Null and overt se constructions in Brazilian Portuguese and the network of se constructions

1 Introduction

Portuguese se constructions have a constructional counterpart in which the clitic is absent. The null clitic construction, observed in all se constructions – namely reflexive, reciprocal, middle, anticausative, passive, and impersonal constructions – is more frequently used in Brazilian Portuguese (BP) than in European Portuguese (EP) and is mostly used in the informal register. Null se constructions are usually interpreted in essentially morphosyntactic terms, as the result of an ongoing general tendency in BP towards the morphological loss of clitics (Cyrino 2007; Galves 2001). Soares da Silva et al. (2021) showed empirically, however, that the alternation between overt and null se constructions is motivated by semantic and lectal factors.

Taking as the point of departure Soares da Silva et al.’s (2021) corpus and their profile-based and sociocognitive analysis of overt-null se variation, the article discusses why the null se construction needs to be included in the taxonomic network of se constructions and hypothesizes where in the taxonomy it is placed, addressing specifically the extent to which the overt and null se constructions are allostructions.

2 The network of middle constructions

Se constructions in Portuguese are a set of polysemous marked constructions that produce valency-related effects, generally referred to in the literature as middle voice (Kemmer 1993; Maldonado 1999; among others). Portuguese is a one-form language, displaying one single middle marker to encode all situation types or semantic domains, that is, all semantic and pragmatic contexts associated with middle-marked constructions (Kemmer 1993: 7), albeit with optional additional markers to encode reflexivity (a si próprio/mesmo ‘to oneself’) and reciprocity (um ao outro ‘one another’). In Portuguese, the relevant semantic domains, as identified by Kemmer (1993), are those provided in Table 1.

In this article, we follow the classification adopted in Soares da Silva et al. (2021), unless we are discussing Kemmer’s model. In our classification, the situation types in Table 1 have been rearranged as reflexive, reciprocal, middle, anticausative, passive, and impersonal. The middle is then divided into semantic subclasses: cognition, emotion, perception, union, denomination, and bodily action. The less frequent situation types in the corpus analyzed in Soares da Silva et al. (2021) were grouped as “other”.

Kemmer (1993: 208) proposes the relative elaboration of events as the overarching semantic property of all se constructions, defined as “the degree to which the facets in a particular situation, i.e., the participants and conceivable component subevents in the situation, are distinguished”. This semantic property interacts with transitivity (considered to be prototypical two-participant and one-participant events). The middle marker, therefore, in relation to some situation types, performs a valency changing function (Geniušienė 1987; Kemmer 1993), specifically that of an intransitivizer (Givón 2001: 116).

Figure 1 shows Kemmer’s (1993) network of middle constructions in form of a conceptual map. Based on diachronic evidence, a direct semantic connection between nodes linked by straight lines is established. The closer the nodes are physically, the closer the semantic connection is. The relative distinguishability of participants correlates with the situation types placed vertically at the center of the semantic map (including grooming) between prototypical two-participant and one-participant active events, but it does not correlate with all situation types. For situation types such as the passive middle, the Agent and the Patient are two very distinct entities, although the Agent is suppressed. Even disregarding the relative distinguishability of participants, the passive event is less elaborated, as it corresponds to an alternative conceptualization of a particular event, i.e. profiling the endpoint of the event.

Figure 1:

Semantic relations among middle and other situation types with the Portuguese se constructions plotted onto the conceptual map, indicated through the gray shading. Based on Kemmer (1993: 202). ©John Benjamins. Reprinted with permission.

The low elaboration of events associated with se constructions, resulting in intransitivization, is readily observable in those situation types that offer the possibility of alternative conceptualizations, that is, the oppositional middles. Oppositional middles in Portuguese are the constructions associated with passive, impersonal, anticausative, reflexive and reciprocal, grooming (e.g. pentear alguém vs. pentear-se ‘comb someone’ vs. ‘comb oneself’), emotional and cognitive middles (e.g. preocupar alguém vs. preocupar-se ‘worry (someone)’ vs. ‘worry’), body posture (e.g. levantar alguém vs. levantar-se ‘pick someone up’ vs. ‘get up’), and translational and non-translational motion (e.g. aproximar algo vs. aproximar-se ‘move something closer’ vs. ‘move closer’, virar algo vs. virar-se ‘turn something’ vs. ‘turn around’). The gray area in Figure 1 corresponds to the se constructions plotted onto the semantic map of the middle domain.

Inglese (2022) raises the question whether the middle marker also encodes intransitivization in the case of non-oppositional middles, that is, in those situations in which the middle marker shows a systematic relationship with classes of verbs. Given that the middle marker does not exhibit an obvious valency-related function, Inglese (2022: 496) concludes that oppositional and non-oppositional middle constructions should be approached separately. In Portuguese, non-oppositional middles are associated with the following situation types: cognition and emotional middles (arrepender-se ‘regret’, ensimesmar-se ‘be self-absorbed’); behavior middles (comportar-se ‘behave’); perception middles (deparar-se com ‘come across’); relational middles (queixar-se ‘complain’); body posture (ajoelhar-se ‘kneel’, espreguiçar-se ‘stretch’); and translational and non-translational motion (escapulir-se ‘slip away’, voltar-se ‘turn around’). To build a conceptual map for non-alternating middles is more complex, as these are related to the lexicon, and often language specific. It is a typological challenge to achieve generalizable results (Haspelmath 2003: 224).

The variation between overt and null constructions that, in Soares da Silva el al. (2021), is observed for BP for all types of se constructions (see Section 3) – that is, with both oppositional and non-oppositional middles – raises the question of the place of null constructions in a taxonomic network, as we will discuss in Section 4.

3 The overt-null se alternation

The data used in this study were extracted from three informal sub-corpora of BP: (i) C-Oral (263,000 words), which consists of spontaneous oral language transcripts; (ii) Museu da Pessoa (1,182,544 words, referred to here as “Pessoa”) which consists of interview transcripts about life stories; and (iii) Fóruns (263,772 words), which consists of manually compiled message board posts showcasing written informal language. The data set includes 1,313 occurrences of se constructions: 514 overt se constructions and 799 null se constructions. All collected occurrences of overt and null se constructions in BP were manually annotated according to semantic, syntactic, and pragmatic factors (see Soares da Silva et al. [2021] for details on the factors).

The overt-null alternation is observed in all construction types, but, in each case, the event is construed in a different way if it is encoded by the overt or by the null construction. Examples (1) and (2) denote the same reciprocal event of getting married, but the conceptualization of this event in (1) is different from that in (2).

In (1), the clitic encodes the crucial moment or condition of change of state and, consequently, the energetic event of getting married profiling the agency and the affectedness of the participants. On the contrary, the absence of the clitic in (2), which is marked with Ø, correlates with an absolute, non-energetic construal of the reciprocal act of marriage, viewing this event as a whole and as an object of the speaker’s conceptualization. Importantly, there is independent evidence to argue for energetic versus absolute construal (Langacker 1991: 389–393) as an explanatory conceptual criterion for the presence versus absence of the clitic se. In (1), the presence of dynamic adverbs (aí … aí ‘then … then’) and adverbial clauses (quando menstrua ‘when she menstruates’), as well as sequences of dynamic events (eles vão e ‘they go and’), are independent markers that profile the moment of change, putting forward an energetic construal, encoded by the overt se construction.

In examples (3) and (4), the emotional reaction verb preocupar-se ‘worry’ equally occurs in an overt and in a null middle construction.

The overt construction in (3) profiles the Experiencer’s shift of emotional state, and marks the force-dynamic psychological experience. This reading is reinforced by the several markers of change focus: the temporal dynamic adverb agora ‘now’, the epistemic adverb mesmo ‘really’, the focus marker é que ‘it is … that’, the aspectual periphrasis ir preocupar ‘going to get worried’, as well as dynamic verbal expressions (foi, olhou para mim ‘turned, looked at me’). In contrast, the clitic deletion in (4) defocuses the force-dynamic psychological experience and the energetic change and profiles the new resulting emotional state, that is, that ‘you are worried’. This interpretation is supported by the absence of the marker focusing on the change of emotional reaction in (4).

Examples (5) and (6) illustrate the alternation in the impersonal construction.

In (5), the profiled moment of change is encoded both by the clitic se and by independent markers, namely the temporal (future) verbal periphrasis vai se chegar and the aspectual verbal periphrasis está se chegando, which indicates a change in progress. In contrast, the event in (6) is construed as non-energetic. As the energy source is not profiled, there is as a result no marker of change focus. Additionally, (6) includes the interrogative adverbial como ‘how’ (and the imperfect past tense) indicating a routine procedure, marking the event as impersonal.

Figure 2 shows the conditional inference tree for the full BP data set (1,313 tokens, combining all se constructions).^[1] The most important variable overall is “change focus”, that is, the profiling of the moment of the change of state and its markers, such as the markers illustrated in examples (1), (3), and (5). The tree splits the “no change focus marker” off from “change focus marker”. It shows the importance of the “change focus” semantic factor as a predictor of the overt-null constructional variation in all se constructions: tokens with a focus on change prefer the overt se construction while tokens without a focus on change are associated with the null se construction.

Figure 2:

Conditional inference tree for the full data set (Soares da Silva et al. 2021: 169).

The overt-null se alternation in BP is, therefore, mainly determined by semantic differences of construal, specifically the cognitive grammar (Langacker 1991, 2008) distinction between energetic or force-dynamic and non-energetic or absolute construals of events. The overt se construction typically encodes an energetic construal in which the focus is on the pivotal moment of the force-dynamic, unexpected change of state. When the moment of change is not profiled, the null se construction is typically produced: it encodes an absolute construal, profiling a self-contained whole, a single participant thematic process conceptualized autonomously, without explicitly invoking a force-dynamic interaction.

Other semantic factors, such as verbal semantic class (node 12), lexical aspect (nodes 13 and 3), and semantic role of the NP (node 6), play a role, but speakers’ choices are now somewhat constrained to specific contexts. Consequently, they appear low in the tree in Figure 2.^[2] The lectal factor of register in some specific se constructions is another factor for the Brazilian speakers’ choice between overt and null se constructions: it is an important factor for the anticausative construction (see Figure 3), and the only predictor in passive construction (Figure 4).

Figure 3:

Conditional inference tree for anticausative constructions in the data set (Soares da Silva et al. 2021: 180).

Figure 4:

Conditional inference tree for passive constructions in the data set (Soares da Silva et al. 2021: 187).

In summary, two domains from which overt-null se alternation factors may emerge can be pointed out. The first and most important domain is the distinction between energetic or force-dynamic and non-energetic or absolute construals of events. Specifically, when Brazilian speakers, in informal contexts, want to construe a reflexive, reciprocal, middle, anticausative, passive, or impersonal event as profiling an energetic, force-dynamic interaction and the corresponding pivotal moment of the change of state, they are more likely to use the clitic se. On the other hand, the absence of the clitic functions as the alternative for a non-energetic, absolute, and objective construal of the event, detached from the energetic elements and from the conceptualizer and focused on the resulting state.

The second domain of origin for the overt-null se alteration is the language-external, lectal context in two dimensions. First, there is a geographic factor: the overt-null se alternation is productive in BP in all types of se constructions, but restricted (to anticausative and some middle se constructions) in EP. Second, there is a stylistic factor: there is a tendency in BP for the null construction to be produced mostly in informal and spontaneous register. Thus, the overt-null se alternation shows a pattern of constructional divergence in the two national varieties of Portuguese, with BP showing a restructuring of the network of se constructions.

Overall, the key aspect to consider is that a difference in construal is what determines the alternation between overt and null se constructions. This means that the overt-null se alternation consists of two alternate constructions that both present a choice point for an individual Brazilian language user and exhibit a systematic difference of meaning in expressing the same reflexive, reciprocal, middle, anticausative, passive, or impersonal event. Hence, the two alternating constructions are interchangeable in the sense that they express the same referential event, but necessarily involve semantic differences in construal. Lectal factors also play a role, but a less prominent one, in determining the alternation.

4 Overt-null se alternation and the network of se constructions

All reflexive, reciprocal, middle, anticausative, passive, and impersonal situation types are encoded by overt and null se constructions, as discussed in Section 3. The results from Soares da Silva et al. (2021) showing a systematic difference in construal in the vast majority of the null se constructions types^[3] in nonstandard BP justifies the inclusion of the null variants in the network of se constructions.^[4]

Two aspects need to be discussed. The first concerns the potential impact that the alternation may have on the semantic map of se constructions. We argue that, given that there is no split between the marked (overt marking) and unmarked (null clitic) se constructions in relation to the different situation types, both the overt and the null marking of se constructions occupy the same region in the conceptual map, as Figure 5 shows.

Figure 5:

Semantic relations among middle and other situation types with the Portuguese overt and null se constructions plotted onto the conceptual map, indicated through the gray shading. Based on Kemmer (1993: 202). ©John Benjamins. Reprinted with permission.

The second aspect relates to the taxonomy of se constructions. Both overt and null se constructions should be included in the taxonomic network, but the question is where in the network (cf. Pijpops 2019, 2020) – that is, at what level of abstraction – the alternation is observed.

We argue that the overt and null se constructions are allostructions, as defined by Capelle (2006) and elaborated by Perek (2015): a taxonomic relation with an overarching construction from which properties (both in terms of form and function or meaning) are inherited and shared by the allostructions. It may appear, at first glance, that the variation will be found at all levels of abstraction, given that the change of form (null clitic vs. overt clitic) is associated with a systematic change of construal (non-energetic construal vs. energetic construal). At the level of concrete instances, the form and meaning distinctions are observed, often with the same verb co-occurring with the null and overt se constructions (see examples in Section 3). The same form-meaning distinction is observed at the immediate superordinate level of abstraction, as Figure 6 shows for middle constructions.

Figure 6:

Alternation at lower levels of abstraction for examples (3) and (4).

The specific constructions in Figure 6 are instantiations of the allostructions [SUBJ SE Middle V (PP)] and [SUBJ Middle V (PP)], inheriting both form and function. The allostructions inherit form and function from the more abstract constructeme (Capelle 2006) at the top, which is more underspecified for form. In the corpus, the majority of the middle verbs occurring with the null se construction are verbs of cognition (e.g. preocupar(-se) ‘worry’, lembrar(-se) ‘remember’, esquecer(-se) ‘forget’), followed by verbs of union, denomination, and bodily action (Soares da Silva et al. 2021). In relation to the latter three, the association relates to the very high frequency of specific verbs in the corpus. For verbs of union, the most pervasive verbs are casar(-se) ‘get married’ and separar(-se) ‘separate’. Chamar(-se) ‘be called’ (i.e. entities’ proper names) is the only denomination verb co-occurring with the null se construction, and body posture verbs such as levantar(-se) ‘get up’, sentar(-se) ‘sit down’, and deitar(-se) ‘lie down’ are the examples of bodily action verbs that most frequently co-occur with the null se construction. Figure 7 shows the partial network of constructions focused on the middle types.

Figure 7:

Partial network of middle constructions for oppositional middles.

In relation to the other types of se constructions, those also exhibit the null-overt alternation at the same level of abstraction as the middle alternation, that is, at the level of the constructeme and lower levels of abstraction. At the highest level of abstraction is the hyper-schematic se construction [V(-SE)], from which the different types of se constructions inherit their form and meaning.

There is no question that, formally, the different construction types inherit their form from the more schematic superordinate constructions, but the semantic relatedness and inheritance are more difficult to grasp. As became apparent in Section 2, the relative elaboration of events and related intransitivization presents shortcomings as the semantic characteristic of all situation types, as non-oppositional middles, are less obviously marked for intransitivization, which has led scholars such as Haspelmath (1995: 373) to hypothesize that there may not be “a real common meaning that all situation types share”. Following Inglese (2022), the meaning of the hyper-schematic construction [V(-SE)], which is related to the positioning of se constructions within the transitivity continuum, should be posited in relation to oppositional middles that exhibit a relative reduction of transitivity. Therefore, we separate the network into oppositional se constructions and non-oppositional se constructions. Given that the latter are associated with specific verbs, nodes in the network with lexically filled verbs need to be added to the network.

Figure 8 shows the network of se constructions with the highest level of abstraction. The allostructions cannot be posited at the highest level of abstraction, as the difference between the variants is related to propositional meaning associated with the events (i.e. of the reflexive event, reciprocal, middle, etc.), rather than with intransitivization. In other words, the energetic and non-energetic construals are not instantiations of the relative reduction of transitivity, but are, instead, related to each particular construction type, because they impose a particular perspective on each event. As Perek (2015: 153) states, “the constructemes capture the level at which constructions are semantically equivalent and the allostructions specify exactly how these constructions differ”. The proposed network encodes the premise that the conceptual differences observed between variants do not compromise the referential or propositional meaning (e.g. reflexivity, passivization, impersonalization, etc.).

Figure 8:

Partial network of oppositional and non-oppositional se constructions.

Despite the separation between oppositional and non-oppositional middles in the taxonomy, we postulate that the non-oppositional middles must, nonetheless, enter the taxonomy of se constructions. The few situation types in the middle category that are non-oppositional are instances, with lexically filled verbs, of cognition and emotional middles, behavior middles, perception middles, relational middles, and verbs of body posture or translational and non-translational motion (see examples in Section 2). Partial inheritance links between non-oppositional middles and a superordinate, hyper-schematic se construction at the top of the hierarchy are established, which means that specific instances of a category can be placed in the network as exceptions (Goldberg 1995). The marker se, at the top of the hierarchy, assumes an unspecified semantic value that is instantiated at the level immediately below as either an intransitivizer (left branch) or lexically determined (right branch).

Besides the taxonomic relationship between the constructions – that is, at different levels of abstraction – the allostructions also hold horizontal relationships within the network (Diessel 2019). These horizontal relationships capture the similarity and contrast in terms of form and meaning between constructions at the same level of abstraction. In other words, the overt se construction and the null se construction are not only instantiations of a superordinate construction, but they also exhibit contrast links within each type of construction. Furthermore, a horizontal relationship is also posited between the different null se constructions for each construction type. As an example, the middle null se construction variant is not only related semantically to the middle overt se construction, due to the differences in construal, but it is also related, given the similarity of form and, to a certain extent, function (i.e. absolute construal), to other null se construction variants in the network, which are at the same level of abstraction. This is also the case for the overt se constructions. The functional similarity, however, is constrained by the type of event underlying the variants, as similarity links are more plausibly established between certain types of constructions, such as some reflexives and emotional middles, like focar(-se) ‘focus oneself’/‘get focused’ and preocupar(-se) ‘worry oneself’/‘become worried’; middles with union verbs and anticausatives, like dividir(-se) ‘become divided’/‘get divided’ and integrar(-se) ‘integrate’/‘become integrated’; and impersonal and passive events, like ainda se passava muito aquela ideia romântica trabalhada ‘that well-crafted romanticized idea was still being put across’/‘people were still putting across that well-crafted romanticized idea’ (Pessoa) and lá não pode tirar foto não ‘photos are not allowed there’/‘one cannot take photos there’ (C-Oral).

Finally, this study only considered the alternation based on form, that is, the formal realizations of the se construction. The choice of null and overt se constructions as alternations was a methodological choice taken by the researchers, but it is by no means the only possible alternation, if the point of departure is not the constructional form but the function. For certain se construction types, other types of horizontal links are established with other constructions that perform similar functions but which are formally distinct. For example, the impersonal se construction establishes horizontal links with other formally distinct constructions such as the impersonal use of personal pronouns and the periphrastic passive construction, among others (Afonso 2008). The motivating factors and their underlying causes for preferring a particular impersonal construction over others have not been studied, but Soares da Silva and Afonso (2022) hypothesize that any semantic differences between the constructions reside in the degree of impersonalization, with the impersonal null se construction performing the maximal degree of impersonalization, given the deprofiling of the energy source.

5 Conclusions

This article has discussed the impact of the null versus overt variation of the se constructions in BP for the network of se constructions. The results of the usage-based study by Soares da Silva et al. (2021) showed that, contrary to previous studies that decoupled the variation and semantic factors, one semantic factor is the main predictor for the variation for the vast majority of construction types, namely the energetic versus non-energetic construal of the event. When the event is construed as energetic, the moment of change being profiled, the overt se construction is preferred. On the other hand, when the event is construed as non-energetic or absolute, the null se construction tends to be produced. Lectal factors also interact with the choice of variants; the systematic semantic difference between them occurs in BP and in more spontaneous, unguarded speech, or oralizing registers such as internet forums.

The driving factors for the variation and the mechanisms that facilitate the predictors to operate determine the configuration of the network of se constructions. Together they form compelling evidence that the null se constructions must be present in the network. Considering the conceptual map of middle constructions, both overt and null se constructions can be plotted onto the same region in the map. On the other hand, in relation to the taxonomic network of se constructions, we propose that the alternation is an allostruction which is positioned at relatively lower levels of schematicity, that is, at the level of the constructeme and below. For each event type, we posit a constructeme whose subordinate constructions constitute the variants. The overt and null se constructional variants inherit the propositional or referential meaning from the constructeme and instantiate the form (which is partially underspecified in the constructeme). The constructions at the subordinate levels, therefore, inherit the intransitivizing function of the middle marker. However, as we have discussed, following Inglese (2022), the marker does not seem to function as such in non-oppositional types. If the latter are to be included in the network, as we argue, we propose that a partial link is established in relation to the top node [V(-SE)] for which the marker is unspecified. In the immediate subordinate constructions, it is either specified as an intransitivizer or is lexically determined (cf. the syntax-lexicon continuum).

Within the same taxonomic network, and following Diessel (2019) and Capelle (2006), horizontal links between the allostructions within the same level of abstraction are posited, not only considering the functional contrast between the allostructions, but also considering the functional similarity across the network.