2.1. The notion of concession

Broadly speaking, concessive relations arise from the assertion of two events in a context of incompatibility. The typology of the relations is vast, as is the array of the syntactic structures which perform them. In this paper, three widely recognized concessive types (see Latos, 2009; Moeschler & Spengler, 1982; Pander Maat, 1999; Rudolph, 1996) shall be discussed: concessives of negated causality, restrictive concessives and argumentative concessives (see Spevak, 2005, for a complete typology), as in (4) to (6). The concession through negated causality is based on the negation of a causal implication (normally if p, not q), and the restrictive concession, on a rectification or mitigation of the validity of a given content or part of it (q, even though p). The argumentative concession, in turn, presumes a dialog situation and a double argumentative move in which a fact is granted and immediately presented as irrelevant (it is true that p, but q).

In (4), the concessive relation is established between world facts. The fact that it rained very much allows us to pragmatically presume, based on experience and world knowledge, that the river is going to overflow as a result, and a contrast is established precisely by the frustration of this assumption, since the river did not overflow. This concessive type is canonic and is closely linked to anteposition of the concession. In (5), there is no assumption of causality. This relation is established by means of restricting the validity of the previously stated content, a kind of correction. Following the logic of this way of functioning, the concessive is always postposed. In (6), conversely, the relation is established between a move of accepting a content’s truth, for whom the other(s) or the public opinion etc. are accountable, and a second move which argumentatively weakens this content in order to support a different position than might be expected. In these three types, the concessive meaning emerges from the complex construction in which the connective is an essential element, given its procedural character.

The third type, the argumentative concession, makes the affinity explicit which exists between the ideas of assent and concession. Assent is inherent to the concessive argumentative move, as it was of the rhetoric figure of speech concessio, which, being circumscribed especially to the legal sphere, consisted in a strategy of simulating agreement with the opponent in order to gain advantages, thus counteracting potential objections to the opposite view (Letoublon, 1983; Spevak, 2005). As we shall see, in the historical development of the concessives with licet, bem que and embora, evidence is found that the genesis of change lies in argumentative moves which include assent. Although each of the sources has its specificities, as I will seek to show, they allow us to concede (for various reasons) a fact which afterwards will be presented as being irrelevant.

2.2. The derived nature of concessives and the issue of cyclicity

The domain of contrastive relations, which encompasses the concessives and the adversatives, is among the most prone to renewal over time, which is explained to a great extent by the subjectivity potential which is inherent to the constructions (Mauri & Ramat, 2012; Meillet, 1948[1912]). The derived nature of concessivity has been made evident in cross-linguistic studies, which have identified universal and free-choice quantification, emphatic assertion, temporal coexistence, volition, causal and conditional relations and negative human feelings as the most productive semantic sources (Harris, 1988; König 1985a, 1985b, 1988; Kortmann, 1997; Letoublon, 1983, among others).

Portuguese has not inherited Latin concessive connectives. In order to express concessive values, in addition to juxtaposition and intonation, Portuguese has developed a range of typically phrasal connectives (ainda que, bem que, apesar de (que), por mais que, sem que, mesmo se, etc.) through grammaticalization processes at various stages of its history. Even in Latin, concessivity was already a derived notion (Bertocchi & Maraldi, 2009; Spevak, 2005), which is reflected in the complex morphology of Latin connectives (except licet), such as quamquam, quamuis, etsi, etiamsi, tametsi and quamlibet, formed by the amalgamation of two or three grammatical words.

In view of these facts, the hypothesis for which I seek empirical support is that, although Portuguese has not inherited any Latin concessive conjunction, the means of concessive expression which emerge in different states of Portuguese could replicate analogous tendencies to those of the constitution of connectives in Latin, and in the historical derivation between these two languages, these tendencies would stand for indicators of continuity in the scenario of deep remodeling undergone by the concessive system, in addition to being indicators of cyclicity in diachronic developments.

Meillet (1948, pp. 140–141) was one of the first to conceive the existence of renewal cycles in grammatical change. To him, grammaticalization consists in a permanent procedure of linguistic renewal impelled by expressivity: les langues suivent ainsi une sorte de développement en spirale: elles ajoutent des mots accessoires pour obtenir une expression intense; ces mots s’affaiblissent, se dégradent et tombent au niveau de simples outils grammaticaux; on ajoute de nouveaux mots […] l’affaiblissement recommence, et ainsi sans fin. [languages follow a kind of spiral development: they add accessory words to achieve an intense expression; these words weaken and become mere grammatical tools; new words are added […] weakening resumes, without end]

In particular, the history of conjunctions, according to Meillet (1948, pp. 171–173) develops as a repeated and continued effort in pursuit of expressivity. However, not all conjunctions would have the same predisposition. He states that Romance languages feature a stable set of conjunctions which dates back to Latin forms, as is the case in Portuguese of the conjunctions e, ou, se and que; and another group of conjunctions with a special meaning, such as the contrastive ones, which is marked by instability. In the second group, creations would occur constantly due to the need of expressing contrast with an ever-renewed emphasis. In this perspective, the history of concessives, which displays periodic renewals of their expression means in Portuguese and other languages, seems to be a locus to observe cyclical changes.

Recent research, such as Hansen’s (2018, 2020), has suggested the existence of onomasiological cycles at the semantic and pragmatic levels. These are recurrent evolution processes in either one or two languages which descend from each other, in which a particular meaning at the content level gives room to one or more meanings at the contextual level. Hansen (2018) defines linguistic uses at the content and contextual level respectively as:

“(…) bear saliently either on a state-of-affairs in some real or imagined world that is referred to in their host clause or on the relation between that state-of-affairs and other (real or imagined) states-of-affairs.

(…) express speakers’ comments on the relation between described states-of-affairs and the discourse itself (including, but not limited to, the subjective attitudes to the states-of-affairs in question that may be entertained by either the speaker, the hearer or some relevant third party).”

(Hansen, 2018, p. 129)

Two or more phrases could be involved in a cycle and the older one could disappear or be circumscribed to certain text genres and particular language registers. As an example, Hansen illustrates the repeated evolution of a deictic temporal adverb to a contrastive marker undergone by three etymologically different items: Latin nunc (‘now’), or, from Old French, and maintenant, from Modern French.

2.3. A prototype of argumentative concessive schema

When it is present, the connective is always part of a larger connection mechanism. In concessive relations, the articulated parts pose restrictions on interpretation, and the connective provides the instructions to interpret the relation between the parts. These instructions include a supposed incompatibility, which is contextually induced (Lang, 2000; Moeschler & Spengler, 1982), resulting from the speaker’s or writer’s assessment and leading to an expectation being abandoned.

The generic idea of incompatibility underlying the concessive nuances requires refinement in order to achieve a better understanding of the potential factors that motivate semantic change. From Rossari (2014, 2015), we can draw a refinement proposal which is particularly relevant for the phenomena discussed herein. The author’s focus lies on correlative concessive constructions from French of the type certes p, mais q, which are common in dialogical contexts, where the first member starts with an assent adverb (certes, en effet, d’accord, soit), and the second, with a contrastive connective (mais).

Rossari proposes a more general concessive schema, in which the level of contribution of each construction component to the concessive meaning is defined. The schema helps explain not only certes p, mais q, but also other ways of expressing concession, such as, for instance, constructions with peut-être and the epistemic future in Italian. In that schema, the contribution of the members is made in such a way that the sequence certes p expresses assent, and therefore retrieves a previously mentioned content or one which is available in the speakers’ mental universe. In other words, p is presented as factual. Nevertheless, by itself certes p does not enable a concessive reading; it must be followed by mais q, which adds the communicatively most relevant content to the context at hand. That means that q is focal. Thus, there is a difference between the discursive statuses of p and q, this difference being explored in order to express concessivity. For Rossari, the incompatibility at the core of the concessive meaning has an informative character; more precisely, it is a discrepancy between the informational statuses (presupposed vs. focal) of the segments p and q. The argumentative move which sustains this discrepancy can be synthetized in (i) and (ii):

1. The assent marker presents the content of p as being agreed or granted by Speaker 1 (S1); however, this content has been conceived previously by a Speaker 2 (S2), thus being a known piece of information, so that polyphony is established: two discursive positions meet and the assent of S1 regarding the enunciation of S2 prevails.

2. The content of q is endorsed by S1, being the main content.

In other words, in Rossari’s proposal (2014, 2015), the achievement of a concessive interpretation between p and q requires the concomitant existence of two traits: the contrast between the discursive/informative statuses of p and q and the possibility of interpreting p as previously conceived or as already existing in the discursive background. The distinction between agreed (p) and endorsed (q) is essential: the first is background information, the second is central.

4.1. Licet: from modal verb to concessive subordinator

The modal verb licere existed throughout the Latin period, in syntactic constructions with monovalent or bivalent predicates, being combined primarily with the infinitive. Initially, it functioned as a modal deontic, having the value of permission to do something, and later developed the epistemic value of possibility. The data in (7) and (8), extracted from Spevak (2005, p. 84), provides examples of deontic and epistemic values, respectively:

Recent studies on Latin linguistics (Bertocchi & Maraldi, 2009; Spevak, 2005) show that the present tense third-person singular form, licet, in co-occurrence with the subjunctive, became a concessive conjunction over the course of a slow and gradual evolution process, indicated by ambiguous uses of licet between modality and concession. In the classical period, the conjunction role is rare, but in late Latin, licet was among the most frequent concessive connectives.⁵

The data in (9) and (10) represents ambiguous contexts, in which pragmatic enrichment takes place (bridging context, according to Heine, 2002; critical context, according to Diewald, 2002). In (9), which has been discussed by various authors (Bertocchi & Maraldi, 2009, Ernout & Meillet, 1951; Spevak, 2005), the permission or possibility value is used to grant the other(s) the freedom of saying or doing: ‘let them object, I accept it, no matter what!’. In (10), discussed in Spevak (2005, p. 87), licet dicat (‘may say’) does not refer to an actual process, but to a possibility, ‘in case he says’, which is plausible given its habituality (sicuti solet dicere, ‘as is usual’). In these contexts, the concessive nuance is of an argumentative type, presuming a dialogism whose arguments have different weights.

According to Spevak (2005), the possible concessive reading of (9) and (10) is based on the relation between licet, which marks assent, and other contextual factors, such as the contiguity between two pragmatically contrastive propositions, the occurrence of the subjunctive and the position of licet necessarily on the first member of the construction. The concessive reading related to assent is also strengthened by the fact that licet had an autonomous use to express assent, as in the dialog in (11):

At a later stage of change, licet becomes a hypothetical concessive subordinator, as in (12). These are cases in which the hypothetical value is associated with protases referring to circumstances which would have had the power to affect the content of the nuclear clause, but eventually prove to be insignificant. Licet undergoes one more semantic change, losing the conditional nuance and beginning to express factual concession, as in (13), in which the subjunctive mode prevails, with the only function of marking a syntactic dependency (Spevak, 2005, p. 156).

Evidence of the historical development between modality and concession in other languages is mentioned by Bybee et al. (1994, p. 227). Among other phenomena, the authors highlight the synthetic future tense in Spanish, where the concessive meaning would have been derived from the probabilistic use in interactional contexts.

4.2. Bem que: from emphatic adverb to concessive conjunctional periphrasis

Table 1 shows in absolute numbers the mapping results of the adverbial and conjunctional patterns of bem in the longitudinal sample.⁶ The adverbial uses of bem are very frequent in all time spans. The first uses of the concessive conjunction bem que have been documented, with reduced frequency, in the transition from Middle to Classical Portuguese, a period in which, according to the data, Portuguese already had a range of concession markers (como quer que, pero, ainda que, posto que, conquanto, sem embargo de) and displayed other renewal processes under way, such as, for instance, apesar de (que), which derives from the negative emotion expressed by the noun pesar (‘sorrow’) and develops into a factual concession, and nem que, which derives from an emphatic negative source, nem (lat. nec, ‘and not’) and develops into a hypothetical concession.

The adverb bem (Lat. bene, ‘well, fortunately’) is multifunctional. Depending on the context and scope relations, it can act as a modal, intensifier or focalizer. The modal adds a positive evaluation about a process, state or quality, such as in “sentir-se bem (‘feel well’); estar bem vestida (‘being well-dressed’); unhas bem tratadas (‘well-cared nails’)”. The intensifier indicates the highest degree in a scale and can affect predicates, circumstances or qualities, such as in “constipação bem forte (‘quite strong cold’); vida bem cheia de aventuras (‘life quite full of adventure’); chegar bem cedo (‘arrive quite early’)”. The focalizer singularizes the piece of information to its right and can affect nouns, adverbs and verbs, highlighting or specifying them, such as in: “os mouros eram bem quinze mil (‘the Moorish were well fifteen hundred’); estavam bem ali (‘they were right there’); é bem capaz de estar gostando (‘it is quite likely that he/she is enjoying it’)”.

All three usages are found from Old through Modern Portuguese. I will discuss only the focalizer, which is the most frequent in the data set and seems to be related to the change.⁷ Following Ilari’s proposal (2002), I take it that the focalization carried out by bem implies verification operations which depend on the comparison to some parameter which is either explicit or recoverable in the context. Such operations consist in verifying a number; the identity of individuals, places and times; of properties and relations; and even in verifying the factuality. In the latter, which we are interested in exploring, bem emphasizes the speaker’s commitment to the truth of the utterance, suggesting that endorsing other opinions would not come into question. According to Ilari (2002, p. 194), this emphatic use is a polyphony marker: “(…) sugere-se que é possível fundamentar a afirmação na observação imediata dos fatos ou em premissas facilmente compartilhadas, e evocam-se, polifonicamente, opiniões divergentes” [“(…) it is suggested that the statement can be supported by the immediate view of facts or easily shared assumption, and, polyphonically, divergent opinions are summoned”].

In the investigated data, the emphatic adverb bem typically occurs in contexts with complement constructions exhibiting the following characteristics: (i) bem is in the nuclear clause and the clause content falls under the scope of bem; (ii) the clause subject is in the first person; (iii) the predicates are most frequently cognition verbs (saber, entender, compreender, crer) and, to a lesser degree, verbs of speaking (dizer, jurar, aconselhar) and volition (querer, sofrer); (iv) the present indicative prevails. As an example, consider (14) to (17), which is data from Old Portuguese,⁸ prior to the emergence of the conjunctional use:

The data from (14) to (17) illustrates the function of assertive reinforcement. In all cases, the speaker’s level of adherence to the content is high, meaning that he explicitly states his belief in the truth. The emphatic marking of the truth implies in the existence of potential reasons to question it, establishing a dialogical scenario, in which antagonistic positions naturally coexist.⁹ Being something disputable, the frequency of justifications with bem along the enunciations seems understandable, such as in (14), (15) and (17). Bem is commonly found in adversative construction with mas, as in (16), which displays a correction strategy.

The data from (18) to (20), in turn, enables a concessive reading. These are correlative constructions of the type bem p, mas q, which realize the concessive schema (see Section 2.3). The concessive reading is pragmatic, being achieved compositionally based on a particular combination of bem, mas and other contextual traits, as I will discuss next.

The excerpt in (18), which tells the stories of Christian martyrs, presents the sayings of Saint Nereus about the importance of virginity. In the complex construction, p and q are contents which are stated and presented with different discursive statuses. In p, the speaker accepts the possibility of a woman cleansing herself from sin through penance, which is conceived by another instance, the religious institution; however, in q the speaker endorses the fact that the purification is insufficient to restore the woman’s integrity. These moves are assured by linguistic means: (i) the emphatic adverb bem indicates assent to the content of p, which conveys background information (that is, religious rules); (ii) the modal verb poder along with bem indicates that this is something which has not come to pass in the moment of enunciation; and (iii) mas indicates that the content of q is the most important and is in a contrast relation to the enunciation p. In this perspective, the emphatic adverb and the modal verb act together in assigning a lower discursive status to p, whereas mas (associated to negative polarity) ensures the centrality of q. In this case, the possibility is granted before being qualified as less relevant.

The data in (19) is part of the Colóquio de simples e drogas e cousas medicinais da Índia [‘Colloquium of simple medicines and medicinal things from India’], by the physician-botanist Garcia de Orta. The work is built as a dialog, in which Dr. Orta reveals his thoughts to Ruano, an imaginary listener. The concessive schema is performed in a quite similar way to that in (18), with a convergence of the roles of bem and the modal verb; the difference lies in the fact that p takes up a part of the immediately preceding enunciation (by Ruano); and also in the role of the modal, which is aimed here at mitigation of the assent. Again, the prevalent move is that of accepting what has previously been said by the other and endorsing a different position. A similar interpretation also seems to apply to (20), taken from the play Labyrintho de Creta, which is representative of the Portuguese comic theater from the 1700s.

In these correlative structures, such as (18)–(20), a fundamental change takes place: the assertive strength of bem is weakened. The adherence is weaker now. Additionally, the speaker is not the entity in charge of the asserted content anymore; (an)other enunciator(s) are included, so that there is polyphony. Polyphony is thus a trace shared by the source and bridging context, but it is performed in different ways. In source constructions, by using the emphatic adverb bem, the speaker emphasizes his position, and for that purpose, polyphonically raises potential objections with the intention of counteracting them. In correlative structures, conversely, the speaker emphasizes the point of someone else’s position, in order to weaken it for the sake of the position endorsed by him.

The occurrences of the periphrasis bem que in (21) to (23) are among the first to be documented in the 16^th and 17^th century. There, bem que is a grammatical word, standing at the left border of p, and the connection p and q falls under its scope; additionally, it provides instructions for a concessive interpretation.

In a possible interpretation for (21), the content of the concessive clause with bem que (although they exist in other parts of Malabar and in Jaoa) introduces a reservation towards the previously stated content (both exist in India and in greater amounts from Calcutta to Kannur), indicating that both types of cardamom at hand also exist in other parts of the Malabar coast and not only in the two mentioned cities. In (22) and (23), the concessive segments are parenthetical (prepositional phrase and adjective phrase, respectively) and also add reservations to the content of the main clause. These pieces of data already anticipate traces of concessive constructions with bem que which are frequent in the 18^th and 19^th century: express a restrictive nuance in 90% of the data; they can articulate clause and non-clause segments and, in case of clause segments, they co-occur with the subjunctive mode, which expresses argumentative irrelevance rather than eventuality or hypothetical status.

As previously said, the concessive with bem que is falling into disuse since the 20^th century; nowadays, mainly in informal genres, uses of bem que are common to express a desire in autonomous utterances, such as in (24).

The investigated data suggests that, while the concessive bem que recedes, another concessive periphrasis which is also formed by the adverb bem is conventionalized as a productive option for expressing factual concessivity in Portuguese. We are referring to se bem que, which emerges at a later time than bem que, according to Table 2, which presents the absolute frequency of bem que and se bem que from a longitudinal perspective (for a comparative study on the evolutions of bem que and se bem que, see Longhin, 2022):

The occurrences (25) to (27) provide usage examples of se bem que in modern Portuguese:

The data in (25) to (27), along with the other pieces of data gathered in the corpus, point out particularities of complex constructions with se bem que, such as possibility of realization in indicative and subjunctive mode ((25) and (26), respectively), the varying clause order, with a preference for postposition, and prevalence of the concessive meaning of restriction, which supports functionalist studies in various samples of Brazilian Portuguese (see Neves, 2011).

The constitution of se bem que puts into play the modification expressed by the adverb bem and the conditionality of se, another productive source of concessive connectives in languages (König 1988). The starting point appears to be the use of the conditional mold to express opposition, weakening the conditional implication and the hypothetical value of the protasis p in order to express contrast, as suggested by the data from (28) to (30), extracted from texts of the 17^th century. In these pieces of data, chronologically preceding the occurrences of se bem que, the concessive reading depends on specific contextual arrangements, such as, for instance, anteposition of the protasis, structural paralelisms with lexical opposites, bipolarity or emphatic correlations of the type bem… (muito) melhor (‘(much) better’).

In (28), the writer supports the point of view that the novices’ study is impaired due to the various duties and occupations which they have to fulfill, following the rule of the house where they live. By means of two parallel predicates which ser (‘to be’), the conditional se ponders favorable and unfavorable aspects. The presence of bem in p reinforces the writer’s adherence to observance, to submission to the house rules; however, in q, he argues that the rules hamper the study, reinforcing the initial argument. In (29), the construction has an explicit negative in the apodosis q. From the argumentative point of view, bem modifies the protasis, providing assertiveness, but the negation of the apodosis is the decisive argument. In (30), bem is part of an emphatic correlation with another adverb “bem… muito melhor…” (‘though… much better’), and both adverbs indicate positive values in a potential evaluation scale. In contrast to (28) and (29), in (30) bem is an intensifier which, on an evaluation scale, increases the importance of the content of the protasis, whereas melhor underscores the superiority of the content of the apodosis, establishing a concessive reasoning in which something important is considered to be insufficient. In all three cases, the irrelevance resulting from the protasis is taken as the less favorable argument, which afterwards becomes one of the traces codified in the semantics of se bem que.

Despite the presence of bem, the emergence of se bem que seems to deviate from the assumed cyclicity which I discuss in this text, given that the potential motivating contexts involve different uses of bem, and in these contexts conditionality is prevalent. However, the emergence of concessives with se bem que helps to elucidate, at least to a certain extent, the retreat of bem que. Both periphrases specialize in expressing restrictive concessive relations, thus concurring, but se bem que is more flexible than bem que, since it occurs with both the indicative and the subjunctive mode. In this respect, I relate to the explicative hypothesis of Montero Cartelle (2000), for whom the continuity of a connective in language has a correlation to more flexible constructions in terms of form and function, as seems to be the case of the history of concessive connectives in Spanish.

4.3. Embora: from a volitive adverb to a concessive subordinator

Ali (1931), a Brazilian philologue, links the concessive conjunction embora to the phrase em boa hora (‘in_Prep good_Adj time_{generic noun}’), which is used to wish success, a proper time for something, whether sincerely or out of courtesy, given the common sense that success of human acts depended on the time at which they were performed. The negative counterpart em má hora (‘in bad time’), to curse¹⁰, also existed. According to Ali (1931, p. 217), the semantic change was triggered when the adverb, in the context of optative clauses, began to indicate that an individual grants the possibility of a fact or, at least, does not object to it. As an example, Ali mentions the utterance “Ria embora quem quiser, que eu em meu siso estou” (‘Whoever wants may laugh as much as they want, because I am in my senses’), an excerpt from a theater play by Gil Vicente. In other words, embora is acting in the function of assent.

Two more research works on embora have provided important support. Guimarães (1987), from a pragmatic perspective, interpreted the change undergone by embora as a product of a difference of perspective between speakers. From the time category (proper time to perform something), the constructions would have slipped into the category of quality (something is good) and to the category of opposition. For Guimarães, the transition between quality and opposition would be caused by a displacement of argumentative positions. Lima (1997) focused on the semantic-pragmatic changes which led to the constitution of the concessive meaning from the perspective of grammaticalization. Based on sparse and dated data, he proposed stages of evolution, where initially a transition from time expression to approval would have occurred; and, afterwards, out of the implication between the ideas of approval and concession (i.e., I approve implies I do not oppose), the contrastive meaning would have emerged. For the author, the concession begins to be conveyed at the pragmatic level, by conversational implication, and only later, at the semantic level.

In the present study, the investigation in texts from Old Portuguese yielded scarce data for the meaning of good fortune or cursing, of which (31) and (32) are representative. However, there is data in which em boa hora is linked to the expression of a positive evaluation of the state of affairs, such as (33). In these cases, em boa hora is often followed by a reason, which can be taken as the speaker’s perspective superseding the common sense view.

The data mapping of embora in the corpus of Middle to Modern Portuguese has evidenced a variety of uses, which possibly are involved in more than one path of change. The uses which seem related to the emergence of the concessive meaning are displayed in Table 3.¹¹ The frequencies point towards important facts. The use of the adverb embora as an assent marker is always more frequent than that of success and prevails until the second half of the 18^th century, a period in which the first concessive uses have been identified. The concessive conjunction has been documented from the 19^th century on and its frequency is high in texts from the 20^th/21^st century. In the expression of concessivity, embora oscillates between the categories of adverb and conjunction.

In the following I shall discuss the contexts in which embora expresses assent, the contexts which display double compatibility with assent and concession, and the contexts with a concessive value.

The data from (34)–(37) provides instances of embora in the role of assent marker. The contexts are marked by conflict situations, in which the acceptance expressed by embora, which can be paraphrased as yes, reinforces an unexpected or unwanted situation or fact.

In (34), an excerpt of reported speech, the imprisoned warrior emphatically claims with pride and courage during the ritual in which he is to be killed that they should kill him, and justifies this with the argument that he has done the same to relatives and friends. A contiguous explanation is also found in (37), in direct speech. In this dialog, in view of the impending risk of losing respect, the individual does not oppose and qualifies it as being irrelevant: ‘let it be lost, it doesn’t matter, because little is lost’. In (36) the acceptance of an unlikely fact, ‘cutting body parts’, is presented, yet under the condition that obedience is kept (‘as long as obedience remains whole’). In (35), unique data is found: there is a lexical cue, the verb conceder (‘grant’). Here, as well, the conflict context is prominent, since it is explicitly granted that the evil spirit expelled from a body may remain in the world. In sum, these are contrastive contexts in which embora indicates little relevance or non-opposition.

In (38) and (39), below, embora is part of the concessive schema in correlative constructions which are similar to those discussed for licet and bem. In the complex construction, there is a recurrent move whereby the first member p presents emphatic acceptance (with embora) of an available content; and the second member q, headed by a contrastive connective, provides the focal information. Again, the concessive reading emerges pragmatically from a particular combination between p and q. In Table 3, this data is computed as assent adverbs.

The occurrence in (38), taken from Father Antonio Vieiras’s writings, is part of a dialog in which a dispute without final agreement takes place about the letters of the alphabet. At a given point, speaker C accepts the arguments of A, as an act of courtesy, then rejects them. (39) is taken from a theater play; it presents a scene in which a conflict is installed when the character Simic trespasses DL’s property using a rope and is caught. DL reacts and uses the same rope to arrest the trespasser, who resists and begs not be arrested. The utterance with concessive value is produced by Simic as a reply to DL’s question about the reason for trespassing. Simic emphatically accepts the prospect of DL arresting him, which is surprising in view of the previous context, but is understandable considering the pressure to confess the reason of the crime. In other words, both situations are negative for Simic, but he assigns different statuses to them, ‘better to be arrested than to confess’. At the end, in the last utterance, another contextual cue confirms the conflict scenario: the speaker uses a cleft sentence, a focal strategy which triggers contrast by exclusivity, and thereby highlights the centrality of the content of the second member (Por iſſo he que eu te aperto, ‘That is why press you’, in which that anaphorically retrieves the question about the motive of the crime).

Data such as (38) and (39), which is compatible with the concessive reading, features the following characteristics: embora is always in a post-verbal position and is frequently intensified by muito, which shows its adverbial character; the verb occurs in the subjunctive or imperative mode; embora in most cases is amalgamated, equaling a single morphological form. These characteristics suggest that the change in syntactic category occurs at a later point than the semantic change.

As a conjunction, embora expresses a factual concessive relation; in particular, the nuances of negated causality and restriction. The constructions may include articulation of clause segments such as (40), and non-clause segments, as in (41), a type of parenthesis which is prosodically independent from the nuclear clause. The concessive segment, whether a clause or not, has a variable position, meaning that it can be anteposed, interposed or postposed. The construction also displays variability in the tense-mode selection, yet the subjunctive tenses prevail. Cases such as (42), which retain traits of the adverbial source, as suggested by their non-initial position, have been computed as concessive adverbs.

4.4. Discussion

The data has indicated a quite similar correlative structure for the three phenomena, which enables similar types of inference directed towards concession, based on an old argumentative schema between what is apparently accepted (out of irony, attenuation, kindness or credibility) and what is endorsed. In this structure, the assent marker is a decisive element in the construction of the concessive relation: thereby, the speaker approves and reinforces an argument which is available in the co-text or the universe of shared human knowledge, weakening this argument for the sake of another; the factuality required in concession is ensured; and incompatibility is created.

The histories of Latin and Portuguese show, with a distance in time spanning centuries, that different linguistic elements – licet, bem, embora – have been recruited for marking assent in the correlative structure, (re)activating an argumentative schema which dates back to classical rhetoric, in the figure of concessio. Thus, licet, bem and embora participated in the concessive schema and served as a systematic resource for this reasoning, promoting inferential enrichments based on the conceptual affinity between assent and concession.

Obviously, this is not exclusively a phenomenon of Romance languages; the emphatic assertion of truth, approving or agreeing to something, is acknowledged as being one of the cross-linguistic historical sources for concessive relations (König, 1988).¹² In any case, it is significant that this is a recurrent channel of change in Latin and in Portuguese. Furthermore, achieving some understanding of change phenomena which recurrently have expanded the concessivity system of one or more languages can enable the analysts to make predictions about future paths of change within this domain of meaning.

Rossari (2014, p. 22) concludes, from a diachronic perspective, that for certes, in French, no semantic change has occurred, since “their concessive use is not the result of an evolution in their meaning, but the result of their integration in one peculiar discourse configuration”. The case of the constructions examined here seems to be different. The data suggests that the use of assent markers in the correlative structure led to an initial change, the weakening of the assertive potential. Afterwards, as this usage was repeated, the concessive value, which previously depended on the particular arrangement of the parts p and q, is internalized as a conventionalized trait in those items themselves, which gradually take on other connection traits, such as, for instance, the increasing establishment of the position at the left margin, the widened scope over the relation between states of affairs and the instructional marking.

One last aspect to be highlighted refers to the main research question in historical semantics and pragmatics, which is the interpretation validity of linguistic data produced in other times, in genres which most likely have already changed. In fact, this is a challenge for historical sciences. The interpretations presented herein are the result of the analyst’s work, being explicative hypotheses, which gain plausibility when one makes use of methodological resources such as: the identification of distributional traits which correlate to the different readings; the consideration of the text as the unit of analysis, given that the access to the full texts of the sample and not just to the excerpt obtained by a computer tool is essential to understand communicative purposes, argumentative positions and relations between the subjects. The data reinforces the importance of the interaction contexts in grasping cues of change and in the manifestation of concessive relations.