In this section, the distribution and properties of Bùlì ká are compared with the Q-particles of other languages including sá in Tlingit (Cable 2007; 2010), da in Sinhala, and ka in Japanese (Hagstrom 2004; Kishimoto 2005) leading to the conclusion that ká is Q in Bùlì. As will be illustrated below, ká only attaches to nominals, and may not separate a DP from a functional head that selects it. The relevance of this particle to the discussion in this paper is seen in section 4 where I argue that the ká headed phrase undergoes covert movement.

Before considering the distribution of ká, a quick note on the claim that it attaches to nominals. This immediately raises a question of what the categories of wh-phrases are in the language, especially “where”, “why”, and “when”. Their composition suggests they are all nominals. As shown in (8), there are nominals corresponding to “direction”, “body”, and “day” in the meaning equivalent of bɛ̄-gēŋ ‘where’, b^wān-ɲíŋ ‘why’ and dā-dīnā ‘when’ respectively.4 As will be shown later in the discussion, nominal constructions like “relative” clauses, and “because” clauses permit ká to appear at their edge while non-nominal clauses including “if-clauses” and “without” clauses do not.

Returning to the distribution of ká, like Q in Tlingit and Sinhala, it is obligatory in a wh-question, as seen in (8).

Secondly, the particle ká must c-command a wh-phrase. It can be directly prefixed to the wh-word, as the above examples show, or it can be attached to a larger wh-containing phrase, as the (a) examples in (9)–(11) show. If the Q doesn’t precede the wh-phrase, the constructions are ill-formed.

One important property of Q-particles observed in many languages is that, although a wh-phrase is allowed inside a complex-DP, the Q-particle must appear at the edge of the island (Hagstrom 1998; Kishimoto 2005; Cable 2007; 2010). This same property can be observed for Bùlì. Ká cannot appear inside but must appear at the edge of the complex DP.

I argue, based on the distributional similarities of ká to the Q-particles da in Sinhala, ka in Japanese and sá in Tlingit, that ká is Q in Bùlì, and, as such, it should be given an analysis parallel to the Q-particles in these languages.

Despite these similarities, there are some points of divergence among the languages in terms of distribution and function of their Q-particles. The Q in Bùlì only attaches to nominals, and, unlike its counterpart in Tlingit, it does not also appear on wh-words to mark indefinites. It can also be used in declarative sentences to mark focus. The selectional requirement of Q thus independently rules out (12b), where Q is attached to the relative clause, and explains why the Q-particle cannot appear at the left-edge of the matrix clause, as in (13b), or subordinate clause, as in (13c)–(13d), unlike in Tlingit where Q can mark subordinate clauses.

Another point of divergence is that, in Bùlì, there can be at most one instance of ká per question, even in multiple wh-questions, (14), (again unlike Tlingit, which can have multiple Q-particles in multiple wh-questions).

Cable (2007; 2010) draws a distinction between Q-projection languages (15) and Q-adjoining languages (16). Only in a Q-projection language will movement triggered by a Q probe yield phrasal movement of the type usually called wh-movement.

The Q-particle is barred from appearing in certain environments in Q-projection languages: (i) between a wh-possessor and the possessed NP, (ii) between a wh-determiner and its NP complement or (iii) between a pre or postposition and its complement. To the extent that these constructions are replicable in Bùlì, we should observe similar restrictions on the distribution of ká if the language is a Q-projecting language. As will be shown below, similar restrictions are observed for the distribution of ká.

In possessive constructions, ká cannot appear between a possessor NP and a possessed wh-phrase (17).

It cannot appear between a wh-determiner and its NP complement (18).

It cannot also appear between a preposition and its complement (19b), nor can it precede the preposition (19c). Example (19c) contrasts with the distribution of Q in Tlingit, which allows Q to take a preposition as its complement.

The data from (17)–(19) show that ká is restricted from appearing in the structural positions that the Q in Q-projecting languages are also banned from occurring. This is thus evidence that the language is a Q-projection language.

As the discussion shows, ká must precede the wh-phrase, it cannot occur inside an island, and it may not separate a DP from a functional head that selects it. We also saw that ká heads its own projection i.e it can either directly take a wh-phrase or a wh-containing phrase as its complement (Q-projection language). Since ká has the same range of restrictions as Q, I conclude that ká is Q. This serves as the background for next section where I argue that ká diagnoses covert movement.

In this section, we look at the wh-in situ immediately preceded by ká which I argue undergoes covert movement. Does this phrase show the properties that are diagnostics of movement? I present evidence from islands, binding theory, the absence of intervention effects, and, arguably, scope that answers this question in the affirmative. I show that this covert movement is comparable in many ways to the overt movement of languages like English.

Studies over the years have converged on several properties for diagnosing whether a syntactic operation involves movement. A stable diagnostic for movement is island-sensitivity. If the ká-phrase involves any kind of movement, then we expect the ká-phrase to show island sensitivity. This expectation is borne out, as shown earlier in (12), repeated as (20), and (21). Although wh-phrases are allowed inside an island, attaching the particle ká to the wh-phrase inside the island renders the construction ungrammatical (20a). It must appear at the edge of the island (20b).

When a wh-phrase is coordinated with another element, ká must precede the entire conjunction phrase, (21b). Placing it immediately before the wh-phrase in the second conjunct, (21a), results in ungrammaticality. This is consistent with earlier studies which show that covert movement obeys the CSC (Ruys 1992).

An adjunct “because” clause behaves like a nominal: the because has the final particle la: a determiner-like particle found with relative clauses and ɲīŋ ‘body’ which is the same form we find with the wh-word “why” in the language. All these factors suggest that it is nominal. As a result ká cannot appear inside, (22b), but may appear at its edge, (22c).

Also, ká cannot appear inside an “If-clause”, (23), or “Without-clause”, (24). It cannot appear at the edge of these clauses because of the subcategorization restriction that it appears with nominals.

Assuming that complex DP phrases, coordinate structures, conditionals and adjunct clauses are islands for movement, these observations support the claim that there is covert movement involved in these constructions.

A second piece of evidence for covert movement comes from Binding Theory. Nissenbaum (2000) shows that covert wh-movement feeds condition A of the binding theory. Consider the data in (25) from Nissenbaum (2000). The covert phrasal movement of the in-situ wh-phrase in (25a) licenses the anaphor. In contrast, since there is no such movement in (25b), the anaphor is not licensed resulting in the ungrammaticality of the sentence.

In-situ wh-phrases in Bùlì can feed condition A of the binding theory. Consider the examples in (26). The anaphor needs an antecedent within the same clause in order to be licensed, (26a)–(26b). When the antecedent fails to c-command the anaphor, the result is ungrammatical, (26c).

The examples in (27) replicates the English paradigm from Nissenbaum given in (25), though the judgments in Bùlì are perhaps stronger. The ungrammaticality of (27b) is expected because the antecedent of the anaphor is found in the matrix clause while the anaphor is in the embedded clause. The grammaticality of (27a) is surprising since the antecedent and the anaphor are in different clauses. The contrast between (27a) and (27b) is expected if covert movement is involved—and importantly successive cyclicity in (27a). The covert movement of the ká-phrase into the matrix verb will license the anaphor. The example in (27b) is ruled out because there is no such movement. It will later be shown that we lose the Nissenbaum effects in instances where the in-situ wh-phrase does not move covertly.

In the preceding section, we looked at the first kind of wh-in-situ phrase, the one preceded by ká, and I argued that the ká-phrase undergoes covert movement. In this section, we take a look at the second kind of wh-in-situ, the one without ká. At this point, there are two possibilities that might, in theory, be attested with the ká-less in situ phrase: either (a) they pattern like in situ wh-phrases with ká and show signs of undergoing covert movement, or (b) they do not show such signs, and therefore could be concluded not to undergo covert movement. The evidence leads to the conclusion that (b) is correct.

Ká-less wh-phrases occur in multiple questions. In multiple wh-questions, ká appears to the left of the highest wh-containing phrase, (40).

Having multiple ká, (41a), or placing it on the lowest wh-phrase, (41b), results in ungrammaticality. An open question for the current account is why there is one ká which appears on the highest wh-phrase.5

An initial point that suggests that a ká-less in situ wh-phrase doesn’t undergo covert movement is that it can occur inside an island, (42). As has been noted above, the in-situ phrase with ká is island sensitive since movement out of islands is generally barred. The fact that the ká-less phrase is allowed inside an island could be taken as evidence that it doesn’t undergo covert movement.

Another suggestive piece of evidence tha ká-less phrase does not undergo movement comes from binding theory. Relying on a previous observation that covert movement can feed condition A of the binding theory, we noted that the in-situ phrase with ká, can feed condition A of the binding theory. A related question is: can the wh-phrase without ká feed condition A of the binding theory? It should be able to feed the binding theory if it involves movement just like its counterpart with ká.

Multiple questions are possible in embedded clauses, (43). In (43b), there is only one wh-phrase ká wà-dēk_i foto kuna ‘which picture of herself’ in the embedded clause. Because this wh-phrase is preceded by the Q-particle ká, it moves covertly into the matrix clause, thereby licensing the anaphor. Example (43c), on the other hand, involves a multiple question. In this construction, the Q-particle ká is on the wh-phrase ká núr bānā ‘which people’, and it is that phrase which moves covertly. The anaphor in the second wh-containing phrase is not licensed in this context. We can explain this by arguing that the inability of an antecedent in a matrix clause to license an anaphor in an embedded wh-phrase without ká is because it does not involve movement, hence the ungrammaticality of the sentence.

Another argument for the non-movement of the ká-less wh-phrase comes from intervention. We saw that placing an intervener above a wh-phrase preceded by ká doesn’t trigger intervention. In multiple wh-contexts where the second wh-phrase is not preceded by ká and doesn’t move, we might predict intervention effects. This is indeed the case, as shown in (44) below.

Although the ká-phrase can move across the intervener in (44a), the second wh-phrase will not, resulting in an intervention effect. In example (44b) where negation is below Wh₁ and above Wh₂, the result is ungrammatical, as an instance of an intervention effects. Example (44c) shows that if all the wh-phrases are above the intervener we do not get intervention effects. Note that this extends to other interveners including “only” and “also”.

We have seen in this section that there are basically two kinds of wh-in-situ in the language: the one preceded by ká and the one without a ká. As argued in the previous section, drawing from different kinds of evidence, the in-situ wh-phrase preceded by ká is the only one that undergoes covert movement. The ká-less one, on the other hand, does not undergo any kind of movement. This is reminiscent of wh-in situ in Vietnamese (Bruening & Tran 2006) where it has been argued that wh-in situ phrases can be interpreted either by covert movement or by unselective binding without movement. In Vietnamese, however, the covert movement strategy occurs in matrix questions when there is no question particle, while unselective binding (lack of movement) is employed when there is a question particle and in embedded questions generally.

In this section, we turn to constructions like (45), repeated from (4), where the ká-phrase is found at the left-periphery of the clause. I show that, as far as wh-movement is concerned, Bùlì is a language where both overt and covert movement behave nearly identically except for the phonology—with both interacting with Islands, intervention effects, and binding theory in perfectly identical ways, once extraneous factors like linear precedence are taken into account.

In these constructions, as noted earlier, the wh-phrase is obligatorily followed by ātì if the wh-phrase is a non subject, (45a)–(45b), and ālì if it is a local subject.

I follow Ferreira & Ko (2003) and Hiraiwa (2003; 2005) in assuming that these constructions involve overt movement of the goal to the Spec of the probe, ātì, thus making overt movement optional in the language. The fact that the gap can be found several clauses away and is subject to islands is evidence that this construction involves movement.

The gap can be found several clauses away. The wh-phrases originate from within the embedded clauses in (46).

Like the in situ construction discussed above, the left-peripheral Q behaves like movement, sensitive to familiar islands as seen in (47)–(51).

Another piece of evidence for overt movement comes from binding theory and reconstruction. Wh-movement does not obviate binding relations from the position from which movement took place. Consider the following examples in English. An R-expression cannot co-refer with a c-commanding pronoun, (52a), so John in the object DP cannot co-refer with the subject pronoun. When we move the object containing the R-expression John to a position where the subject no longer c-commands it, it still does not make the construction better, (52b). This follows from the observation that wh-movement does not obviate binding possibilities (in this case principle C violations). This observation is easily explained under the copy theory of movement (Chomsky 1995) where a moved element leaves behind a copy of itself, rather than a trace.

Reconstruction effects are observed for the wh-phrase in the left periphery. As expected, in (53b), moving the wh-phrase to the left does not bleed the satisfaction of principle A.

It is important to note, however, that the antecedent for the anaphor must be a pronoun. The construction becomes bad if the subject is replaced by an R-expression, (54b).

A constraint like (55) seems to be responsible for the impossibility of demonstrating reconstruction effects with R-expressions.

Other data that support (55) independent of reconstruction are given below. In all the ungrammatical cases below, an R-expression follows a co-indexed pronoun or a reflexive. Even though the pronouns and anaphors do not c-command the R-expressions in (56), the sentence is still judged to be ungrammatical.6