The aim of this book was to provide a new perspective on the semantics–pragmatics interface by combining insights from Construction Grammar and Relevance Theory. The main driving force behind this work was the commitment to obtaining cognitively accurate descriptions of language use. As mentioned in the introduction, such a commitment to cognitive plausibility and descriptive accuracy led to the development of the two frameworks. Given their respective domains of interest, however, they sometimes make different predictions and provide opposite analyses of the same linguistic phenomena. In Construction Grammar, more room is given to knowledge than in Relevance Theory, in which greater emphasis is placed on inferential processes. As a consequence, this naturally questions their capacity to achieve descriptive accuracy. The main challenge of this book was therefore twofold. Primarily, this contribution to the field can be viewed as an attempt to identify where descriptive accuracy actually lies. More generally, though, the aim was also to identify the extent to which the integration of the two frameworks generates further and better insights into the underlying mechanisms of verbal communication. It was not the aim of this book simply to act as a judge as to which framework is better at achieving descriptive accuracy, and therefore to draw up a list of concessions that the other theory has to make. Rather, my aim has been to enhance the explanatory power of each framework through their combination; the new model presented consists of more than just the sum of the two different approaches.
Needless to say, this integration first requires a thorough examination of the two theories and their respective strengths and weaknesses. This is why I introduced each of the frameworks individually in detail in Chapter 2.
In Chapter 3, the main focus was on lexical semantics–pragmatics. In this case, it was shown that CxG and RT have radically different understandings of the degree to which semantics and pragmatics actually contribute to the interpretation of a lexeme. First, it became apparent that a critical analysis of the notion of ‘concept’ was necessary. Both frameworks use this term, but in radically different ways. Moreover, it was shown that in RT the standard (i.e. referential atomic) view is incompatible with some of the central tenets of the theory. I argued that the best alternative is to define concepts in terms of (a rich body of) encyclopedic knowledge. The main challenge from this perspective is to explain exactly how this type of conceptual information enters the interpretation process of a lexeme. I have shown that viewing the interpretation of a lexeme as depending on both very rich semantics and strong pragmatic principles is possible and, in fact, actually accounts for the various observations made in the literature on lexical processing. This view was articulated around the notion of lexically regulated saturation (a term borrowed from Reference Depraetere, Cappelle and WadaDepraetere, 2010, Reference Depraetere2014), which captures the complex relation between semantics and pragmatics: in spite of the very rich semantics that a lexical item makes accessible, its actual function is systematically reconstructed in context in accordance with the principle of relevance. This naturally requires adopting a particular view on what counts as semantic content and pragmatic inferencing. The suggested analysis makes for a more powerful analysis than either of CxG or RT alone can.
In Chapter 4, the aim was to understand more specifically how the direct linguistic environment in which a lexeme occurs can affect the interpretation process of lexically regulated saturation. In RT, little attention is given to larger linguistic structures. By contrast, constructionists prefer to identify the larger structures in which lexical constructions are embedded. In the first part of the chapter, I focused on more grammatical (i.e. schematic) types of constructions. In CxG, the term coercion is typically used when a lexical item inherits part of its content from the grammatical construction in which it is used. The only problem with this notion is that CxG fails to explain in detail how the lexeme inherits the meaning associated with the grammatical construction. It was shown that RT can help to describe the exact nature of coercion. First, I argued that coercion is itself also primarily a pragmatic process carried out by the language user. More precisely, I argued that the actual nature of the interpretation process behind coercion is the same as in non-coerced cases, namely that of lexically regulated saturation, and that the function of the constructions in which lexemes occur is simply to act as an additional constraint on the recovery of the intended interpretation. The next step was to spell out exactly what the role of the constructions involved in the process of coercion is. It was shown that coercive constructions are formally (semi-)schematic (i.e. grammatical) and have a procedural function. This view in turn required defining what counts as procedural meaning. It was shown that there is as yet no clear and uncontroversial definition of this notion. I compared various approaches to procedural meaning (and grammatical encoding more generally) and put forward the tentative claim that the procedural nature of schematic constructions follows from their encoding meta-conceptual information, i.e. information about the type of concept which is expected to occur in a given position of a construction. This actually resulted in adjusting the strong views adopted in both CxG and RT with regard to the nature of coercion and procedural encoding (that the meaning of the grammatical constructions usually, though not systematically, wins over that of the lexemes found in them). In the final part of that chapter, I focused on more idiomatic (i.e. lexically fixed) constructions. It was shown that the interpretation of a lexeme is also often determined by larger sequences in which it occurs. The main challenge was to understand how these constructions enter into the interpretation process and how they affect the process of lexically regulated saturation usually performed for the individual words that occur in these patterns. In accordance with the approach developed in RT, I argued that interpreting utterances that contain idiomatic constructions also primarily involves pragmatic processes that are guided by the principle of relevance. That is, interpreting a sentence containing an idiomatic construction consists in a context-sensitive reconstruction of the most relevant interpretation, the function of the construction being a strong candidate but not the only possibility. When the hearer reconstructs the meaning of the idiomatic patterns, there are felt to be contextual effects whereby lexically regulated saturation is ‘suspended’. What the chapter therefore shows is that the pragmatic process of lexically regulated saturation is central to the interpretation of lexical constructions. This process is guided by the procedural semantics of the grammatical constructions in which lexemes occur, which can lead to coercion effects. Also, lexically regulated saturation may be suspended when the hearer recognizes the use of an idiomatic construction in which the lexeme is embedded. In each case, this involves an intricate interaction between one’s rich linguistic knowledge and relevance-guided pragmatic processes. This approach was then represented as in Figure 5.1.
‘Lexically regulated saturation’ and utterance comprehension (2)
Once more, this book was therefore meant to be an illustration that combining insights from the two theories helps to increase the overall explanatory power of each framework. Of course, this book contains a mainly theoretical investigation, and critics might rightly argue that the resulting model still needs to be put to the test. However, in a chapter in Depraetere et al.’s recent book, Models of Modals, I actually used this theoretical apparatus to look at modal verbs in English (Reference Leclercq, Depraetere, Cappelle and MartinLeclercq, 2023), and the discussion reveals that greater precision can be obtained when combining CxG and RT as I have done in this book. While emphasizing the inferential nature of the interpretation of modal verbs (in terms of lexically regulated saturation), I showed that a vast network of modal constructions in fact intervenes in the process. For instance, speakers of English also know a great number of modal ‘idioms’, such as those in (117) and (118), which may short-circuit lexically regulated saturation.
(117)
Subj must surely VP She must surely be the most beautiful in the world. (Reference Cappelle, Depraetere and LesuisseCappelle, Depraetere and Lesuisse, 2019: 231)
(118)
I can’t tell you how … Sweet loyal Jack. I can’t tell you how good it is to see you. (Reference LeclercqLeclercq, 2022: 239)
The ‘Subj must surely VP’ construction typically triggers a strong epistemic interpretation of must, as is the case in (117). Likewise, the verb can in (118) is not simply used (if at all) to communicate the speaker’s inability to be specific about all sorts of (metaphorical) quantities, but it actually puts emphasis on their extremely positive feelings with regard to the situation at hand, a function that is associated with the ‘I can’t tell you how … ’ construction. So the interpretation of these verbs is not just a matter of lexically regulated saturation. At the same time, understanding sentences like these does not simply consist of the recovery of these constructions and their function but rather results from the context-sensitive reconstruction of the speaker’s intended interpretation in accordance with one’s expectation of relevance. That is, understanding sentences like (118), for instance, is more complex than simply recovering the function of the pattern ‘I can’t tell you how … ’, which then only gets canceled in inappropriate contexts. Rather, understanding this sequence of words consists in a parallel, context-sensitive reconstruction of the speaker’s intended interpretation via considerations of optimal relevance.
In addition, Reference Leclercq, Depraetere, Cappelle and MartinLeclercq (2023) shows that modal verbs can also be used in more schematic modal patterns from which they inherit their function (i.e. whose function can be viewed as being ‘coerced’ onto the modal verbs). This is the case of the sentences in (119) and (120), for instance.
(119)
Subj Mod be Compl It’s so incredible what your mother did. She must be a saint. (COCA)
(120)
I don’t think we Mod VP I don’t think we may consider Trump as a “coronabuster”, on the contrary, but at least the cartoon is well rendered.Footnote 146
The pattern in (119) is typically associated with an epistemic interpretation, which the verb must here receives. Also, comparing corpus data from different sources, Reference Leclercq, Depraetere, Cappelle and MartinLeclercq (2023) establishes that speakers of English most probably store the construction identified in (120), which typically conveys an invitation not to perform the action denoted by the VP. In this pattern, it is usually necessity verbs that are used (e.g. I don’t think we should change things now, Reference Cappelle, Depraetere and LesuisseCappelle, Depraetere and Lesuisse, 2019: 232; And let me be clear on this. I don’t think we need to have our boots on the ground, Reference Leclercq, Depraetere, Cappelle and MartinLeclercq, 2023: 74), which means that the modal value is typically rather strong. The sentence in (120) inherits the general function of this construction, and the verb may, whose modal value is typically rather weak, seems to express a stronger kind of possibility meaning. Once again, this means that the interpretation process is not solely the result of lexically regulated saturation. At the same time, the resolution process involved, although undeniably being constrained by the semantics of both the modal verb and the construction in which it occurs, is again context-sensitive and only contextual relevance enables the hearer to derive the exact interpretation intended by the speaker. As mentioned in Chapter 4, the pragmatic (i.e. inferential) understanding of may in this sentence follows naturally from the procedural semantics encoded by the (grammatical) construction in which may occurs, the exact function of which precisely constrains the inferential process of lexically regulated saturation involved while interpreting the lexeme.
Modality is of course one domain of application, and further research is needed to show the benefits of combining CxG and RT. Still, beyond the various theory-internal developments suggested throughout this book, I hope to have shown that a better understanding of linguistic communication (and human cognition more generally) can be achieved when integrating insights from frameworks whose primary commitment is cognitive plausibility.
A number of further conclusions fall out from my investigation. On the question of whether it is possible to combine insights from CxG and RT, I hope to have convinced the reader that the answer is positive. More importantly, concerning the explanatory power of this integration, the various arguments developed show quite clearly that a deeper understanding of the semantics–pragmatics interface (in particular) and of verbal communication more generally is possible and that the respective approaches can indeed benefit from one another in order to achieve even more descriptive accuracy. It was shown that this sometimes requires articulating various aspects of language use in more complex ways than is sometimes the case in either framework. The aim of this book was not to provide a simple model, however, but one which makes more cognitively accurate predictions than CxG and RT manage to do on their own. Instead of reducing the domain of application of each framework, I believe this work therefore opens up a wide array of new research possibilities within (and across) these two already far-reaching theories. Here are some of the directions for future research.
Concerning the semantics–pragmatics interface, it might be important to carry out additional experimental work to test the hypothesis that the function of the (idiomatic) patterns does not necessarily have priority over that of the individual lexemes found inside them. The literature in psycholinguistics already seems to support this view, yet more evidence might be needed to further substantiate this claim. In addition, the discussion in Chapter 4 on the nature of grammatical constructions and procedural encoding also leaves open a number of questions. It was shown that providing a specific definition for what grammatical/procedural encoding consists of is necessary. I put forward the hypothesis that this type of meaning consists in a form of meta-conceptual encoding. This particular hypothesis also needs to be elaborated further and has to be tested experimentally. More generally, there are other areas of research that I think could benefit from integrating CxG and RT. For instance, I think it would be interesting to see how much the principle of relevance affects the gradual process of language acquisition and, therefore, the development of the construct-i-con. This observation applies equally well to the domain of language variation and change. In a similar spirit, it might also be interesting to investigate the extent to which combining CxG and RT can contribute to a better understanding of the cognitive strategies used by individuals with particular language impairments (e.g. dyslexia, aphasia, autism, Alzheimer’s, etc.). Finally, in more applied approaches to language, this integration might also be fruitfully applied in the context of discourse analysis when looking at the various strategies that speakers use in order to direct their speech in a particular direction.
Open access