1.1. Sociohistorical background

The role of the Portuguese language in Africa was limited during the first centuries that followed the expansion of the Portuguese seaborne empire from the 15^th century onward, mainly because Portuguese settlers were a minority whose economic activity was mostly restricted to the coastal areas and islands. Despite the steady growth of the Portuguese presence over time, for example in the Luanda area, the independence of Brazil (1822) and the Conference of Berlin (1884–1885) brought about the need to further occupy and exploit the continental spaces in an effective manner. The expanding colonial presence in Angola and Mozambique in particular also meant a greater presence of Portuguese in the 20^th century, through administration, economic activities, and education (e.g., P. Gonçalves, 2010, 2013; Inverno, 2011), even though it remained to be the language of a mostly privileged minority.

The role of Portuguese was clearly reinforced after the independences, when Portuguese became the only official language of the new countries, supported by factors such as democratization of education in this language, increased social mobility, and language usage in the media. While Portuguese was predominantly an L2 language before the independences, in the subsequent period it quickly became more widespread, with an increasing number of L1 speakers, primarily in urban areas of Angola, Mozambique, and São Tomé and Príncipe.

With respect to Angola, according to estimates provided by Endruschat (1990), after the exodus of most of the metropolitan population in 1975, which represented some 10% of the total population in 1973, only 1 or 2% of the Angolan population spoke Portuguese natively, whereas some 15 to 20% were L2 speakers. Half a century later, according to the 2014 national census (INE-A, 2016), Portuguese was the most spoken language at home for 71,15% of the population, a number that rises to 85% in urban areas. The combined numbers of the other languages spoken at home (63,5%) further suggests that nowadays a substantial percentage of the Angolan population is monolingual in Portuguese.

According to the Mozambican censuses, the percentages of L1 Portuguese speakers rose from 1,2% in 1980 to 16,6% in 2017, whereas the combined percentages of L1 and L2 Portuguese speakers in the same period went from 23,2% to 58,1% (INE-M, 2019). Again, it is in the urban environment that Portuguese is more often used at home. The number of Portuguese speakers has grown at the expense of the Bantu languages, whose L1 speakers represented 83,3% of the total in 2017 (INE-M, 2019), against 98,8% in 1980. The slower spread of Portuguese in Mozambique can be assigned to differences in the colonization process, which was less intensive and widespread than in Angola.

The situation of São Tomé and Príncipe is somewhat different from the previous two countries, since roughly 63% of the population declared to speak Portuguese in the 1980 census, a percentage that rose to 98,4% in the 2012 census (INE-STP, 2013), without distinguishing between L1 and L2. The fact that in the latest census only about half of the Santomean population declared to speak one of the creole languages confirms that Portuguese in São Tomé and Príncipe is now predominantly an L1 variety, which is a unique situation for a former colonial language in Africa.

Growing nativization of the former colonial and exclusive official language in these three countries can be assigned to common and country-specific factors. Among the common factors are the neutral and overarching status of Portuguese when compared to the linguistic and geographic dispersion of the coexisting indigenous languages, the almost exclusive role of Portuguese in administration and education, and the insufficiency of language policies regarding the non-official languages. These factors promoted the prestige of Portuguese as the language of upward mobility and its spread as a lingua franca.

However, country-specific aspects should also be factored in. In Angola, the significant Portuguese presence in the 20^th century, up until the independence, especially in urban areas, created long-lasting roots for Portuguese (Inverno, 2011). This presence was further boosted by the intense armed conflict following the independence (1975–2003), which led to an exodus of Bantu-speaking populations from the rural areas to the cities, in particular the capital Luanda, where Portuguese already functioned as a medium of interethnic communication between speakers with a linguistic Bantu background.

In Mozambique, the Portuguese presence was more limited, both in numbers and geographically, and the armed conflict following the independence was somewhat less destabilizing than in Angola. In addition, the independence movements in Angola and Mozambique used Portuguese as their primary means of communication. It was readily declared the exclusive official language after 1975, although language policies favoring Bantu languages have been established since the 1990s (Chimbutane, 2018).

Finally, in São Tomé and Príncipe, a small island state which had been effectively settled and occupied since the late 15^th century, creole languages Angolar, Forro, and Lung’Ie were predominantly spoken until the late 19th century. However, the abolition of slavery in 1875¹ led to labor shortage in a period of booming coffee and cacao cultures. To overcome the labor crisis, the Portuguese created the Curadoria Geral dos Indígenas, a labor bureau through which large numbers of indentured laborers (serviçais or contratados) were recruited mainly from Angola, Cape Verde, and Mozambique. The demographic impact was so massive that this new population quickly outnumbered the islands’ native population (Bouchard, 2019; Hagemeijer, 2018). Since the Forros, the native population, were deliberately kept segregated from these contractors and their descendants, the so-called Tongas, the newly arrived population on the islands mostly adopted Portuguese as an L2 and lingua franca, to the detriment of the creole languages (Hagemeijer, 2018). As in the case of Angola and Mozambique, the consolidation of Portuguese in São Tomé and Príncipe is therefore a 20^th century phenomenon, further boosted after the independence.

In sum, common and country-specific sociohistorical factors explain the rise of Portuguese in Angola, Mozambique, and São Tomé and Príncipe. The process of nativization capitalizes on an increasingly urban, schooled population for whom Portuguese becomes the primary or dominant language, which is then transmitted to the following generation. All being equal, the prediction is therefore that the numbers of L2 and L1 Portuguese speakers will steadily continue to grow, whereas the coexisting, non-official languages will be gradually absorbed in the process of language shift. While in the past Portuguese was primarily the language of a minority including colonizers, local elites, and those who assimilated to the colonial regime, it has now become the language of the masses.

1.2. Research and data

The colonial period did not produce meaningful scientific research on the development of Portuguese in Africa. Since the 1980s, however, African varieties of Portuguese (AVPs) have become the object of a growing body of descriptive and theoretical work, which focuses mainly on the properties of these new linguistic realities, mostly in comparison with European Portuguese (EP), the official standard in the former Portuguese colonies in Africa, and on the factors that underlie language variation and change.

During the first decades following the independences, linguistic research on AVPs concentrated mostly on the major countries, Angola and Mozambique, whereas the other former colonies only have come into the picture more recently, especially after the turn of the millennium.² There is also still a notable unbalance with respect to the linguistic domains being investigated, with a stronger focus on lexicon (e.g., loanwords), syntax (e.g., clitics, dative objects) and morphosyntax (e.g., agreement patterns), and comparatively less research on phonetics and phonology.

The data sets used for the study of AVPs are quite varied, but a primary distinction should be made between written and spoken data.

Written data have been used in several forms. Especially in the case of Angola, several scholars (e.g., Chavagne, 2005; Mingas, 2000) have made (partial) use of literary work by authors who seek to reproduce informal, spoken language use as a stylistic device, such as Luandino Vieira, who gives couleur locale to his characters by introducing features of the speech of the musseques (slums) of Luanda.³ Some studies are based on originally handwritten materials, such as essays written in an educational setting by children or adults (e.g., Afonso, 2008; Cabral, 2005; Lima Afonso, 2009).

Most studies on the grammar of AVPs, however, privilege the use of spoken data, such as interviews with speakers, in order to study patterns of variation and change. These data should, ideally, correspond to more demanding processes that lead to the constitution of corpora (recordings, transcriptions, annotation, etc.), but sometimes they also include native intuitive data or non-recorded, overheard data. Some examples of larger spoken corpora are the corpus of the project Panorama do Português Oral de Maputo (Stroud/P. Gonçalves 1997–2000), Corpus África (e.g., Bacelar do Nascimento et al., 2008), and the PALMA corpora (Hagemeijer et al., 2022a).⁴ Other corpora have mostly been collected within the scope of doctoral research (Chavagne, 2005; Inverno, 2011; Jon-And 2011, a.o.) and are generally not publicly available.

Experimental research involving grammaticality judgments has also gained some traction (e.g., Justino, 2011; Mapasse, 2005) and can be especially useful to confirm grammatical tendencies that have been previously established within a given (sub)variety.

Most of the research on AVPs focuses on urban varieties, in particular those spoken in the countries’ capitals, where L1 and L2 Portuguese is typically more widespread, but a growing body of work centers on other regions or locations, for instance Lubango, Angola (Cabral, 2005); Almoxarife, São Tomé (Figueiredo, 2010), and Dundo, Angola (Inverno, 2011).

Irrespective of the descriptive or theoretical framework employed, as a rule of thumb studies on AVPs seek to describe and analyze features that are divergent from standard EP or hint at a possible continuum with other varieties, in particular Brazilian Portuguese, due to the role that the Atlantic slave trade played in the (linguistic) making of new societies (e.g., Avelar & López, 2018; Brandão & Vieira, 2018; Petter, 2008).

1.3. Language variation

Since AVPs have their roots in second language acquisition and therefore rely on a wide array of (socio)linguistic variables, they exhibit a large spectrum of variation in many domains of grammar. For example, with respect to Mozambique, P. Gonçalves & Chimbutane (2004, p. 7) acknowledge the existence of basilectal subvarieties whose speakers have little exposure to EP (or Portuguese in general), as well as subvarieties of educated speakers whose language use is closer to the standard. In addition to the more widely acknowledged and investigated role of interspeaker variation, P. Gonçalves (2013, p. 161) also emphasizes the role of intraspeaker variation, which remains an understudied domain:⁵

“[t]his variability is evident both in the somewhat irregular and unsystematic way in which these features manifest themselves in the linguistic production of speakers […], and in the frequent lack of coherence between speakers’ oral or written productions and their grammaticality judgments when subjected to experimental procedures, in particular grammaticality tests.”⁶ [my translation]

The research on AVPs has traditionally focused on features that diverge from (standard) EP, but the methodologies and the underlying data (cf. previous section) that are used are highly variable, ranging from impressionistic to full-blown quantitative studies. The latter type of studies shows that at least in the domain of (morpho)syntax AVPs often still exhibit substantial convergence with EP regarding specific features, as shown by studies on number agreement (e.g., Brandão, 2013), dative structures (P. Gonçalves, 1991; R. Gonçalves, Duarte & Hagemeijer, 2022, in this special collection), or the expression of Goal arguments (Hagemeijer et al., 2022b, in this special collection).

Therefore, the distinct morphosyntactic patterns found in AVPs are generally indicative of language variation, but do not necessarily reveal language change, in the sense that it is not necessarily the case that new or innovative patterns have become privileged or even fully crystallized. It is of course often unknown how certain tendencies compare to the past, for which often no quantitative information is available, and how they will trend in the future. One exception are two recent papers on Mozambican Portuguese (MP) (R. Gonçalves, Duarte & Hagemeijer, 2022, 2023), which conclude that the VP-internal changes discussed in older work by P. Gonçalves (1991) have apparently not become more generalized.⁷ The role of nativization and schooling also needs to be factored in, as this may ultimately favor standard uses.

Therefore, regarding the large spectrum of variation that characterizes AVPs, intrinsically associated with its L2 or recent L1 status in multilingual settings, the availability of detailed sociolinguistic information, quantification, including statistical treatment, and research on a broad range of subvarieties are aspects that will allow for more fine-grained research and results.

1.4. Language contact

Given its pluricentric nature, Portuguese has been in contact with a wide range of typologically different languages in different parts of the globe. In Africa, it has been historically in contact with languages from different branches of the Niger-Congo language family, most prominently with Bantu languages spoken in Angola and Mozambique, and with creole languages in Cape Verde, Guinea-Bissau, and São Tomé and Príncipe. As a consequence, language contact has been argued to play a prominent role in the shaping of AVPs, in the sense that it is widely assumed that grammatical features from the speakers’ L1 languages have been transferred to L2 varieties of Portuguese.

The most elaborated version of the contact-induced hypothesis has been proposed in the work by P. Gonçalves (e.g., 1991, 2002, 2010), where it is argued that L1 Bantu parameter settings carry over to the initial stages of L2 acquisition of MP, with adjustments over time in light of ambiguous properties in the Portuguese input and insufficiently robust primary linguistic data. P. Gonçalves underscores this claim with case-studies on urban MP, which is primarily in contact with the Bantu language Changana (S53, within the Tsonga cluster). A case in point is the occurrence of double object constructions with Recipients (IO-DO) in this variety, whereas EP exhibits prepositioned IO (DO a IO). New parameter settings arguably trigger further changes, for instance dative passives in MP.

While certain features attested in AVPs show all signs of being triggered by language contact, these features are typically a subset of the variation and generally not the full picture in light of the available empirical data. Moreover, the presence of contact-induced features tends to correlate with sociolinguistic variables and will be more prominent in subvarieties of lesser schooled speakers who are less exposed to Portuguese. The use of defective copies found in PP-relativization in Santomean Portuguese (STP) is an example of a highly specific contact-induced feature related to Forro, where this is the standard pattern, but its distribution is very limited, especially in comparison with pied-piping and preposition-chopping, which are not available in Forro (Hagemeijer, R. Gonçalves & Alexandre, 2024).

Finally, the traditional assumption was that the AVPs are L2 varieties, which favored the playbook of language contact. However, the situation is currently much more complex, with monolinguals in Portuguese and large numbers of bilinguals (in the broad sense of the term) with different degrees of proficiency in the languages they speak (P. Gonçalves, 2012). With respect to these bilingual speakers only information gathered through detailed sociolinguistic questionnaires (such as the ones sometimes used to assess the linguistic proficiency of heritage speakers) will be able to place speakers on a spectrum of proficiency. Since the metadata of the PALMA corpora only include the languages that informants declared to speak, a split between monolinguals and non-monolinguals in Portuguese could be applied, but it is not expected that these two groups perform differently (inasmuch as the other social variables are held constant).

1.5. The role of general principles of grammar

In some work it has been argued that the direction of change should be assigned to general properties of the grammar of Portuguese, leading to convergent paths of restructuring, i.e., the emergence of similar structures across different varieties.

Research on PP-relativization provides perhaps one of the clearest examples of such a tendency, since it has been established that in all varieties of Portuguese, including EP, pied-piping of the preposition, the canonical strategy, is increasingly competing with preposition-chopping and, to a lesser extent, with resumption and defective copying (e.g., Alexandre, 2000; Hagemeijer, R. Gonçalves & Alexandre, 2024; Tarallo, 1985). The common ground between these different non-canonical or substandard strategies is that they privilege the invariable relativizer que to the detriment of relativizers which exhibit phi-features (qual, cujo, etc.). In the case of AVPs, this change cannot be assigned to the role of language contact, since the main contact languages exhibit distinct strategies of PP-relativization, but crucially not preposition-chopping.

There are several other domains of grammar where some degree of convergence between varieties of Portuguese has been observed, in particular the tendency toward losses in the functional domain, including number marking, accusative clitics, and weak semantically opaque prepositions (e.g., a). However, since this tendency does not affect EP, it is suggestive that general principles that guide second language acquisition may have played a meaningful role, since AVPs and in part also Brazilian Portuguese have been historically shaped by second language learners. Also, with respect to these features, a contact-induced hypothesis cannot be upheld because, as in the case of PP-relativization, the main contact languages exhibit diverging strategies.

1.6. The role of schooling

The research on AVPs has quite unanimously led to the conclusion that schooling is a decisive sociolinguistic variable promoting greater convergence with the standard language (EP).

In the earlier work on MP, in the 1990s, several Mozambican scholars, whose work is summarized in Moreno & Tuzine (1997, pp. 71–72), emphasize the role of schooling as a primary factor with regard to access to standard Portuguese. In their chapter, based on quantified data from the corpus of the pioneering project Panorama do português oral de Maputo [Panorama of spoken Maputo Portuguese], Moreno & Tuzine (1997) show that non-standard features in the domain of (morpho)syntax are much more prevalent in lower educated speakers.

The sociolinguistically oriented work on MP of the 1990s also highlights the importance of variables which strongly correlate with schooling, such as socioeconomic class and the urban environment, as indicators of exposure/access to and use of norms that lean closer to the standard. The terms ‘basilectal’ and ‘acrolectal’ have been used to describe the spectrum of variation which encompasses the subvarieties of those speakers who have little exposure to standard Portuguese and those who use subvarieties closer to EP, typically schooled, urban speakers with high socioeconomic status (P. Gonçalves, 2010; P. Gonçalves & Chimbutane, 2004). Moreover, it is among (bilingual) speakers with this profile that Portuguese tends to become the primary (or dominant) language in Mozambique (P. Gonçalves, 2012).

Subsequent research on MP, and also on other AVPs, has confirmed the pivotal role of schooling. This has been demonstrated in, for instance, variationist studies on number marking in the NP (e.g., Brandão, 2013; Jon-And, 2011; Pessela, 2020) and in the domain of (accusative) object clitics (STP) (R. Gonçalves, Duarte & Hagemeijer, 2023). Quantitative research (R. Gonçalves, Duarte & Hagemeijer, 2023; Miguel, 2003) and also experimental work (Mapasse, 2005) shows that more schooling correlates with greater use of accusative clitics.

Note, however, that the degree of schooling does not always appear to be the main factor driving variation. Research on rhotics (<r, rr>) in STP shows that variation is primarily age-related (Bouchard, 2017; Pereira, Hagemeijer & Freitas, 2018). In fact, it is quite possible, and in need of further research, that the effect of schooling is stronger in the domain of (morpho)syntax than in the domain of phonetics and phonology. While on the one hand language attitudes appear to favor EP grammar (i.e., syntax, morphology, morphosyntax) (e.g., P. Gonçalves, 2005), on the other hand it has been observed by several scholars that EP pronunciation can be perceived as awkward or snob (Gerards & Meisnitzer, 2024; P. Gonçalves, 2005).

Schooling also increases metalinguistic awareness and the capability to use different registers according to social context. Ultimately, it may be the case that populations which are overall more and better schooled and the growing number of L1 users of Portuguese or speakers whose primary language is Portuguese increasingly counter the (historical) role of language contact.