We begin with ‘immediate situation’ uses, where uniqueness of the relevant predicate is presupposed relative to the immediate context of use. Ṭuroyo allows the definite article in such cases; the following sentence is acceptable in a context where both the speaker and listener know the dog in question (cf. Schwarz 2009, citing Ebert 1971):

The definite article can also be used to express the uniqueness of ‘larger situations’, where a description is unique relative to a local context (cf. Schwarz 2009, citing Ebert 1971):

Doron & Khan’s (2016) ‘king’ example in (11b) can also be classified as a ‘larger situation’ use.

A case involving ‘global uniqueness’—where the descriptive content is unique relative to the whole world—is the following (cf. Schwarz 2009):

So even if the description is inherently unique, and not mentioned in prior discourse, the definite article is used.

We turn now to part-whole bridging. Bridging is a phenomenon in which the introduction of one entity into the discourse raises other, associated entities to a sufficient level of salience that they can be referred to using an anaphoric expression. The following example illustrates a type of bridging known as ‘part/whole bridging’ (cf. Ebert 1971; Schwarz 2009). In the following example, the first sentence establishes a context, and the second sentence includes a definite description licensed by an indefinite in the first sentence:

Let us suppose that Ṭuroyo’s definite article has the meaning in (2). Then, according to Schwarz (2009), we can explain its appearance in part/whole bridging contexts like the above as follows: The definite description ‘u burgo ‘the tower’ is interpreted relative to the situation consisting of the church introduced in the previous sentence. Since the tower is part of the church, the the tower is guaranteed to exist and in the relevant church-situation, it is the only one. Hence, this data is consistent with a ‘weak’ analysis of the definite article.

To summarize, Ṭuroyo’s definite article is used in all of the environments where Germanic weak articles are found: cases where uniqueness is satisfied in the immediate, larger, or global situation, and in part-whole bridging examples. But as we will see in the following section, it is used in all of the environments where Germanic strong articles are found, as well.

The following example, based on Schwarz (2009), illustrates another type of bridging, known as ‘producer/product bridging’; cf. John bought a book today. The author is French (Schwarz 2009 p. 6).

This use of the definite article is not directly predicted based on the ‘weak’ lexical entry in (2), where the ⟿ λsλF .ιx .F_s(x). As Schwarz (2009: 54) puts it, “when considering wholes and their parts, it is clear that there is a containment relationship between the two, which in turn ensures that whenever we are looking at a situation that contains the whole, it will also contain the part. This is not the case for the relationship between products and their producers. A situation containing a book does not generally contain the book’s author.” In German, the strong article is used for producer/product bridging.

The reasoning is complicated a bit by the fact that the familiarity analysis does not directly predict that these uses should be possible. To account for them, Schwarz (2009) advocates a relational variant of that lexical entry for cases like this, after arguing that the ‘producers’ in producer-product bridging must be described using relational nouns. But this is one respect in which the Ṭuroyo article matches the German strong article, which is analyzed as a familiarity article.

The article in Ṭuroyo also exhibits prototypically anaphoric uses. Example (17) (based on Arkoh & Matthewson 2013) shows an anaphoric use of the definite article.

This example is much like Heim’s glass example (4), where an anaphoric use of the glass does not carry a uniqueness implication; there is no implication that there is only one orange here. Anaphoric examples abound; here are two more, based on Schwarz (2009) and Ebert (1971), respectively:

As with the orange example, there is no implication that there is only one politician here. Under Schwarz’s (2009) reasoning, these examples would suggest an analysis of Ṭuroyo’s article as a strong article, as in (5), where the_i ⟿ λs .λF .ιx .[P_s(x) ∧ x = v_i]. However, we argue below that this data is also consistent with an analysis as a weak article.

Recall from above that Coppock & Beaver (2015) argue for a ‘super-weak’ analysis of the English definite article, one where uniqueness is presupposed but existence is not, based on anti-uniqueness effects in examples like:

where what is implied is that Scott is an author of Waverley, but someone else is as well, so there is no satisfier of the predicate ‘only author of Waverley’.

With this in mind, we set out to determine whether Ṭuroyo’s definite article appears in noun phrases with exclusives that exhibit anti-uniqueness effects. And indeed, we found that it does. We found anti-uniqueness effects not only for nominals in predicate position, as shown in (20), but also for nominals in argument position (object position in particular), as shown in (21).

Example (20) implies that the book has multiple authors, and (21) implies that more than one person sang a song at the party. Regarding (21), for example, our consultant said, “To me, it sounds like there were many songs; Sona was not the only one who sang a song over there.” Hence we do observe anti-uniqueness effects with this type of example. From Coppock & Beaver’s (2015) perspective, what this shows is that a ‘super-weak’ analysis of the sort given in (9) is appropriate: There is no satisfier of the predicate ‘only author of that book’ if there are multiple authors of the book, so existence of a satisfier of that predicate cannot be presupposed by the sentence.

We have seen that the distribution of Ṭuroyo’s definite article spans across strong, weak, and super-weak environments, just like its English counterpart. Should we then conclude that it is ambiguous between all three of the lexical entries given above, or is its meaning sufficiently general that it applies in all three types of cases? The latter hypothesis certainly has the advantage of parsimony. Can it be maintained?

Beaver & Coppock (2015) propose a way of resolving the tension between uniqueness and familiarity for English definite articles with a single, unified analysis. Under Beaver & Coppock’s (2015) view, familiarity is a special case of uniqueness, in the sense that if familiarity holds, then uniqueness automatically holds as well. To account for Heim’s glass example above, Beaver & Coppock posit that nouns may carry an index, so that the glass_i requires uniqueness relative to the property ‘being a glass labelled i’. It turns out that if i is a familiar index, then uniqueness of glass_i is guaranteed, no matter how many glasses there are, under the system that they define. Under these assumptions, an article that carries a uniqueness presupposition should always be usable in anaphoric environments—environments where the strong article is licensed in German—unless it is preempted by another form. A simple way of implementing this idea within static semantics would be to borrow the idea from Hanink (2017) that an index-introducing element idx can head the complement to a determiner, as follows:

This idx_i can combine via Predicate Modification with glass before it combines with the definite article, yielding λx.[GLASS(x) ∧ x = v_i]. Since v_i picks out a particular individual, this predicate is guaranteed to be unique, relative to a given assignment.

Given a super-weak semantics for the definite article as in (9), where the ⟿ λP. λx .∂(|P|≤ 1) ∧ P(x), this analysis can accommodate the full range of uses, with the exception of product-producer bridging (which requires a separate treatment even under Schwarz’s theory). Even if we assumed that Ṭuroyo’s definite article were ambiguous between weak and strong, we still would not be able to accommodate the full range of uses; in particular, we would not be able to accomodate the super-weak uses. Hence we propose that Ṭuroyo’s definite article is a super-weak uniqueness article, but nouns may carry an index; hence the anaphoric uses. Cross-linguistic predictions of this view are discussed briefly in the conclusion.

Before moving on, let us round out the picture of where the definite article occurs by briefly mentioning a number of additional aspects of its distribution. Following Schwarz (2009), we will set these observations aside and not try to over an account for them.

The definite article has proprial uses. Example (18) above illustrated a proprial use of the definite article, modifying a proper name, as in ‘u Sargon ‘the Sargon’. This is one respect in which English and Ṭuroyo differ. As we are not in a position to offer a unified analysis, our conjecture is that this usage involves a separate, subtly different sense of the definite article that is not available for English the. See Muñoz (2019) for a recent treatment of the semantics of proprial articles that might apply here as well.5

As in English, singular definites may denote kinds:

For generic interpretations, definite plurals can be used:

To the extent that kinds constitute a particular type of individual, these uses may somehow be related to the proprial uses, as Schwarz (2009: 66) suggests, but this is an issue that we will not investigate further here.

Unlike in English, we did not find evidence for so-called “weak definite” interpretations (where uniqueness is not implied); for the following example, the consultant we asked reported that Bob and John must be reading the same newspaper.

But given that weak definite interpretations are highly lexically specific (cf. the newspaper vs. the magazine in English, where only the former gives rise to a weak definite interpretation), we hesitate to conclude that weak definite interpretations are unavailable without a fuller investigation of weak definiteness per se, and we leave this to future research. We take all of these uses to be compatible with a uniqueness-based analysis of the definite article, though we will not spell out any particular treatment of them.

As Doron & Khan (2016) point out, definiteness spreading is found in noun phrases featuring both demonstrative determiners and attributive adjectives:

It is also found in noun phrases with possessives:

The previous examples use the possessive suffix -idi/aydi, which co-occurs with the definite article. There is another possessive suffix, -i, which is typically used with nouns denoting close familial relationships, and this one does not co-occur with the definite article (Jastrow 1993: 52–53). However, modifying adjectives are still marked by the definite determiner in conjunction with this kind of possessive:

Like the examples with two definiteness markers above, this construction does involve auxiliary marking of the adjective with the definite article.

Doron & Khan (2016) argue that definiteness spreading (at least with demonstratives) imposes a contrastive interpretation. They give the following example:

We presented informants with a visual scenario in which there was only one car, which was red, and our participants reported that example (29) (with ‘the red car’) would be felicitous in such a scenario, even in the absence of a linguistic antecedent.

Similarly, we found no requirement of contrast in definiteness-spreading examples with possessives. In a context with only one rosebush, the following sentence was judged acceptable by our informants:

Hence, at least for our informants, definiteness spreading is not conditioned by contrast, unlike Doron & Khan (2016) would predict; it seems to be conditioned purely syntactically. Following Danon (2010), we assume that a morphological definiteness feature can be positively specified on both N and A in some languages, and that this fact underlies the phenomenon of definiteness spread in languages like Hebrew. We assume further that Ṭuroyo is one of those languages. In the cases just discussed, the intervention of the demonstrative or possessive phrase between the head noun and the modifier triggers a second realization of definiteness on the modifier, inherited from the dominating DP, as grammatical agreement.

It may seem at this point that almost nothing can limit the distribution of Ṭuroyo’s definite article. But we turn now to superlatives, which will finally put a stop to it.

To express superlative meaning, one option is to use a positive-form adjective along with an explicit comparison class. In this case, a definite article appears:

Waltisberg (2016: 61) reports that this is the most common strategy for expressing superlativity. Superlative meaning presumably comes about in this construction through a combination of the uniqueness requirement imposed by the definite article and flexibility with respect to the threshold for the positive-form gradable adjective. Finding a threshold that satisfies the uniqueness requirement for the definite article leads to a superlative interpretation, such that the referent holds the gradable property to a greater extent than any other individual in the comparison class. Our consultants did not offer this strategy when translating English sentences containing superlatives, perhaps due to a preference for explicit, morphosyntactic expression of the superlative meaning.

When superlativity is morphosyntactically expressed, a form that is morphologically indistinguishable from a comparative is placed syntactically before the noun (whereas adjectival modifiers are normally placed after the noun). There is no morphological distinction between comparatives and superlatives in Ṭuroyo; the distinction is made purely through syntax. There are three ways of forming comparatives in Ṭuroyo, one native, one borrowed from Arabic, and one borrowed from Kurdish.

All of these comparative forms can also be used in superlatives.

The native Ṭuroyo strategy is shown in (36), with the adjective rabo ‘big’ shortened to rab.

The Kurdish strategy, which was quite commonly used by our informants, is seen in the following examples:

Notice that the superlative adjective precedes the noun in these examples. Adjectives normally follow the noun. It is only when a -tër-marked adjective is interpreted as a superlative that it appears prenominally. Here is a minimal pair showing that adnominal superlatives are prenominal, while adnominal comparatives are postnominal:

The same pattern holds for all three of the morphological comparative/superlative-marking strategies: with a superlative interpretation, the adjective is placed before the noun, and otherwise it is placed after the noun. These examples illustrate how the marking of superlative meaning is achieved through a combination of morphology and syntax.

There are quite a number of other Semitic varieties that exhibit this superlative fronting phenomenon. Other Aramaic varieties that do so include Syriac and the Jewish dialect of Zakho; see Gutman (2018: 86, 123) for examples. Superlative fronting occurs in Arabic as well; Elghamry (2004) and Hallman (2016) offer theoretical takes on it, which we will discuss further below. Plank (2003: 361–362) discusses the case of Maltese (heavily influenced by Arabic, if not strictly speaking Semitic) within the context of a general discussion of superlatives in Romance languages, pointing out that because the superlative is fronted, one of the definite articles that would otherwise surface is lost.

But fronting of superlatives in this fashion is not ubiquitous among the Semitic languages. Urmi, for example, a dialect of Assyrian (alternatively, ‘Eastern’) Neo-Aramaic—much more closely related to Ṭuroyo than Arabic or Maltese—treats superlatives on a par with other adjectival modifiers, that is, after the noun (Khan 2016: 67) (although Gutman (2018: 7) notes that superlative adjectives can occur prenominally as part of the Construct State construction in Urmi).

Superlative fronting (an instance of what Gutman (2018) calls ‘inverse juxtaposition’) is “clearly an areal phenomenon” according to Gutman (2018: 123, fn. 10), and it may be due to contact with Persian languages, where superlatives, along with ordinals (which are morphologically superlative) are generally placed before the noun, although adjectives canonically appear after the noun (see e.g. Samvelian 2007 for Persian, MacKenzie 1961: 68 for Sorani Kurdish, and Thackston 2006: 28 as well as MacKenzie 1961: 215 for Kurmanji Kurdish).

Gutman (2018: 123, fn. 10) writes, “One reviewer suggested this [superlative fronting] is semantically motivated, as superlatives establish a unique reference similarly to determiners which are typically pre-nominal.” While superlative fronting may be semantically motivated, it cannot be motivated based on uniqueness. The case of Mauritian Creole (Wespel 2008) is instructive as a point of comparison. In Mauritian Creole, a distinct pattern (namely, absence of definiteness-marking) is found for adjectives that “establish a unique reference”, including sèl ‘only’, superlatives (although here there is a split, described further below), and menm ‘same’. If the fronting of superlatives in Ṭuroyo were due to their inherent uniqueness, then we would expect fronting with all uniqueness-implying adjectives, including exclusive adjectives. As we saw above, exclusive ‘only’ in Ṭuroyo is not fronted, nor is it in any of the other geographically-related languages with superlative fronting, as far as we know. If the fronting is semantically motivated in some other way, the relevant factor is highly specific to superlatives.6

We also do not see the split among superlatives that we see in Mauritian Creole. The Mauritian Creole sentence corresponding to (42) lacks a definite article while the one corresponding to (43) has one. Ṭuroyo’s definite article is persistently absent across these contexts, and superlatives are consistently fronted.

Superlative fronting in Ṭuroyo therefore does not reflect the distinction that governs definiteness-marking on superlatives in Mauritian Creole.

Nor is the fronting of superlatives, and concomittant absence of the definite article, tied to whether they have an ‘absolute reading’ or a ‘relative reading’ (Szabolcsi 1986; Heim 1999). The following example illustrates a relative reading:

(We consider this a relative reading because the comparison class, made explicit with the among phrase, is a set of mountain-climbers, rather than mountains.) One might expect the definite article to be absent only with relative readings, as superlatives are semantically indefinite on such readings (Szabolcsi 1986).

Hallman (2016) offers another somewhat semantically-based theory of superlative fronting in Syrian Arabic in terms of Heim’s (1999) analysis of superlatives. As he points out, the syntactic behavior of superlatives in Syrian Arabic matches Heim’s (1999) posited Logical Forms (LFs), in which superlatives undergo a raising operation within the local noun phrase. He posits that Heim’s LF movement takes place overtly in Syrian Arabic. While this view is attractive, it cannot be maintained for the present case, at least not in a totally unconstrained form: If superlatives always surfaced in their LF position, then superlatives should surface far from their local noun phrase on relative readings too, assuming Heim’s (1999) scope theory of relative superlatives. On relative readings, superlatives undergo a movement operation that lands them in proximity to the focused consituent, far away from the modified noun. As we have just seen, with example (44), superlatives do not move so far away from the noun they modify on relative readings. Thus the surface position of superlatives does not always reflect their LF position.

So, while we cannot definitively rule out the possibility that the fronting of superlatives is semantically motivated, we can safely conclude that it is not based on their inherent uniqueness, nor is it due to a reliable correspondence between surface position and position at LF. There may be no more parsimonious generalization than just that it is superlatives that are fronted.

It is only superlatives that front, and more importantly, it is only with superlatives that the definite article is absent. If the definite article were systematically absent with semantically unique descriptions, then the absence of the definite article with superlatives might be taken as an indication regarding the meaning of the definite article. But given that the definite article is compatible with semantically unique descriptions so long as there is no fronted modifier, we find it rather more likely that the absence of the determiner with superlatives is due to syntactic circumstances, namely preemption by the superlative. It appears that the fronting of the superlative somehow displaces the definite article, leaving no room for it.

As mentioned above, Arabic exhibits the same fronting phenomenon with superlatives. Here is a pair of examples from Elghamry (2004: p. 900, ex. (2b) and (3e)), with a positive and superlative adjective, respectively:

Elghamry (2004) points out that superlatives generally pattern with quantifiers like ‘all’ with respect to morphology and word order, and unlike demonstratives and typical adjectives. They do not exhibit agreement in features like number and gender, they combine with a noun that surfaces in genitive case, and they can precede construct state constructions.

To account for superlative fronting in Arabic, then, Elghamry (2004) argues that superlatives are not of category A, but rather head a nominal projection. Like ‘all’, superlatives head a QP (Quantifier Phrase) projection under DP, and undergo movement to the head of the DP projection. The associated nominal, which surfaces in genitive case, occupies the specifier of this phrase. A parallel analysis in Turoyo would yield the structure in (46) for ‘(the) highest mountain’:

Given that superlative fronting is an areal phenomenon, it is reasonable to expect that similar underlying structures would be involved throughout the Sprachbund. As we have seen, superlatives do not inflect for number and gender, even though ordinary adjectives do. As Elghamry (2004) argues for Arabic, this fact lends a bit of support to the view that superlatives are not adjectival, and rather some other category such as Q. The other evidence given for the analysis Arabic cannot be ported to Ṭuroyo, unfortunately: Ṭuroyo nouns do not inflect for case, so we cannot use evidence from case-marking, and there is no clear evidence for a construct state construction in Ṭuroyo. But as far as we can see, Elghamry’s (2004) analysis of superlative fronting in Arabic can be extended to Ṭuroyo. If this analysis is correct, then the absence of the definite article with superlatives is due to the fact that the superlative is occupying its syntactic position, and not because of an obstacle presented by its semantics.