2.1 Introduction to Yorùbá

The Yorùbá language belongs to the Niger-Congo language family and together with the language groups of Igálà and Itsẹkiri, it forms the Yoruboid branch of the Volta-Niger group. It is spoken mainly in Nigeria, Benin and Togo and has, according to ethnologue.com, about 50 million native speakers.

The phonology of Yorùbá is comparably well-studied and has been discussed in much detail in works by Akinlabi (1985); Pulleyblank (1986); Archangeli & Pulleyblank (1989); Akinlabi & Liberman (2001). Yorùbá has seven vowels which differ in [±ATR].

Words in Yorùbá including compounds show intricate harmony effects with respect to [ATR].

The tonal system of Yorùbá is discussed in Akinlabi (1985); Pulleyblank (1986); Akinlabi & Liberman (2001). According to these descriptions, Yorùbá has three tones:

Akinlabi (1985) and Pulleyblank (1986) provide a number of arguments that mid-tones are underlyingly toneless. In some configurations, tones are subject to an OCP-like dissimilation process that, in some configurations, extends beyond words and which will be used as a diagnostic in Section 6.

As for the syntactic properties of the language, we note that Yorùbá is an SVO-language with relatively rigid, head-initial order in most domains. Tense, negation and aspect appear in between the subject and the verb. In line with its head-initial status, the language makes use of prepositions and clause-initial complementizers (see e.g. Awobuluyi 1977; Ilori 2010). The examples below illustrate some of these features.

One feature that will play a role in the discussion below is that pronominal arguments as well as many of the grammatical particles such as negation, tense, mood and aspect interact with each other in a way that their surface form is conditioned by the other particles in the relevant configuration. Awobuluyi (1977) lists five different surface forms for the first person singular subject pronoun depending on whether it precedes certain tense, negation or mood markers.

Another feature that we briefly want to mention for now is the so-called high-tone syllable, glossed as hts in example (3b). The hts is an element following the subject in some configurations. If the subject ends in a low tone, the hts copies the final vowel of the subject and projects its high tone as in (3b). If the subject ends in a mid-vowel, this mid-vowel is simply overwritten (4a), which is one of the arguments why Akinlabi (1985) and Pulleyblank (1986) argue that mid-tones are underlyingly toneless. If the subject ends in a high vowel, the hts does not surface (4c). There is a fair amount of literature on the precise function of this element (see e.g. Awobuluyi (1977); Bisang & Sonaiya (1999)) but, as far as we can tell, the standard account is that it expresses some features relating the subject to the predicate (see Dechaine (1993)) in a way that is often attributed to subject agreement and that it is restricted non-future tense (see Ilori (2010) for discussion) as it is incompatible with the future marker yóò (4b):

As with the dissimilations of tone, we will briefly come back to the hts in Section 6. For now, we will move to taking a closer look at the syntax of coordination in Yorùbá. We will see more intricacies of Yorùbá syntax as we go along.

2.2 Coordination in Yorùbá

The grammatical markers used to indicate coordination in Yorùbá distinguish between the clausal and the non-clausal domain. In the non-clausal domain, the markers àti and pẹ̀lú are all used to coordinate nouns but according to Awobuluyi (1977: 104f) only àti is used to coordinate other elements such as PPs (6a).²

In the clausal domain however, these markers cannot be employed. Two straightforward coordinators used for clausal coordination are àmọ́ and ṣùgbọ́n, both of which express an adversative relation between the two conjuncts that is usually translated with the English coordinator but:

Simple, non-adversative coordination of clauses usually makes use of the element sì. Curiously, however, sì does not occur in the same position as the above-mentioned elements àmọ́ and ṣùgbọ́n. Rather it appears in a position somewhere between the subject and the verb of the second conjunct.

Given examples as in (9) and (10), the literature is divided as to whether sì should be treated as a regular coordinator or not. Awobuluyi (1977) and Ilori (2010) both reject the notion of sì as a coordinator and call it a consecutive adverb. Accordingly Ilori (2010) glosses sì as ‘in-addition’ and Awobuluyi (1977) sometimes translates it to English ‘then’. Yusuf (1980), Abimbola (2017) and Givón (2018) on the other hand take sì to be a coordinator.

In what follows, we will present three initial arguments that defend the claim of Yusuf (1980); Abimbola (2017); Givón (2018) that sì is a proper coordinator. The first argument is that sì is by no means restricted to cases of consecutive actions. This is shown in the following example taken from Abimbola (2017). Here, the two conjuncts are not in a consecutive relation; rather they express two states that both hold simultaneously. A consecutive adverb like English then is arguably not felicitous in such contexts.

What we take this to mean is that Yorùbá sì can be used in all contexts in which English and can be used. Consecutive readings of coordination are simply pragmatically inferred in some cases. A simple notion of sì being a consecutive adverb can thus quite easily be shown to be untenable.

The second argument that sì is a proper coordination is based on the fact that sì is also in complementary distribution with the more regular conjunctions ṣùgbọ́n and àmọ́ that we saw above. Examples like (12) and (13) are perceived as extremely marginal or are flat-out rejected by the speakers we consulted.

Such behavior would be completely unexpected if sì were really an adverb because we do not expect a vP/VP-adverb to depend on the presence or absence of a clausal coordinator. In English for example, a consecutive adverb can easily cooccur with an adversative conjunction (‘John went to bed early but then he still woke up too late’). In fact, we would like to submit that this piece of data strongly suggests not only that sì is not an adverb but rather that sì should be treated as a proper conjunction.

The third argument that sì is a proper coordinator is that it licenses coordination-specific processes; a diagnostic developed in detail by Weisser (2022). One such process is VP-ellipsis. In (14a), we see that VP-ellipsis is licensed by the regular adversative coordinator ṣúgbọ́n. The example in (14b) shows that it is also licensed by sì. Crucially, as shown in (14c), one of the coordinators has to be present; VP-ellipsis is not possible in cases where there is no coordinator at all. Again, this shows quite clearly that, for syntactic purposes, sì counts as a proper coordinator.³

Against the background of all of these arguments, we would like to put forward that configurations involving sì should be treated like instances of proper coordination and that sì is a proper clausal coordinator with a somewhat unusual distribution.

In the course of the next sections, we will see some more arguments that sì differs from adverbs in that it has a fundamentally different distribution as well as a fundamentally different behavior regarding scope. This serves as additional arguments that the notion of sì as an adverb is empirically not adequate. In the next subsection, we will turn to discussing the distribution of sì in simple clauses.

2.3 The distribution of sì

In this section, we will describe the general distribution of sì. As noted above, sì is, in simple clauses, located somewhere between the subject and the verb. Accordingly, we find two statements about the actual position of the element in question. Awobuluyi (1977: 69) calls sì a preverbal modifier whereas Givón (2018: 187) calls the position a “post-subject position”. And while these statements are clearly not wrong, we can refine them and be more precise if we employ sì in slightly more complex configurations. For example, we can see that sì not only follows the subject but rather also follows a number of modal and temporal elements (15) as well as negation (16):⁴

On the other hand, we see that sì also does not appear to be immediately preverbal. Preverbal adverbs can intervene between sì and the verb. Similarly, Yorùbá allows for two kinds of preverbal PPs both of which intervene between sì and the verb.

In (18), the benefactive PP bá mi (‘for me’), must appear in between sì and the verb.

Finally, we also note that the aspectual marker indicating perfect appears after sì and before the verb.

For now, we can thus summarize the position of sì within the second conjunct of clausal coordination schematically as follows:

We should note that there is very little flexibility in this ordering with respect to sì. sì can never precede the subject or tense, negation or modal elements even if there is other preverbal element it could right attach to. In (21), we see the coordination of two conditional clauses and again, the position of sì is fixed to the position to the right of the modal element bá glossed as may. It cannot appear before the subject (21b) even though there is a clause-initial complementizer tí it could potentially lean on. It also cannot occur between the subject and the modal. It must appear in its designated position below the modal and the verb.

Similarly, we note that the position of sì with respect to the adverbs is fixed. Non of the above-mentioned adverbs tun or jàjà can precede sì.

And, finally, in (23) we see that the benefactive PP bá mi (‘for me’) cannot precede sì either.

So, given that sì has a fixed position in the clausal spine of the Yorùbá clause (see schema in (20)), we could hypothesize that there is a fixed clausal projection that must host sì. And, as we will see in Section 4, this actually has been suggested by Ilori (2010), who assumes that sì obligatorily adjoins to VP/vP. And even though such an account would certainly need some additional assumptions to capture the entire picture (e.g. we would need to explain the fixed order of sì and other vP/VP-adverbs as well as the perfect marker tí), such an account according to which sì occupies a fixed, yet to be defined position in the clause, certainly seems promising at this point. However, as we will see, the data points in the next section do not fit particularly well with this picture.

3.1 Adverbial clause constructions

In the preceding section, we have seen that the alleged coordinator sì in Yorùbá seems to occupy a somewhat unusual but ultimately well-defined position in the second one of its conjuncts: It seems to occupy a certain middle-field position somewhere below the higher verbal categories such as tense, negation and modals but above perfect aspect and vP/VP-adjuncts. In this brief subsection, we will coordinate somewhat more complex constructions involving clauses containing a clause-initial adverbial clause. Abstractly, the pattern is as in (24). We conjoin two clauses C1 and C2, both of which contain a clause-initial adverbial clause but which otherwise follow the standard syntax of Yorùbá.

All things being equal, we would expect that the conjunction sì should appear in between the subject and the verb of the second conjunct (C2). This, however, is not what we find. Consider (25), which illustrates an actual example of the abstract structure in (24). The two clauses conjoined translate to If the sun shines, Ade will go to Lagos. (C1) and If it rains, Olú will go to Ibàdàn. (C2).

What we would have expected is to find sì after the phrase Olú yóò (‘Olú will’) and before the phrase lọ sí Ìbàdàn (‘go to Ìbàdàn’) as this is the vP/VP-edge of the second conjunct. What we find instead is that sì is found inside the embedded clause-initial conditional clause. Attempting to place sì in the expected position results in a significantly degraded sentence:

This is a puzzling fact from a syntactic point of view because the presence or absence of a clause-initial adverbial clause should not make a difference for the purposes of sì-placement if sì simply were a grammatical particle that is restricted to a given syntactic position in the second conjunct. Even more striking perhaps, the adjunct clause should not be transparent for sì-placement itself because it is generally taken to be an island that is opaque to syntactic processes altogether.⁵

We will provide a full explanation for this curious placement pattern in Section 5. For now, we want to restrict ourselves to two comments about what these examples show us. First, we interpret these facts such that approaches which seem to explain the placement of sì inside the second conjunct by purely syntactic means are either doomed to fail or they at least must make some very unusual assumptions about the syntactic structure of adverbial clauses and their syntactic status as islands. The second comment concerns the placement of sì inside the adverbial clause. Just taking the position inside the adverbial clause, we find that sì appears again at what could be described as a VP/vP-edge position. This seems to indicate that sì seems to pick out the leftmost VP/vP-edge of the second conjunct regardless of syntactic embedding or constituency. Intuitively, we take this to mean that even though sì can never appear in the position in between the two conjuncts, it still has some abstract connection to that position and does not want to be pronounced “too far away” from that position. This intuition will reappear in the discussion about relative clauses and focus constructions below.

3.2 Relative clause constructions

In Section 2 we have seen that, in standard clauses, sì appears at a specific position inside the second conjunct. Ilori (2010) described this position as the VP/vP-edge and, for simple cases, this description was indeed sufficiently accurate. In the preceding subsection about configurations involving adverbial clauses, we saw however, that the placement of sì seems to be somewhat oblivious to syntactic constituency and islandhood in that it seems to pick the leftmost VP/vP-edge in the linear order irrespective of whether that edge belongs to the matrix clause or an adjunct. In this short subsection, we look at configurations where the subject of the second conjunct is modified by a relative clause.

Relative clauses in Yorùbá appear in postnominal position and are introduced by the multifunctional complementizer tí, which we already saw heading conditional clauses in the previous section. The relative clause itself contains a gap in the position where the head noun is interpreted unless that position is located inside a syntactic island; in which case, speakers use a resumptive pronoun inside the relative clause. In (27a), we see relativization from a direct object position, which results in a gap (here indicated with a t). In (27b), we see relativization from a possessor of a direct object, which requires the use of the resumptive pronoun rẹ́.

Given the discussion in the previous sections, the question that arises is of course, what happens if the subject of the second conjunct is modified by a relative clause. Again, we can abstractly schematize the structure we want to test as in (28):

Against the background of the previous sections, we can ask ourselves whether sì attaches at the vP/VP-edge of the matrix clause or of the relative clause. The answer is, in this case, that sì has two options. Both positions are acceptable.

In (29a), sì is located in between the subject and the verb of the relative clause modifying the subject of the second conjunct (‘who Adé saw yesterday’). In (29b), sì is located in between the complex subject (i.e. after the postnominal relative clause) and the perfect marker ti.

Again, we can briefly highlight that the availability of (29a) seems problematic for an account which seeks to explain the placement of the coordinator simply by means of syntactic constituency. Syntacto-semantically, the coordinator should take scope over the entire second conjunct and therefore, it should not be able to appear inside the relative clause. The availability of (29b), which is, from a syntactic (i.e. Ilori 2010’s) point of view, presumably the expected position of sì sets relative clauses apart from conditional clauses.⁶

3.3 Focus constructions

The last of the more complex configurations we want to discuss concerns the so-called focus construction. Focus in Yorùbá is expressed by having the focussed item in a sentence-initial position followed by the element ni. With object focus, the object position is simply a gap:

Subject focus on the other hand in Yorùbá involves the use of a resumptive pronoun in subject position:

As in the other cases, we can now conjoin two clauses involving focus fronting as abstractly schematized for object focus in (32).

Somewhat surprisingly, the coordinator sì will then surface between the focussed XP and the focus marker ni:

Having sì in the standard position between the subject and the verb of the second conjunct is very much dispreferred:

Again, this presents a straightforward argument against sì begin a simple VP-adverb. Regular VP-adverbs like the ones we have seen above, are not licensed in such a position:⁷

The fact that sì surfaces in between the focussed XP and the focus marker ni clearly sets it apart from adverbs and other elements in Yorùbá. But in order to fully evaluate the consequences for the analysis at hand, we need to take a small detour to discuss the syntactic properties of Yorùbá focus constructions.

Focus constructions across Niger-Congo and West African languages in general are the subject of a long-standing debate. One of the core questions about these constructions are whether they involve biclausal, cleft-like structures or whether they are monoclausal. For Yorùbá, as far as we know, the existing analyses Awobuluyi (2008); Ilori (2010); Aremu (2021) uniformly point to monoclausal structures where ni realizes a specific functional projection in the left periphery.

Bisang & Sonaiya (2000) have argued that the focus particle ni is identical with one of the copulas in Yorùbá, which might be interpreted to pointing towards a biclausal cleft-like structure. The example in (36) illustrates the similarity between copula-like uses of ni and its use in the focus construction above:

However, despite this superficial similarity, several facts point to the conclusion that what we are dealing with is in fact not a cleft-structure but indeed a monoclausal structure.

The first argument is that biclausal cleft-like structures like ‘It was clothes that I bought’ usually employ relative clause constructions and the case in (36) does not. As we have seen in Section 3.2, relative clauses in Yorùbá are introduced by the complementizer tí, which is absent in examples like (36).

The second argument comes from the fact that if ni in examples like (36) were a real copula, then we would expect it to behave like a regular verb in that it could be modified with adverbs. This however, is not possible. Example (37) (repeated from (35), above) below shows that it is not possible to modify the alleged copula with adverbs.

The third argument is that it is not possible to have ni supplemented with future tense or perfect aspect morphology, which we would expect if ni were a real copula (38a,b).

Finally, we note that there is a proper copula jẹ́ in Yorùbá, which retains the canonical SVO word order of the language, as shown in (39a). This contrasts with the alleged copula ni (39b) that always require the fronting of the object.⁸

Unlike ni, jẹ́ can occur with these future tense or perfect aspect morpheme indicating that it is actually a regular verb. As the examples in (40) show, it also patterns like other verbs with respect to its placement relative to these markers as well as adverbs.

Note also, that examples like (39b) optionally allow for jẹ́ to appear in its expected position indicating that alleged cases of predication with the particle ni are merely instances of focus out of a predicative clause with a covert copula.

For all of these reasons, we concur with existing analyses in Awobuluyi (2008); Ilori (2010); Aremu (2021) that the focus particle ni is the realization of a functional projection in the left periphery.

What this means for our purposes here is that if sì precedes the focus marker and the focus marker is a functional head in the left periphery, then it is not hard to see why an analysis of sì as a vP/VP-adverb is hard to maintain. Consider the following tree depicting an object focus derivation taken from Ilori (2010):

If a tree like (42) is intended to cover the second conjunct of (33), repeated below for simplicity, then it is clear that sì cannot be a VP-adverb as Ilori (2010) claims it to be. If it were a VP-adverb, we would incorrectly expect it to appear in the position between the subject and the verb (43b).

In fact, we would like to submit that what these examples suggest is that not only is an analysis in terms of sì being a VP-adverb empirically inadequate but also that it is not straightforwardly possible to define a uniform syntactic position of sì altogether. sì sometimes appears in the left periphery and sometimes it appears as low as a VP-adverb. Any attempt to describe the placement of sì in such cases would either make some adhoc construction-specific assumptions where sì goes in focus and non-focus constructions or it would need to stipulate several different elements with the same form and the same function, which only differ in their placement properties. We will discuss this observation in more detail below.

5.1 Describing the pattern

One of the underlying intuitions that popped up throughout the discussion in Section 3 about the clause-initial adverbial clauses and the relative clauses was that sì seems to attach to what looks like the first vP/VP-edge it can find from left to right. In simple clauses, this inevitably will be the vP/VP-edge of the matrix clause but in more complex constructions where there is an embedded clause to the left of the matrix vP/VP-edge, then the vP/VP-edge of the embedded clause will be chosen. This intuitive analysis raises two questions: (i) What is so special about the position of the vP/VP-edge? and (ii) What about the position in Section 3.3 between the focussed element and the focus marker? As it turns out, the answer to both questions is the same one. We would like to defend that the following distributional statement about sì holds:

In the rest of this subsection, we will go through the crucial derivations to show that the rule is empirically adequate. We can begin with some simple clauses and for now we simply submit that Yorùbá simple clauses consist of two prosodic domains that coincide with the domains we saw were relevant for the placement of sì in Section 2:¹⁰

Of particular importance is the fact that the first domain contains complementizers the subject and the higher verbal projections.¹¹ At this point, this assumption is admittedly not well supported as there is, unfortunately, not a lot of work on Yorùbá prosody that we can build on. We will look at a number of phonological and morphophonological processes in Section 6 and we will argue that their patterns of application and under-application indeed suggest that the prosodic domains are such that they fit with our assumption above. In addition, we want to note that we think that this sort of prosodic bipartition is not completely adhoc. Prosodic literature on better studied languages has often noted that a prosodic bipartition along the VP/vP-edge (or more specifically some XP-edge including aspect but excluding higher verbal projection such as tense) seems to be a frequently attested pattern crosslinguistically (see e.g. Kahnemuyipour 2009).

So, if we take this assumption at face value, virtually all of the patterns fall out without further ado. The following examples indicate some of the second conjuncts of examples above with the prosodic phrasing we assume.

Similarly, the more complex constructions discussed in Section 3, can receive a similar representation. As for clause-initial conditional clauses, we will assume that they are prosodically integrated and their prosodic phrasing is therefore accessible for sì-placement. Again the pattern is the same, sì will simply attach right after the first prosodic phrase.

With relative clauses, we saw actually two patterns in Section 3. In one of the patterns, we do see a similar effect as with the adverbial clauses in that the linear order simply trumps the syntactic embedding and again, the coordinator will attach after the first prosodic phrase.

The second placement pattern with relative clauses on the other hand did have sì in its expected position at the vP/VP-edge of the matrix clause. We believe that this kind of optionality is actually due to optionality in prosodic phrasing. It has been known for languages like German that relative clauses can be prosodically integrated or unintegrated (see e.g. Holler 2005; Truckenbrodt 2005; Kaland & van Heuven. 2010; Féry 2017) and while this difference often coincides with the status of the relative clause as restrictive or non-restrictive, it does not necessarily have to all the time (see Kaland & van Heuven. 2010; Weisser 2022). We will thus hypothesize that the pattern in which the coordinator sì appears in the matrix clause is due to a non-integrated relative clause modifying the subject. By assumption, non-integrated relative clauses indicated by angled brackets are not accessible for sì-placement.

The final pattern that we need to take a look at are the so-called focus constructions. We saw that with focus, the coordinator does not appear in its expected position in the clausal middle field but rather in between the focussed XP and the focus marker. We believe that this is due to the fact that focussed XPs are prosodically prominent because they receive their own prosodic phrase. This explains why they serve as a host for the coordinator sì:

Again, we want to refer the reader to Section 6 for an argument that the prosodic phrasing in (56) is indeed correct. In particular, we will show that the focus marker is indeed prosodified with the following subject.

This concludes the discussion of the individual patterns. We hope to have shown that a plausible prosodic phrasing pattern can describe the placement of sì in all cases. Admittedly, for now, the prosodic phrasing we assumed to derive the pattern, while plausible from a crosslinguistic perspective, lacks concrete evidence from a language-internal perspective. As noted above, we will provide evidence for the prosodic phrasing in Section 6 below. Before we do that, we want to briefly discuss some technical assumptions of the model we entertain to derive the placement generalization about sì attaching to the first prosodic phrase of the second conjunct.

5.2 The displacement rule

In the previous section, we have shown that a plausible prosodic phrasing pattern for Yorùbá actually allows us to maintain the following placement generalization about the distribution of sì (repeated from (48) above):

In this section, we will discuss how we think this pattern should be derived. In order to do this we will briefly refer to similar cases of morpheme placement being sensitive to prosodic phrasing that have been noted in the literature but for a more comprehensive picture of crosslinguistic evidence, the reader is referred to Section 7.

In general, the fact that small grammatical elements are placed in what can be described as a second position of some sort is crosslinguistically speaking not surprising. Second-position clitics have been reported in countless languages from all continents. Following literature such as Marantz (1988); Bošković (2001); Embick & Noyer (2001); Anderson (2005), researchers have started looking at how to define this second position and as it turns out, the term second-position is not a coherent, uniform notion crosslinguistically. In some cases, the second position clitic attaches to one syntactic constituent and in other cases, second position means that the second position clitic attaches to the first prosodic word. More recently, people also found a number of cases which look more like the case at hand, namely where the second-position clitic attaches to an entire prosodic phrase (see e.g. Inkelas & Zec 1990; Chung 2003; Herd 2003; Legate 2008; Dawson 2017; Weisser 2022; Belkind 2023) rather than to a prosodic word or a syntactic phrase.

The placement of such clitics, whose position is defined with respect to phonological or prosodic constituents, is typically derived by means of some dislocation mechanism that displaces the element in question one position to the right from the perspective of its syntactic position (see amongst many others Halpern (1995); Embick & Noyer (2001); Chung (2003); Legate (2008) and many others. Applied to the case study at hand, this gives us a straightforward way to derive why the Yorùbá coordinator sì shows up in the place that it does. Maintaining the assumption that it is a proper coordinator for all syntactic and semantic purposes, it makes sense that it is syntactically base-generated in the position in between the two conjuncts. And, from that position, the prosodic dislocation rule applies, displacing it exactly one prosodic phrase to the right. This is illustrated below. (58) shows the syntactic base-position of the coordinator in between the two conjuncts. We saw in Section 2 that sì is only available for clausal coordination and if we assume that the syntactic base-position of sì is in between the two conjuncts, we can model this as simple syntactic selection.

Further we assume that this structure is then sent to the interfaces LF and PF. On the LF side, this predicts that, for the purposes of interpretation, the coordinator still takes scope in the usual position in between the two conjuncts. We have seen that that prediction is borne out because even in cases where sì is deeply embedded inside an adverbial clause inside B, it will still always express coordination between the two conjuncts A and B. On the PF side, this structure will then be linearized and mapped onto prosodic categories. We can abstractly represent this as follows with a_x and b_x referring to placeholders for prosodic subconstituents of A and B, respectively:

Based on this kind of prosodified structure, the clitic displacement rule of sì can apply. It will dislocate sì exactly one prosodic phrase to the right.

For the sake of concreteness, we can assume that it will be integrated into an adjoined position in the prosodic phrase preceding it (see Ito & Mester 2006; 2009; 2012; Elfner 2012; Bennett et al. 2016) leading to a recursive prosodic phrase.¹²

This derivation predicts that the placement of sì should not have any bearing on syntactic processes inside the clause it finds itself in; in other words, there cannot be any morphosyntactic processes inside the second conjunct that are sensitive to the presence of sì. For all we know, this prediction is borne out.

So to sum up, in this section, we have proposed a new generalization to capture the placement of sì. In previous sections, we had established that syntactic rules are not particularly adequate to capture the distribution as the pattern seems to ignore syntactic islands and, to a certain extent, scope and constituency. Building on that insight, we proposed in this section, that the correct distributional generalization is one in terms of prosodic phrasing: sì consistently appears after the first prosodic phrase of the second conjunct. We then went on to indicate how we think the prosodic phrasing is in the crucial examples from Sections 2 and 3. The last section briefly discussed the displacement rule from a more abstract perspective.

In the next section, we will take a look at a number of phonological processes that span over individual words and particles. The goal of this section is to find independent evidence for the prosodic phrasing we assumed above.

6.1 Fusion

We begin with instances of phonological fusion. As noted in a number of works (see e.g. Akinlabi 1985; Pulleyblank 1986; Akinlabi & Liberman 2001), there are quite a few environments in Yorùbá, when two adjacent elements are fused.

Verb-Object: 

One of the most productive cases of fusion comes from verb-object combinations, which are frequently fused when the verb ends in a vowel and the object begins with one. Akinlabi & Liberman (2001) give quite some examples involving different types of vowel and tone combinations including the following ones:

As we can see, in some cases, the vowel quality of the verb is preserved (64c) and in some cases, the vowel quality of the object (64a,b,d,e). Similarly, we see that the tones of the verb is sometimes preserved (64a,b,c,d) and sometimes the tone of the first syllable of the object is preserved (64e). Note that these properties do not necessarily have to go together indicating that the two segments have indeed fused. Clearly we cannot do justice to the process of vowel fusion in Yorùbá in the present section but we want to note that this indicates that the verb and the object are phonologically fused in the same domain.

comp-Subject:

Another configuration where we see fusion of two grammatical elements is with combinations of various question-introducing complementizers in combination with a subject pronoun in third person singular.

In addition to the data in (65) listed in Akinlabi & Liberman (2001) we note that this is not only found in polar questions but also with a proper complementizer and a subject which undergo fusion.

foc-Subject:

It might be a surprising assumption that the focus marker ni is not prosodified together with the focussed XP but rather with the following subject. However, this assumption is supported by the fact that the focus marker and an adjacent 2sg or 3sg subject pronoun is contracted the portmenteau forms lo or ló is used:

Subject-neg:

The final case of fusion or contraction applies with third singular pronouns and negation. Consider the minimal pairs in second singular and third singular in the affirmative and the negative:¹³

6.2 Association of floating tones

Subject-tense: 

We have already seen in the introduction to this paper that the non-future tense morpheme, the so-called high tone syllable (hts) acts like a phrasal affix attaching to the subject position. As the name of the suffix already suggests, non-future tense is expressed merely with a floating high tone.

This suggests that the subject and the tense element are in the same prosodic domain.

6.3 OCP-Effects

Next, we move on to so-called OCP-effects as discussed by Akinlabi & Liberman (2001). As these authors observe, tonal combinations are in some configurations restricted in the sense that the same tone cannot occur on two adjacent syllables. As a result, one of the tones will be deleted. Akinlabi & Liberman (2001) liken the process to so-called effects of the Obligatory Contour Principle (OCP).

Verb-Object: 

The crucial configuration we want to look at involve verbs and pronominal objects. Usually object pronouns all come with a high tone as illustrated in the left and the middle columns in (73) where they appear after a low tone or a mid tone verb. In the rightmost column, where they appear after a high-tone verb, the high tone on the object pronoun is deleted.¹⁴

We take this as another piece of evidence that verbs and object pronouns form a close morphophonological complex.

6.4 Assimilation

Comp-Subject: 

We have seen that the third person pronoun fuses with the final vowel of preverbal particles and complementizers. The first and the second person pronouns do not fuse but still the final vowel of the particles assimilates to the vowel of the pronoun:

In (74), the contexts with a first person plural pronoun a, we see that the particles all end in a, and in (75), a second person singular context where the pronoun is o, we find that they all end in o.

Neg-Modal:

Negation in Yorùbá shows a number of allomorphs and one alternation that is particularly insightful for our purposes at hand is the alternation between kò and kì. According to (Ilori 2010: 235f), kò, the default choice, changes to kì when it precedes the habitual modal í (75b). When it precedes another morpheme ń expressing habitual actions, it does not change (75c).

According to Ilori (2010), the change from kò to kì is the result of regressive assimilation of the vowel /i/ from the habitual modal to the negation. This, we take it, suggests that negation and the habitual modal are in the same prosodic domain.

6.5 Allomorphy

Subject-neg: 

The final process that we briefly want to mention is that of allomorphy between the first person subject pronoun and the negation. The first person singular morpheme mo changes to mi when preceding the negation:

Again, we take this to indicate that the subject position and the negation are part of the same morphophonological domain.¹⁵

6.6 Summary

In the previous section, we have discussed some of the phonological processes that could give us a hint about the notion of phonological domains. The illustration in (78) summarizes the processes and the configurations they apply in.

We have seen that the preverbal particles such as question markers and the focus marker ni shows interdependence with the subject position. The subject position is also closely related to the tense position and the position of negation. Negation in turn is dependent on the presence or absence of the habitual modal í.

Finally, we have seen that the verb and the object position are closely related. They are fused in some cases and also, we see the application of the tonal OCP deleting the tone of the object depending on the tone of the verb.

All in all, we see a lot of interdependencies between various elements in the clause but crucially for the claims put forward in this paper, we do not see any interdependencies across the domains we assumed for the purposes of sì-placement. There is no interdependency between the subject and the verb or between negation or a modal and the perfect marker or the verb.¹⁶

The absence of any interactions between the focussed XP and the focus marker as well as the fact that focussed pronouns must be expressed with the independent, bisyllabic form because the have no host to cliticize to strongly suggests that the focussed XP is indeed, as we predict, prosodified in its own domain. Similarly, the absence of interactions between the higher domain containing subject, tense, negation, etc and the lower domain containing aspect, the verb and the object, similarly suggests, in our view that these are prosodified in two different domains.

We take this as strong evidence that the phonological and prosodic domains are indeed along the lines of what we assumed for sì-placement. In the next section, we show that even though the Yorùbá pattern of a clause-internal conjunction might seem surprising at first sight, it patterns nicely with other instances of clause-internal coordinators and with typologies of clitic placement more generally.