Talk into vs convince to r Talking as a cause leading to containment, convincing as a cause leading to a result *-* Patrick J. Duffley Universitě Laval This paper explores the causative constructions 'talk NP into + -ing vs 'convince NP to + infinitive by means of a collection of attested occurrences. It shows the connection between the characteristics described by Wierzbicka (1998), Gries & Stefanowitsch (2004) and Rudanko (2006) and the linguistically-signified semantic content involved in these structures. Wierzbicka's account and the related Construction Grammar approach are shown to be wanting on both the descriptive and explanatory levels due to a distancing from the level on which a stable relation exists between meaning and linguistic form. An explanation of the distribution and semantics of the two constructions is proposed based on Langackers (1987) semiological principle, i.e. on the semantic content associated with each of the linguistic signs involved in these sequences. Keywords: causatives; container metaphors; path metaphors; Construction Grammar; Natural Semantic Metalanguage; embodiment i. Introduction Causative constructions have generated an abundant body of literature over the past 40 years (cf. among others Aissen 1979; Bardzokas 2012; Baron 1974; Comrie & Poiinsky 1993; Gilquin 2010; Givón 1975, 1980; Gries & Stefanowitsch 2004; Hollmann 2005; Kim & Davies 2015; Rudanko 2006, 2015; Shibatani 1976; Song 1996; Talmy 1976; Verhagen & Kemmer 1997; Wierzbicka 1998). The first task to be undertaken in the present study is to refine the observation of the empirical data using modern tools not available to earlier researchers such as Wierzbicka, in this case the Internet and on-line electronic corpora. Besides the World Wide Web, two corpora will be used for this study: the 553-million-word Bank of English (henceforth BOB) and the 450 million word Corpus of Contemporary American English (COCA). *3SCS Publishing - ffiQt ^ll&aäöif ipSSMh^sfcj t piisteä on 2,''20/2'2i0 7:4Q ii£ via ISS^SCtKCFTK UlZTJZ&ZZIr. Äccoimc. a_43ia *fl-f^^fog John Benjamins Publishing Company Our second task will be to argue, contra a constructionist type of approach, that the observations made by the authors who have studied the causative 'talk NP into + -ing and 'convince NP to + infinitive' sequences can all be explained by the meanings of the items of which these sequences are composed. This leads into the third more theoretical section of this paper, which will involve a discussion of what these findings imply for those who have attempted to apply a constructionist approach to the sequences examined here and of the limitationrof this type of approach in general. 2. Refining the observation of the empirical data The first major discussion of the semantics of the English causative constructions investigated in the present study concerns the structures 'get + object + to + infinitive,' as exemplified by J got Joe to unlock the door-, versus Verb + object + into + gerund-participle,' as in I talked Joe into unlocking the door, and is provided by Anna Wierzbicka (1998). Since a number of her claims are problematic from the empirical point of view, it seems logical to start by examining them using modern electronic tools in order to clarify the nature of the data concerning these constructions before attempting to propose a hypothesis to explain them. The first problematic claim in Wierzbickas discussion is the following: In the case of the into construction, the causee originally didn't want to do what he or she dids whereas in the case of the get construction there is no such assumption. (Wierzbicka 1998:125) This claim is belied by the following attested use from COCA of the sequence 'verb + object » into + gerund-participle' in which the causees have presumably been longing to perform the event denoted by the gerund-participle for a considerable time: (1) The premise is simple: Four unmarried women with love life "issues" are followed around by two love coaches (relationship columnist Teresa Strasser and JD Roberto, reality-show host of "Outback Jack"), who seek to guide them into folding the man of their dreams by giving them advice. (COCA) Wierzbicka adds that "in the into construction, the causee's action is 'triggered' by the causer's will, not by the causees own will, whereas in the get construction, the causee is acting in accordance with both his or her own will and the causer's will" (1998: 125-126). This claim is also problematic with respect to (1) above, in which the causees' will is part of the reason why they are aiming to perform the event denoted by the gerund-participle (finding the man of their dreams). EBSCOIvvSm - printed OH. £>£0/222G 7:4G ill vis £I&£2iP.£H073. Q"K±7£;£L£21.&. All ose flTsfajtct: t& fc-ccpa://wv.ebjlto ,C Here the person referred to is explicitly described as having wanted to do art full-time and was gradually able to achieve this goal under the teadierVguidance and encouragement. A minor empirical problem concerns Wierzbicka's characterization of the 'get + object + ro-infinitive" construction as implying no assumption that the causee originally didn't want to do what he or she did. It should be pointed out to make things clearer that the get construction does not necessarily imply that the causees will is involved at all, as can be seen from (3) below, which refers to a young baby: (3) Basically, to get him back to bed I gave him 10 oz of formula and got him to burp a few times (it was pretty though, with him screaming). (answers.yahoo.com) Here the child evidently did not want to burp, and, In any case, burping is not a voluntary action for a young infant. A second contention made by Wierzbicka that is not borne out by usage is that "in the into construction, the causee is unaware of what is happening (namely, that his or her action is 'triggered' by the causers will), whereas in the get construction, there is no such assumption" (1998: 126). The incorrectness of this claim is demonstrated by examples (1) and (2) above in which the causee expressly wishes to perform the action denoted by the gerund-participle and is aware that his or her performance of this action is due to the causer. Wierzbicka makes another empirical claim that is problematic when she states that "one cannot 'encourage' or 'induce' someone into doing anything" (Wierzbicka 1998: 125). Examination of real usage data in the BOE shows however that the sequences 'encourage + into + gerund-participle' and 'induce t into t gerund-participle' are in fact both attested: (4) The new PRSA legislation is aimed at encouraging more people into making private provision for their retirement. (BOE) (5) I am fairly confident in my bowling now - I try to induce batsmen into making mistakes. (BOE) Another author who has dealt with the two patterns that are the object of our attention here is Juhani Rudanko, whose use of a corpus-based approach gives a sounder empirical basis to his observations. Thus he offers the very pertinent - printed on 3/&0,/2C££ "7r4Q £H via U$i~>7£RZITn, All ti3= j-ubjeCrfc fco hrSfcpa i/fmnr. daaco. com/tanaa—of-use consideration that "the into -ing pattern places an emphasis on the extent of the movement from the initial state to the final state and also carries the idea that the movement may have been by stages" (Rudanko 2006: 328). This observation can be illustrated by (6) below, which Rudanko cites in support of his analysis: (6) You were mirroring a phenomenon that was happening in many gay lives across America during that time - an explosion of self-invention. Did this somehow lead you into studying the Native American berdacne? (BGE, cited by Rudanko 2006: 322) Support for Rudanko's intuition can also be found in the fact that if the to + infinitive construction were used in the context of (6) above, the impression of a gradual waxing of interest in berdache would disappear and only the resultant impact on the interviewee's area of study would be evoked. Further support of his description is provided by the fact that the 'to + infinitive' construction would not be compatible with the notion of gradual painless transition denoted by the verb ease in a context such as (2) above. A second observation by Rudanko also hits the mark, this time regarding the infinitival construction: "The to-infinitive pattern focuses more on the resulting state, regardless of how it may be reached" (Rudanko 2006: 328). This remark is supported by the absence of the into + -ing construction with the verb cause (cf. Rudanko 2006: 319): since this verb presents the complement's event as a mere effect brought about by a cause, there is no concern with the means used to achieve the resultant effect. Further confirmation of Rudanko's observation is provided by the verb gel, which tends to focus very strongly on the mere idea of achieving a result, as illustrated by uses such as She got the job and We finally got to the top of the mountain. While this verb is minimally compatible with the into + -ing construction, usage is preponderantly in favour of the to-infinitive structure. A search in the Corpus of Contemporary American English for the strings 'get [verb] + them!her + to + infinitive' obtained 2468 attestations, while the corresponding sequence with 'into + -ing only found 3. Another corpus-based study of the 'into + -ing construction, by Cries and Stefanowitsch (2004), provides a further relevant observation concerning this sequence: The verb most strongly associated with the [into] construction is trick, followed by fool, coerce, and force. (Gries & Stefanowitsch 2004:227) This statement is based on a measure of "collostructional strength," a calculation of association strengths between items performed on the basis of "a cross-tabulation of the individual frequencies of the word and the construction in question as well as their joint frequency" (Cries & Stefanowitsch 2004: 227). This measure places »3 - printed, oci 7:4-3 AM via H2£S3.YK07jI U?*TVEHZIT&. All isae j-ohject &D b'Scpa; //«rw. ebjaco -cQn/fceiau-af-nM trick ahead of talk, which is the most frequent collocate of the into construction in terms of raw frequency (in COCA, talk occurs 887 times followed by into + -ing, as compared to 536 for trick). One point of interest with respect to the above observation concerning the verb force is that, although this verb is high on the collostructional-strength scale with respect to other verbs in its association with the into + -ing construction, it is much more strongly associated with the to + infinitive structure than with into + -ing: a search in COCA for thejtrings 'force + them/her + to + infinitive' obtained 2113 attestations, while the corresponding sequence with 'into + -ing only found 19. This distinguishes it sharply from its semantic neighbour coerce, of which 21 occurrences of the string 'coerce [verb] + them/her + into + -ing" were found in COCA versus only 10 of the sequence 'coerce [verb] + them/her + to + infinitive'. 3, Explanations anyone? Although the three authors discussed above make many valid empirical observations and generalizations about the two structures under study, no explanations are proposed as to the origin of the semantic intuitions and statistical observations concerning these two causative constructions. That is what 1 will attempt to do in the rest of this article. The proposal put forward here will argue that the sequences in question do not constitute "constructions" in the sense of Goldberg (1995) in that observed usage is fully predictable from the semantic content of the items combined in these sequences. The first question that we will attempt to answer is why the 'into + -ing con struction tends to suggest "manipulation" (Wierzbicka 1998) or "trickery" (Cries & Stefanowitsch 2004). in our view, this is due mainly to the meaning of the preposition into, which denotes movement leading to containment. This can be illustrated by its spatial use in a context such as (7): (7) He walked into the room. The notion of movement leading to containment signified by this preposition holds the potential for conveying the impression of entrapment, i.e. of maneuvering someone into a situation that they did not want to be in. In this respect, it is significant that the second most frequent pair of verbs found with this construction after trick and fool is force and coerce, both of which imply that the person denoted by the direct object does not perform the action expressed by the gerund-participle willingly. This impression is also observed with the verb induce in (5) above. On the grammatical level, the gerund-participle's grammatical meaning also contributes to the expression of this message: Duffley (2006: 19-20) argues that the schematic meaning of the gerund-participle puts the implicit subject of the -ingfoma in relation with the inferiority of the event. This is most evident in the progressive, where the subject of the auxiliary be is situated within the inter i or-ity of the event expressed by the gerund-participle at the moment in time evoked by the auxiliary, as in: (8) When I saw her, she was playing tennis with her sister. In its gerundive use, the gerund-participle evokes the event's interiority as a homogenous whole made tip of ail of the instants contained between its beginning and its end points, a construal which is aspectually neutral (see Duffley 2014: 35-36, De Smet 2010: 1169) and can denote both an event that is completely accomplished (as in I remember playing tennis with her sister) and one that is in progress, as in the use below involving the preposition in: (9) She was so engrossed in playing tennis with her sister that she didn't even notice my arrival. In the into + -ing construction, the object of the mam verb ends up being involved in the event denoted by the gerund-participle, as illustrated by the pair of sentences in (10)-(11), the second of which could describe the situation immediately ensuing upon that depicted by the first: (10) Nita is a huge fan of Jason Becker, and her boyfriend/manager tricked her into playing a Jason Becker song for her hero! (wgrd.com/alice-cooper-guitarist-nita-straiiss-was-tricked-into-playing) (11) She was engaged in playing the Jason Becker song when I walked in. Thus both the meaning of the preposition into and the way the gerund-participle construes the event that it denotes make a contribution to the effect of entrapment and the attendant impressions of manipulation or trickery observed by Wierz-bicka and Gries & Stefanowitsch. It should be noted, however, that this impression is not present in all occurrences of this construction: (12) She finally wore me down and talked me into buying a package of cigarettes. In this use, the main idea is that of pushing the person into an action that they did not want to perform. The next question to be addressed is the explanation of Rudanko (2006)s observation that the Unto + -ing" pattern places an emphasis on the extent of the movement from the initial to the final state and gives the impression that the movement may have been by stages. It will be argued here that this is also principally due to the meaning of the preposition into, which represents the movement evoked by EESCOhc-il: - priuT:«i an 2/2HJ2£:-Za =o _ccm/fccnu-Of-tLit to as leading to the complete containment of the trajector within the landmark. In his study of the distinction between in and into, Tutton (2009: 20) notes that "into seems to accentuate path to a greater extent than boundary-crossing in!' This is due to the to-component of its meaning. Due to its in-component, into adds fo the extension of the movement leading up to the container the additional stages of penetrating the container's external limit and getting all of the trajector inside. This is illustrated in Figure 1: Figure 1. Into as movement leading to containment It is revealing to compare the impression produced by the Into -»- - frig construction with that observed with 'to + infinitive'. Rudanko remarks quite judiciously that the to-infinitive pattern focuses more on the resulting state, regardless of how it may be reached, but does not propose any explanation as to why this should be so. It will be held here that this is due to the meaning of the components of this construction, the preposition to and the bare infinitive. The preposition represents the terminus of the movement that it denotes as a point rather than a container, as illustrated in Figure 2: Figure 2. To as movement leading to a point The bare infinitive, for its part, denotes the complete actualization of its event in time (cf. Duffley 2006: 28-30), or in Langacker's terms (1987: 250-252), the "full instantiation" of the event. This semantic configuration focuses on the mere achievement of the infinitive event's full actualization as a result of the matrix verb's event. There is no notion of penetration of an external limit or entry into the interior of a container, and so no impression of stages in the movement leading to the complement event's realization. This account also provides an explanation of why the verb ease is not found with the 'to + infinitive* construction. Since this verb denotes a gradual, painless transition into a new state, it is not semantically compatible with the notion of the mere achievement of resultant realization expressed by the 'to + infinitive' pattern. Encouragement, on the other hand, can be construed as gradually coaxing someone into a new way of doing things, and this explains why the 'encourage + into + gerund-participle' pattern is attested in uses like (4) above. In addition, an explanation can also be offered for the fact that a verb like talk is not compatible with the 'to + infinitive' construction. The reason is quite simple: talking is not goal- or result-oriented. On the other hand, the verb convince, which is very clearly result-oriented and focipes on obtaining the adhesion of the convincee to the opinion of the convincer, is almost exclusively construed with to + infinitive: only 5 mfo-constructions are attested in the 553-million-word Bank of English corpus vs 6851 attestations with the infinitive, and only one occurrence of this pattern is found in the 450-million-word Corpus of Contemporary American English, and that in a parallel construction with the verb force: (13) A few of the young women reported that they had been more forcibly coerced into sexual intercourse, thai is, that their boyfriends had used pressure to convince or force them into having sexual contact against their will. (COCA) The řo-infinitive pattern, on the other hand, is attested 3718 times with the verb convince in COCA. Persuade, in comparison, is more amenable to the Unto + -ing construction: there are 22 occurrences of this pattern in the Bank of English (cf. Hunston & Francis 2000: 103) and 12 in COCA (versus 4288 of the to-infinitive structure). This is because convince is result-oriented while persuasion involves both process and result. Convince derives from the Latin verb meaning 'conquer, overcome'; to convince someone of something is to cause that person to believe the truth of something by means of facts, arguments, etc.; persuade comes from the Latin 'advise, urge until the attainment of a result' and can be defined as 'to bring (someone) to do or believe something through reasoning, argument or appeal to emotions'. Further evidence of the result-focussedness of convince versus the inclusion of the idea of process with persuade is found in two facts of usage. The first is that if we are absolutely sure about something we say I'm convinced rather than I'm persuaded: (14) - Are you sure he's guilty? - Yes, Tm convinced. Here the speaker wants to express solely the resultative state of mind that he is in with respect to the person's guiltiness. The second fact is that one can say to use persuasion on someone, but not *to use conviction on someone: COCA has 22 occurrences of the former sequence and none at all of the latter. Here persuasion ESSCŮhcs-S - jprinwd on 2j';a/2C-iCI 7:40 AM via OKITERZITa. Ml life srJijairc to cpa r / J vww. sba=o . c o &/ ■& cejils -o£- tU-± is construed as the means employed in order to get someone to do something. The subtle semantic difference between these two verbs is thus reflected in their relative frequency with the 'into + gerund-participle' pattern. t" An explanation can also be proposed for the constructional difference noted above between force and coerce based on the lexical meanings of these two verbs. The fact that coerce is more frequent with the into -ing structure than with the to-infinitive is a reflection of the original ideal of hemming someone irj.(ct>- + arcere 'to shut up, enclose') expressed by this verb, which according to the OED involves "the application of force to control the action of a voluntary agent" and according to Webster's places emphasis on "domineering and overriding resistance": to coerce is to forcibly constrain someone so that they are funneled into a situation against their will. The verb force, for its part, evokes the exercise of the power required to move an object and so is goal- or result-oriented, which explains its marked preference for the fo-infinitive. This aspect of its meaning is also reflected in its occurrence in structures such as to force a door open in which the adjectival object complement denotes the resultant state into which the door is put by means of the exercise of force. Table 1 provides a summary of the comparative frequencies of the sequences examined thus far: Table 1. Summary of comparative frequencies Sequence Frequency get + them/her + to-infinitive (COCA) 2468. get + them/her + into + -ing (COCA) - 3 talk + NP + into + -ing (COCA) 887 trick + NP + into + ing (COCA) 536 force + them/her + to-infinitive (COCA) 2113 force + them/her 4- into + -ing (COCA) 19 coerce + them/her + to-infinitive (COCA) 10 coerce + them/her + into + -ing (COCA) 21 convince + NP + to-infinitive (COCA) 3718 convince + NP + into + -ing (COCA) 1 convince + NP + to-infinitive (BOE) 6851 convince + NP + into + -ing (BOE) 5 persuade + NP + into + -ing (BOE) 22 persuade + NP + into + -ing (COCA) 12 persuade + NP + to-infinitive (COCA) 4288 - priEM-d on. 2,'*0,'2C5:Q 7:40 SM via SRSSSltBaUS, HNIVEHZITA- ill -ese ai^jje-rt; fco brcpa;//vw. eba=o _ccni/ce=iaa-ö=-u-* 4- Construction grammar, embodied cognition and the basic design architecture of human language George Lakoff and his co-researchers (Lakoff 1987; Lakoff & Johnson 1980, 1999; Lakoff & Turner 1989) have developed several lines of evidence in support of the hypothesis that language involves embodied cognition, i.e. that people use their understanding of familiar physical objects, actions and situations such as containers, spaces and trajectories to understand other more complex domains. The data examined in this study provides further evidence in favour of this hypothesis, as it shows that causing someone to do something by means of convincing is construed in terms of a trajectory leading to a point, while causing someone to do something by means of talking is conceived in terms of a trajectory leading to containment. There is, however, another dimension of embodiment relevant to our discussion that is often overlooked, and that has to do with the basic design architecture of human language itself: Ronald Langacker (1987: 11) notes the fundamental fact that "language is symbolic in nature," by which he means that the foundational relation on which human language is based is the association between a mind-engendered meaning and a bodily-produced sign. It is our contention that many of the problems that we have pointed out in Wierzbicka's, Rudanko's and Gries & Stefanowitsch's analyses stem from a neglect of this basic principle. A large part of the cause of this oversight stems from treating meaning on the level of the overall construction rather than of the linguistic signs of which the construction is composed. This approach is characteristic of Construction Grammar, which works on the assumption that: (1) "constructions are taken to be the basic units of language" (Goldberg 1995: 4) and (2) "constructions are independent of the lexical items which instantiate them" (Goldberg 1995: 1, 4). Regarding the first postulate, it should be remarked that it only makes sense if one follows Goldberg in expanding the notion of'construction to cover morphemes, based on the fact that the latter are "pairings of meaning and form that are not predictable from anything else" (Goldberg 1995: 4). This, however, amounts to an abuse of the term 'construction/ as on the syntactic level this term corresponds to a combination of items that all have meanings, whereas on the morphemic level it denotes the combination of a meaningful meaning with an as yet meaningless form. As for the second assumption, it needs to be handled with care as it has the potential for introducing a certain disconnection between linguistic meaning and linguistic form. This occurs, for example, in Wierzbicka's analysis, where the meanings attributed to the 'get + object + to + infinitive' and 'verb + object + into + gerund participle' constructions are much too specific and consequently unable to cover all of the uses of the two patterns in question. There are two main reasons for this. The first is that in the majority £E5C0£|q=13 - £.£i£ifc=d. on £,J£0/2C2a 7;4Q iH vis M££1ry«e0v?. UEZVERZITA- ill isae aiihject to h -ft fcp j i /vmr .d>=--o . cam,/ ttnnj £-a_= c of cases constructions are discourse entities, assembled on-line for specific communicative purposes.1 Since the most salient aspect of discourse entities is the specific message that they serve to communicate and not the linguistic means used to convey this message, there is a high risk of confusing linguistic meaning with the specific message intended when one analyses meaning on the level of the construction. The intended message, however, is the end-product of an inference that the speaker wants the hearer to make, and is based not just on the linguistic meanings contained in the utterance but also on the utterance situation, on what has been said previously by the interlocutors, on shared knowledge of the world, etc. In Wierzbicka's case, the use of Natural Semantic Metalanguage as a tool for describing meaning constitutes a second confounding factor in the analysis, as it distorts the description of meaning by forcing the latter into the mold of a limited set of universal semantic primitives, a procedure which is somewhat analogous to translating the English constructions into a foreign language in order to describe their import. This compounds the confusion between linguistic meaning and intended message, as the equivalence between the Natural Semantic Metalanguage paraphrase and the original English construction lies on the level of the message and not on that of the linguistic means used to convey that message (cf. the equivalence in the message communicated between the Spanish utterance Tengo hambre and the English 1 am hungry, where there is no notion of 'tener' or 'having' in the English utterance, and no notion of'being' in the Spanish, in addition to the conceptualization of hambre as a noun in a direct object relation to the verb tener in the Spanish sentence vs. the construal of hungry as an adjective in subject complement function in the English one). Dissociating linguistic meaning from linguistic form represents a failure to respect the principle of embodiment as it applies to human language on the most basic level, namely the fact that language is symbolic in nature. Generative Grammar disregards this principle by attempting to deal with form in abstraction from meaning on the syntactic level. Chomsky (1957: 141) famously argued that since meaning is "notoriously difficult to pin down," if it were to be shown that it played a central role in linguistic analysis the latter's results and conclusions would "become subject to all the doubts and obscurities that plague the study of meaning" thereby striking a serious blow at the very foundations of linguistic theory. In his view it is preferable therefore to set meaning aside and focus on formalizing the distributional configurations of the physically observable linguistic signs that make up the i. This has been known for a long time: Hermann Paul observed in 1880 that "one simply has to admit that only very few sentences have been memorized as such. Most sentences are composed on the spur of the moment" (pp. 109-110). t - p=iDt«d on 7:40 ££d v±s. StkSASEftQVS. nZT ZVERZ1TA HI -j» t-D b~ i : / .■' inW. az o . torn,-' fceiaaa —o=-u- e sentence, i.e. to build a formal syntax of the language. Since, however, meaning cannot be kept out of the picture completely, as it is obvious that people speak in order to express their ideas, it must be re-introduced into the analysis at some point. In a generative model, this is done only after the syntactic component has generated a string of linguistic forms, which is then sent to the semantic component for inter pretation. This theoretical model raises a number of important problems. Firstly, as pointed out by Seuren (2004: 161), there is the crucial question^f what drives the syntactic sentence generator. To claim that the generation of a sentence is "a process, activated by a start signal, that randomly selects lexical items and casts them into a grammatical structure" is absurd according to Seuren. Equally absurd is "the notion that a randomly generated sentence structure should be taken to pass an instruction to the cognitive system of the same organism for the sentence to be interpreted''': this would entail that the speaker does not have any idea in mind to express before he starts generating a syntactic string of linguistic forms. Serious problems thus arise when form is treated autonomously from meaning. The opposite pitfall also exists, however, i.e. treating meaning in abstraction from form. This is a significant risk for approaches that situate meaning on the level of the utterance or the construction such as that of Evans (2009: 25), who holds that "words do not in fact have meaning (...). In my account, meaning is a function of an utterance, rather than a given lexical representation associated with a word or other symbolic (i.e. linguistic) unit." The problem with this type of account is that the relation between message and utterance/construction is not usually stable: some sentence- or construction-level assemblages clearly do have a stable semantic content (e.g. All that glitters is not gold; to cat the mustard); in most cases, however, that is not the case. Thus, as noted by Bardzokas (2012: 29), in an utterance of the sequence John is too tired the speaker is not understood to be simply predicating of John the property of excessive tiredness, but to be putting his degree of tiredness into relation with some action that John might be expected to perform in the situation to which the utterance refers. The message conveyed thus varies in function of what it is that John is too tired to do. Correlating meaning and form on the level of the utterance leads therefore to a proliferation of different meanings for the same form. This is exemplified in Evans' analysis of the adjective long as having "at least two conventionally established lexical concepts" (p. 299) -[EXTENDED IN HORIZONTAL SPACE J, as in a long stick, and [EXTENDED IN DURATION], as in a long kiss. This amounts, however, to treating as part of the adjective's lexical concept something which is contributed by the lexical concept of the noun following it. Something similar occurs in some applications of Construction Grammar. A case in point is Stefanowitch (2001)'s construction-grammar approach to English analytic causatives. This author posits three "causation event types" as being relevant to these constructions: the MANIPULATE type, where an animate causer st - printed on £/iO/2C£G £1-1 via iC-JZ^^KOVr. tTTiTSRZITi, £11 l»e 5fihje*?t: fcn hrrpa; //mrs.efcaco.ceat/ps-raij—p:E—mc o:f'^:f--"iiitentionalIy' acts on a causee in a way that influences the causee such that he or she performs some activity (e.g. This guy was taking women from the teller and making them give him money); the TRIGGER type, where an event occurs that influences the causee in such a way that the latter inevitably undergoes some process (e.g. Seeing people abuse the system makes me feel bad); and the PROMPT type, where an event occurs and the causee perceives this event and decides to react by performing some activity (e.g. What made her decide to go into an old peoples home?). Stefanowitch himself observes (2001: 311) that the MANIPULATE, TRIGGER and PROMPT configurations "are not tied to any particular constructions," but occur across structures with the four main causative verbs make, have, get and force. In the illustrative examples just given above, the very same verb make occurs in all three types. This entails that make is in fact indifferent to Stefanow-itchs causation event types and must be treated as having a more abstract type of meaning that has the potential to correspond to all three of them. With regard to the two constructions examined in the present study the same thing is true. Both talk into and convince to can correspond to the MANIPULATE type: (15) Hello can someone give me some tips? For a third day straight this guy (friend's friend) talked me into giving him some money. (16) Some bum named Slim followed me around and convinced me to give him some money for a local bus to another town. In addition, convince + to-infinitive can convey the notion of PROMPT: (17) It bugged him for a while, but her reaction convinced him to drop it until he could sleuth it out on his own. Such causation event types appear therefore to be tangential to the type of meaning expressed by the 'convince + object + to-infinitive' and 'talk + object + into + gerund-participle' constructions. This is not surprising, in that the event types in question were posited independently of any reference to linguistic sign-meaning units in the first place. Another Illustration of the metalinguistic character of some constructions can be found in Rudanko (2015)'s treatment of the transitive into -ing structure. Following on Goldberg (1995)'s general theory, he classifies this sequence as a "caused motion construction." This leads him, however, to exclude the following attestation from the scope of his study: (18) Just as he once battled for supreme fitness, he has poured his energy into learning to speak again. (BNC) EE3C0ilO--5 - printed, on 2/20/2T: M ill via JEL5iJ.'iKj',JJ; OSZuTEElZITA. All m* iiijttC to hTTp-: „eh;i=e^ ccjn/nft=ma-oi-u-tt The reasons given for this exclusion are that "the direct object designates a resource at the disposal of the referent of the higher subject, and the pattern may be viewed as one of subject control." Regarding the allusion to subject versus object control, since it has been demonstrated that such distinctions are substantially pragmatic in nature (Duffley 2014), one wonders whether they can be validly invoked as evidence of the existence of different grammatical constructions. More fundamentally, the question is raised by this Construction Grammar approach as to why He charmed the child into learning to speak again represents a construction, whereas (IS) above does not. It is argued here that both are simply possible exploitations of the semantics and syntax of the Verb + NP + into + -ing sequence, and that the difference between them is merely a product of the differences in meaning between the verbs charm versus poury the direct objects the child versus his energy, and the pragmatics of their interrelations. A similar problem of the Construction Grammar approach leading to a rei-fication of the meaning of constructions due to the fact that it postulates a stable form meaning correlation on the level of the overall syntactic configuration is found in Kim and Davies (2015: 81), where the transitive into -ing sequence is once again treated as a "caused motion construction" distinguished from other such constructions by the fact that "it entails that the situation denoted by the gerundive phrase actually happened." While this is usually the case with the sequence in question, nevertheless it depends crucially on the semantic content of the matrix verb. Thus just as Mary invited the woman into her house does not necessarily entail that the woman actually came into her house, the sentence below does not necessarily entail that the addressee actually explores what may keep them from love: (19) When you stop to listen, you hear a small but persistent voice that urges you towards healing. This voice invites you into exploring what may keep you from love, belonging, and satisfying relationships. According to Kim and Davies' criterion, this means that some instances of the sequence invite + NP + into + -ing should be treated as caused motion constructions and others not. Such a division seems artificial and non-linguistic. The general methodological point that I wish to make here is that one should first investigate whether the message conveyed by a construction can be explained by the stable form/meaning units of which it is composed before attributing a meaning to the whole construction or attempting to fit it into a set of sign-independent notional categories. The case of Wierzbickas analysis of the 'get + object + to + infinitive' and Verb + object + into + gerund- participle' constructions examined here is an example of 'jumping the gun,' where the analysis skips over the level where the form-meaning relation is stable to the level of the % - pr±nt«d on. 3/5'V2£2a 7:43 EM via MS5SR5KD\Tfc nBTVEEtZITA^ All tim s-ubjecc to hctpa ://vw. abjco. ccm/fcSTnn-p£-ajE utterance-message relation, where it is not. Stefanowitch's approach, on the other hand, starts with abstract non symbolic categories and works back towards the linguistic forms, but does not succeed in connecting the latter to these categories in any direct or stable manner. This paper has attempted to show the explanatory gain that can be achieved by starting one's analysis on the level on which meaning is stably embodied, which is normally that of the word or morpheme, where a linguistic sign is stored in a stable, permanent and direct relation with its meaning outside of any particular context. References Aissen, Judith. 1979. The Syntax of Causative Constructions. New York NY: Garland. Baron, Naomi S. 1974. The structure of English causatives. Lingua 33: 299-342. doi:iO.1Ol6/0O24-384l{74)90044-8 Bardzokas, Valandis. 2012. Causality and Connectives. From Grice to Relevance [Pragmatics & Beyond New Series 216]. Amsterdam: John Benjamins. doi:io.i075/pbns.2i6 Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. Comrie, Bernard & Polinsky, Maria. 1993. Causatives and Transitivity [Studies in Language Companion Series 23]. Amsterdam: John Benjamins. doi:io.i075/s!csJ23 De Smel, Hendrik 2010. English -ing clauses and their problems: The structure of grammatical categories. Linguistics 48:1153-1193. doi:io.i5i5/ling.20io.038 Duflfky, Patrick J, 2006. The English Gerund-participle. A Comparison with the Infinitive. Frankfurt: Peter Lang. DufHey, Patrick J. 2014. Reclaiming Control as a Semantic and Pragmatic Phenomenon [Pragmatics & Beyond New Series 251]. Amsterdam: John Benjamins. Evans, Vyvyan. 2009. How Words Mean. Lexical Concepts, Cognitive Models, and Meaning Construction. Oxford: OUP. doi:io,io93/acprof:oso/97Soig9234660.ooi.oooi Gilquin, Gaetanelle. 2010. Corpus, Cognition and Causative Constructions [Studies in Corpus Linguistics 39]. Amsterdam: John Benjamins. dohi0.iO75/scl-39 Givon, Talmy. 1975. Cause and control: On the semantics of interpersonal manipulation. In Syntax and Semantics, Vol. 4, John P. Kimball (ed.), 59-89. New York NY: Academic Press. Givon, Talmy. 1980. The Binding Hierarchy and the typology of complements. Studies in Languages. 333-377. do'i:io.t07S/sl-4.3.03giv Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago IL: University of Chicago Press. Gries, Stefan T. & Steianowitsch, AnatoL 2004. Covarying collexemes in the /« to-causative. In Language, Culture, and Mind, Michel Achard & Suzanne Kemmer (eds), 225-236. Stanford CA: CSLI. Hollmann, Willem B. 2005. The iconicity of complementation in present-day English causatives. In Outsidc-in- inside out [Iconicity in Language and Literature 4], Constantino Maeder, Olga Fischer & William J. Herlofsky (eds), 287-306. Amsterdam: John Benjamins, d oi: 1 o.i075/ill4.2ihol Hunston, Susan & Francis, Gill. 2000. Pattern Grammar: A Corpus-Driven Approach to the Lexical Grammar of English [Studies in Corpus Linguistics 4]. Amsterdam: John Benjamins. doi:io.i07s/sd.4 - prijit*d on 2,J20y2CEG -140 Hi via MiE£jlYKJV?. IISI^RZITS. list ai&jecc to h?rjgj:/twv.eha-o^ccmj'^ffTTn?i-or-i^e Kim, Jong Bok & Davies, Mark A. 2015. The iV/to-causative construction in Englisli: A construction-based perspective. English Language and Linguistics 20: 55-83. Lakoff, George. 1987. Women, Fire, andDangerous Tilings: What Categories Reveal About the Mind. Chicago IL: University of Chicago Press, doi:io.72Q8/chlcago/978022647ioi3.ooi,QOOi f Lakoff, George & Johnson, Mark. 1980. Metaphors We Live By. Chicago IL: University of Chicago Press. Lakoff, George & Johnson, Mark. 1999. Philosophy in the Flesh: The Embodied Mind and its Challenge to Western Thought. New York NY: Basic Books. ^ Lakoff, George & Turner, Mark. 1989. More Than Cool Reason: A Field Guide to Poetic Metaphor. Chicago IL: University of Chicago Press. doi:io.7208/chscago/9780226470Q86.o01.oooi Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, Vol. I: Theoretical Prerequisites. Stanford CA: Stanford University Press. Paul, Hermann, 188011920]. Prinzipien der Sprachgeschichte. Halle: Max Niemeyer. Rudanko, Juhani. 2006. Emergent alternation in complement selection. The spread of transitive into -ing constructions in British and American English. Journal of English Linguistics 34: 312-331. doi: 10.1177/0075424206293600 Rudanko, Juhani. 2015. Wheedled me into lending him my best hunter, comparing the emergence of the transitive into -ing pattern in British and American Englisli. In Perspectives on Complementation: Structure, Variation and Boundaries, Mikko Höglund, Paul Rickman, Juhani Rudanko & Jukka Havu (eds), 128-140. Houndmüls: Palgrave Macmillan. Seuren, Pieter A. M. 2004. Chomsky's Minimalism. Oxford: OUP. Shibatani, Masayoshi. 1976. The grammar of causative constructions. In Syntax and Semantics, Vol. 6: The Grammar of Causative Constructions, Masayoshi Shibatani (ed.), 261-312. New York; Academic Press. Song, Jae Jung. 1996. Causatives and Causation. A Universal-typological Perspective. London: Addison Wesley Longman. Talmy, Leonard. 1976. Semantic causative types. In Syntax and Semantics, Vol. 6: The Grammar of Causative. Constructions, Masayoshi Shibatani (ed.), 43-116. New York NY: Academic Press. Tullon, Mark. 2009. When in means into: Towards an understanding of boundary-crossing in. Journal of English Linguistics 37; 5-27. dol:io.ii77/oo75424208329308 Verhagen, Arie &Kemmer, Suzanne. 1997. Interaction and causation: causative constructions in Modern Standard Dutch. Journal of Pragmatics 27: 61-82. d oi: 10.1016/S037S-2166{96)00003-3 Wierzbicka, Anna. 1998. The semantics of English causative constructions in a universal-typological perspective. In T?te Mew Psychology of Language, Vol 1, Michael Tomasello (ed.), 113-153. Mahwah NJ: Lawrence Erlbaum Associates.