The literature on postcolonial, nativizing African varieties of Portuguese (AVPs) spoken in Angola (AP), Mozambique (MP), and São Tomé and Príncipe (STP) refers a more widespread and non-standard use of locative preposition em ‘in’ selecting the Goal argument of directed motion verbs (e.g., Adriano, 2014; Avelar, 2017; Chavagne, 2005; Mingas, 2000; P. Gonçalves & Chimbutane 2004; P. Gonçalves, 2010; R. Gonçalves, 2010), which contrasts with the use of the standard directional prepositions a ‘to’ and para ‘to, toward’ used in European Portuguese (EP), the official standard in the former Portuguese colonies in Africa. In the case of AP and MP, the role of language contact with Bantu languages has been at the forefront of the descriptive and theoretical explanations regarding these patterns (e.g., Avelar & Álvarez López, 2018; Marques, 1985; Mingas, 2000; P. Gonçalves & Chimbutane, 2004).

Since these previous claims regarding the expression of Goal arguments are typically not based on quantified data nor on comparisons across AVPs, we aim to provide a corpus-based assessment of the expression of Goal arguments of two high frequency verbs of inherently directed motion (cf. Levin, 1993), ir ‘to go’ and chegar ‘to arrive’, in the three varieties at stake, in order to discuss the factors that drive language variation in this domain. Since language contact has been a primary explanation, the point of departure will be a cross-comparative approach between the three AVPs and their main contact languages in the urban areas where the corpora were collected: Bantu languages Kimbundu (KB) for AP (Luanda) and Changana (CH) for MP (Maputo), and creole language Forro (Santome) for STP (São Tomé). Since it will be shown that the contact-induced hypothesis does not offer an encompassing explanation, we will shift our focus to an analysis that takes into account semantic features of the constructions under discussion.

The paper is organized as follows. Section 2 sets the stage with respect to the three varieties at stake and the corpus-based data used in this paper; section 3 discusses the expression of Goal arguments in the three AVPs with motion verbs ir ‘to go’ and chegar ‘to arrive’ based on quantified data, showing the patterns of variation within and across these varieties; section 4 provides an overview of the properties of Goal arguments in the main contact languages of the AVPs, namely Kimbundu, Changana, and Forro; section 5 relates and discusses the findings of sections 3 and 4 from the perspective of language contact, arguing that this factor lacks explanatory power to account for the observed patterns; finally, in section 6, we offer an alternative explanation for the data at stake that takes into consideration the semantics of the predications of the verbs ir and chegar and of the complements of the prepositions, focusing on features such as durativity and the type of location.

Portuguese in Africa has been historically acquired as an L2, a phenomenon that essentially has its roots in the late 19^th and 20^th century when the former Portuguese colonies were effectively colonized (e.g., Hagemeijer, 2016; P. Gonçalves, 2013). In addition, Portuguese in Angola, Mozambique, and São Tomé and Príncipe also shows an increasing tendency toward nativization since their independences in 1975, boosted by the choice of Portuguese as the only official language and its widespread democratization in all these postcolonial societies, hereby reinforcing its role as a lingua franca. Portuguese is by far the most spoken language in São Tomé and Príncipe (98,4%), whereas the predominant creole, Forro, is nowadays only spoken by 36,2% of the population (INE, 2013); in Angola, 71,15% of the population indicates that Portuguese is the most spoken language at home, especially in urban areas, whereas Kimbundu, one of the main Bantu languages which traditionally competes with Portuguese in the capital Luanda, is only spoken by 7,82% (INE, 2016); in Mozambique, Portuguese is spoken by around half of the population, which includes 16,6% of L1 speakers (INE, 2019), also with prevalence in urban areas. Despite the lack of direct information on L1 and L2, it can be inferred that Portuguese in Luanda and São Tomé is generally L1 and/or the primary language, with increasing monolingualism, especially among the younger generations, which make up a substantial part of the total population.

This case study is based on spoken urban corpora of MP, AP, and STP that were prepared within the project Possession and Location: microvariation in African varieties of Portuguese (PALMA) (Hagemeijer et al., 2021; Miguel et al., 2021; R. Gonçalves et al., 2021). The corpora were collected in the capitals Maputo, Luanda, and São Tomé between 2008 and 2020, by several researchers of the Center of Linguistics of the University of Lisbon, as well as doctoral students, and were as much as possible balanced according to level of education, age, and gender (cf. Hagemeijer et al., 2022). The interviews that integrate the corpora are predominantly semi-structured. Portuguese is the L1 or primary language of the majority of the informants who contributed to the corpora, especially in the case of urban AP and STP. Table 1 summarizes the basic information of the corpora used.

For the data set used in this paper, we extracted all the occurrences of ir ‘to go’ and chegar ‘to arrive’ from the three corpora on the CQPweb platform and then proceeded to exclude the following contexts:

The remaining constructions were annotated with respect to (i) the preposition selecting the locative argument (V PP_Loc) and cases without preposition (V NP_Loc), (ii) the physical or non- physical nature of the propositions described by predications, and (iii) the nature of the NP (area or container) of PPs headed by locative em.

The available literature on AP, MP, and STP refers the use of a non-standard prepositional strategy with locative em ‘in’ to introduce arguments with the thematic role of Goal1 (e.g., Adriano, 2014, pp. 334–340; Cabral, 2005; Chavagne, 2005, pp. 223–227; Mingas, 2000, pp. 75–77; P. Gonçalves & Chimbutane, 2004; P. Gonçalves, 2010; R. Gonçalves, 2010) instead of the standard prepositions a and para. Chavagne (2005, p. 227) mentions that the use of em with verbs of movement represents a strong tendency, comparing it with Brazilian Portuguese (BP). Based on a smaller spoken corpus, R. Gonçalves (2010, p. 48) shows that in STP there are two main tendencies: ir preferentially selects PPs headed by para (even in contexts where EP selects a), whereas chegar shows a tendency toward em. Our data show Goal arguments of ir and chegar (among other verbs) headed by em are indeed attested in all three AVPs, as illustrated in (1) below.2

In our corpora of the three AVPs, however, Goals introduced by prepositions a or para, the standard strategies in EP, overall represent the most common strategy, which can be seen in examples (2–3).

The overall results for the three AVPs with respect to the use of a, para, em, até, Ø (null preposition), and other prepositioned contexts introducing Goal complements of the verb ir in our corpora are presented as percentages and the corresponding raw numbers (between parentheses) in Table 2 below.

The overall results for the three AVPs with respect to the use of a, em, até, Ø, and other prepositioned contexts introducing Goal complements of the verb chegar are shown in Table 3 below.

The tables show in the first place that the varieties exhibit substantial disparity with respect to the patterns selected by ir and chegar, both within and across varieties. It also follows that the complements of ir and chegar are predominantly PPs and not NPs. STP, however, is exceptional in the sense that this variety exhibits a rather expressive number of cases where chegar selects an NP argument (22/22%).

In the specific case of ir (cf. Table 2), the standard EP strategies with a and para are most frequent (940/83,56%), with a strong preference for para in the case of STP (259/61,52%) in comparison to AP and MP.3 With respect to the use of the non-standard strategy with em, which only represents a relatively small number of overall cases in these varieties (115/10,22%), AP is the variety that stands out for its more widespread use of this preposition (63/29,30%). The application of Fisher’s test4 (Eddington, 2016) to the data in Table 2 shows that significant differences can be observed between varieties, since the p-value is much lower than 0.05 in all cases: The pairs AP/MP, AP/STP, and MP/STP all have a p < .001.

In the case of chegar (cf. Table 3), the variation in the AVPs mainly occurs between two prepositional strategies, either with a or with em. In contrast to ir, the non-standard strategy with em, corresponding to 144/51,25% of the occurrences, is more widespread than the canonical strategy, which accounts for 93/33,10% of the cases. While the strategy with em is dominant in MP (86/67,19%), it is also common in AP (21/39,62%) and STP (37/37%). Differently from AP and MP, as mentioned above, STP also shows a fair number of cases where chegar selects an NP (22/22%), a tendency that was not observed with regard to ir (2/0,47%). Applying Fisher’s test to the data in Table 3 shows, again, that significant differences between varieties can be observed: the pair AP/MP has a p-value = 0.006181; the pair AP/STP a p = 0.004861; and, most notably, the pair MP/STP a p < .001.

All in all, the statistical tests suggest significant differences between the three varieties with respect to the selection of Goal complements with both ir and chegar.

In order to assess the role of language contact, we will now turn to the strategies used to express Goal arguments in the main contact languages of the three urban AVPs at stake, which are the Bantu languages Kimbundu (Mbundu) and Changana, for AP and MP respectively, and the creole language Forro for STP. If contact plays a role, as has been claimed by several authors, the role of these languages is expected to be both direct and indirect. Direct, in the sense that there are speech communities and/or bilinguals of these languages within the urban environment and among our informants, especially in the case of MP; indirect, in the sense that these languages – or sometimes languages with a similar typology spoken by migrants moving to urban areas – have been historically in contact with Portuguese, even if speakers do not necessarily speak them (fluently or actively).

For Proto-Bantu, the ancient ancestor of the contemporary Bantu languages, three locative noun class prefixes have been reconstructed, corresponding to classes 16, 17, 18 (*pà-, *kù-, *mù-), which have a reflex in many contemporary Bantu languages but have been lost or only occur as vestiges in other ones (e.g., Zeller, in press).

Kimbundu, a western Bantu language, preserves reflexes of the three proto-Bantu locative noun classes (e.g., Araújo, 2013; Chatelain, 1888–1889; Diarra, 1990; Miguel, 2019).5 The marker that is standardly used to introduce the Goal argument of verbs of directed motion is ku, which stems from class 17, as illustrated in (4–5).

Differently from Proto-Bantu, however, the locative markers in Kimbundu exhibit properties of free morphemes, i.e., prepositional items, instead of prefixes, as shown in (4–5). The prepositional use of locative morphemes has been demonstrated for a number of Bantu languages in different areas (e.g., Marten, 2010; Zeller, in press). Although the prepositional status of locative markers in Kimbundu has been assumed in different studies without being substantiated with linguistic evidence (e.g., Diarra, 1990, p. 59), it is, for instance, supported by the fact that the locative marker does not trigger locative agreement in object position. In (6), it is the noun ngeleja ‘church’ which triggers class 9 agreement with demonstratives and possessives and not the locative related to class 17.

Furthermore, when the locative marker selects an animate object, a connective morpheme -a is required, as illustrated in (7).6 This feature is not compatible with a prefixal status of ku because the intervention of this type of morphological material between a noun class prefix and the nominal root is not attested in Bantu languages.

The expression of location in Changana, a southeastern Bantu language, is quite distinct from the facts observed for Kimbundu, since the language only exhibits remnants of the three locative classes reconstructed for Proto-Bantu (Chimbutane, 2002; Ngunga & Simbine, 2012; Sitoe, 2001). Instead, Goals are characterized by other morphological strategies (Chimbutane, 2002; Sitoe, 2001), typically the attachment of suffix -ini (or its alomorph -eni) to the NP, accompanied by optional prefixation of á- or é-.

Sitoe (2001, pp. 6–7) mentions that Changana locatives are generally vague with respect to their localization. He illustrates this claim with the example nambzeni (nambu ‘river’ + -eni), which “can refer to a location inside the river, on the river, at the bank of the river, or in the general area of the river.”

Changana further displays a number of inherently locative nouns, which includes borrowings, that lack any overt marking, as mánánga ‘desert’ in (9a), or can be prefixed with á- to reinforce their locativity, as ákáyá ‘home’ in (9b).

For endogenous toponyms, on the other hand, prefix ká- is employed (10a), whereas toponyms of foreign origin are not marked or are optionally marked by prefix á- (Sitoe, 2001, p. 10) as shown in (10b).

Finally, Goal arguments in Forro, the main creole language of S. Tomé, are uniformly expressed as NPs (e.g., Hagemeijer, 2004).

In sum, the three main contact languages of the urban AVPs analyzed exhibit fundamentally different solutions to encode Goal arguments with verbs of directed motion: Kimbundu exhibits PPs headed by ku; Changana uses a strategy in which NPs are morphologically marked; in Forro, Goals are also NPs but lack any (overt) morphological marking.

A cross-comparison between the patterns available in the AVPs and their respective main contact languages shows that the contact-induced hypothesis runs into several problems.

First, the AVPs in this study predominantly use a PP strategy to introduce Goals of the direction motion verbs ir and chegar, whereas two of the contact languages, Changana and Forro, exhibit an NP strategy, respectively with or without overt morphology. Second, while the prepositions heading PPs in the AVPs show considerable variation, a clear-cut distinction arises between ir, where standard a and para constitute the dominant pattern, and chegar, which shows a preference for the non-canonical strategy with em. The main contact languages (Kimbundu, Changana, and Forro), on the other hand, use a language-internally uniform strategy to express the complement of these and other directed motion verbs.7

The contact-induced hypothesis is also unable to explain why ir in STP shows a strong preference for para and hardly any cases of em, and chegar a primary pattern with em, since Forro systematically uses an NP argument in both cases. In AP, however, the more pronounced overall presence of em, although far from being the exclusive patterns, would be more consistent with a more emphatic role of language contact, since Kimbundu also displays a generalized and uniform PP strategy with locative ku, which could further facilitate syntactic transfer and convergence.8 This line of argumentation can arguably not hold for MP though, since it was shown that its main contact language Changana exhibits morphological strategies.

To overcome this problem in the case of MP, P. Gonçalves and Chimbutane (2004) and P. Gonçalves (2010) propose an analysis, based on a scenario of L1 Bantu speakers who are L2 learners/acquirers of Portuguese, whereby preposition em is reanalyzed as an NP-internal locative Case marker, [em-NP], i.e., as a prefix, inspired in the Bantu noun class tradition. This hypothesis, they argue, would then be able to account for examples like (13a-b), where preposition no in (13a) and em (in 13b) would be incorporated into the NP as follows: [_NP no-encontro da igreja] and [_PP para [_NP em-casa]].

We concur with P. Gonçalves and Chimbutane (2004) that examples of cooccurring prepositions, in (13b), in particular para and em are rare. In our data, we identified only 5 such occurrences, produced by 4 different speakers. In 4 of these occurrences, the locative NP was casa ‘house’. The scarcity of these constructions would therefore not only weaken the hypothesis that em is a prefix, but would also require further explanations as to the workings of the grammaticalization of a preposition into a bound morpheme in MP. As shown in section 4, the primary locative strategy in Changana relies on a suffix (-ini or -eni), and not on a prefix. Moreover, there is no evidence that other Bantu affixes, for example negation prefixes or verbal extensions (applicatives, etc.), induce reanalysis of Portuguese morphemes as bound morphology in MP (or AP). This raises the more general issue as to whether or how morphology in a L1 is processed and transferred into an L2 target language in contact situations. In the case of Changana-Portuguese it would be a costly operation, requiring both a categorial and syntactic reanalysis of em.

Furthermore, Bantu locative morphemes are typically prefixed to the head noun, as in the case of Changana’s ká- and á- described in section 4, which means that in a strong version of the contact hypothesis the strategies with em are not expected to contract with definite articles9 or demonstratives nor to be separated from the head noun by possessives, numerals, prenominal adjectives, etc., contrary to fact (e.g., há quem chega no meu salão ‘some people arrive at my beauty salon’, MP corpus data). This type of noun modifying material is generally postnominal in Bantu, but in those languages where it occurs preverbally, the locative marker is a free morpheme (preposition).10

The contact hypothesis for MP was also adopted by Avelar and Álvarez López (2018) for the (L2) Cabinda variety of AP, which is primarily in contact with Ibinda, a western Bantu language belonging to the Kongo cluster, which exhibits prenominal locative morphemes. While this would be syntactically consistent with contact-induced hypothesis above, it still faces the problems mentioned in the previous paragraph regarding the NP-internal syntactic differences between Bantu and Portuguese.

Finally, the NP-internal prefixation hypothesis proposed originally by P. Gonçalves and Chimbutane (2004) is unable to account for AP, since it was shown that Kimbundu locative markers introducing Goals are free morphemes (prepositions). Since Kimbundu and Portuguese both exhibit prepositions, it would be counterintuitive to claim that Ps incorporate in AP with verbs of directed motion. In fact, as expected, we did not find any instance of double prepositions in our AP data.11

The gradual withdrawal of the contact languages caused by shift, especially in the urban environment in Angola and São Tomé, where Portuguese is not only widely spoken but also frequently the native or primary language of the speakers, with increasing monolingualism in Portuguese, especially among the younger generations, constitutes yet another argument that increasingly weakens the role of language contact with respect to the observed patterns. This tendency has become stronger since the independences in 1975 and can be correlated with the massification of education in the official language and, concomitantly, higher educational attainment. Even though the spoken local (non-native) varieties constitute the main input (e.g., Stroud, 1997), the role of education is expected to raise the speakers’ linguistic awareness and to promote wider use of patterns found in the standard language.12 While the historical relevance of language contact cannot be ignored, the contemporary urban environments have become the playground of thriving language shift toward Portuguese, diminishing the role of transfer of substrate patterns.

Finally, compared to EP, the AVPs appear to privilege prepositions para and em over a with the verbs at stake, even though ultimately a finer-grained analysis is required to assess the use of para and a, for example with regard to implications concerning a longer or a shorter stay at the Goal, as in the difference between ir para Luanda ‘to go to Luanda’ vs. ir a Luanda ‘to go to Luanda’, where the former implies a longer stay in Luanda than the latter. The partial loss of a is an expected outcome in contact situations where L2 acquisition plays or played a prominent role. Phonetically weak and semantically opaque functional items, such as directional preposition a, definite articles, or accusative clitics, are prone to loss and/or partial replacement by other, often less functional material. In contexts of ambiguous and insufficiently dense input (e.g., P. Gonçalves, 2002, 2004), it is expected that directional para or locative em have the upper hand in comparison to a. For instance, the latter, unlike the former two, did not make it into the Portuguese-related creoles in West-Africa belonging to the Upper Guinea and Gulf of Guinea clusters. Moreover, in (spoken) BP, which also has a history of substantial L2 acquisition and partial restructuring (e.g., Lucchesi, 2001), directional preposition a has given rise to variation, favoring in particular em in the structures at stake, and does not occur in Afro-Brazilian subvarieties which underwent substantial restructuring (see Avelar, 2017, for discussion).

In sum, the empirical evidence does not underscore a major direct role for language contact regarding the selection properties of directed motion verbs ir and chegar, even though there may be – on a limited scale – individual grammars that privilege the grammatical properties of the contact languages. It was also shown that a morphological analysis based on language contact, proposed initially for MP, cannot be applied across the board, primarily because it is not consistent with the data from the contact languages and the AVPs.

In light of the above, we aim to develop a different, semantic-oriented approach to the data, as an alternative (and complement) to the contact-induced hypothesis.

In the remainder of this paper, we discuss Goal PPs introduced by em, a non-standard strategy in EP, focusing primarily on the lexical semantics of verbs and their objects and on the semantics of predications.

The main goal of this paper was to assess the use of preposition em introducing Goal phrases in urban AVPs, in order to test the hypothesis of language contact as the major cause for this use. To do so, we analyzed examples extracted from spoken corpora of three AVPs (AP, MP, and STP) with two inherently directed motion verbs ir ‘to go’ and chegar ‘to arrive’ and we compared the strategies to express Goal phrases in these AVPs with the ones used in the main contact languages (Kimbundu for AP; Changana for MP; and Forro for STP).

In general, we found that AVPs typically use PPs as Goal phrases and that there is a difference between the verbs: ir generally occurs with the standard EP strategies (a and para), whereas chegar exhibits strong variation between the standard EP strategy with a and the non-standard EP strategy with em. In addition, STP stands out in comparison to AP and MP because of a strong tendency for ir to select para and due to the fact that this variety also exhibits a fair number of cases where chegar selects an NP argument.

It was further shown that the language contact hypothesis lacks explanatory power with respect to the use of prepositions with the verbs ir and chegar across AVPs. The three main contact languages exhibit internally uniform patterns for the verbs ‘to go’ and ‘to arrive’ (PPs headed by ku in Kimbundu; morphological strategies in Changana; and NPs in Forro), which do not match the patterns in the respective AVPs, where PPs are the dominant pattern and ir and chegar exhibit different patterns of variation. Although the semantic hypothesis was not explicitly applied to the contact languages discussed in section 4, they lack the kind of sensitivity to the semantics of the predicate and to the nature of the Goal (area or container) observed in the AVPs with regard to the verbs of directed motion at stake. At best, there is only a very mild correlation between some of the patterns that occur in the AVPs and in the contact languages. In addition, we argued in section 5 that the contact-induced morphological incorporation hypothesis of preposition em inspired on the Bantu locative classes, originally proposed for MP, lacks empirical motivation.

Given the weaknesses of the contact-induced hypothesis, we put forward a different hypothesis. We assumed, as in Nikitina (2008), that the use of locative prepositions heading Goal phrases can be explained taking into consideration semantic/pragmatic factors. To test this hypothesis, we analyzed the data using the following parameters: the semantic nature of the predication, i.e., whether it describes an event of movement that can be traced using spatial coordinates; the semantic nature of the complement of the preposition, i.e., whether the NP denotes an entity with well-defined boundaries; and the lexical semantics of both verbs, i.e., whether the verb describes a durative or non-durative change of place.

The main findings of this approach are as follows. In the first place, and as expected, there are no clear-cut rules regulating the use of em in AVPs. Instead, directional readings of preposition em in AVPs are favored by the following properties, where i) was shown to be the most prominent and (iii) the least relevant:

The main conclusion that emerges is that in the case of the contemporary, urban AVPs the contact-induced hypothesis is seriously flawed by the empirical data, whereas the occurrence of Goal phrases headed by locative prepositions (such as em) is attested in different, typologically unrelated languages. In other words, we are dealing with a cross-linguistic phenomenon that relies on the mechanisms available in each language to express motion and on general semantic and pragmatic principles.

There are, however, some shortcomings of our analysis that require future research, in particular the amount of analyzed verbs – only two. Different types of verbs of movement (durative and non-durative verbs; manner of motion and inherently directed motion verbs) will have to be added to the sample in order to test if our general conclusions hold regarding the durative vs. non-durative nature of (the lexical semantics of) the verb and its relative weight in the licensing of Goal phrases headed by em. Other aspects of Nikitina’s proposal, namely the hypothesis that Goal-em phrases cannot cooccur in the same predication together with Source phrases or with verbs that describe highly specific manners of motion (e.g., to crawl), will be difficult to evaluate, because they occur only scarcely in the spoken corpora. A workaround would be to carry experimental tasks with native speakers of the AVPs. The increase in the number of contexts will also allow us to further assess whether the parameter “type of NP” is indeed the less influential factor in determining the occurrence of Goal phrases headed by em in AVPs.