1 Introduction

The variation in (1–3), English, French, and Spanish, respectively, has been a source of debate over the past forty years with different studies suggesting different characterizations in terms of verb classes to account for the different patterns. Unlike previous works, the focus of this paper will not be on the verb (i.e., verb classes) but on the elements that appear in these constructions, which I think provide an interesting case study.

From the data in (1–3), we can conclude that precipitation verbs are able to participate in different syntactic frames, an unsurprising property, as this is also the case with many other verbs. However, there are two features that make the variation in (1–3) unique. First, the range of possibilities that it encompasses is very broad: it spans from configurations where the verbs appear by themselves, (a)-sentences, to configurations that include DPs in syntactic positions usually linked to arguments, (b) and (c)-sentences.² Second, the elements that appear in these constructions, despite lacking argument properties (in most cases), seem to behave as such. It is on this property that the present study focuses through the examination of two theoretical questions that can shed some light onto their seemingly unusual behaviour: how do we determine whether a given element is a syntactic argument?, and how are arguments and other elements introduced in the syntax?

Taking precipitation constructions as a case study, the first aim of this work is to show that, in some cases, to establish the argument properties of a given element we need to go beyond generally accepted tools of syntactic assessment such as syntactic distribution and syntactic behaviour, and consider the interpretational properties of both the construction in which the element is found and the elements that appear in such constructions. The second aim is to show that the pattern found in precipitation constructions, if taken as a possible outcome as opposed to an exception, offers support for a constructivist model of the grammar, one based on the so-called naked root approach, as in Borer (2005).

Generally speaking, and borrowing the characterization from an anonymous reviewer, precipitation constructions present us with a case of non-arguments masquerading as arguments. Since, as will be shown, neither the DP nor the pronoun can be considered arguments of the verb, I will argue that their presence in these structures is a by-product of the combination of general language constraints along with the particular features of the functional elements in these languages, and general properties of the architecture of the grammar. Specifically, a constraint like the EPP and the properties of functional lexical items—expletive types: overt vs. covert, and whether they bear or lack phi features—can account for both the presence and cross-linguistic variation of the pronominal items. On the other hand, the possibility to merge (and pronounce) the same element in different syntactic positions, a general property of the grammar, can account for: a) the denominal character of precipitation verbs (Hale & Keyser 1991), from which the possibility that a DP that shares the same root as the verb (or a hyponym) realized in the structure simply follows, and b) the argument-like behaviour of the DP in (1c), and its non-argument properties in the (b)-sentences.

This work is organized as follows. The next section provides an overview of previous analyses about precipitation constructions in the three languages, summarizes the general properties of these structures, and introduces some of the theoretical issues that will be addressed in Section 3. It also offers the empirical motivation behind data selection. Section 3 comprises a detailed examination of the syntactic and interpretational properties of weather-it and the DP. In Section 4, I present an alternative analysis of precipitation constructions based on a constructivist model, which incorporates Hale & Keyser’s (1991) intuition of precipitation verbs being denominal. The paper ends with a brief summary and general remarks.

2 Overview of precipitation constructions in the three languages

Precipitation verbs have received different classifications depending on both the language and the structure analyzed. In Romance, some studies consider them to be unaccusatives (Ruwet 1987, French; Calzado Roldán 2000; Meulleman & Stockman 2013, Spanish), while others propose that they are unergatives (Zubizarreta 1985, French and Romance in general). The possibility of dual behaviour within the same language has been noted in Italian (Benincà & Cinque 1992). Recently, the same has been suggested for English (Levin & Krejci 2019).

The existence of these various characterizations represents a theoretical problem for projectionist analyses where a mapping between arguments and syntactic positions is assumed, and arguments are supposed to project from the verb. This is true not only in the case of cross-linguistic variation, but also in cases of variation within a language. For instance, several questions arise from (1): how do we go from (1a) to (1b)?; is there an implicit object in (1a)?; are (1b) and (1c) alternating structures?, if so, what type of alternation do they represent?

While most of these questions have been addressed in previous studies, a less discussed topic is the argument status of the elements that appear in them, Paykin (2010), and Levin & Krejci (2019) being two exceptions to this claim. Interestingly, their characterizations are different. Paykin proposes that precipitation verbs in their non-metaphorical uses are avalent in French, i.e., they lack arguments. Specifically, she argues that the DP that appears in non-metaphorical constructions is not a real object, because, unlike objects of unaccusative verbs, its presence is not obligatory, (4a–b), it does not move to the preverbal field, (5), and it may bear an adverbial interpretation, (6). In her analysis, as well as in Ruwet (1987), il is considered an expletive.

Levin & Krejci (2019), on the other hand, argue that both weather-it and the DP are arguments in English, and propose that precipitation verbs can be construed as substance emission, (7), and directed motion events, (8). In their analysis, the alternation between the two configurations arises from the fact that the root lexicalizes only a small number of the attributes of the happenings it is used to describe.

While it is in principle possible for precipitation verbs in these languages to not pattern alike with respect to arguments, there are strong similarities among (1–3) that suggest this may not be the case.³ From an interpretational perspective, the most salient property is that in the three languages precipitation verbs by themselves seem to be enough to convey the meaning of what I will call for convenience a canonical precipitation event. This is the case of a natural raining event that involves raindrops falling due to gravity, which can be paraphrased as ‘there is rain falling.’⁴ In terms of arguments, this means that precipitation verbs lack arguments. Another common feature is that a DP whose root is the same as the verb’s may appear in non-metaphorical precipitation constructions. In such cases, there is also a common requirement: the DP needs some modification. When there is compatibility between the DP modifier and the properties of the event that the verb encodes, an adverbial interpretation emerges in the three languages. When there is no compatibility, the DP has a classificatory function specifying the properties of the precipitation substance. The DP can also be a hyponym that specifies the type of substance, e.g., drizzle, sleet, etc. Table 1 summarizes the properties of precipitation constructions in the three languages and the classifications that precipitation verbs have received.

A combination of theoretical assumptions and the methods employed to evaluate argument status underlie the characterization that the elements that appear in these constructions in English have received. In other words, some of the procedures followed to determine whether an element is a syntactic argument impact the conclusions obtained. Thus, to understand these constructions, we must reconsider how we determine the argument status of a given syntactic element.

While providing a definition of what an argument is has proved a difficult topic (see Williams 2015), there are two generally accepted features that are used to characterize arguments: semantic content and referentiality. Specifically, arguments are considered to bear a general semantic interpretation, i.e., a semantic role, and they are taken to be referential. Additionally, roughly speaking, arguments are said to be necessary to convey the meaning of the verb—at least the meaning that is intended in the context they are used. In other words, their absence results in an ungrammatical structure. Depending on the approach, arguments are said to project from the verb, the projectionist view, or they are elements merged in syntactic positions associated with specific functions and interpretations, the constructivist view. The two approaches make different predictions with respect to the elements that can be found in a given structure, something that will be discussed in the light of precipitation constructions in Section 4. Both approaches, however, make use of the idea that there is a link between syntactic positions and the argument status of an element.

In addition to referential and semantic content, some syntactic properties associated with specific argument configurations are also used to determine or confirm the status of elements. For instance, since structural accusative case is dependent on the existence of an external argument (Burzio’s generalization), the fact that a DP bears accusative case is taken as evidence that the DP that appears in the subject position is an external argument. The combination of semantic requirements and syntactic distribution is also used to establish the argument properties of elements. While the use of these methods is not a problem per se, there are instances the require us to dig deeper to be able to obtain a proper characterization. I believe this to be the case with precipitation structures. The problem is that, as will be shown, if we only consider distributional evidence and so-called syntactic behaviour, or if we take the presence of some syntactic property without considering all possible options, false positives abound. Before showing how this happens, some observations about the data used in this work are in order.

3 On the status of weather-it and the DP in non-metaphorical precipitation constructions: re-examining the syntactic evidence

In this section I examine the interpretational and syntactic evidence that previous studies have provided in favour of the argument status of weather-it and the DP in English. Previous proposals about the semantic content of weather-it and new data about the DP will be considered. A central feature of the discussion is that it incorporates the meaning of the constructions, a factor that has been overlooked in most analyses. In some cases, data from French and Spanish are employed to provide a cross-linguistic perspective.

4 Precipitation constructions and their components: an alternative analysis

Since neither the DP nor the pronoun can be categorized as arguments, the next question that we have to answer is how do we account for their presence in precipitation constructions? I will leave aside for a moment the pronoun, which can be taken as a general language requirement, and focus on the DP. As suggested at the beginning, one hypothesis is that precipitation verbs are denominal verbs (Hale & Keyser 1991); thus, the possibility of a DP from the same root of the verb to appear in the construction may emerge from their underlying structure. This by itself, however, does not explain the properties of these constructions, or their elements. Actually, Hale and Keyser’s structure for weather verbs, (60) (1991: 22), makes the wrong predictions regarding the properties of the elements that appear in it.²⁸

Since their framework maintains the standard assumption that the element that is merged in the complement position of the verb is an argument, the NP in (60) is an argument. Second, given the relation between NP and the verb, the verb-complement configuration in this case implies that precipitation verbs have an external argument similar to unergatives like laugh, sleep, run, etc. Hale & Keyser (1991), aware of the problem that arises from the presence of an expletive at surface structure in precipitation constructions, suggest two possible solutions: the subject of precipitation verbs is a true subject selected by the verb, or the subject position is empty; thus, rain is unaccusative.²⁹ Unsurprisingly, the first solution is the same as the reasoning found in Levin & Krejci (2019). The second solution corresponds to the analysis of precipitation verbs in French developed by Ruwet (1987), which, as Paykin (2010) shows, is not correct either. Overall, the two options demonstrate that this model of the architecture of the grammar either leaves out precipitation constructions or makes the wrong predictions about them. Thus, even though it accommodates most instances, it may not be correct.

I am aware that precipitation verbs constitute a rather small set, but if we look at this problem from the arguments’ perspective, the possibility that there are verbs that lack arguments does not seem exceptional, providing that they by themselves are sufficient to convey the meaning of the event that they encode.³⁰ An additional property of precipitation constructions that this model cannot accommodate is the existence of cases like (1c), repeated as (61). In this case the DP a light rain occupies the subject position, a position associated with arguments, but if the DP is not an argument of the verb, how is (61) derived?

Crucially, constructivist models of the architecture of the grammar can handle the dual behaviour that the DP displays in precipitation constructions. Since, in these models, arguments are merged in specific functional projections that determine their argument status, and, to some extent, their interpretation, it is possible for the DP in (61) to be merged in a non-argument position first and then to be remerged in a position associated with arguments. Thus, in this work, I will only maintain the spirit of Hale & Keyser’s (1991) analysis—the fact that precipitation verbs are derived from the incorporation of a nominal root into an abstract verbalizing head—and combine it with a constructivist approach.

The specific constructivist model that I adopt is the one proposed in Borer (2005). The main reason is that it espouses the so-called naked root approach, which straightforwardly accommodates the fact that precipitation verbs have no arguments.³¹ In Borer’s framework, roots are not associated with any syntactic content (e.g., syntactic category or argument structure). The position where they are merged in the structure determines their syntactic category. As in other constructivist models, the projection of arguments is based on event structure. Her model includes an event phrase (EP), a functional projection located above IP that establishes a mapping from predicates to events (Borer 2005: 82). The element that appears in the specifier of EP is the argument that may receive the interpretation of the originator of the process. However, such interpretation emerges by entailment; thus, the projection of EP does not imply the existence of an originator (Borer 2005: 83). By the same token, merging an element in the specifier of EP does not make it an originator.³² The other functional projection (FP) where an argument is merged is the Aspect Phrase (AspP), which is related to telicity and accusative case marking. The exact properties of this projection are not relevant for our analysis; what matters is that it is the locus where the default participant, the argument linked to the theme interpretation, the one that undergoes some kind of change, is merged. (62) depicts the structure that will be used in this work.

Before providing the relevant derivations, a last theoretical detail should be pointed out. In this work, I adopt Haugen’s (2009) revision of Hale & Keyser’s (2002) proposal of incorporation, whose central features are the treatment of movement as Copy and the adoption of the idea of late insertion from Distributed Morphology. As Haugen argues, treating movement as Copy allows us to leave a copy of the moved item in its original merging position, which is essential for the possibility to spell out a nominal related to the verb. Additionally, it allows us to maintain the identity between the two copies via co-indexation without breaking the chain. This along with the idea that morpho-phonological material is inserted post-syntactically (late insertion) allows the option to spell out different lexical items in the two copies, making possible the insertion of a non-cognate nominal into the lower copy. This is possible because the encyclopedic information attached to roots is irrelevant to the syntax (Haugen 2009). (63), from Haugen (2009), illustrates the incorporation process in Hopi, as in the CO construction in (64). He suggests that (63) can be extended to English, which is exactly what we do in (65).

(65), corresponds to the VP of examples in (66), in English, French, and Spanish. As can be seen, the √rain is copied onto the v slot to form the (denominal) verb rain. The representation v-√rain, which includes a “v”, is used to make the structure more transparent. Likewise, I am not committed to the insertion of any lexical content in the root and the use of rain is just for illustration.

The insertion of the expletive proceeds at the IP layer as a sentence requirement (e.g., nominal feature) in the three languages alike. The only difference is that in English and French its realization is overt, while in Spanish it is covert. Language internal motivation for proposing a null expletive in Spanish comes from differences in agreement patterns between non-metaphorical and metaphorical precipitation constructions with postverbal DPs, as suggested in Section 2.1. However, the analysis developed here does not depend on the existence of a null expletive. The same agreement facts can be captured with further specifications on the Infl head regarding phi features in so-called default environments. (What matters is that there is no argument in this case.) (67) includes the complete derivation. Note that the specifier of the intermediate FP related to the theme argument does not project in this case. Additionally, only the incorporated copy, the one that forms the verb, is spelled out. We might argue that, in this case, the NP is defective, i.e., it has less syntactic functional structure; thus, the copy associated with it is not pronounced. (Functional projections that are not relevant for the analysis are not included in the derivations. Likewise, verb movement is not represented.)

(68) represents the actualization of (67) when a DP from the same root of the verb or a hyponym is included. In this case, unlike (67), the lower copy of the root is also spelled out, as in (69), English, French and Spanish, respectively. These cases are analogous to Haugen’s example in Hopi.³³

Notice that an AP has been added to the structure.³⁴ However, its presence is not a structural requirement (e.g., part of a more complex structure like a small clause). Note that while it is the case that DPs that share the same root as the verb appear with a modifier in the three languages, when the DP is a hyponym no modification is required, (70). Thus, if we are proposing that both instances correspond to the same structure, then the presence of the adjective cannot be taken to be the result of a more complex structure.

Regarding the adjective, we might argue that when there is an identity between the verb and the DP (i.e., when both share the same root), the spelling-out of two identical copies would be redundant. Thus, cases that do not include modification like (71) are filtered out at LF. These sentences, however, are not ungrammatical.

Massam (1990) offers a similar explanation for (72). She proposes that “since the lexical entry for the intransitive verb […] and the lexical entry for the CO verb derived from it are logically related, they are truth conditionally the same” (182). Thus (72a) does not add any new information with respect to (72b).

Crucially, this requirement is not universal. As Haugen (2009) suggests, “restrictions on hyponymous and cognate objects may depend on language-specific pragmatic conditions, relating to the avoidance of redundancy (Grice’s Maxim on Quantity)” (260). A potential piece of evidence for this hypothesis comes from languages where the lower copy without modification is spelled out. Haugen (citing Klaiman 1990) provides two examples in Hindi where the unmodified copy must be spelled out: khaannaa khaa- food-eating and gaanaa gaa- song-singing. Examples of precipitation constructions that seem to follow a similar pattern are found in Eriksen et al. (2010). For instance, these authors note that in Turkish the verb yağmak is used for precipitation events in general. However, unlike in Finnish where the DP is dropped when it refers to rain, in Turkish it is usually realized, (73). Other cases are (74) and (75), in Mwotlap and Bislama, respectively.

The last construction that concerns us is (1c), repeated as (76), where the DP appears as the subject.

As in the previous context, the modification of the DP in (76) is not the result of structural requirement, but an interpretational or pragmatic one. Since the DP occupies a higher position than the VP in the structure, i.e., it takes the VP as its predicate, a predicative relation is established between the two. The oddness of cases that do not include modification arises because the DP and the verb are logically related; thus, the predicative relation is not interpretable. This is also a case of redundancy, compare (77) and (78), which includes a hyponym and no modification is required.

Additionally, one can hypothesize that the presence of the postverbal PP in this context could also be linked to the impossibility to interpret this predicative relation. Since the DP and the verb are logically related and the predicate does not include anything but the verb, the predicative relation turns out to be vacuous, (79). Thus, descriptively, we can say that the inclusion of the PP allows us to extract some meaning from these constructions.

As a side comment, notice that under special informational conditions, as in (80), from Gardelle (2015: footnote 19), it is possible for the DP to appear with no modification. Moreover, a PP is not required in the predicate. Thus, the grammaticality of (80) can be taken as evidence that the presence of the DP modifier and the PP are not structural but interpretational or pragmatic requirements. The possibility to spell out a hyponym without modification goes in the same direction.

Turning back to (76), note that it could be paraphrased as ‘a light rain fell over the city’. This is similar to the interpretation found in metaphorical constructions. However, even though it may look like there is a bleaching process taking place in this case too, I think it is actually the combination of the structural configuration and the specific elements that appear in the structure what makes it as if there were one. In any case, the interpretation is similar to a light verb construction, except that, in this case, the process works in reverse: part of the verb meaning seems to be bleached, and only the dynamic part, i.e., the movement interpretation, is preserved. As a result, the verb (the predicate) works as a placeholder of the subject DP, which now carries the information about the substance that is part of the event, and has an argument-like function. Derivationally, we can capture this by remerging (copying) the DP into the specifier of the Functional Projection (FP) related to the theme argument. Then the DP then raises to fulfil other clause requirements like the EPP (and licensing the Event variable in EP), and receive or valuate case (depending on the approach). As standardly accepted, the higher copy of the DP is spelled out, as shown in (81).

Before finishing this section, I would like to call the reader’s attention to two related issues. The first one is that in the initial examples there are no French or Spanish counterparts of (76). One property that has been systematically noted about non-metaphorical precipitation constructions in French is that the DP does not move to the preverbal field (Ruwet 1987; Paykin 2010), (82).

However, examples with a DP that could be analyzed as an ellipsis of a phrase that contains the noun rain (DP: Des grosses gouttes (de pluie), similar to the compound raindrops in English) are possible in French, (83).

The acceptability of (83) should make us wonder whether movement to the preverbal field is impossible or if it is just infrequent in French. A similar point can be made about Spanish. Although examples like (84) are probably difficult to find in corpora, they cannot be ruled out as ungrammatical.

Thus, the relevant question is why there may be a lack or low frequency of this configuration in French and Spanish. The tentative answer that I am going to offer is that this behaviour is related to more general properties of these languages. Generally speaking, both French and Spanish allow postverbal subjects with so-called unaccusative verbs, a configuration that carries a thetic interpretation. Since the two options are possible in these languages, when the DP is realized preverbally the structure is interpreted about it. However, given that the two possibilities exist and given the properties of the predicate (specifically the fact that verb and the noun are related), DPs in precipitation constructions are not good candidates to be realized preverbally. One can speculate that since in Spanish and French the realization of a DP in the preverbal field in some contexts carries some interpretation, in instances where the two options exist, preverbal realization would be more frequent, or limited to interpretable cases. Since movement to the specifier of inflection happens across the board in English (only a group of unaccusative verbs allows there-insertion), this type of structure is probably less marginal in English. This last point brings us to the second issue: are there any differences between (85a) and (85b)? Can we take them to represent some type of alternation?

Although these questions are from trivial, based on the previous discussion, I believe the short answer to the second question is no. The way I see it, (85a) and (85b) are two possibilities that emerge from the combination of the properties of precipitation verbs and the specific properties of English. Thus, the two options are, in a sense, equivalent. The fact that the truth conditions of the sentences are the same is compatible with this hypothesis. Even though it could be argued that in order to obtain (85b) more information is needed, I am not sure whether we can identify the exact source of such a need. The possibility that (85b) is interpreted about the DP, while (85a) is a thetic judgement is not easily demonstrated either. Note that both sentences seem to be possible in a context where two people with slightly wet clothes arrive at a friend’s house and are asked about what happened to them. This is a scenario where a broad focus statement would be felicitous, which takes us back to the ambiguity that is found in English with respect to the interpretation of preverbal subjects.

I think that the closest we can get to find a possible difference between the two options is to take the structures in isolation and focus on the fact that the PP is not required in one of the cases. Thus, we can propose that (86a) is compatible with a context where less information needs to be provided, e.g., there was a raining event and the rain was light. In this case the location of the event is probably known by the interlocutors.

Yet, it might be argued that the fact that the PP is a requirement in (86b) but not in (86a) makes a strict comparison between the two sentences unattainable. Likewise, the idea that the interpretation of the verb is to some extent different in the two cases seems to go against the possibility of comparing the two sentences. Overall, the behaviour and the interpretation of precipitation constructions and their components suggest that they cannot be assimilated to other instances in the language. Moreover, if we take their different instantiations as a possible output and not as an exception, it is possible to argue that they offer support for a constructivist model of the grammar, one where not only arguments are merged in specific functional projections, but also roots are devoid of any syntactic information.

5 Conclusions

Taking non-metaphorical precipitation constructions as a case study, I have shown that, in some cases, to obtain an accurate characterization regarding an element’s argument status, both the general interpretation of the structure in which the element appears, and the interpretational properties of the element(s) need to be considered along with traditional methods of syntactic analysis. Specifically, it was demonstrated that the syntactic distribution of weather-it, if taken in isolation, proves to be an insufficient diagnostic tool of its argument status. Moreover, the use of methods that combine syntactic distribution and semantic requirements may give us the wrong idea that meaning is being taken into account. To this end, the behaviour of control verbs was quite revealing. Although weather-it is compatible with a number of control verbs, their interpretation in such contexts corresponds to that of aspectual verbs. A general conclusion that emerged from most contexts analyzed is that the eventive nature of the predicate in which weather-it occurs may be the source of its spurious argument behaviour, a fact that has been largely overlooked.

I have also shown that some theoretical assumptions that derive from a fairly standard representation of the architecture of the grammar within the generative framework make the wrong predictions regarding the characterization of both weather-it and the postverbal DP in non-metaphorical precipitation constructions. This, I believe, speaks in favour of the need for a more articulated architecture of the grammar. Additionally, the fact that the DP that appears in non-metaphorical constructions displays a dual behaviour regarding its argument status (which correlates with the positions in which it is realized in the structure) provides support for a constructivist model of the architecture of the grammar, one where verbs (more specifically, roots) are not linked arguments. Finally, although this study focused on non-metaphorical constructions, the properties and interpretation of metaphorical constructions also point in the same direction. Note that the meanings that emerge in such cases clearly depend on the combination of the elements introduced in the structure and the syntactic positions that they occupy which suggests that meanings are indeed built in the syntax.

Abbreviations

1 = first person, 3 = third person, ACC = accusative, ART = article, CAUS = causative, CIRCG = cirumgressive, CL = clitic, DAT = dative, EXPL = expletive, FUT = future, INF = infinitive, IPFV = imperfective, PFV = perfective, PL = plural, PRED = predicative, PROG = progressive, PTCP = participle, SG = singular

Competing interests

The author has no competing interests to declare.