5.2. Discussion
As to why the correlation found in Araraquara between the type of sg.imper and notions related to deferential/non-deferential address is not as robust in the negative imper, we think that this could be because the distinction is an emergent one. Therefore, in keeping with the principles of koineization, such new and emergent morphological distinctions are more likely to be realised in the more frequent and unmarked positive imperatives than the less frequent and marked negative imperatives. Note also that, cross-linguistically, there is a tendency for distinctions present in positive imper to be neutralised in negative imper (Aikhenvald, 2010, pp. 165–197; Van Olmen, 2024). As to why v-imper is more prevalent in negative imperatives, this is a historical legacy. The v-imper corresponded etymologically to the imper belonging to the paradigm in which você was the subject pronoun; having the você pronoun correspond to the t-imper is an innovation. As noted, this innovation has managed to make clear inroads in unmarked positive imperative contexts but not so much in negative contexts.
Regarding the prevalence of t-imper in the first conjugation versus the other conjugations and especially the irregular verbs (what we classed as conjugation 4), in which there were significantly more v-imper, these effects could be related to frequency effects and the difference between producing imperative forms from memory vs. relying on generalisations/rules present in the morphological system. Frequent morphological forms typically become lexically entrenched and resistant to change (Bybee, 1988, 1995, 2010). The v-imper of the irregular verbs used in the present study (see Table 6) is undoubtedly lexically stored due to their frequency. Therefore, if we accept that the substitution of the v-imper by the t-imper is an innovation caused by the process of koineization/new dialect formation, then it follows that the most frequent, entrenched forms will be most resistant to this change. This conclusion comes from the assumption that irregular verbs have weak associations with other lexemes and so are lexically autonomous and entrenched (Bybee, 1995, 2001, p. 124; Caramazza et al., 1988; Hare et al., 2001, p. 4; Mowrey & Pagliuca, 1995; Stemberger & MacWhinney, 1988) and therefore unlikely to be subject to morphological overgeneralisation (Corbett et al., 2001) and resistant to morphological change. Conversely, those verbs which are most likely to be subject to change are the less frequent ones, which are most likely to be produced in accordance with the morphological generalisation/rules of the language (as opposed to being lexically stored). As to why these verbs should be those of the 1st conjugation, as opposed the those of 2nd or the 3rd conjugations (remember that in both Araraquara and Belo Horizonte, there were significant differences between the 1st conjugation and both the class of irregular verbs and 3rd conjugation verbs), there is independent evidence to suggest that there are qualitative differences between these conjugations. Veríssimo and Clahsen (2009) provide experimental evidence to suggest that first conjugation verbs are constructed in accordance with morphological rules (root + thematic vowel + desinences), whereas the stems of third conjugation verbs are stored as a unit. Accordingly, the prediction would be that any innovations in the morphological system would first target those verbs which rely more on the morphological system for their production.
Regarding the lack of reliable effects for context in Belo Horizonte, it could be that the origin of this change in progress is the city of São Paulo, which, with its 11.5 million inhabitants, is the largest city in Brazil and also the most important city in economic terms. Pending experimental confirmation, were this to be the case then it is understandable that this innovation has reached Araraquara, which is in the interior of the state of which São Paulo (289 km away) is the capital and the economic, cultural, and prestige focus point for residents. Belo Horizonte is not only further away from São Paulo (583 km), but it is also the capital and prestige centre of its own state, Minas Gerais, which historically was in conflict with São Paulo for political and economic hegemony of Brazil. Given that in both regions você, which corresponded etymologically to the v-imper, is the dominant subject pronoun, then there is evidence to support the aforementioned idea of a change in progress originating in São Paulo but which is yet to reach other capitals of the South-East region of Brazil. This evidence comes from the overall numbers of imper (see Table 7) which show a preference in Belo Horizonte for v-imper (56% of tokens in Belo Horizonte vs. 46% in Araraquara).
The general conclusion of our study could be that the reliable effects for context in Araraquara suggest that there is a change in process whereby BP is developing a nuanced term-of-address system. Alongside the established differences between deferential/non-deferential address marked by the use of the subject pronouns (o senhor/a senhora vs. você respectively) there is emerging within the você paradigm a way to mark social distance/power/solidarity via the use of t-imper or v-imper. The latter being correlated with more deferential situations. This socially determined use of the sg.imper is an emerging generalisation and therefore is only present in positive imperatives and effects those verbs whose imperative forms are most likely to be ‘morphologically created’ as opposed to having a dedicated memorised form. The emerging nature of this generalisation and the supposition that it originated in São Paulo, helps to explain the lack of reliable effects in Belo Horizonte. The change has not reached other cities of the South-East yet.
Whilst the previous paragraph could be a valid conclusion based on our study, we would like to remain circumspect with our conclusions, however, due to the significant inter- and intra-individual variation in the data. For example, whilst the percentages do indicate a slight preference in Belo Horizonte for v-imper, it should be noted that t-imper occurred in 44% of cases; this is an extremely high number. Note also that when the data sets for both cities were combined there were no reliable effects for the variable city or the variable context. Also, although there were reliable effects in Araraquara for the variable context, confirming our hypothesis, when one analyses the data more closely (see Table 9 and Figure 3) two things stand out: (a) subjects produced an imperative which went against our hypothesis in roughly a third of cases; (b) there remains much individual idiosyncratic variation. For example, in Araraquara speaker F3 behaved almost categorically in accordance with our hypothesis (t-imper in non-deferential contexts and v-imper in deferential contexts) whilst speaker M4 displayed v-imper consistently in both contexts. In the following section, we entertain the possibility that both imperatives exist in a state of stable free variation, and we discuss the concept of morphological idiosyncratic variation.
5.2.1. Idiosyncratic variation
Idiosyncratic variation (henceforth IV) has been given different names in the linguistic literature (e.g., Ellis, 1985, p. 136: “random variability”; MacKenzie, 2019: “individual differences”; Guy, Oushiro et al., 2022: “intra-speaker variation”; Dorian, 2010a: “idiosyncratic variation”; Bülow & Pfenninger, 2021: “intra-individual variation”; Bolinger, 1956; Kopf & Weber, 2023 i.a.: “free variation”). It relates to what Moser (2023) calls true optionality, this is, cases where speakers consistently alternate between two or more syntactic variants without a clear preference, and where neither form is associated with specific grammatical, pragmatic, or sociolinguistic constraints. This means that the variation is not driven by external pressures (such as register, dialect contact, or language change) but is instead an inherent property of the speaker’s grammatical system. It can also be linked to the concept of overabundance (Thornton, 2012, 2019), the situation in which two (or more) inflectional forms are available to realize the same cell in an inflectional paradigm. On the topic of overabundance, Thornton (2019, p. 232) notices that “most linguists assume that overabundance can exist only as a transitional stage during diachronic change, and that any single speaker only uses one of the cell mates available in a community’s repertoire; besides, many assume that cell mates always differ according to geo-socio-stylistic conditions, or in meaning”. This is because the possibility of this variation goes against a number of proposed linguistic principles: Clark’s Principle of Contrast (Clark, 1987), the Pāṇini’s principle (Cardona, 1997), or the idea within theories of koineization/dialect mixing that after a period of variation grammars will become stable and focussed (Britain & Trudgill, 2005; Kerswill, 2004; Siegel, 1985; Trudgill, 2001a; 2001b). Most significantly, however, is the fact that the presence of IV is in direct contradiction to one of the most fundamental and guiding principles of research in variational linguistics, namely, the principle of “orderly heterogeneity” (Weinreich et al., 1968). This principle states that language variation is not random or idiosyncratic but, when not grammatically governed, is structured along social dimensions whereby linguistic variables can be indexical of some social meaning or category. There is a plethora of studies which show how variables at all linguistic levels of analysis can be correlated with grammatical and social meaning (see Beaman & Guy, 2022; Werth et al., 2021). However, the fact that some variables in some languages can have social or grammatical correlates does not necessarily mean that all variation is correlated this way. Indeed, it has been suggested that the assumption of orderly heterogeneity has made researchers blind to idiosyncratic variation (Dabrowska, 2012; Dorian, 2010a, 2010b). Moreover, there is evidence (Bülow et al., 2019) to question studies in the area of correlational sociolinguistics for adhering to what Wolfram and Beckett (2000) have termed “the homogeneity assumption”, whereby sets of speakers are treated as a homogenous group with respect to their variation.
Studies within what has been termed the third-wave of variationist sociolinguists (Eckert, 2012) have moved away from such homogenizing tendencies and view linguistic variation as not marking fixed social categories but as a tool for constructing different personae, styles and identities which, in turn, are related to larger social structures (Moore & Carter, 2015). However, the orderly heterogeneity of linguistic variation is assumed (although it is accepted to be instantiated in complex social ways), and idiosyncratic variation is largely disregarded. This tendency is also related to the shift of analysis in sociolinguistics from the individual level, i.e., that of idiolects, to the population level (Weinreich et al., 1968). The necessity of this analysis was further enforced by Labov (2001, p. 34) who highlighted that “linguistic analysis cannot recognize individual grammars or phonologies … individuals are not the final units of linguistic analysis, but the components that are used to construct models of our primary object of interest, the speech community.” However, more recent research, especially within sound change, has highlighted the importance of the individual in linguistic change (Anthonissen, 2021; Blythe & Croft, 2021; Stevens, 2022; Stuart-Smith & Timmins, 2010).
In spite of the orderly heterogeneity assumptions lasting relevance (see Beaman & Guy, 2022), there is a growing body of research that stresses the prevalence of IV which is not correlated with any social meaning within language communities (Bülow et al., 2019; Bülow & Pfenninger, 2021; Dorian, 2010b; Entringer, 2021; Fasold, 1975; Hymes, 1979; Johnstone, 1996; Le Page & Tabouret-Keller, 1985; O’Neill, 2024; Schilling-Estes, 1998; Wolfram & Beckett, 2000). Dorian (2010a) highlights the disregard for analyzing IV, noting that researchers often miss it due to their backgrounds in highly standardized languages and ingrained assumptions that speech variation is tied to social categories (e.g., age, gender) or grammatical constraints (see also Milroy & Milroy, 1999; Schieffelin et al., 1998).
Werth et al. (2021, p. 2) postulate different causes for the lack of particular attention on the matter. Firstly, they highlight how the methods for carrying out research on IV using insights from variationist and complexity theories “have remained relatively elusive”. Secondly, they note that IV goes against Chomsky’s (1965) view of generative grammar, which depicts the competence of an ideal speaker-hearer in a homogeneous speech community. Speakers’ convergence is taken as a fact (“almost [as] an article of faith”, Dąbrowska, 2018, p. 222). Such stances lead to disregarding IV, before even asking what it could mean, or supposing that there is a need for more research to uncover the underlying correlations. Note that Guy and Hinskens (2016, p. 1) state, “a fortiori there has to be a unifying, albeit maybe slightly more abstract, overall system in the coherence of language varieties, otherwise we would all be living in an eternal tower of Babel.”
We would like to suggest that the variation in sg.imper in Belo Horizonte, at least, is a matter of IV. And, despite the reliable effects of context and polarity in our study and immediacy in other studies, we highlight that these effects are tendencies and not absolutes. There still remains much IV at the population level, which needs to be acknowledged. These conclusions for BP, however, are problematic in the light of previous studies, most notably Dorian (2010a), who suggests that the prevalence of IV can be a result of social factors, and identifies the following social traits:
the language has a small population of native speakers;
there is an absence of socioeconomic differentiation between speakers;
the language can be classed as a minority language or endangered language;
the language is used as the traditional means of communication within the home, but it is under pressure from another more dominant language for communication in society, as a whole;
the language lacks a written standard (see also Entringer, 2021).
These social traits are categorically not characteristic of Brazilian society, and yet, the results of our study on sg.imper and also a study on pl.imper (O’Neill, 2004) reveal much IV in BP. This latter study, which was based on data from the current experiment but focussed on pl.imper, found that although speakers displayed certain tendencies or preferences for pl.imper (e.g., in negative contexts in Araraquara) these tendencies were not displayed overall by all speakers across both cities. The only variable which was significant in both cities was that of speaker, suggesting that there exists much idiosyncratic variation in the pl.imper. We suggest that the same could also be true of sg.imper. In what follows we argue that the relationship with the written standard could be a significant ecological factor in the maintenance of idiosyncratic variation. Therefore, languages need not lack a written standard to display IV, rather IV can appear in social contexts in which speakers feel alienated from the standard language and therefore do not actively engage with it or aspire to learn it.
5.2.2. The formation of BP and the rise of its standard
Spoken BP is considerably different from the official standard for the country. So much so that it has even been claimed that BP should be considered the language with the most native speakers that does not have a written system (Perini, 2003). Space restrictions do not allow for a full discussion of the origins of the marked differences between spoken and written BP or the history of the establishment of the BP written standard (for an overview, see O’Neill, 2022). What follows is a summary of the main points.
In addition to the universal tendency of languages to change over space and time, in Brazil there were also a number of socio-historic and political factors which worked as a catalyst for language change. Politically, there was, from colonial to modern times, a significant disregard for education and literacy. Historically, the Portuguese spoken in Brazil was subject to prolonged and extensive (a) contact with other languages, (b) acquisition as a second language by adult speakers (indigenous peoples, African slaves, and non-Portuguese speaking migrants), (c) regular exponential increases in the number of native speakers, and (d) dialect mixing due to population migrations. Given such social circumstances, the predictions of linguistic research is that of rapid language change resulting in the emergence of new varieties of the language, or koinés (Bentz & Winter, 2013; Bentz et al., 2015; Britain & Trudgil, 2005; Kerswill, 2013; Kerswill & Williams, 2000; Lupyan & Dale, 2010; McWhorter, 2007; Nettle, 2012; Nichols & Bentz, 2017; Sampson et al., 2009; Siegel, 1985; Trudgill, 1986, 2001a; 2001b, 2010a, 2010b, 2011a, 2011b). These new varieties with their mixed paradigms and abundant morphological variation are the basis of spoken BP.
The standard written language for BP is markedly different, however, since it was artificially created at the end of the nineteenth century on a largely EP model (Faraco & Zilles, 2017). The reasons for this were ideological and political. In the newly established Brazilian Republic, the elites were grappling with the social realities of their country, especially the racial ones, and were highly influenced by ideas originating in Europe. A continent which represented modernity and progress and was the model to imitate. In linguistic terms, this translated into basing the standard language (largely though not exclusively) on that of the old metropolis, even though this was linguistically very distant from virtually all of the Brazilian population. The rational for this decision was that it was considered that the standard language should have fixed forms and rules and constitute something which has been elaborated and endorsed by an elite group of experts. The differences between this standard and spoken BP were not seen as a problem. On the contrary, it was considered a positive aspect that people might have to strive to learn this standard since it was hypothesized that this would also bring about positive civilising effects. Unfortunately, however, there was never the political will or public investment to ensure that this standard was taught to the illiterate masses from the establishment of this standard until the modern era. At this point, the differences between spoken BP and the written standard were numerous. This fact is not, in itself, a problem for speaker’s engagement with and acceptance of the standard language. Many languages display significant linguistic differences between the official written standard and the actual every-day speech of its speakers, for example German.6 In Germany, however, the standard (Hochdeutsch) is accepted as the legitimate form of written language. Moreover, in Germany, there is, and historically has been, the political will and material resources to ensure that this standard is successfully taught in schools. This has not been the case with the standard norm in Brazil historically (O’Neill, 2022, p. 13–16), which ultimately resulted in its legitimacy being compromised since its artificiality was fiercely condemned by linguistics and teachers alike. In fact, the legitimacy of the standard-norm became so compromised that teachers claimed that they did not teach this norm but another type of standard language, termed the norma culta ‘educated norm’ (Soares Gomes, 2019, p. 22). This educated norm (Faraco, 2008) is typical of the speech of the educated and affluent who live in the major urban centres, and it does not usually correspond, not even in formal, monitored contexts, to the grammatical characteristics of the traditional standard norm. This educated norm is the de facto norm which many speakers aspire to. However, due to a lack of comprehensive, codified accounts of the features of this norm (Faraco, 2017, p. 23; Moura Neves, 2002, p. 239) and its lack of official status, it is a nebulous concept (O’Neill, 2022, p. 15–16) and so cannot really be termed a ‘norm’ but rather a set of linguistic tendencies which roughly correspond to the written language in (some) newspapers and (some) other formal materials and to the spoken linguistic habits of the educated. However, given Brazil’s size and population, there are many educated individuals and thus there are many different regional educated norms = linguistic tendencies (Faraco & Zilles, 2017, pp. 184–186), including different tendencies for speaking and writing.
What the foregoing reveals is that due to a number of factors (artificiality of the written standard, loss of legitimacy of this standard, lack of investment in education) Brazilian speakers came to become alienated from the traditional official standard. For many speakers, it did not represent something to be actively acquired, since this position was occupied by the educated norm. This norm, however, was not fixed or prescriptive but an organic development characterised by a number of linguistic tendencies. Tendencies, by their nature, are variable and elastic. It is this relationship with the written standard that, in our opinion, approximates Brazilian society to those societies characterised by idiosyncratic variation. It is not the case that there is no official standard for BP, or speakers are excluded from the written standard. Rather, they are alienated from it and do not actively engage with it. In monitored contexts, they aim to speak or write a form of the language which is flexible and highly variable.