2. The empirical picture

As mentioned in the introduction, the IPP-effect in Afrikaans is often taken to be optional. However, there is a clear difference in the frequency of the IPP- and the past-participle form in IPP-contexts if one considers the various V₂ subclasses. This is shown in a recent study by Dirix et al. (Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020), who conducted a corpus study using the Taalkommissie (‘Language Commission’) corpus and the Wikipedia corpus (https://viva-afrikaans.org/). Their study shows that aspectual verbs (begin ‘begin’, gaan ‘go’, kom ‘come’, bly ‘stay’, aanhou ‘continue’, and ophou ‘stop’), subject control verbs (probeer ‘try’, durf ‘dare’, and leer ‘learn’), causative laat ‘let’, perception verbs (sien ‘see’ and hoor ‘hear’), and benefactive (help ‘help’ and leer ‘teach’) show very high frequencies of the IPP-form, ranging from 81.25 to 100 percent. These are also all subclasses of verbs which exhibit the IPP-effect in Dutch (Schmid Reference Schmid2005).Footnote ⁵

Two subclasses of Afrikaans verbs which show different IPP-behavior compared to Dutch, however, are (i) motion verb loop ‘walk’ and the three cardinal posture verbs, sit ‘sit’, staan ‘stand’, and lê ‘lie’, and (ii) the root modal verbs moet ‘must’, kan ‘can’, wil ‘want’, and mag ‘may’. In the following subsections, we discuss these subclasses in detail.

2.1 Motion and posture verbs

In Afrikaans, the motion verb loop ‘walk’ and the posture verbs sit ‘sit’, staan ‘stand’, and lê ‘lie’ occur in pseudocoordination constructions (see footnote 2 and De Vos Reference De Vos2005 for detailed discussion). Consider again (2) above. In Robbers (Reference Robbers1997), De Vos (Reference De Vos2001, Reference De Vos2005), and Dirix et al. (Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020) it has been noted that this set of verbs in particular exhibits optional IPP. Cavirani-Pots (Reference Cavirani-Pots2020) additionally shows that this optionality is reflected both in corpus data and in large-scale native-speaker judgments. The latter is an important addition to the discussion surrounding the optionality of the IPP-effect in Afrikaans because corpus results typically cannot tell us anything about speaker-internal optionality. Cavirani-Pots’ data are based on the collected judgments of 201 Afrikaans native-speakers who assessed three-verb clusters featuring the above four verbs in the correct IPP-context.Footnote ⁶ They reveal a high degree of intraspeaker optionality regarding the IPP- and non-IPP-forms in IPP-contexts. An adapted version of the relevant data table is given here in table 1.Footnote ⁷

Table 1.

Optionality of ge- per motion or posture verb (%) (Cavirani-Pots Reference Cavirani-Pots2020:192)

As the table shows, for the majority of speakers surveyed, ge- is truly optional in IPP-contexts featuring motion and posture verbs. Interestingly, the extent to which speakers permit both IPP- and non-IPP-forms mirrors the extent to which the verb in question has been grammaticalized (De Vos Reference De Vos2005, Breed Reference Breed2017, Cavirani-Pots Reference Cavirani-Pots2020):Footnote ⁸ Most speakers permit both IPP- and non-IPP-forms with lê, the least grammaticalized posture verb, with sit and staan less generally permitting both options, and strongly grammaticalized loop doing so least of all. Similarly, the number of speakers requiring an IPP-form (no ge-) is highest for loop and lowest for lê with sit and staan behaving more like this least grammaticalized verb.

For the four verbs under consideration here, then, IPP evidently can be truly optional for many modern-day speakers. However, as discussed in Cavirani-Pots (Reference Cavirani-Pots2020:276), these data do not represent all varieties of Afrikaans, and exclude especially those regions that are hard to reach via online questionnaires (e.g. the Northern Cape province). We return to consider IPP in colloquial varieties of Afrikaans, including, notably, some that were not covered in Cavirani-Pots’ survey, in section 2.4.

2.2. Modal verbs

The class of modal verbs in Afrikaans behaves significantly differently to the other subclasses of IPP-verbs in the language. Modal verbs are morphologically special as they have a dedicated past-tense form, which the other IPP-verbs, thanks to Präteritumschwund (‘preterite loss’; see Conradie Reference Conradie, Salmons and Dubenion-Smith2007), do not: compare moet–moes ‘must’, kan–kon ‘can’, wil–wou ‘want’, sal–sou ‘will’, and, marginally, mag–mog ‘may’ (mog is an archaic verb, absent from the active lexicon of most present-day speakers of Afrikaans). Furthermore, they lack a perfect-participle form (Donaldson Reference Donaldson1993:242). Dirix et al. (Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020) show, based on a corpus study, that the classic IPP perfect-tense construction (MOD₂-V₃-het₁) is virtually non-existent in the corpus. Specifically, they tested Robbers’ (Reference Robbers1997:56–57) claim that a sentence like (3) can have five different past-tense forms associated with different degrees of acceptability:

Table 2 presents the associated corpus results from Dirix et al. (Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020:130–131)Footnote ⁹ showing the frequencies of the construction type in (4a) – pret₁-inf₂ – versus all constructions with het ‘have’ – i.e. (4b–e) taken together – per modal verb. As can be seen from this table, the construction type in (4a) occurs in the vast majority of the relevant data points (95.2%). This means that modal verbs generally occur only very infrequently in IPP-contexts.

Table 2.

Frequencies of PRET₁-INF₂ construction versus those with het (Dirix et al. Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020:130–131)

Table 3 presents the corpus results from the same study for the frequencies of the constructions in (4b–e) and shows that within the constructions with het, the construction type in (4d) is by far the most frequent.

Table 3.

Frequencies of different het-constructions (Dirix et al. Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020:132)

Insofar as we can label one of the options given in (4) as being ‘IPP-like’ (compare the Dutch IPP-form: Jan heeft ₁ hard kunnen ₂ werken ₃), it would at first sight be (4c): Here we have a modal that is selected by perfect auxiliary het ‘have’, with the modal itself selecting a third verb, and not surfacing as a participle (*gekan), but in ge-less form. The fact that the modal does not appear as a participle is unsurprising as most varieties of Afrikaans lack modal participles.Footnote ¹⁰ It, however, turns out that exactly this option – which speakers judge as being available in the context of grammaticality judgmentsFootnote ¹¹ – is virtually absent in the corpus search executed by Dirix et al. (Reference Dirix, Augustinus, Van Eynde, De Vogelaer, Koster and Leuschner2020). This can be seen from table 3: 0.2 percent of the datapoints relate to this option. This replicates the findings of De Schutter’s (Reference De Schutter2001) smaller, fiction-based corpus study. Like De Schutter, Dirix et al. also found that (4a)-type structures are very common in contexts where an IPP-structure could have surfaced in Dutch; that is, Afrikaans speakers favor the use of a two-verb past-marked modal (MOD₁-V₂) pattern where Dutch speakers harness the classic IPP perfect-tense MOD₂-V₃-AUX₁ pattern.

The other commonly occurring pattern is (4d) (93.9% of the datapoints), a construction in which an apparently past-marked modal selects the perfect auxiliary, which in turn selects V₃.Footnote ¹² These so-called preteritive assimilations (Ponelis Reference Ponelis1979) are very common whereas (4e)-type structures, where the modal is present/unmarked, exhibit a much more restricted distribution (4.7% of the datapoints). This is true in more standardly oriented varieties; it is worth noting that (4e)-type structures are not uncommon in Kaaps (see again Hendricks Reference Hendricks, Hendricks and Dyers2016, Reference Hendricks, Carstens and Bosman2024).

We also see past-doubling in (4b), where both auxiliary het and kon superficially express the past tense. This structure is interesting as kon could in fact be an infinitive form: as part of the reanalysis of perfect-auxiliary het (see section 3 below), Afrikaans has innovated a past-tense modal infinitive structure which permits speakers to produce structures like (5) (see Conradie Reference Conradie, Salmons and Dubenion-Smith2007, Reference Conradie, Carstens and Bosman2024 for extensive discussion):Footnote ¹³

(4b) therefore constitutes another IPP-innovation. Like (4c), though, this structure barely surfaces in Dirix et al.’s study (0.2% of the datapoints). What speakers prefer instead is (4d), … om daar te kon gewerk het, the ge-containing variant of (4c) (93.9% of the data points for (4d) versus 0.2% for (4c)).

In sum, given that Afrikaans modals essentially (see note 10) have no perfect-participle form, morphological restrictions categorically rule out optional IPP for this class of verbs from the outset: Only the IPP-form is expected to be possible. In practice, modals do not seem to occur in IPP-contexts, however, as speakers instead prefer either a past-tense form with just one selected main verb (4a) or so-called preteritive assimilation constructions, in which the modal is no longer the cluster’s V₂, with the result that it cannot ‘show IPP’ (4d). This class thus behaves very differently from its cognates in Dutch. This is a significant consideration, given that modals constitute the core class of IPP-verbs in that language, and in West Germanic more generally (Schmid Reference Schmid2005).

2.3. IPP verbs and quirky V2

So far, we have seen that there are six subclasses of Afrikaans verbs that show the IPP-effect either (almost) obligatorily or optionally:

Interestingly, the majority of these verbs have another type of morphosyntactic behavior in common, namely that they are able to occur in the so-called quirky Verb Second (henceforth quirky V2) constructions (De Vos Reference De Vos2005; see also Ponelis Reference Ponelis1993:325–330 on so-called complex initials and Roberge (Reference Roberge1994) for diachronically oriented discussion of so-called verbal hendiadys). An illustration of the construction is given in (7a); compare this with the standard V2 configuration in (7b).

In (7a), the entire pseudocoordination(-like) complex (loop (en) vertel) appears in V2-position; this is quirky V2. In (7b), only the finite verb, loop ‘walk’, appears there. In Dutch, quirky V2 is completely ungrammatical. Interpretively, the Afrikaans quirky V2 structures in (7) overlap, i.e. the alternation between them can be semantically vacuous, but this may be a restricted alternation; further research is required on this matter, so we leave it aside here.

The only verbs listed above that cannot occur in quirky V2-structures are the perception verbs, and aanhou ‘continue’, ophou ‘stop’, and durf ‘dare’. The nature of the excluded verbs suggests that the possibility of participating in quirky V2 depends on the size of the verbal complement a given verb takes. That is, verbs that do not require their own subject and that select a structurally very restricted complement (potentially, in the sense of Wurmbrand Reference Wurmbrand2010, Reference Wurmbrand2024) can combine with a lexical verb in order to create a quirky V2 verbal complex (effectively a verbal compound; see section 3 below for brief further discussion); verbs that independently select for an external argument and that require a larger complement cannot.

For perception verbs, then, we assume quirky V2 to be ruled out because these verbs select their own subjects, which requires the projection of independent vPs (i.e. thematic/argument-structure domains). This seems justified, given that the relevant perception verbs, hoor (‘hear’) and sien (‘see’) retain their core lexical semantics – one cannot employ the relevant verbs in IPP-structures without the subject hearing and seeing what is encoded by V₃ and its associated arguments and modifiers; hoor and sien are therefore not (partly) grammaticalized evidentials of the kind found in many languages (Aikenvald Reference Aikenvald, Aikhenvald and Dixon2014). Aanhou ‘continue’ and ophou ‘stop’, in turn, are particle verbs, which obligatorily strand their particles under V2. Consider (8):Footnote ¹⁴

As their component parts are necessarily separated under V2, these verbs are independently incapable of forming a unit with the lexical verb that undergoes movement to the V2-position. As for durf ‘dare’, the semantic connection between this verb and the class of root modals, and the fact that Dutch durven groups with the root modals in respect of its fixed plaatscategorie (‘positional category’) in verb clusters (Coussé & Bouma Reference Coussé and Bouma2022:126) lead us to expect that this verb will pattern with the modals. And this is correct: the modals can never occur in quirky V2 structures:Footnote ¹⁵

It is of theoretical importance that IPP-verbs that lack an independent subject requirement and take small verbal complements feature in quirky V2 structures while those, like the individual verbs just discussed, that require an independent subject, take larger complements and/or consist of separable parts do not: This shows us that Afrikaans IPP-verbs are verbs that combine very closely with the lexical verbs they select. That is, in order to be able to co-occur in V2 position, the two verbs combined must appear to the syntax as one complex verb,Footnote ¹⁶ and, moreover, as one in which the component parts are more tightly bound than those of particle verbs (which, as shown in (8) above, are separated under V2). This perspective picks up on earlier discussion of IPP-phenomena which views it as the reflex of the co-occurrence of two or more verbs which have to share a domain which is in fact too small for them (see Kjeldahl Reference Kjeldahl2010 for discussion and references). A full theoretical analysis of the Afrikaans IPP-effect and how the relevant subset of IPP-verbs facilitate quirky V2 is beyond the scope of this article. In section 3 below, we will, however, offer some initial thoughts, focusing specifically on the interaction between pre-existing Dutch-derived properties and Afrikaans’ development in a contact environment.

2.4 IPP in colloquial Afrikaans

The last component of our empirical sketch concerns a discrepancy between more and less standard-oriented colloquial varieties of Afrikaans. Donaldson (Reference Donaldson1993:225-226) notes for colloquial varieties in general that the presence of ge- on V₂ in IPP-contexts is strongly preferred. Other scholars also mention specific varieties that seem to prefer ge- on V₂ in IPP-contexts, e.g. Griqua/Griekwa Afrikaans, Knysna Boswerker (‘Knysna Forest worker’) Afrikaans, and Kaaps (De Vos Reference De Vos, Garding and Tsujimura2003, Conradie Reference Conradie, van Kemenade and de Haas2012, Hendricks Reference Hendricks, Hendricks and Dyers2016, Reference Hendricks, Carstens and Bosman2024).

In the absence of detailed empirical studies of any of these varieties, the true extent of the preference for ge- must be viewed as an open question. What can already be established at this stage, however, is that the strong preference for ge- on V₂ seems to correlate with other non-standard behaviors of the prefix, and also with some further relevant morphosyntactic properties. Three of the properties discussed in De Vos (Reference De Vos, Garding and Tsujimura2003) and Conradie (Reference Conradie, van Kemenade and de Haas2012) are briefly presented here, namely (i) the combination of ge- with other verbal prefixes and particles, (ii) ge- occurring on V₃ rather than on V₂ in IPP-contexts, and (iii) auxiliary het (V₁) being dropped in the presence of ge- on V₂ in IPP-contexts.

In Griqua Afrikaans, and also in less standard-oriented varieties more generally, ge- frequently occurs on perfect-participle forms which already contain a verbal prefix like be-, er-, her-, ont-, or ver-. Rademeyer (Reference Rademeyer1938:62-63) gives gebegene ‘begun’, geërken ‘acknowledged’, geherken ‘recognized’, geonthou ‘remembered’, and geverneem ‘enquired’ as examples. These forms indicate that ge- has greater freedom in the relevant varieties than in standard Afrikaans, where ge- is obligatorily absent in the presence of these prefixes (Donaldson Reference Donaldson1993:225-226, Conradie Reference Conradie, van Kemenade and de Haas2012). In Griqua Afrikaans, ge- exhibits even greater positional freedom: where ge- consistently appears between the particle and the verbal stem in Dutch and also in most varieties of Afrikaans (e.g. opgebel – up.ptcp.call –‘called’), it can attach to the outside of the entire particle-verb complex in Griqua Afrikaans, as in geopbel ‘called’. Examples from Rademeyer (Reference Rademeyer1938) are geneersit ‘put down’ and geaanteel ‘reproduce’. Further, according to Rademeyer, ge- can even occur in both positions simultaneously. Consider (10) in this connection:

The second deviation from standard varieties with respect to ge-placement that Griqua Afrikaans shows – and this it shares with other less standard-oriented varieties such as Knysna Boswerker Afrikaans, Baster Afrikaans, Velddrifse Vissertaal (‘Velddrif Fishermen’s Language’; De Vos Reference De Vos, Garding and Tsujimura2003) – is that ge- can also occur on the lexical verb (V₃) in IPP-contexts rather than on V2. Consider (11):

This option may be more widespread than just the varieties mentioned by De Vos (Reference De Vos, Garding and Tsujimura2003). Cavirani-Pots’ (Reference Cavirani-Pots2020) questionnaire study did not target dialectal varieties per se but found that ge- on V₃ was accepted by 14 speakers in pseudocoordination constructions in IPP-contexts with loop ‘walk’ as V₂, and by 35 speakers in similar constructions with sit ‘sit’ as V₂. Future work should probe this in more detail.

The final morphosyntactic property that Griqua Afrikaans specifically has that combines with its preference for ge- in IPP-contexts is that it often drops auxiliary het ‘have’. Consider (12) ([het] marks the position of the omitted V₁ auxiliary):

Taking the data in this subsection together, we see that there appear to be colloquial varieties of Afrikaans, including Griqua Afrikaans, in which the IPP-effect is much less strongly attested than in standard and standard-oriented varieties of Afrikaans. Significantly, for at least some of these varieties, this fact correlates with further distinctive morphosyntactic behavior, notably of ge- and of the perfect auxiliary het. In the following section, we consider how the “un-Dutch” behavior of these and other elements discussed above may shed light on the nature of IPP in contemporary Afrikaans.

3. Inheritance and contact in the making of Afrikaans IPP

In the preceding discussion, we have seen that the IPP-effect is still a feature of modern Afrikaans, albeit to varying extents in different varieties. The internal make-up of the most commonly attested IPP-structures is quite different from that found in Dutch and West Germanic more generally, however. Where modal-centered IPP is common to all Continental IPP-systems, constituting an IPP-“baseline” (Schmid Reference Schmid2005), this option seems only to be rarely attested in Afrikaans production (section 2.2). On the other hand, Afrikaans does feature a comparatively speaking wide range of IPP-verbs, including several innovated ones (e.g. aanhou ‘continue’ and motion loop ‘walk’). Furthermore, these IPP-structures can alternate with ge-marked structures in a way that is not possible in Dutch varieties (sections 2.1, 2.3, and 2.4). In this section, we will offer some initial consideration of factors – both “internal” and “external” – that may have played a role in creating the IPP-picture that has emerged from our empirical investigation.

Firstly, it is important to note that Afrikaans IPP-clusters consistently require 231-clustering.Footnote ¹⁷ This contrasts with the usual Dutch pattern, which is 123, with 231 being an additional option in some dialects (Wurmbrand Reference Wurmbrand, Everaert and van Riemsdijk2017). The fact that Afrikaans permits only the 231-order plausibly follows from two factors:

The loss of optionality in two-verb clusters may be a contact effect. In the sense that free-variation-type optionality, of the kind available in Dutch verb clusters, entails the existence of more than one linguistic form per meaning, it instantiates complexity (Forker Reference Forker2021; see also Kroch Reference Kroch, Beals, Denton, Knippen, Melnar, Suzuki and Zeinfeld1994 on the instability that frequently arises in contexts featuring doublets where only one is a productive part of the system). This type of complexity is known to be vulnerable in situations involving adult L2-learners (Smith & Wonnacott Reference Smith and Wonnacott2010, Perfors Reference Perfors2016).Footnote ¹⁹ More generally, simplification has been suggested to be the outcome in contact situations involving significant numbers of adult L2-speakers (Trudgill Reference Trudgill2011), a scenario which certainly held at the Cape during the period of Dutch rule (1652–1806; see Ponelis Reference Ponelis1993:chapter 1, and Groenewald Reference Groenewald, Carstens and Bosman2024).

While contact may therefore have played into the loss of ordering optionality in Afrikaans, the way in which the innovated aspectual and pseudocoordination verbs were integrated into the existing 231 IPP-pattern, however, reflects the internal organization of the grammar. More specifically, the innovated IPP-verbs exhibit the 1-2 ordering associated with modal verbs (see (14a) above); for verb-clustering purposes, they therefore became part of the existing grammatical class to which modal verbs belong. More precisely, the innovative IPP-verbs – which are all aspectually and perspectivally oriented in semantic terms (see below) – joined the class of modal-patterners, which require their selected verbal complement to follow them, i.e. mod₁-inf₂ as in (14a).

This can be interpreted as a consequence of the acquisition biases that drive L1-acquirers to make maximal use of minimal means (MMM; Biberauer Reference Biberauer2019b).Footnote ²⁰ That is, where possible, newly grammaticalized verbs will be assigned to already-existing grammatical categories, adopting their characteristic properties. This is in preference to the newly grammaticalized verbs exhibiting distinctive behavior and initiating a novel category of their own. In our case, all the innovative verbs therefore adopted the 1-2 ordering characteristic of Afrikaans modals, as in (14a), with the light verb (V₂ in the cluster) preceding the verb it selects.

Further, the fact that acquisition of three-verb clusters “piggy-backs” on the already-acquired ordering of two-verb clusters (see again Van Kampen Reference Kampen2017) reflects MMM in two additional ways. Firstly, it reflects the general computational fact (see footnote 20) that syntactic structure is always built up via successive binary Merge operations; superficially three-member structures will therefore always consist of nested binary structures, as illustrated in (15) for a familiar three-membered derivational form and in (16) for the 231-structures that are our focus here:

This structural fact then plays into the second MMM-reflex, namely the already-mentioned acquisitional fact (see again footnote 20) that three-part structures are acquired by combining two pieces of earlier-acquired knowledge (see (13b) above): That the modal is initial in relation to its selected verb (1-2 ordering) whereas the auxiliary is final in relation to its selected verbal complement (2-1 ordering). Binary Merge produces two nested substructures (e.g. gaan ₂ koop ₃ het ₁ as in (16) is [[gaan ₂ + koop ₃] het ₁]), which acquirers learn independently and then combine to give the IPP-ordering.

Having considered the internal and external factors that appear to have shaped the 231-ordering that characterizes Afrikaans IPP-clusters, let us consider why this ordering is significant in understanding their distinctive properties. That IPP-structures are 231 is significant as this order ensures (i) the cluster-initiality of the IPP-verb (V₂), (ii) the adjacency of the IPP-verb (V₂) and the lexical verb (V₃), and (iii) the cluster-finality of het (V₁). (ii) matters as it creates the structural conditions required for the “derivational univerbation” (see below) observed in many Afrikaans IPP-clusters, while (i) ensures that a perspectival element surfaces cluster-initially, and (iii) facilitates key tense-related developments that have not occurred elsewhere in Germanic. Let us consider these points in a little more detail.

All the IPP-verbs (V₂s) are in some sense point-of-view/speaker-oriented (i.e. perspective- or stance-marking elements), and there has been significant innovation in the Afrikaans point-of-view-centered aspectual domain. More specifically, comparison with Dutch shows that “light” verbs familiar from Dutch (e.g. gaan ‘go’, kom ‘come’, loop ‘walk’, and staan ‘stand’) have become more grammaticalized in Afrikaans than is the case for their Dutch counterparts (see, among others, Breed Reference Breed2017, Biberauer Reference Biberauer2019a, Cavirani-Pots Reference Cavirani-Pots2020). At the same time, aspectual pseudocoordinations have also been established (De Vos Reference De Vos2005, Biberauer & Vikner Reference Biberauer, Vikner, LaCara, Moulton and Tessier2017, Biberauer Reference Biberauer2017a, 2019a, Cavirani-Pots Reference Cavirani-Pots2020), and additional lexical items (e.g. aanhou ‘continue’ and ophou ‘stop’) have joined the class of IPP-verbs, potentially as an early step in an incipient grammaticalization process. These developments can be thought of as ‘internal’ to the extent that grammaticalization is an internal process.

Here it is worth noting, however, that the aspectual developments in Afrikaans are strongly tied to intersubjective meanings, that is to meanings that are particularly strongly associated with interactional contexts, which, in the case of Afrikaans, would also have been sociolinguistically complex throughout its history. In such contexts, it is useful for speakers to have a robust inventory of interactionally oriented lexical and grammatical structures facilitating effective communication.Footnote ²¹ Plausibly, then, the impetus for the grammaticalization – or, registering Traugott’s (Reference Traugott and Hickey2009) subjectification > intersubjectification pathway and important work by, among others, Diewald (Reference Diewald2011) and Müller & Axel-Tober (Reference Müller and Axel-Tober2025), plausibly, the pragmaticalization – observed in Afrikaans was both internal and external.

Importantly, the quirky V2 phenomena presented in section 2.3 demonstrate the extent to which some of the aspectual forms – crucially, those internally/grammatically licensed to do so (see again section 2.3 on the contrast between aanhou/ophou and the other forms, and the discussion to follow) – have grammaticalized: they effectively serve as adverbal modifiers adjoined directly to the finite verb and creating a syntactically indivisible unit that is effectively a compound verb (see Biberauer & Vikner Reference Biberauer, Vikner, LaCara, Moulton and Tessier2017 for some initial discussion of what this might entail in formal terms). Had (ii) not held, the formal conditions for the rise of quirky V2 would have been compromised, with knock-on effects for the rise of the IPP-patterns seen in Afrikaans.

More specifically, quirky V2 arises where verbs that are combined via Merge – that is, verbs that are structurally adjacent (Merge sisters) – come to be treated as a single syntactic unit (via a derivational univerbation process, the formal details of which we leave aside here). This applies to the lower verbs in IPP-structures (V₂ and V₃) wherever V₂ directly selects V₃. The “size” considerations discussed in section 2.3 therefore become relevant: only V₂s selecting very small complements (specifically, subjectless vPs) will produce V₂-V₃ Merge sisters, and only these verbs are therefore predicted to be compatible with quirky V2 (where they, of course, are part of a two- rather than three-verb cluster).Footnote ²²

Hierarchy aside, the linear aspect of the structural adjacency between V₂ and V₃, highlighted in (ii) above, is also important. The initial placement of V₂ (i.e. the initial verb in a Quirky V2 verbal complex) arguably enables these verbs to serve as modifiers in a complex verb, parallelling what we see in other attributive compounds in Afrikaans more generally (Van Huyssteen Reference Huyssteen2020, Reference Huyssteen and Breed2023). In practice, then, many Afrikaans 231-clusters are therefore actually two-verb clusters, with V₂-V₃ constituting an initial verbal compound and V₁ the second cluster-member, that is, [[gaan ₂ + koop ₃] het ₁] as in (13) is effectively [[gaan-koop]₂ het ₁]. This pattern matches the invariant 2-1 ordering found with auxiliary het in verb clusters (see again (14b)). Compressed verbal structures of this kind are systematically unavailable in Dutch, where the presence of verbal inflection (contrasting finite and non-finite forms, which require additional functional structure) presumably precludes the derivational univerbation observed in Afrikaans. Internal grammatical factors have therefore served to integrate the 2-3 component of Afrikaans IPP-clusters in a distinctive way.

Turning to the placement of cluster-final het, (iii) above was, in turn, crucial to the developments that have affected clause-final het more generally (see Conradie Reference Conradie, Salmons and Dubenion-Smith2007, Zwart Reference Zwart2017). The precise formal status of final het need not concern us here;Footnote ²³ what matters is that the absolute requirement that non-V2 het be verb-cluster-final and necessarily adjacent to its selected participle/IPP-infinitiveFootnote ²⁴ (see (14b)) has clearly fed into its formal reduction: as is well known, rigid placement of this kind feeds morphologization (Hopper & Traugott Reference Hopper and Traugott1993:132, Conradie Reference Conradie, Salmons and Dubenion-Smith2007). And this reduction has, in turn, led to a significant restructuring of the Afrikaans tense system, which we will now briefly consider.

In Dutch, ge- is often analyzed as a completive-aspect marker (Zwart Reference Zwart2007). In Afrikaans, it appears to have undergone further grammaticalization (Asp > T, effectively), becoming a tense-related marker in modern-day Afrikaans (De Vos Reference De Vos, Garding and Tsujimura2003).Footnote ²⁵ The proposal that ge- may in fact be a tense-marker of some kind would, for example, account for dialectal data like that from Griqua Afrikaans, which permits het-dropping (see again the data in (12) above). The ge-to-tense grammaticalization process may have its origins in the Cape Dutch pidgin that fed into the structure of Afrikaans: Roberge (Reference Roberge and Mesthrie2002:93) notes that ge-/ga- in the Cape Dutch pidgin thought to have been spoken in the Cape from the early eighteenth century “marked events situated in the past,” and he mentions the possibility that this usage was reinforced by Khoekhoe preterital particles of similar form. In most modern Afrikaans varieties, however, ge- still systematically combines with het, which has also undergone further grammaticalization in all varieties of Afrikaans.

Focusing first on ge-: its distribution in colloquial varieties (other than Griqua Afrikaans) suggests that it has become part of a superficially circumfixal tense-marking structure (ge-V-het).Footnote ²⁶ As such, the tendency for speakers to include it in (some of the) IPP-structures can be understood as a tendency to regularize the expression of past tense in Afrikaans such that it includes both the het- and the ge-components. This impetus to regularize may well be reinforced by the existence of superficial 231-structures that are in fact two-verb 2-1 clusters, namely, those containing V₂s that participate in quirky V2. Example (17) illustrates the relevant structural configuration:

As 2-1 structures are participial structures in Afrikaans (see (14b)), the occurrence of ge-marking in IPP-structures featuring quirky V2 V₂-verbs is not unexpected: This pattern simply reflects the verbal complex being treated in the same way as the majority of simplex verbs in Afrikaans, which take ge- in participial contexts.

Given the fact that ge- is, however, not consistently realized in non-IPP contexts either – standard-oriented varieties in particular retain (versions of) the Dutch-derived ge-prefixing constraint – optionality is to be expected. According to the ge-prefixing constraint that is active in standard Afrikaans (as it is in standard Dutch), verbs with unstressed initial syllables do not take ge-: thus het (*ge)probeer/(*ge)verstaan/(*ge)begin (‘have tried/understood/started’). Interestingly, Conradie (Reference Conradie, van Kemenade and de Haas2012) points to phonological considerations that appear to condition the realization of ge-marking in standard-oriented varieties: Participles are characterized by a rising stress-pattern, in the absence of which ge- is obligatory; thus gegáán ‘went’ versus (*ge)probéér ‘tried’. In less standard-oriented varieties, this phonological consideration is loosened under the influence of what can be viewed as “competing” morphosyntactically based ge-generalization/regularization pressure. This then gives rise to forms like those introduced in section 2.4.

Since IPP-structures effectively feature a two-part participle in Afrikaans (see again (17)), with V₃ serving as a stressed component (e.g. gaan éét ‘go eat’), we expect the ge-less structure to be preserved in more standard-oriented varieties where ge-marking is prosodically conditioned. This, then, supports the retention of IPP-structures. Where morphosyntactically driven ge-generalization is in play, however, optionality emerges. Against this backdrop, the fact that Afrikaans IPP-structures may be ge-marked therefore follows from internal factors (the considerations determining the realization of ge-) and the differing formal generalizations that speakers of different varieties postulate – possibly to varying extents in different registers – regarding those internal factors.

Returning to het: Its formal reduction has not just affected the realization of participial structures; it also seems to be an important consideration in the reorganization of the modal system. Specifically, if het has become a tense suffix, as tentatively proposed by Zwart (Reference Zwart2017), it becomes possible to analyze past-marked modals as suppletive forms which will not therefore co-occur with het: They are already tense-marked. To the extent that speakers’ grammars contain inflectional het, both (4b) and (4c) are therefore expected to be absent from production, as observed. Similarly, (4a) is expected to be common, again as observed. Space considerations preclude full engagement with the patterns discussed in section 2.2, but what again seems clear is that the differences between Afrikaans and Dutch in the modal domain are at least partly the result of contact (the factors affecting the reanalysis of het and ge-) and partly of the kind of internal reorganization that is familiar from systems in which key elements undergo reanalysis.