2.1. Properties of incoordinate sentences

For Kuteva et al, an incoordinate sentence is composed of a sequence of a coordinative connector and a sentence, which would correspond to a second coordinate conjunct:

Thus, incoordinate differ from coordinate sentences, as the sentence that would correspond to the first conjunct is missing. Bearing this in mind, Kuteva et al (2017) attribute three properties to incoordinate constructions:

• a) they involve the coordinating connectors but and and;³

• b) these connectors appear at the beginning of the sentence (vs. typical position in coordinate sentences);

• c) they result in the independent use of formally coordinated sentences.

Kuteva et al propose that, in incoordinates, as we do not have a sentence corresponding to the first conjunct, it is possible to pragmatically infer it. The authors state that the discursive-pragmatic context provides the interlocutor with information that is equivalent to the first conjunct, and that is necessary to interpret the content of the incoordinate construction.⁴

On the other hand, Giorgi (2018: 4), in defining the properties of counterexpectational questions, as (9), lists the following characteristics, the first two being typical of root sentences in direct discourse:

• – There is a characteristic falling-raising intonation.

• – The sentence is accompanied by gestures of the hands and of the head, plus brow raising.

• – Presence of the imperfect form of the indicative mood.⁵

• – Presence of the particle ma.

• – Presence of negation.⁶

For these constructions, Giorgi (2018: 5) suggests an extension of her (2015) analysis for Pending Topic, considering that these sentences also correspond to bi-sentential structures, or discourses. In terms of syntactic structure, Giorgi (2018: 8,16) offers an analysis in which ma fills the head of the discourse projection (dis), the specifier is empty, and the content of the second conjunct occurs in complement position, as below:

It must be noted that Giorgi does not seem to adopt the idea that these constructions should be characterized as a type of coordination, since there is no reference in this representation to Conj or ConjP. On the contrary, the idea in this representation is that ma is a discourse marker that occupies Disc⁰ and projects Discourse. In the author’s proposal, Disc⁰ is a head that connects two textual units (which correspond to two CPs, or to a situation and a CP), i.e., Disc⁰, and ma that instantiates it, selects a CP as complement, and the specifier of Disc⁰ can be either a situation or a CP. This shows, according to the author, that some syntactic representations “must be enriched with information traditionally considered as belonging to other modules of grammar” (Giorgi, 2018: 5), namely to Discourse.

Regarding the reticence marks in specifier position, Giorgi (2018: 16) affirms that it corresponds to the expectation that is denied by the content of the second conjunct.

Kuteva et al do not present a syntactic structure for these sentences. However, the authors investigate the status of but and and in these constructions and state that these elements lose their status as coordinative conjunctions, namely because, unlike the conjunction but, which articulates two segments (cf. (12)), but, as a sentential particle, resulting from the incoordination process, expresses the speaker’s attitude and does not articulate segments, as in (13).

As for a sentential particle, the authors point out that “it may – but doesn’t have to – even relate to a linguistic chunk, i.e., it is not bound to a previous assertion by an interlocutor.” In their analysis, Kuteva et al (2017) consider that the possibility of a sentential particle not being bound to a previous assertion by an interlocutor is due to speakers having no difficulty in providing, given the previous pragmatic-discursive context, possible equivalents to the absent sentence, as below:

Apart from determining the structure of these sentences, it is important to ascertain its pragmatic value, which, in our analysis, is linked to the value of mas and its conjunctional status. Furthermore, by determining the status of mas, we can more accurately propose a syntactic configuration.

2.2. Pragmatic value of incoordinate sentences

Regarding the value of incoordinate sentences, Kuteva et al suggest that they are associated to the value of mirativity, which was defined by DeLancey (1997, 2001) as a way of conveying information that is new or unexpected to the speaker. Mirativity is often expressed through the use of evidential markers and, for this reason, the work of DeLancey (1997) was pioneering in distinguishing evidentiality from mirativity. Also, Kim & Aleksova (2003) present evidence from Bulgarian to show that these concepts are distinct, since, in this language, they are expressed by using different markers.

Also, according to the work of Aikhenvald (2012), mirativity can express various values in relation to the speaker, the interlocutor, or the main character of a narrative. The values reported by Aikhenvald (2012: 437) are: sudden discovery, surprise, unprepared mind, counterexpectation, or new information. However, taking into account that Aikhenvald (2012: 457) refers that not all values that make up the mirativity range can be used in all languages, Kuteva et al propose surprise as the central value of the concept of mirativity.

Niclot’s (2018) work assumes the existence of incoordination and also proposes the association of mirativity with the adversative connector, considering that ma and aber (adversative conjunctions in Italian and German, respectively) can be used to convey surprise.

The author mentions that this construction is also available in English, providing examples such as “but that’s great!” and “but you cook terribly!” and considers that “(…) the availability of both an adversative and a mirative reading of but in these two languages legitimizes the hypothesis that polysemy rather than homonymy is involved” (Niclot 2018: 1).

Kuteva et al (2017) argue that but and and, in incoordinate sentences, constitute a group of non-specialized expressions for conveying mirativity. By defining but and and as mirative particles, the authors distinguish their values from those of coordinating conjunctions, referring to examples (12) and (13), repeated below, for convenience:

Regarding the function and value of but in these examples, Kuteva et al (2017) propose that in (17), as but occurs as a conjunction, it has its source meaning of counterexpectation. Differently, sentence particle but, occurring in (18), conveys mirativity, which is the resulting meaning of the incoordination process.

On the adversative and mirative values of the conjunction, Niclot (2018: 1) assumes in his analysis that mirativity is a value that derives from adversativity: “More generally, on grounds of the attested adversative function of the cross-linguistic equivalents of but (henceforth buts when relevant), we can assume that this meaning is diachronically primary and that the sense of surprise has arisen from it along a process of grammaticalization.”

We understand this proposal, as, in EP, one of the main values of mas is also counterexpectation (cf. Canceiro, 2023), and there is also evidence that in some languages there are particles dedicated to conveying mirativity. However, in the case of incoordinate sentences with mas, we believe it is difficult to distinguish counterexpectation from mirativity, since both correspond to the value of surprise or unexpectedness. We, thus, propose counterexpectation as the main value conveyed by incoordinate sentences with mas.

As mentioned, Kuteva et al (2017) consider that, when going through the process of incoordination, but loses its conjunctional status and becomes a sentence particle. According to the authors, this change of status is due to the, already mentioned, change of meaning from counterexpectation to mirativity.

Although we agree with the association of incoordination to surprise or unexpectedness, we believe that surprise arises from an expectation that is denied, as below:

Thus, denying an expectation always results in a surprise, even though the opposite is not true. In example (20), the sentence conveys surprise, but it does not necessarily deny an expectation:

Also, we propose that in some incoordinate sentences that occur without previous linguistic material the content of the expectation (that is denied by the incoordinate) is generated through R-inferences (related to cultural/social questions, cf. Schwentter, 2000; Niclot, 2018), as below. In (21), the incoordinate must be uttered by a speaker whose cultural background creates the expectation of a bride wearing white. Thus, this example with mas conveys that, for this speaker, a bride wearing red goes against their expectations.

We believe our proposal is further corroborated as, in languages as Spanish or German, in which there are two adversative conjunctions, pero/sino in Spanish and aber/sondern in German, it is the one that conveys counterexpectation (pero and aber, respectively) that occurs in incoordinates.

Niclot (2018: 62) even mentions, regarding the example below, that it could be analyzed as resulting from the ellipsis of the first conjunct, whose content would be similar to “I didn’t expect this”, which further strengthens our idea.

Thus, although Kuteva et al (2017) and Niclot (2018) consider mirativity to be the main value of incoordinate but, we believe there are arguments that support an analysis in which it maintains its source meaning of counterexpectation. Kuteva et al also admit that DeLancey (c.p.) considers that when but occurs in an incoordinate sentence it still conveys counterexpectation.

In the next section, we will present our analysis for EP, in which we will argue for a conjunctional status for mas, while bearing in mind the maintenance of one of its source meanings, i.e., counterexpectation.

3.1. The status of mas and types of incoordinate sentences

The occurrence of mas in incoordinate sentences led us to question its conjunctional status and rethink the possibility of treating it as a Discourse Marker, since the function of a coordinating conjunction is typically to articulate two conjuncts, forming a complex constituent. We will start by defining the status of mas, since the properties of the head must be preserved during the derivation of the structure of the sentence.

Colaço (2013: 256), analyzing constructions in which the copulative conjunction seems to cross discursive boundaries, points out that, in certain cases, the particle e (‘and’) seems to present a behavior that is both conjunctional – as a connecting element – and discursive – since it is associated to progression of the narrative.

Similarly, Giorgi (2016, 2018) notes some properties that distance the connector ma from the conjunction ma, such as the fact that it is a discursive head, which connects textual units.

Bearing in mind that there are different types of incoordinates (cf. (4) vs. (5), repeated below as (28) and (29), for convenience), it would be possible to consider that, in some types, the behavior of mas is closer to that of a discourse marker.

However, we believe Coutinho’s (2008: 208) work should be considered as the author states that it is unfortunate to use the expression discursive markers, as this might reiterate a dissociated conception between language and discourse. The author also notes that it is problematic to sustain the existence of linguistic units especially designed for discourse, thus implying the existence of others that are not (Coutinho, 2008: 196).

Still, to evaluate the hypothesis that the conjunction is transformed, through incoordination, into a discourse marker (DM), it is relevant to take into account the work of Fielder (2008). Regarding the distinction between coordinating conjunction and discourse marker, Fielder (2008: 80), for Bulgarian, considers that these connectors “can be plotted along a continuum of connectivity that operates on multiple levels of linguistic structure: syntactic, semantic, and discourse-pragmatic.” Fielder (2008: 82) also questions whether the distinction between coordinating conjunction and discourse marker is valid or useful.

Fielder’s work is motivated by the fact that in Bulgarian there are three adversative conjunctions/discourse markers whose conjunctional status is not fully defined. The author analyzes various corpora and concludes that, for example, the medial position of the coordinating conjunction in the sentence is not a sufficiently robust argument to determine the status of the constituent.

Bearing in mind the possibility of mas being a discourse marker in incoordinates, let us look at the behavior of adversative connectors, like porém, contudo, in EP, classified as DMs by Macário Lopes (2016: 447). In the examples below, we can see, for example, the possibility of DMs co-occurring with coordinating conjunctions. In a construction with mas such co-occurrence would cause ungrammaticality (cf. (30c)):⁷

We can also verify the behavior of incoordinate mas against some of the prototypical properties of DMs (cf. Jucker & Ziv, 1998), which show that incoordinate mas is not a DM.

A DM is prototypically associated with a special intonation, occurs between commas, or followed by a comma, but in opposition to porém and contudo (cf. (30a)–(30b)), mas does not occur in these contexts:

As the insertion of the comma yields (31b) ungrammatical, mas does not seem to be associated with a special intonation nor should it be followed by a comma.

Also, a DM can prototypically move to several positions within a sentence, which is not possible with mas, as shown below:

This patterns with the typical behavior of conjunctions, and contrasts with prototypical adversative discourse markers, as porém/todavia/contudo, which as shown below, present great mobility in the sentence they occur:

If mas (‘but’) is a DM, its alternation with other elements that are unanimously contrastive DMs, such as porém (although) and contudo (however) should be possible:

This exchange is not possible given the pragmatic/semantic value of incoordinates (counterexpectation), i.e., these DMs, porém/todavia/contudo, cannot occur in an incoordinate because they do not convey the same value. Furthermore, we also consider that this alternation is not possible due to different status of these elements: mas is a conjunction, and porém/todavia/contudo are discourse markers. This different behavior further strengthens our argument for mas being a conjunction.

Bearing this in mind, as it will be mentioned below, the assumption that mas is a discourse marker would imply considering a new category of DMs.

If mas (‘but’) is a DM, its occurrence should be optional, and its omission should not affect the pragmatic value of the sentence:

The example (35b), although not ungrammatical, displays a different meaning in relation to (35a). In our analysis, only in (35a) it is possible to infer that a different behavior is expected from the addressee of the speaker. We propose the presence of mas is necessary, as the contrast above illustrates, to convey counterexpectation (one of the main values of this conjunction).

However, a property mentioned by Giorgi (2018) allows the approximation of mas to a discourse marker. This DM would have as distinctive property, regarding conjunctional mas, the impossibility of counterexpectational sentences occurring in subordination contexts, as (36) from Giorgi for Italian, and (37) provided by us for EP:

This means that these expressions typically occur in direct speech and that we could have direct speech reported as:

However, DMs as porém, todavia or contudo are not excluded from subordination domains, as shown in (39), an example from Mateus et al (2003: 571), in which we include mas to evidence its contrasting behavior:

So, even if it is assumed that mas in these constructions is a DM, it is necessary to specify to which category of DMs it belongs, since mas, although it is associated with the adversative value, seems to behave differently from other discourse markers with this value.

When it occurs in incoordinates, mas makes it possible to connect two overt discourse units/sentences (cf. Giorgi’s analysis of ma), or a sentence to a situation. Given this behavior, if we define these elements as DMs, they should belong to the class of coordinative conjunctive DMs. This class would include DMs that occur in incoordinate sentences, and we suggest that, for this reason, a separate class should also be considered to include DMs that occur in insubordinate constructions, as in the following examples:

In examples (40a) and (40b) que seems to act as a complementizer introducing new discourse yet conveying different values (a wish vs. an order). This would raise the question of how many DM categories we need to account for all these cases?

Also, given that there is no projection dedicated to DMs in current discourse approaches to languages, unless we accept Giorgi’s proposal (2016, 2018), what would be the consequences of assuming mas as a DM in terms of syntactic structure?

Assuming incoordinate mas is a conjunction, we can observe it shares three properties with the regular conjunction that occurs in adversative coordination:

• i) mas occupies the initial position of the conjunct;

• ii) mas cannot be moved within the conjunct;

• iii) mas cannot compete for a structural position occupied by another conjunction – the head of the structure.

Regarding the possibility of the conjunction co-occurring with complementizers in coordinating subordinate sentences, we acknowledge, with Giorgi (2016, 2018) that the examples provided mainly involve direct “speech” or dialogue, introducing full sentences, i.e. CPs, thus bringing mas closer to a discursive element. The CP nature of the constituent selected by incoordinate mas is overtly attested in some examples where mas cooccurs with the complementizer que (‘that’) in EP:

As for the sensitivity to the finite or non-finite nature of the sentences that mas introduces, we assume that this is a consequence of the previous property, since the infinitive typically occurs in subordination contexts.

Regarding the properties of conjunctions/discourse markers, we find interesting the question posed by Aarts (2007: 225) about the properties that define the status of constituents:

How can we be sure to identify all the relevant properties, and are all the properties equally important?

From what we have shown, we admit, with Fielder (2008), that the distinction between mas as a conjunction or a DM in incoordinate sentences is not necessary or useful. By discarding this distinction, we propose that, at the syntactic level, mas is a regular conjunction, and, at the discourse level, it apparently acts as a discourse marker.

Regarding the differences between incoordinates, even though Kuteva et al (2017) mention that the incoordinate can, but doesn’t have to, be linked to a previous assertion, we assume this possibility to be linguistically relevant. Let us look at the examples of incoordinate sentences below:

Although these incoordinates seem different, as they occur in different contexts, i.e., in response to an addressee (cf. (42) and (43)) or out of the blue, as (44), all convey the denial of an expectation the speaker had. To clarify this reasoning, let us analyze each of the examples: (i) in (42), the incoordinate is similar to the second conjunct of integrated coordination, i.e., the expectation that is denied is obtained from the content of the previous linguistic material and our knowledge of how the world works; (ii) in (43), João’s utterance does not occur only as a response to Ana’s ordering of a chocolate mousse, but also because he expects that only those who are not allergic to chocolate can enjoy chocolate mousse; (iii) in (44), Pedro’s utterance expresses that he didn’t expect his daughter to be so strong and being able to lift a big box. Hence, the linguistic context that precedes the incoordinate in (42) is the target of Paulo’s utterance, because Ann’s sentence creates the expectation that 8 months is not enough time to speak fluently; in (43), Ana’s sentence is not the target of João’s utterance, since the incoordinate occurs as a reaction to what is in the speakers’ common ground (shared knowledge), Ana’s allergy to chocolate, which, in turn, generates the expectation of her not being able to eat it. On the other hand, in (44), as there is no linguistic context, the expectation has to be inferred from Pedro’s utterance. In sum, regardless of their context of occurrence, incoordinates arise as reactions to unexpected events, which are inferred by the speaker and that are contrary to their expectations.

Bearing in mind this behavior, we propose that incoordinates cause inferences that act in a way similar to cataphors, as “comprehenders arguably engage in a type of active linguistic prediction” (Kush & Dillon, 2021: 1). In fact, only when we have access to the content of the incoordinate do we know what expectation is denied:

In these examples, the expectation is not overtly realized, and its content can only be established in a backwards process. Thus, in (45a,b), which are Type I incoordinates, only after the incoordinate is uttered do we know that speaker B did not expect the food to be good (45a), or that given past events they expected speaker A to not eat meat (45b). Considering traditional cultural beliefs, we can assume that in (46) the expectation would be for the bride to wear a dress instead of pants. Thus, the incoordinate has the function of establishing the content of the expectation, and also denying it. In these cases, we assume a reconstruction phenomenon happening at LF, similar to what Cyrino (1996) proposes for null object. However, we distance our proposal from Cyrino’s regarding the existence of ellipsis, which we consider that preferably does not occur in incoordinates licensed by situational context. Also, admitting a bottom-up approach to structure derivation, it is plausible that the expectation in these cases can only be obtained after the incoordinate is built.

However, the fact that cases as (42) and (43) appear in a context that has previous linguistic material grants these constructions different properties.

Assuming that sentence proforms have linguistic content and can be legitimized by their connection to a previous linguistic context, we consider important to note that only incoordinates as (47) and (48) allow the recovery of the content of the first conjunct with a proform such as Sim (‘Yes’). On the other hand, incoordinates as (49), that occur out of the blue, do not legitimize the occurrence of Sim:

Schwenter (2000: 2), analyzing the different behavior of adversative conjunctions in Spanish (pero and sino), also notes this possibility with the conjunction conveying counterexpectation (pero):

According to Schwenter (2000: 2), the contrast/opposition value of pero (‘but’) is corroborated by the occurrence of Sí (‘Yes’), which asserts that both agree on Juan being intelligent. And, as the author states:

The role of pero is that of introducing an argument that is stronger than the one put forth by A; more explicitly, then, B’s response can be understood as conveying “while it is true that Juan is intelligent, and that this is an argument in favor of hiring him for the position, it is also true that he doesn’t know anything about linguistics, and this is a decisive argument for NOT hiring him”.

This possibility is also present in Italian, as noted by Giorgi (2018: 12), that shows that ma can occur as a conjunction (51) and, if uttered by different speakers (cf. (52)), the conjuncts correspond to discourses, and the second conjunct can be preceded by Sì (‘Yes’), which reports to the previous discourse fragment:

Since we need to obtain contrast in adversative constructions, as it is the basic value of the conjunction, the occurrence of Não (‘No’), instead of Sim, is precluded from sentences as (53), (54) and (55). Also, the expectation that the incoordinate denies is presupposed by the speaker, so Não would imply negating the speaker’s own psychological process.⁸

The possibility, or not, of recovering the content of the previous discourse fragment will thus be one of the distinguishing properties between these two types of incoordinates. It is possible to consider, in line with Evans (2007), that both constructions are incoordinate, but that they are at different points in the incoordination process, in which those of the type of (42) and (43), vs. (44), are still dependent on the recovery of the linguistic material that legitimizes them. This would also be in line with Aarts (2007: 241) proposal of syntactic gradience, being in this case what the author defines as subsective gradience, as it allows “for members of a class to display the properties of that class to varying degrees.” Thus, as mentioned, we propose the inclusion of paratactic constructions into a continuum of connectivity, which includes integrated coordinate sentences but also incoordinates (Type I and II).

Thus, Type I incoordinates are at an earlier stage of incoordination and still dependent on a previous linguistic context; differently, type II incoordinates correspond to a more independent stage, in which there is an extension of the sentence domain to the pragmatic, motivated/guided by the context. Drawing a parallel with Evans’ (2007) analysis of insubordination, we could assume (i) that Type I incoordinates are in the first stage of incoordination; and (ii) Type II incoordinates, are in the last stage of incoordination and can, therefore, occur independently.

Note that in Evans (2007) proposal there are four stages and that we are reducing it in our proposal to two. We chose to eliminate Evans’ stage I, as it would correspond to an integrated canonical coordinate sentence, and we combined stages II and III, because stage II is defined by the author as the one in which an ellipsis occurs. In our analysis, however, in Type I incoordinates, the linguistic material is preferably inferred from the previous discourse fragment, or from information shared between the speakers (common ground); recovered by a sentential proform (Sim); or by an elliptical constituent.

In Type I incoordinates, we assume that it is possible to partly recover linguistic material, although not in a strict sense. Regarding example (56), we admit it is possible i) for the incoordinate to occur as a reaction to an expectation of the sort “I expect this to be bad” or ii), alternatively, given that ellipsis can correspond to material recovered from a situational context, it is possible that speaker B reconstructs something as “És mau a cozinhar, mas isto está ótimo!” (‘You are bad at cooking, but this is great!’). If this second alternative is assumed, then we have elided linguistic material that corresponds to the first conjunct.⁹

As for Type II, since they occur out of the blue it is impossible to assume that there is ellipsis, i.e., taking into account its classical definition, as the omission of a linguistic expression that may be recovered by the linguistic or situational context that, usually, precedes the elided material. In this case, the expectation is inferred from the content of the incoordinate.

Bearing in mind the different characteristics of Type I and II incoordinates, we will adopt Cristofaro’s (2016) proposal for the development of insubordinate sentences as it is the most appropriate to account for incoordinates. The author’s work combines several processes, as ellipsis and clausal disengagement (or extension, in Mithun (2008)), which can be applied to different types of sentences. The concept of extension was proposed by Mithun (2008), regarding insubordinates, which are legitimized through an extension mechanism that allows them to occur independently in the absence of the matrix clause because there is an extension of the sentence domain into the discursive and pragmatic domains (cf. Mithun, 2008: 69). We assume that, particularly, in Type II incoordinates, as they are related to cultural beliefs, there is an extension of the sentence domain to the pragmatic one, which allows the sentence to be felicitous.

Below, we present Table 1, whose aim is to systematize our proposal for incoordination. As mentioned, there are two types of incoordinates, depending on their occurrence in relation to a previous discourse fragment (Type I) or regarding a situation (Type II). As for Type I, these sentences can also be divided into two subtypes: i) Type I – A: the incoordinate is a reaction to an unexpected situation in relation to the content of the previous discourse fragment; and ii) Type I – B: the incoordinate is a reaction to an unexpected situation regarding information that is in the common ground, i.e., it entails shared knowledge between speakers. As for Type II, these are related to cultural beliefs or preconceived notions of our knowledge of the world.

As we stated, incoordinates occur as reactions to events that are unexpected to the speaker. The speaker’s expectations have different origins: in Type I – A, the expectation is created only by the content of the previous discourse fragment; in Type I – B, it arises from the previous discourse fragment but also from information in the common ground; and, in Type II, it is the incoordinate in the situational context that occurs that allows us to infer what was expected.

3.2. The structure of incoordination

The syntactic structures of coordinate sentences have been the subject of several proposals in the literature: tripartite structures that do not consider the conjunction to be the head of the structure (Johnson, 2002; Borsley, 1994, 2005; a.o.); binary structures, which also consider the conjunction not to be the head of the structure (Chaves, 2007); binary structures that obey the Specifier-Head-Complement configuration, each conjunct occurring as a specifier or a complement of Conj (Kayne, 1994; Johannessen, 1998; Matos, 1995, 2003; Colaço 1998, 2005); and, finally, binary Adjunction structures (Munn, 1992, 1993, 1999), which assume that the first conjunct is outside of the structure of ConjP (in Munn, BP), in Syntax.

Nevertheless, incoordination poses problems for both configurations, since the specifier position, which in integrated coordination attributes to ConjP its features, is typically occupied by a non-overt category. As we showed before, the incoordinate, even when there is a previous discourse fragment, does not occur as a response to that sentence, so it is not possible to assume the existence of elided material, corresponding to the first conjunct. Considering this, the issue arises concerning a configuration lacking a specifier position and how ConjP’s categorical features will be assigned, as per the widely accepted proposal (cf. Johannessen 1998), they are established through Agree between the first conjunct and the conjunction.

In this respect, Matos (1997, 2000) proposes that ConjP assumes its categorial features by Agree with its specifier, except when the specifier is not overt, in which case ConjP receives its features from the complement, a proposal that is further explored in Matos (2009) for parenthetical coordination. In this paper, the author, explores the operation of Pair Merge in parentheticals and states that “in an adjunct configuration, the underspecified categorial value of Conj is fixed by the head of this complement” (Matos 2009: 167).

Colaço & Matos (2010: 14) also highlight that the proposal incorporating the specifier projection and the one that does not are not inherently incompatible. In other words, for standard binary coordination, the traditional Specifier-Head-Complement configuration, [ConjP [XP] [Conj’ [Conj] [YP]] is the unmarked option. On the other hand, for structures with only one conjunct in complement position, Colaço & Matos (2010) propose a configuration as [ConjP [Conj’ [Conj] [YP]]. The authors also claim that there is nothing to stop Agree from operating in both the [ConjP [Conj’ [Conj] [YP]]] and the [ConjP [XP] [Conj’ [Conj] [YP]]] configurations.

In our proposal, the sentence that occurs in complement position is a CP, as it bears exclamative/interrogative illocutionary force (cf. Giorgi, 2018). Our proposal bears in mind the gradience associated to paratactic constructions (cf. Matos, 2005, for EP), that include coordination and juxtaposed sentences. In juxtaposed sentences, as (57a), there is no explicit conjunction, even though, in some cases, it can be inferred (57b).

Bearing in mind that despite the presence of the conjunction, these sentences fit into the paratactic spectrum, we admit that incoordinates can also be included in this group of constructions. In incoordinates, as we showed in the beginning of section 3, the content of the utterance gives us information regarding the expectation that is not overtly realized and the denial of that same expectation. Note that Type II incoordinates, even though they occur in the absence of previous linguistic material, still need an adequate situational context to be felicitous. Consider the example presented before about the bride wearing a red dress, the incoordinate can occur in the context of a wedding, but not in the context of the father watching his daughter lift a big box. So, taking this into account, both in Type I and II, the incoordinate occurs juxtaposed to the previous linguistic fragment (Type I) or to the situational context (Type II).

This idea that some discourses and texts may be governed by the same or similar conditions that govern the combination of constituents is not new. Giorgi’s (2016, 2018) work is clearly inspired by Cinque (2008), that analyses examples as (58), assuming Kayne’s (1994) Linear Correspondence Axiom (LCA) to hold of Discourse Grammar. According to LCA, linear precedence in a discourse must also reflect asymmetric c-command.

Cinque proposes (59) as the syntactic representation for (58), explaining that, in order to account for the LCA, it is needed to “merge the linearly preceding sentence in the specifier of an (empty) head, which takes the following sentence as a complement” (Cinque, 2008: 118).

The author mentions that the constituent in specifier position can also be a discourse fragment, corresponding to a DP, in a sentence of the type A pink shirt? I will never wear any such thing in my life!, in a configuration as below:

In Cinque’s configuration, the H projection is used to signal the blocking of any Sentence Grammar (e.g., c-command) relation between specifier and complement, which is the general rule for sentences in discourse. We will adopt this projection, as between juxtaposed sentences/discourses we also want to block the relations between specifier and complement. As for mas, as shown, it maintains its conjunctional status and, thus, in our proposal, we preserve the projection of ConjP.

We present below the syntactic configurations for Type I and II incoordinates:

Note that Type I incoordinates are at an initial stage in which previous linguistic material is needed to legitimate their occurrence. In (61.B), the previous linguistic content is not recovered and, thus, is not projected in ConjP. In this analysis, as Giorgi (2018) proposes, the expectation is in specifier position and the incoordinate is contrasted with it. As mentioned, that linguistic content can be recovered through the sentential proform Sim. The representation presented above is similar to Giorgi’s (2018: 17), as it assumes the possibility of Sim occurring in specifier position, which “refers anaphorically to the first conjunct uttered by the other speaker.” The derivation proposed in (61.B’) considers the possibility of recovering linguistic material, later elided, through the situational context, which in this example we assume to be something along the lines of “you are bad at cooking”, that, regarding its content, it is similar to what the sentential proform Sim recovers.

In the representation above, we have the syntactic configuration for Type I – A incoordinates, as it occurs as a reaction to the previous discourse fragment. In our analysis, the structure for Type I – B is the same except for the expectation in the specifier position of ConjP, as it would be the information in the common ground.

Below, we have the syntactic structure for Type II incoordinates, which is similar, but instead of having two discourses juxtaposed, we have the incoordinate juxtaposed to a situation. We emphasize that the projection of the situation is relevant as not all contexts legitimize the incoordinate.