American Spanish dizque from a Functional Discourse Grammar perspective

2.1 Peninsular Spanish

The word dizque originates from the third person singular present indicative form dize ‘[he/she/it] says’ followed by the complementizer que. The apocopated form diz occurs from the earliest texts onwards. The evidence comes from medieval juridical texts from the period between 1200 and 1400 in the diachronic CORDE corpus, which contains 530 occurrences of diz against 2,359 of dize (cf. also Eberenz 2004, 142–4). Before going into more details, note that the apocopy in the case of dize does not occur in isolation; rather, the elimination of final /e/ seems to have been quite fashionable in medieval times. In fact, the very same set of medieval juridical texts from 1200 to 1400 contains several other forms in which final /e/ is omitted: (i) noche ‘night’ (669 tokens) is reduced to noch in 169 cases, (ii) por ende ‘therefore’ (1,587 tokens) occurs without the final /e/ in 86 cases, and (iii) the relative pronoun donde ‘where’ (413 tokens) occurs as dond in 31 cases. Torreblanca and Blake (2002) provide a detailed study of apocopated verb forms. Taking this into account, the apocopy of dize is less noteworthy than claimed in certain accounts (e.g. Magaña 2005), particularly given the high frequency of this verb.

However, with respect to the combination of diz with the complementizer, it is remarkable that the frequency relation between the variants is different: there are 470 cases of dize que, against 277 of diz que, which constitute more than a third of the total. In the juridical texts, both dize que and diz que are used with human referents and also to refer to the contents of laws and similar, as illustrated in (3) and (4), respectively:

In both cases, there is no personal subject, a situation which Miglio (2010, 14) considers a step towards the development of modern dizque, although it is not clear which of the two forms is preferred in these contexts.

It is only in diz que that the apocopated form survives in post-medieval Spanish, whereas final /e/ is restored in all other cases mentioned. This indicates that diz que has come to form a petrified unit, as can be seen in the following example, quoted from Eberenz (2004, 149):

The most probable interpretation of diz que in this context is not that of a finite verb followed by a complementizer because diz fails to agree in tense and number with the preceding verb dezían “they said.”

The following example is similar to modern usage, not only with regard to the orthography of dizque as a fused form, but also due to its low pragmatic prominence (Olbertz 2007, 154), as it appears in a relative clause which forms a third level embedding in relation to the main clause:

Although in this clause dizque follows the relative pronoun, its scope is the same as in (5), i.e. the entire clause. This scope relation is less clear in the following case, where the adverb precedes a noun phrase:

• ‘[The captain mentioned earlier sends a nobleman and notorious tyrant […] together with certain Spanish men]

• a. in order to punish as he puts it the rebellious Indians who were fleeing from such infectious diseases and massive death’

• b. in order to punish the what he calls rebellious Indians who were fleeing from such infectious diseases and massive death’

• (Bartolomé de las Casas, Brevísima relación de la destrucción de las Indias, 1552)

Both are possible interpretations, although, given the context, (7b) may be a bit more probable; there is no way, however, to be sure, because in sixteenth-century Spanish, dizque does not yet occur inside the noun phrase. Irrespective of how one interprets dizque in (7), this case exemplifies a general characteristic of the use of dizque, i.e. that it tends to go hand in hand with the speaker’s dissociating himself or herself from the reported content.

2.2 Grammaticalization or lexicalization?

How should this process be accounted for? The answer from almost all recent studies on this phenomenon is that dizque undergoes a process of grammaticalization. Those who argue explicitly in favour of this analysis mention (a) univerbation, which implies the loss of verbal properties of diz, (b) semantic change, (c) the loss of the possibility of being negated, and (d) the possibility of being preceded by que (as in example (6) above) (e.g. Magaña 2005, 64–6). Considering dizque the result of a grammaticalization process is based on the idea that univerbation is necessarily related to grammaticalization, which is, however, a misunderstanding: univerbation is in fact a necessary feature of lexicalization, which consists in the fusion of two or more words, or, as Lehmann (2002, 13) puts it: “Only complex units can be lexicalized”. As regards the relation between grammaticalization and lexicalization, Lehmann (2002, 15) explains that the two processes have certain properties in common, as “[b]oth lexicalization and grammaticalization are reductive processes which constrain the freedom of the speaker in selecting and combining the constituents of a complex expression”.

If we consider dizque to be the result of lexicalization, the four morphosyntactic and semantic properties mentioned above are a consequence of the petrification of two frequently co-occurring words, diz and que, to form the new word dizque (property a). Being a new word, diz has undergone a change of word class, i.e. the new word is an adverb, which again entails semantic change (property b). This semantic change is such that negation is indeed excluded (property c), a point to which I will return in Section 4.3. Finally, being an adverb, dizque is syntagmatically flexible, i.e. it can occupy various slots in the clause, among which that at the beginning of relative or complement clauses, and as a consequence, it may be immediately preceded by complementizers or relative pronouns, both realized in Spanish as que (property d).

There are two additional properties of dizque that favour the lexicalization hypothesis. First, being an adverb, the meaning of dizque can be expressed by alternative adverbial expressions such as según dicen ‘as people say’ or supuestamente ‘supposedly’ or any other contextually appropriate alternative, and its use is certainly not obligatory, rather, it is a free choice item; in Lehmann’s terms, its “paradigmatic variability”, i.e. “the freedom with which the language user chooses a sign” (Lehmann 2015, 146), in other words, the ease with which an item can be substituted by another item or just be left out, is not affected as would have been expected in the case of grammaticalization (Lehmann 2015, 147–52). Second, as briefly mentioned above, dizque is not bound to a specific slot in the clause or sentence; had this been the case, that could have warranted the claim of grammaticalization. Rather, in Lehmann’s terms, dizque has not lost its “syntagmatic variability”, i.e. “the ease with which it can be shifted around in its context” (Lehmann 2015, 167), a fact that will be the central issue of Section 3.1.

Therefore, I consider dizque an adverb, more specifically a reportative adverb in most of its uses. Reportativity is the indication that “the Speaker is relaying the view of others” (Hengeveld and Mackenzie 2008, 103), which may involve second-hand information, hearsay, or folklore (Willet 1988, 57). By marking information for reportativity, the Speaker indicates to the Addressee that she or he is not to be held responsible for the relayed content, which means that reportative marking may serve as a hedge.^[3] This is why dizque will be treated as a modifier at the Interpersonal Level.

3.1 Colombian and Mexican Spanish dizque

In modern usage, dizque scopes over the entities which it immediately precedes. This section describes the variation in scope of dizque as an interpersonal modifier. The presentation of the data will proceed from the highest towards the lowest layer of modification. Consider the following example:

In (8), the reportative adverb dizque modifies the entire Communicated Content. However, this does not always imply that it occurs in clause initial position:

The occurrence of dizque in second position here is due to the topicality of the personal pronoun ella in (9) (cf. e.g. Cruschina and Remberger 2008, 111; Demonte and Fernández-Soriano 2022, 9).

In the following example, dizque takes a Communicated Content in its scope that modifies another Communicated Content:

In this example, dizque modifies por respeto, which, again, modifies the Communicated Content no mostraron el cadáver. That por respeto is indeed a Communicated Content on its own is indicated by means of the commas separating this phrase from the remainder.

It should be noted that (10), more clearly than (8) and (9), implicates a negative Speaker attitude with regard to the content modified by means of dizque. The same holds in (11), where the reportative adverb modifies a noun phrase, i.e. a Referential Subact, which is a relatively infrequent phenomenon (cf. Section 3.2, Table 1).

The scope of dizque is further reduced, when dizque modifies nominal or adjectival predicates that occur within a noun phrase, which occurs more frequently in Mexican than in Colombian Spanish (Table 1). In (12), dizque modifies an attributive adjective, estable, i.e. the expression of a modifier of the Ascriptive Subact restricting a Referential Subact (coded as parejas estables):

In (13a), dizque modifies the only Ascriptive Subact heading the Referential Subact, realized as a noun in the plural, adelantos ‘advances’:

As dizque immediately precedes the noun adelantos in this example, it is used here as if it were an adjective, which it is not because it cannot be inflected. (13b) is evidence of the fact that true adjectives agree in gender and number with the nominal head:

From a syntactic point of view, having an adverb fulfil an adjectival function is quite a strange situation. A possible explanation of this situation will be dealt with in the next section.

3.2 The gradual scope decrease of dizque

In this section, I will argue, on the basis of contrastive and diachronic data, that the application of the modifier dizque has generalized from Communicated Contents (C) to modifying Communicated Contents (mod C) and Referential Subacts (R) and only later to Ascriptive Subacts (T). This means that, unlike what would be expected in grammaticalization (Hengeveld 2017), the scope of dizque tends to decrease rather than increase.

As I mentioned earlier in passing, there is a difference between Colombian and Mexican Spanish with respect to the applicability of dizque to nominal, adjectival, and verbal predicates; this relation is specified in Table 1. In this table, each of these uses is related to a particular interpersonal category, where mod should be read as “modifying element of”, and the label “main clause” includes all finite clauses with the exception of adverbials.

As this table shows, the use of dizque with nominal predicates is much more advanced in Mexican Spanish than in Colombian Spanish; the same holds, to a somewhat lesser degree, for attributive adjectives. This difference becomes clearer when we consider this in terms of FDG: in Mexican Spanish, there is a strong tendency towards applying dizque to Ascriptive Subacts (T), whereas the modification of Communicated Contents (C) by means of dizque is considerably less frequent than in Colombian Spanish.

That this really is a matter of development becomes more obvious when we consider diachronic data.^[4] Table 2 compares written texts from Colombian Spanish from CORPES XXI (2001–2019) with Colombian data from CORDE (1896–1956).^[5]

Whereas in the older texts, dizque was almost exclusively used as a modifier of Communicative Contents (C), in the present-day texts it has extended towards the modification of modifiers of Communicative Contents (mod C); it has also come to modify of Referential (R) and Ascriptive Subacts, i.e. (T) or (mod T), albeit to a much lesser extent.

Interestingly, it turns out that none of the cognates of dizque in present-day Brazilian Portuguese (diz que, cf. Casseb-Galvão 2011) and Galician (disque, cf. Sousa 2012) modifies Ascriptive Subacts (cf. also Sanromán Vilas 2020). Similar reportative markers in other Romance languages are constrained in the same way (cf. Cruschina and Remberger 2008 on, among others, Sardinian nachi and Sicilian dicica, Dumitrescu 2012 on Romanian cică, and González Vázquez 2021 on the – etymologically deviant – Galician marker seica). This means that, with respect to scope decrease, Spanish dizque is more advanced than its Romance cognates.

In sum, it has become clear how the application of dizque to Ascriptive Subacts must have come about. This implies for FDG that the development of modifiers is not necessary on a par with the development of grammatical items, such as aspectual and modal operators,^[6] which increase their functional scope in the process. Although lexicalization may advance in the same direction, as shown by Fontes (2016, 206–28) for Brazilian Portuguese ainda bem ‘just as well’, the case of dizque shows that a development of lexicalized items in the opposite direction is also possible.^[7]

4.1 First person contexts

Consider the following examples, in which dizque is used with a first-person referent:

In examples (14) and (15), the source of the reported Communicated Content is external to the first-person narrator, with the difference that dizque is used redundantly in (15) as it occurs in a speech act complement. In (14), there is no implicature with respect to the evaluation by the Speaker of the Communicated Content, and in (15), there is no implicature either; rather, the truth of the possible claim is explicitly rejected. In (16), however, the case is different since the first-person narrator himself is the source of the reportative. More concretely, he is commenting on his own behaviour, thus implicating that the proposition presented within the Communicated Content is a lie. The disparity of these examples indicates that the use of first-person subjects with dizque is not necessarily related to an implicature of doubt as several authors claim (e.g. Travis 2006, 1290; Demonte and Fernández-Soriano 2013, 218; De la Mora and Maldonado 2015, 171). Such a relation exclusively exists when the first person is the source of the report and the context is such that he/she describes his/her actions.

Despite the fact that there is no systematic relation between first-person reference and the use of dizque, the examples in (14)–(16) have something else in common, which is unrelated to first-person reference: in each case, dizque modifies a full Communicated Content. Therefore, I take this opportunity to provide a general FDG representation that holds for these cases, i.e. the combination of first-person reference with dizque as a modifier of the Communicated Content:

• (17) (A_i: [(F_i: ILL (F_i)) (P_i)_S (P_j)_A (C_i: [(T_i) (+id R_i: [+S, –A] (R₁))] (C_i): dizque_Adv (C_i))] (A_i))

This representation should be read as follows: (A) corresponds to the Discourse Act. As the elements of the Discourse Act form a non-hierarchical, i.e. equipollent, configuration, they are enclosed between square brackets. (F) represents some illocution, which would be declarative in the cases of dizque we have seen so far. The two variables (P) represent the participants of the interaction, whose functions are indicated by means of the subscripts “S” for Speaker and “A” for Addressee. The Communicated Content (C) contains the Subacts of Ascription (T) and Reference (R), which, again, form an equipollent configuration. The latter contains the operator + id indicating identifiability, and the content of the Reference variable is the feature [+S, –A] indicating first-person reference. In order to avoid massive sequences of closing brackets, each entity (also called “layer”) contains a closing variable. The first closing variable of (C_i) is followed by a colon, which introduces the modifier; in this case dizque, the subscript Adv indicates that it is an adverbial modifier.

4.2 Properties of dizque specific to Mexican Spanish

In this section, I come back to the Mexican bundle of “others” mentioned in Table 1. More specifically, there are a number of non-reportative tokens in the Mexican sample, which will be discussed here, and their semantics will be contrasted with the reportative usage of dizque.

Example (18) is a further case in which the Speaker is the source of dizque. Nevertheless, this case is entirely different from those presented in the previous section, in the sense that there is no Communicated Content involved.

This case crucially differs from those we have seen so far, in that dizque is not related to communication but just means ‘pretend’ or, in adverbial terms, ‘seemingly’, i.e. it is used beyond its originally reportative meaning. As is to be expected, this use of dizque is not restricted to first-person subjects, as the following examples illustrate.

• ‘Remember how Ana, my cousin, hit you with her bag because, while pretending that you helped her cross the road, you took her by her upper arm so that you could touch her left breast.’

• (Rafael Tovar y de Teresa, Paraíso es tu memoria, 2009, Mexico)

Since (19) and (20) do not involve speech reports, dizque has no longer an interactional function but is used in the description of events. In these cases, dizque serves as what could be called a ‘pretence marker’, indicating that the event described serves an unexpected purpose. Therefore, in FDG, the use of dizque in these three examples should be accounted for as a modifier of an event (aka “state-of-affairs”):

• (21) (e₁: [(f₁)…] (e₁): (f₂: dizque_Adv (f₂)) (e₁))

The system of this structure at the Representational Level is the same as in the Interpersonal Level representation in (17) above. As the content of the event is different in each case, the event variable (e₁) is given just a numeral subscript instead of an identifying letter. The same holds for the other variables. In this case, dizque is marked as descriptive lexical predicate represented by means of the variable (f).^[8] This formula should be read as “any state-of-affairs such that it is modified by means of the pretence marker dizque”.

De la Mora and Maldonado (2015, 176) claim that “[i]n current Mexican Spanish the pure [sic] evidential reportative is practically lost”, but that this is being recovered by the innovative complex marker que dizque. Indeed, que dizque does occur in my corpus, of which the following example is representative:

However, my sample only contains five such cases, excluding those in which que is a complementizer or a relative pronoun.^[9]

Despite De la Mora and Maldonado’s (2015) claim concerning the gradual loss of the reportative meaning of dizque, there are quite a number of cases that presuppose a verbal act by some source, be it external or internal to the speaker. Consider the following two examples:

In (23) dizque precedes a bare noun and in (24) an attributive adjective, which according to De la Mora and Maldonado (2015, 175) corresponds to what they call the “disqualifying function” of dizque, and for Martínez Levy (2019, 167–9) such expressions have an ‘ironic meaning’.^[10] However, in both cases, dizque reflects a claim that has been made explicitly by someone in (23): either the individuals in question themselves or the referent of el cabrón (an individual named Luismi); in (24), the attribution of exorbitante was probably made by the men who asked for the prices of the brothel.

In the following example, dizque precedes an adverbial purpose clause, which, as such, also belongs to the expressions of the “disqualifying function” identified by De la Mora and Maldonado (2015, 173).

If the meaning of dizque really were “disqualifying”, then the contrastive clause starting with cuando en realidad ‘while in fact’ would make no sense, because dizque itself would already express falsity. Rather, what is marked linguistically in these examples is just reportativity, nothing more. ‘Disqualification’ and ‘pretence’ are conversational implicatures, inviting the Addressee to infer that either the Ascription is inappropriate, as in (23) and (24), or that the propositional contents evoked by the modifier of the Communicated Content is false, as in (25).

It is probable that the frequent use of dizque with a relatively low scope, i.e. as a modifier of Ascriptive Subacts and as a modifier of Communicative Contents, in Mexican Spanish has led to a reinterpretation of the conversational implicature of ‘pretending’ or ‘deceit’ as part of its meaning, which, again, will have led to the use of dizque outside reportative contexts, where it expresses these very lexical meanings: pretence and deceit (cf. also Travis 2006, Olbertz 2007, Sanromán Vilas 2020).

4.3 Properties of dizque specific to Colombian Spanish

This section is dedicated to two kinds of deviant uses of dizque: (i) the occurrence of dizque in negated rhetorical questions and (ii) the use of dizque in a non-reportative mirative function. Both form the Colombian group of “others” in Table 1.

5.1 Distinguishing lexical items from grammatical formatives

In FDG literature, it has been claimed that grammatical formatives can be identified by their lack of modifiability (Hengeveld and Mackenzie 2008, 59). In their discussion of the lexicon, the authors are more specific: “Only those elements that are susceptible to modification can be fully lexical. This clearly applies to full verbs, nouns, adjectives and adverbs.” (Hengeveld and Mackenzie 2016, 1157).

I believe to have made clear that dizque is a lexical item. However, it shares with other so-called “sentence adverbs” not only the impossibility of negation (cf. example (26) in Section 4.3), but also the property of not allowing for modification, which again is not specific to dizque but also applies to clearly lexical adverbs such as supuestamente ‘supposedly’. For evidence, I reuse example (26a) here, renumbered as (34a) for convenience. As both adverbs are of an interpersonal nature, only interpersonal modifiers, based on Keizer (2018a, 75), have been tested.

This means that, although as a general rule, the claim that modifiability serves as a criterion for the identification of lexical items may hold, it is certainly not watertight, particularly not for interpersonal adverbs.^[14] Rather, I believe that in cases of doubt purely morphosyntactic criteria, such as the ones discussed in Section 2.2 (based on Lehmann 2015, among others), are a reliable alternative.

5.2 Dizque with nominal predicates

As mentioned earlier, there is a morphosyntactic problem when dizque modifies the head of a Referential Subact, i.e. the Ascriptive Subact, because such Ascriptive Subacts are by default realized as nominals. As dizque does not have adjectival properties (cf. Section 3.1), and in this case, cannot be an adverb either, it must be a particle in this context. As opposed to adverbs, particles belong to a (relatively) closed class, and therefore are, in a way, closer to grammatical formatives than to lexical items. But, on the other hand, they do allow for focalization, as illustrated in this example, quoted from Olbertz (2007, 168):

In this example, también ‘also’ focalizes dizque, which means that even if it is grammaticalized to a certain extent, it is certainly not a true grammatical formative. Because, in general, grammatical formatives cannot be focalized, I consider it an intermediate case.^[15] A possible solution offered by FDG is to account for dizque as a lexical operator (Keizer 2007), which would correspond to the following representation, in which dizque takes the position of an operator of the Ascriptive Subact within the Referential Subact, but at the same time retains lexical properties:

• (36) (R₁: (dizque T₁) (R₁))

5.3 Dizque scoping over modifying elements

There are two cases in which dizque has scope over modifying items at the Interpersonal Level: (i) within Referential Subacts, it scopes over the modifying ascriptions (encoded as attributive adjectives and similar), and (ii) as regards Communicated Contents, it scopes over modifying Communicated Contents (encoded as adverbial phrases or clauses). For ease of exposition, I will begin with the former.

Examples of the use of dizque in modifying ascriptions embedded in Referential Subacts have been quoted in (12) and (24). A further example is the following:

Example (37) differs from slightly those quoted in (12) and (24) in that the modifying Subact is not realized as an adjective but as the prepositional phrase de concienciación nacional. As regards the Interpersonal function, there is no difference. The appropriate way of representing such cases consists of having the reportative modifier apply to the second Ascriptive Subact embedded within the Referential Subact. Therefore, the following representation corresponds to the use of dizque illustrated in (37):

• (37) a. (R_i: [(T_i  – movimiento – (T_i)

•   (T_j:  – de concienciación nacional – (T_j): dizque_Adv (T_j))] (R_i))

As this representation is specific of (37), I have added the lexical content for clarification purposes. Obviously, this lexical content does not form part of the Interpersonal Level. This fact is indicated in (37a) by means of the n-dashes, which indicate that all relevant steps pertaining to the Representation and to Encoding have been skipped here.

With regard to the second case, I propose to introduce a general modifying function to Communicated Contents into the theory. Indeed, FDG allows for assigning functions to all entities at the Interpersonal and Representational Levels, but function assignment to Communicated Contents has not yet been explored systematically.^[16] On the other hand, based on empirical evidence, FDG does distinguish between independent and dependent entities at the layer of the Discourse Act. In my view, the distribution of dizque is evidence in favour of distinguishing between independent and dependent Communicated Contents as well, the latter would then correspond to additional content that modifies in some way the primary Communicated Content.

Cases in which dizque modifies a modifying, i.e. dependent, Communicated Content quoted so far are (10) and (25). Example (38) is similar:

In this case, it is tempting to assume that the function of the modifying Communicated Contents is “Reason”, which would also apply for example (10), whereas in the case of (25) it would be “Purpose”. However, such functions form part of the semantics of these examples and therefore correspond to the Representational Level. What we need in the present case is a general non-semantic function, which at the same time avoids making use of highly general concepts, as would be “modification”. An appropriate label could be “Elaboration”, to be abbreviated as “E”. The assignment of this function would mark the dependent Communicated Content as such, thus distinguishing it from independent Communicated Contents, which bare no such function. A simplified interpersonal representation of (38) would have to look as follows:

• (38) a. (C_i:  – en la agencia nos prohíben decir cómo nos llamamos – (C_i))

•    (C_j:  – por seguridad –  (C_j): dizque_Adv (C_j))_E

Note that the dependent Communicated Content is not embedded in the independent one, but, as the following more general representation will show, both form part of one and the same Discourse Act:

• (39) (A₁: [(F₁: ILL (F₁)) (P₁)_S (P₂)_A (C₁: [(T₁) (R₁)] (C₁)) (C₂: […] (C₂): dizque_Adv (C₂))_E] (A₁))

This representation is meant to account for all cases in which dizque scopes over a modifying Communicated Content, independently of the mutual ordering of the two types of Commicated Contents, which is a matter of morphosyntactic encoding.^[17]

Summing up, accounting for the use of dizque to modifying Ascriptive Subacts does not require any adaptation of the tools FDG offers. Concering the application of dizque to modifying Communicated Contents, I have introduced the concept of dependent Communicated Contents marked by a general function “Elaboration” (E), thus distinguishing it from independent Communicated Contents.

5.4 How to link the lexical item dizque to its interpersonal frame?

Although dizque as an adverbial modifier is quite flexible, there are some restrictions. In order to illustrate the available options, I will reuse four of the examples provided in this article, renumbered for convenience and each followed by the corresponding frame.

These representations correspond to the most frequent uses in my sample. Therefore, I refrain from including the infrequent modification of Referential Subacts and the mirative use, which, at least for the time being, can be considered the result of coercion.

The association of lexical items with the frames into which they are to be inserted is a matter of discussion in FDG. Hengeveld and Mackenzie (2016, 1143) propose the following procedure: “The restrictions on the compatibility between lexemes and frames need to be represented somehow in the lexicon. This can most simply be done by numbering the finite set of frames and indexing each lexeme with the numbers of the compatible frames. All the lexemes that share index n will then be insertable into frame n”. Applying the simple example of the above-mentioned four frames to dizque yields a lexical entry of dizque in the Interpersonal and Representational lexicon, as something like: dizque ^1,2,3,4. This open-ended approach leaves room for the addition of further frames when some innovative use of dizque becomes part of common usage.

Uncommon abbreviations in glosses

dmrk

discourse marker (pues)

dom

differential object marking

pol

polite

Abbreviations in representations

Interpersonal Level:

A

Discourse Act; Addressee (function and operator)

C

Communicated Content

E

Elaboration function

F

Illocutionary predicate

id

identifiability operator

P

Participant

R

Subact of Reference

S

Speaker function and operator

T

Subact of Ascription

Representational Level:

e

event (=state-of-affairs)

f

lexical predicate