1. Introduction
Afrikaans has a curious grammatical phenomenon illustrated in example (1) from Van der Merwe (Reference Merwe1997:29).
This extract from a magazine discusses a small aeroplane (vliegtuigie) and shows two ways of referring to this object: by the masculine pronoun hy and by the neuter pronoun dit. The variation is puzzling: Why does the same noun trigger different pronominal gender values, even in the same utterance? The relation between masculine and neuter anaphoric pronouns in Afrikaans, and the context in which their alternation and competition can be understood, are the main concern of this article.
The variation shown in (1) is well known, and there is a body of literature available, prominently Scholtz (Reference Scholtz1966), Ponelis (Reference Ponelis1979), Van der Merwe (Reference Merwe1997), and Kirsten (Reference Kirsten2016). The present article is written with the following intentions in mind. First, it discusses Afrikaans in the context of pronominal gender systems, which are typologically special (section 2). Second, it provides an overview of the existing research on pronoun use in Afrikaans, not all of which may be accessible to interested readers (section 3). Third, a modest set of new data is presented, drawing from the NWU-Kommentaarkorpus. Section 4 of the article discusses the observed variation in the light of the typology of gender systems, showing that the issue is of wider interest and relevance. Section 5 concludes the article. The study focuses on present-day Afrikaans; the history of Dutch and Afrikaans gender are mentioned only in passing.
2. Pronominal gender
Afrikaans has a gender system of a special type, known as a pronominal gender system (Corbett Reference Corbett1991:168). As the name says, this implies that gender is marked only on pronouns, not on other potential agreement targets such as articles, adjectives, or verbs. The pattern is familiar from English, where gender is expressed only on the personal pronouns he/him, she/her, and it and the possessive pronouns his, her, and its.Footnote 1 Pronominal gender systems are non-canonical gender systems (see Corbett & Fedden Reference Corbett and Fedden2016 for criteria and discussion). Gender is poorly visible: Being limited to pronouns, it appears less frequently in an utterance or text than it would if a richer array of agreement targets were available. Where it manifests itself, pronominal gender is often non-redundant. This means that it does not repeat properties already encoded elsewhere, the way agreement typically does. It can even introduce novel information into an utterance. Think of English friend or guest, which can take masculine or feminine pronouns, thereby revealing the gender of the referent (example (2) from a post on X).
As will be shown, Afrikaans pronouns have even more expressive power than those of (Standard) English.
English aside, pronominal gender systems are not well studied. Audring (Reference Audring2008, Reference Audring2009) discusses a number of languages with this type of system, offering a tentative account of how pronominal gender manifests itself in the languages of the world. Of special interest for our purposes is the observation that pronominal genders are typically organized around basic semantic principles. This sets them apart from other gender systems, where it can be much harder to explain why a certain noun belongs to a particular gender, and the motivating factors may be phonological and/or morphological as well as semantic (see, e.g., Fedden, Guzmán Naranjo, & Corbett Reference Fedden, Guzmán Naranjo and Corbett2025 on German).
Most pronominal gender languages discussed in Audring (Reference Audring2008, Reference Audring2009) have quite straightforward assignment. Most conform to some version of the Animacy Hierarchy (Silverstein Reference Silverstein and Dixon1976), in the sense that each gender value is associated with a single section on the scale. Figure 1 shows a simple variant of this hierarchy; a more complex variant is shown in section 3.2.
A variant of the Animacy Hierarchy.
Two gender systems often split the hierarchy into animate and inanimate or into human and other. In the latter case, one pronoun is literally a personal pronoun, referring only to persons, while the other is used in all other contexts. To give a slightly more complex example, the Niger-Congo language Zande as described in Claudi (Reference Claudi1985) has four singular pronouns, one used for male persons (masculine gender), another for female persons (feminine gender), the third for animals (animal gender), and the fourth for inanimate referents (neuter gender). In the plural, masculine and feminine are merged. Again, the semantics of the system aligns with the Animacy Hierarchy.
While the prevalence of basic semantic patterns may reflect (over)simplifications in the descriptive sources, Audring (Reference Audring2008, Reference Audring2009) proposes that the pattern makes sense considering the circumstances under which pronominal genders are likely to develop. As far as the process is understood, gender arises from nouns that grammaticalize into demonstratives and personal pronouns (Corbett Reference Corbett1991:310–312). On the one hand, then, pronominal systems may be young gender systems, as pronouns are among the earliest agreement targets a language acquires. On the other hand, pronouns are known to retain gender longer than other word classes (Corbett Reference Corbett1991:143, 259). Loss of agreement markers, therefore, can also lead to pronominal gender.
Both in youth and in decline, gender systems are expected to be meaning-based, albeit for different reasons. Young gender systems should be semantic because they arise from nouns in a (semantically) classifying function. Declining gender systems, in turn, should be semantic because pronouns appear to be incapable of supporting more complex assignment systems alone (see Audring Reference Audring2008, Reference Audring2009, Reference Audring2014 for discussion). When agreement is reduced, the gender system’s survival may depend on its ability to revert to simple, meaning-based assignment.
In the light of this brief (and necessarily condensed) account of pronominal gender systems, it can be pinpointed more clearly what’s puzzling about Afrikaans. As will be shown, the principles that govern gender assignment in Afrikaans are mainly semantic. This is the expected situation. However, the language seems reluctant to settle on a clear and stable alignment between meaning and gender. Instead, usage is variable, opportunistic, and diachronically unstable. This data situation requires a more nuanced understanding of pronominal gender.
We return to this puzzle after a look at the morphology of gender in Afrikaans, and at what is known about the distribution of the pronouns.
3. Afrikaans pronominal gender
3.1 Gender in Afrikaans
Afrikaans belongs to the pronominal gender languages. This is evident from the paradigmatic oppositions that indicate the presence of gender. These are found in the third-person singular of the personal, the possessive, and the reflexive pronouns (table 1, after Donaldson Reference Donaldson1993:123, 290).
Paradigm of Afrikaans third-person singular personal, possessive, and reflexive pronouns
The demonstratives come in various forms, among others dit, dié, hierdie, and daardie. While dié and dit suggest a two-way gender split, the two forms also differ in other respects: dit doubles as a personal pronoun, while dié is used for emphasis (e.g. Donaldson Reference Donaldson1993:145). Therefore, the distribution of the demonstratives is not, or not only, a matter of gender. There is no evidence for gender on the articles, the adjectives,Footnote 4 or the relative pronoun. The difference between interrogative wie and wat is not usually considered a gender distinction (parallel to who and what in English; recall footnote 1).
For present purposes, the possessives and reflexives are set aside, as this study is mainly concerned with the choice between masculine and neuter, which are formally identical in these paradigms. Similarly, contracted forms that appear instead of a pronoun and a preposition, such as daarvan (lit. ‘therefrom’) or hiervoor (lit. ‘herefor’), are not considered, as daar- and hier- can replace any personal pronoun in inanimate reference, i.e. the masculine, the neuter, and the plural (Donaldson Reference Donaldson1993:128). This leaves the personal pronouns, i.e. hy/hom and dit, as the forms of interest.
As to the distribution of the forms, the feminine pronouns are systematically used for female persons (and occasionally for female animals)Footnote 5 while the masculine and neuter forms appear to be in competition. An example was shown in (1); two more are given in (3) and (4), also from Van der Merwe (Reference Merwe1997:29). The excerpts are from the Gospel of Luke and the Gospel of Matthew. Both contain the same allegory. In (3) the central noun huis ‘house’ is picked up by a masculine pronoun, in (4) by a neuter pronoun.
In addition, the masculine and the neuter pronouns both overlap in usage with plural hulle, as shown below.
Again, the variation in (3) and (4) is surprising given what is known of other pronominal gender languages, where the pronouns appear to be tied to basic semantic classes, leading us to expect consistent choices. While the existing literature gives no clues on how this situation arose in Afrikaans, it does provide insights into the extent of the competition between the masculine and the neuter and suggests a number of factors influencing the choice. Both are reviewed next.
3.2 Variation: extent and factors
3.2.1 Extent
The earliest accounts highlight the fact that the use of pronouns in Afrikaans is no longer related to the gender of the respective nouns in Dutch. Dit is used more extensively in Afrikaans than in Dutch; according to Scholtz (Reference Scholtz and du Plessis Scholtz1963), this is already the case in the second half of the eighteenth century. On the other hand, hy is used in more contexts than English he (Scholtz Reference Scholtz1966:123).
The variation between the pronouns is reported for a wide range of antecedents, starting from lower animals and babies and reaching all the way down the Animacy Hierarchy to abstracts and substances. While humans are normally referred to with masculine or feminine pronouns, Scholtz (Reference Scholtz1966:124) gives an example of neuter dit in reference to a baby. At the other extreme, he quotes the writer Langenhoven who shows a remarkable preference for hy, even in reference to substances such as water (5), where a neuter pronoun would be expected (Scholtz Reference Scholtz1966:126).
The distribution between masculine hy/hom and neuter dit is discussed in Scholtz (Reference Scholtz1966), Ponelis (Reference Ponelis1979), Van der Merwe (Reference Merwe1997), and Kirsten (Reference Kirsten2016).Footnote 6 Scholtz covers literary language, while Ponelis adds observations about colloquial speech. Van der Merwe provides a more systematic investigation, based on print media and a questionnaire, though the report remains largely qualitative. Kirsten’s dissertation is the first study based on (written) corpus data and including quantitative results.
3.2.2 Semantic factors
All scholars agree that the most evident factor in pronoun choice is, as expected, semantics. In particular, the following properties are identified as relevant:
Animacy
Individuation (individuality, uniqueness)
Agentivity (action, movement or resistance against movement; Scholtz Reference Scholtz1966:131)
Concreteness
Countability (bounded shape/countable – mass/uncountable)
The association of animate referents with masculine and inanimate referents with neuter pronouns is already mentioned in the earliest sources (Bouman & Pienaar Reference Bouman and Pienaar1924:179) and is considered the primary or basic function of the two pronouns (Scholtz Reference Scholtz1966:125). Individuation and agentivity can be seen as extended manifestations of this basic split: (higher) animates are more likely to be perceived as unique individuals than objects or substances, and they are more likely to be moving or acting entities.
Concreteness and countability are splits within the inanimate domain. They are often discussed as joint properties, but they can in fact cross-cut one another, as shown in table 2 with four example words.
The orthogonality of concreteness and countability
Ponelis (Reference Ponelis1979:586–590) develops a similar classification, with an additional class of pure abstracts (“suiwer abstrakta”), which is not defined but appears to comprise nominalized actions and events. He states that the use of masculine hy decreases steadily from concrete to abstract and from count to mass, with neuter dit as the only choice for the pure abstracts. As an alternative, he notes that dit is available for all inanimate referents.
All of these factors can be insightfully combined with the help of an enriched variant of the Animacy Hierarchy as given in figure 2 (also called the Individuation Hierarchy; cf. Sasse Reference Sasse, Joachim Jacobs, Sternefeld and Vennemann1993:659; Audring Reference Audring2009; De Vos, De Sutter, & De Vogelaer Reference De Vos, De Sutter and De Vogelaer2021). This hierarchy has additional subclasses within the inanimates, based on concreteness and countability.Footnote 7 Masculine pronouns are more likely to be found towards the higher end of the hierarchy (left), neuter pronouns towards the lower end (right).
An extended Animacy Hierarchy.
While individuation and agentivity are not represented in the figure, they are clearly associated with the higher end of the hierarchy, although objects and abstracts can be promoted to agents (this is sometimes interpreted as personification). (6) shows an example from Scholtz (Reference Scholtz1966:134). As expected, the pronoun reflects this reading by being masculine.
Example (6) is particularly instructive, as it shows that the placement of a certain referent on the hierarchy can depend on the construal of the situation by the speaker. This opens the door to a second major factor in pronoun choice, which might be subsumed under “pragmatics,” understood in a broad sense.
3.2.3 Pragmatic and stylistic factors
There is general agreement in the literature that semantic factors do not fully explain the distribution of the masculine and the neuter pronouns. Pragmatic and stylistic factors also play a role, which can be roughly sorted into attitude, register, and personal preference.
With regard to attitude, studies report an increase in masculine pronouns whenever the speaker wishes to express emotional involvement, be it positive or negative. Van der Merwe (Reference Merwe1997:29) reports masculine pronouns for huis ‘house’ and wyn ‘wine’ in contexts of appreciation, but also cites a student saying that he or she uses masculine pronouns “[a]s iets byvoorbeeld ‘n probleem veroorsaak” (‘if, for example, something causes a problem’). The neuter, in turn, is used when speakers wish to be objective, distant, or disapproving (the latter means that both pronouns can be used to express negative attitude). Scholtz (Reference Scholtz1966:131) summarizes similar observations in stylistic terms: a calm and sober style goes with neuter gender, while a more dramatic, lively, and animated style favors the masculine.
For Ponelis, the difference between “saaklik” (factual) and “lewendig” (lively) is accompanied by a difference between formal or written and informal or spoken language. He characterizes dit as the more formal and conservative choice, while hy is preferred in informal registers and has a more modern ring to it. In fact, he interprets the choice for hy in reference to objects in colloquial speech entirely in terms of register, denying that it has any affective value at all (1979:590). Generally, he claims an “aggessiewe opmars,” an aggressive advance, of masculine hy (1979:585), which the later sources Van der Merwe (Reference Merwe1997) and Kirsten (Reference Kirsten2016), as well as the data presented in the next section, fail to confirm: hy remains, or has developed back into, the marked choice.
Furthermore, pronoun use can be a matter of personal preference. This is pointed out by Scholtz (Reference Scholtz1966), who describes individual writers as strong hy- or dit-users, with others showing less marked tendencies. Interestingly, Van der Merwe reports a skewed distribution: a third of the 50 respondents to her questionnaire consistently opted for dit, while nobody used hy throughout (1997:28). This supports the view that hy is marked. However, we may also see an effect of prescriptivism: some students (and their teachers) appear to regard dit as the only correct choice for inanimate referents in Standard Afrikaans (Van der Merwe Reference Merwe1997:27). It’s not entirely clear what this judgement is based on. Carstens (Reference Carstens2013:73), for example, treats both hy and dit as the norm for inanimates.
A final factor of influence disregarded here is regional or dialectal variation. The interested reader is referred to Du Plessis (Reference Du Plessis1984) and Spruyt (Reference Spruyt1993); both mention Griekwa-Afrikaans especially.
What is lacking in the literature is a systematic account of pronoun choice in colloquial speech. While the ideal data source, a corpus of spoken Afrikaans, is not (yet) available, the Virtuele Instituut Vir Afrikaans, VIVA (https://viva-afrikaans.org/), offers the NWU-Kommentaarkorpus, a collection of unedited user comments to news and discussion websites. This resource is probably the closest approximation to spontaneous language currently available. The next section shows what this corpus tells us about contemporary pronoun usage.
3.3 New data
The NWU-Kommentaarkorpus is available with special permission through the VivA corpus portal. Version 1.4.1, which I had at my disposal, contains data from the year 2020. For the purpose of the present article, a section of 214,500 words was searched for the personal pronouns hy, hom, and dit in anaphoric use.Footnote 8 All examples given in the present section are from the corpus.
The search yielded 2,772 instances of dit and 1133 instances of hy/hom. This set had to be filtered manually in order to extract, in order of sequence,
(a) the personal pronouns,
(b) the pronouns used as anaphors with a nominal antecedent/referent, and the antecedent itself,
(c) the pronouns referring to inanimate entities.
Filtering condition (a) was especially relevant for hom, as this form doubles as a reflexive. Condition (b) excluded a great majority of the neuter pronouns, as dit is used in a wide variety of functions. Most instances of dit were dummy pronouns in copula constructionsFootnote 9 or referred to entire propositions. These uses were not considered, as dit has no competitor here (but see footnote 13). Ambiguous instances such as (7) were retained because they do offer some degree of choice. In (7), dit can refer to the proposition ‘to earn respect’ but also to the noun respek itself. There were 13 such cases in the data set.
Condition (c) also had a great impact on the size of the data set, as the masculine pronouns overwhelmingly referred to humans. In six cases, the referent was an animal of unspecified gender, while two instances occurred with the noun liggaam ‘body’. These cases were excluded. The filtering eventually yielded 301 instances of dit and no more than 36 instances of hy/hom.
The antecedents were classified according to their semantics (pragmatic categories were too difficult to operationalize, and the data was relatively homogeneous in terms of liveliness). The categories were inspired by the literature (see section 3.2.2), but had to be simplified due to the small number of masculine tokens. Also, sorting abstracts into countable and uncountable proved to be difficult in practice. On the other hand, three categories needed to be added because they were attested with some frequency: collective/place, event/activity, and plural/coordinate. The following examples illustrate the groups.
The merged groups were based on conceptual similarity. Collectives/places are entities occupied or formed by groups of humans; they were included because they can also be interpreted as abstract bodies. Events/activities have a verbal character and often a temporal dimension (this class is reminiscent of Ponelis’ “pure abstracts” mentioned in section 3.2.2). Plurals and coordinates go together for another reason: here the singular pronouns enter into the domain of the plural, where the competitor is the third-person plural pronoun hulle.Footnote 10
The results are given in table 3; figure 3 visualizes the numbers in a bar chart.
Distributions of pronouns over semantic classes (number of tokens)
Distributions of pronouns over semantic classes (visualization).
The results show a number of clear tendencies. First, despite the relative informality of the material, masculine pronouns are rare if reference to male humans and animals is excluded. Where they appear, they occur with object nouns such as vlag ‘flag’, masker ‘mask’, glas ‘glass’, minibus ‘id.’, or with nouns denoting collectives such as regering ‘government’, staat ‘state’, or land ‘country’. In these contexts, the masculine almost draws level with the neuter.
In the object class, the sample contains one instance where the same referent is picked up by pronouns of different genders (8).Footnote 11
Among the collectives, the greatest variation can be seen with the antecedent land ‘country’. Examples (9) and (10) show an instance with a masculine and a neuter pronoun, respectively.
In reference to abstracts, there are hardly any masculines. One appears after the noun siekte ‘illness’; the other three tokens are all questionable: two refer to leuen ‘lie’ and appear in a proverb (the same proverb in both cases), the third refers to spyt ‘regret’ but is used in a poem, i.e. in literary language. For all the other classes, only neuter pronouns were found. Example (11) gives an example from the class “event/activity.” Some of the antecedents in this class could also be classified as abstracts (zwangerschap ‘pregnancy’, verandering ‘change’). This does not change the overall results, as both classes came with neuter pronouns exclusively.
Two additional observations are noteworthy. The first is the comparatively large number of abstract referents. This may be due to the popularity of two topics: politics (with nouns such as apartheid ‘id.’, korrupsie ‘corruption’, or oorlog ‘war’) and the Covid-19 pandemic.Footnote 12 The second striking result is the high percentage of neuter pronouns for plural and/or coordinate antecedents. Note, however, that this use also belongs to the functions of the neuter in other Germanic languages (see, e.g., Romijn Reference Romijn1996 on Dutch) and is probably no innovation.
Summing up, the data in this corpus suggest that dit is the unmarked and preferred choice across the board. There are two areas of competition: the objects, where the masculine competes with the neuter, and the collectives, where both pronouns compete with one another and with plural hulle. Generally, the frequency of masculine pronouns was too low to allow any further inferences.
4. Discussion
The results from the Kommentaarkorpus are very similar to Kirsten’s (Reference Kirsten2016) and confirm her observations. She concludes that “hy and hom are increasingly reserved for male referents, with a decrease in inanimate reference” (2016:189, my translation). The findings contradict Ponelis’ account from around forty years earlier, which partly reports, partly predicts a growing invasion of the masculine.
What sense can be made of these observations? Before trying, two caveats are worth pointing out. On the one hand, Ponelis, who had no quantitative data at his disposal, may have overestimated the frequency of the masculine, which as an innovative choice must have attracted attention. On the other hand, the Kommentaarkorpus still contains written data, however informal, and may not be representative of colloquial speech. If, however, the observations are correct and Afrikaans has witnessed first an expansion of the masculine at the expense of the neuter, followed by the reverse pattern, it is interesting to consider how this might have happened. To this aim, it is helpful to pull apart two conceptual metaphors for the forces underlying the relation between grammatical gender and semantics.
On the one hand, linguists usually conceptualize gender values as mapped onto semantic classes. The mappings that capture the usage patterns in Afrikaans can be schematized as in figure 4 (the feminine is now included).
Extended Animacy Hierarchy with gender values in Afrikaans.
The arrows indicate the associations; I call them “vertical” forces for the purpose of the discussion. This model does not explain why some classes show more variable gender choices than others: the forces should be equally strong for all classes.
Here is where another conceptual model is helpful, involving what could be called “horizontal” forces. “Horizontal” forces arise from opposing properties. Figure 5 sketches the model, with an example set of properties inspired by section 3.2. In Afrikaans, the left-hand list of properties “pulls” towards the masculine gender, the right-hand list attracts the neuter gender.
Opposing properties linked to gender choice.
While it seems a little artificial to spell out our conceptual model in such detail, it helps to understand the situation in Afrikaans. The “horizontal” forces in particular can account for the locus of stability and the locus of variation.
All empirical studies report a stable alignment between semantic class and gender at the outer ends of the Animacy Hierarchy: the left-hand side is the domain of the masculine (and the feminine), the right-hand side is the domain of the neuter. This makes sense, as the properties cluster together at the ends of the scale. Humans are animate, agentive, individual, concrete, etc., while uncountable abstracts are inanimate, patientive, and so on.Footnote 13 Hence, the “horizontal” forces are in agreement and lead to consistent choices of gender values.
Variation, on the other hand, is expected in two situations: for referents belonging to more than one semantic class, and towards the middle of the scale. The first situation applies to the collectives, which fit both the human and the abstract section of the Animacy Hierarchy, as they can be construed as a group of persons or as an abstract entity. Depending on the construal, masculine (and plural) or neuter pronouns are used, respectively.
The second situation applies to the class of objects. Objects are inanimate, but represent concrete and potentially individual units; they can sometimes move or resist against movement etc. Hence, they show properties pointing in opposite directions, again with variable outcomes in gender choice.
As a consequence, the middle section of the hierarchy appears to be a vulnerable area, where pressures of any kind may sway the system. What forces may have caused the expansion of the masculine in Ponelis’ time can only be guessed at. For the expansion of the neuter, however, there is an obvious candidate: English. English is the primary language of contact, with bilingualism as the norm (see, e.g., Coetzee-Van Roy Reference Coetzee-Van Rooy2013). It also has a pronominal gender system very much like Afrikaans, only with a larger domain for the neuter (figure 6).
Extended Animacy Hierarchy with gender values in English.
If English indeed has an influence, the greatest effect can be expected in the domain of object reference, where (a) the two systems differ, (b) Afrikaans shows the greatest variation, and (c) we are in the middle section of the hierarchy, which may be most vulnerable to change. Thus, the argumentation has led us to a point where the instability of Afrikaans pronominal gender is not as surprising as it seemed. Especially towards the middle of the hierarchy, external and internal forces, from expressive intentions to pressures from the contact situation, may find little resistance.
A remaining difficulty is intra-sentential variation as in (1) and (8). In (1), the masculine pronoun precedes and the neuter pronoun follows the noun. In (8), the masculine pronoun stands closer to the noun than the neuter pronouns. However, how this might matter here is not easily explained. Distance and position effects are known to influence agreement choices, but the relevant accounts either appeal to a difference between syntactially versus semantically motivated choices (Corbett Reference Corbett1979) or to different degrees of semantic or pragmatic motivation (Corbett Reference Corbett2023). Neither seems to apply here, so the examples may merely reflect fluctuations in usage.
5. Conclusions
This article discusses the competition between masculine and neuter anaphoric pronouns in Afrikaans. The fact that masculine hy/hom and neuter dit are used interchangeably in various contexts is interesting and unexpected given what is known about pronominal gender languages: such systems are usually organized according to simple semantic principles, often aligned with the Animacy Hierarchy. Afrikaans, however, seems reluctant to settle on a uniform pattern with consistent choices.
The article reviews the available literature and reports additional data from a corpus of unedited user comments to news sites. Competition is mostly found in reference to objects and to collectives, where masculine and neuter pronouns appear with roughly the same frequency.
It is argued that the Animacy Hierarchy, enriched by “horizontal forces” of a pragmatic nature, helps to make sense of this pattern. Variation is found in two situations: where a referent can be anchored to different points of the hierarchy, and in the middle section of the hierarchy where the “pull” forces are weakest. This explains the variation for collectives and for objects, respectively. Pronoun use for referents at the outer ends of the scale, that is in reference to humans and animals, or to abstracts and masses, shows the greatest stability.
By providing an occasion to sharpen our conceptual metaphors for the relation between grammatical properties and semantic classes, the puzzle of Afrikaans pronominal gender is potentially instructive for the analysis of other phenomena and other languages. Hopefully, this article serves as an inspiration to further explore the applications and implications.
Open access
