2.1. Semantic incorporation in Brazilian Portuguese (and beyond)

Research from the last decades has shown that contrary to other Romance and Germanic article languages, Brazilian Portuguese allows for bare singular nominal phrases in a wide range of configurations, including the subject position of episodic sentences, as in (3), sometimes even with definite/specific interpretations, as in (4). For a detailed discussion of the literature and proposals for alternative analyses to what is presented below, see Wall (2017, chapter 2).

In many accounts, ignoring the somewhat special definite/specific interpretations such as in (4), bareness has been related in some way to number neutrality, analyzing BP bare nominal phrases as mass or kind-denoting expressions (Müller, 2002; Müller & Oliveira, 2004; Müller & Paraguassu-Martins, 2007; Paraguassu-Martins, 2010; Pires de Oliveira & Rothstein, 2011; Pires de Oliveira & Rothstein, 2013). As mentioned above, there is no consensus in the literature on the theoretical status of bare nominals in BP, but it would go beyond the scope of this paper to provide a thorough discussion as the one provided in Wall (2017). For the present purpose, number neutrality is assumed because it fits naturally into the Carlsonian incorporation proposal for weak indefinites.

The strong connection of bare nominal phrases to number neutrality has been corroborated by research on incorporation, where it has been shown that many article languages show semantic incorporation of a verb and an adjacent direct object when this occurs bare, whereas the presence of a definite or indefinite article usually blocks such readings (Klein et al., 2013; Stvan, 2009; Taveira da Cruz, 2008; see Wall, 2017, pp. 85–87, for an overview of the literature on Brazilian Portuguese).²

The semantically incorporated direct objects in (5a–7a) all contrast with the non-incorporated direct objects in (5b–7b) in not allowing for a referential interpretation and in only being interpreted existentially, not as generics. They are also number-neutral and therefore do not necessarily receive a singular interpretation.³ When the event or the object nominal is interpreted as singular, this occurs due to context. In a contemporary ordinary discourse, for example, (6a) would most likely be read as meaning that the stated person is wearing one (unique) skirt today. However, the sentence would also be perfectly compatible with a situation where the speaker wants to highlight that all of the several outfits that the mentioned person is wearing today include a (possibly different) skirt, in contrast to (6b), which is about a very specific one. In 2.3, a series of overtly indefinite objects will be introduced that have very similar properties. Before that, it is important to mention that there are other overtly articulated objects that have been analyzed as semantically incorporated.

2.2. Weak definites as incorporated structures

While cross-linguistic evidence for bare nominal semantic incorporation is quite robust, a special case has been highlighted, so-called weak definite nominal phrases. In these, overtly definite singular objects receive readings that resemble incorporated objects despite the presence of a definite article (Aguilar-Guevara, 2014, pp. 24–26; Aguilar-Guevara, Le Bruyn & Zwarts, 2014, among others). In (8), object NPs are not referring to a uniquely defined instance of the concept expressed by the noun in the context of the utterance.

Going into the details of the properties of these special kinds of definites is not possible within the scope of this paper. While they all share certain properties, there are also cross-linguistic differences that have been discussed in the literature (see Leonetti, 2019, and references therein). One important trait is that only certain verb-noun combinations receive these interpretations. ‘The newspaper’ in read the newspaper can be interpreted as a weak definite but ‘the book’ in read the book not. These readings are not restricted to object positions; they can also appear as subjects or within prepositional phrases. It is vital to highlight that weak definites do not refer to a specific entity as objects for the purposes of this study. Klein et al. (2013) propose that they are interpreted as part of the event expressed by the verb. Sentence (8d) expresses the plan to go to the bar for ‘coffee-drinking.’ The meaning of (8c) is not that Lola happens to be in a prison for some random reason (visiting, working, etc.) but that she is imprisoned. Thus, the verb-noun combination gives rise to an enriched meaning. In (8a) and (8b), the bus and the newspaper may be or even typically are different ones every day, the verb-noun combination in these cases describes some sort of routinized activity. This might be part of the explanation why read the book does not receive weak definite readings. Klein et al. (2013) do not provide a detailed semantics for weak definites but rather sketch what such an analysis might look like within the Carlsonian incorporation approach. They propose that the interpretation of weak and regular definites differs as shown in (9).

According to this analysis, (9a) and (9b) are syntactically identical but (9b) has a different semantic structure. Carlson et al. (2013, p. 28), elaborating upon Klein et al. (2013) points out that weak indefinites could be interpreted along the same lines, giving both incorporated interpretations the compositional structure ARTICLE(V(N)), where definite and indefinite article contribute their usual semantics, however not by combining with the nominal but with the whole VP. This could be a starting point for a unified treatment of weak definite and indefinite incorporation and it is especially relevant to the Brazilian Portuguese data discussed in this paper, mainly for two reasons. The first and more obvious one is that in this incorporation account, weak indefinites are actually expected (Klein et al., 2013, p. 201), which is not the case in alternative accounts for weak definites. Thus, prima facie, the BP data fits most naturally into this account. The second reason is that, although not discussed at length in the present paper, the incorporation analysis also straightforwardly allows for a unified treatment of similar effects in more and less idiomatic V+N combinations. Gehrke and McNally (2019) provide an explicit account for determiner variation and modification in idioms, building on the ideas in Carlson (2003), among others. In their account, there is a compositional step for the content semantics of verb and noun ‘before’ the semantic contribution of the determiner is calculated. This point is crucial because in BP, highly idiomatic V+N expressions may have weak indefinite versions with similar effects on interpretation. The details of this problem must be left open for future investigation since this would go beyond the rather empirical focus of the present paper. For relevant examples of weak indefiniteness in idioms, however, see (17b) below and the stimulus materials in the Appendix, Section 1.

2.3. Number-neutral indefinites

In the following, I will introduce a phenomenon that gives rise to weak and number-neutral interpretations in Brazilian Portuguese similar to many of the examples given above, but where the object nominal phrase is headed by an indefinite article. To the best of my knowledge, this phenomenon has not yet been described in the literature (but see Section 4.2 for discussion of possibly related constructions). Similar to weak definites, these number-neutral indefinite nominal phrases show a certain degree of idiomaticity or special selectional restrictions between nominal and verb, but they are not completely fossilized idiomatic expressions. Although the number of licensed nominals is generally limited, there is a certain range of systematic paradigmatic variability. Furthermore, some of the examples include idiomatic expressions that are generally constructed with a bare object. There, the addition of an indefinite article seems to have the same effect as in less idiomatic uses.

The following paragraphs provide a first (and possibly superficial) systematization of the examples observed thus far. Some of these V+N combinations are given in (10)–(12). The examples in (a) give the original attested quotation with an approximate translation. Examples (b) and subsequent ones isolate the relevant V+N combination and provide word-to-word glossing and translation.

Example (10) is particularly clear since the sentence only makes sense when interpreted analogously to assistir televisão (‘TV-watching’), there is no interpretation available where the indefinite article would be interpreted as quantifying the described event. It could be argued that some kind of vague event quantification might be involved, but this issue requires further study (cf. discussion in Section 4.1). However, for talking about units of TV-watching or of individual programs, other expressions must be used. In the following, reference to quantification is made in this stronger, non-vague sense. The only available interpretation here is that the insensible neighbor threatens any time span and event of TV-watching.

As in (10), assistir uma televisão in (11) is not interpreted as quantified. A series of activities are proposed, one of which is TV-watching; another one is playing videogames. Regarding the last activity, the literal meaning of eating just one stuffed zucchini is possible in principle. However, it would be odd to prepare only one zucchini for a meal when several persons are involved, and stuffed zucchini are not served as entire vegetables but cut into halves and only coincidentally one might get both pieces of one previously entire zucchini. So, the natural interpretation, also in line with the other proposed activities, is simply zucchini-eating, without any specification of the number of individual vegetables.

In (12), several number-neutral indefinite objects occur, some of them showing an elevated degree of idiomaticity. Thus, pegar uma boa piscina is not interpreted literally, namely that people come to pick out a good pool, which does not make sense in this context. Rather, the interpretation is that of pegar piscina ‘stay at/in some pool’. The expression bater uma pelada is even more opaque and also exists in the form bater pelada, both expressing the activity of playing football, without necessarily implying a proper or unique complete game. The same holds for the following activities where it is not said that people come to the club to go to only one sauna or play only one game of basketball, snooker, or bowling, but just that people practice these activities there. Thus, in examples (10)–(12), the object nominal phrases um(a) N never denote a quantified event or entity. While in some cases the literal interpretation does not make any sense at all, the activity may involve, or, for some verb+noun combinations even typically involve one entity, as for instance in the case of taking a sauna.

Furthermore, if contrasts like bater pelada vs. bater uma pelada were presented to some speakers, it is not inconceivable that they might have the intuition that the first case denotes the activity while the second one is a single match. Despite these observations, it will be claimed in this study that individual unit or entity interpretations arise because of the context (in a broad sense, possibly including world knowledge). The indefinite article may support or emphasize such interpretations, but it does not derive them. This can be shown with nouns denoting individual concepts in the sense of Löbner (2011). Such individual concepts like sun, moon, or pope are normally used with the definite article, whereas the use of the indefinite article is “incongruent” in Löbner’s terminology and gives rise to type-shifting in the semantics. It is pragmatically odd to refer to the central star of our solar system or to the head of the catholic church with an indefinite article because this implies⁴ that there is more than one entity in the domain denoted by this concept (e.g., # a pope, a sun). This is usually also the case with the Brazilian Portuguese article system. However, it is possible to use individual concept nouns with an indefinite article without giving rise to pragmatically infelicitous interpretations when an activity reading is intended, as in (13) and (14).

In (13), reference is made to the morning sun, not some particular morning sun which would be odd, and (14) does not refer to some sky among several but simply to the sky. Thus, the meaning intended by the verbal complex in (13) is that the person is involved in the activity of sun-bathing, and (14) addresses those persons that are able to engage in the activity of sky-marveling. Here the indefinite article again does not behave as expected in the regular composition but rather gives rise to an incorporated activity reading of the type IND(marvel’(sky’), or, informally ‘some sky-marveling’. These examples seem to fall neatly into the “possible class of weak indefinites” hypothesized by Klein et al. (2013, p. 191). However, since these number-neutral indefinites have not been discussed in the literature yet, the rest of this paper is dedicated to better empirical coverage of them, combining different experimental methods.

For a first assessment of these unexpected uses of the indefinite article, there are at least three empirical issues that have to be addressed. First, it is not clear whether the anecdotic evidence presented in (10)–(14) constitutes a systematic trait of Brazilian Portuguese grammar or whether these are isolated and accidental uses. Second, it must be demonstrated that native speakers’ interpretation of such sentences does not involve quantification over the nominal, as asserted above. Finally, if the indefinite article does not contribute to individuation or quantification, it is unclear what the indefinite article contributes to these phrases. The following section aims at providing first answers to all three questions.

Since the first question is basically about the acceptability of such constructions, an Acceptability Judgment Task (AJT) has been conducted. The AJT not only assesses the acceptability of these constructions in a controlled setting with controlled experimental stimuli, but it also tests a first hypothesis about the contribution of the indefinite article, namely that of creating communicative proximity or informality. Several native speakers that have been consulted with regard to the contrast between the bare and the indefinite number-neutral structures had reported this pragmatic effect. The second question is semantic in nature. A Truth Value Judgment Task (TVJT) was utilized to systematically separate the various quantificational interpretations from the number-neutral interpretations. Finally, a metalinguistic interview was conducted where all three issues were discussed openly with the experiment participants: the acceptability of the sentences, their meaning, and possible functions of the indefinite article.

3.2. The Truth Value Judgment Task

The intuition underlying the incorporation analysis for the indefinite objects in (10)–(14) is that these nouns are interpreted as number-neutral and hence the combination with the verb results in an unquantified activity interpretation. Thus, assistir uma televisão can mean anything from watching one or several movies to shortly zapping through the program (possibly multiple times), and jogar um basquete equally can describe several games or just a couple of shots at the basket (which need not be temporally contiguous). This intuition was tested with a Truth Value Judgment Task. In this task, participants had to decide if a target sentence in a given context was true or false. These situations were designed to reflect the intended interpretation, which is either a repetition of the activity or atelic interpretations (or both). Thus, if speakers rejected the intended interpretation, they would have to declare these sentences as false in the given scenarios.

A 2X2 factorial design was used for this study. The first factor was the presence vs. absence of the indefinite article in the object nominal phrases and the second consisted of two different types of context sentences that introduced the scenario in which the sentences had to be judged. One context reflected the atelic and/or iterative reading and the other described a situation in which the sentence was false because of some unrelated contradiction. An example is given in (16).

The two kinds of context in (16) are different in that in (16a) Sérgio is relaxed and in (16b) he is stressed because he has to train. Thus, (16b) is incompatible with (16c) which says that Sérgio usually relaxes on the weekends. As for the manipulation of the object nominal phrase, the bare nominal version is predicted to be compatible with Context Sentence 1 because it just expresses the activity pegar piscina. For the indefinite object version, if the interpretation is not of an incorporated construction but a quantified one, going several times to the pool would be incompatible with it and the target sentence is predicted to be judged as false. If, however, the verb-object combination is interpreted as incorporated and the nominal as number-neutral, the sentence is expected to be judged as true in Context 1.

The present TVJT was embedded in a larger elicitation task where participants had to produce 40 sentences in total. Eight of these were critical items for the TVJT. Six of the critical items introduced contexts where the activity was iterated and two implied incomplete events. The eight items in their different conditions were distributed over four lists according to the Latin Square design. The elicitation task was implemented with the Speech Recorder software (Draxler & Jänsch, 2004) which randomized and presented the stimuli and recorded the responses. Participants were instructed to form a complete sentence that repeated the target sentence, beginning with their judgment. Thus, the reaction to (16c) should be É verdade dizer que nos fins de semana, Sérgio costuma relaxar e pegar (uma) piscina (‘It is true to say that on weekends, Sérgio usually relaxes at the pool’) or É falso dizer que … (‘It is false to say that …’). Asking for a repetition of the target sentence had two reasons. First, it helped to ensure that participants rated the target proposition rather than another component of the background sentence. Second, repeating the sentence allowed us to see if speakers had issues with the indefinite article in these constructions and, for example, whether they changed or dropped it.

In total, 352 responses were recorded and transcribed (8 sentences by each of the 44 participants), resulting in 88 judgments for each of the four experimental conditions. The analysis of the transcription showed that 23% of the responses judged a different proposition (82 out of 352). These sentences were excluded from the analysis. 37% judged the correct proposition but used a different determiner or none in the indefinite object conditions. 63% of the responses were fully cooperative in the sense that the stimulus sentence was repeated correctly after the judgment. The results in Table 1 summarize all responses that judged the correct proposition, irrespective of article use. Table 2 provides more detailed information about the use of the determiners in the responses and the judgments per experimental item.

The results show that there is a bias toward ‘true’ responses in the data, around 35% of the contexts intended to produce ‘false’ responses were judged as ‘true.’ One reason for that might be that participants could consider the sentence as true if the information was partly compatible. Perhaps in a case like (16), where the target sentence is incompatible with Context 2 because Sérgio cannot be relaxed and stressed at the same time, some speakers might have focused on the fact that in both cases Sérgio is involved in some activity in water and hence considered the target sentence as ‘true.’ Nevertheless, in more than two-thirds of the trials, participants were able to identify the incompatibility as expected. At the same time, speakers overwhelmingly found the target sentences compatible with the context sentences that induced an incomplete activity or iterative reading. Even though some speakers failed to get the intended reading in these trials, the large number of ‘true’ replies strongly suggests that speakers were able to reach the number-neutral interpretation of the indefinite object rather robustly.⁷

The robustness of the availability of the number-neutral interpretation could be even higher as expected because, on closer inspection, the ‘false’ judgments are strongly concentrated on two out of five items (4 and 8, highlighted in grey in the ‘true’ column of Table 2). These items received ‘true’ judgments in 35% and 67%, respectively, while the remaining five items all were judged as ‘true’ in almost 100% of the trials. This suggests that the two particularly diverging items presented some additional problems, and the ‘false’ judgments here did not aim at the unavailability of the number-neutral reading of the indefinite object. Upon closer inspection, in these items, the context sentences indeed could be understood as incompatible with the target sentences for independent reasons. As for the NP type of the elicited direct objects, Table 2 shows that indefinite objects in the stimulus materials robustly elicited indefinite production (64% of observations), and that sporadically, even with bare object stimuli, indefinite objects were produced (cells highlighted in grey in the ‘bare’ column of Table 2).

To summarize, the TVJT offers support for the predicted number-neutral interpretation of indefinite objects in the tested constructions, resulting in pure activity readings that may be modeled as the outcome of semantic incorporation of the noun into the verb. This still leaves open the question of the indefinite article’s role or contribution in these structures. The sections that follow give some preliminary hypotheses in this regard.

3.3. The metalinguistic interview

The metalinguistic interview which concluded the experimental series had several additional purposes which are not relevant here (biographical information of participants, their linguistic attitudes, de-briefing, etc.). Toward the end of the interview, the minimal pairs of sentences were presented to the participants, and they were asked which of the sentence they preferred, whether one of the sentences sounded bad to them and if they accepted both, what the difference between the two versions was (if any). The minimal sentence pairs are given in (17), together with the general reaction with respect to the indefinite object versions.

As can be seen in the reactions given above, even if presented in direct contrast with bare objects, the indefinite objects have been accepted and, in many cases, they were even the preferred option in these constructions. For a majority of participants, furthermore, many of the minimal pairs were synonymous. Besides considering the sentences as synonymous, a very frequent reaction was that the indefinite objects were perceived as being part of informal and spoken language or more intimate communicative situations. When explicitly asked, participants mainly agreed that their interpretation of the indefinite object was number-neutral, admitting that it may denote numerous instances of the stated event or merely fragments of the event. Only rare reports of other interpretations were made. The cardinal interpretation, connotations of large or small quantities or duration times, and links to positively connoted events were among the occasional reactions. Some participants related the examples to a light verb construction (dar uma x-ada), which will be discussed in Section 4.

Summing up, the number-neutral interpretation has been corroborated and the association to informality appeared again, and it was even the strongest of the reported associations. These findings and also the more sporadic interpretations will be put in perspective with the results from the two experimental studies in the following section. In this section, the incorporated objects will also be discussed in a wider context of related constructions.

4.1. The status and interpretation of the ‘incorporated’ indefinite objects

Regarding the first two questions, the evidence from all three data types is overwhelmingly converging and points toward broad acceptability of the investigated phenomenon and the availability of the number-neutral readings. In the AJT, thirty lexicalizations of the construction were tested in a controlled environment including distractors. Speakers rated the sentences with indefinite objects consistently very high and made no notable difference between bare and indefinite objects. These very unconscious judgments are consistent with those made in the metalinguistic interview, in which participants were made aware of the contrast and could even signal a preference if they had one. In this case, the vast majority of the indefinite objects was readily accepted and, in some cases, even preferred over the bare object version. These findings suggest that the type of indefinite objects investigated here is not marginal or accidental in Brazilian Portuguese, on the contrary, it might even be quite a productive trait of the grammar. The results do not provide evidence for unrestricted productivity, however, in the sense of allowing for free combinations of verbs and objects. Since the verb-object combinations were not created randomly in the study but were inspired by observed patterns, productivity can only be considered as partial. A plausible assumption would be that restrictions similar to those on weak definites also govern the availability of weak indefinites with number-neutral interpretations.

The TVJT showed quite clearly that number-neutral interpretations are available. Overwhelmingly, participants judged both sentences with bare as well as indefinite objects as true in contexts of iterative or incomplete events. For most items, the ‘true’ judgments were even categorical and what looks like variation in the overall judgments has been identified as problems with two specific items that received very low judgments for independent reasons. Again, it should be highlighted that the indefinite objects in these sentences were treated very similarly to the bare object sentences. Furthermore, in the metalinguistic interview, participants confirmed the availability of the number-neutral readings. Concerning the reported alternative readings, it should be noted that they happened extremely infrequently and that not all are incompatible with the number-neutral interpretation. As a result, connotations of longer or shorter duration intervals do not contradict a reading of pure activity. However, those interpretations that would contradict a pure activity reading, such as the cardinal interpretation or big or small numbers of events, are manifestly contradictory among themselves. One plausible explanation would be that some few speakers felt forced to come up with ad hoc interpretations in trying to motivate the contrast between the bare and the indefinite objects here. The observation related to positively connoted activities could be explained by the fact that most, if not all examples described such activities. Hence, the generalization was correct, but not necessarily specific to the indefinite objects. The alternative would be that such readings might indeed exist marginally, but they were not readily available in these sentences for the vast majority and further research would be needed to examine them.

Finding answers for question three turned out to be more difficult than for the first two. On the one hand, the previous intuition that indefinite objects were in some sense marked as informal or uses of more intimate spoken communication has been repeated by many participants in the metalinguistic interview, but on the other hand, no effect for informality had been found in the AJT, making the results inconclusive. As already suggested in the presentation of the results of the AJT, the absence of evidence cannot be taken as evidence for absence, and some possible reasons for the lack of an effect have been put forward. Because the informality connotation emerged as the most powerful ‘extra’ meaning in the metalinguistic interview, this intuition should be given more consideration. Two areas of further inquiry are briefly sketched in the next section by examining potentially relevant phenomena in Portuguese and English.

4.2. Pragmatic and constructional sources for the informal flavor

This section discusses some phenomena from Portuguese and English that might be relevant for a better understanding of the pragmatic effect of the number-neutral indefinite objects as they might indirectly shed light on the reported sensation of informality associated with this construction. First, it is important to keep in mind that configurations as those presented in Section 2.3 have not been reported in any article language and do not seem to be possible in these. As an illustration, compare the contrasts between English and Brazilian Portuguese in (18)–(20).

As with other article languages, European Portuguese also does not seem to have such indefinite objects with number-neutral readings. Colloquial Spanish has the construction in (21), featuring an indefinite object. However, this seems to be highly idiomatic. The verb has to be echar, and omitting the indefinite article leads to degraded acceptability, bare and indefinite do not seem to be interchangeable here.

Examples (18)–(20) repeat the idea that Brazilian Portuguese indefinite articles may appear in configurations where other article languages only allow for bare objects. In English, at most a mass quantifier like some may yield similar readings. Interestingly, English some has another use which also is reported to have informal and ‘affective’ connotations.

According to Esposito and Potts (2020), (22c) has this affective connotation, while neither (22a) nor (22b) have it. The authors point out that this additional effect only appears on some when the indefinite article would be a ‘competitor’ in the sentence. Being truth-conditionally equivalent sentences, the authors suggest a pragmatic account for the relevant readings of (22a) and (22c); in their case some as the more marked form would signal lack of engagement in the relevant modes of identification. The competition between English a and some in (22) is surely of a different kind than the one between the bare and the indefinite objects in Brazilian Portuguese in (18b)–(20b). Note that under an incorporation analysis, the bare and the indefinite object would, in any case, have the same semantics, and under such an analysis a pragmatic account for the role of the indefinite article in these sentences could be motivated along the line of reasoning in Esposito and Potts (2020). The specifics of such a pragmatic account must be investigated further. For the time being, the incorporation analysis could provide a solid foundation for a clearer understanding of the pragmatic contribution of the indefinite article in these constructions.

A further source of the informal flavor of the number-neutral indefinite objects might be a somewhat similar and very productive light verb construction, which has also been associated with the indefinite bare singulars by some participants in the metalinguistic interviews. The light verb in this construction is always dar (‘give’), and its general form is that in (23a). Into this general form, either verbs or nouns can be inserted (23b, c), however, depending on the word class, the construction has different properties.

Crucially, the construction derived with a noun (23c) does not show signs of incorporation, whereas the verbal one (23b) does. For instance, it is possible to replace the indefinite article in (23c) with numerals and by this to count the uses of the fork, spoon, or backpack. This is not possible with the verb-based construction in (23b), according to the literature (Figueiredo Silva, 2002; Lisboa de Liz, 2005). Furthermore, the nominal versions can be passivized without problem, which is evidence that they are regular arguments of the verb. Again, this is not possible for the verbal versions. Thus, in (23b) we have another construction in Brazilian Portuguese that seems to have an incorporation semantics and that features an indefinite article. Such similarities between this construction and the indefinite objects discussed here might produce associations between them in the minds of the speakers. And given that the construction in (22c) is highly productive in spoken language but not in written language it might ‘feel’ more informal, and this association could be passed over to the indefinite object constructions. Again, investigating the precise relationship between the construction in (23b) and the indefinite objects remains work for the future. Finally, I would like to emphasize that the two alternative sources of informal flavor are not mutually exclusive and that they both may contribute to the pragmatic effect. The first source would be a pragmatically interpreted markedness effect, and the second would be a usage-based association of the construction with the settings of its recurrence and that of similar and very productive ones.