2 Tagalog basics

Tagalog is a head initial and predicate initial language. It allows predicates of any category type, and word order following the predicate is generally flexible. Simple active clauses may be realized in one of two ways, depending on which of the verb’s arguments functions as the subject. In an actor-subject clause like (1), the external argument (i.e. the agent/experiencer) is the subject of the clause. In a theme-subject clause like (2), it is the verb’s internal argument (i.e. the theme) that is the subject of the clause.Footnote ^[3]

Full noun phrases as well as proper names are marked by a case particle that precedes the noun phrase. Tagalog contrasts three cases, which I refer to here as Genitive, Oblique, and Subject. The genitive case is an elsewhere case which marks the direct object of an actor-subject sentence, the agent argument of a theme-subject sentence, and possessors. The oblique case marks indirect objects in addition to certain types of direct objects (see Section 3.2). The subject case marks the subject of the clause.Footnote ^[4] Pronouns have distinctive case forms reflecting each of the three cases.

There is no definite or indefinite article in Tagalog corresponding to English the or a. Modulo the presence of an obligatory case marker, noun phrases in Tagalog are therefore often bare nominals which are morphosyntactically underspecified as to whether they are specific or non-specific in their interpretation. Nevertheless, there is good reason to believe that bare nominals can function either specifically or non-specifically. Consider, for instance, sentence (6) compared with the sentences in (7). The direct object in (6) is the complement of an intentional transitive verb. Characteristically, verbs of this type do not commit the speaker or the subject of the sentence to the existence of a referent for their object (Moltmann Reference Moltmann1997, Hallman Reference Hallman2004), as is evidenced by the felicitousness of the possible continuation of the sentences which explicitly denies the existence of a referent for the object. In the sentences in (7), by contrast, the direct object is the complement of an extentional verb, which generally does commit the speaker to the existence of a referent for the object in most contexts. In these examples, a continuation of the sentence that explicitly denies the existence of a referent for the object is infelicitous.

It is difficult to see how contrasts like this could be explained if bare noun phrase could only function either specifically with an intended referent in mind or non-specifically with no intended referent in mind.Footnote ^[5] Following von Heusinger (Reference von Heusinger2002, Reference von Heusinger and Reich2011a, Reference von Heusinger, Maienborn, von Heusinger and Portnerb), I will assume a general notion of specificity according to which a specific noun phrase is one that is referentially anchored to the speaker of the sentence or to some other referring expression in the sentence. This conception of specificity is similar to the characterization of specificity as ‘certainty of the speaker about the identity of the referent’, but it is also broader in that the specificity need not be linked to the speaker of the sentence but may instead be linked to other referents found in a sentence.Footnote ^[6] It is also general in that it covers the three different types of specificity (epistemic, partitive, and scopal) distinguished by Farkas (Reference Farkas, Nash and Tsoulas1994) (see von Heusinger (Reference von Heusinger2002, Reference von Heusinger and Reich2011a, Reference von Heusinger, Maienborn, von Heusinger and Portner2011b) for details).

Following much of the literature on specificity, I assume that noun phrases that are semantically definite are a subtype of specific noun phrases. Definite noun phrases, in addition to being referentially anchored to the speaker or other referring expression in the sentence, must be familiar in the discourse (Karttunen Reference Karttunen and McCawley1976, Kamp Reference Kamp, Groenendijk, Janssen and Martin1981, Heim Reference Heim1982) and referentially unique (Roberts Reference Roberts2003, Abbot Reference Abbot and Brown2006). Noun phrases can be indefinite but specific, as specificity does not require familiarity of a discourse referent to both Speaker and Addressee or even necessarily to the Speaker. Because Tagalog lacks a definite or indefinite article, it is unclear whether there is a grammatically based difference between definite noun phrases and specific indefinite noun phrases as there is for English and other languages.Footnote ^[7] If only for descriptive purposes, however, it is possible to draw this type of distinction by examining the function of a noun phrase in its discourse context. Consider, for instance, the sentences in (8) and (9). In both examples, the direct object occurs with a demonstrative ito (‘this’), although the object in (8) is arguably definite while the object in (9) might better be described as indefinite but specific. Sentence (8) occurs in the context in which several women who have been nominated for a comedy actress award are being interviewed, and this sentence is a quote from one of these women. In this context, then, the referent of the direct object (award na ito ‘this award’) is plainly both familiar and unique (the award has been previously mentioned, and there is only one award to be given out) and therefore semantically definite. By contrast, there is no previous mention of the object nitong dakilang pangitan (‘this great vision’) in the context in which sentence (9) is uttered. This noun phrase is therefore unfamiliar (hence, not definite), but presumably still specific in that the speaker clearly has a (unique) referent in mind.Footnote ^[8]

In what follows, I will use the term specific as a cover-term for noun phrases that can be characterized either as semantically definite or as indefinite but specific. Following Enç (Reference Enç1991: 11), I will assume that at least a subset of quantified noun phrases may also be characterized as specific (see Section 3.1.5 for discussion). In terms of semantic types, I assume that non-specific noun phrases are property denoting non-referential expressions of type ${<}$ e,t ${>}$ , while specific noun phrases that may be characterized as either definite or indefinite but specific are referring expressions of type ${<}$ e ${>}$ . Following standard assumptions, I take quantified noun phrases to be type ${<}<$ e,t ${>}$ ,t ${>}$ .Footnote ^[9]

3 Specificity and the morphosyntax of the theme

As mentioned in the introduction, much current work on Tagalog syntax asserts that if the theme argument of a transitive verb is specific, then it must be realized as the subject of a theme-subject sentence – or, equivalently, that the object of an actor-subject sentence must be non-specific. The existing literature has not been terribly precise about what counts as ‘specific’, but given the assumptions laid out at the end of the previous section, we can suppose that what previous works have had in mind is something like the restriction stated in (10).

The evidence that is cited for this restriction is typically based on the elicited interpretation of an actor-subject sentence containing two unmarked nominals compared with a theme-subject sentence containing the same two unmarked nominals. The reported fact, as exemplified by the translations for sentence pairs like (1) and (2), is that the theme argument is interpreted non-specifically in actor-subject sentences and specifically in theme-subject sentences. These previous studies, however, have not considered the type of contrast with bare nominals and intentional and extentional verbs documented above, nor have they gone much farther with the investigation to ask whether noun phrases that are explicitly marked to indicate specificity in some way may actually occur as the direct object of an actor-subject sentence. In other words, while previous works seem to tacitly assume a negative answer to this question, none to my knowledge have actually demonstrated that an unambiguously specific noun phrase (e.g. one containing a demonstrative, or some other indicator of specificity) cannot function as a direct object in an actor-subject sentence.

The main goal of this section, then, will be to broaden the empirical base surrounding the issue of the relationship between specificity and the morphosyntactic realization of theme arguments in transitive constructions. We will see, on the one hand, that specific direct objects are in fact attested in actor-subject sentences, contra the expectations of the restriction in (10). On the other hand, we will see that the semantic function of the theme does still play an important role in how theme arguments are realized, but in a more intricate way than previous studies have yet observed.Footnote ^[10]

4 Objects and object positions in Tagalog

The goal for the remainder of this paper will be to provide a formal account of the patterns just summarized. Our starting point will be the analysis of actor-subject and theme-subject sentences and their inter-relationship proposed in Rackowski (Reference Rackowski2002) and also Rackowski & Richards (Reference Rackowski and Richards2005). For Rackowski (Reference Rackowski2002: 81–84), clauses in Tagalog are headed by a inflectional head T(ense), which must Agree, in the sense of Chomsky (Reference Chomsky, Martin, Michaels and Uriagereka2000, Reference Chomsky and Kenstowicz2001), with a local DP – namely, whatever DP happens to be the closest to T within T’s c-command domain. In an actor-subject sentence, the external argument that resides in the specifier of a functional head (vP) immediately dominated by tp is the closest DP c-commanded by T and therefore the DP that T Agrees with. This is represented in (49). (The dashed line indicates the Agree relationship.)

I will assume in what follows that Agree between T and a DP involves valuing a Case feature (e.g. Nominative) on the DP and concomitant valuation of Phi-features for T.Footnote ^[18] This will be important shortly.

theme-subject sentences for Rackowski are derived when the verb’s internal argument is moved out of the VP, as schematized in (50), and placed in the highest specifier position of the projection where the external argument is merged. As a result of this process, which Rackowski refers to as Object Shift, the internal argument winds up closer to T than the external argument and hence becomes the closer target for Agree. The result of this derivation is what we have been referring to as theme-externalization, whereby the theme argument is realized as the subject of a theme-subject sentence.

According to Rackowski, object shift of the verb’s theme argument is triggered by an [epp] feature on v, which selects a DP within its c-command domain to be merged in its specifier. This feature is absent on v in actor-subject sentences, and so object shift does not apply.Footnote ^[19] Footnote ^[20]

Of particular interest for our purposes, Rackowski also suggests that the often-cited specificity restriction also follows from the analyses sketched in (49)/(50) in conjunction with an assertion appearing in Chomsky (Reference Chomsky and Kenstowicz2001: 33) that configurations resulting from displacement of the object from the VP (e.g. object shift) have particular semantic properties that correlate with (e.g.) specificity. Chomsky claims, in particular, that certain types of objects (pronouns, definite DPs, etc.) may be incompatible with the interpretations that would be assigned if they had not undergone object shift. Although not acknowledged explicitly as such by either Chomsky or Rackowski, the relationship between object shift and the semantic properties alluded to by Chomsky is what is pre-figured explicitly by Diesing’s (Reference Diesing1992) Mapping Hypothesis.Footnote ^[21]

According to the Mapping Hypothesis, there is a unique mapping from the Logical Form of a sentence to the tripartite semantics of quantification, whereby syntactic material outside the VP is mapped onto a quantifier’s restriction, while material inside the VP is mapped into the nuclear scope. Following Heim (Reference Heim1982), Diesing assumes that indefinite as well as other types of ‘weak’ (non-presuppositional) DPs are interpreted as free variables that must acquire their quantificational force from a quantifier or some other type of operator that binds them. Among the possible binders is the existential quantifier which, according to Diesing, is introduced into the Logical Form representation by an operation of existential closure, which takes the VP as its syntactic domain.

As discussed in detail in Diesing & Jelenik (Reference Diesing and Jelinek1995) and Diesing (Reference Diesing1997), the Mapping Hypothesis predicts certain restrictions on the syntactic distribution of direct objects. Concretely, because the VP is the domain of existential closure, only those DPs functioning as objects that introduce a free variable – i.e. those that are type ${<}$ e,t ${>}$ – may appear within the VP. A non-specific indefinite DP will always meet this condition – the variable introduced by an indefinite DP is bound by the existential operator that is introduced by existential closure, which yields an existential (indefinite, non-specific) interpretation. Specific indefinite and definite DPs being of type ${<}$ e ${>}$ , as well as quantified noun phrases being of type ${<}<$ e,t ${>}$ ,t ${>}$ , may not appear within the VP since the result of existential closure over a noun phrase of either type would be semantically ill-formed. Likewise, pronouns and proper names, which are also of type ${<}$ e ${>}$ , may not appear within the VP. Diesing & Jelenik and Diesing hypothesize that DPs that are not type ${<}$ e,t ${>}$ must escape the effects of existential closure by moving out of the VP either in the overt syntax or by Logical Form (de Hoop Reference de Hoop1992, Runner Reference Runner1998, Carlson Reference Carlson, Coene and D’Hulst2003, Hallman Reference Hallman2004, López Reference López2012).Footnote ^[22]

It should now be clear that the Mapping Hypothesis, when wedded with Rackowski’s analysis of Tagalog actor- and theme-subject sentences ((49)/(50)), predicts only a single pattern relating to the morphosyntactic realization of the theme argument in Tagalog. Concretely, only a non-specific theme (themes of type ${<}$ e,t ${>}$ ) can function as the direct object of an actor-subject sentence, while all others types of themes (pronouns, proper names, and all types of specifics) must be realized as the subject of a theme-subject sentence. The former conclusion follows because a non-specific theme must remain within the VP (given the Mapping Hypothesis), in which case it is too distant from T to enter into an Agree relationship with it (T therefore agrees with the external argument in this case). The former conclusion follows because specific themes obligatorily move out of the VP – outside of the domain of existential closure – to a position where they become the closer target for Agree with T.

Given our observations from the preceding section, it is plain that this relatively simple picture does not account for the full range of facts. The analysis predicts the right pattern for pronouns and proper names as far as theme-externalization is concerned (i.e. the realization of the theme as the subject of a theme-subject sentence, or, in Rackowski’s terms, object shift followed by Agree between T and the theme), but incorrectly precludes all other specific themes from being realized as genitive marked objects in actor-subject clauses. Put in other terms, Rackowski’s analysis does not presently account for the more fine-grained differential behavior of themes with respect to theme-externalization. Furthermore, Rackowski’s analysis, being solely concerned with the relationship between actor-subject and theme-subject clauses, does not account for the possibility discussed in Section 3.2 of oblique case marking for specific themes in actor-subject sentences.

A solution to the problem of the differential behavior of themes vis-à-vis theme-externalization that is consistent with Rackowski’s basic proposal would be to stipulate that pronoun and proper name DPs undergo object shift overtly (with the ‘narrow’ syntax), while object shift for other specific DPs may apply overtly or covertly (at L(ogical) F(orm)). If we make the further assumption that covert movement in contrast to overt movement does not feed into the Agree relation initiated by T, then the correct pattern follows. The problem with this approach, however, is that there is no principled reason why the covert/overt distinction should work in this particular way, as opposed to, say, the other way around, with object shift (optionally) covert for pronouns and proper names and obligatorily overt for non-pronoun/proper name specifics. In other words, there is no obvious reason why overt movement should specifically ‘privilege’ pronouns and proper names.

It would be preferable, then, to maintain that all themes that must raise out of the VP to escape existential-closure – all themes that are not ${<}$ e,t ${>}$ – do so overtly. As alluded to earlier, a way to do this and account for the difference between pronouns and proper names on the one hand and all other specific themes on the other is to suppose that there are two distinct positions external to the VP that Rackowski’s object shift targets. Concretely, suppose following proposals of Johnson (Reference Johnson1991), Collins & Thráinsson (Reference Collins and Thráinsson1996), Basilico (Reference Basilico1998), Hornstein (Reference Hornstein, Epstein and Hornstein1999), Hallman (Reference Hallman2004), and Travis (Reference Travis2010), among many others, that there is an intermediate derived object position located above VP but below vP. Suppose that this position exists in addition to the position that Rackowski identifies as the sole target of object shift for Tagalog, i.e. the outermost specifier of vP. If we suppose, finally, that non-pronoun/proper name specific themes minimally target the lower object position, while pronoun and proper name themes obligatorily target the higher position (see (51)), the contrast between the different types of themes begins to follow.

Given the proposal already in place that T Agrees with the closest DP within its c-command domain, T will obligatorily Agree with either a pronoun or a proper name that has undergone object shift to the highest derived object shift position (DP $_{1}$ in (51)). The intermediate object position (DP $_{2}$ ), on the other hand, is high enough for the class of DPs that raise to this position (i.e. non-pronoun/pronoun noun specifics) to escape existential-closure, but crucially low enough that they will not be targeted for Agree(ment) with T.

Supposing this much, three questions now open up.

I will defer a complete answer to Questions 1 and 2 until Section 6. I will do for the moment with stipulating that pronoun and proper name themes obligatorily target the highest position within vP, while non-pronoun/proper name specific themes minimally raise to the intermediate object position DP $_{2}$ in the structure in (51), although they may also target the higher position DP $_{1}$ as well (in which case, they will be Agree with T and be realized as the subject of a theme-subject clause). In Section 6, I offer a more formal and hopefully more illuminating proposal. I now turn directly to proposing an answer to Question 3 and offering some positive support for the key assumptions associated with the analysis surrounding object positions in (51).

5 Oblique marked themes

As we have observed, the option of oblique marking for the theme argument in an actor-subject clause operates in a way that completely parallels the option of theme-externalization. Specifically, although these two morphosyntactic strategies for realizing a theme argument are mutually exclusive of one another (a point we will return to shortly), the option of oblique marking, like the option of theme-externalization, is obligatory for pronoun and proper name themes, and optional for other specific themes. The correspondence between these two strategies emerges naturally if we suppose that the rule that assigns oblique case to a theme argument applies to the same syntactic configuration that underlies theme-externalization – namely, to a syntactic configuration like the one in (52) in which the theme has raised to a functional projection located above the projection wherein the external argument is located. (Evidence for this high position for oblique marked themes is reviewed below.)

Furthermore, the rule that assigns oblique case to the theme should not apply in a configuration like (53), where the theme is located below the external argument in the lower derived object position that we have posited.

A rule that meets these needs is given in (54).Footnote ^[23]

I will take the domain referred to by this rule to correspond to the Spell-out domain of a Phase. Following Chomsky (Reference Chomsky, Martin, Michaels and Uriagereka2000, Reference Chomsky and Kenstowicz2001), I take the relevant Phases to be, at least, vP and CP, and the Spell-out domain to correspond to the complement of the Phase heads – C and v, hence TP and VP. Given this, the oblique case rule can apply in the configuration in (52) because the theme and the external argument are within the same Spell-out domain. By contrast, the rule will not apply to the configuration in (53) because the external argument and the theme argument are within distinct Spell-out domains.

By hypothesis, the rule in (54) exists as an option alongside (but in complementary distribution with) the other route mentioned earlier by which a raised DP may gets its Case feature valued – namely, through Agree(ment) with the functional head of the clause T. Concretely, if the rule in (54) does not apply then T will Agree with the raised theme with the concomitant effect of valuing the Case feature of the theme (e.g. as nominative) and having its own Phi-features valued by the theme. (The result of this strategy for valuing the raised theme’s Case feature is a theme-subject sentence.) On the other hand, if the rule in (54) does apply, then T will Agree with the external argument and the result will be an actor-subject sentence. One additional anaytical detail is needed here to see how this works – in particular, to explain why assignment of [oblique] to a raised theme blocks Agreement with T. Put in more concrete terms, what we need is to rule out a sentence like (55), where the verb (a theme-subject form of the verb) appears to Agree with an oblique marked theme.

The ungrammaticality of (55) follows from specific assumptions about the mechanisms of Agree postulated by (Chomsky Reference Chomsky and Kenstowicz2001). Concretely, Chomsky proposes that a head (a Probe in his terms) can enter into an Agree relationship with another element (a Goal) if and only if the Goal has uninterpretable and unvalued features that need to be valued and checked. This condition is referred to as the activity condition. Consider, for instance, the Agree relationship discussed above involving T and the highest DP within the vP (either the external argument or the theme if the theme has moved to this position). As discussed above, I assume that Agree between T and this DP has two functions: valuing the (uninterpretable) Case feature of the DP and valuing the (uninterpretable) Phi-features on T. Given the activity condition, these two functions are inextricably linked – if DP has its Case feature valued from a source other than T, then it will be inactive for purposes of Agree with T and will therefore be unable to supply the value for the Phi-features of T.

Given this, consider the situation schematized in (56) (cf. (52)), in which the theme argument has raised to the highest specifier of vP (just above the external argument) and has been assigned [oblique] by the rule in (54). Here, T has unvalued Phi-features that could in principle be supplied by either of the two DPs within its c-command domain – either DP $_{1}$ (=the theme), which is closest to T, or DP $_{2}$ (=the agent/external argument).

By the activity condition, T can Agree only with the external argument in this configuration, since the Case feature of the theme has been valued (as [oblique]) by (54), leaving it with no uninterpretable/unvalued features that would make it active for Agree with T. T can therefore only Agree with the external argument since only the external argument has an unvalued Case feature making it active for Agree. In short, the activity condition derives for us the results that if a theme is assigned [oblique] by (54), then T cannot Agree with the theme (thus precluding (55)), but must instead Agree with the agent/external argument. Conversely, if a theme is not assigned [oblique] by (54), then the locality condition associated with the Agree operation (i.e. closest c-command) will require T to Agree with the theme.Footnote ^[24]

With the mechanism of oblique case assignment to raised themes now in place, let us turn to consider the most crucial claim associated with this analysis, namely that oblique marked themes, as well as themes that function as subjects in theme-subject clauses, are structurally higher than the external argument. This contrasts with genitive marked objects in actor-subject clauses, which are located below the external argument (though external to the VP if specific by earlier hypothesis). There are two pieces of positive support for this claim. The first comes from variable binding. In an actor-subject clause, a quantified noun phrase serving as the external argument and subject may bind a pronoun contained within a genitive marked direct object, yielding a bound-variable interpretation for the pronoun. Unsurprisingly, a genitive marked quantified noun phrase serving as the object cannot antecede a pronoun contained in the external argument/subject and license a bound variable interpretation for it. Assuming c-command as a requirement for bound variable anaphora, the contrast between (57a) and (57b) demonstrates that the external argument asymmetrically c-commands the (genitive marked) object in actor-subject clauses.

When the theme argument is realized as the subject of a theme-subject clause, as in (58), it crucially may bind a pronoun contained in the external argument, as the following example from Richards (Reference Richards, Philips, Paul and Travis2000) demonstrates.Footnote ^[25]

Crucially, an oblique marked theme in an actor-subject clause may also bind a pronoun contained within the external argument.Footnote ^[26]

Example (59), compared with (57b), positively supports the claim that an oblique marked theme (in an actor-subject clause), in contrast to a genitive marked theme, is structurally higher than the external argument. This fact also supports the claim that an oblique marked theme in an actor-subject sentence is structurally at least as high as the subject marked theme in a theme-subject sentence.

Additional evidence for this claim comes from scope. Consider first the sentence in (60), an actor-subject clause with a genitive marked object.

The most salient reading for (60) is one in which there is a single police officer who has arrested every thief – i.e. the reading where the external argument/subject has wide scope over the genitive marked object. In fact, this interpretation seems to be the only one available. When asked to judge the plausibility of this sentence in a context where bawat magnanakaw (‘each thief’) denotes every thief in a single large geographical area (e.g. the state of Texas), speakers noted that sentence (60) does not seem very plausible because a single individual is unlikely to be able to carry out so many arrests all on their own. Consider next, then, the sentences in (61) and (62). Sentence (61) is an actor-subject sentence with an oblique marked theme object, while sentence (62) is a theme-subject sentence.Footnote ^[27]

As predicted, both of these sentences have an interpretation that is absent in (60), according to which, more plausibly, different police officers were responsible for the arrest of individual thieves. This interpretation is a straightforward result of a syntactic configuration in which the theme has scope above the external argument. Curiously, though, sentence (61) with the oblique marked theme has both an interpretation where the theme has wider scope than the external argument as well as one (like (60)) where the external argument has wider scope than the theme. By contrast, the theme of the theme-subject sentence in (62) seems to only have the interpretation where the theme has wider scope than the external argument.

The difference between (61) and (62) follows, I claim, from a hypothesis that the noun phrase that functions as the subject (i.e. the agent in an actor-subject clause, the theme in a theme-subject clause) ultimately winds up outside of vP in Spec, TP (Sabbagh Reference Sabbagh2014: 64–67). If this is correct, then for the structure of a sentence like (61) the external argument occupies at least two distinct positions in the course of the derivation. One of these positions is the specifier of vP, below the hypothesized position to which the oblique-marked theme has raised. The other position is the specifier of TP, which is above the position of the oblique-marked theme.

Given this, we can now conjecture that the availability of the two interpretations for (61) arises because there are two locations where the external argument can be interpreted. Concretely, if the external argument is interpreted in the specifier of TP, then we obtain the a ${>}$ each interpretation. On the other hand, if the external argument is interpreted in the specifier of vP, then each ${>}$ a is obtained. For sentence (62), by contrast, the external argument does not raise to TP’s specifier. Instead, it is the theme argument that, by hypothesis, raises to this position. In principle, then, there are two positions where the (externalized) theme may be interpreted, as schematized in (64), but both of these positions crucially scope above the external argument and, hence, the only interpretation available given the relative scope of the theme and the external argument will be the each ${>}$ a interpretation.Footnote ^[28]

The same scope facts hold for a slightly more complex example involving the interaction between quantified noun phrases headed by (the Tagalog equivalents of) many and most.Footnote ^[29] First, it will be important to note that a quantified noun phrase headed by the quantifier marami (‘many’) in subject position elicits a subtle ambiguity between what Cohen (Reference Cohen2001) refers to as a ‘linear’ and a ‘reverse’ reading. Sentence (65), for instance, can be understood either as an assertion that among Filipinos, many are living in debt (=the linear reading) or as an assertion that among those living in debt, many are Filipino (=the reverse reading).

For the examples to be discussed immediately below, we will be primarily concerned with the reverse readings associated with quantified noun phrases containing the quantifier marami (‘many’). Consider, then, the minimal pair of sentences in (66) and (67) and the two scenarios described below the examples. Note that the difference between (66) and (67) resides solely in the case marking associated with the object – it is marked genitive in (66) and oblique in (67).

Scenario 1: There are 15 police officers, five FBI agents, and 10 crimes. There are 10 police officers and one FBI agent who investigated a total of (at least) seven crimes, while the remaining five police officers and four FBI agents each investigated only a single crime.

Scenario 2: There are 15 police officers, five FBI agents, and 10 crimes. There are four police officers and one FBI agent who investigated a total of (at least) seven crimes, while the remaining 11 police officers and four FBI agents each investigated only a single crime.

Sentence (66) was perceived to be true given the first scenario above, but false under the second scenario. By contrast, sentence (67) was perceived to be true given either of these two scenarios. This is expected if the object can be interpreted as having wider scope than the external argument only when it is oblique marked (as it is in (67)) but not when it is genitive marked (as it is in (66)). Concretely, the second scenario is one that is only true on the interpretation paraphrased in (68b) in which the object (‘most crimes’) is interpreted as having wide scope over the external argument (‘many police officers’). The other scope arrangement, paraphrased in (68a), could only be true of the first scenario.

In sum, the binding and scope facts reviewed here offer a fairly clear piece of evidence that there is a relationship between the morphosyntactic realization of a theme argument and its relative syntactic scope – concretely, themes that are either marked oblique or realized as the subject of a theme-subject clause are syntactically more prominent than the external argument, while genitive marked themes (of actor-subject sentences) appear to be structurally less prominent than the external argument. In the absence of simple word order evidence, this is the type of evidence one would hope for to confirm that the basic analysis sketched in the first part of this section is on the right track.

6 The definiteness hierarchy and the architecture of the clause

A pronoun or proper name theme must be expressed either as an oblique marked object in an actor-subject clause, or as the subject of a theme-subject clause. Other (non-pronoun/proper name) specific themes, by contrast, may be expressed in either of these ways, or they may be expressed as a genitive marked direct object in an actor-subject clause. In terms of the account of these patterns sketched above, these differences relate to the claim that pronoun and proper name themes obligatorily raise to the higher of two VP-external positions, while other specific themes raise minimally to the lower of the two VP-external positions (although they may raise higher). The question we turn to now is why there should be this particular distribution. The key to answering this question, I propose, is the definiteness hierarchy in (69) from Aissen (Reference Aissen2003), cf. Comrie (Reference Comrie1979, Reference Comrie1989) and Croft (Reference Croft1991).

The empirical basis for (69) according to Aissen is established by cross-linguistic patterns of Differential Object Marking (DOM) – the phenomenon whereby certain types of objects, but not others, may receive special types of case marking (Bossong Reference Bossong, Wanner and Kibbee1991). Concretely, Aissen demonstrates that the different patterns of DOM summarized in (70) flow from the definiteness hierarchy, which is taken to express the implication that if a language has DOM at a certain point on the scale, then it will have DOM for all points higher ranked on the scale.

Note that if we consider the partially ordered definiteness hierarchy in (71), the hierarchy also provides a useful basis for stating the generalizations concerning the Tagalog patterns we have encountered so far, relating, in particular, to the different morphosyntactic strategies for expressing a theme argument. Concretely, a theme argument that is higher than the point on the hierarchy labeled ‘Non-Specific’ may be realized either as the subject of a theme-subject clause or as an oblique marked object of an actor-subject clause. Pronouns and proper names form a natural class in that they must be realized in one of these two ways, while noun phrases that may be characterized as either definite or indefinite specific form a natural class in that they need not be realized in either of these ways, but may instead be realized as a genitive marked object in an actor-subject sentence.

Viewed from this perspective, the morphosyntactic strategies available for expressing a theme argument in Tagalog come close to following the pattern of DOM in Romanian, where, as described by Farkas (Reference Farkas and Farkas1978), pronoun and proper name objects must be marked (by the preposition pe), but marking is optional for definite objects and impossible for non-specific objects.

In addition to the empirical motivation for this hierarchy involving its relevance to DOM, Farkas (Reference Farkas2000) argues that the general rankings that the hierarchy stipulates can be deduced from a notion of determined reference. Summarizing Farkas’ view informally, DPs whose restrictive conditions narrow down their referent to a single individual have determined reference. Farkas refers to DPs with determined reference as no-choice DPs. Working with the partially ordered definiteness hierarch in (71), Farkas argues that pronouns and proper names form a natural class because they are inherently no-choice DPs. Definite and indefinite DPs for Farkas are descriptions that can function as no-choice DPs just in case the set identified by their descriptive content (i.e. the NP) is a singleton. The function of the definite article (for languages that have one) according to Farkas is to mark a DP as a no-choice DP, while a specific indefinite is a no-choice DP because the Speaker ‘has a particular value in mind for the variable associated with the DP’, even though ‘the context and descriptive content are not sufficient to narrow down the choice as far as the Addressee is concerned’ (Farkas Reference Farkas2000: 17). Indefinite specifics therefore differ from definites in that they have determined reference only relative to the Speaker rather than to both the Speaker and the Addressee. Finally, non-specific indefinites impose no restriction on the the value of the variable they introduce beyond their descriptive content, meaning that they do not have determined reference.Footnote ^[30]

For Aissen (Reference Aissen2003), the definiteness hierarchy is part of Universal Grammar. Operating within the framework of Optimality Theory, she makes use of this hierarchy as a crucial ingredient in the formulation of a family of constraints which have the effect of requiring certain types of objects (e.g. definite objects) to be realized with case marking. I will not review the details of Aissen’s analysis here, principally because it is concerned only with deriving surface morphosyntactic patterns. As was argued in Section 4, the morphosyntactic patterns associated with the realization of theme arguments in Tagalog systematically correlate with certain types of structural prominence. It is the interconnectedness of the morphosyntax and the structural prominence relations that I wish to account for, and so the question that is now opened up is what role the definiteness hierarchy might play in such an account.

A particularly straightforward answer to this question emerges from recent work which explores the hypothesis that markedness hierarchies including, but not necessarily limited to, the definiteness hierarchy in (69) might be expressed in the geometry of clause structure (see, in particular, Jelinek Reference Jelinek, Bobaljik and Phillips1993, Jelinek & Carnie Reference Jelinek, Carnie, Harley, Carnie and Willie2003, Carnie & Cash Reference Carnie, Cash and Johns2006, Merchant Reference Merchant and Bunting2006). Broadly speaking, this hypothesis amounts to the claim that the markedness relations among individual points on the definiteness hierarchy, or natural classes of points defined by this hierarchy, correspond to c-command relationships in the syntax. Holding off for the moment the question of the exact nature of the correspondence, consider how this hypothesis might be implemented to offer an answer to the question posed at the beginning of this section relating to the syntactic positions available to the theme argument.

According to (69)/(71), pronoun and proper name phrases form a natural class of noun phrases in the sense that they are contiguous points on the hierarchy and outrank all other noun phrase types. Definite and specific indefinite noun phrases also form a natural class as they are likewise contiguous points on the hierarchy – both are outranked by pronoun and proper name noun phrases and ranked above non-specific (indefinite) noun phrases.

From the perspective of the hypothesis that the markedness relations described by the hierarchy correspond to c-command relationships, it follows, assuming that there are at least two phrase structure positions above the VP that a theme argument may occupy (as proposed in Section 4), that the phrase structure position occupied by pronoun/proper name themes will be higher than (i.e. c-command) the position occupied by definite or specific indefinite theme arguments. If this is correct, then the distribution of object positions stipulated earlier for (51) (Section 4) now follows.

Given this much, we can now turn to the question of how the correspondence between the definiteness hierarchy and the phrase structure is formally achieved. A very clear answer to this question is provided by Merchant (Reference Merchant and Bunting2006), who argues that the relationship between the definiteness hierarchy in (69) and the clause structure is direct. Concretely, Merchant proposes that the functional architecture of the clause includes a set of functional heads, as in (72), whose purpose is to host in their specifiers phrases (DPs) with the particular prominence properties described in the hierarchy.

One of the goals of Merchant’s proposal is to provide a reanalysis of Aissen’s account of DOM. His proposal involves two key claims. First, all noun phrases must move (overtly) to the appropriate functional projection for their type (i.e. a pronoun moves to FP’s specifier, etc.).Footnote ^[31] Second, a functional head responsible for case assignment (e.g. v) is interpolated at a specific point in the hierarchy (e.g. above JP but below HP). Objects that must move to a functional projection above this head will be assigned case.

To illustrate with a concrete example, consider the case of DOM in Hebrew. In Hebrew, pronoun, proper name, and definite objects are marked by the preposition ‘et, while specific indefinite and non-specific indefinite objects are not marked (Gívon Reference Gívon and Greenberg1978). Merchant’s proposal accounts for this pattern by proposing that the functional head responsible for case assignment (e.g. v) is interpolated in the hierarchy below the functional head associated with definite noun phrases (=HP) but above the functional projections JP and KP associated with, respectively, specific indefinite and non-specific indefinite noun phrases, as in (73).

Cross-linguistic variation with respect to which types of objects are marked and which ones are not falls out from the different locations where the case assigning head is merged. Thus, for a language where, say, only pronouns and proper names are case marked (e.g. Pitjanjatjara, as cited in Aissen (Reference Aissen2003: 452)), v would be merged below GP but above HP.

The account of differential object marking supplied by Merchant can be used also to provide us with an account of the differential behavior of themes in Tagalog vis-à-vis theme-externalization or oblique assignment. Concretely, we can propose that the functional heads in Tagalog that host pronoun and proper name DPs in their specifier are located above the position of the external argument (in vP), and that the functional heads that host other specific DPs (i.e. definite and specific indefinites) are located below this position. This is schematized in (74).Footnote ^[32]

Note that the outcome of this is exactly the same as the analysis sketched earlier in Section 4 (in particular, (51)) – pronoun and proper name themes move to a position above the external argument where they may be the target for either Agree with T (=theme-externalization) or assignment of oblique case by the rule in (54). By contrast, all other specific themes (definite and specific indefinite themes, as well as quantified noun phrase themes (see footnote 30)) move to the functional projections located below the external argument, where they will not be able to Agree with T or be assigned oblique case by (54), yet, by hypothesis, they will still be outside of the VP and therefore outside of the domain of existential closure. Although the outcome is the same as the proposal initially sketched out in Section 4, the advantage of the current proposal is that the differential behavior of themes follows as a consequence of the clausal architecture rather than needing to be stipulated. This is a welcome result.

One might at this point reasonably worry about the number of functional projections that are introduced by this proposal. More specifically, one might worry that although the analysis captures the fact that pronouns and proper names pattern as a natural class (by virtue of the claim that the functional projections associated with both noun phrase types occur above v), there is no direct or even indirect evidence for two separate functional projections – one for pronouns and one for proper names. The same is true for the separate functional projections hypothesized to occur below v for definite and specific indefinite noun phrases. Merchant suggests in passing that it may be possible to reduce this worry by supposing that the different functional heads may be fused into a single functional head. The structure in (74) might therefore be reduced to the relatively less complex structure in (75).Footnote ^[33]

Accepting this much, we can now also say more about the optionality of theme-externalization or oblique case assignment for non-pronoun/proper name specific theme arguments. Up to this point, this option has been accounted for by the stipulation that non-pronoun/proper name specific may optionally target the higher of the two positions in a structure like (48). From the present perspective, we can provide a solution to this optionality that is grounded in the way that Merchant’s approach proposes to handle cross-linguistic variation in the domain of DOM. Concretely, the optionality of theme-externalization or oblique case assignment can be accounted for by supposing that there is variability concerning the point where vP is folded in with respect to the hierarchy of functional projections corresponding to the definiteness hierarchy. For instance, suppose that in addition to the structure in (72), vP may be merged as in (73) below the functional projection that hosts definite and specific (indefinite) noun phrases.Footnote ^[34]

Given (76), a non-pronoun/proper name specific theme will obligatorily raise to the specifier of HP, where it will either be assigned oblique case by the rule in (54) or become the closest target for Agree with T. Overall, then, the distribution of syntactic positions for theme arguments and hence the differential morphosyntactic behavior of themes receives a principled account given the hypothesis that the definiteness hierarchy is represented as part of the clausal architecture.

At this point, one may wonder whether the suite of functional projections corresponding to the definiteness hierarchy is projected specifically for the theme argument or whether there may also be a set of functional projections that are somehow specifically related to the external argument as well. In other words, does this suite of functional projections appear in the clause only once or twice? As far as I can tell, there seems to be no reason to posit a separate hierarchy for the external argument in Tagalog. The primary reason for this is based on the evidence discussed in Section 5, that oblique marked themes or themes realized as subjects of theme-subject clauses are structurally more prominent than the external argument. In order to account for this fact on the assumption that there are two separate sequences of functional projections associated with the definiteness hierarchy (one theme-related, the other external argument-related), it would have to be the case that all external argument-related functional projections are located below the theme-related functional projections that appear above vP in (75)/(76). If these external argument-related projections occur this low in the structure, however, we would not expect to see any particular effect or evidence of them with respect to, for example, Agree with T or oblique Case assignment.

Evidence for a separate hierarchy of functional projections related to the external argument would be supplied, on the other hand, by morphosyntactic evidence (related to Agreement or Case assignment) that indicated that the external argument-related projections occurred above the theme-related projections. Suppose, for instance, that there is a separate external argument-related set of functional projections. Suppose furthermore that the functional head associated with pronouns and proper names for the external-argument related hierarchy (FP-ext in (77)) occurs above the position identified in (75)/(76) of the functional head associated with pronouns and proper names for the theme-related hierarchy (FP-theme in (77)), but below T. Combing these assumptions would give us the schematic structure in (77).

Given this structure, we would predict one of two outcomes. Supposing a clause with pronominal external argument and theme, the oblique Case assignment rule in (54) could assign oblique Case to the external argument and T could Agree with the theme. This would yield a sentence like (78) which is hopelessly ungrammatical. Alternatively, T could Agree with the external argument and the theme could be realized in the genitive case (the conditions for oblique Case assignment to assign oblique to the theme are not met in this configuration). This derivation would result in (79), which is also impossible as we have already observed (see example (37a) from Section 3.2).

Overall, then, there does not seem to be much in the way of positive support for positing a separate suite of functional projections associated with the definiteness hierarchy and specifically related to the external argument. All of the facts that we have encountered support just one hierarchy in the architecture of the clause. I will leave it here as an open question whether there is evidence from other any other language to support multiple instances of the same functional hierarchy in a single clause. Here, we have looked at the type of evidence that would support this for Tagalog, which may serve as a basis for locating the right types of evidence for other languages.