‘Don’t you ma’am me!’: A Construction-based Analysis of the Schema ‘don’t you V me’ Expressing Disapproval in English

The cognitive construction grammar (CCxG) approach can be used to examine the correlation between a constructional schema (CS) and an illocutionary force, in spoken discourse. This study aims to explore the construction [D(Y)[X]viM]j as in “Don’t you ma’am me” in the expression of disapproval or reprimand through the CxG-based examination of data obtained from three corpora: The Movie Corpus (TMC), The TV Corpus (TTVC) and the Corpus of American Soap Operas (CASO). Five constructional schemas (CS-0 to CS-4) have been identified, and they pertain to a network of constructions in which low-level CSs are more unambiguous and productive than highlevel ones. Although constant elements of such constructions contribute to a more solid correlation of form and meaning, the variable (verb) undergoes a process of functional shift to guarantee the formulaic constituency of these constructions and the expression of disapproval in a given communicative situation. A distinctive feature of the converted verb is its connection (or anchoring) to the preceding move, which can be either semantic or morphological (or echoic), the latter being, on some occasions, detached from the original meaning of the verb.


Introduction
The notion of Speech Acts (Austin 1962;Searle 1970) and the macrofunctions of language (Halliday 1970) have allowed for a better understanding of how speakers or writers construe meaning upon the lexical choices available within specific communicative situations. Central to the discipline of discourse analysis is the correlation between the concepts of idiomaticity and illocutionary force. Highly idiomatic expressions show that their constituents are not as semantically relevant as the unit they generate. These expressions are not decomposable, and their meaning is constructed upon discourse functions "since their surface meanings can be readily decoded" (Moon 1997: 47). Examples (1) and (2) illustrate how the semantics of ma'am and ethic indirectly contributes to the output meaning of an idiomatic expression in a specific communicative act.
(1) A: They stopped paying you, didn't they? B: Yes, ma'am. A: Don't you ma'am me. What do you need? (Criminal Minds 2009, TTVC) (2) A: I didn't know about any of the other stuff. Anyway, he's hired me as his lawyer now, so I'm bound to client confidentiality… B: Don't you fuckin' ethic me. Don't you ethic me. (Rake 2016, TTVC) These two examples are particularly dialogic, and their analysis is based on the typology of three elements: a verbal constituent (e.g. ma 'am, ethic), the preceding context, and the illocutionary force. Both stretches of talk show a feeling of disapproval, or reprimand, towards the other interlocutor's attitude by resorting to the same frame: 'don't you V me'. The intriguing nature of this construction is not new: Ruiz de Mendoza and Gómez González (2014) used the example of "Don't you "honey" me?" to describe the temporary breach between addressee and addresser in the use of the vocative honey, and they claim that this construction can be more complex because it seems highly connected to a previous context. This construction is not entirely connected to disapproval of vocatives; it also appears to show speakers' general feeling of disapproval, which is seemingly detached from the lexical meaning of the verb used. Thus, a preliminary observation of such forms shows that these verbs (ma'am, honey, or ethic) are constituents of constructions conveying a sense of disapproval in spoken English.
These examples also confirm that the study of these constructions should be addressed from a discourse construction grammar approach (Antonopoulou and Nikiforidou 2011; Östman and Trousdale 2013; Yuan 2019), in which specific morphosyntactic patterns are linked to a communicative function. Discourse constructions such as "Don't you stare at me" and "Don't you ma'am me" share some syntactic commonalities, and speakers use them to disapprove of a hearer's saying or behavior. However, a closer look into the compositionality of these constructions shows that the verbs (ma'am and stare) are semantically and discursively different. Whilst ma'am calls for some preceding context as it might refer to vocatives or attitudes, stare is more semantically transparent, and needs no context to understand its communicative goal. This can lead to a series of interesting questions concerning the degree of 'compatibility' of a certain syntactic construction with a communicative function: (i) are some constructions more unambiguous (or univocal) than others in the expression of a certain illocutionary force? (ii) How determinant can some noncomponential factors such as preceding context or construction frequency be? To answer these questions, two constructional schemas (CSs) have been identified at this point, in which a specific form of a discursive construction is associated with a specific meaning (Booij and Audring 2015): CS-1 and CS-2 possess similar semantic value ([disapproval of SEMi]j) but their morphosyntactic information differs in the type of verb (Xi) that is used to conform to the general construction. In these two CSs, the constituents D (don't), Y (you), and M (me) make up the steady configuration of the constructional schemas. The variable constituent (Xi) is always a verb or a verb phrase, which can be a denominal (converted) type as in ethic (Xi). Some of these word types are not generally found in English, which explains why the utterances that comply with the construction mentioned above are considered overtly nonce expressions.
This study focuses on spoken discourse and it seeks to explore the morphosemantic and syntactic features of a construction network that includes CS-1 and CS-2. Thus, there are two primary aims to this study: (i) to investigate the features of lexical inheritance, ambiguity, productivity, and replicability in this construction network following Halliday and Matthiessen's (2004) approach of systemic functional linguistics and Booij's (2010) systematization of construction morphology; and (ii) to examine the morphosemantic traits of the denominal verb [Xcvi], and see how its functionality within these constructions complies with the prerequisites of syntagmatic simplicity and paradigmatic flexibility (Adolphs 2008: 29). The variable (verb) constitutes an interesting element in these constructions as it seems to reflect the intrinsic value of syntactic constituency and communicative intention. This case-study project might provide a valuable opportunity to advance the understanding the interface of discourse construction grammar and the learned pairing of form and meaning.

Limiting the Scope of Study 2.1 On conversation analysis from a construction-based perspective
Austin's speech act theory (1962) and Searle's interpretation of the theory (1970) have represented a breakthrough within the domain of discourse analysis as it stands today. One of their most outstanding contributions is precisely the strict association between the illocutionary force of a speech act and the syntactic and semantic characteristics of an utterance. However, this has been somewhat controversial as it has also been argued that other factors, such as context or co-occurrences, can be more relevant than a syntactic frame. For example, Levinson claims that how people use the sentences is not restricted by the surface form (1983: 265). In addition, Schlegloff argues that the sequential unpredictability of a conversation stretch is disregarded by the traditional properties of the speech act theory (1988: 61). An interpretative methodology is also based on the assumption that speakers' reactions are dependent on previous utterances, and not necessarily on the structure of utterances, i.e. it is founded on "receivers' recognition of the producers' intention" (Stoll 1995(Stoll -1996. If the interpretative methodology is used in (1), don't you ma'am me is understood as a reproach of a prior 'trigger' or initiative, and both (reprimand and initiative) are structurally linked through the word ma'am.
The idea of relying heavily on pre-established formulas to explain the routines of language and the variability of constituents is connected with the notions of patterns of lexis and lexical priming by Hoey (1983; but on a more structural plane. Lexical priming is based on the predisposition, or priming, of words to occur in, or avoid, specific grammatical functions or positions within discourse (Hoey 2005: 13). Cognitively speaking, the property of priming also explains how speakers associate a sequence of words with a particular semantic context, allowing them to recognize "the similarities between what they want to say at a particular point and what they have heard or read at other times" to advance in (re)producing the priming (Hoey 2005: 20).
Following this approach, in CS-2 [D(Y)[X]cviM]j, three types of elements are identified: constant [D, (Y), M], variable [Xcvi], and irrelatives (i.e. units that are not necessary or relevant in the constant frame, e.g. like that in don't you honey me like that). This construction-based categorization corroborates the premise that some elements in the utterance might be more prone to substitution than others (Keller 1979: 239).
As commented earlier, the illocutionary force that is discussed here is that of disapproval. The study of said function through a spoken corpus is built on the idea that recurrent patterns or routines are a reflection of these linguistic forces or the presupposed notion that sequential or contextual factors are as relevant (cf. Aijmer 1996). Identifying such patterns could also lead to grammatical and vocabularybased criteria that characterize discourse frames or constructions of all sorts.
However, the analysis of illocutionary forces and specific constructions should not be merely structural. The sequential order of the construction within a dialogic text is critical to also comprehend its textual macro-function. For example, the binary sequential labeling of 'initiative' and 'reactive' could be applied to a number of situations. These moves are dependent on the criteria of "making a claim and fulfilling this very claim" (Weigand 2017: 182), which is reflected in either accepting, rejecting, or negotiating the prior decision or claim. The reactive component of 'rejecting', which seems closer to CS-2 [D(Y)[X]cviM]j, is pragmatically motivated and it is linked to a previous sequence: in (1) the preceding utterance yes, ma'am (a reactive to the initiative They stopped paying you, didn't they?) is also the initiative to the reactive phrase don't you ma'am me. A corpus-based examination of data can help determine if a reactive criterion is also involved, alongside with a fixed syntactic pattern, in the interplay between an illocutionary force and a given dialogic context.

On Construction Grammar in spoken discourse
The analysis of linguistic constructions through the theories of construction grammar (CxG) and conversation analysis (CA) encompasses a great deal of complementarity between the fields of cognition and discourse (or interaction). If constructions in CxG and turn-types in CA are examined independently, the relevance of language usage or social practices to both approaches implies that "full understanding of linguistic expression, as well as grammatical organization, cannot be reduced to 'form' only, 'meaning' only, or 'function' only" Östman 2005: 1755). Although CA differs from CxG in how meaning is negotiated through a number of ethnographic and linguistic aspects, the fact is that both attest the role of context-based paradigmaticity, that is, the emergence of recurrent patterns that can be described morpho-syntactically and are related to a specific discursive function.
The concept of construction that is used in this study is based on that by Jackendoff (2008) and Booij (2010), whereby a construction consists of pieces of syntactic structure with associated meanings. This association is necessarily dependent on a cognitive model to explain how lexical or syntactic structures can become instructions that modify a given discursive form (Langacker 2001: 143). On a more grammatical and lexical level, the notion of construction is characterized by (i) a morphosemantic arrangement of elements; (ii) a degree of productivity; and (iii) an unambiguous (or univocal) pairing of meaning and form (Ruiz de Mendoza 2013: 236). However, although the term constructional schema can be used to qualify any piece of syntactic structure, their varying length, semantic compositionality, and idiomaticity (Goldberg 2006: 5) indicate that not all these meaning-form pairs convey the same level of semantic uniformity or ambiguity. Therefore, CA represents a promising approach to check how negotiation of meaning through interaction reflects the correlation between a 'meaningful construction' (Jackendoff 2008) and an illocutionary force. This correlation is related to the lexical constructional model (LCM) approach, in which a construction is defined "as a form-meaning (or function) pairing where form affords access to meaning and meaning is realized by form to the extent that such processes have become entrenched in the speaker's mind (…)" (Ruiz de Mendoza 2013: 237). This opens the question of "to what extent" some constructions are more entrenched than others, and whether some constructions can be more predictable or transparent than others.
Using CxG in CA can help understand argument structural meaning, conventional implicatures, illocutionary forces, or pragmatic presuppositions (Fischer 2015). In this particular case, the examination of constituents in CS-2 can shed light on their specific morphosemantic properties, and on how these values depend on "what linguistic items have preceded it, what are expected to follow, and what do follow" (Sinclair and Coulthard 1975: 34).
The introduction of cognitive construction grammar (CCxG) in this study, especially after Goldberg's book Constructions (1995), contributes to understanding how constructions lead to new proposals of constructions. The concept of 'a network of constructions' is not new (cf. Booij 2010: 25-50), and it is based on the fundamental notion that explains how new constructions follow the same general architecture of unified ones; that is a combination of a given form with a specific meaning (Boas 2013). The case study in this analysis can help understand the correlation between morphosemantic composition of constructions and productivity in a high-level construction (HLC) that is inherited by a low-level construction (LLC).

Methodology
The use of corpus-based data has a relatively long tradition within discourse analysis (cf. Adolphs 2008). However, there are certain drawbacks associated with the use of corpora in the annotation of construction-based utterances: the spoken nature of the phenomenon studied here, and the compilation of all the units (both standard verbs and converted ones) complying with CS-1 and CS-2. As regards the former, to guarantee that all the data compiled was dialogic, the following corpora were used: The Movie Corpus (TMC), The TV Corpus (TTVC), and the Corpus of American Soap Operas (CASO). The latter problem was tackled through the implementation of a search query string (SQS) to ensure that the resulting hits fit into the construction pattern [D(Y)[X]cviM]j: 'do n't you [N*]'. This SQS guarantees that converted denominal verbs are collected, as opposed to non-converted ones, for their being nonce units is interpreted by the search engine as 'nouns', not verbs. This part of the process also involves manual annotation as unwished constructions such as those starting by why (as in why don't you text me?) were left out: they did not comply with either CS. However, to obtain as much data as possible, the SQS 'do n't you [V*]' is also introduced in case the corpora recognized some converted verbs as verbs, not nouns. The data could also help address the domains of syntactic framing, idiomaticity, and lexical nonceness. However, the three corpora used in the study represent scripted speech, and not necessarily spontaneous, naturally-occurring one. These texts are worth studying as they majorly reflect the constructions of spoken speech, and they are useful resources to look at nonstandard, or very informal, language.
The second part of the study encompasses the examination of context-based utterances in dialogic form to explore the conversational value of these constructions according to the binary set of moves: initiatives and reactives. Two annotators manually disambiguated each set; and inter-annotator agreement (or IAA) shows high figures of recall: 96% for Part of Speech (POS) and 91.5% for word senses. Also, data is used to examine if there is a functional correlation between the preceding/subsequent co-text and the CS-2 variable [Xcvi] in all the hits. This analytical stage follows Schleppegrell's premise that "discourse analysis seeks patterns in linguistic data" (2012: 21), which reflects the tight linkage between syntactic framing and illocutionary values. With the aid of corpus-based data, constructional schemas are elaborated and used in a taxonomic network in order to explore their syntactic constituency. This evidence is expected to shed light on the features of inheritance, ambiguity, and productivity, as expressed by resembling constructions.
A relevant premise in the data-gathering stage is the discursive nature of the constructions under scrutiny and that of the variables that are compiled from the corpora. Based on this, CS-2 [D(Y)[X]cviM]j is the product of syntactic framing and word nonceness through the process of conversion (or functional shift). The converted denominal verbs within the frame are generally nonce because "they are being used for the first time to solve an immediate problem of communication" and their usage is perhaps more important than its input meaning as they are proposed with "future standardized status in mind" (Crystal 2000: 223). Denominal conversion is a highly productive process in contemporary English (Bauer et al. 2015: 277), and the quantification of its productivity and token frequency is a difficult task, for corpora might not identify converted units as clearly as affixed ones.
Although CS-2 [D(Y)[X]cviM]j is considered a primary analytic unit in this analysis, the phrase might be subject to various interpretations. The construction is a well-structured discursive unit that is associated with the act of disapproval as in standard (or more general) CS-1 [D(Y) [X]viM]j such as don't you mock me, don't you yell at me, etc. Alternatively, CS-2 [D(Y)[X]cviM]j could fall into the structural category of 'phrasal lexical items' (Kuiper 2007), in which the issue of idiomaticity is examined through the properties of syntactic flexibility and slot restriction. Kuiper finds that interlocutors should be aware of what perceptual cues of the idiomatic expression are salient to recognize their 'deformation' more naturally (Kuiper 2007: 95).

Findings and Discussion 4.1 Morphosemantic peculiarities of the constructional schemas under analysis
As suggested earlier, CS-1 and CS-2 possess identical internal composition and semantic correspondence. Their complexity does not lie in the general syntactic constituency of the frame, but in the process of 'anchoring' to internal or external constituents. Whilst in CS-1 and CS-2 the verbs stare and ethic are anchored in a previous action (stare) or verbal move (ethic), the converted verb ma'am in (3) is linked to a specific lexical unit. In (3), the variable ma'am is recognized in both the initiative move (Ma'am, this is an official crime scene. I can't let…) and the reactive move (Don't you ma'am me!). As expected, it is a denominal verb, whose grammatical category has been shifted through a process of conversion. However, the relevant finding in this example is that the lexical unit ma'am is semantically opaque as the meaning activated here is not related to the semantic content of the lexeme ma'am. It is a mere replication or echoing of a previous lexical unit that fits into the CS in the expression of disapproval. Following Hoey's structural predisposition of lexis (1983; 2005), ma'am, as in 'don't you ma'am me!', is not a meaningful unit; its meaning is only conveyed through the construction to which it belongs, similarly to the constituency of set phrases or idioms.
A distinctive feature of the variable [Xvi] is its connection to a previous element (word, discursive frame or attitude) to complete the semantic compositionality of the construction. Following this initial premise, it is presupposed that CS-1 and CS-2 inherit constructional qualities from a source (unified) construction, or CS-0, in which 'l' stands for either an internal or external element to the construction, and P stands for any personal pronoun. Note that CS-1 and CS-2 have been conveniently restricted to M (me) to make data more manageable in the compilation process.
However, the echoic anchoring of ma'am differs from a number of examples obtained from data. While ma'am is morphologically anchored in the initiative move, the verb ultimatum in (4) is not. This indicates that there are two general types of constructions [D(Y)[X]viM]j expressing disapproval that are distinguished according to the nature of the variable: those that are morphologically anchored (ma'am), and those that are semantically anchored. The latter, as in ultimatum, is not semantically opaque. In contrast, the variable ultimatum encapsulates a general attitude that is perceived by B in (4), and it is used to express disapproval of such an attitude. In fact, ultimatum is a perfectly meaningful unit as opposed to ma'am in (3). Both are used in a similar frame to express disapproval, and they are both morphologically primed to conform to the general structure of [D(Y)[X]viM]j. Therefore, variables (verbs), such as ma'am and ethic, can undergo a process of semantic shift, and they also possess distinctive values of lexical or semantic anchoring. The particularities of ma'am are that it is used in the initiative, and replicated in the reactive to show disapproval. Like CS-2, CS-3 also represents the converted nature of ma'am, but its reformulation specifies whether the converted verb has been echoed or replicated in the model ([[X]cvi]pre). In this vein, 'pre' stands for the preceding syntactic function of the verb in the initiative. In the case of ethic, although it shares with ma'am their denominal source, it is not morphologically anchored in the initiative.
Interestingly, several syntactic constructions with a converted verb were generated by the search queries. In general, 26 out of the 63 variables are morphologically anchored (see Appendix 1). This number shows that there is no clear-cut preference of one type of construction in English. One unanticipated finding was that the majority of the morphologically anchored variables were vocatives, which are literally implemented in the initiative utterances or moves to fulfil any of the three communicative functions identified by Leech (1999: 111-112): to summon the hearer's attention, to identify the hearer, and to maintain the social implication of both hearer and speaker in the conversation (i.e. endearment, politeness). The cases of endearment (honey), kin treatment (daddy) and personal names (Danny) are the most common ones, and they pertain to CS-3, as shown in (5). These vocatives comply with any of their functional categories in the initiatives, but as being morphologically anchored in the reactives, they move towards disapproval. Morphologically anchored vocatives might convey a sense of disapproval towards the use of the vocative itself.  However, on many occasions, there is no direct connection between the use of a vocative and its disapproval, as seen in (6) and (7). It could be seen as a means of construing meaning upon pre-established lexical priming, recognized by both hearer and speaker through a specific construction. In the examples (6) and (7), the communicative function of the vocatives John and Mom in the reactives are not precisely a reprimand due to the misuse of these words by one of the interlocutors; they constitute, in fact, a discursive tool that fits into a specific construction in the expression of disapproval towards a preceding action or comment. These unusual constructions in which one of the constituents (converted verb) is possibly used as an adjacent means of connecting both initiative and reactive through the expression of disapproval, proves that the syntactic force inherited from higher-level constructions can be more relevant than the semantics of constituents. CS-4 shows a constructional schema of these verbs (e.g. ultimatum), in which SEM does not refer to any of the constituents in the construction, but the initiative move in general. Therefore, the main aspect that differentiates CS-4 from the other constructions rests on how the verb in CS-4, being a denominal verb that is linked to a preceding utterance ([[X]cvi]pre), does not apply to the same verb used in the preceding utterance (SEMi), but to the general initiative move (SEMk). The use of two different semantic marks (i.e.'i' and 'k') allows for a finer-grained perspective of how a preceding utterance can generate lexical echoing (SEMi) or just a nonce word that is used to abstract a preceding move or attitude (SEMk).

CS-4 < [D(Y)[[X]cvi]preM]j ⟷ [disapproval of SEMk]j >
The semantically anchored cases are more difficult to detect in corpora as they are not linked to previous lexis. The nexus of these examples lies in a communicative situation, which explains the variability of these forms. However, discursively, they are more transparent than the morphologically anchored ones as they are construed upon a previous context, and these constructions are meaningful in isolation. That is, the utterance don't you ultimatum me in (4), without explicit information to the previous context, is transparently related to the idea that someone reprimands their interlocutor for having set conditions, demands, or deadlines. Conversely, in (5), the phrase don't you John me is overtly opaque. Still, the idea of disapproval remains, but not its causes. As suggested, some semantically anchored units, e.g. (8) and (9), show that the variables guilt and lawyer clearly (or transparently) characterize the interlocutor's attitude or comment. The evaluative force, alongside with the sense of disapproval, is one of the most essential features that define semantically anchored forms, as opposed to morphologically anchored ones. The judgement in these two cases encompasses a new variable, particularly a denominal verb, that generates both disapproval and evaluation. This is different from some of the cases mentioned earlier, say (5) and (7), in which no evaluation is rendered. The morphologically anchored case in (6) is probably midway as don't you honey me might also express that B does not approve of the endearment vocative honey, and finds that the vocative is inappropriate. Therefore, the latter communicative situation can also be interpreted as a 'guise' of evaluation, for B evaluates A's phrases, and expresses disapproval. (8) A: I realize that you were raised to be a self-absorbed, entitled prince of a man by your father, but you have a mother, too. Although semantically anchored frames are transparent when they are examined out of context, this is not always the case as there might be opacity that depends on the lexical unit itself, and not necessarily on the communicative situation. As seen in (10), the variable agent might not be as precise and transparent as (8) and (9), which is why an additional explanatory comment tags along. Converted denominal verbs can show high degrees of polysemy, particularly in cases in which common nouns are used. That is the case of honey in (6), in which don't you honey me can be understood as 'stop putting honey all over me' or in a more metaphorical reading, 'stop using that sweet talk with me'. In examples (11) and (12), the structures clearly comply with the discourse frame studied here, and accordingly, they convey the idea of disapproval. In this case in particular lip and railroad are used in English as 'to back talk' and 'to accuse someone without enough evidence' (MWD11) respectively, so they are not as discursively innovative as (8)  As shown in (6) and (7), a particular case of morphologically anchored construction is that in which variables are implemented randomly. This corroborates the idea that the output semantics of the variable in CS-4 [D(Y)[[X]cvi]preM]j is less dependent on random morphological anchoring, and far more connected to the meaning conveyed by a previous utterance. The examples (13) and (14) show that the anchoring process syntactically depends on the last content lexical unit, e.g. language, providence. The fact is that if a slight substitution were carried out in (14), say don't you indeed me, the illocutionary force would not be necessarily softened. The morphologically anchored examples (13) and (14), in which no semantic paradigm is followed, just a replication or echoing of a preceding word, from the initiative into the reactive, are a reminder of the adjacency pair approach. Both the variable and its precedent are interpreted by interlocutors as discursively relevant. Their discursive meanings are also dependent on their replication, and the pairing is seen as a significant strategy in the construction that is discussed here.

On inheritance and ambiguity
Through the property of inheritance, high-level constructions can contain some properties of lower-level constructions, including those features that pertain to LLCs and differentiate them from other LLCs. This hierarchical rearrangement of construction networks through inheritance is primarily characterized by constructions that share a set of features, and thus "form clusters of mutually related generalizations about linguistic competence, going from more abstract and unconstrained to more restricted" (Fried and Östman 2004: 72). The introduction of this principle allows for a general understanding of how certain constructions are embedded in more general ones, particularly in the discursive frames that might depend on external units (lexis or moves) to have constituent slots filled. Table 1 shows a sketchy summary of two differentiating values ([X] and [SEM]) in the four constructions introduced in this study. These four CSs inherit the learned pairing of form and meaning from CS-0, and they are recognized as constructions as "some aspects of its form and function is not strictly predictable from its component parts or from other constructions recognized to exist" (Goldberg 2006: 5).  Table 1 succinctly informs that these constructions (CS-1 to CS-4) are recognized as distinct, for some varying (or differentiating) elements in the schema (i.e. [X] and [SEM]) can be either present (Y) or absent (N). The four schemas are abstracted through five global aspects of their constructions: (i) the verbal nature of the schema (Xvi), (ii) the converted property of the verb ([X]cvi), (iii) the presence of the denominal verb in the preceding clause ([[X]cvi]pre), and finally, the semantic correlation that is established either between (iv) the converted verb and an explicit lexical unit (or an explicit action, e.g. stare) in the preceding clause (SEMi), or (v) the connection between the converted verb and a semantically-detached initiative (SEMk). The last aspect, in other words, implies that the cases of don't you ultimatum me or don't you ethic me do not refer to specific words (ma'am) or actions (stare) that are explicitly expressed in the initiative. These cases convey the sense of disapproval of a preceding move, and the variables used in this type of construction (e.g. ultimatum, ethic) are meant to evaluate the initiative without any sort of lexical/referential linkage. Table 1 also reveals which aspects in the four schemas are relevant to further elaborate on the hierarchical arrangement of constructions and on the property of inheritance.
[X] and [SEM] are itemized to explore which properties are inherited. For instance, data in the table suggests that [X]cvi and SEMi constitute common features in most of the schemas, which means that the properties of denominal conversion and disapproval of a preceding nouns are bound to inheritance and higher level of abstraction.
The absent or present features in Table 1 do not negate the fact that such elements are embedded in an HLC, and therefore an element is not really absent but rather is unified in a more general construction. Inheritance is particularly visible in the constructions that are distiguished through one feature only. In Figure 1, the variable [Xvi] in CS-0 represents the tertium comparationis in the analysis as it leads to a hierarchical detachment of constructions. At the next level of abstraction, CS-1 and CS-2 are differentiated through the morphology of the verb: as opposed to CS-1, the verb in CS-2 is denominal. This distinction is paramount in the network because it reflects the backbone of morphological variability in the constructions under study. Alternatively, CS-3 and CS-4 represent the lowest-level constructions, in which denominal verbs are morphologically (CS-3) or semantically (CS-4) anchored in the preceding (initiative) move. Although both denominal verbs are used in a preceding move, they are differentiating at this point not through the morphological constituency of the verb, but through the semantics of the schema. Whilst CS-3 is referring to the idea of disapproving of a preceding noun [SEMi], particularly a vocative (e.g. ma'am), CS-4 does not relate to the lexical semantics of the anchor, but to that of the preceding discursive construction or of the speaker's attitude [SEMk]. This shows that a network of constructions cannot be necessarily devised through one differentiating feature only, for LLCs can involve changes that affect other constituents of the schema. This diverging point in the network is what we call here a node, and it is used to isolate the points at which constituency construction is altered. Figure  1 shows two nodes in the network: node A for the denominal nature of the verb and node B for the semantic correspondence between the anchor and a preceding unit. The hierarchical network also contributes to understanding the correlation between the degree of ambiguity and the level of construction. The degree of ambiguity, understood in this study as the potential of a construction to generate more than one sense, can be assessed through a corpus-based examination of hits. HLCs, due to their level of abstraction and generalization, are expected to be more ambiguous (or polysemic) than LLCs. However, corpus-based data also suggests that even constructions pertaining to the same level (e.g. CS-1 and CS-2) can show disimilar values of ambiguity. All of 354 tokens or variables complying with CS-1 are generated through the SQS 'don't you [V*] me'. A preliminary analysis of the top fifty variables or tokens on TTVC (Appendix 2) shows that 70.1% of the hits belong to a completely different construction expressing 'request' or 'advice' ('Why don't you [V*] me'). The rest of the hits comply with the form of CS-1 but they are semantically (and prosodically) different: 18.6 % ('request') as in Don't you like me? and 9.3% ('disapproval'). These results indicate, on average, that the SQS '[why] don't you V* me [irrelatives]' generates mostly directive moves in spoken speech, which is perfectly compliant with traditional studies of grammar on the use of imperatives and requests (Rupp 2003: 18-20), in which the discursive function of this formula has been pointed out. Alternatively, the majority of the hits (91.2%) in Appendix 1 are reactives expressing 'disapproval' (CS-2), and the rest correspond to denominal verbs that have been standardized in English (e.g. text, blackmail, email, beep). Therefore, nearly all the examples extracted from the corpora through the SQS 'don't you [N*] me' only generate CS-2. This finding confirms that CS-1 and CS-2 possess different degrees of ambiguity, with the former being less univocal than the latter.

On replicability and productivity
The question of frequency has also been associated with the capacity of some constructions to entrench more easily than other resembling constructions. Therefore, rather than frequency, Ruiz de Mendoza refers to the notion of 'replicability' (R) to explain how replicable (or frequent) constructions are characterized by meaningfulness and minimum alteration (2013: 237). Stated differently, in similar contexts these minimally varying constructions can maintain the function or illocutionary force, and its high frequency (or replicability) can be an indication of well-instantiated cognitive schemas.
R is easily measured through the number of construction types generated by the corpora. Construction types are semantically equivalent frames that are distinguised through the variable [Xvi] as in don't you stare at me and don't you hit me. These types coincide in form and meaning. Therefore, in the case of CS-1, R is 110 (R1) whereas R is 59 in CS-2 (R2). This difference is reasonable as CS-1 inherits (from CS-0) some grammatically-compliant features that are natural to English whereas CS-2 introduces a functional shift process that generates nonceness. Although conversion is recognized as a highly productive word-formation mechanism in English (Lieber 2005: 418), converted verbs such as vocatives or personal names are rarely seen in verbal constructions. This seems to be related to the tenet of embedded productivity (Booij 2010), which denotes the phenomenon "that a wordformation process is normally unproductive, but is productive when it co-occurs with another word-formation process" (47). Although this definition is introduced in morphological construction, it can be used in syntactic CSs as well. Data also confirms that there exists a correlation between the number of schemas unified in a single construction and the degree of productivity.
Like R, type productivity can also be imported from morphological studies to measure the corpus-based degree of nonceness (hapaxes) of a certain constuction in corpora (Bauer 2001). The measurement of productivity has been traditionally questioned by the inconsistencies of the formulas and procedures (Fernández-Domínguez 2015). Type productivity is used in the present study as it involves the phenomena of nonceness and word-type variability. It is based on the concept of potential productivity (Baayen 2009), partcularly on Baayen's earlier approach to productivity measurement (Baayen 1993), in which P (productivity index) can be measured by dividing n (number of hapax legomena) by N (number of occurences of a CS in the corpora). In the case of n, this study is especially interested in the types that only occur once in the corpus.  Table 2 shows that the productivity index of constructions (Pcs) is not proportional to the index of replicability (Rcs). Pcs, however, reflects the high levels of nonceness that characterize LLCs: P3 and P4 indicate that their corresponding constructional schemas (CS-3 and CS-4) are more 'open' frames than CS-1 and CS-2, and this seems to be in connection with their low degree of ambiguity. Highly productive constructions demonstrate that once the learned pairing of form and meaning has been discursively acquired (or 'fixed') by users, variables, such as [Xvi], can be changed without altering the whole semantics of the construction. These changes also confirm that productivity is also linked to lexical creativity and construction potentiality.

Conclusion
Although the study focuses on four constructions expressing diapproval, the findings may well have a bearing on the general implications of construction constituents, notably variables. A variable, either semantically anchored or morphologically anchored, is expected to be the headword or nexus between the reactive move and its prior initiative one. This binary approach in conversation analysis has confirmed that even though the variable is generally expected to fall into the 'converted denominal' category, it exists on a three-layer network: (i) the variable is a new denominal verb that evaluates the interlocutor's comments or ideas (i.e. semantically anchored as in ultimatum); (ii) the variable is also found in the initiative and the speaker uses it in the CS to show disapproval, particularly as a vocative (i.e. morphologically anchored as in ma'am); (iii) the variable is also found in the initiative, and its use in the construction does not respond to a specific criterion, it frequently originates from the last words that make up the initiative (i.e. morphologically anchored as in providence).
All of the CSs under study (CS-1 to CS-4) conform to the formulaic constituency of a more general construction (CS-0). The changing nature (and semantic value) of the variable is in fact a syntactic means that guarantees the semantic and discursive adequacy of the constructions in the expression of the illocutionary force of disapproval. This shows that some features of HLCs are inherited by LLCs, e.g. constant units of the frame, and the variable [Xvi] is expected to undergo a functional shift process to fit into the inherited frame. Interestingly, the process of conversion restricts the degree of unambiguity of lower-level constructions, for the verbs that originate from the new constructions are generally nonce elements. In other words, LLCs are less general and more univocal because the morphological constituency of variables is highly dependent on communicative situations and preceding moves. Baayen's productivity index also confirms how the LLCs, due to their feature of nonceness, are highly productive whereas HLCs are more ambiguous and less productive. However, the level of replicability (or frequency) in the latter is much higher. The replication of a given CS contributes to both a more strict correlation of form and meaning, and a more significantly learned encoding of constant and variable constituents of the construction.
This study, aimed at examining the internal (morphosemantic) and communicative value of CSs expressing disapproval, has also found that whilst discourse constructions are syntactically decomposable, their variable units (denominal verbs) simply conform to the syntactic patterning to maintain the communicative function of the phrase. Although this study focuses on scripted spoken corpora, the findings show how the internal structure is '(de)formed' to be communicatively