1. Introduction: Simple copulas and complex copulas

Spanish has two well-known copular forms, ser and estar, each with its own unique syntactic and semantic characteristics. For instance, ser but not estar can take predicates like americano (1a) which denotes a property of the subject. This inherent property reading is also possible when ser has a predicate like ‘red’ (1b).

Alternatively, estar but not ser accepts predicates like cansado in (2a) which expresses the state in which the subject finds himself at a particular time or place. This same ‘current state’ reading can be seen in (2b) with the predicate ‘red.’ Unlike (1b), predicates of estar like (2b) cannot be interpreted as inherent properties of the subject.

Due to contrasts like (1–2) some scholars (Schmitt 1992; Leonetti 1994) have proposed that ser takes individual-level predicates while estar takes stage-level predicates (Milsark 1974; Carlson 1977). In other words, ser takes individual level predicates which assert a property of an individual (1) while estar takes stage-level predicates which have to do with the stages of the individual (2).

Similarly, Gumiel-Molina et al. (2015) explain that estar combines with adjectives with a within-individual comparison class. These compare the subject to counterparts of itself. For instance, (2b) is evaluated with respect to the typical or prior state of ‘his eyes.’ In contrast, ser combines with adjectives with a between-individual comparison class. These make a comparison of the subject “with respect to a class of comparison comprised of individuals sharing some property with the subject of predication, which determines the standard value” (2015). Thus, in (1b) the predicate implies a comparison with a different person’s eyes. Other work, however, has shown that ser and estar do not always correlate neatly with the individual and stage level distinction (Fábregas 2012).

The difference in meaning between ser and estar has also been linked to viewpoint aspect (Luján 1981). This is based on data like (2a) where estar is used for ‘being tired,’ a temporary or ‘perfective’ condition. In contrast, being American (1a) is an ongoing or ‘imperfective condition, so it requires ser. Gallego & Uriagareka (2016) also described estar constructions as perfective and ser as imperfective. Their analysis attributed this aspectual difference to a preposition, or some other contextually anchoring spatio-temporal head called X, that incorporates into the copula. Thus, estar is often seen as a complex form including ser plus an additional element.

In contrast, Arche (2012) and Arche et al. (2017) have argued that ser and estar cannot be reduced to a difference of viewpoint (or situation) aspect. They mention, for example, that both copulas are compatible with perfective or imperfective aspect (3).

For Arche (2012), the sense of temporality one finds with estar constructions may be connected to the fact that their meaning is linked to external circumstances or situations, while ser predicates have a sense of permanence because they are not linked to external circumstances. Arche (2012) derives estar from a non-central coincidence head, namely an ellative head like from (Hale & Keyser 2002). This ellative head is said to explain some of the aspectual characteristics of estar because it captures the fact that “when a property holds of an individual with estar [it] is as a result or consequence of some previous eventuality…” (2012: 127). An analysis of the way in which the ellative head achieves this ‘resultative’ reading is not developed in detail.¹ In turn, Arche et al. (2017) provide additional evidence showing that viewpoint aspect cannot solely explain the ser-estar distinction.² They observe that estar is linked to utterance time, similar to Arche’s (2012) observation that estar constructions are linked to external circumstance. Like Arche (2012), they attribute the temporal boundedness of estar to the coincidence value of some aspectual or prepositional head in the structure.

One recurring idea regarding estar is that it is more complex than ser. Like Gallego and Uriargareka (2016) Arche and colleagues (Arche 2012; Arche et al. 2017) see estar as a complex element derived from the presence of an additional head or element, P, and this head is responsible for the temporal/aspectual characteristics present in estar and absent in ser. This head is also said to explain why estar appears to be linked to a particular external circumstance or some specific spatiotemporal location. In this sense, estar is more marked than ser. Indeed, Fábregas et al. (2023) argue that, unlike ser, estar is more marked because it contains an interpretable terminal coincidence feature that can check the uninterpretable aspect feature of a P or Asp head that merges above a PredP: [_AspP Subject [Asp⁰_[uT] [_PredP …[Pred⁰ [Predicate]]]]]. In this analysis, ser appears when there is no head with an uninterpretable terminal coincidence feature in the structure. Ser can also appear when such a prepositional/aspectual head is present, but its features are checked by another head. Thus, estar is present when it needs to license some terminal coincidence head in the structure, and ser is the elsewhere copula.

Recent studies have also looked at other ‘copular’ forms in Spanish, such as possession verbs, to see if these are also sensitive to the so-called stage and individual level contrast. This line of inquiry emerges from the abundance of research that treats possession verbs like tener or have as copulas. The claim has been traced back to Benveniste (1966), and it proposes that have is a complex copula, namely a copula with an additional element. This can be a dative P (Belvin & den Dikken 1997; Torrego 1999), a locative P (Freeze 1992) an abstract P (Harley 1995; 2002), a comitative P (Avelar 2004; 2009; Levinson 2011), a P that projects a DP (Kayne 1993) an Appl head (Kim 2011) or a variety of heads (Boneh and Sichel 2010). More recently, Myler (2016; 2018) has proposed that have type verbs are basically transitive copulas. That is, have is inserted into v_be when v_be is embedded below a transitive Voice head.

If have-type verbs are essentially complex or transitive copulas, and if copulas are argued to have different forms depending on whether they merge with stage or individual level predicates, then one would expect that possession verbs may also have different forms, one for individual level predicates and the other for stage level predicates. Following Gallego and Uriagareka (2016) Ruvalcaba (2020) suggested that tener, like ser, was Spanish’s individual level possession verb while traer was Spanish’s stage level possession verb. This claim builds on the observation that traer appears to be ambiguous between a dynamic verb of motion and a stative verb of temporary possession (4b) in some varieties. The possessive reading of this verb was acceptable among the speakers consulted by Ruvalcaba (2020) all of whom were from northern Mexico or the southwestern U.S. The current analysis draws from the data in Ruvalcaba (2020) as well as new grammaticality judgements from speakers of Mexican Spanish varieties. Thus, the data may not be licit for all Spanish speakers.³

Ruvalcaba (2020) argues that possessive traer contrasts with tener in the same way that estar contrasts with ser. This opposition is reflected in the different types of possessive relations each verb can express. For example, in (4a) tener can express an ownership reading. The meaning of (4a) could be that Lupe is wealthy, namely that the predicate is expressing a property about Lupe as an individual. However, with traer in (4b) this reading is not possible. The reading in (4b) is one where Lupe has money on her, not that she is wealthy. One can follow it with a phrase that explicitly negates the ownership reading. Thus, the act of having money (4b) is referring to a situation or some external circumstance rather than a property of Lupe.

Finally, in (5a) one finds that traer is incompatible with a characterizing part-whole relations. This resembles the ‘permanence’ or individual level-ness of ser constructions.

Section 2 discusses problems with Ruvalcaba’s (2020) interpretation of data like (4–5). For instance, tener can actually take both individual and stage level predicates.

Additionally, similar stage and individual level contrasts among possessive verbs have been noticed in the Bantu language Mashi (Finholt 2024). An interesting finding in the Mashi data is that the distribution of Mashi’s possession verbs is conditioned by tense and modality. In other words, the possession verb that takes an individual-level predicate is different from the one that takes stage-level predicates, as in (4), but this distinction is only made in the present tense and in the indicative mood. The question of whether this occurs in Spanish as well is discussed in Section 3.

Apart from previous research on these constructions, this paper will present new data to motivate a new derivational account of the tener and traer distinction. This new account proposes that tener and traer are the transitive counterparts of ser and estar, as proposed for have by Myler (2018). This helps to explain the wide range of meanings possible with tener, which is analyzed as the elsewhere possessive copula, as has been argued for ser (Fábregas et al. 2023). In this way, the current analysis manages to preserve and strengthen Ruvalcaba’s (2020) intuition that tener and traer correspond to ser and estar while also accounting for data not explained by the previous analysis. This novel data includes sentences with complex predicates (Section 5).

2. An overview of tener and traer: insights and limitations

The verb tener can express a range of relations. One of these relations is ownership or legal possession (6a). For instance, in (6a) Lupe is a car owner, and in (6b) she is wealthy. A phrase where the ownership reading is contradicted by a conjoined phrase (‘but it’s not hers’) degrades the sentence. Ruvalcaba (2020) treats this as evidence that tener, like ser, is limited to individual level predicates.

In contrast, traer can express temporary possession, and it is not restricted to individual level predicates. This is evidenced by the fact the ownership or wealth readings can be negated by the conjoined clause in (7a–b).

Again, Ruvalcaba (2020) argues that the ownership reading of tener (6) is like the imperfective reading of ser constructions. Likewise, the temporary possession of traer (7) is like the perfective or temporary readings of estar. This supports the idea that tener is derived from ser and traer is derived from estar.

An issue with this contrast is that sentences like (6) can be licit if additional context is provided. For example, the temporal (8a) and locative (8b) modifiers in the following sentences make the sentences acceptable again. This suggests that, when context is provided, tener can have stage level predicates and express temporary relations. In other words, tener is not restricted to permanent or individual-level predicate readings.

Ruvalcaba’s (2020) idea that tener is restricted to individual level predicates likely stems from the fact that it is more compatible with individual level predicates than traer, but it is not restricted solely to this type of meaning.

If one examines the data provided in Ruvalcaba (2020), one can see that tener can take all the predicates that traer can, but traer cannot take all the same predicates that tener can. For instance, tener can express kinship (9a) and part-whole relations (9b) but traer cannot.⁴ Tener is also compatible with predicates like diabetes (9c) which express long term ailments, but traer is not. On the other hand, traer can merge with a noun that refers to a more temporary ailment or condition, like ‘headache’ or ‘cold’ in (9d), and tener can too. As noted above, traer can express temporary control or possession (7) and tener can too (8). This suggests that tener is less restricted to certain meanings than traer, and this conflicts with the analysis of tener as a verb that takes individual level predicates and only expresses permanent or characterizing relations.

To summarize, (8–9) suggest that tener is not necessarily restricted to a permanent or imperfective meaning as suggested in Ruvalcaba (2020). On other hand, traer appears to be restricted to temporary possession, and this resembles the temporal or non-permanent qualities that have been observed for estar.

Another limitation of Ruvalcaba’s (2020) analysis of tener and traer involves its claim that these verbs are the result of the incorporation of the preposition con into the copula, a claim that stems from Avelar (2004; 2009) and Levinson (2011). For tener, Ruvalcaba proposes that the preposition incorporates into ser. For traer, it incorporates into estar.⁵ Ruvalcaba follows Levinson (2011) in assuming that the possessor is introduced by a V head.⁶ These derivations are illustrated in simplified trees in (10–11).

As evidence for the structures in (10–11), he points to constructions where traer appears in its decomposed form. For example, the constructions in (12) are argued to be instances of a traer structure in which the preposition did not incorporate into the copula. Just like their traer counterparts in (9), these so-called estar con constructions can express temporary control or possession (12a) and abstract possession (12b). In addition, like traer sentences, estar con constructions appear to be restricted to temporary relations, as shown by (12c–d).

The comparison between traer (9) and estar con (12) constructions demonstrates that they have similar semantic restrictions. This parallel is thus used as evidence for an analysis like (10). Given that both tener and traer are both thought to have an incorporated con (10–11), one would expect a similar ‘decomposed’ counterpart for tener, namely a ‘ser con’ type construction. However, as shown below, this is not the case.

If traer decomposes into estar con, then tener should be able to decompose into ser con constructions. Nevertheless, ser con constructions are limited. They cannot express ownership (13a) nor short or long-term ailments (13b–c). In contrast, tener can express all these relations (7–9). The sentence in (13e) shows ser con allows some part-whole relations, but not all (13d), while tener can express both types of part-whole relations (9a–b).

The data in (13) demonstrates that ser con constructions do not share as many parallels with tener constructions as one saw between traer and estar con constructions.

The unavailability of a productive ser con construction is unexpected since there is nothing in Ruvalcaba’s (2020) analysis that prevents its productivity. More specifically, Ruvalcaba (2020) explains that estar con constructions are formed when a functional prepositional layer, small p (similar to little v but in the prepositional domain), merges with a PP headed by con. This functional layer above the PP blocks incorporation of the comitative P, con, into the copula, thus allowing v_be to be spelled out as estar. If the p is absent, the P con must incorporate into the copula, as in (10), yielding have. Ruvalcaba does not mention anything about the ser structure that prevents it from merging with a pP structure headed by the same p that intervenes between con and the copula in estar con constructions. In other words, there is nothing structural that restricts ser con constructions from being as productive as estar con constructions.

Ruvalcaba (2020) also explains that the p that blocks incorporation in estar con constructions denotes an underspecified relation. Thus, its presence (or absence) has no effect on the meaning of the sentence; the possession reading comes from the comitative preposition itself. In other words, the same ‘ingredients’ that make up a tener sentence are present in ser con constructions; the only difference is that the latter includes a meaningless functional p that maintains the copula and the P separate. Thus, within Ruvalcaba’s analysis, all of the sentences in (13) should be acceptable. This casts doubt on the notion that tener is composed of ser and con.⁷ This observation combined with the fact that tener is not restricted to individual level predicate readings present a challenge to Ruvalcaba’s original analysis of the tener and traer.

Subsequent sections show that this data is better explained when one abandons the idea that a preposition con incorporates into various copular structures (i.e., ser and estar) to form tener and traer. A more productive analysis is one where tener is a transitive, elsewhere copula (Myler 2018) that is exponed in non-marked contexts, similar to Fábregas’s (2023) analysis of ser. In contrast, traer can be analyzed as a transitive copula that is spelled out in ‘marked’ syntactic contexts, namely contexts where there is terminal coincidence head in the structure.

3. A comparison between Spanish and Mashi

In his analysis of Mashi’s possession constructions, Finholt (2024) finds that there is a verb, dwiire, that takes stage level predicates (14a) and a different verb, jira, that takes individual level predicates (14b). For instance, the sentence in (14b) refers to the state of a car within a particular spatiotemporal location or context, while (14a) refers to a characterizing property of cars. Similarly, Finholt points out that in sentences like (14c), the verb that takes stage-level predicates, dwiree, can express a meaning in which someone is holding the child. In contrast, the verb jira can only express an infelicitous reading where the child is interpreted to ‘live in his arms.’

This contrast is similar to the tener-traer distinction described above. In Spanish, traer is restricted to a context like (14b).⁸ Similarly, tener tends to appear in construction that express permanent possession. Nevertheless, as described above tener is not restricted to individual-level predicates or readings, and, as such, it would be licit in both of the contexts in (14a–b). Likewise, tener would be licit in a context like (14c). This data (14) suggests that jira is more restricted than tener.

Apart from having two possession verbs that appear to line up with the individual vs. stage-level opposition, Mashi also has two copulas that seem to behave in the same way. In other words, Mashi has two copulas that resemble ser and estar. As (15) shows, li is the copula that takes stage-level predicates like ‘sick’ (15a), while ba is the copula that takes individual-level predicates like ‘tall’(15b).

Similar to Ruvalcaba’s (2020) analysis, Finholt (2024) proposes that the possession verb in Mashi that expresses temporary possession dwiire is derived from combining a comitative preposition with the structure of a stage-level copula li. The other possession verb jira is derived from combining a comitative preposition with the structure of an individual level copula ba.

The details of Finholt’s analysis are illustrated in (16). Instead of proposing an X head as the source of the stage-level reading, as was done by Gallego and Uriagareka (2016) and Ruvalcaba (2020), Finholt adopts Myler’s (2018) notion of a PredP with a head that encodes individual or stage level semantics.⁹ Moreover, with the stage-level possession verb dwiire, instead of a V below v_be, the subject is introduced in the specifier of a PP. The PP proposed in this analysis follows Harley’s (1995; 2002) notion of an abstract preposition P_have which incorporates into be to yield have. In Mashi, however, Finholt (2024) argues that P_have is spelled out as a comitative prepositions, similar to Avelar (2004; 2009), Levinson (2011) and Ruvalcaba (2020).

In short, Finholt’s (2024) analysis is similar to the one proposed in Ruvalcaba (2020). Indeed, Myler (2018) also mentions that the Pred_stage head, like the one seen in (16a–b), is similar to other heads argued to derive stage level copulas (e.g., estar) such as Gallego & Uriagareka’s (2016) X head. Due to these similarities, if one applied Finholt’s (2024) analysis of Mashi to Spanish, one would run into the same problems already discussed for Ruvalcaba (2020).

Finholt’s (2024) Pred head which only takes individual level predicates accounts for Mashi’s jira, but it would overly restrict tener. Moreover, like Ruvalcaba (2020), Finholt (2024) supports an incorporation analysis by pointing to possession constructions in Mashi that include an overt copula (li or ba) as well as a comitative preposition (the spellout of P_have).¹⁰ As mentioned above, however, the lack of a productive ser con construction suggests there is no underlying P_have in Spanish’s tener verbs.¹¹ Moreover, it is unclear in Finholt’s (2024) analysis why P_have incorporates into v_be in some contexts and why it fails to incorporate in other contexts (where it surfaces as a comitative preposition). Apart from this, extending Finholt’s analysis to Spanish makes certain predictions about Spanish copulas ser and estar, namely that they emerge from the meaning of the Pred head. The problem with this assumption is that it conflicts with data of ser and estar that shows it does not neatly correspond to the individual vs. stage level opposition (Fábregas 2012). Additionally, it would not capture the notion that estar appears to be the more complex or marked form and that ser appears to be a more simplex and unmarked form, as has been argued by Gallego and Uriagareka (2016), Arche et al. (2017), Fábregas (2023) among others.¹²

An interesting aspect of Mashi is that the individual-level and stage-level contrast only happens in the present indicative context. This is true for copular constructions (16a) and possession constructions (16b). Finholt (2024) explains that in past tense, the temporary or stage-level copula li is used in all copular constructions, even when expressing properties of an individual. Similarly in possession constructions in the past tense, the stage-level possession verb dwiire is used in all cases, temporary or not. Furthermore, in subjunctive cases, the individual-level copula ba is overgeneralized. Similarly, for possession constructions in the subjunctive, the individual-level possession verb, but in its decomposed form, is used for all types of possession relations. By decomposed, Finholt refers to an individual-level copula and a comitative preposition (na, ‘with’). Finholt (2024) derives these patterns from the rules of exponence he proposes for Mashi (17).

These rules allow Finholt (2024) to capture the distribution of the ‘temporary possession’ verb dwiire and the ‘permanent possession’ verb jira. In this analysis, given the overgeneralization of dwiire in irrealis moods, it is seen as the elsewhere possession verb. Likewise, the ba copula is the exponent of the individual level copula (in the present indicative context) and there is a homophonous copula that is inserted elsewhere (in non-indicative contexts), regardless of which Pred head (stage or individual level) is present in the structure. These rules thus explain what form, whether copular or possession verb, will be used in which context.

The sensitivity of the individual-level and stage-level distinction in Mashi to tense and mood brings up the question of whether this is also true in Spanish. An interesting aspect of Spanish is that the individual-level and stage-level distinction is preserved in past tense, specifically in imperfective context. This is demonstrated in (18). In the sentences in (18), one can see that traer cannot express possession in the preterite (trajo) but it can express possession with imperfect(ive) past (traía). Because both are past tense, it suggests that the perfective aspect of trajo may be what prevents the stative possession reading.

Indeed, as one reviewer points out, stative verbs lack stative perfective readings. For example, (19a) lacks a stative reading. One may be able to get a dynamic reading from (19a) where somehow Juan went from a non-English speaker to an English speaker in the span of a moment. Likewise, (19b) lacks a stative reading of admirar, which is to hold in high esteem. Perhaps one may be able to get a stative reading in sentence like (19a–b) if one assumes that Juan is dead, but it is odd when uttered out of the blue.

The data in (19) suggests that the perfective aspect of the preterite is not incompatible solely with traer but with stative verbs in general.

Interestingly, modifiers like ‘as a child’ or ‘all his life’ can repair the data in (19). Thus, the stative verbs that appeared to be incompatible with perfective tense in (19) are acceptable in (20).

The same appears to be true for traer. The sentences that were illicit in (18) can are repaired in (21) shows with the addition of the phrases ‘for a long time’ in (21a) and ‘in her youth’ in (21b).¹³

In short, this data suggests that, unlike dwiire in Mashi, Spanish’s traer can express stative possession in the past tense.

Mashi’s copulas and possession verbs also lose the individual/stage-level distinction in irrealis moods. It is common knowledge that Spanish maintains the copular distinction across various moods. An example of the subjunctive is included in (22a). Like the copulas ser and estar, the possession verbs tener and traer also maintain the distinction in the subjunctive (22b).

Thus, one can see that Spanish’s verbs are also less restricted than the verbs and copulas in Mashi when it comes to mood.

In conclusion, there are several similarities between Mashi and Spanish. Mashi has copulas that resemble ser and estar, or ba and li respectively; the latter (estar and li) seem to be temporally bound or take stage level predicates. Mashi also has possession verbs that resemble tener and traer, or jira and dwiire respectively; the latter verbs (traer and dwiire) have similar restrictions as the so-called stage level copulas. There were also many differences between the two languages. Mashi’s possession verb, jira, is restricted to individual level predicates. Its Spanish counterpart, tener, is not; it takes stage level predicates when adequately contextualized. Additionally, the distribution of dwiire and jira only aligns with the stage/individual level opposition in the present tense and indicative mood. In contrast, the distribution of tener and traer can be found in the past tense and subjunctive mood. As is discussed above, these differences make it difficult to extend Finholt’s analysis of Mashi to account for the Spanish data of tener and traer.

4. A new way of deriving tener and traer from ser and estar structures

To account for the data discussed so far, the current analysis proposes that tener is the elsewhere possession verb and traer is a possession verb that is used in marked contexts. This idea stems from an analysis of ser and estar by Fábregas et al. (2023). They propose that some of the qualities of estar can be traced to a terminal coincidence head. In other words, estar is spelled out in ‘marked’ structure that contains a terminal coincidence head. In environments with no terminal coincidence head, ser is used. The tree in (23) provides an illustration of this idea.¹⁴ This tree shows that an Asp head with an uninterpretable terminal coincidence feature ([uT]) merges with PredP. For this Asp to be licensed, another head with a terminal coincidence feature [T] merges higher in the structure. In (23), the head with [T] is the copular head represented as V.

Fábregas et al. (2023) argue that estar has a [T] feature and ser does not. Consequently, whenever an Asp[uT] is present in the structure, it will 1) introduce a bounded or temporal reading, and 2) require estar to be present so it can be licensed. Whenever the Asp[uT] head is not present in the structure, or if said feature is checked by another head, then estar will not be required and V will be spelled out as the default ser.

The idea that the ser-estar opposition and its semantic contrasts are related to the central-terminal coincidence opposition has been proposed in other analyses (Brucart 2010; Arche 2012; Arche et al. 2017). All these analyses propose that both or one of the copulas are derived from a head, specifically a preposition, that has a central coincidence (e.g., in, on, with) or terminal coincidence (e.g., to, into, from) feature.¹⁵ For Arche et al. (2012) and Fábregas et al. (2023), estar is derived from the presence of a terminal coincidence head that ser lacks. This would make estar a ‘marked’ or more complex form, not unlike the complexity of estar proposed by Gallego & Uriagareka (2016). As described above, Fábregas et al. claim that this “markedness might reflect an extra prepositional specification of terminal coincidence, a feature that imposes a presupposition of boundedness, association to an external situation, etc…” (2023: 41). As such, this notion can also be extended to traer. In other words, if a terminal coincidence feature explains the temporal or bounded features of estar, one could argue that it could also explain similar characteristics in the possessive verb traer.

Within this approach, there is no intrinsic meaning of permanence for ser. Nevertheless, it tends to be used in constructions that denote permanence because such constructions lack a terminal coincidence feature, the same feature that is linked to temporality or presuppositions of boundedness or ties to an external circumstance. This unmarked ser structure is illustrated in (24). In contrast to the structure in (23), example (24) shows the lack of an Asp head with the uninterpretable terminal coincidence feature. As a consequence, V would be spelled out as ser.

In other cases, a different head (not the copular head) checks the uninterpretable feature in Asp. As such, ser is not necessarily restricted to a specific permanent or ‘individual-level’ reading. This is because ser is an underspecified form that appears when estar is not needed. An example of this is seen in (25). In this case, the uninterpretable terminal coincidence feature is checked by la boda, an eventive subject (Brucart 2010; Fábregas 2012; Fábregas et al 2023).¹⁶ Other examples like (25), where ser is not restricted to individual-level predicates, are discussed in Fábregas et al. (2023).

As mentioned above, tener is compatible with a range of interpretations, including permanent or individual level predicate interpretations. However, like ser, it is not restricted to these readings, and it can also denote temporary possession.¹⁷ In contrast to Ruvalcaba’s (2020) claim that ser and tener were restricted to permanent or individual level predicates, an alternative approach where ser and tener are analyzed as elsewhere forms allows us to account for instances where they do not take individual-level predicates.

Another feature that a revised analysis of tener and traer needs to account for is the lack of a ‘decomposed’ tener construction. Although Ruvalcaba’s account also predicts the existence of ser con, as a decomposed version of tener, none are presented in his analysis. In response, the current analysis argues that tener and traer are not formed by a comitative preposition.¹⁸ Instead, the current analysis argues that tener and traer are the transitive forms of ser and estar. This analysis follows the independently supported analysis of have verbs developed in Myler (2016). For Myler, have is the exponent of a v_be in contexts where it is embedded below a transitive Voice head. The rule of exponence for have (26b) and the structure (26a) is shown below. As one can see, have is spelled out when v_be is below a so-called expletive Voice head. This refers to a type of Voice that 1) has phi features, 2) introduces a subject in its specifier ({D}), and 3) fails to discharge a thematic role for this subject. It is important to note that, in this analysis, the transitivity of the Voice head derives the allomorph have, but the possession relation itself is introduced by a Possessor Phrase (PossP) lower in the structure; one process is largely independent of the other. In other words, Myler (2016) draws from previous analyses (Szabolcsi 1983; Kayne 1993; Partee 1999) in his claim that the possession relation stems from a PossP within the copular predicate. The copular predicate is essentially a possessive DP (e.g., ‘John’s car’) without a possessor. The lack of a possessor argument is demonstrated in (26a). As such, an unsaturated or ‘unsatiated’ relation is ‘passed up’ the structure until the transitive Voice head merges with vP and takes a subject. Thus, the subject in the specifier of expletive Voice is interpreted as the sentential subject and it also gets the thematic role introduced by the possessor head (Poss) in the DP.

If one were to propose an analysis like the one in (26) for Spanish, one could derive tener or traer without the presence for a comitative con preposition. This would explain why there is no link between tener and ser con constructions; there is no con in tener.

Myler (2016) pointed out that the con was unnecessary in the derivation of have within Levinson’s (2011) analysis. Since Ruvalcaba (2020) uses Levinson’s structure in his analysis of Spanish, Myler’s (2016) points are particularly relevant. Myler pointed out that the presence of a V that introduces the subject in Levinson’s (2011) account of have makes it an essentially transitive construction. Thus, con is not needed:

…the suggestion that P Incorporation is what triggers the spell-out of be as have is not necessary. One could equally well claim that it is the presence of v (which is independently necessary for Case-assignment, on Levinson’s assumptions) that triggers the realization of be as have. This would equate to the claim that have is the transitive form of be (Myler 2016: 391).

Thus, if con is unnecessary, we can propose (27) as the rule of insertion for tener. This rule states that tener is an exponent of v_be in the environment where it is embedded below an expletive voice head. This not only aligns with the previous observation that con is not present in tener but also that tener is the elsewhere possessive verb.

Note, for instance, that in this rule there is no restriction regarding the meaning or the types of predicates that tener can take. It is simply the spell out of a v, a head that Myler (2016; 2018) analyzes as an identity function. Based on this, one can propose a structure for tener like the one illustrated in (26). As the tree in (28) shows, the possessor role of Juan stems from the PossP in the DP. Since the PossP lacks a specifier,¹⁹ the possessor-less possessor relation is passed up the tree until it is saturated by the subject in the specifier of the expletive Voice head. Given the fact that v is below expletive Voice, tener is spelled out due to the rule of exponence in (27). As a default possession verb, tener will be glossed simply as have from this point on.

While this accounts for the underspecified possession verb tener in Spanish, it does not explain its alternative verb traer.

Recall that traer, like estar, is limited to temporary or stage-level predicates. Fábregas et al. (2023) proposed that estar was inserted to license a head with an uninterpretable terminal coincidence head. Indeed, one could propose a rule of exponence for estar like the one in (29a). This would contrast with the rule for the copula ser (29b).

Based on this analysis of estar, one could propose that traer is the spell-out of a v below an expletive Voice head, like tener, but also above a head with an uninterpretable terminal coincidence feature, like estar (30).²⁰ Note that the terminal coincidence feature on v and the presence of a transitive Voice are independent aspects of this derivation, and their only tie is that they create an environment in which traer is exponed.

This captures the fact that traer appears in more restricted environments, namely those with particular temporal properties introduced by Asp_[uT]. This rule, and the one for tener, also captures the fact that the heads are not restricted to a particular modality or tense like possession verbs are in Mashi. The proposed structure for traer is illustrated in (31). Here, I adopt Finholt’s practice of glossing traer-like verbs with have_temp.

The Asp with the terminal coincidence feature below v could also vary, or it could be different types of terminal coincidence P’s present, like de (from) or a (to) etc. This could explain other verbs Ruvalcaba (2020) claims are ambiguous between dynamic and stative possession verbs (32a–b).

Alternatively, one could propose that the v is distinct in (32).²¹ A more in-depth account of other stative possession verbs will be set aside for future research.

Lastly, there is the question of how to derive estar con constructions. While the current analysis does not assume that traer and estar con constructions are related, it aligns with Ruvalcaba’s analysis (2020) that estar con constructions are derived from a pP structure (36).²² The subjects in these constructions are introduced by a functional p that merges with a PP headed by con, as shown in (33). Additionally, the possessor relation can be encoded in the p or it can emerge from the accompaniment semantics of con.²³ The only difference between the current analysis and the one proposed by Ruvalcaba (2020) is that now there is an uninterpretable terminal coincidence feature that triggers estar.

To reiterate, the current analysis does not treat traer and estar con as derivationally related. The former gets its possessive reading from a PossP and the latter gets it from the p and P heads in the pP. The fact that the possessive meaning has different sources in each construction can help to explain some of the discrepancies between the two constructions. As one reviewer notes, some traer constructions (e.g., *los niños traen diabetes, the children have diabetes) are less acceptable than their estar con counterpart (i.e., */?los niños están con diabetes, the children are with diabetes). It is possible that the difference in the level of acceptability of these two constructions may be due to their unique structures.

Here again, it is important to note that terminal coincidence heads vary (Hale and Keyser 1984; 2002). If a different Asp or P headed AspP, one could have a different type of terminal coincidence being expressed. Arche describes two subclasses of non-central coincidence prepositions: allative and ellative. The former includes to, up to, onto, into and is associated with a future/prospective reading (2012: 124). The latter includes from, out of, of, and is linked to past/perfect readings. Indeed, estar is believed to be formed from the ellative type by Arche (2012). As mentioned for llevar constructions (32) it is possible that v may be spelled out as a different form according to the type of terminal coincidence feature it merges above. For instance, as a reviewer points out, there is another construction that resembles estar con construction with the verb venir (‘to come’) as shown in (34a). Another construction one could add is the andar con sentence in (34b).

Here, as in the cases with estar con constructions, the preposition con is the source of the possessive meaning. Thus, one could propose a structure similar to (33). As mentioned above, the difference in verbs could be attributed to different Asp or P heads below v, each expressing a different type of terminal coincidence. More precise details of such an analysis are left for future work.

5. The tener-traer distinction and complex predicates

Additional evidence of the ser-tener and estar-traer parallels can be found in sentences with complex predicates. Small clauses are known to be restricted by the matrix verb. Jiménez-Fernández (2012) mentions that Spanish verbs considerar (to consider) and esperar (to expect) each restrict for a particular type of small clause. Specifically, the former restricts for an individual level predicate in the small clause and the latter restricts for a stage level predicate in the small clause. This is illustrated in (35–36). As one can see, the verb considerar (‘to consider’) combines with a small clause that has an individual-level predicate (35a) but not with small clauses that have a stage level predicate (35b). Jiménez-Fernández (2012) also shows that the opposite is true for the verb esperar (36a–b).

The verbs tener and traer can also take small clause complements. Interestingly, the choice of tener or traer influences the reading of the small clause.

Some varieties of Spanish can express inalienable relations between the subject and object without a possessive pronoun.²⁴ For instance, the translation for the English Mary raised her hand would literally be Mary raised the hand. In the sentences in (37a,c,e) the subject of the matrix clause and the subject of the small clause are inalienably related. In these cases, the small clause predicate expresses a property or a state of the small clause subject, namely the subject’s eye color or condition. Whether it is a property or a state can be determined by whether the main verb is tener or traer. The examples in (37b,d) include comparable copular sentences to demonstrate the similarities of these two constructions. One can see that with estar and traer, the readings are restricted to states rather than intrinsic properties of the individual. Thus, having red eyes can only mean that they are bloodshot with traer (37c) or that the individual is wearing contact lenses. This is comparable to the estar construction (37d). In contrast, tener, allows the adjective rojos to function as an individual-level predicates where the person’s eyes are naturally red (37a).²⁵ This meaning can be interchangeable with ser constructions (37b). The meanings of tener sentences can also have temporary or stage level readings. This aligns with the current analysis which treats tener as the elsewhere possessive verb. In contrast, traer cannot express a reading where the color refers to the person’s natural eye color. With a different color (37e) the bloodshot reading is not possible, and the only meaning available is that Juan recently put on contact lenses that changed the color of his eyes.

The presupposition of change in (37e) is similar to the way in which estar constructions can presuppose a different or changed state (Romero 2009, as cited in Fábregas 2012). For example, (38a) means that a property of the house is the color white. The sentence in (38b) can express a similar meaning, but it presupposes that the house was previously a different color. This may be linked to Arche et al.’s (2017) claim that estar is tied to utterance time.²⁶

A similar pattern is observed for traer in (37c) and in the data in (39). For example, when traer is present, having black hair in (39c) it implies that Juan colored it black, but it may have been a different color in the past,²⁷ similar to the estar construction (39d).²⁸ With tener (39a), the small clause predicate can refer to the property of the individual, Juan, like ser in (39b). For this reason, the meaning is one where black is the natural color of his hair. Additionally, tener is not incompatible with the reading where Juan colored his hair either. This aligns with our analysis of tener as the elsewhere possessive verb.

Given the underspecification of tener, the temporary reading of tener constructions can stem from a variety of sources, including the surrounding context and discourse. This contrasts with a similar reading in traer sentences, which stems from the terminal coincidence feature below v. Thus, the obligatory temporary reading and the presupposition of change one finds only in estar and traer constructions further support the idea that they are derivationally related.

Additionally, one can see that in (40) the predicate of the embedded small clause can express a property of the small clause subject. In other words, tener allows the small clause to have an individual-level predicate reading (40a). The opposite is true with traer. With traer in (40b) the reading is one where the object, Pedro, is currently a cook, but that is not necessarily his profession. In other words, it refers to their current state.

In contrast to (37a,c,e), the data in (40) have complex predicates with an embedded PP Another instance of small clauses with PP predicates are those in (41).

The predicates of the small clauses in (41) appear to refer to a state of the students in cases when traer is used, but they can have a reading where the locative PP can be a property of the students when tener is used. With tener, Juan can be in a different city (41a). For example, he can be the director or founder of a study abroad program in Tucson that helps students travel abroad for their studies. In this way, he is indirectly responsible for their presence in Tijuana, and the locative phrase can express a property about the students, namely that they are study abroad students located in Tijuana. In contrast, in (41b) Juan has the students with him at utterance time. They are not necessarily study abroad students. In this case, ‘in Tijuana’ refers to the state of the students; they are under Juan’s care in Tijuana at utterance time or at some specific spatio-temporal location.²⁹

The link between the state denoted by traer (41b) and utterance time resembles the connection between estar and utterance time, as described by Arche et al. (2017). Note, also, that these sentences do not necessarily express possession. As will be discussed in more detail below, the subjects of these sentences (40–41) appear to have a causative reading, a thematic role not observed in previously discussed constructions.

The data in (42) are more straightforward causative have constructions. One can see that traer constructions behave like estar with respect to their incompatibility with perfective aspect or habituality (see Arche et al. 2017 for discussion on this incompatibility). In (42) tener and traer take a small clause with a VP predicate. One can see that bare plural objects are preferable for traer (42a,c). A definite object like ‘a dozen tamales’ degrades the sentence (42b,d). Nevertheless, a definite object is fine with tener.

If one agrees with Arche et al. (2017) that estar, with regards to viewpoint aspect, tends to be continuous, and if one assumes that traer is derived similarly to estar, then the behavior of traer in (42) would be expected. This is because the bare plural allows a repetitive or ongoing action, such as (42a) where the students are making an unspecified number of tamales; they are just involved in the ongoing action of tamale making. When one has a definite object, the action has an endpoint that may conflict with the aspectually continuous nature of traer. Nevertheless, definite small clause predicates are more acceptable with tener. This adds to the ser-tener and estar-traer parallels. Moreover, similar to estar constructions and traer sentences with simple predicates, examples like (42d) are more acceptable when further context is provided. For example, the sentence in (43) is the same as (42d) except that this time additional information is included. Now, the clause is linked to external circumstances, namely a particular spatiotemporal location that contrasts with a prior state.

This, like the other complex predicate examples, further demonstrate the ser-tener and estar-traer parallels. In addition, the structural analysis proposed in the previous section can be extended to these constructions.

One way to account for traer sentences like (43) is to propose that the Asp head with a terminal coincidence feature merges with a small clause or PredP, as shown in (44). As explained previously, the terminal coincidence feature on v licenses the uninterpretable feature on the Asp head. To capture the causative reading, one can propose that there is a Voice head that introduces an external argument ({D}) with a causer role. This contrasts with the expletive Voice head for tener and traer with simplex predicates which did not assign a thematic role to its external argument (28, 31). Nevertheless, the Voice head in (44) is still transitive and, consequently, traer, is spelled out. If the v lacked an interpretable terminal coincidence feature, it would be spelled out as tener. Thus, the same sentence with tener would differ from (44) in the sense that the tener construction would lack an AspP (as well as a v with a terminal coincidence that licenses the Asp head). A similar structure can be proposed for the causative readings one finds with PP predicates (40–41). However, with respect to the inalienable sentences in (37a, c) and (39a, c), an analysis like (44) is possible but with a different Voice head, namely the expletive voice head which does not assign a thematic role to the argument it introduces.

The current analysis does not make a strong claim about why the matrix verb is sensitive to the predicate of the lower small clause. Jiménez-Fernández (2012) proposed a topicalization analysis to account for the stage vs. individual level predicate interpretation of small clauses. In said analysis, small clauses with individual level predicates are simple structures consisting of only the subject and predicate. As such, the subject of the small clause moves into the specifier of a Topic Phrase [_TopicP subject_i [Topic [_PredP subject_i [ predicate]]]. In contrast, small clauses with stage level predicates are said to be more complex, and so these sentences allow the small clause subject or some other element in the structure to occupy the topic position [_TopicP X [Topic [_PredP …X…]]]. Another potential explanation is that the lower small clause head incorporates into v. Here, one could follow den Dikken’s (2006) proposal that the minimal domain of a head movement chain (α, t) “includes the maximal projection of the raised head (which is included in the minimal domain of the host β of the raised head α” (den Dikken 2006: 114). Thus, in (44) the complement of the trace is equidistant from the complement of v since v hosts the trace’s antecedent. Thus, the VP in (44) is no different than typical objects of tener or traer. As such, one could propose that the temporal restrictions of traer with simplex complements are linked to predicate of the small clause.³⁰ Beyond this, a more complete analysis of this process is left for future research.

Whichever way one captures the sensitivity of tener and traer to the embedded predicate, the data presented so far brings to light new questions about analyses like the ones proposed by Ruvalcaba (2020) and Finholt (2024). If one takes a look at estar con constructions, one can see they cannot express a causative reading. For example in (45a–b) Juan is simply accompanying the students in their chores.

In Ruvalcaba’s (2020) analysis, the verb traer is essentially estar without a comitative preposition incorporated into it. In other words, the Voice head has nothing to do with whether traer or estar is spelled out. In such an analysis, it is unclear why traer can have causative reading with complex predicates but estar cannot. In other words, the lack of with incorporation cannot account for its incompatibility with causative readings.³¹ In the current analysis, estar cannot express a causative reading because when there is a causative Voice in the structure (44), v is spelled out as traer or tener.

To further support the presence of a causative Voice head in (44), one can see that the subject of traer and tener can be non-agentive. For example, in (46a) one has an inanimate subject ‘house.’ This is licit in a context where prospective buyers are visiting an open house, and the house of interest turns out to be constructed like a maze. Likewise, the sentence in (46b) has an inanimate subject, yet it is interpreted as the causer of the state denoted by the small clause.

The verb tener would be equally acceptable in (46a–b). As such, one can see that there is a clear causer role in these sentences. One would therefore be able to extend an analysis like (44) to these sentences. Here again, one would have a Voice head that introduces a causer subject. This contrasts with tener or traer sentences with simplex predicates where an expletive Voice head merges with vP and introduces a subject but does not discharge a thematic role. Since the Voice heads for simplex and complex predicates are both transitive, tener or traer can be inserted into v in both cases.

This analysis has shown that the tener and traer verbs are sensitive to the predicate of a small clause complement. This further demonstrated the nuanced differences between tener and traer. The data aligned with what was observed for tener and traer sentences with simplex predicates. First, traer functions like tener in that it can take a small clause complement without expressing a dynamic ‘bring’ reading. Moreover, in contrast to tener, traer is limited to sentences with complex predicate that are temporally bound and linked to external circumstances. The verb tener is not restricted, which aligns with its analysis as the elsewhere possession verb. The section also discussed so-called causative have constructions in Spanish where the matrix subject is the causer of the state of affairs denoted by the PredP. These are derived by a transitive Voice head, namely a causative Voice head, that merges with vP. This contrasts with the expletive Voice head proposed for non-causative sentences.

6. Discussion and Conclusion

This paper has proposed a new way of explaining the possession verbs tener and traer in the Spanish spoken in northern Mexico and the southwestern U.S. This explanation captures the parallels between these verbs and the ser and estar copulas. It showed that, contrary to the claims made in Ruvalcaba (2020), tener is not restricted to individual level predicates. Instead, following the analysis of ser by Fábregas et al. (2023), the current analysis argued that tener is the elsewhere possession verb. In contrast, it showed that traer is temporally bound and linked to external circumstances, as has been argued for estar. This is due to the presence of an uninterpretable terminal coincidence feature that must be licensed by a head with an interpretable terminal coincidence feature. The head with this feature is spelled out by estar in intransitive contexts and traer in transitive contexts. Likewise, the v without the feature is spelled out as ser in intransitive contexts and tener in transitive contexts (when the vP is a complement of an expletive Voice head).

This analysis accounts for the data discussed in Ruvalcaba (2020) in addition to the new data with complex predicates in Section 5. Moreover, it explains some aspects of the data that are not explained in analyses like those proposed in Ruvalcaba (2020) and Finholt (2024) both of which propose that the possessive verbs are copulas with a comitative preposition incorporated into them. For instance, the current analysis accounts for the discrepancies between traer and estar con constructions, such as the inability of the latter to express a causative reading with a small clause complement as well as the differing levels of acceptability between each construction despite having the same predicate. It also accounts for the absence of ser con constructions which is predicted by the idea that tener consists of ser and con, as well as by the analysis of estar con sentences, which are more productive, as decomposed traer constructions. All this is due to the fact that con is not involved in the derivation of tener or traer. These verbs derived from a transitive Voice head.

Section 3 discussed the parallels and differences between Spanish and Mashi. For example, it showed that Mashi, like Spanish, has a copula, li, that is restricted to stage-level predicates. Like traer, Mashi also has a possession verb, dwiire, that appears to be restricted to temporary possession or a possession relation that is tied to external circumstances. In contrast, Mashi has a copula, ba, that is restricted to individual level predicates. This is related to Spanish ser, which tends to appear with individual-level predicates, although this is due to the fact that it cannot license a terminal coincidence feature in the structure (Fábregas et al. 2023). Likewise, Mashi has a possession verb, jira, that is restricted to individual-level predicates. This is similar to tener which can appear with individual level predicates, such as kinship, part-whole relations, ownership, and other forms of permanent possession. However, it was shown that tener does not appear to be as restricted as jira, its Mashi counterpart. This is because tener is an elsewhere possession verb. In Finholt’s (2024) analysis, Mashi’s individual-level copula, ba, and possession verb, jira, are spelled out in the syntactic context where there is a PredP headed by a Pred head that only takes individual level predicates. In the current analysis, the individual and stage level distinction is not assumed to be encoded in the Pred head. Instead, it proposes that estar and traer constructions are spell-out in marked structures where a terminal head feature or head present. Both ser and tener are analyzed as elsewhere forms.

If Mashi’s individual level forms, ba and jira, are indeed categorically restricted to individual level predicates, then it is possible that these are derived by the presence of a central coincidence feature or head in the structure. Hale and Keyser connected the ability of certain prepositions to express stasis or static relations in small clauses (e.g., “with the baby in/*to/*into bed, we can relax”) to atelicity and central coincidence. They stated that prepositions that can express this stative relation in small clauses “share the property of expressing the relation of central coincidence, holding between the figure…and the place.” (1984; 2002: 218). In this way, they link stativity to a property of certain heads, namely prepositions such as in, on, with. Likewise, Brucart (2010) argued that ser, because it tends to appear to individual level predicates, was spelled out in the presence of a central coincidence feature. While the current analysis assumes that ser is a neutral element, it is possible this explanation applies to Mashi. In other words, Mashi’s copula, ba, is spelled out in the presence of a central coincidence feature. This would account for the difference between ser and Mashi’s copula. This can also explain the restrictions demonstrated by Mashi’s permanent possession verb, jira, which are not found with Spanish tener.

Finally, the analysis showed that Spanish lacks some of restrictions Mashi had with respect certain inflectional contexts. For instance, the distinction between the individual and stage level copulas (ba and li) as well as the individual and stage level possession verbs (jira and dwiire) was only made in the present tense and indicative mood. Spanish, on the other hand, allowed the ser-estar and tener-traer distinctions beyond this tense and mood. It was shown that Spanish differentiated between these forms in the subjunctive and past tense. While the preterite appeared to disallow the distinction between tener and trear, it was shown that this was resolved when the sentence was sufficiently contextualized.

While several questions remain about possession in Spanish and Mashi, the aim of the current analysis is to not only offer potential answers but to also open up new lines of inquiry. Future analyses will need to provide a more in-depth account of the aspectual and pragmatic properties of estar and traer. Likewise, a more systematic account is needed to describe the contexts that allow tener to have temporary readings in some cases and permanent readings in others. Finally, a more detailed analysis is needed of the heads involved in possession verbs as well as their connection to the stage versus individual level readings of complex predicates. However, it is hoped that the current analysis has presented enough observations and interesting questions to advance our understanding of Spanish possession verbs.

Abbreviations

Acknowledgments

I would like to thank Andrew Carnie, Heidi Harley, Ivy Sichel, and Neil Myler for helping me with questions that came up in the process of writing this paper. I would also like to thank the three anonymous reviewers for their valuable and constructive feedback.

Competing interests

The author has no competing interests to declare.