultaneity. In Northeast Linguistic Society (NELS) 25, ed. Jill Beckman, 473–487. Amherst: GLSA.

    Google Scholar 

  245. Thoms, Gary. 2011. From economy to locality: do-support as head movement. Ms. Strathclyde University.

  246. Titov, Elena. 2019. Morphosyntactic encoding of information structure in Akan. Glossa 4 (1): 27–136.

    Google Scholar 

  247. Torrego, Esther. 1984. On inversion in Spanish and some of its effects. Linguistic Inquiry 15 (1): 103–129.

    Google Scholar 

  248. Trommer, Jochen. 2011. Phonological aspects of Western Nilotic mutation morphology. Habilitation thesis, Universität Leipzig.

  249. Truckenbrodt, Hubert. 1999. On the relation between syntactic phrases and phonological phrases. Linguistic Inquiry 30 (2): 219–255.

    Google Scholar 

  250. van Urk, Coppe. 2015. A uniform syntax for phrasal movement: A case study of Dinka Bor. PhD diss., MIT.

  251. van Urk, Coppe. 2018. Pronoun copying in Dinka and the Copy Theory of Movement. Natural Language and Linguistic Theory 36 (3): 937–990.

    Google Scholar 

  252. van Urk, Coppe, and Norvin Richards. 2015. Two components of long-distance extraction: Successive cyclicity in Dinka. Linguistic Inquiry 46 (1): 113–155.

    Google Scholar 

  253. Vergnaud, Jean Roger. 1974. French relative clauses. PhD diss., MIT.

  254. Wiland, Bartosz. 2010. Overt evidence from Left-Branch Extraction in Polish for puncuated paths. Linguistic Inquiry 41 (2): 335–347.

    Google Scholar 

  255. Willis, David. 2000. On the distribution of resumptive pronouns and wh-trace in Welsh. Journal of Linguistics 36 (3): 531–573.

    Google Scholar 

  256. Yip, Moira. 2002. Tone. Cambridge: Cambridge Universty Press.

    Google Scholar 

  257. Zaenen, Annie. 1983. On syntactic binding. Linguistic Inquiry 14 (3): 469–504.

    Google Scholar 

  258. Zaenen, Annie, Elisabeth Engdahl, and Joan M. Maling. 1981. Resumptive pronouns can be syntactically bound. Linguistic Inquiry 12 (4): 679–682.

    Google Scholar 

  259. Zentz, Jason. 2014. Morphological evidence for a movement analysis of adverbial clauses. In Chicago Linguistic Society (CLS) 47, eds. Carissa Abrego-Collier, Arum Kang, Martina Martinović, and Chieu Nguyen, 373–385. Chicago: Chicago Linguistic Society.

    Google Scholar 

  260. Zimmermann, Eva. 2017. Morphological length and prosodically defective morphemes. Oxford: Oxford University Press.

    Google Scholar 

Download references


We would like to thank three anonymous NLLT reviewers for their valuable and challenging comments. For helpful feedback on various stages of this work, we would like to thank audiences at Morphosyntactic Triggers of Tone at Universität Leipzig, Syntax Brown Bag at NYU, LingLunch at UConn, NELS 45 at Concordia University, Montréal, SyntaxLab at the University of Cambridge and colloquia at University College London, Leibniz-Zentrum für Allgemeine Sprachwissenschaft, Berlin and University of California Santa Cruz. Thanks also go to Augstina Pokua Owusu and Joana Serwaa Ampofo for discussion of the data. This research was completed as part of the DFG research training group Interaction of Grammatical Building Blocks (GRK 2011).

Author information



Corresponding author

Correspondence to Andrew Murphy.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Korsah, S., Murphy, A. Tonal reflexes of movement in Asante Twi. Nat Lang Linguist Theory 38, 827–885 (2020).

Download citation


  • Movement
  • Phases
  • Successive-cyclic movement
  • Tone
  • Syntax