The syntax of non-canonical coordination in Jordanian Arabic: An experimental investigation

2.1 Gapping

Gapping occurs when verbal material in a non-initial conjunct is elided:

There are two major analyses of gapping: reductionist and non-reductionist analyses. Reductionist analyses derive gapping via reduction mechanisms, such as movement and ellipsis, and in this case, coordination always operates on full-fledged constituents. A subset of these analyses assumes that in this case coordination operates at the complementizer phrase (CP) level (e.g., Jackendoff 1971, Ross 1967, Sag 1976, Pesetsky 1982, Williams 1977, Jayaseelan 1990). Another assumes that coordination operates at the VP/vP level (e.g., Coppock 2001, Yatabe 2001, Lin 2002, Crysmann 2003, López and Winkler 2003, Beavers and Sag 2004, Chaves 2007, Johnson 2009, Toosarvandani 2016).

Non-reductionist analyses, on the other hand, account for gapping without resorting to movement or ellipsis. These analyses do not assume that all forms of coordination where it appears that non-constituents are being conjoined are derived from larger constructions (i.e., CPs or little verb phrases). An example of this type of analysis is those proposed within the Categorial Grammar framework (e.g., Kubota and Levine 2015). We will not address this set of analyses here as this contribution does not aim to argue against non-generative accounts of non-canonical coordination. The major aim is to present empirical evidence for non-canonical coordination in JA and an analysis of this phenomenon within the Generative Grammar framework.

One type of criticism of reductionist analyses of gapping is that these analyses are unable to account for peculiarities of scope ambiguity in gapping (e.g., Siegel 1984, 1987, Oehrle 1987, McCawley 1993, Johnson 2004, Kubota and Levine 2016, Potter et al. 2017), as exemplified in (4):

Here, the ambiguity arises from the semantic scope of the modal and negation operators. In one reading, the modal and negation scope above negation. In this case, the sentence reads as: it is not possible for the two events of Jason moving to Japan and his wife moving to Germany to occur at the same time. In another reading, the modal and negation operators scope locally within each conjunct, and in this case, the sentence reads as follows: it is not possible for Jason to move to Japan, and it is not possible for his wife to move to Germany, as separate events independent of one another.

A major challenge to these analyses (which was reported in the literature) has two manifestations: (i) the untenability of vP-sized conjuncts proposal to account for the distributive scope reading analyses, and (ii) the untenability of CP-sized constituents proposal to account for the wide scope reading (Kubota and Levine 2016, Potter et al. 2017).

One major analysis of gapping within the generative framework is Johnson (2009) who derives the construction from vP-sized conjuncts. In his analysis, verbal elements from within the conjuncts undergo an across-the-board (ATB) movement to a higher verbal projection, namely Pred. This is in addition to other movements, such as the movement of the subject of the initial conjunct from spec-vP to spec-TP, as illustrated in (5) for (4) (Johnson uses a multiple branching structure for coordination; we use binary branching structure here. Clearly, Johnson’s analysis accounts for the wide-scope reading of negation and the modal):

(5)

As can be seen, an ATB movement analysis accounts only for the distributive scope reading of gapping. More recently, Potter et al. (2017) argue that gapping can be derived from two sources: a clause-sized source and a vP-sized source. Thus, the vP source will give us the wide-scope reading of semantic operators, whereas the CP source will give us the distributive reading of these operators. Potter et al. provide evidence for this argument from the fact that gapping interacts with independent factors, such that a particular reading of the two possible readings is syntactically constrained. To clarify, consider the examples in (6):

In Potter et al.’s analysis, a constituent appears in a Topic Phrase in examples like (6a, b) (Rizzi 1990). Hence, necessarily, the gapping here should be derived from a CP-sized source because it is impossible for the topicalized phrase to scope over the coordinate structure unless the coordination is clausal. Given this, the availability of a distributive-scope reading in particular is due to a syntactic constraint on the structure. On the other hand, a constituent appears in a Focus Phrase in examples like (8a, d) (Rizzi 1990, Haegeman 2000). Here, only a wide-scope reading is available and the coordination is necessarily at the vP level because the focus phrase appears within the TP. As we can see, the readings are syntactically constrained. In Section 4, we will adopt a modified version of Potter et al.’s analysis to account for gapping in JA.

2.2 NCC

As noted earlier, NCC is among the topics that have received special attention in the recent work in coordination (e.g., Osborne 2008, Hofmeister 2010, Sailor and Thoms 2013, Bruening 2015, Kubota and Levine 2015). Just like gapping, NCC received reductionist (e.g., Wilder 1997, Crysmann 2003, Beavers and Sag 2004, Hofmeister 2010, Sailor and Thoms 2013, Bruening 2015) and non-reductionist analyses (e.g., Dowty 1988, Steedman 1989, Kubota and Levine 2015). The former attempts to derive NCC via movement or ellipsis.

Again, scope ambiguity in NCC was among the major challenges to reductionst analyses; NCC exhibits the same type of scope ambiguity observed in gapping. For instance, in (7a), the negative operator in no girl can have a wide scope reading in which the sentence means: it is not the case that Sam sent a girl chocolates today and flowers yesterday; this reading is unavailable in the non-elliptical version of the structure in (7b), where the negative operator has a distributive scope. The problem here is that the non-elliptical version of the structure allows for only the latter reading, namely the distributive scope reading of negation. This is a serious problem for reductionist analyses.

To account for this peculiarity, Sailor and Thoms (2013) argue that NCC is derived from two sources: a vP source and a CP source. These two sources undergo a series of movements in the non-initial conjunct plus ellipsis of the remnants to derive the surface constructions. For instance, in examples like (7a), the distributive scope reading of negation occurs when NCC is derived from conjoined CPs as in (7b). Sailor and Thoms call the constituents that move to the edge of the non-initial conjunct chunks. The derivation is illustrated in (8) for (7a).^[1] The derivation above accounts for the distributive scope reading of NCC. As for the wide scope reading, it occurs when coordination is low at the vP level. In this case, the negation will necessarily outscope coordination. As before, a series of movements occur in the non-initial conjunct followed by an ellipsis of the remnant of the movement.^[2] Thus, for Sailor and Thoms, the ambiguity in NCC is structural.

(8)

3.1 Experiment (1): The status of gapping in JA

3.1.3 Results

The results were analyzed using R (R Core Team 2021). To simplify the results, we converted the Likert scale values for Rating to binary values of ‘Acceptable’ (4–7) and ‘Unacceptable’ (1–3) and named the new variable Acceptability. To test for any possible effects of gapping on Acceptability, a generalized linear model (GLM) was fit to the data, using the (glm) function from mlogit package (Participant was not included as a random factor). The results are shown in Table 1. The intercept, as thankfully highlighted by an anonymous reviewer, represents the forecasted value of the dependent variable when all independent variables are set to zero. Importantly, it does not signify interaction. Conversely, interaction is specifically observable between two factors, namely Gapping and Acceptability in this context. We present the distinctions between the two levels of acceptability.

Table 1

Results of GLM: gapping ∼ acceptability

	Estimate	Std. error	Z value	Pr(>|z|)
(Intercept)	1.23327	0.09341	13.203	<2 × 10−16***
Gapping_No Gap	0.12656	0.13455	0.941	0.347

*** indicates high significance.

As shown in Table 1, there is a main effect of gapping in the positive direction (χ ² = 13.203, p < 0.001). Figure 1 shows the percentage of Acceptability values for Gapping factors (Gap and NoGap). As can be seen, the general population finds gapping acceptable. (The y-axis shows the overall number of test items *participant.)

Figure 1

Acceptability percentages for gapping constructions in JA.

3.2 Experiment (2): The status of NCC in JA

3.2.3 Results

Similarly, the results of Experiment (2) were analyzed using R (R Core Team, 2021). To simplify the results, we converted the Likert scale values for Rating to binary values of ‘Acceptable’ (4–7) and ‘Unacceptable’ (1–3) and named the new variable Acceptability. As before, to test for any possible effects of NCC on Acceptability, a GLM was fit to the data, using the (glm) function from the mlogit package. The results are shown in Table 2.

Table 2

Results of GLM: NCC ∼ acceptability

	Estimate	Std. error	Z value	Pr(>|z|)
(Intercept)	1.9621	0.1276	15.371	<2 × 10−16***
NCC_NoNCC	−1.4887	0.1541	−9.662	<2 × 10−16***

*** indicates high significance.

As can be seen, there is a main effect of NCC in the positive direction (χ ² = 15.371, p < 0.001). Figure 2 shows the percentages of the Acceptability options for the two NCC factors (NCC and NoNCC).

Figure 2

Acceptability percentages for NCC constructions in JA.

4.1 Assumptions

We adopt the view that coordination targets constituents (i.e., phrases or clauses) of the same category. Thus, where it appears that non-constituents are being conjoined, some reduction processes (movement or ellipsis) have been applied. This is a long-standing assumption within phrase structure grammars, such as the Generative Grammar framework. We also adopt the view that non-canonical coordination (i.e., gapping and NCC) can be derived from two sources: a CP source and a vP source (Potter et al. 2017, Sailor and Thoms 2013). We will show below that the CP source captures the distributive scope reading of scope-taking elements in non-canonical coordination, whereas the vP source captures the wide-scope reading of these elements. The last assumption has to do with the way syntactic structures are derived. We assume, following Phillips (1996, 2003), Bruening (2014), and Osborne and Gross (2017), that syntactic structures can be derived from left to right, reflecting how sentences are processed. Within this framework, displaced elements are merged before their lower copies in the base position. This way of merge and derivation allows for lower copies to be minimal, which means that some elements in the higher copies can be skipped in the lower copies (more clarification below).

4.2 Derivation of non-canonical coordination

The empirical investigation of NCC was not concerned with the ambiguity debate. However, any analysis of NCC should account for the ambiguity issue – given that no systematic study was done to establish the acceptability of NCC in JA, as far as we know. We will thus explain the derivation of examples that give rise to ambiguity below. Moreover, it is ambiguity that motivated the two-source analysis that we adopt here. Consider the following example from JA:

As can be seen, the sentence gives the right to ambiguity as explained in Section 2. The non-elliptical version allows only for a distributive reading:

We propose that the structure involving NCC is derived from full-fledged sources: a CP source and a vP source. The former gives rise to a distributive-scope reading. That is, the girl to which Ali did not give chocolates or flowers is not the same girl or does not have to be the same girl. The CP source undergoes ellipsis, giving us the surface structure of NCC. We adopt the analysis advanced by Sailor and Thoms (2013), where the chunks in the non-initial conjunct undergo topicalization to the left edge of the conjunct, followed by an ellipsis of the remnant of movement (Section 2). To explain how the analysis works for (11), the derivation begins from left to right by constructing the first conjunct. The structure of coordination we adopt is the one proposed by Collins (1988) where coordination projects its own phrase (i.e., &P). For simplification, we will use the English translation of the Arabic words in the tree.

(13)

The derivation continues with merging the constituents from left to right. The non-initial conjunct is merged. Following Sailor and Thoms, the chunks that survive deletion in the non-initial conjunct focus move to the left edge of the non-initial conjunct. Then, the remnant of the movements is elided:

(14)

As can be seen, negation occurs in all conjuncts; hence, the distributive scope reading of negation. The wide-scope reading of negation, on the other hand, is derived when negation outscopes the coordination. We propose that this reading is derived from the vP source of NCC, where negation occurs outside coordination. In this case, ATB movement applies to the vP source of non-canonical coordination. As assumed earlier, in a left-to-right syntax, when ATB movement applies, the lower copies of the moved phrase can be minimal, as an effect of distance during processing. To explain, the derivation of the vP source of (11) proceeds as follows. The subject ‘Ali’ is merged followed by T. Following Johnson (2009), there is a PredP projected above vP. The verb undergoes ATB movement to Pred. The indirect object no girl undergoes movement to the edge of vP. Note that displaced copies are merged first in left-to-right syntax:

(15)

The derivation continues with building &P. Only minimal copies of the moved phrases are merged in the lower positions as an effect of distance from the landing site. Therefore, the quantifier ‘no’ is skipped from lower copies and only ‘girl’ is merged:

(16)

Note that the negative quantifier occurs outside negation, accounting for the wide-scope reading of negation. Note also that the same type of analysis applies to gapping to derive the ambiguity.