Generally, a declarative sentence may begin with a simple, compound or a complex nominal subject. A simple subject may comprise a single nominal, a compound subject may contain two nominals and a complex subject may include three or more nominals. In Bantu languages, the nominals may belong to similar or different noun classes. In Xitsonga, also a Bantu language, when the compound or complex subject is composed of lexical items belonging to the same noun class, agreement morphology replicates such noun class, creating a relationship of verisimilitude between them. However, where members of the compound or complex noun subject are lexical items from different noun classes, the generic agreement morpheme or marker swi-, or its variants swa- or swo-, takes precedence (PanSALB 2019) (see Box 1 for a typical example) Box 1.

In Box 1 the generic agreement morpheme swi agrees with the compound subject ximanga na kondlo [a cat and a mouse], where ximanga is classified under noun class 7, which takes the agreement morpheme xi, and kondlo under noun class 5, which assumes the agreement morpheme ri. The use of any of the two agreement morphemes, that is xi or ri, will result in a grammatically deviant sentence (see Box 2).

The aim of this study was twofold: (1), to determine the noun class associated with generic agreement morphology in Xitsonga; and (2) to find out whether such noun class and its agreement morpheme can be regarded as the basic class and basic morpheme, respectively. It is hypothesised that in line with the markedness framework, there is a basic noun class and an agreement morpheme that can be considered unmarked unlike other noun classes and their agreement morphemes. The study therefore elucidates several core questions:

By examining the interface between subject noun class agreement and generic morphology, the study opens a debate not only about subject agreement in Bantu languages but also, more importantly, generic morphology. However, currently, the methodological framework provides a compelling foundation for the study’s analytical trajectory.

Data were collected and analysed through the autoethnography narrative design, which encompasses three dynamics: ‘auto’ (or self), ‘ethno’ (culture) and ‘graphy’ (the act of writing) (Belbase, Luitel & Taylor 2008:88). Autoethnography, which is grounded in postmodern philosophy and permits the gathering and analysis of emergent data (Munro 2011:162), denotes a textual representation of a researcher’s own personal experiences and understanding (Belbase et al. 2008:88). Stated differently, it allows a personalised style of writing, where the researcher draws on their experience to extend understanding about the issue in question (Wall 2006:146). According to this view, autoethnography is a method of research, which is characterised by concrete actions, self-consciousness, introspection and self-critical reflection (Belbase et al. 2008:95). It focuses on fluidity, intersubjectivity and responsiveness to particularities and shows a commitment to refigure, refashion and question normative discourses (Adams & Jones 2010:197). Moreover, autoethnography examines the dialectics of subjectivity and emphasises the various elements of the self, the socio-cultural dimension and the research process and its representation (Allen-Collinson & Hockey 2008:212). In this study, the emphasis is on the use of self as a source of data and the subsequent reflective account of engaging with noun class 8 and agreement morphology. The markedness theory underpins the analytical framework of this study and is examined in the subsequent section.

The term markedness originates from phonology to describe ‘the central observation that not all elements in a phonological system are of equal status’ (Rice 2007:79). Generally, markedness postulates that in the languages of the world, there are certain linguistic elements that are more basic, natural and frequent than others (Yan-giu & Feng-juan 2015:667). Three types of markedness are identified: formal, distributional, and semantic. This study draws on the concept of semantic markedness in which one of the lexical items is described as unmarked because it has less specific meaning than the marked member (Yan-giu & Feng-juan 2015:669).

According to Andersen (1989:15), a semantic relation occurs where one of the terms is opposed to the other while subsuming it. A pertinent instance can be found in the words mbuti [goat] and munhu [person], which are unmarked. When these words are compared to their respective hyponyms, xiphongo [a buck] and wanuna [a man], they are less semantically specific. In other words, while mbuti and munhu function as superordinate terms referring broadly to goats and people, respectively, xiphongo and wanuna serve as marked lexical items, denoting specifically the male goat and male person. It is anticipated that the theoretical framework will facilitate the identification of the noun class linked to generic agreement morphology and further enable an assessment of whether both the noun class and its associated agreement morpheme may be construed as unmarked in relation to other noun classes and their respective agreement markers. In the next section, the nature of noun classes is examined.

In markedness, the unmarked member possesses a more general and neutral meaning and is more widely used than the marked one (Yan-giu & Feng-juan 2015:669). The marked member therefore tends to be more restricted in distribution, that is in the range of contexts in which it occurs, than the former (Andersen 1989:13). However, it is noted that differences in distribution are a matter of grammar. The words mbuti and munhu are, for example, construed as positive and unmarked because they are more used in questions than their apparent hyponyms xiphongo and wanuna. The latter is deemed negative and marked as they are mainly used in certain circumstances as a result of their limited meanings. Prior to engaging with generic morphology, it is essential to examine the noun class system and its associated agreement morphemes in Xitsonga.

Table 1 illustrates noun classes, their prefixes, examples and their respective agreement morphemes in Xitsonga.

It is important to note that several ‘noun classes’, which appear in Golele’s (2013) classification, have been omitted from Table 1 as a result of their problematic nature. Initially, noun classes tend to appear in pairs, with the second member of the pair serving as the plural-inflected counterpart of the first (Miti 2006). Noun classes 11, 14, 15, 16, 17 and 18 cannot be discussed in terms of ‘number’. They are therefore also omitted from the table. Kubayi (2022), for instance, has argued that ‘noun classes 16, 17 and 18’ do not exist in Xitsonga. Torrend (1936) further notes the locative status of noun classes 16, 17 and 18 in Rhodesian Tonga in contrast to their nominal classification. Similarly, Msaka (2019:41) has observed that noun classes 15, 16, 17 and 18 are excluded from the Chichewa noun classification system on the basis that they do not correspond to nouns that make a distinction between singular and plural.

It is also clear from the table that some nouns in Xitsonga lack prefixes although it is possible to categorise them under specific noun classes. A demonstrative case is the word kondlo [mouse], which belongs to noun class 5. Other members within the same category are gama [eagle] and chela [frog]. Other examples of class 9 include homu [cow] and huku [chicken]. It may thus be tentatively argued that noun class distinctions in Xitsonga are primarily grounded in agreement morphology, rather than in prefixal form or semantic content.

The data in Table 1 further show that the agreement morphemes of the respective noun classes do not necessarily resemble the prefixes of their noun classes. The prefix for noun class 1, for instance, is mu- and yet its agreement is u. Consistent with this, noun classes 3 and 4 are prefixed by mu- and mi-, but their agreement morphemes are wu and yi, respectively. Similarly, in Swahili, the prefix for noun class 1 is m- as in mtu ‘person’, and yet its agreement morpheme is a/yu (Broohm 2017). More importantly, it is apparent that the only noun class that takes the agreement morpheme swi, which is similar to the generic agreement morpheme as identified by PanSALB (2019), is noun class 8, which is an inflection of noun class 7.

A closer examination of the data presented in Table 2 indicates that the agreement morpheme associated with noun class 8 replicates its corresponding prefix swi-. Moreover, this noun class does not appear to be semantically motivated, as it encompasses a heterogeneous range of referents, including inanimate objects, humans, animals, birds, insects and abstract entities. In fact, many scholars have pointed out for long that the membership of noun classes in general is arbitrary rather than semantically motivated (Creider & Denny 1975; Broohm 2017). This raises the question of what underlying principles govern the noun class system in Xitsonga. As stated before, there is a similarity between the prefix and the agreement morphology of noun class 8, but some nouns lack prefixes. However, noun classes project specific agreement morphemes with which they are associated. It may therefore be safe to argue that the noun class system in Xitsonga is based on agreement morphology irrespective of whether a specific nominal possesses a prefix. As noted above, the prefix for noun class 3, for example, is mu- and yet its agreement morpheme is wu. Similarly, the agreement morpheme for noun class 4 is yi, which differs from its supposed prefix mi-. Notably, in certain noun classes, such as class 8, the prefix swi- appears to serve as a consistent morphological base, replicated in its corresponding agreement morpheme swi-.

Table 1 further shows that noun class 8 also has a generic member, namely swilo ‘things’, which is an inflection of xilo ‘a thing’, which belongs to noun class 7. This suggests that, in a strict sense, members of noun class 8 – whether human or non-human – may be construed as ‘things’. Apart from its generic semantic role, like other noun classes, class 8 also performs certain syntactic functions. It is well established that in Bantu languages, for instance, head nouns can control the structure of other categories, including adjectives, pronouns, demonstratives and numerals as much as they determine the structure of their agreement morphemes (Broohm 2017:88). Consider Box 3.

The data in Box 3 demonstrate that the noun swihlangi [children], like other nouns, controls the structure of the numeral swimbirhi [two], the adjective leswintsongo [young]and the demonstrative leswi [these] through prefixation, infixation and suffixation, respectively, thus yielding them phonological forms (see Fromkin, Rodman & Hyams 2014). This pattern, however, is not exclusive to Xitsonga. In Latege, spoken in Gabon, modifiers occurring in the noun phrase agree with the head noun (Linton 2013:5). Still, in Sεlεε, a Ghana-Togo-Mountain language, determiners, numerals and interrogative pronouns agree with their controller nouns (Broohm 2017:84). Consistent with this, Jiang (2023:1) has observed that noun prefixes do not only appear in other syntactic categories, including verbs, adjectives, prepositions, numerals, determiners, quantifiers and adpositions, but they also agree with the noun in question. This pattern is likewise observed in Tshiluba, a Bantu language spoken in the Democratic Republic of Congo. Jiang (2023) writes:

What are the syntactic mechanisms through which the noun class 8 subject agreement morpheme is realised in Xitsonga?

In terms of the extended projection principle, all languages have (noun) subjects (Cook & Newson 1988; Kayne 1994), which control and determine the structure of the agreement morpheme so that the two may take the same form (Tshikalange 2000; Carnie 2007). Consider the noun class 8 subject swihlangi [children] shown in Box 4.

The example in Box 4 shows that it is the extended projection principle that ensures that the agreement morpheme swi- is controlled and determined by the noun subject swihlangi. Put differently, it is the subject that triggers the feature of agreement, giving it form and structure. This explains why agreement takes a particular form even in nouns that lack the prefix. For instance, in Kondlo ri tshama enceleni [a mouse lives in a hole], the unprefixed noun subject kondlo [mouse] triggers the correct noun class 5 agreement morpheme ri-. Speakers of the language possess unconscious knowledge of noun-subject agreement marking in a language.

Unlike English, which retains subject-verb agreement, the verb is not involved in the subject agreement system in Xitsonga. Before proceeding, it is essential to consider the concept of agreement. Forms agree based on a formal relationship between elements, whereby a form of one lexical item requires a corresponding form of another (Crystal 2008). Agreement occurs when two expressions possess the same value for the relevant grammatical feature, such as number, that is singularity or plurality of the subject as well as the feature of person (Randford 1997). The latter requires the subject to be characterised as first, second or third person. As shown in Table 1, in addition to number and person features, Bantu languages exhibit a third grammatical property, namely noun class, in terms of which a particular noun is classified. For a diagrammatical summary of the agreement of the sentence Swihlangi swi dlaya nyoka enceleni [children kill a snake in a hole] (see Figure 1).

Figure 1 clearly demonstrates that the subject swihlangi [children] functions as a third person plural noun within noun class 8, while the agreement morpheme swi- encodes grammatical information pertaining to person, number and nominal class (Radford 1997). Radford (1997) does not construe agreement as a lexical item or category in English but rather as an abstract inflectional element (e.g. -s as in ‘Mary rides a bicycle’). However, from Figure 1, it is evident that agreement, located between the subject and the verb, is a separate and independent lexical item that can stand on its own in Xitsonga. It is a rich element with a high functional load with the capacity to carry the workload of the noun subject; hence, its capacity to suppress it (Carnie 2007; Radford 1997; also Van Valin 2001; Williams 1994). In Box 5, for example, swi- has taken over the function of the noun subject.

Agreement may also be conceptualised through the distinction between surface structure (S-structure) and deep structure (D-structure). The suppression of the noun subject via agreement morphology, as illustrated in Box 5, is characteristic of surface structure (S-structure), since in deep structure (D-structure), the subject is universally present across all human languages (Cook & Newson 1988; Kayne 1994). To account for the empty subject in the S-structure as evident in Box 5, the subject position is filled by the empty category pro (see Box 6).

Over and above subject suppression, the rich agreement morphology enables noun subject inversion, resulting in inverted declarative sentences, such as in Box 7b.

While Box 7a represents a declarative construction that follows the SVO configuration, Box 7b demonstrates subject inversion, whereby the subject swihlangi [children] appears in the object position, yielding a VOS configuration. In this structure, the agreement morpheme swi- remains in the subject position, reflecting syntactic displacement. However, swihlangi remains the subject and nyoka the object of the sentence. This is facilitated by the agreement morpheme’s capacity to encode the grammatical features of the noun subject swihlangi, regardless of its syntactic position within the sentence. Unsurprisingly, the sentence is not semantically affected by the subject inversion. Both the normal declarative in Box 7a and the inverted declarative in Box 7b can be represented in the D-structure as in Figure 2 and Figure 3, respectively.

In the depiction of the D-structure in Figure 2, agreement morphology is located between the NP and the VP, which is the normal position in declarative constructions. However, in the D-structure in Figure 3, it occupies the subject position without assuming the status of the subject of the sentence. This illustrates the ability of the agreement morpheme to take on the functional roles of the noun subject. In this instance, the agreement morpheme is capacitated by the subject, which has moved to the object position to assume its functions in its absence. The fact that the subject in Box 7b now occupies the object position implies that the VP is now modified by two NPs, one serving as the object and the other as the subject, leading to dual nominals in the object position. The vertical broken lines on either side of the diagram in Figure 3 indicate that the empty NP pro and the NP swihlangi form a unified whole. This is also illustrated by the broken curve underneath the diagram.

When the agreement morpheme swi- is inflected into swa- or swo-, the sentence gets semantically slightly altered.

The expression in Box 8a suggests that the form of the agreement morpheme swi- represents an SVO declarative construction stating that ‘the children kill a snake in a hole’. This can be regarded as the basic form of the agreement morpheme. The sentence in Box 8b indicates that the form swa- expresses ability, with the meaning ‘the children can kill a snake’. The morpheme swo-, as illustrated in Box 8c, conveys a sense of bravery attributed to the referent noun subject. Implicit in this construction is the notion that only a very limited number of children – those capable or courageous enough to kill a snake – exist, figuratively suggesting they could be counted on one hand. Apart from being the agreement morpheme for noun class 8, swi- can also serve as a generic morpheme.

Generic reference occurs where a lexical item refers to an entire or whole class of entities rather than to specific or individual members (Crystal 2008; Richards & Schmidt 2010; Brown & Miller 2013). In this regard, swi- is not only a noun class 8 agreement morpheme. One of its extended functions is to mark compound and complex noun subjects whose nominals belong to different noun classes. Consider the following:

In Box 9, the compound subject consists of two nouns, vavanuna [men] and vavasati [women]. They both belong to noun class 2 and have the agreement morpheme va-. However, the use of the agreement morpheme va- in Box 10 produces a syntactically deviant sentence. It demonstrates that va- can only be used to denote noun class 2, which is the human class and not in a generic context. Ironically, this is despite the fact that both the two members of the compound noun subject (swihlangi [children] and swisiwana [poor people]) are humans. In Box 11, the compound subject mirhi na maribye ‘trees and stones’ comprises members belonging to noun classes 4 and 5, respectively, and yet, the sentence has not lost its grammaticality. This signifies that the agreement morpheme swi-, which belongs to noun class 8 as noted before, has a generic function. By the same measure, the agreement morpheme swi- in Box 12 is employed because the complex subject comprises nouns from multiple noun classes, specifically classes 2a, 8 and 10. Notably, all constituents of this complex noun phrase refer to human entities.

The function of swi- as a marker of generic agreement morphology in Xitsonga is further evidenced by its use in everyday conversational contexts. Consider the expressions in Boxes 13a and 13b and the question in Box 13c.

The expression a swi ve tano [let it be so] is commonly employed in everyday discourse to signal an expectation that one of the interlocutors will honour a commitment or fulfil a promise. The expression a swi na mhaka [it does not matter] functions to convey that a particular state of affairs is not cause for concern, as a positive outcome is anticipated. As shown in Box 13c, the morpheme swi is used in the everyday interrogative expression swi lo yini? [what is wrong or what is the matter?], which serves to express concern about the current state of affairs or to elicit information regarding a given situation. Likewise, idiomatic expressions can also fall within the scope of generic morphological marking as evident in Box 14a-c.

The idiom in Box 14a is typically used to express appreciation for newly acquired knowledge, often gained through travel, exposure to unfamiliar environments or interactions with new people. In contrast, Box 14b conveys wonderment at life’s marvels or a response to cultural surprise, while Box 14c is employed upon hearing a startling statement or witnessing an unexpected event. Now consider the sports commentaries in Box 15a-c.

The two utterances in Box 15a-b are commonly used within South African soccer media discourse – the first predominantly on television and the second on radio by prominent commentators – to express admiration for well-executed goals. Notably, Box 15b is also frequently employed to celebrate the victorious team.

Generic morphology also features prominently in academic writing, particularly at the beginning of sentences, where it serves to introduce assertions or advance lines of argumentation, as illustrated in Box 16a-b.

In summary, the agreement morpheme swi not only aligns with noun class 8, which encompasses both human and non-human referents, but also extends to other plural noun subjects across various noun classes.

The aim of this study was twofold: (1) to determine the noun class associated with the generic agreement morpheme swi-; and (2) to find out whether both the particular noun class and its agreement morpheme can be described as basic, implying that they are unmarked. It has come to the fore that the agreement morpheme finds expression in noun class 8, whose membership consists of both human and non-human subjects. It has further emerged that the agreement morpheme is broadly employed to encode plural noun subjects across diverse noun classes and to structure a range of communicative functions, including everyday expressions, soccer commentary, idiomatic formulations reflecting general states of affairs, assertive statements in academic discourse and general inquiries concerning well-being.

It is unsurprising that noun class 8 and its corresponding agreement morpheme in Xitsonga may be interpreted as the default or unmarked forms, given their relatively broad semantic scope and reduced specificity compared to other noun classes and agreement markers. Moreover, both noun class 8 and its agreement morpheme appear to encompass other noun classes and agreement markers in distinct ways, without suggesting any oppositional relationship between noun class 8 and the rest of the noun class system. Similarly, swi should not be construed as standing in contrast to agreement morphemes associated with other noun classes. Accordingly, and in line with Andersen (1989), noun class 8 and its associated generic agreement morpheme are better understood not as contrastive elements but as superordinate categories relative to their subordinate counterparts.

This study is expected to offer a substantive contribution to the broader morphosyntactic analysis of linguistic phenomena in Bantu languages while also foregrounding the theoretical insights afforded by markedness theory. It is therefore reasonable to conclude that this study holds theoretical significance for Bantu languages more broadly, particularly with regard to linguistic phenomena that may be characterised in terms of contrast or hierarchical relations such as hyponymy. This study carries specific implications for subfields such as lexicography, particularly regarding the treatment of agreement morphology within the central entries of dictionaries. Additional areas of impact include grammatical analysis and language pedagogy. In the context of the latter, instructional focus has traditionally centred on nominal class agreement, with comparatively limited attention to generic morphology. From this perspective, a more comprehensive understanding of syntactic agreement systems calls for further in-depth research into both nominal and generic morphology and related phenomena across Bantu languages.