Is free enrichment always free? Revisiting ad hoc-concept construction

In relevance-theoretic pragmatics, lexical adjustment, or ad hoc-concept construction, is considered a case of free pragmatic enrichment. It is hence included within the group of non-linguistically mandated processes that are automatically carried out during comprehension. Its output is an inferentially-specified representation of the conceptual content encoded, or activated, by open-class words like nouns, adjectives, verbs or adverbs. Ad hoc-concept construction is a necessary step to obtain a fully propositional, truthevaluable form that captures in a more precise manner the portion of meaning that the speaker communicates explicitly. However, this paper argues that this lexical pragmatic process needs not always be described as completely exempt from linguistic mandate: it may sometimes be enacted and steered by a series of linguistic elements. These fall within two categories: (i) the evaluative morphemes that some lexical items may receive in some languages, and (ii) lexical and phrasal items adjacent to content words. These elements would finetune the representations arising from content words as regards more specific, contextdependent features or shades. Moreover, they could add to such representations information about the speaker's attitude towards, or stance about, what the speakers refer to. Therefore, that information needs not always be derived as implicatures. © 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Relevance-theoretic pragmatics portrays verbal comprehension as an intricate and incredibly fast process contingent on interpretative hypotheses about the content explicitly and implicitly communicated. Hypotheses about the former depend on various sub-conscious, almost automatic, non-sequential tasks. Linguistic decoding yields a logical form, or a minimally parsed and structured string of conceptual representations. Since this form is not yet fully propositional and truth-evaluable, it needs to undergo a series of pragmatic enrichments. They make up the process known as mutual parallel adjustment Wilson, 1986, 1995;Carston, 2000Carston, , 2002Sperber, 2002, 2004).
One of these enrichments is disambiguation of the possible senses of lexical items and sentential constituents. Other enrichments are linguistically mandated, as they are enacted by specific elements (Carston, 2000;Jary, 2016). The procedures, or processing instructions, which some of those elements encode even constrain their output in precise manners (Blakemore, (i) Assigning reference to deictics, proper nouns and pro-forms like personal, anaphoric or cataphoric pronouns. (ii) Delimiting the time and duration of the actions alluded to by verbs. (iii) Completing semantically sub-propositional sentences (Carston, 2000(Carston, , 2002(Carston, , 2009Carston and Hall, 2017;Hall, 2017). (iv) Establishing temporal or causal relations between the referred states of affairs.
In contrast, other enrichments are non-linguistically mandated because they are automatically performed as a prerequisite to turn the logical form into a fully propositional form. Known as free enrichment, they include: (i) Supplying the unarticulated, or non-verbalised, constituents necessary to get a meaningful proposition, such as the location and/or time of an event, or the instrument with which an action is realised (Carston, 2000). 1 (ii) Constructing ad hoc concepts, or adjusted conceptual representations from open-class words like nouns, adjectives, verbs and adverbs.
Relevance theorists acknowledge that the output of the latter type of free enrichment is often determined by the interaction of open-class words with co-occurring linguistic material, paralinguistic elements and contextual information (Carston, 2000(Carston, , 2010a(Carston, , 2010bWharton, 2009: Ch. 6). However, its portrayal as a cognitive process that is completely exempt from linguistic mandate should be revisited: the need for ad hoc-concept construction may be overtly indicated by certain types of linguistic elements that accompany, or are added to, content words. Unfortunately, the list of such elements has thus far been underexplored by relevance theorists.
Furthermore, as regards co-occurring linguistic material, research from a relevance-theoretic perspective has not specified whether such material is only limited to other lexical items or could also include certain morphological elements and stylistic choices. Therefore, a typology also seems most needed, since such elements and choices can constrain the direction and output of ad hoc-concept construction in precise manners by prompting the addition of a wide array of beliefs. This involves admitting that ad hoc-concept construction is not always a totally free pragmatic enrichment process.
This paper aims to show and discuss the variety of linguistic elements that may work as ostensive, intentional markers or signals indicating that ad hoc-concept construction is mandatory. Such elements will be argued not only to point to a need for specific conceptual adjustments, but even to direct them. Thus, this approach lends support to a conception of ad hoc-concept formation as an extremely context-sensitive process that operates on the grounds of a variety of stimuli. Furthermore, this approach stresses the complexity, heterogeneity and richness of the output of this process. Indeed, it may encompass varied information about a portion of the speaker's meaning other than specific features, properties or shades of what she refers to through content words: information concerning her attitudes and psychological states. 2 Lastly, this approach involves reconsidering the status of information that has often been claimed to be derived as implicatures. In fact, it contends that such information may be represented as a consequence of lexical pragmatic processes and may feature as part of conceptual entities provided two criteria are satisfied: such information must fulfil a defining function and contribute to identifying what the speaker actually denotes or refers to by means of a specific lexical item at a particular moment of speaking.
The paper continues by offering some notes on ad hoc-concept construction and the two approaches to it. Next, it classifies, discusses and illustrates the list of linguistic elements triggering and guiding this pragmatic process. The illustration will be done mainly through examples from Spanish and English. To conclude, this paper offers some remarks on ad hocconcept construction and suggestions for future research.

Ad hoc-concept construction
In relevance-theoretic pragmatics, lexical items are differentiated depending on whether they are endowed with a conceptual or a procedural semantics (Blakemore, 1987(Blakemore, , 1992(Blakemore, , 2002Wilson and Sperber, 1993). Items with the former semantics are said to encode concepts. These are denotational mental entities comprising three types of entries Wilson, 1986, 1995: 88e89): a) Logical entry, which includes the defining properties of what the concept denotes. b) Lexical entry, which subsumes data about the natural-language word needed to verbalise it and its pronunciation. c) Encyclopaedic entry, which groups personal and/or cultural information about what it refers to.
Relevance theorists initially endorsed the assumption that the concepts encoded by words belonging to open grammatical classes are not unalterable and do not enter inferential processes in a static manner. Rather, they are malleable and what they represent on particular occasions is amenable to context-sensitive specification (Barsalou, 1983). Accordingly, nouns, adjectives, verbs and adverbs just pack some sort of representational skeleton, or pro-concept, and procedures triggering the finetuning of their denotation (Sperber and Wilson, 1998;Wilson, 2011;Hall, 2017). In other words, concepts were thought to be fairly schematic and to need inferential adjustment. 3 Relevance theorists also distinguished two conceptual, or lexical, adjustments. On the one hand, broadening or loosening of the denotation towards a less specific notional space. This involves elimination of some defining property in a concept's logical entry (Carston, 2000(Carston, , 2002Carston, 2006, 2007). In (1) below, the resulting concept FLAT* would be deprived of the requisite of absolute flatness, as the mentioned plateau does have hills and elevations: 4 (1) The Spanish central plateau is flat.
On the other hand, narrowing or strengthening of the denotation towards a more restricted notional space. This involves turning idiosyncratic information in the encyclopaedic entry into a logical, defining characteristic. In (2) below, SING* may be specified as referring to singing a particular type of songs, singing alone or in a choir: (2) Peter likes singing.
Both adjustments were posited to be contingent on the concepts encoded by other adjacent words and assumptions about the communicative situation and the speaker's intentions (Carston, 2002: 339;Hall, 2017: 93). Moreover, both adjustments were believed to not be mutually exclusive, but to be sometimes operated jointly, as when words are metaphorically used. Here, a defining property was considered to be dropped while some idiosyncratic characteristic was promoted to a definitional, almost quintessential one. To put it simply, conceptual adjustment consists of a rearrangement of stored information (Hall, 2017: 93). Its result was labelled 'ad hoc concept' and was pictured as an occasion-specific mental representation. Ad hoc concepts were not seen as composite objects, but as atomic, monolithic, perhaps one-off, entities occupying a moment in an individual's cognitive activity.
More recently, some relevance theorists deny that open-class words possess denotational content. They regard such words as conceptually underspecified, which enables them to communicate diverse meaning shades or nuances depending on context and available information (Carston, 2013a(Carston, , b, 2016Wilson, 2016). 5 They treat such words as purely procedural elements that activatedor open, so to saydtemporary mental files, folders or labels amenable to housing diverse beliefs (Hall, 2017: 97). The amount, type and content of those beliefs will depend on other sentential material, paralanguage, available manifest information and general knowledge. In this view, ad hoc concepts are not the result of a mental task modulating already packaged raw denotational material, but specifically-created addresses in memory that connect with, or give access to, context-dependent information. Accordingly, the mental file CHEESE* that a hearer would create upon processing (3) below would store beliefs concerning a type or brand of cheese that the speaker enjoys, perhaps at room temperature, cut in cubes, very thinly sliced or accompanied with nuts, raisins or quince, while she has a glass of sherry and listens to one of Chopin's nocturns in her favourite armchair during a rainy autumn evening: (3) Oh, this is cheese! These beliefs may feature in a mental file as a consequence of the speaker's pitch, tone of voice, paused pace when speaking and facial expression of delight or satisfaction, as well as because of a paralinguistic element like the interjection, a sentential property like exclamatory syntax and a prosodic feature like the sentence stress falling upon the last word. Obviously, changes in paralanguage affect the content of such beliefs. Moreover, a mental file may also house beliefs about affective attitudes like love, dislike or scorn; emotions or feelings like happiness, sadness or irritation, or cognitive states such as (un)certainty, caution, scepticism, etc. (Padilla Cruz, In press a).
Regardless of whether ad hoc concepts amount to adjusted conceptual material or mental files, some linguistic items seem to overtly indicate that their construction is called for. The role of such items would be to point the direction that conceptual adjustment should take or the sort of beliefs to be entertained and sent to a mental file. What follows classifies and exemplifies these varied elements. Yet, the list is not intended to be exhaustive, as its members are language-specific. It should therefore be understood as an open catalogue that may subsequently increase with elements from other languages fulfilling similar functions.

Linguistic triggers of ad hoc-concept construction
The linguistic elements that may enact and contribute to ad hoc-concept construction fall into two categories. While the first one groups morphological elements like evaluative morphemes, the second one includes a variety of lexical and phrasal elements, as well as some stylistic choices. They will be referred to as 'triggers'.

Morphological triggers
Languages like Spanish, Italian, Modern Greek or Jordanian Arabic avail themselves of a wide variety of morphemes, among which feature the diminutive and the augmentative. Classified as derivational morphemes, they are highly productive and can be added to nouns, adjectives, adverbs and even verbs. The diminutive indicates appraisals about smallness, littleness or scarcity, whereas the augmentative expresses assessments about largeness or excessiveness. Hence, both morphemes are regarded as evaluative (Volek, 1987;Wierzbicka, 1991;Bosque, 2009). They are considered to change the denotation of the word to which they are attached in terms of size or amount.
However, the diminutive and the augmentative serve other functions (Sifianou, 1992;Mendoza, 2005;G omez Torrego, 2002;Badarneh, 2010 In friendly, close and laidback relationships, both morphemes may also be used to (iv) voice socially-beneficial attitudes and feelings like affection, endearment, admiration or surprise towards the hearer or something connected with him (Mendoza, 2005;N añez Fern andez, 2006). Thus, both morphemes boost his positive face (Brown and Levinson, 1987) and maintain or enhance rapport (Spencer-Oatey, 2008). They suggest that the interlocutors belong to the same group whose members share similar viewpoints, experience similar feelings and/or hold similar attitudes: Social proximity furthermore facilitates the use of both affixes in order to (v) suggest, assert or emphasise camaraderie. A (fairly) smooth, intimate, joking or playful relationship may even prompt their employment for mild insults as tokens of mock impoliteness (Mendoza, 2005 When face is not at stake, the diminutive and the augmentative may finally (vi) aggravate the offensiveness, abusiveness or aggressiveness of actions like insulting, offending or criticising: 6 (9) a. ¡Eres un coñazo! 'You are a pain [þAUG] in the arse!' b. ¡Vaya apartamentucho ha alquilado María para el verano! 'What an apartment[þDIM] Mary has rented for the summer!' they are attached in terms of littleness/largeness or low/high amount or degree. Both morphemes could be thought to encode some feature that enables them to specify the denotation of the term receiving them in a 'non-serious', perhaps non-literal manner (Dressler and Merlini Barbaresi, 2001). Although that feature would trigger more specific concepts, the resulting concepts need not always be 'non-serious' or non-literal. Rather, they could be highly idiosyncratic, complex and contextsensitive (Padilla Cruz, 2020a). When the diminutive is attached to 'coche' ('car') in (4a), the noun seems to allude to a modest, inexpensive or average sub-compact vehicle, probably lacking (many) extras, while attachment of the augmentative causes it to refer to the opposite type of car. In (4b) the diminutive makes the adverb 'cerca' ('close/nearby') denote an unexpected or surprising closeness that will not involve much travel time. Similarly, in (5a) the diminutive conveys the idea that Mary came more dressed up than usual, than expected or than required for the occasion; in (5b) the augmentative suggests that the baby is well-built, big for his age or larger than other babies, and in (5c) the diminutive gives the impression that the speakers went to bed (much) later than usual. 7 In turn, in (6a) the diminutive helps the speaker communicate that the shop where she got some drinks was perhaps old, rundown, gloomy or seedy, whereas in (6b) it signals that she ascribes to the attacking bird any of a range of qualities like dangerousness, fierceness, wretchedness, hideousness, nastiness or despicability.
In contrast, in (7a) the diminutive would hint at the speaker's love, admiration or affect towards the hearer's boyfriend, while in (7b) the augmentative would evidence the speaker's astonishment, admiration or surprise at the hearer's new car. Likewise, in (8) addition of either the diminutive or the augmentative makes manifest that the hearer is regarded as the type of sly, cunning, witty, mischievous, restless swindler that the speaker adores. Finally, the augmentative in (9a) unveils that the speaker considers the hearer an outstanding, despicable, annoying or irksome moron, while the diminutive in (9b) evinces the speaker's dislike, scorn, disgust or aversion towards the apartment in question.
Following the initial relevance-theoretic approach, these two morphemes could be said to effect, and somehow steer, the adjustment of the concepts encoded by the nouns, adjectives, adverbs and verbs to which they are attached. They somehow signal that beliefs pertaining to subjective or objective judgements, idiosyncratic properties or specific nuances of what those words denote, or even to the speaker's feelings, emotions or attitudes towards, or because of it need to be given prominence in its conceptual representation. The result would be finetuned, more delimited and perhaps occasion-specific, mental objects on the grounds of such beliefs.
Since these two morphemes would trigger and constrain conceptual modulation, they would be procedural elements (Padilla Cruz, 2020a, Padilla Cruz, In press a). They encode some content, but that content is not a stable concept or notion; it is computational and amounts to processing instructions yielding occasion-specific, context-sensitive and highly idiosyncratic conceptual representations. Even if temporarily, the diminutive and the augmentative raise (an) item(s) of encyclopaedic information concerning the speaker's views, opinions, standards, appraisals, feelings, emotions and/or attitudes towards the status of a defining property, while preserving information about other properties included in the logical entry of the concepts encoded by the words to which they are appended. Therefore, they guide conceptual narrowing in such a manner that specific characteristics, components, considerations, feelings, emotions and/or attitudes are brought to the fore, thus restricting the denotational potential of such words.
In the more recent view, the semantics of the diminutive and the augmentative would also be considered procedural. However, it would contribute to the creation of purpose-built notional files, folders or addresses in memory. The type, content and amount of the beliefs to be harboured in them would be somehow determined by these morphemes. Accordingly, the file TIENDA* (SHOP*) resulting from the attachment of the diminutive in (6a) could store beliefs pertaining to the characteristics of the shop in question, or the poor variety of bad-quality products it sold. Among the beliefs that an occasion-specific mental file or folder may house, there may feature beliefs about the speaker's feelings, emotions and/or attitudes towards what she refers. 8 The need for lexical pragmatic processes may be indicated, and their result constrained, by evaluative morphology. The diminutive and the augmentative may highlight certain beliefs concerning specific properties of what speakers refer to or the speakers' psychological states. In the classical relevance-theoretic approach, those beliefsdthough perhaps not salientdwould likely feature in the encyclopaedic entry and be shifted to the logical entry. In the more recent approach, they would be incorporated in a mental file. Since they occupy a place in the notional conglomerate activated by content words, where they play a delimiting role, and are brought to the fore by these affixes, they would not count as implicatures.
The enactment of ad hoc-concept construction does not seem limited to morphological constituents of content words. A further group of linguistic elements may also trigger it.

Lexical and phrasal triggers
The second group of linguistic elements indicating and steering ad hoc-concept construction gathers (i) lexical items occupying the determiner positiondarticles, demonstratives and possessivesdand (ii) prototypical qualifying adjectives, as 7 Access to encyclopaedic information about the speakers and/or cultural information about certain types of people or habits in a particular place could enable the hearer to infer the reasons why the speakers went to bed late: perhaps because they watched TV, chatted, drank a lot or partied, for example. 8 Clearly, these examples presuppose a neutral or positive intonation. The content of the beliefs to be stored in a conceptual file may certainly be affected by factors other than intonation: pitch, facial expressions or gestures. For more details about the impact of evaluative morphemes on lexical pragmatic processes, see Padilla Cruz (2020a).

M. Padilla Cruz
Journal of Pragmatics 187 (2022) 130e143 well as lexical and phrasal items fulfilling adjectival functionsdnouns, noun phrases and participles. Moreover, this group also includes the so-called (iii) expressive expletives and (iv) evidential participles, which, despite their resemblance, do not behave as prototypical adjectives. Lastly, (v) certain stylistic tactics, such as lexical repetition and rephrasing, also feature in this group.

Determiners
Determiners accompany nouns and specify their reference. Central determiners precede nouns in languages like English or Spanish. They include articles, demonstratives and possessives (Greenbaum and Quirk, 1993).
Traditionally, the definite article is claimed to situationally identify the nominal referent as unique, while the indefinite article indicates that it cannot be uniquely identified (Gundel et al., 1993). In relevance-theoretic pragmatics, both articles have recently been argued to contribute to lexical pragmatic processes (Scott, 2019). The indefinite article in (10) below, for instance, adjusts the concept LADY in such a way that it refers to an unknown woman. Or, alternatively, it causes the addition of beliefs concerning the fact that the lady is unknown to the speaker to the mental file that it generates: (10) There is an old lady standing at your office entrance.
Demonstratives are usually portrayed as guiding the hearer to the intended referent by situating it in terms of proximity to, or distance from, the speaker and/or the hearer (Ariel, 1990;Reboul, 1997). They somehow signal the space where the hearer must search for it (Strauss, 2002). In relevance theory, they have been analysed in purely procedural terms: they encode procedures facilitating reference resolution when the intended referent cannot be uniquely identified in the discourse context (Powell, 1998(Powell, , 2001. Such procedures indicate its spatial, temporal, emotional, mental or discourse proximity or distance to a deictic centre, as opposed to that of competing referents (Scott, 2009(Scott, , 2013(Scott, , 2019. Thus, demonstratives affect the derivation of explicit meaning.
Demonstratives may be accompanied by overt or subtler clues of the speaker's signalling intention, like pointing gestures, gaze direction or directional nodding, and fulfil a truly demonstrative function. When such clues are missing, demonstratives work as indexicals. Nevertheless, if demonstratives assist referent identification by directing the hearer, they may somehow contribute to specifying the concept associated with it. In addition to the location of what it denotes, they would send assumptions about its characteristics to the logical entry of its associated concept. In the more recent account, such assumptions would find a place in the mental folder opened by the adjacent noun.
When the intended referent is highly accessible, there may be no need to locate and/or identify it. Demonstratives have then been argued to generate distinct effects: suggesting the speaker's attitude towards, commitment to or involvement with what she alludes to. This is possible thanks to the contrast that they would implicitly set with alternative referents in the discourse context. Thus, if a proper name refers to an easily identifiable individual, the demonstrative focuses on just one conception of that individual and creates a contrast with other conceptions. It is this emphasis that would enable the speaker to communicate her attitude towards, or feelings about, that individual: (11) This Robin de Locksley is amazing! (adapted from Scott, 2009) Yet, such a contrast could not arise unless the activated concept was narrowed down: out of the possible features, properties, facets, aspects or shades connected with its denotation, a specific one, or a subset of them, needs to be selected as quintessential, even if only temporarily. Accordingly, the demonstrative in (11) would specify the concept alluding to the person in question in terms of a (series of) feature(s), such as his bravery, intrepidity, fairness, generosity or impartiality, which may be contextually salient or inferentially determined on the grounds of contextual information. Albeit momentarily, such (a) feature(s) would be given prominence and presented as definitional. Hence, ROBIN DE LOCKSLEY* would refer to a known person who is remarkable for his intrepidity, for example.
But the concepts activated by the nouns with which demonstratives co-occur may be adjusted on the grounds of the speaker's feelings about and/or attitudes towards their referents. Consider, for instance, the sort of texts often accompanying some photograph postings on social networks: (12) a. These guys! b. Those magical days… If (12a) appears below a picture showing a gorgeous group of happy friends, arms on each other's shoulders, as they were having some beers, partying or in front of an attraction they visited, instead of fulfilling an identifying function, the demonstrative probably adjusts GUYS on the grounds of assumptions to the effect that the guys in question are very special to the poster, she loves or admires them. Likewise, if (12b) accompanies a portrait of a happy and loving mother with her two children in the countryside on a beautifully bright autumn day, the demonstrative could help specify that the charm of the alluded days stems from a (mix of) feeling(s) of togetherness, complicity, happiness, excitement, amusement or satisfaction. In the recent approach, the demonstratives would send beliefs about such feelings to a conceptual file. 9 Surely, the resulting concepts and the fact that the texts accompany visual postings might also give rise to a wide array of weak implicatures. In the 9 As a reviewer points out, the interaction of the poster's beliefs and attitudes with the images significantly affects the resulting ad hoc concept. If the group of guys in the former picture exhibited the trophy of a hunting weekend, and the poster was against that practice, GUYS* might subsume beliefs about her negative attitude towards them. Likewise, if the latter picture showed toys on the floor, peanut butter on the curtains and cracker crumbs all over the room, MAGICAL DAYS* might surely house beliefs about different attitudes and feelings of the poster.

M. Padilla Cruz
Journal of Pragmatics 187 (2022) 130e143 case of (12a) these could concern the poster's feelings of pride in people like these or elatedness with memories about them, while in that of (12b) such implicatures might regard her nostalgia or longing for those days. Lastly, possessives express some relationship between a nominal referent and any of the interlocutors. A similar argument could be extended to possessives. In some cases, they would modulate concepts in such a way that their referent is presented as owned by (a) particular individual(s), thus identifying the owner. In other cases, possessives may specify referents in such a way that they are related to the speaker or some other individual(s) in a different manner. Accordingly, the ad hoc concept resulting from (13) below would not allude to a roommate that is the property of the speaker, but to a person with whom she is sharing a room at a particular time: (13) My roommate is very funny and easy-going.
Moreover, in some southern varieties of Spanish, as well as in Modern Greek, possessives may precede or follow proper names as a way of conveying overtones of social proximity, familiarity, endearment, affection, fondness or tenderness. In these cases, the possessive could cause the storage of assumptions concerning the speaker's feelings about the person she refers to within the mental folder activated by the proper name: (14) Mi Cristina es muy lista.
'My Christine is very clever'.
Along with determiners, nouns may be modified by adjectives or adjective phrases. These may also finetune conceptual representations.
Four features characterise adjectives (Greenbaum and Quirk, 1993: 129): a) Occurrence in attributive position as noun premodifiers. 10 b) Occurrence in predicative position as subject complements. c) Possibility of being intensified by adverbs like 'very' or 'highly'. d) Availability of comparative and superlative forms.
The main function of adjectives is to modify the designation of the noun they accompany: while adjectives in attributive position tend to designate a relatively permanent property of its referent, those in predicative position are neutral as regards the temporariness or stability of the property of the nominal referent (Cruse, 2011). Thus, the information contributed by an adjective restricts the set of potential referents of the concept encoded by the noun they modify, so adjectives can also serve an identifying function (Scott, 2008). This restriction has been depicted in semantics as a process of intersection between both the set of all possible entities denoted by a noun and the set of entities amenable to having the property denoted by an adjective (Griffiths, 2006: 37;Cruse, 2011: 313).
Many combinations of adjectives and nouns are endocentric, as what they denote is of the same ontological type as what the noun alone denotes. Hence, the adjective merely constrains the applicability of the noun. But there are fully restrictive adjectives that limit the reference of the noun exclusively, particularly or chiefly (Greenbaum and Quirk, 1993: 143): (15) a. The only occasion. b. A certain person. c. The chief excuse.
Other adjectives have heightening or lowering effects on the referent of the noun, so they can work as emphasisers (16a), amplifiers (16b) or downtoners (16c) (Greenbaum and Quirk, 1993:  Still, other adjectives must be interpreted relative to some standard pertaining to the category to which the referent of the noun belongs. Therefore, the noun determines the precise interpretation of the adjective, while this also limits the applicability of the noun. These adjectives are known as relative descriptors (Cruse, 2011: 314): 10 Some adjectives are restricted to this position or occur predominantly in it and do not characterise the referent of the noun directly, as in "an old friend of mine".
In contrast, other adjectives negate the referent of the noun and indicate where to look for the intended one. Hence, they are negational descriptors (Cruse, 2011: 314): 11 (18) a. A former President.
b. An imitation fur coat.
These possibilities suggest that adjectives could also modulate the denotation of the concept encoded by the noun they cooccur with. Thus, 'true' in (16a) would cause SCHOLAR to designate a highly committed, hard-working, fruitful, insightful and inspiring researcher, while 'former' in (18a) would make PRESIDENT simply denote a person who no longer has the presidency. In the more recent view, the adjective in (16a) would endow the conceptual file opened by 'scholar' with assumptions concerning the characteristics and capacities of the scholar in question, while that in (18a) would place assumptions concerning the expiry of the term of office in the mental file created by 'president'.
Instead of the properties denoted by adjectives being individually represented in a logical form, adjectives would narrow the concept that the noun activates. The material packed in them, so to say, would be integrated in, or merged with, that encoded by the noun, thus enriching it. Accordingly, the prenominal adjective and adjectival noun phrase in (19) would not give rise to slots in the logical form to be filled by concepts like FABULOUS and FIVE-STAR. Rather, perhaps they delineate HOTEL as an astonishingly luxurious, top-quality, trendy lodging, probably with a private shopping gallery, an enormous swimming pool and incredibly well-equipped, elegantly-decorated rooms with spacious terraces facing unsurpassable views: (19) Last summer we stayed in a fabulous five-star hotel on the seafront.
In the more recent account, that information would feature in an occasion-specific mental file. Similar effects could have elements with adjectival functions like nouns, noun phrases, present and past participles, and even defining and non-defining relative clauses. Likewise, other linguistic elements occurring in adnominal position without properly adjectival functions, namely expressive expletives and evidential participles, may flesh out concepts.

Expressive expletives
Languages like English or Spanish also utilise seeming nouns and adjectival participles that appear to modify nouns or adjectives: 'fucking', 'damn(ed)', 'bleeding', etc. With a fairly elusive semantics, they do not behave as adjectives because they do not refer to qualities or conditions. Nevertheless, they resemble adverbs like 'extremely', 'incredibly', 'surprisingly' or 'shockingly', and have heightening or intensifying values (Huddleston, 1988;Greenbaum and Quirk, 1993): (20) They served fucking/damn good nibbles at the reception. This elusiveness places them within the category of expletives and enables them to work as expressives (Potts, 2007). They can thus voice a wide variety of affective attitudes, feelings and emotional states ranging from shock, surprise, unexpectedness, puzzlement, pleasure, happiness, joy or euphoria, to irritation, annoyance or even wrath: (21) I forgot to return the fucking/damned book! In (21), the target of the affective attitude, feelings or emotional states that the speaker verbalises by means of the expressive expletive would be the referent of the head of the noun phrase working as the direct object, as the expletive immediately precedes it. However, the speaker might be interpreted to voice her attitude, feelings or emotions about the referent of another sentential constituent, namely the subject. For this interpretation to be possible, the expressive expletive would need to somehow 'hop' towards the left in the sentential structure (Gutzmann, 2019;Bross, 2021). 12 Moreover, the speaker might be thought to express her attitude, feelings or emotions not about the referent of a particular sentential constituent, but about the state of affairs described in the utterance. This would require argument-extension and license a sentence-level interpretation (Gutzmann, 2019;Bross, 2021). 13 A pragmatic approach known as the speech-act hypothesis accounts for these possibilities by surmising that an expressive expletive can accomplish a quasi-independent speech act which conveys a not-at-issue content that is different from the content of the sentence where the expletive appears, which amounts to at-issue content (Frazier et al., 2015). This approach also postulates that, for the speaker's affective attitude, feelings or emotions to have been expressed about the referent of the subject constituent of a sentence, regardless of the position of the expletive within the sentence, the referent of this 11 Additionally, some adjectives may give rise to different readings, like 'beautiful' in "a beautiful dancer": a predictable reading can be that someone is a dancer and is beautiful, but another may refer to someone who dances beautifully. Other adjectives vary as regards the characterisation of the whole entity referred to by the noun or just a part of it, as in the case of 'red' in "a red hat"di.e., the whole hat is reddor "a red book"di.e., only the cover is red (Cruse, 2011: 314e315).
12 Gutzmann (2019) labels this syntactic movement argument hopping and this particular case left hopping. 13 Still, two further interpretative possibilities could arise when the sentence where an expressive expletive occurs is subordinated to a matrix clause, as in "Mary said that I forgot to return the fucking/damned book". On the one hand, the attitude, feeling or emotion that the expletive expresses may be felt to have been voiced about the referent of the subject of the matrix clause. This yields a matrix-subject interpretation. On the other hand, that attitude, feeling or emotion could be perceived to have been expressed about the whole propositional content of the entire complex sentence, regardless of whether the expletive appears in the subordinated clause. This involves a matrix-clause interpretation (Bross, 2021). constituent needs to have agentive-causer nature. In other words, the subject must denote an individual who may be held responsible for the action described in the sentence. This is the so-called culprit hypothesis. 14 The speech-act hypothesis, however, has been seriously challenged from a syntactic perspective, which contends that expressive expletives can only have local interpretations. Interpreting that the speaker's attitude, feeling or emotion are expressed about the referent of a constituent other than the constituent next to which an expressive expletive is placed, or about the state of affairs described in the whole utterance, are illicit possibilities because the expletive carries an expressive feature (Gutzmann, 2019). This feature automatically triggers a local interpretation whenever the expressive expletive precedes a nominal head. For the referent of another syntactic constituent to be the target of the speaker's attitude, feeling or emotions, the expletive would need to be deprived of this feature, which would have to be assigned to the head of a determiner phrase in order to be subsequently eliminated. In turn, sentence-level interpretations are considered to not be syntactically motivated, but to originate solely as implicatures (Gutzmann, 2019: 107). Indeed, a sentence-level interpretation may be ruled out through the addition of an attitudinal adverbial that is semantically opposed to the expressive expletive appearing in a sentence: 15 (22) Fortunately, I forgot to return the fucking/damned book.
Recent tests of the claims made by both the pragmatic and syntactic approaches about the distinct interpretations that utterances containing expressive expletives could receive have yielded inconclusive results, thus pointing to the need to consider the role of pragmatic factors and mechanisms in the interpretation of such utterances (Bross, 2021: 12). 16 Regardless of whether the presence of an expressive expletive may give rise to a sentence-level interpretation, its attachment to a phrasal head could also be thought to effect and assist the lexical pragmatic process so far regarded as free enrichment. If open-class words possess denotational content, these expletives will adjust it. Accordingly, the adjustment in (20) would yield a narrowed concept referring to outstanding, extremely delicious, varied and sophisticated nibbles. If open-class words lacked that content, expressive expletives would help create a mental file and determine the amount and content of beliefs to be stored in it.
Some beliefs might pertain to the speaker's attitudes, feelings and/or emotions. Thus, in (21) the expletive could cause the file BOOK* to carry assumptions about the speaker's scorn or disparaging attitude towards the book in question. In addition to the aforementioned syntactic criteria, the fact that such attitudes are held towards the book, and not towards the proposition expressed, is evident from: 'I forgot to return the fucking/damned book/s' (24) En la recepci on sirvieron aperitivos jodidamente buenos. 'They served fucking/damn good nibbles at the reception' Along with pitch, intonation, facial expression and gestures, the expletive may nonetheless bring about a representation of the speaker's overall irritation because of her forgetfulness or any other negative emotion that she is thought to experience while speaking. That propositional-attitude representation does take the proposition expressed under its scope.

Evidential participles
English and Spanish also possess some past participles which, unlike typical adjectives, do not denote permanent, inherent, temporary or accidental properties or states (Huddleston, 1988;Haegeman and Gu eron, 1999;Collins and Hollo, 2000;B€ orjars and Burridge, 2001). Rather, they indicate that a nominal referent should not be definitively alluded to by means of a particular term, as further evidence or more trustworthy information should confirm that the referent actually did something or can be regarded as a particular type of person or state of affairs. Examples are 'alleged', 'supposed' or 'suspected'. They convey the speaker's epistemic stance about the adequacy of a linguistic label for that referent: uncertainty, scepticism, doubt, reservation, caution or reluctance to admit something about it. Therefore, such participles are evidential participles (Padilla Cruz, 2020b): 14 Both the speech-act and the culprit hypotheses were informed by the responses to a questionnaire administered to 48 undergraduate students (Frazier et al., 2015). 15 Gutzmann's (2019) view found support in the assessments made by 60 undergraduate students in a questionnaire. 16 For a discussion of some of the drawbacks of these approaches, see Padilla Cruz, In press a. 17 See Padilla Cruz, In press b for a more detailed account.
Unlike prototypical adjectives, which evidential participles resemble to some extent, evidential participles do not tend to be intensified and can only occupy attributive position: (26) a. *Very alleged murderer arrested.
b. *The murderer is alleged These participles cannot have comparative or superlative forms either: (27) a. *The murderer is more alleged than suspected. b. *The murderer is most alleged.
The transitive nature of their corresponding verbs would be expected to license passivisation. However, replacement of evidential participles with a passive form yields an odd sentence (28a). Moreover, despite the transitivity of the corresponding verbs, the individual alleged to have performed some action cannot be their direct object (28b). For the corresponding verbs of evidential participles to be passivised, the participles would need to be followed by an infinitival clause (28c): (28) a. *The murderer who was alleged.
b. *The police alleged the murderer. c. The murderer who is alleged to have killed three children.
Evidential participles alert the audience to the believability of an assertion, or more precisely, of a portion of an assertion: a particular characterisation or labelling of some individual or state of affairs. They point to the (in)adequacy of a particular label on the grounds of the supposed trustworthiness or reliability of the information source available to the speaker upon using it (Dendale and Tasmowski, 2001;Aikhenvald, 2004;Matthews, 2007). Since that information source may be perceptual or epistemological (Cornillie, 2007)dvisual, non-visual but directly perceptible, acquired through some other person's testimony, etc. (Aikhenvald, 2004)devidential participles ultimately indicate the speaker's degree of commitment to the adequacy of an employed characterising term and, hence, to an asserted content where that term appears. Indeed, the origin of the knowledge upon which the speaker relies when asserting that content and utilising that term might be another person's testimony or that knowledge might even be incompatible with her own doxastic universe (Crystal, 1991;Nuyts, 2006). Thus, such participles caution the audience to question or be prudent about a linguistic label and a propositional content where that label occurs (Wilson, 2012;Piskorska, 2016).
Evidential participles behave like a wide array of evidential or hearsay adverbials, particles and parenthetical phrases, which have also been found to reveal the type of evidence that a speaker takes into account when making a claim: (29) a. Reportedly, the murderer killed the three children.
b. Evidently, the murderer killed the three children. c. The murderer killed the three children, they say.
Such adverbials, particles and phrases have been included in the broad category of evidentials (Ifantidou, 2001). Their members are believed to raise the activation of an ensemble of mental mechanisms assessing the benevolence and trustworthiness of informers, as well as the reliability of the information that they dispense (Wilson, 2012(Wilson, , 2016Padilla Cruz, 2020b): epistemic vigilance mechanisms . Such mechanisms scrutinise a variety of factors affecting judgements about the reliability of an informer and imparted information (Origgi, 2013). As a consequence of their operation, hearers can create mental representations of the speaker's (un)certain, sceptical, cautious or dubious attitude towards the content that she communicates, which are known in relevance-theoretic pragmatics as higher-level explicatures: (30) Because of unreliable testimony the speaker (firmly) disbelieves that p.
These representations differ from other representations of the speaker's attitude towards the factuality of a proposition, which hearers construct as a result of a variety of markers of epistemic modality: (31) The speaker (firmly) believes that p.
Although evidential participles also increase the activation of epistemic vigilance mechanisms, perhaps they do not trigger mental representations of the speaker's commitment to the truth of some state of affairs alluded to through an entire proposition. As in the case of expressive expletives, the fact that such representations are not enacted by evidential participles would be attested by the morphosyntactic characteristics of these participles: namely, placement in adnominal position, possibility of co-occurrence with an evidential or hearsay adverbial, and gender and number agreement with the modified noun in inflectional languages. Rather, evidential participles would make manifest assumptions about the source of information available to the speaker when referring to such a state of affairs in a particular manner and thus about her commitment to it actually being the case. Those assumptions could settle in the conceptual file activated by the label that the speaker applies to it while showing her (un)certainty that the state of affairs in question can actually be referred to by means of that label. Thus, evidential participles contribute to fleshing out that cognitive entity and give rise to the formation of an occasion-specific, highly idiosyncratic conceptual element.
As with expressive expletives, evidential participles would determine the sort of assumptions that may be accommodated in idiosyncratic mental files. Some of these assumptions would certainly concern the speaker's uncertain, sceptic, doubtful, reserved, cautious or reluctant stance about the fact that a person actually did something or an event really amounts to a particular state of affairs. Others would pertain to personal estimates, judgements or considerations as to the reasons why the speaker is not yet prone to considering someone or something in a particular manner: lack of supporting evidence, dependence on unreliable testimony or untrustworthy individuals, need for confirmation or investigation, etc. Their actual number and content would also depend on available contextual and encyclopaedic information. Thus, the evidential participle in (25) above may give rise to some sort of complex notional conglomerate paraphrasable as "the person who is thought/ surmised/hypothesised/suspected/ventured to have killed a young lady because he was seen at the crime scene right before the victim was found dead". 18 Lexical pragmatic processes may be affected by stylistic choices, too. Speakers often repeat certain lexical items and reword. Repetition and rewording may then yield context-sensitive concepts.

Stylistic choices
As the reappearance of some formal element, repetition may involve individual lexical items, phrasal constituents or larger stretches (Bazzanella, 2011). Regardless of its monological or dialogical nature, its appearance in discourse has often been negatively assessed (Ochs and Schieffelin, 1983) due to its associations with redundancy, disfluency or sloppiness (Schegloff, 1987;Bazzanella, 2011). Yet, it may have a variety of cognitive, textual, stylistic, argumentative, conversational, interactional and ethnic purposes (Bazzanella, 1993). For instance, in adultechild interaction, it is a tactic deployed by adults to check that children understand and by children to ratify understanding (Clark and Bernicot, 2008). In second language learning, it facilitates access to target language forms, enables noticing of specific features and benefits automaticity (Duff, 2014). In turn, in storytelling repetition of a 'first saying' fulfils a resumptive function (Wong, 2010), while in administrative language it may aim to avoid or deflect blame, signal democratic openness and perform swiftness (Hansson, 2015).
As a lexical phenomenon, repetition is common in poetry or rhetorics, and is exemplified by epizeuxis, reduplication or the so-called distant repetition, among others. In relevance-theoretic pragmatics, it was initially approached as a resource resulting in a wide range of cognitive effectsdweak implicaturesdwhich are achieved as a way of offsetting an increase of cognitive effort Wilson, 1986, 1995: 219). In the well-known example (32) below, relating what is said to contextual information yields contents alluding to feelings of regret, sadness or longing, or to the fact that those days were innocent, carefree, happy, etc.: (32) My childhood days are gone, gone.
More recently, repetition has been approached as a deliberate ostensive phenomenon, which is sometimes intended for emphasis or intensification, while in other cases it seeks to attain a wide array not only of propositional, but also of nonpropositional effects (Jackson, 2016). These are cognitive effects which involve the activation of perceptual, emotional or sensorimotor mechanisms, give rise to images, feelings and psychological states, and cannot therefore be precisely paraphrased (Wilson and Carston, 2019: 32). Furthermore, repetition is also accounted for on the grounds of the showing-saying continuum (Wharton, 2009) and treated as a case of indeterminate showing (Sperber and Wilson, 2015), by means of which the speaker provides the hearer with relatively direct evidence of her intention to achieve a variety of such effects (Jackson, 2016).
This view should not rule out the possibility that repetition also results in ad hoc concepts: it could indicate the peculiarity of a notion or concept that the speaker intends to communicate. It might enable the representation of certain beliefs, or even give rise to a series of non-propositional effects, which may be subsumed by a concept, thus delimiting its denotatum or contributing to its specificity. Hence, repetition would somehow expand or enrich the encyclopaedic entry of the concept in question, or enlarge a conceptual file. Consider the following examples: (33) a. Marta vino despacio, despacio.
'Make me a coffee coffee'.
The repeated adverb 'despacio'/'slowly' in (33a) could cause its associated concept to denote a pace that is actually slower than average or usual, or shockingly or irritatingly slow. It could even encourage the hearer to create some image of the alluded person walking at an unusually slow pace. In turn, the repeated noun in (33b) suggests that what the speaker wants is a real coffee, one that is prototypically good, tasty, fragrant and rich in caffeine (Bazzanella, 2011: 250). In the standard account, the speaker invites the deduction of implicit contents. In this approach, in contrast, repetition makes manifest assumptions and may originate a cascade of non-propositional effects that become an integral and defining part of a concept, even if momentarily. In other words, repetition would flesh out a concept.
Repetition is "[…] characterized by an unstable balance between variance and invariance, sameness and difference, old and new" (Bazzanella, 2011: 248). A special type of repetition is rewording, which involves the replacement of a previously used word with a semantically related, more precise term. Therefore, rewording is a variant repetition that implies reformulation or paraphrase, and satisfies simplifying or clarifying functions (Bazzanella, 2011: 246).
In relevance-theoretic pragmatics, rewording is analysed as an assistive mechanism facilitating the hearer's better understanding of the speaker's meaning through in-depth exploration of the encyclopaedic entry of concepts (Blakemore, 1993: 110). Arguably, each of the consecutively-used terms is paired with a concept, so access to their respective encyclopaedic entries enables the hearer to grasp certain meaning shades or peculiarities. Accordingly, the activation of HOUSE and PALACE in (34) leads the hearer to infer that the speaker means a quite spacious, magnificent, perhaps luxurious or even sumptuous dwelling: (34) Paul has bought a new house, a palace.
The second word, however, might be deemed to canceldor erase, metaphorically speakingdthe concept activated by the first one and to create a new, richer notional entity on the grounds of the material in it. Alternatively, the second word might endow the concept activated by the first one with a constellation of beliefs; these would sit in its logical and encyclopaedic entries, thus enriching and modulating it, or would simply be incorporated into the corresponding mental folder. Be it as it may, the two closely-related terms would not result in two conceptual entities, but co-construct only a single, occasionspecific one.

Conclusion
Although lexical adjustment has been portrayed as an automatic process exempt from linguistic mandate, this paper has argued that this portrait needs to be retouched. It has illustrated that the formation of particularised concepts may be indicated and assisted by a number of linguistic elements. These include the evaluative morphemes available in some languages and lexical and phrasal elements. Among the latter feature determiners, qualifying adjectives and words and phrases fulfilling adjectival functions, expressive expletives, evidential participles, repeated lexical items and substitutive terms.
Such elements work as signals or pointers that a specific notion is needed in order to apprehend the speaker's meaning. Although they have been presented individually, they may obviously co-occur, thus providing additional evidence for, and pointing in the same direction of, the intended notion. Yet, the proposed list of elements is far from exhaustive. Future research should elucidate whether certain language types or (sub-)families may contribute additional elements. Furthermore, it should be ascertained whether paralanguage plays a similar signalling and assistive role.
Describing concepts as mental constructs that are compartmentalised into a set of information entries, or as some sort of file storing varied beliefs, and depicting ad hoc-concept creation as rearrangements or refinements of their content has some theoretical implications. Firstly, occasion-specific concepts must be understood as complex and dynamic cognitive entities subsuming more or less manifest, or fluctuating, information. Secondly, some of that information, and in particular that pertaining to attitudes and psychological states, may become manifest, and be stored in them, as a consequence of the characteristics of linguistic input. Therefore, it does not always need to originate as implicatures. Thirdly, lexical pragmatic processes may perhaps play a more powerful role in the comprehension of the speaker's meaning than what has thus far been envisaged.
Finally, the picture that has been sketched in this paper of ad hoc concepts as mental entities amenable to integrating or storing varied types of information also raises a question: what sort of information is likely to find a place in the encyclopaedic entry of a concept or in the mental folder that a word activates? Or, to put it differently, how might the mind decide on the beliefs that may be stored in, and associated with, an ad hoc concept? A first criterion identifying concept-integral information could be its defining or specifying function: certain items of information may feature in an occasion-specific conceptual entity if they contribute to specifying, delineating or fleshing out the raw cognitive material that a particular word activates. Another related criterion could be its aboutness, or the fact that it concerns what the speaker is actually thought to mean through a word: certain beliefs may feature in an occasion-specific conceptual entity if they contribute to identifying, determining or restricting what the speaker actually denotes or refers to by means of a particular term at a particular time of speaking. Still, a further criterion may be the occurrence of some linguistic element indicating that such information should be entertained and be accommodated by the concept. For instance, the addition of morphemes like the diminutive or the augmentative to specific words, or the appearance of determiners, possessives or expressive expletives could entitle hearers to entertain beliefs about speakers' psychological states and include them within a conceptual element. Similarly, the insertion of an evidential participle could encourage hearers to make assumptions about speakers' epistemic stance and send them to a conceptual folder. This would mean that the assumptions that arise from the joint interaction between context and what the speaker is thought to communicate by means of an entire utterance or a portion of it would not be part of a concept, but implicatures.

Declaration of competing interest
The author has no conflicts of interest to declare.