Force dynamics as the path to the Spanish subjunctive

2.1 Mejías-Bikandi’s (2014) hypothesis: the “out of reach” complement

Mejías-Bikandi (2014: 653) admits that his previous model faced trouble in accounting for mood choice in causative predicates, which never allow for the use of the indicative in their complements, not even in those cases where it seems to be required by the speaker’s both epistemic and pragmatic assessment.

To overcome this inconvenience, he first – building on both Langacker (1991) and Cristofaro (2003) – makes a distinction between two types of subordination: grammatical and conceptual subordination (Mejías-Bikandi 2014: 660). And then, poses a specific explanation for the complements of causative verbs, by which mood choice is presented as the result of these complements being both grammatically and conceptually subordinated to the main clause.

As for the notion of conceptual subordination, both Mejías-Bikandi (2014) and Cristofaro (2003) adopt Langacker’s definition of a subordinate clause as “one whose profile is overridden by that of a main clause” (1991: 436). Cristofaro (2003: 29), though, admits that Langacker provides no criteria to identify the cognitive properties of such kind of subordination independently of the clause types in which they are manifested; and so, she opts for a pragmatic definition of the term, as ‘lack of assertiveness’, as her starting point. Mejías-Bikandi, in contrast, makes both assertion and conceptual subordination ultimately dependent on grounding: “a necessary condition, although not a sufficient condition, for a clause to be asserted is that it receives an autonomous profile, and in order to receive an autonomous profile, it must be independently grounded” (2014: 661).

According to Langacker, the term ground indicates “the speech event, its participants (speaker and hearer), their interaction, and the immediate circumstances (notably, the time and place of speaking)” (2008: 259). A finite clause constitutes an instance of a grounded process (438), tense and modality being regarded as its grounding elements: “Tense indicates the temporal relation holding between the process described and the speech situation, and the presence or absence of a modal gives an indication of the epistemic status that the process described holds for the participants” (whether it is part of their view of reality) (Mejías-Bikandi 2014: 658). Non-finite verb forms, on the contrary, do not constitute instances of a grounded process (Langacker 2008: 438).

As for subjunctive clauses, there seems to be a considerable degree of disagreement among cognitive linguists regarding their particular status in relation to grounding. Thus, Maldonado (1995: 406) considers them not to be grounded. Achard, in contrast, claims that they are, though not to the speech event, but to some mental space (see Fauconnier 1985) set up by the matrix clause (Achard 1998: 70, 2000, 2002). Langacker, for his part, goes no further than stating that they “are intermediate between finite and non-finite clauses” (2009: 303). Finally, Mejías-Bikandi’s position is that they are generally grounded, except for the case of the complement of directives and matrices of volition and causation.

To hold this view, he first resorts to the idea that “with non-finite complements there tends to be a fixed temporal relation established between the process described by the complement and the process described by the main clause. Specifically in the case of infinitival complements, the event described by the complement is future with respect to the one expressed by the main clause” (2014: 658). And next, he points to the fact that directives and matrices of volition “determine, by virtue of their meaning, the temporal relation that holds between the event described by the complement and the reference point introduced by the main verb” (Mejías-Bikandi’s 2014: 657). Having established a correlation between this kind of temporal relation and ungrounded, non-finite complements, he, therefore, concludes that the complements of these matrices, as well as those of causative ones, are not independently grounded either, since “their temporal relation to the speech situation is not independently indicated by the complement verb, but rather, is established via the matrix verb” (659). Not being independently grounded – so the argument goes – these complements can neither be said to have an autonomous profile nor be asserted (651).

Through this line of argument, Mejías-Bikandi adds a new scenario to his previous model: some clauses are not asserted, even when they include information regarded as both true and new, just because they lack an autonomous profile, and so they somehow remain “out of reach” – so to say – for the speaker to assert. This is in no way a minor addition. Remember he started by defining mood as the direct consequence of the speaker’s assessment of the information given. Finally, though, he ends up making a sharp distinction between, on the one hand, those predicates that allow for a choice of mood consistent with the speaker’s assessment of the subordinate proposition, and, on the other, those that do not. The general view, then, seems to be not that of grammatical mood as always a consequence of the speaker’s assessment, but instead that of different syntactic and conceptual contexts showing different possibilities and rationales concerning mood choice in Spanish, one of them mirroring the speaker’s assessment, but not the other.

2.2 Criticism on Mejías-Bikandi’s (2014)

As we have just seen, Mejías-Bikandi’s whole account of mood choice in the context of the complements of directives and matrices of volition and causation ultimately depends on his argument that these complements cannot be considered as independently grounded because their “time relation to the ground is established via the matrix verb” (2014: 661). However, this last (and crucial) point in his reasoning seems highly problematic. Let us look at why.

First of all, as previously shown, he presents both modality and tense as the grounding elements for clauses. Grounding, on its turn, is defined in direct relation to the speech situation. However, as Langacker (2008: 445–453) points out, the conceptualizer invoked by a grounding element, as far as modality is concerned, is just a virtual entity, which cannot generally be identified with the speaker or the hearer. Complement clauses, in fact, are frequently not presented as assessed epistemically from the speaker’s viewpoint, but from that of the matrix subject (Achard 1998; Langacker 2008: 449–450; Maldonado 1995: 406). Being aware of this fact, Mejías-Bikandi admits that, in a sentence like (3), the complement clause is “located as part of the immediate reality of some conceptualizer” which does not correspond to the speaker, but to “my brother” (2014: 658–659). Nonetheless, he still considers it to be grounded – indirectly, via the matrix clause.

The necessary conclusion here is that, according to Mejías-Bikandi’s conception, modality per se is always to be considered as a grounding element, either direct or indirect. Unless it is blocked by tense, as it will be seen next.

In contrast to modality, and despite its acknowledged status as a grounding element, tense is not regarded by Mejías-Bikandi as always serving this function. And so, he considers the complement clause in a sentence like (1) as not being “independently grounded,” because, allegedly, its “temporal relation to the speech situation is not independently indicated by the complement verb, but rather, is established via the matrix verb” (2014: 659). Note that, in claiming this, he not only appears to deprive the past subjunctive tense of the ability to directly link an event to the speech situation but, curiously enough, he also refuses to acknowledge indirect reference as grounding for tense – whereas that same phenomenon was accepted in (3) as proper grounding for modality.

According to his analysis, then, the complement clause que Pedro llegara antes de medianoche in (4) should be considered to be independently grounded – even though it is not presented as epistemically assessed by the speaker, but by Luis – because there is no pre-determined temporal relation between the event described by the complement and the matrix clause.

Conversely, that same complement clause in (5) should be considered as not independently grounded, not because of the lack of an epistemic assessment by the speaker on that clause, but because the temporal relation between the event described by the complement and the matrix clause is said to be pre-determined by the matrix predicate.

Ultimately, Mejías-Bikandi’s analysis requires us to accept that the complement in (5) is not grounded despite the facts that (1) it includes a subjunctive verb form, generally understood as an expression of modality; (2) it is also marked by tense; and (3) he defines both modality and tense as grounding devices.

Another problem arising from Mejías-Bikandi’s (2014) analysis pertains to the meaning of tenses. Spanish tenses, including those of the subjunctive, have been shown – contrary to what Achard (2000: 159) claims – to always convey information on the location of a process as seen from the speaker’s viewpoint (see Ruiz Campillo 2019; Suñer 1990: 77–105; Suñer and Padilla-Rivera 1987).^[8] However, Mejías-Bikandi holds that the causative complements’ “temporal relation to the speech situation is not independently indicated by the complement verb” (2014: 659). He insists that, by stating this, he is not claiming “that the tense of the subordinate verb is determined by the form of the matrix verb” (657). But then, if tense in these complements does not indicate the “temporal relation to the speech situation,” what sort of information, if any, is it supposed to provide? Should we conclude that the past subjunctive in llegara ‘arrive’ means ‘previous to the speech situation’ in (4) but ‘posterior to an event located in the past’ in (5)? Or, instead, should we assume that the past subjunctive means just nothing in (5)?

Regarding the former alternative, it seems unreasonable to consider that a verbal form such as the past subjunctive, which usually conveys some meaning, does not in some other context, even when there is no reason to think so: it equally indicates distance from the ground in both (4) and (5).^[9] As for the latter, we can indeed find this line of thought within the framework of European structuralism, where meaning is understood as the sole result of paradigmatic oppositions (see Saussure 1991 [1916]: 147). As a consequence, if a certain linguistic unit does not contrast with any other in some context, that unit is considered by some grammarians to be void of any function or meaning in that context (see, for instance, Bull 1965: 174). However, such reasoning is incompatible with cognitive linguistics’ conception of the linguistic sign as an association of a linguistic form^[10] with a meaning, such that one is able to evoke the other (Evans and Green 2006: 12–13; Langacker 2008: 5).

In any case, the point here is that the complement clause of the matrices analyzed by Mejías-Bikandi does allow for some contrast between tenses. Indeed, the complements of the so-called verbs of influence generally refer to an event understood as “future with respect to the moment in which the request, command, advice, or want is expressed” (2014: 65; see also Kempchinsky 1990: 238; RAE-ASALE 2009: § 26.4e–f).^[11] However, this does not mean that the subordinate verb tense in those cases loses its ability to locate the subordinate process directly, from the vantage point provided by the speech situation, as he concludes. On the contrary, these constructions, when including a past tense matrix verb, still allow for a choice between the past and the present subjunctive in the complement (Bosque 1990: 58–60; Hummel 2001: 257–260; Kempchinsky 1990: 238–239; Suñer and Padilla-Rivera 1987: 639), as shown in (6).

These alternatives are also available in the case of causatives, as evidenced by the real examples of use presented in (7) and (8):

As can be seen, in (6) the past subjunctive indicates distance from the ground, as opposed to the present subjunctive. In sentences such as this, as well as in (7) and (8), what is meant by the use of the present subjunctive, as Suñer and Padilla-Rivera put it, “is that the subjunctive action is not just subsequent with respect to the main verb action (which could be expressed with a past subjunctive), but also subsequent to the time of communication (which cannot be expressed with a past subjunctive). Hence, the present subjunctive is also used here in a meaningful way” (1987: 639–640; emphasis mine). This evidence appears to refute Mejías-Bikandi’s argument that the temporal relation between the complement of causative matrices and the speech situation can only be established via the matrix predicate.

Finally, Mejías-Bikandi’s (2014) proposal, as a whole, can be said to be not only problematic from a theoretical point of view, but also counterintuitive, to the extent that it portrays the use of the subjunctive in complements of causative verbs as a marginal phenomenon, excluding them from the conceptual category they naturally belong to, that one which some grammarians (as will be shown later) have rightly understood as the central use of the Spanish subjunctive.

3.1 Force dynamics

Within the framework of cognitive linguistics, a linguistic unit is regarded as a conventional assembly of a semantic structure with a linguistic form, also known as construction (Goldberg 1995). This is so irrespective of their level of formal complexity – from minimal symbolic elements (morphemes) to larger conventional patterns such as the ditransitive construction or the CNICCC, here analyzed (Evans 2007: 42–43; Evans and Green 2006: Chs. 17, 19–20; Langacker 2008: Ch. 6).

On the other hand, this school holds that semantics is conceptualization. According to this view, linguistic units are associated with a wide range of construal operations, whereby an objective situation can be conceptualized in different ways (see Croft and Cruse 2004: 40–73; Langacker 1987, 2008: cap. 3; Talmy 2000: 40–84; Verhagen 2007). Force dynamics (FD henceforth) can be regarded as one important construal operation.

FD was proposed in the late ’80s by Talmy (2000: Ch. 7) as one of the basic schematic systems which make up the human conceptual structuring system. Understood as a generalization over the concept of causality, it presents a situation as a force interaction between two participants: the agonist (the one that shows an intrinsic tendency towards either action or inaction), and the antagonist (the one that either exerts or keeps from exerting its force against the agonist’s intrinsic force tendency). Crucially, as Boye highlights, a force not only needs something to act on (the agonist), but also “a goal, against which it is directed, and a resultant” (2005: 58). This resultant – Talmy adds – is one “either of action or of inaction, and it is assessed solely for the agonist, the entity whose circumstance is at issue” (2000: 415).

From these components, Talmy derives two major categories of FD patterns: the causative category, which always involves the agonist’s resultant state of activity being the opposite of its intrinsic actional tendency: action, as in (9), or inaction, as in (10); and the letting/concessive category, which always involves the agonist manifesting its intrinsic actional tendency: action, as in (11), or inaction, as in (12). Here I offer some examples of these categories:

All these previous examples constitute quite simple instances of the FD system since they all (1) refer to the domain of physical force interactions; (2) make explicit reference to all the factors involved in the FD system; and (3) include a non-sentient entity as the antagonist. However, as Talmy (2000: 435–436) explains, FD patterns are commonly not so simple. On the one hand, the FD system not only applies to “physical force interactions but, by metaphoric extension, also to psychological and social interactions, conceived in terms of psychosocial ‘pressures’ ” (409) (cf. Sweetser 1984: 61). On the other hand, FD patterns have the conceptual structure of a script, i.e., that of a complex pattern changing through time. In the particular case of FD patterns including a sentient entity as the antagonist, this script necessarily starts with a volitional act, which is nothing but another force interaction (of a psychological kind), embedded within the main one (Talmy 2000: 421). As for the main interaction, Talmy distinguishes the following phases:

… the first phase, (a), is a stretch during which a sentient antagonist, foregrounded as subject, impinges extendedly on a stronger agonist, intending that this will make it act as shown in the subsequent phases. The (a) phase may include a latter portion, (a’), during which the agonist weakens or the antagonist strengthens. In the punctual (b) phase, a criterial shift in relative strength takes place. Phase (c) is the aftermath of this shift, with the agonist now forced to manifest the intended action. (436)

As a result, expressions referring to an FD pattern can be distinguished in relation to two factors: (1) phase, i.e., the location along the temporal sequence at which focal attention is placed, and (2) factivity, that is, the occurrence or nonoccurrence of portions of the sequence and the speaker’s knowledge about this (Talmy 2000: 436). Thus, for example, “the lexical verb try involves focus at the initial phase without knowledge of its outcome, while succeed and fail focus on a known occurrent or nonoccurrent outcome” (436). Matrices of volition and intention (either positive or negative), as well as Spanish lexical items such as intentar ‘try’, resistirse a ‘resist’, ceder a ‘give in to’, conseguir ‘get’, etc. are particularly specialized in alluding to different phases of the kinds of FD patterns including a sentient entity as the antagonist. Here we have some examples:

As it can be seen, in all Talmy’s examples reproduced in this section, the complement clause takes the subjunctive in its translation into Spanish.^[13] By systematically describing the underlying conceptualization common to all of these sentences, Talmy’s (2000) FD system helps to clarify the rationale they follow with regard to mood choice in Spanish. Remarkably, this relationship had already been described by Marathon M. Ramsey, back at the end of the nineteenth century, in terms not so different from Talmy’s: “The principal use of the subjunctive is after verbs expressing an action calculated to cause another person or thing to act … What is true of the above is equally applicable to verbs of opposite effect, tending to prevent another from doing something” (Ramsey 1894: 321).^[14]

From the previous analysis, one could be tempted to conclude that, within an FD pattern, the verb expressing the goal process always takes either the subjunctive or the infinitive in Spanish. However, that is not the case. In fact, in some of the FD patterns presented by Talmy (2000), the verb referring to the goal process – the one involving the agonist – can only take the indicative when translated into Spanish, as shown in the following examples:

What is then the crucial factor determining mood choice in the case of the CNICCC? To answer this question, two other concepts will be introduced in the next section: focusing and viewpoint.

3.2 The CNICCC, focusing, and viewpoint

3.3 Directives, verbal force, and beyond

Bull shows that the relationship between causation and the Spanish subjunctive may concern verbs of speech as well, to the extent that speaking can be presented as causing or leading “to the resulting event” (1965: 187; see also Ahern 2008: 27–30; Terrell and Hooper 1974: 490–492). Crucially, besides directives, other verbs (and expressions) of communication, even those describing the manner of speaking (such as gritar ‘shout’, insistir ‘insist’, clamar ‘cried out’, etc.) or how information is communicated (such as escribir ‘write’, notificar ‘notify’, etc.) can also be used in Spanish to convey the idea that “the subject of the main verb attempts to influence (cause) the action of the subject of the dependent verb by speech: Insisto en que vaya [‘I insist him to go’]” (Bull 1965: 188; see also RAE-ASALE 2009: § 25.3n, 25.4d). Thus, the same verb or expression (such as susurrar ‘whisper’) can trigger both the indicative, as in (30), and the subjunctive, as in (31).

To explain this phenomenon, those approaches to mood which understand this verbal feature as the consequence of the meaning of some kind of matrix need to attribute two different senses to all of these verbs, one related to the notions of assertion and report, and another for commands. However, resorting to homonymy is not a satisfactory solution in this case, as both Bosque (1990: 44–45) and RAE-ASALE (2009: § 25.4f) state, unless it manages to account for both the reason for such a homonymy and its systematicity in Spanish, i.e., to provide a proper explanation for the fact that all (affirmative forms of) verbs of speech are invariably understood as causative when complemented by a subjunctive clause.

In contrast, this kind of linguistic behavior constitutes no problem for the model presented here, since – as shown in Section 3.1 – Talmy also refers to the metaphoric extension of the FD framework to interpsychological interactions between sentient entities, whereby “one sentient entity’s production of stimuli, including communication, … is perceived by another sentient entity and interpreted as reason for volitionally performing a particular action” (2000: 438).

As it has been previously shown, within any kind of FD pattern, foregrounding the force exerted by the antagonist triggers the use of either the subjunctive or the infinitive in the complement referring to the goal process. Following this rationale, it is not hard to explain how expressions describing acts of communication behave as directives when (1) their subject is presented as the antagonist of an FD pattern – i.e., as exerting (or keeping from exerting) some kind of force against another entity – and (2) its complement takes the subjunctive, as corresponds to a proper FD patterns goal process. This way, the subjunctive is not presented here as the mere consequence of the matrix verb: it is the combination of these two traits that associates with the CNICCC.^[18] Thus, Talmy’s (2000) FD system is shown to also account for the use of the subjunctive in the context of directives, even if no specific matrices – such as the ones referred to in Mejías-Bikandi’s (2014: 656) paper – are used for that purpose.

This same rationale can also easily account for the fact that, in Spanish, expressions such as those of propositional attitude (such as pensar ‘think’, ocurrírsele a alguien ‘occur to someone’, etc.) or even those primarily describing the cognitive process of attention (mirar ‘look’, vigilar ‘watch for’, velar por ‘see to it’, poner cuidado en ‘take care to’, etc.) also take an FD sense when followed by a subjunctive complement (RAE-ASALE 2009: § 25.4g–k, n), as in (32).

As it was the case with the notions analyzed in Sections 3.1 and 3.2, this kind of relationship between FD patterns and the subjunctive does not only apply to verb matrices, but also to nominal heads:

3.4 The wish modality

Let us now consider the next example:

Martínez Mancía includes matrices such as esperar ‘hope’ or desear ‘wish’ in the category that he labels wish modality, and claims that they exhibit a force dynamic configuration: they represent an entity (the speaker, any other sentient entity functioning as the main subject, or else “a generalized social authority”) both as the conceptualizer of the subordinate clause and as an antagonist exerting a psychological force – “based on the antagonist’s own personal, subjective ideals of what is beneficial and what could bring happiness, i.e., personal convenience” – on a sentient or non-sentient agonist (2021: 129–130). I hold that this same analysis can be also applied to other verbs, such as confiar en ‘expect’ and temer ‘fear’, as in (35).^[19]

The force to which expressions of wish and fear refer is of a volitional kind, roughly equivalent to that of querer ‘want’. In contrast to querer, though, it is not aimed at changing the course of events. This is because wishes and fears are (at least) partly epistemic (Palmer 2001: 13). As Fillmore puts it, verbs such as hope and wish involve an epistemic stance plus “a positive interest in the state of affairs expressed as the complement clause” (1990: 154).

In epistemic FD patterns, as Langacker explains, the force exerted is of an internal kind, residing in the mental effort expended by the conceptualizer in simulating the growth of his/her conception of reality “along a path by which it comes to encompass the ground process” (2008: 306). Due to their internal nature, these patterns – in contrast to those including real-world force (Sweetser 1982: 495) – impose no time restriction over the goal process, which, as a result, can be understood as posterior, as in (34) and (35), as simultaneous, as in (36), or even as previous to the antagonist’s location in time, as in (37).

However, being of an epistemic nature, they require the factual status of the goal process to be unknown (Palmer 2001: 13) by the antagonist when exerting its force, as shown in (38).

The contrast between these matrices and that of querer ‘want’ may be hard to see when the goal process is conceived as posterior to that referring to the antagonist’s force, but becomes evident when it is presented as simultaneous or previous. In these latter cases, the antagonist can only be seen as exerting its force, not on the agonist to act, but instead for (esperar ‘hope’, desear ‘wish’, confiar en ‘expect’) or against (temer ‘fear’) accepting the goal process as real (Langacker 2008: 306) – or, as Carrasco Gutiérrez puts is, as (not) wanting “la factualidad del evento subordinado” [the factuality of the subordinate event] (1999: 3084; translation mine).

From the previous analysis, it follows that wish and fear matrices also lie within the scope of the model here: they do not only associate with FD patterns but also imply foregrounding the antagonist’s force as the main clause verb. Such a conceptual arrangement, once more, is shown to trigger the use of the subjunctive (or the infinitive) in the complement, and prevent that of the indicative, as predicted.^[20]

Once again, the corresponding nouns and adjectives follow the same rationale:

This way, the present model is shown to account for the use of the subjunctive not only in the complements of directives and matrices of volition and causation but also of those expressing wish and fear, neglected in Mejías-Bikandi’s (2014) paper.