1 Introduction

The uncertainty surrounding the lexical aspectual properties of posture verbs in English such as sit, stand, or lie in their ‘simple position’ sense (i.e., when they denote disposition of a stationary object relative to a ground) has led linguists to the creation of different categories to account for their seemingly incompatible features: on the one hand, they resemble events in that they pass some of the same tests that set eventive verbs apart (1a); yet on the other, they also behave like states in that they are devoid of both change and agency (1b). To add to their ambiguity, unlike most stative verbs, they can also be used in causative contexts such as the causative meaning illustrated in (2a) and the ‘assume position’ reading illustrated in (2b).

Another important issue this paper aims to tackle is how to reconcile the fact that posture verbs do not show the same properties cross-linguistically, so that, for example, Spanish sentar ‘sit’ cannot be used in the same way as English sit to express stationary disposition. That is, sentar ‘sit’ requires the verb root to appear as a participle preceded by the copula estar ‘be’ (3–4).

The questions that arise regarding these verbs are related to their argument structure and event structure. Specifically, we propose that the challenges these verbs pose stem from the meaning codified in their lexical entries pertaining to causation and path encoding, which is subject to cross- and intra-linguistic variation. The differences in causation and path encoding in Germanic and Romance languages are introduced in Section 1.1. On the other hand, event structure notions like dynamicity and stativity are important to understand how posture verbs may instantiate both causative events and stationary disposition in English, as well as these verbs’ tendency to become copulas cross-linguistically. These matters are introduced in Sections 1.2 and 1.3, respectively.

The rest of the paper is structured as follows: in Section 2, we discuss the internal aspect of posture verbs and propose a different approach to overcome the limitations encountered by previous work on posture verbs. Section 3 presents Ramchand’s (2008; 2014; 2018) framework, on which our proposal is based. In Section 4, we explain how the differences in causativity and path encoding can lead to a better understanding of these verbs’ properties. In Section 5, we disentangle the co-event and copular uses of posture verbs and shed light on their propensity to grammaticalize into copulas. Finally, Section 6 concludes.

2 The internal aspect of posture verbs

Dowty (1979) discusses a class of English “stative” verbs denoting posture that, counter to expectation, are compatible with the progressive tense, which should only be acceptable with verbs that include the notion of change or agency in their meaning. Yet, these “stative”, and therefore non-dynamic, verbs can bypass this restriction on the use of the progressive tense (12).

Despite being amenable to being used in the progressive, posture verbs are ungrammatical with the phrase “what x did was…”, used to test for agency, which might not be that surprising after all since these events have an inanimate entity as subject (13).

Dowty notes that these verbs are acceptable in the progressive if the entity is moveable or has recently moved or might move somewhat soon. This seems to be the reason behind the differences in acceptability for the sentences in (14), where a locative relation is predicated of different non-animate figures and a ground.

To account for these contrasts, Dowty assumes Carlson’s (1977) thesis that the progressive can only appear with stage-level predicates, that is, those involving a predication of a stage of an individual. By contrast, object-level predicates such as know, love, like, believe, etc., in other words, those predicated of an individual, are incompatible with the progressive. We are then to conclude that posture verbs are stage-level predicates, a conclusion we will endorse here. For that matter, the progressive should only be compatible with stage-level predicates. Dowty further concludes, against Taylor (1977), that not only activities and change of state events require an interval (i.e., a sequence of temporal moments) to be true, but also the English posture verbs sit, stand, lie, etc. should be included in this list. Thus, he proposes a three-way classification of stative predicates depending on whether they can be predicated of intervals or moments (15).

While Dowty’s identification of the properties of posture verbs is insightful, he only deals with their non-agentive use, what we are calling here the simple position meaning, and thus provides a somewhat limited snapshot, as it does not explain why these verbs may be used under other circumstances in contexts requiring an agentive entity (compare (13) with (16)) and, furthermore, denoting a change of state. Yet, Dowty shows that even though some predicates behave like eventive verbs in that they allow the progressive tense, they are still unable to show dynamism, a point that will become relevant at a later point in our discussion.

Another attempt to comprehensively account for these verbs was undertaken by Maienborn (2005), who created a category labeled ‘Davidsonian-states’ which encompasses posture verbs along with other verbs such as sleep, gleam, and wait. The rationale behind this decision is again the contradictory behavior of posture verbs when a new set of diagnostic tests is applied. Maienborn builds an ontology of eventuality types distinguishing between Davidsonian eventualities and Kimian states. Kimian states (17) are found in stative verbs such as love, hate, and copular constructions (i.e., predicates in combination with a copular verb such as be in English or ser and estar in Spanish). In contrast, Davidsonian eventualities (18), typically include process verbs such as play, flicker, flap, and the class of Davidsonian stative verbs (what Dowty called interval predicates), along with others such as sleep, gleam, and wait.

Building on the definition of Davidsonian eventuality, Maienborn devises a set of ontological properties and diagnostics to test for the presence of this type of eventuality in predicates. First, since Davidsonian eventualities are perceptible they should be acceptable as infinitival complements of perception verbs. Second, they can be located in space and time, so, they should be compatible with locative and temporal modifiers. Finally, since they can appear with argument participants, they should also be compatible with manner adverbials, instrumentals, and the like. The application of these diagnostics shows that posture verbs pattern with process verbs, and differ from statives, which in this proposal are subsumed under the Kimian state category (2005:283–297). As an example, consider the contrasts with perception reports in (19–22), which clearly set posture verbs apart from other stative predicates.

Nevertheless, even though posture verbs are different from stative verbs in perception reports, they still behave like other stative verbs in that they cannot be embedded in the phrase “what happened was …”, which is only acceptable with event predicates (23–25).

This last diagnostic leads Maienborn to group posture verbs together with stative verbs and conclude that stative verbs must differ in terms of the entity they contain. Specifically, stative verbs feature a Kimian state, while posture verbs have a spatio-temporal unit in their denotation, just like process verbs, thus constituting in themselves the class of Davidsonian-states. Therefore, what sets Davidsonian-states apart from process verbs must be the size of the interval of which they are true: in other words, while process verbs require an interval to be truthfully predicated, Davidsonian-states, like any other type of stative verb, do not since they “hold at atomic times” (Maienborn 2005:285). Note that the view advocated by Maienborn opposes Dowty’s account, for whom these verbs require an interval in order to be truthfully predicated. We will not dwell on this matter, but we would like to call attention to the fact that the equivalent Spanish stative construction with posture verbs uses the copula estar ‘be’ and a participle, so that, for example, sit in its stative meaning would be equivalent to estar sentado ‘be seated’ (26). If Maienborn’s proposal were correct, these predicates in Spanish, and other Romance languages, should be an instance of a Kimian state, that is, a property predicated of an object instead of a spatio-temporal unit, while English and German posture verbs would be an instance of a Davidsonian event. Furthermore, predicates such as estar sentado ‘be seated and estar enfermo ‘be ill’ would contain different types of events: a Davidsonian event and a Kimian state, respectively, even though they both require the same copula (for further discussion, see Silvagni 2017).

Overall, the literature regards posture verbs as a hybrid combining features of both stative and eventive verbs: they are stative-like in that there is no dynamism or change involved in the situations they denote, but, at the same time, they seem to be event-like in that they can be predicated of a spatio-temporal unit, although they lack dynamism. This is where Silvagni’s (2017) proposal for a new taxonomy of eventualities based on the notion of stage (cf. Carlson 1977; Kratzer 1995), rather than dynamicity, becomes essential. Silvagni draws a sharp line between events and states in terms of the presence or absence of a spatio-temporal point. States are merely properties over individuals, hence spatio-temporal notions are not relevant to them. In contrast, events are defined by the presence of a stage, or stages, as part of their internal constitution. A welcome consequence of the redefinition of eventualities based on the presence of a spatio-temporal unit is that dynamicity is considered an epiphenomenon of events rather than their defining property, thus departing from most accounts of internal aspect, which argue that dynamicity is the deciding feature of eventiveness (cf. Vendler 1957; Smith 1991; Rothstein 2004). When dynamicity is disregarded as the defining property of events, two main classes of eventualities arise, namely, states and events, which may be further subclassified as non-dynamic and dynamic (27).

In Silvagni’s framework, because a lack of dynamicity does not preclude eventualities from being eventive, posture verbs are included in the event category instead of being classified as states. Thus, posture verbs in their simple position sense would not be stative but rather non-dynamic event predicates, which explains why these verbs can appear with the progressive tense in English, are possible in perception reports, and pattern outside stative verbs in all other diagnostics.

Yet, these predicates can also behave like change of state verbs in that an initiating entity can bring about a change of posture in another entity or in itself. A change of state implies the existence of dynamicity in the event. This accords with Silvagni’s definition of dynamicity, which obtains whenever a sequence of stages or spatio-temporal units is triggered by the action of an entity able to produce such an event. Whether the actant fulfills the action intentionally or unintentionally is not relevant; what is important is that the entity can generate such an event, since the property of being acted on belongs to events rather than to the actant. This property of events will be identified here with Ramchand’s initP, which is available to those events liable to being caused or initiated (see Section 3). Events that include this node in their first phase syntax tree will then be interpreted as dynamic if there is a process head in the lexical entry of the verb root. In Section 4, we propose that posture verbs’ assume position and causative meanings include an initP to add the initiating subevent that triggers the process subevent, leading to a result state in both meanings.

3 The framework: Ramchand’s first phase syntax

In Ramchand’s first phase syntax (2008; 2014; 2018), event and argument structure are mutually linked as the verb phrase (VP) is segmented into three phrases corresponding to the initiation, process, and result subevents. Each phrase has available a specifier position to lodge the entities serving as participants of the subevents, except for initP, whose participant appears as the specifier of the evtP, the head that introduces the subject of the predication. This information is codified in roots’ type-A meaning, which instantiates structural meaning relevant to syntax and is used to create the articulated phrase structure. By contrast, type-B meaning encapsulates lexical conceptual meaning with no effect over syntax but ultimately relevant to the mind module as it spreads all over the syntactic structure, when the derivation is sent off to the conceptual-intentional interface. The number of projections is determined by the category labels included in the verb root’s lexical entry, which is inserted into as many terminal nodes as category labels are specified. That is, terminal nodes may not necessarily be realized by a single verb root, but rather a lexical item may identify a terminal node or a chunk of structure in line with Nanosyntax’s premises (Starke 2010; Caha 2019). Each category label corresponds to a subeventive projection identifying the subevents of the macro-event, namely, a causative subevent, a process subevent, and a result subevent, respectively (28).

Like other theories of argument structure, Ramchand’s first phase syntax presumes the introduction of an event variable along with the verbal head. The event variable may consist of several subevents, which together form a macro-event via the ‘leads to’ relation (29).

In Ramchand’s framework, a process denotes an eventuality with internal change unlike states, which lack it. Only procP is assumed to contain an event variable, whereas initP and resP instantiate states (30).

The interpretation of state subevents is contingent on its merging position; thus, the composition of argument structure is syntactically determined. Ramchand proposes two rules to determine its interpretation: if the state precedes the process subevent, then it will be interpreted as an initiation subevent (31); in contrast, if it follows the process subevent, it will be understood as instantiating a result subevent (32).

This notwithstanding, we will assume Silvagni’s (2017) redefinition of the notion of event in terms of the presence of a spatio-temporal unit, or stage, and discard dynamicity as a necessary property of events. This conception allows the existence of both non-dynamic and dynamic events, while states become properties of an entity, to which they apply regardless of spatio-temporal notions (see (27b) and (27a), respectively; repeated below as (33)).

Consequently, it is necessary to modify the denotation of the Ramchandian process as a subevent containing solely a spatio-temporal unit, which in the present terms constitutes an event. On the other hand, a distinction between initP and resP is drawn according to which only the former denotes a proper state of being the cause of the initiation of an event (i.e., a process), whereas the latter is regarded as an event, hence containing a spatio-temporal unit. Under these assumptions, dynamicity results from the concatenation of at least an initiation and a process head, or the combination of the process head with a scalar head, in other words, a result or a path phrase.⁶ The latter possibility has already been discussed in Fábregas & Marín (2017), where they put forward a proposal which also endeavors to separate eventivity and dynamicity and which derives dynamicity from the type of complement selected by the process head in Ramchand’s first phase syntax. Note that the changes introduced do not substantially modify Ramchand’s event composition rules. The interpretation of two subevents in immediate vicinity follows from the previously stated rules, which resolve that a state subevent appearing before an event, or process head, is an initiation subevent (34), and that two consecutive events are interpreted as a process and a result subevent (35), each in order.

Ramchand’s theory also accounts for the possibility of having event participants as expressed by NPs, DPs, PPs, etc., and how they contribute to the composition and interpretation of the event. These elements may occupy the specifier or complement positions of the initiation, process, and result heads, thereby determining the interpretation of the entities as participants in the subevents. A predication relation is stablished between heads and their specifiers. A DP appearing in the specifier position of the result head will be interpreted as resultee, while if it appears in the specifier position of the process head, it will be interpreted as undergoer. The same reasoning follows for the interpretation of DPs in the specifier position of the initiation head in Ramchand (2008). This notwithstanding, following Harley (2013), Ramchand argues for the existence of a functional head evtP, hierarchically higher than initP, which merges with the resulting first phase syntax and whose specifier lodges the external argument of the predicate (see Kratzer 1996; Alexiadou, Anagnostopoulou & Schäfer 2006; Pylkkänen 2008, among others). The entity occupying the specifier position of this head may take up this position via external or internal merge. evtP also closes the first phase syntax and deploys its content to denote a property of events (see Ramchand 2018 for further discussion).⁷ Importantly, evtP is independent of the init head, introducing the causative semantics in the first phase syntax. It is the causative head that legitimizes the introduction of an initiator in the external argument position, that is, an entity whose inherent properties allow it to generate the event regardless of intentionality. The subevent introduced by the initiation head is characterized as a state, in other words, an eventuality without internal change. The denotation corresponding to each subeventive head along with the instructions for interpreting each DP in the corresponding specifier position, regardless of the verb, is detailed in (36–38).

Ramchand (2014) considers that cross-linguistic variation should encompass exclusively type-A meaning, that is, information that is structurally relevant for the first phase syntax of verbs. Among the meaning components included in the repository of type-A meanings, Ramchand includes scalar change, understood as the expression of incrementality, path of motion, and property change, which appear as complements of procP or as result phrases. Assuming the widely held view that languages differ in terms of the possibility of instantiating a path, or result, in motion and property change events by means of an element independent of the verb (Mateu 2002; Acedo Matellán 2010; Folli & Harley 2020), satellite-framed languages like English will allow the possibility of instantiating the res head of the first phase syntax by means of a satellite (or a verb root, if a given root encodes that type of meaning), whereas verb-framed languages such as Spanish will only be able to instantiate the path, or result, via the verb root merged in the result head of the VP and no other element will be available to provide this type of meaning.

In conclusion, when posture verbs denote a dynamic kind of event, they should minimally contain an initP and procP in their phase syntax, to which a resP should be added to denote the result state deriving from the change of state. In Section 4, we further motivate this assumption by drawing on cross-linguistic causation and path encoding patterns.

4 Causation and path encoding

In Section 1.1 we put forward two reasons for the differences in the properties of posture verbs in English compared English with other languages. First, English uses a labile strategy to encode causation, that is, it employs the same verb root for causative and inchoative structures with no morphological difference between them, whereas the other Germanic languages mentioned here, as well as all Romance languages, mainly follow the anticausative pattern whereby the assume position meaning is created by means of a reflexive pronoun (see Table 1). Second, languages may encode path information either in the verb root or by means of a satellite independently from the verb. These differences translate into two patterns that posture verbs may exhibit: either a complex first phase syntax consisting of initP, procP, and resP, or a simple one comprising a procP. This section will deal with each of these patterns in turn.

From a cross-linguistic point of view, variation in the expression of causativity is commonplace. Languages vary in the way they mark the relationship between inchoative and causative verb pairs that share the same meaning. This idea is found in Haspelmath (1993), who thinks of an inchoative/causative verb pair as essentially expressing the same change of state but diverging in that the causative form contains information about the causer, while the inchoative lacks it suggesting that the event occurs spontaneously. Whether a situation can be conceptualized as occurring spontaneously determines the existence of an inchoative version of a causative verb. Based on this idea, Haspelmath (1993) proposes a three-way typology of languages to account for the variation in causativity expression. Languages can use a causative, anticausative, and/or non-directed alternation strategy. In causative and anticausative languages one of the verbs is marked as basic and the other one is marked as derived, as we see in posture verbs in Romance languages and some Germanic languages, whereas in non-directed alternation strategy languages like English there is no such derivational relationship between inchoative/causative verb pairs. Note that it is possible for one language to include inchoative/causative verb pairs of different classes. Thus, the typology could be seen as based on tendencies that languages follow while still allowing for other possibilities to express causativity.

This is the approach adopted by Alexiadou, Anagnostopoulou & Schäfer (2006; 2015), who propose a three-way typology of anticausatives based on Greek data. Type A includes anticausatives that are morphologically marked, which in the case of Greek is achieved by means of non-active morphology (39). This class would also include anticausatives marked with a reflexive pronoun, as in Romance languages or German. Type B groups together all unmarked anticausatives (40), which in Greek appear with active morphology. Finally, Type C (41) admits both possibilities: in other words, this class is characterized by the optionality of the marking.

To determine whether an inchoative/causative verb pair will be morphologically marked, these authors resort to Haspelmath’s “scale of increasing likelihood of spontaneous occurrence” (1993:105), which is the likelihood that a situation can occur spontaneously without the involvement of an acting entity. The rationale behind this is that situations that are less likely to occur spontaneously should have marked inchoative forms, while situations that tend to occur spontaneously should have unmarked inchoative forms or optionally marked ones. Although this reasoning may help explain the inchoative/causative verb pairs in their sample, we argue that this is a simplification, which does not encompass all the possibilities languages encode. The spontaneity of an action cannot be the only relevant factor. In Spanish, for example, a reflexive clitic pronoun can accompany posture verbs and, at the same time, express a “high degree of involvement” on the part of the entity initiating the event. Thus, the presence of a reflexive clitic pronoun cannot be said to solely mark the absence of a causer, or spontaneity, when some events require the existence of an undergoer in the conceptualization of the event. To test the non-spontaneous behavior of posture verbs, consider the acceptability of these verbs (42) with adverbial complements such as deliberately or in order to, which should be incompatible with anticausative predicates (Levin & Rappaport Hovav 1995; Mendikoetxea 1999; Chierchia 2004; Alexiadou, Anagnostopoulou & Schäfer 2015, among others). This fact will be reflected in the proposed lexical entry for these verbs.

The data in Table 1 shows that posture verbs in Romance languages (Spanish, Catalan, and French) and Germanic languages (Swedish, German, and Icelandic) tend to follow the anticausative alternation. In stark contrast, English adopts the labile alternation, a non-directed alternation type. As to why this might be so, one can assume that each language’s lexical inventory and its availability of pronouns and verb roots will constrain the adoption of one pattern or another. In other words, the availability of anticausativizing strategies might be contingent on the existence of reflexive pronouns or voice morphology in a language. In this connection, it is pertinent to call attention to studies that address the issue of productivity resulting from satellite-framedness such as Berthele (2004; 2013), Berthele et al. (2015), and Lewandowski & Mateu (2016). On the other hand, each language’s preferred strategy to encode causativity will have a bearing on the type-A information codified in verbs, which can be attested in the properties of posture verbs (see (5) and (6); repeated below as (43) and (44)).

Following Ramchand (2008), we adopt a causativization approach for English based on a null init head that is merged on top of the structure to introduce the initiational subevent and bring about the process subevent. For a verb to causativize, init cannot be specified in the lexical entry of the verb, thus the lexical entry for English posture verbs (45) reflects the observations made above: to wit, due to its labile nature, the construction of the causative forms will require the insertion of a null init head in the first phase syntax, as the verb root by itself only contains a process label. It will be recalled that we redefined the process head using Silvagni’s (2017) conception of eventivity as the existence of a spatio-temporal unit in the denotation of the event. Consequently, the lexical entry specifies that the event denoted by the posture verb sit is a non-dynamic one.

Therefore, it is not surprising that a productive way to express non-dynamic meanings with process posture verbs in English is the progressive (46). We assume with Ramchand that -ing morphology is “a function from event descriptions to event descriptions such that the derived eventuality is an Identifying-State for that event” (2018:61). The -ing morphology can take scope over the event and output a non-dynamic description, allowing speakers to infer “the existence of a more complex extended eventuality in practice, given the right conditions” (2018:62). The resulting structure would have as holder of the state the DP in the highest specifier position of the first phase syntax. The DP would be raised to the specifier position of evtP, which would in turn be spelled out as the auxiliary be (Ramchand 2018), that is, as the materialization of the tense, aspect, and evt nodes.

To express autocausative meanings, English can leverage its satellite-framedness, which makes it possible to codify the path independently from the verb via a satellite. A dynamic event of sitting, for example, will only be possible with the introduction of an init head in the syntax (47). English, being a satellite-framed language, can express the path via a satellite. This is the situation with the assume position meaning in English, which employs satellites such as down or up to express the path of motion. These particles add a result-state subevent to the first phase syntax in (45). By adopting Ramchand’s (2008) result augmentation operation we can account for the properties of the argument structure resulting from the combination of a pure process (posture) verb and a particle. The verb meaning is built compositionally by the addition of a small-clause (SC) structure containing the particle, which can further identify the res head in the verbal ensemble, as it is lexically specified with a res feature and allows the presence of a subject.

The characteristic pattern displayed by Romance languages and Germanic languages such as Swedish or German is reflected in (48), exemplified with sentar ‘sit’ in Spanish. The verb root encodes all three category labels: init, proc, and res, in line with its adherence to the anticausative pattern and its verb-framed nature. At the same time, these languages also have verb roots that obey the characteristic English pattern (45) in verbs such as Fr. gésir ‘lie’, Cat. jeure ‘lie’, Sp. yacer ‘lie’ (see Table 1).⁸

The lexical entry proposed for Spanish sentar ‘sit’ in (48) yields a syntactic structure consisting of an initP, procP, and resP in both the assume position and the causative meaning. Note that the intransitive posture verb sit (45a) denotes a process whose participant can only be the undergoer of the event given that the lexical entry contains no information about the initiator of the event. By contrast, the lexical entry for sentar ‘sit’ (48a) specifies that this verb contains the category labels init, proc, and res, hence the subject is not only the undergoer and resultee of the event, but also the initiator in the assume position meaning. We would argue that the assume position sense and the causative sense differ fundamentally in the fact that the former contains two DPs that refer necessarily to the same entity, since this element is simultaneously initiator, undergoer, and resultee of the event (50) (cf. (42); repeated below as (49)). By contrast, in the causative meaning (51), the DPs possess disjoint reference: one of the arguments plays the role of initiator and the other argument assumes the roles of undergoer and resultee. Therefore, the two argument structures are identical except for the elements lodged at Spec,evtP.

In (50), the function of the clitic is that of a position filler; specifically, it fills spec,evtP. On the one hand, it satisfies the requirement of initP to have a DP that assumes the role of initiator; on the other hand, its reference is determined to be identical with the reference-bearing DP below in the argument structure. We follow Pujalte & Saab (2012) and assume that if a transitive verb with a D feature lacks a specifier in the external argument position, it will lead to a crash in the PF branch, since the verb’s selectional requirements would not be met. To repair the absence of a DP in the syntax, the clitic is inserted post-syntactically as a last-resort repair strategy in the PF branch to meet the D feature requirement of the head. The clitic that appears in this construction has expletive-like characteristics, since it enters the derivation with unvalued phi features that are later valued against a full-fledged DP in the structure through a probe-goal relation.

The existence of a result state in the lexical entry of Spanish posture verbs can be confirmed with the assume position meaning of these verbs (52), where the temporal complement durante una hora ‘for an hour’ is ambiguous between two readings: a result state-related interpretation and an eventuality-related interpretation. The first interpretation of the adverbial measures the length of the result state, whereas the second one indicates for how long the event was repeated (Piñón 1999). In the result state-related interpretation, these sentences could be paraphrased as ‘Elisa {sat down on the chair/lay down on the sofa} and stayed in this state for an hour’.

As to the simple position meaning in Spanish, we noted above that Romance languages do not have available as many process verbs denoting posture as do English and other Germanic languages. The way in which the simple position meaning is expressed with posture verbs following the pattern in (48) in Spanish, and similarly in other Romance languages, is via an auxiliary (estar ‘be’ in Spanish) and the result participle of the posture verb (see (3) and (4), repeated below as (53) and (54)). The lexical restrictions that apply can be explained with reference to the proposed lexical entry for posture verbs in Spanish (48), which entails initiation, process, and result subevents. The posture verb sentar ‘sit’ contains all three subevent components, thus precluding the possibility of its expressing a state. This explains why we can only obtain a non-dynamic meaning out of this verb through the resultative construction with the copula estar ‘be’ (54b). We assume with Ramchand (2018) that the participle instantiates the result subevent of the verb, thereby allowing a non-dynamic interpretation of the verb root (55b).⁹

In conclusion, posture verbs may be associated with two possible structures: a complex first phase syntax consisting of initiation, process, and result phrases, and a simple one comprising a single process phrase. Thus, to express stationary motion with posture verbs instantiating the complex version as is the case with Spanish sentar ‘sit’, the posture verb’s participle, which codifies the result subevent, is used along with the copula estar ‘be’. By contrast, if a posture verb instantiates the simple version, containing only a process subevent, a non-dynamic type of event is immediately obtained since it simply denotes a spatio-temporal stage. In Section 5, we show how posture verbs associated with the simple lexical entry can develop into copulas when the lexical meaning pertaining to the manner of posture disappears.

5 Copular and co-event uses of posture verbs

In Section 1.3 we hinted at the existence of two stable versions of posture verbs: a full-fledged lexical version and a copular one. Talmy (2000) put forward the idea that posture verbs introduce a manner co-event, referring to the posture or disposition of the figure, along with information about its location. This view of posture verbs would correspond to the full-fledged lexical version, a name which brings to the fore the entwining of both type-A and type-B meaning in the verb root (56). This use of posture verbs contrasts with the copular one, which is bereft of manner information and simply locates the figure in space (57).

Under the assumption that posture verbs can express a manner co-event, it should be possible to paraphrase them by expressing the manner independently from the verb root, thus permitting the separation of the stative-locative meaning from the manner component specifying the posture, a prediction which is borne out (58–59).

Clearly, these sentences are altogether impossible in Romance languages such as Spanish, which can only make use of a participle and the copula estar ‘be’ to express a meaning like that of the Germanic structures (60).

This apparent deficiency in Romance languages’ posture verbs is due to lexical availability rather than to the absence or unavailability of an operation in the syntax or semantics to conflate a manner component and a non-dynamic event. Indeed, Romance languages also have a (somewhat limited) number of non-dynamic posture verbs (61–63).

In Den Dikken’s (2010) analysis, the copular use of posture verbs presumes that the verb instantiates a stative aspectual operator be (64a), which is also present in the full-fledged lexical version along with a manner component adjoined to it (64b), an intuition which we think goes in the right direction. However, there are certain implications that need to be considered if we are operating under the assumption that verb roots instantiate different sub-eventive verbal phrases, as we have by adopting Ramchand’s framework.

To deal with co-events in Ramchand’s framework we assume that a verb, realizing at least a process head, may take an additional element, independent from the verb root, to express a result subevent, thus, creating a more complex event. The result obtained mimics the co-event semantics argued for in Talmy (1991; 2000). The framing event would correspond to the bare first phase syntax, to which a co-event may be related by the existence of at least two roots with the relevant category labels instantiating process and result. The so-called manner meaning arises in the conceptual interface as the root contributes its conceptual meaning, or type-B meaning, which is linked to the spatio-temporal unit introduced by the process head. Thus, the relevant distinction between verb-framed and satellite-framed languages is the locus of the lexicalization of the path information, that is, whether the res head is instantiated by a verb root or satellite. The sequences in (65, 66) are clear examples of co-event conflation of a manner or cause event (slide, roll, bounce, or blow) and a result subevent, instantiating a path (down and off).

As for posture verbs expressing the simple position meaning, they cannot be regarded as equivalent to the examples of co-event presented above. Essentially, they merely consist of a manner component codifying the figure’s posture and a location without resultative semantics. The existence of a co-event is based on the availability of at least two subevents, a main subevent of cause or manner of motion and a secondary subevent of path of motion, which put together constitute a macro-event. Following Ramchand (2008), we would argue that the locational phrase that appears along with these verbs is an instance of a rhematic-PP, in which the figure is merged in the specifier position, thus creating a predicational structure without the implication of a secondary subevent (see (55), repeated below as (67), and (68)). Assuming the existence of a result state in this construction is counterintuitive since there is not necessarily a previous action of laying that triggers a result subevent. These sequences simply denote a state of location that includes the disposition of the figure with respect to the ground according to parameters of posture and spatial coordinates. The analysis we propose is based on previous work by Hoekstra & Mulder (1990), who argue for an unaccusative analysis of some uses of unergative verbs, where the surface subject is in fact the subject of a SC in complement position, which we discuss in the following section.

5.2 Copular uses of posture verbs

After Butt & Lahiri (2013), Ramchand (2014) presents the idea that, because of a stable kind of (verbal) polysemy, a verb may have two versions, a full-fledged lexical version and a light one. These forms can be differentiated in terms of the amount of semantic information each contains (Ramchand 2014): full-fledged lexical verbs instantiate type-A and type-B meaning, whereas light verbs exclusively contain type-A meaning.¹¹ The existence of a common source for main and light verbs explains why these concurrent elements share the same phonetic form. As to the semantic content provided by ‘light’ verbs, it is mainly related to event semantics such as causation or the existence of result states, as well as the legitimization of additional event participants such as benefactives or the inclusion of distinctions such as agentive or accidental, also present in the full-fledged form (Ramchand 2014), which is why light verbs are dependent on the presence of a predicative element (Butt & Lahiri 2013).

In the case of posture verbs in the simple position meaning, the ‘light’ verb use is equivalent to copulas, as they contribute to the meaning of predicates not only as linkers between the figure and the ground, but also as aspectual markers of unboundedness or durativity (i.e, type-A meaning) in that they also provide a spatio-temporal unit to the verb’s first phase syntax, which consists simply of a process head and a rhematic PP, thereby conferring on it a non-dynamic event semantics. The diagram in Figure 1 schematizes both the localizing function of the simple position meaning by means of a rhematic PP and the unboundedness of the event denoted by the process head. The procedure whereby process posture verbs become ‘light’, that is, copulas (or auxiliaries over time), is an instance of copularization (Hengeveld 1992), or grammaticalization.¹² During the copularization, which may involve several stages, the posture verb may take on different complements and fulfill as many different functions (see Kuteva 1999).

Figure 1

Stationary motion with posture verbs.

A paradigmatic case of copularization involving a posture verb is that undergone by the Latin process posture verb stare ‘stand’ (74) as it evolved into Spanish estar ‘be’ (Hengeveld 1992). This verb’s path of change involved several stages. In the first stage, the lexical verb began to be used with locative phrases, proof of the verb’s new function as a localizing copula. Next, the verb’s combinatorial possibilities expanded beyond locative phrases and began to include adjectives and nouns. Finally, in the last stage (contemporary Spanish), the posture verb, already acting as a copula, can also be used to express possession (see Batllori & Roca 2011).

1. (74)

1. V
1. →
1. (x_i)_Loc
1. >
1. A
1. >
1. N
1. >
1. (x_i)_Poss                              (Hengeveld 1992:246)

A number of similar stages can be identified in the copularization of English remain, for which Van Gelderen (2015) provides the relevant moments of the reanalysis from its original meaning of ‘stay with’ (75) to its reanalysis as a copula, where it can appear with a complement such as a PP (76). The copularization process usually begins with the loss of part of the verb’s specificity. In the case at hand, remain lost its location and duration meaning, which would pave the way for its subsequent reanalysis as a copula.

1. (75)

1. To the part of this endenture remaynand to the forsaid Alexander.
2. ‘To the part of this agreement remaining to Alexander, mentioned before.’

1. (76)

1. Those lordes of her honorable kinne, which as yet remained vnder arrest
2. should vpon the matter examined, do wel ynough.
3. ‘Those lords of her honorable family, who as yet remain under arrest, should do well enough upon the matter being examined.’
4.                                                                                                                              (Van Gelderen 2015:294–295)

Among the reasons contributing to its reanalysis, Van Gelderen mentions: the presence on closer proximity of a secondary predicate, which could be reanalyzed as its complement (77); the existence of ambiguity as to whether the element following the verb is to be interpreted as an argument or an adjunct (78); and ambiguity as to whether the adjective accompanying the verb is modifying the verb or the subject (79).

1. (77)

1. The Factour with the others did remaine prisoners.
2. ‘The perpetrator with the others remained, prisoners.’

1. (78)

1. I am yor bedman and so shall remayn be the grace of God all the days of myn liff.
2. ‘I am your servant and so shall remain (i) by the grace of God, for all the days of my life.’

1. (79)

1. Since which she was removed to Kimbolton, where she remains now sick.
2. ‘Since when she was removed to Kimbolton, where she now remains sick.’
3.                                                                                                                                                       (Van Gelderen 2015:295–296)

Returning to English posture verbs, the ‘light’ verb use has acquired impoverished semantics, since it does not necessarily imply a physical posture meaning and admits both inanimate and abstract figures (see (50) and (51); repeated below as (80) and (81)). Nevertheless, as shown in the preceding sections, it still applies lexical restrictions on the figure to a certain extent (see (10); repeated below as (82)). Similarly, in Dutch the posture verb zitten ‘sit’ seems to have lost its lexical meaning, realizing now the functions of a stage-level copula (see (11a,b), repeated below as (83a,b)).

1. (80)
1. a.

1. He was lying on the couch.

1.  
1. b.

1. She was sitting on the sofa.

1.  
1. c.

1. They were standing on the corner.

1.  
1. d.

1. The clothes were hanging on the line.

1. (81)
1. a.

1. The problem lies in the fact that S.

1.  
1. b.

1. The operator is sitting in SpecCP.

1.  
1. c.

1. Water keeps standing in the basin.

1.  
1. d.

1. It is hanging in the balance.                                                                                                    (Den Dikken 2010: 49,(46))

1. (82)
1. a.

1. Stand: when long axis is canonically vertical.

1.  
1. b.

1. Lie: when long axis is canonically horizontal.

1.  
1. c.

1. Sit: when there is no major axis, or object has a wide base in canonical position.

1.  
1. d.

1. Hang: when not supported from below.

1. (83)
1. a.

1. Jan
2. Jan
1. zit
2. sit-prs-3sg
1. op
2. on
1. de
2. the
1. bank.
2. sofa
1. ‘Jan is sitting on the sofa.’

1.  
1. b.

1. Jan
2. Jan
1. zit
2. sit-prs-3sg
1. in
2. in
1. Frankrijk.
2. France
1. ‘Jan is in France.’                                                                                                    (Hengeveld 1992:238, (3–4))

This evidence leads us to conclude that there is a stable polysemy for posture verbs in Germanic languages such as English and Dutch. In other words, posture verbs expressing the simple position meaning are undergoing a process of desemantization, whereby type-B meaning is fading, with no implication of physical posture, while type-A meaning endures. The loss of type-B meaning is facilitated by the presence of a PP acting as rheme of the process head, which seems to be a consistent trait of copularization processes cross-linguistically. One could hypothesize that this element, which might have been an adjunct to the posture verb initially, like in the maintain position meaning (see Gómez Vázquez 2020), is being reinterpreted as a complement of the verb validating the localizing function of the construction, a stage like that undergone by remain or estar ‘be’ during its copularization. This reanalysis is detrimental to the physical posture meaning, that is, type-B meaning, which will eventually weaken, to give way to a copula. The durative aspect of the copula will stem from the type-A meaning codified in the verb root, which in the simple position sense associates with a lexical entry containing a single process head, which we have taken to be the locus of eventivity in the first phase syntax, that is, it contains a spatio-temporal unit.

6 Conclusions

We have put forth the hypothesis that posture verbs can be associated with different lexical entries. If they instantiate the simple pattern, consisting of the category label proc, then their aspectual properties are those of a non-dynamic event. This is the case in English, which exploits its satellite-framed nature to create events of change by syntactically including a resP via result augmentation to procP (Ramchand 2008), and causativizes the structure by adding an initP to the first phase syntax. By contrast, the rest of the languages in our sample have been shown to preferentially adhere to the complex pattern, consisting of the category labels init, proc, and res, even though they possess some verbs that follow the simple pattern. Thus, to encode stationary motion, a result participle instantiated by the posture verb root and a copula are used. Unequivocally, the different type-A meaning codified in the lexical entries of each language’s posture verbs determines their ability to encode stationary motion. Another important characteristic of posture verbs is their possibility of having a ‘light’ version of the full-fledged lexical form, which may co-exist with the latter. These stem from a desemantization process, whereby the verb loses its type-B meaning while still keeping its type-A meaning, which in contexts where posture verbs express a simple position meaning amounts to a procP, which we have here identified with the locus of eventiveness.

Abbreviations

Acc = accusative; Act = active voice; Adv = adverbial; Dat = dative; Dom = differential object marking; Intr = intransitive; Nact = non-active voice; Nom = nominative; Pl = plural; Prs = present; Ptcp = participle; Refl = reflexive; Sg = singular; Tr = transitive.

Competing interests

The authors have no competing interests to declare.