Te lo tengo dicho muchas veces. Resultatives between coercion, relevance and reanalysis

Abstract This paper deals with the question of how and why resultative constructions change into anteriors. This discussion will be based on synchronic data concerning tener + past participle, a resultative construction used in modern Spanish. One of the latter's most frequent is te lo tengo dicho 'I have (already) told you'. This is remarkable since decir 'to tell' is a non-transitional verb; te lo tengo dicho thus violates the requirement that resultatives should only combine with transitional verbs. In the literature, such mismatches between the semantics of a given construction and the meaning of its lexical filler have been claimed to normally trigger coercion, i.e. an inferential repair mechanism giving rise to special meaning effects. Thus, coercion - despite being conceived as a purely synchronic mechanism - is a prime candidate for an explanation of the change from resultative to anterior. In line with this hypothesis, occurrences of te lo tengo dicho are attested in my corpus where the latter is specified by quantifying adverbials such as muchas veces 'many times'. However, speaker judgements indiacte that even te lo tengo dicho muchas veces is not an iterative anterior construction, but still a resultative. Based on synchronic data taken from the CREA-corpus, it will be shown that in the vast majority of its occurrences, te lo tengo dicho is part of an dialogal discourse pattern where certain argumentative effects based on its resultative meaning are highly relevant. Crucially, therefore, in such "strong" uses a coercive shift towards an anterior meaning is excluded. On a more abstract level, it will be shown that coercion is controlled by pragmatic factors; in the case of te lo tengo dicho muchas veces, conceptual/semantic plausibility is systematically overridden by pragmatic relevance.

Moreover, Harris' (1982) model opens a diachronic window on the functional polysemy of the perfect in individual languages. Thus, in Modern Spanish, the perfecto compuesto is a perfect at stage 3 (but see Schwenter & Torres Cacoullos 2008;Schwenter 1994), as is the English present perfect. At the same time, however, many of the uses of both constructions reflect earlier stages of their evolution (see (1)).
(1) a. This article focuses on the question of how and why change from stage 1 to stage 2 comes about; this problem is part of the larger question concerning what forces drive the cycle as a whole. A partial answer to this question has been proposed by Rosemeyer (2012: 143) who observes that change from resultative to anterior is "due to the conceptual metonymy between the result of an event and the event causing the result". Even though this observation is basically correct, it is nonetheless insufficient, since the change does not occur by itself. If we take as a criterion for stage 2 the compatibility with iterative adverbials of the type 'many times', examples of anterior uses of the Spanish perfecto compuesto expressing iteration only crop up in the 13 th century, i.e. at a particular moment in time. In other words, it seems that something particular had to happen in order to trigger the shift from resultative to anterior. Table 1 gives an overview by frequency of the first combinations of the construction haber + past participle with the iterative adverbial muchas vezes 'many times' found in the CORDE-Corpus for the period from 1200 to 1300 (the overall number of occurrences of the perfecto compuesto during this interval is roughly 2,0001). As seen in Table 1, the largest share of instances of the perfecto compuesto quantified by muchas vezes 'many times' involves speech-act verbs ('say', 'tell', 'speak', 'order', 'advise', 'debate'), followed by perception verbs ('see ' and 'hear'). Generally, in the textual genres represented in the 13 th century section of the CORDE-Corpus, verbs from these two classes are extremely frequent. Especially in the historiographical prose of the alphonsine period, they usually occur in passages where a narrator addresses his reader with phrases like 'as I have told you now', 'as we have seen now', 'as you have heard now'. Thus, from the 2.000 instances of the perfecto compuesto found for the 13 th century almost 6002 occur in the General Istoria, where they are commonly used in expressions like commo avemos dicho, commo avemos contado, commo avemos departido. However, this does not explain why the change from resultative (stage 1) to anterior (stage 2) affected those two verb classes earlier than others. In order to better understand the change from stage 1 to stage 2, this article focuses on synchronic data concerning tener + past participle, a resultative construction used in present-day Spanish. As I will show, one of the most frequent instances of this construction found in the CREA-corpus is te lo tengo dicho 'I have (already) told you' (section 2). This is remarkable since decir 'to tell' is a non-transitional verb and thus violates the requirement that resultatives should only combine with transitional verbs. In the literature, such mismatches between the semantics of a construction and the meaning of its lexical filler have been claimed to normally trigger coercion, i.e. an inferential repair mechanism giving rise to special meaning effects (section 3). Thus, coercion -despite being conceived as a purely synchronic mechanism -is a prime candidate for an explanation of the change from resultative to anterior. In line with this hypothesis, occurrences of te lo tengo dicho 'I have told you' are attested in the CREA-corpus where the latter is specified by quantifying adverbials such as muchas veces 'many times' (section 5). However, speaker judgements clearly indicate that te lo tengo dicho muchas veces is not an iterative anterior, but rather still at stage 1; it denotes the currently still valid incremental result of a past speech act, thereby emphasizing the latter's strong impact on the current situation, and especially on the addressee. In section 4 it will be shown that in the vast majority of its occurrences, te lo tengo dicho is part of a discourse pattern where this effect -which is an immediate consequence of the construction's resultative meaning -is highly relevant. Crucially, therefore, a coercive shift towards an anterior meaning is excluded in such "strong" uses. On a more abstract level, this means that coercion is controlled by pragmatic factors; in the case of te lo tengo dicho muchas veces, conceptual/semantic plausibility is systematically overridden by pragmatic relevance. In section 6, it will be shown that this outcome is predicted by the inferential model of reanalysis proposed by Detges & Waltereit (2002) and Detges (in print). In the remainder of this paper, I will compare tener dicho with modern uses of haber dicho (section 7) in order to identify context types where a change from stage 1 to stage 2 is more plausible (section 8). As we shall see, this analysis, based on purely synchronic data, will shed new light on the diachronic data in table 1.

Resultatives and anteriors
This article explores the boundary which separates resultative constructions from anteriors. Neatly distinguishing the two types of constructions is therefore crucial to my argument. Resultatives are commonly defined as constructions denoting a "state that was brought about by some action in the past" (Nedjalkov & Jaxontov 1988: 6). More precisely, resultatives denote both the past event (symbolized as e in Fig. 1) and the consequent result r, but it is exclusively the result state r which is asserted and hence specified for tense. Thus, example (1a) for the PC at stage 1, lo he acabado ahora 'I have finished it now', instantiates a present tense use of the construction. That is, in this example, it is asserted that r holds at the moment of speech s (see Fig. 1).

Figure 1. Resultative
Anteriors, by contrast, refer to past events which extend to a present situation or that are relevant for it. Anteriors at stage 2 denote single past events e which extend to the moment of speech s (persistence, Fig.  2a) or events which are iterated towards s and beyond (iteration, Fig. 2b). Resultatives by definition denote secondary states (Nedjalkov & al. 1988: 4-6), i.e. states brought about by a transition ('being wounded', 'being thrown to the ground'). This is directly reflected in their grammatical structure: while primary states (i.e. states not brought about by transitions) are mainly expressed by adjectives, secondary states are usually referred to by past participles, i.e. stative forms derived from transitional verbs, e.g. 'to wound', 'to throw on the ground' etc. -a restriction that does not hold for anteriors. From this, it should follow that only verbs denoting a transition can be construed as resultatives. However, in the following sections we shall see that this is not always the case. As I will argue, such mismatches between the resultant-state meaning of the construction and the non-transitional semantics of the verb set the scene for change from stage 1 to stage 2.
In an exclusively synchronic study on the present perfect in English, de Swart (2000: 11) cogently argues that a stage 2 reading (persistence or iteration) is the natural interpretation that arises if a nontransitional event is presented as a resultant state. The mechanism by means of which this is achieved is type shifting (Michaelis 2004) or coercion (Pustejowski & Jezek 2008;Pustejovsky 1993Pustejovsky , 2001de Swart 1998). Coercion is a synchronic mechanism of reinterpretation (Pustejowski & Jezek 2008;de Swart 2000: 7;Pustejovsky 1993), usually triggered by a mismatch between a given construction and some lexical element that violates the latter's input requirements (see also Michaelis 2004). Coercion is an inferential mechanism that repairs such mismatches and gives rise to special meaning effects. In relevance-theoretic terms, it is an inferential repair of an explicature which is somehow defective (Escandell-Vidal & Leonetti 2002: 178; see also Carston 2004a, b). From a diachronic perspective, this view raises two questions. Firstly, as a mechanism of reinterpretation, coercion is brought about by hearers. But hearers can only reinterpret items that have been presented to them previously by another speaker. Thus, from a diachronic perspective, it has to be explained for what reasons speakers should choose to insert a non-transitional verb into a resultative construction, thereby creating an aspectually defective expression. Secondly, our data in Table 1 support the claim that stage 2 readings are the result of a diachronic change that took place at a particular moment in time. However, coercion is defined as a synchronic mechanism; therefore the stage 2 readings represented by Figs. 2a and 2b cannot simply be the outcome of a coercion effect; while coercion certainly plays a role in the change, the shift from resultative to anterior must involve some further factors and conditions. The topic of this paper is not the perfecto compuesto (i.e. haber + past participle) but tener + past participle, a resultative construction used in modern Spanish. As we shall see, this choice enables us to look more closely at the role of resultative constructions in colloquial discourse. Moreover, access to genuine speaker intuitions will make it easier to precisely define the thin line that separates resultatives from anteriors; tener + past participle provides a test case of what is (or is not) a possible resultative construction. Moreover, this will allow us to identify the pragmatic motivations which can lead speakers to use resultative constructions with non-transitional verbs. Finally, based on an inferential model of reanalysis (Detges & Waltereit 2002;Detges, in print; see also Waltereit, this volume) I will argue that the probability of a coercion process turning resultative tener + past participle into an anterior is quite low.

Tener + past participle
In Modern Spanish, tener is used, among other things, as a full verb meaning 'to have'. In contrast, the form haber is restricted to grammatical functions, among them the role as an auxiliary in the perfecto compuesto (i.e. haber + past participle). This situation is the outcome of a linguistic change which affected the entire Iberian Peninsula. From the medieval period onwards, tener, originally 'to keep', 'to hold', gradually superseded haber, the traditional 'have' verb. In Portuguese, this process led to the substitution of haver by ter (the cognate of Spanish tener) also as the auxiliary of the perfeito composto. In the context of this change, Spanish tener + past participle originally emerged as a variant of haber + past participle, most probably when the latter was still mainly a resultative construction (for more details, see Harre 1990: 94-128). In Modern Spanish however, native speakers categorize tener + past participle and haber + past participle as two distinct constructions. Like haber + past participle, resultative tener + past participle is primarily used in Peninsular Spanish; moreover, it has a pronounced informal flavor. In the CREA-Corpus, it is mainly found in oral passages of literary fiction.3 The form tener + PP represents two distinct resultative constructions. On the one hand, it can function as a kind of dative-passive whose subject is the undergoer of the event denoted by the lexical verb (see (2)); in Detges (2006), this construction type is labeled as "resultative A" (see also Rosemeyer 2014: 56-59;Detges 2000). On the other hand, tener + past participle can take as its subject the agent of the past event and represent it as the participant responsible for the resulting state. Detges (2000; argues that this construction type, referred to as "resultative B", is the starting point of Harris' (1982) anterior cycle.
(2) Resultative A: Subject ~ Undergoer, Dative-Passive Tengo prohibida la entrada en el país. I.have.PRES forbidden the access in the country 'I have the access to this country forbidden', 'entering the country is forbidden for me.' (3) Resultative B: Subject ~ Agent, Starting point of Harris' (1982) cycle Te tengo prohibido entrar hasta que te bañes. you I.have.PRES forbidden to.enter until that you.OBJ bathe.SUBJ.2s 'I have told you not to enter before taking a bath.' It has repeatedly been noted in the literature that resultative constructions preferably combine with particular verb classes (for Old English, see Carey 1995;Detges 2000; for Old Spanish and Old French, see Detges 2006; for a slightly different account, see Rodríguez Molina 2004: 190-200), among them speech act verbs ('to tell'), mental activity verbs ('to think'), perception verbs ('to hear', 'to see') and verbs denoting accomplishments or achievements (e.g. 'to bring about' or 'to achieve'). Detges (2006;2000) argues that these preferences are due to the rhetorical efficiency of resultative B constructions, as these are regularly used for particular argumentative moves (see also below, sections 3 and 4), especially in the first-person singular. These usages have in common that they exploit the relationship between the past event and the current result. Thus, when combined with a speech act verb as prohibido 'forbidden' in (3), the construction represents the 'result' of the foregoing speech act as being fully valid (and hence maximally powerful) at the moment of speech -in spite of having been uttered at some point in the past. By contrast, in constellations with perception verbs ('to see', 'to hear') and verbs of knowledge acquisition ('to learn', 'to find out'), the argumentative benefit of using a resultative B is slightly different. In such a case, the construction refers to a current state of knowledge which is presented as the result of past perception or knowledge acquisition, denoted by the lexical verb. In this constellation, the role of the verb is to invoke the source on which the current knowledge is based. Thus lo tengo visto 'I have seen it' really means 'I know it from having seen it with my own eyes' (Detges 2000: 351-354, Sweetser 1990. Put in more abstract terms, the argumentative benefit of a resultative B with a perception verb is to emphasize the well-foundedness of the subject's current state of knowledge. Other cases may present a mixture of argumentative effects. Thus tengo decidido 'I have decided' presents a current decision as the result of a foregoing process (and thus underpins its wellfoundedness), while at the same time emphasizing the subject's commitment to the decision in question. Such argumentative effects are especially welcome in first-person uses (see Fig. 4 in section 4 below), where the speaker generally insists upon her own knowledge or her own current commitment. Fig. 3 shows the ten most frequent lexical items combined with tener + past participle in the first-person singular. Of the four verb classes mentioned above, three are represented in my CREA data. Remarkably, of the items listed in Fig. 3, two -claro 'clear' and listo 'ready' -are adjectives rather than verbs.
Mental Activity Speech Act Accomplishment The above-mentioned category 'mental activity' is instantiated by the lexical items claro 'clear', entendido 'understood' and pensado 'thought'; when construed with tener + past participle, the items listo 'ready', preparado 'prepared', hecho 'done', escrito refer to accomplishments. The most frequent constellation containing a speech-act verb unsurprisingly concerns the verb decir (dicho 'said'). Interestingly, among the ten most frequent instantiations of tengo + Past Participle found in the CREA-Corpus, two are constructions built with non-verbal lexemes, namely tengo claro que 'I have clearly understood that ...' with the adjective claro 'clear', and lo tengo listo 'I have it ready', built on the adjective listo 'ready'.

Tener + Past Participle and coercion
In both tengo claro que 'I have clearly understood that …' and lo tengo listo 'I have it ready', the construction tener + past participle denotes a resultant state brought about by a foregoing action or activity performed by the subject. But unlike past participles, whereby this action or activity (e.g. entendido 'understood') is made explicit by the respective verb (i.e. entender 'to understand'), the adjectives claro 'clear' and listo 'ready' make no direct reference to any such activity.4 In the case of listo 'ready, finished' it could be argued that despite its nonverbal nature, this adjective has a resultative meaning inasmuch as it presupposes the completion of an (unspecified) action or activity.5 But this argument does not hold in the case of claro 'clear', which refers to a primary state. And yet, speakers of Spanish conventionally understand that tengo claro que means that the subject's current state of knowledge is the result of a previous activity or action. Note that, as a consequence of its high frequency, tengo claro is likely to be a lexicalized item, i.e. its correct interpretation demands no extra effort from the hearer. But originally, a relevant interpretation of the phrase lo tengo claro must have required the insertion of "extra" meaning (Michalis 2004) -in this particular case we may assume that the missing piece of information (i.e. 'result of a foregoing activity') was derived from regular uses of the construction, containing genuine verbal participles. Thus, the non-compositional meaning of tengo claro is an entrenched (and conventionalized) instance of type shifting or coercion. In the cases under discussion here, a mismatch existed between the construction meaning on the one hand ('result of a foregoing action or activity') and the nature of the inserted lexical item (an adjective denoting a primary state) on the other. In these cases, the construction meaning ('state brought about by a transition') has "won out" over the non-resultative meaning of the lexical item claro 'clear'. As has been shown by Escandell-Vidal & Leonetti (2002: 178) and Michaelis (2004: 7), this is a normal outcome of coercion.
An equally problematic constellation involving the resultative B construction and an inserted lexical item is represented by (te)(lo) tengo dicho 'I have told (you)(this)(already)'. Dicho is a past participle derived from the verb decir 'to tell', 'to say'. The latter is not a transitional verb, as the activity of saying or telling does not yield a naturally observable result. And yet, we find tengo dicho among the three most frequent instances of tener + past participle. The reason for this, of course, is the argumentative effect produced by this constellation of construction and lexical verb: tengo dicho emphasizes that the illocutionary effect (i.e. the 'result') of the past speech act still is or should still be valid at the moment of speech. As in the cases of tengo claro que and lo tengo listo, the piece of information originally missing ('result of past speech act') has been interpolated into this string by coercion.
Another inconsistent constellation (not shown in Fig. 3) is represented by tengo visto 'I have seen' or tengo oido 'I have heard'; the activities of seeing or hearing do not produce natural results. What both constructions really mean is 'I know from seeing it myself', 'I know from hearing'. In this case, the somewhat odd constellation of resultative construction and non-transitional perception verb is motivated by the argumentative benefit discussed in section 2. Again, reconciling this obvious mismatch between the construction and the inserted lexical verb is a case of coercion.
Thus, a general answer to the question of why speakers should ever start using resultative constructions with non-transitional verbs or adjectives (thereby deliberately accepting semantic mismatches) is that some combinations of this kind are useful from an argumentative point of view. In such cases, argumentative usefulness (i.e. argumentative relevance) overrides semantic coherence. In the following section, we shall see that this also applies to tengo dicho 'I have told'.

Discourse functions of tengo dicho
As shown in Fig. 4, which gives the frequencies of tener dicho for the entire CREA-corpus,6 first-personsingular present-tense uses are by far the most frequent ones. In this section, I will show that this heavy person asymmetry is a consequence of the latter's functioning in dialogal discourse. As an illustration of tengo dicho 'I have told' used in dialogal discourse, I give a passage from the short story Las noches del iris negro, written in 1991 by the Spanish novelist Enrique Vila-Matas (see (4)). This example contains several elements that are highly typical of how this item is commonly used.  In (4), tengo dicho accompanies a directive speech act ('will you leave us in peace?'). Importantly, this speech act is dominant over the illocution conveyed by the string tengo dicho itself. In this context, the latter's function is to remind the addressee that it's already been a certain time that he has not been complying with what the speaker is currently demanding from him. In such a situational setting, the resultative meaning of the construction, which emphasizes the still lasting "result" rather than the past speech act itself, creates an extra-strong illocutionary pressure on the addressee. An additional, but optional element ('I will not repeat this any more') serves to reinforce the reproach, possibly also by announcing sanctions.
Thus, the scenario given in (4) provides a specific answer to the question of why non-transitional decir 'to tell', 'to say' is used as part of a resultative construction: in 40 out of its 60 first-person occurrences in the CREA-Corpus, i.e. in 66% of those cases, tengo dicho expresses a strong reproach in support of another, dominant directive speech act, as exemplified in (4). In 29 of the 60 occurrences of tengo dicho, i.e. in almost 50%, the dominant directive takes on the more specific form of an interdiction (see below, (5), (6), (8), (9)), i.e. a speech-act type which requires extra-strong illocutionary force.
Despite its ancillary function on the level of discourse, tengo dicho is often found as a matrix clause into which the phrase expressing the directive speech is syntactically embedded as a subordinate clause, as in (6), (7), (8) As can be seen from the examples given in (4)-(9), the situational setting in which tengo dicho usually occurs is one of an extra strong illocutionary force exerted on the addressee. In 31 cases (out of 40), the dominant directive supported by tengo dicho is grammatically expressed in a direct form: its verbal predicate is either in the imperative or in the subjunctive (see (5), (6), (8)), a deontic construction (see (7)) or an impersonal passive (see (7), (8)). In many cases, e.g. in (4), (6) and (9), the situation is characterised by overt impoliteness. In 8 of the 40 cases, tengo dicho itself is introduced by the emphasizer mira que 'look!', as in (4) and (7). All of this shows that the construction tengo dicho is normally used in situations where an extra strong additional pressure on the addressee is required. Its argumentative role is to further dispose the addressee to finally comply with the dominant directive.
As a resultative -which focuses on the current result rather than on the past act itself -the construction does not specify the number of times that the speech act in question was carried out. Thus, it could be that it was uttered a single time (see Fig. 1, repeated here for convenience as Fig. 1a), but it could also have been repeated an unspecified number of times (see Fig. 1b). In the latter case, the resulting state expressed by the construction is the incremental result of repeated speech acts of the same kind, i.e. a conceptually very fragile result. In the following sections, I will argue that the mismatch between the resultant-state meaning of the construction, on the one hand, and the meaning of non-transitional decir 'to say' on the other is accommodated via "ordinary" inferencing, guided by the search for relevance (rather than mechanically triggered by the semantics of the construction and the verb; for discussion, see Asher 2011: 93). In view of this discussion, it is important to distinguish between two levels of interpretation. On the one hand, the combination of resultative construction and lexical verb yields the explicature 'incremental result of a past speech act'. On the other hand, in frequent situations like those represented by (4)-(9), the relevant overall effect is that of a 'strong reproachful reminder in support of a current directive'. As I will argue in the following section, a coerced reinterpretation of the construction as the expression of an anterior (rather than a resultative) meaning is blocked by the relevant effect just described.

At the limits of the resultative construction
In the preceding section, it was pointed out that the act of reproach expressed by tengo dicho can be reinforced by additional linguistic material. In (4), this is done by an independent proposition ('I will not repeat this any more'). A simpler form of reinforcement is iteration. Iteration, in turn, can take two different forms (see Harre 1991): a) reinforcement by predicate-coordination as in (10) and (11), or b) reinforcement by quantification as in (12)-(14). In reinforcement by predicate coordination, the speaker construes the cumulative results of at least two (or more) separate consecutive chains of speech acts. However simple examples like (12)-(14) may be from a grammatical point of view -their semantics is considerably complicated by the insertion of the iterative adverbial. In (12)-(14), tengo dicho still expresses the incremental result of an iterated activity (see Fig. 1c).

Figure 1c. Lo tengo dicho muchas veces 'I have told this many times'
Mapping a quantified, non-transitional past activity onto a current result should lead to coercion by putting the focus on the quantified activity rather than on the abstract result (see above section 1). This interpretation should even be further facilitated in cases where iterative quantifiers like muchas veces 'many times' are added to the construction. Nevertheless, in present-day Spanish, even in cases like (12)-(14), tener + past participle is still a resultative -more precisely, it is a reinforced variant of resultative tengo dicho (see Figs. 1a,1b,1c). Therefore, in cases like (12)-(14), its relevance is exactly the same as in (4). Thus, it seems that the coerced reinterpretation of the resultative as an anterior is not automatically triggered by the constellation of resultative construction and non-transitional verb; nor is it triggered by the iterative adverbial. In the next section, I will argue that the resultative interpretation -despite its conceptual inconsistency -is upheld by the fact that in contexts like (4)-(9), it is the more relevant one.
Note that in (12)- (14), we are maximally close to an interpretation of the construction as an anterior. That it still is a resultative, however, can be understood from the fact that speakers who judge (12) The reason for this is that it is not possible to conceive of (15) as an incremental result. Harre (1999) who has conducted a survey among native speakers formulates the latters' intuitions as follows: "In order for someone to lend their car a second time they must presumably have got it back in their possession after lending it the first time. The fact that they can lend it again automatically cancels the resultant state caused by lending it the first time" (Harre 1991: 59). By contrast, the reason for the grammaticality of (12)- (14) is, of course, the non-transitional aktionsart of decir 'to say'. An incremental-result reading is equally possible in the case of tengo visto muchas veces 'I know from having seen many times'. This construction is normally used in situations where it supports the relevant usage effect of 'safe current knowledge' (16) Lo tengo visto muchas veces. It.OBJ have.PRES.1s seen many times 'I have seen it many times'; 'I know it well from having seen it many times.' Another indicator of the still resultative semantics of predicates like tengo dicho and tengo visto is the finding that they are sometimes modified by the degree word muy 'very'; thus lo tengo my visto. lit. 'I have much seen it' is found once the CREA-data, while lo tengo muy dicho, lit. 'I have said this very much' is documented twice. Like its English cognate 'very', the degree word muy usually accompanies adjectives (muy claro 'very clear') and adverbs (muy bien 'very well'), whereas it is categorically excluded from verbal environments (*me decepciona muy 'it very disappoints me'). The only exceptions to this rule are adjectivelike verb forms such as participles (muy decepcionado 'very disappointed'). The semantic regularity underlying these distributional features is that muy 'very' can modify properties (muy claro, muy bien, muy decepcionado) and property predications (ser muy claro 'to be very clear', resultar muy decepcionado 'to become very disappointed'), but not event predications (*me decepciona muy '*it very disappoints me'). Resultant states can (but need not) be property predications, and can therefore be modified by muy, e.g. lo tengo muy estudiado 'I have studied it very much'. By contrast, event predications with the same lexical items, e.g. lo he estudiado 'I have studied it' or lo estudié 'I studied it' categorically exclude muy. Therefore, combinations like lo tengo muy visto and lo tengo muy dicho are arguments in favour of the view that the respective constructions are resultatives rather than anteriors.
Based on the preceding section, it should be clear that cases like (12)-(14) mark the very border between resultative and anterior meaning. Resultative B constructions with non-transitional verbs are a possible bridging context for the change in question. Despite the semantic mismatches produced by them, these cases nevertheless exist because of their argumentative usefulness, especially in contexts like (4)-(14), as pointed out above. This insight allows us a fresh perspective on the first occurrences of haber + PP with iterative adverbials in the 13 th century (see Table 1 above, reproduced here for convenience in a slightly different form as Table 1a). Given all of the above, we cannot be sure, for most of the cases indicated in Table 1a, whether we are really still dealing with incremental-result readings, or whether the construction already has an anterior meaning. Note, however, that in the 6 occurrences of he dicho muchas vezes 'I have said many times' in Table 1a, the construction does not function as a reproachful reminder of a past directive speech act, but is used to introduce assertions ('I have informed you many times'). Finally, the isolated case of lo he rroto muchas veces 'I have broken it many times' -where an incremental result reading is excluded -proves that a change has taken place. As demonstrated in the previous sections, speakers have clear intuitions regarding the boundary separating resultatives from anteriors. This, in turn, means that, from a diachronic viewpoint, this border can be expected to be relatively stable. Thus, examples like (12)-(14) cannot be taken as evidence for an ongoing change from resultant state to anterior -despite the extremely close metonymic relationship that holds between both readings in such a case. In fact, this change may never happen. In the following section, an explanation will be given as to why this is so.

Why change does not happen: Reanalysis and the Principle of Relevant Usage Effect
According to Detges (in print) and Detges & Waltereit (2002), reanalyses are re-interpretations of a given construction's form and/or meaning. As operations of re-interpretation, reanalyses are usually brought about by hearers. All reanalyses are ultimately based on one (or both) of two pragmatically motivated cognitive principles. The first (and more important) of these, the "principle of relevant usage effect" is given in (17).
(17) Principle of Relevant Usage Effect (hearer) In the simplest interpretation possible, the relevant usage effect observed in the situation is a maximally simple (i.e. direct) instantiation of the explicature (i.e the conventional meaning) of the linguistic item involved.
Hearers who unknowingly carry out reanalyses turn this principle around by applying it as an abductive rule (17'): (17') Abduction (hearer) Given the Principle of Relevant Usage Effect, the explicature of the linguistic item involved in the given situation can be expected to be a maximally simple (i.e. maximally direct) representation of the relevant usage effect.
A simple case in point for rule (17'), based on the Principle of Relevant Usage Effect (17), is the rise of the verb-final interrogative particle -ti in dialectal French (La dame, vient-ti? 'The lady, come-ti', i.e. 'does the lady come?'). This item was brought about by a reanalysis of the masculine weak pronoun il (clustered together with the liaison-consonant -t-). Importantly, as a pronoun, (-t)-i(l) already indicates a yes-noquestion when used in the inversion construction (vient-il? 'comes-he', i.e. 'does he come?'). However, when preceded by a full subject-NP, it agrees with the latter in number and gender (Pierre, vient-il? 'Pierre, comes-he?', Marie, vient-elle? 'Marie, comes-she?'), which shows that it is still processed as an anaphor. This is what qualifies il and its feminine variant elle as pronouns. Nevertheless, even before the change, the relevant usage effect of inverted il is already to indicate the interrogative function of the sentence. But crucially, before the reanalysis, 'interrogative function' is not its meaning, but rather simply a frequently recurring relevant usage effect, given that il (as well as its feminine variant elle) are still anaphoric pronouns. The outcome of the reanalysis, however, is not only that pronominal il turns into the invariable particle -ti. More important is that indicating the interrogative function has now become the meaning of this new element. Thus, what was once a mere usage effect before the change has now turned into a new entrenched meaning. As a consequence, the computational effort necessary for deriving the relevant usage effect from the construction (or the elements thereof) has been lessened. In other words, the meaning of the interrogative particle -ti is a more direct instantiation of the relevant usage effect -interrogative functionthan are the inverted pronouns il and elle.
As has been argued above, 66% of the present-tense first-person singular uses of tener dicho are instantiations of the argumentative pattern exemplified by (4)-(9). These examples therefore represent a recurring situation type where the relevant usage effect of tengo dicho is to reinforce a currently uttered directive. From a purely semantic point of view, a coerced re-interpretation as an iterative anterior would be the most natural one, especially in those cases where tengo dicho is reinforced by an iterative adverbial. But such an interpretation would be a step away from the observed relevant usage effect 'strong reproachful reminder in support of a current directive'. Therefore, a change from resultative to anterior is extremely unlikely in such a situation. In contrast, a reanalysis via coercion will be more likely in situations where the speech act supported by tengo dicho requires less illocutionary force, thereby making the construction's current-result meaning appear less relevant. In my data, 66% percent of the uses of tengo dicho represent situations where the latter's current-result meaning is highly relevant (for the remaining 34%, see below, section 8). Thus, the Principle of Relevant Usage Effect not only explains how and why reanalyses take place, but also, it can be used to make predictions concerning the unlikeliness of a change.

Lo he dicho muchas veces -the illocutionary force of the perfecto compuesto
As mentioned at the beginning of this paper, the grammaticalized present perfect of Spanish, the perfecto compuesto, represents stage 3 of the evolutionary cycle from resultative to anterior perfects. Nonetheless, in many of its uses, it still exhibits features characteristic of a stage 2 perfect, namely, persistence and iteration. In principle, an anterior construction denoting the iterated repetition of a past speech-act with the perfecto compuesto -e.g. te he dicho muchas veces que no hagas eso 'I have told you many times not to do this' -can also be used to express a reproachful reminder in support of a current directive. As follows from a comparison of Figs. 4 (above) and 5 (below), he dicho, with 3,944 instances in the CREA-Corpus, is dramatically more frequent than te lo tengo dicho, with only 60 instances. This is mainly due to the fact that the perfecto compuesto is more grammaticalized than tener + past participle. By the same token, however, Fig. 5 shows that haber dicho is lacking the neat functional profile shown for tener dicho, given that, in this case, the first-person usages are clearly outnumbered by the third person. In principle, however, this does not preclude "strong" uses as those discussed for tener dicho in the preceding sections. As we will see, such uses do exist, but they are far less frequent for he dicho than for tengo dicho. An example of te lo he dicho muchas veces 'I have told you many times' supporting a directive speech act is shown in (18). Given the still close temporal contiguity between the chain of iterated past events and the moment of speech (see Fig. 2c), the illocutionary pressure exerted on the addressee is still part of the construction's conventional meaning and hence an explicature rather than an implicature. However, given that he dicho is less anchored in the moment of speech, it is more often used in cases where this pressure is weaker. In (18) te lo he dicho muchas veces is used as part of the argumentative pattern already familiar to us from section 4. But such cases are not typical of he dicho, and compared to other usages, they are relatively rare. Sometimes, the directive introduced by he dicho is extremely polite (see (19)).
(19) Yo les pediría un poco de prudencia, incluso a Landelino, I you.POL would.like.to.ask.1s a bit of caution, even Landelino 'I would like to ask you for a little bit of caution, even Landelino, que lo he dicho muchas veces. PART it have.1s said many times you will recall that I have said this many times now.' (Feo, Julio, 1993, Aquellos años, cf. CREA) In other cases, the dominant speech act is simply an assertion -i.e. a speech-act type demanding less illocutionary pressure on the addressee.
(20) Eso no es así, Juan y te lo he dicho muchas veces. This not is so, Juan, and you.OBJ it.OBJ PERF.1s told many times 'It's not like that, Juan, and I have told you many times.' (Cacho Cortés, Jesús, 1988 (Mahieu, Roma, 1980, La gallina ciega, cf. CREA) In the vast majority of cases he dicho simply functions as neutral reminder of a punctual or repeated past speech act and represents it as being currently still valid; put more simply, it is a mere marker of reported speech.
(26) La deuda, como he dicho, es demasiado grande. The debt as have.1s said is too big 'Our debt, as I have said, is too much. ' (N.N., La Nación, 17/12/1996, cf. CREA) The comparison of tengo dicho and he dicho shows that the latter is much more frequent, which is, of course, a symptom of its higher degree of grammaticalization. Therefore, he dicho also occurs more frequently in the argumentative pattern described in section 4, at least in absolute figures. However, Table 2 also shows that only a small percentage of the total number of uses of he dicho occurs in such contexts, while tengo dicho is almost specialized in this pattern. These differences between both constructions are an indication that the illocutionary force of tengo dicho -i.e. the construction more strongly anchored in the moment of speech -is greater than that of he dicho. It also follows from Table 2 that tengo dicho is much more regularly used in situations where there is a strong need for its illocutionary force. From among the 40 instances of tengo dicho expressing a reproachful reminder of a past directive, only 3 (i.e. 8%) are supported by an iterative adverbial meaning 'many times' (see Table 3). In the case of he dicho, the number of such adverbials is much higher, not only in absolute figures (38 examples), but also proportionally (25%, see Table 3). This finding confirms the impression that he dicho, i.e. the construction which is less anchored in the moment of speech, has a far weaker illocutionary force and is therefore in greater need of additional support.

Contexts of reanalysis
What are possible context for a reanalysis of tengo dicho as an anterior? From what has been shown so far, it has become plausible that the change is most likely to occur in situations other than those of a reproachful reminder of a directive speech act. In many cases in my data, tengo dicho is used to introduce "strong" assertions. Cases in point are (27), (28) and (29), where the assertions introduced by tengo dicho all have a somewhat challenging undertone. In those cases, too, tengo dicho clearly is used for its illocutionary force. However, in none of these examples, the speaker explicitly addresses her hearer, e.g. by means of a personal pronoun, te oder le 'you'). Therefore, we may conclude that the illocutionary pressure exerted on the addressee is considerably less strong in cases like (27), (28) and (29)  Elaborating on a hypothesis first proposed by Givón (1976;1979), Dahl (2001) identifies "overuse" as a key factor in the evolution of grammaticalization cycles (see also Detges 2001). In the case of tengo dicho, this could mean that the construction be used in situations where its strong illocutionary force is not really necessary, but where its effect is nonetheless welcome. As an unintentional by-product of repeated "overuse", the construction would gradually lose in pragmatic strength; more specifically, in those contexts its illocutionary strength will not be necessary for identifying the relevant usage effect. In light of what has been said in section 6, this means that in such a case, a major obstacle to a coercive reanalysis would be removed, and consequently, the original incremental-result interpretation would eventually be replaced by an iterative anterior one. This scenario ties in with the historical data for haber + past participle documented in Tables 1, 1a. In all of the six cases of haber dicho first found together with muchas vezes 'many times', this construction simply functions as a marker of reported speech (see (30)

Conclusion
In this article, I have argued that mismatches between the meaning of a given construction and that of its lexical fillers are motivated by usage. Speakers willingly accept such mismatches if these help to create useful argumentative effects. In the case of resultative tener + past participle, these effects are based on the relation between past event and current result. When combined with non-transitional decir 'to say, to tell', the resultative denotes the currently still valid incremental result of a past speech act, thereby emphasizing the latter's strong impact on the present situation, and especially on the addressee. In other cases, e.g. with perception verbs, the mismatch is motivated by equally interesting argumentative effects that have to do with the relationship between a current state of knowledge and a past activity of perception. In these cases, coercion, i.e. a re-interpretation aimed at repairing such aspectual mismatches, does not immediately take place; specifically, coercion will be blocked as long as the argumentative effects associated with the construction are immediately relevant in a sufficient number of situations. As I have shown, tengo dicho 'I have said' is frequently used in situations where it functions as a strong reproachful reminder of a past directive, typically an interdiction. As long as this usage pattern is the most frequent single situation type in which tengo dicho is used, no change will take place. Indirect evidence for this claim is provided by a comparison with he dicho, the grammaticalized anterior expressing, among other things, persistence and iteration. Among the uses of this construction, strong reproachful reminders of past directives, although possible, are but an extremely small fraction. We may therefore hypothesize that the reanalysis from resultative to anterior, based on a coercive reinterpretation, is brought about through contexts in which the strong illocutionary force of the resultative is less needed and therefore does not contribute to a relevant interpretation of the construction.