Future Time Reference Alternation in Afrikaans as a West-Germanic Language

Abstract

There are two future time reference auxiliaries in Afrikaans, sal ‘will’ and gaan ‘go’. These auxiliaries are interchangeable in many contexts. In light of the ongoing grammaticalization of gaan, it is pertinent to describe the alternation between sal and gaan in different Afrikaans registers, and contextualize it in the West-Germanic language family where English and Dutch have similar alternating constructions. This is accomplished by analyzing Afrikaans corpus data from the 1970s and the 2000s, both spoken and written. Normalized frequencies and relative frequencies for the use of sal and gaan are reported according to a number of variables, including time, register, lexical verb, syntactic subject, clause type, sentence type, and future proximity. The effect of sentence type and future proximity is consistently present in all the datasets, and a possible change is detected in the effect of subject and clause type. Compared with English and Dutch, Afrikaans future alternation patterns more like that of English, even though it is more closely related to Dutch.

1. Introduction

Afrikaans is a non-radical ‘fort’ creole1 that developed from seventeenth century Dutch through language contact in the Cape Colony at the southern point of Africa (Den Besten 2012). While many Dutch features have been retained, a large extent of deflection and regularization occurred during the formation of Afrikaans (Den Besten 2012, p. 258 and further; see also Ponelis 1993 for a detailed description on the development of Afrikaans). In the early twentieth century, a fully standardized variety of Afrikaans was established, and since then a consistent and widely available record of the written language has been preserved. The focus of this article is on future time reference in Afrikaans, and in particular, the alternation between the two future auxiliaries.

There are two ways to explicitly indicate future reference grammatically in Afrikaans: one is the modal auxiliary sal ‘will/shall’, with a strong future implicature, inherited from Dutch zullen ‘will/shall’ with the same uses. The other is the verb gaan ‘go’, inherited from Dutch gaan ‘go’, which is used as a lexical verb and an auxiliary in both Afrikaans and Dutch. That these auxiliaries (sal and gaan) have been appropriated for future time reference is not surprising, as sal followed the volition path (volition/desire > intention > future) and gaan the movement path (movement toward a goal > intention > future) of grammaticalization, both of which are rather common grammaticalization routes for future auxiliaries (Bybee 2007; Heine 2003).

One of the differences between sal and gaan, apart from their origins, is the modal connotation of each2. On the one hand, sal is traditionally regarded as a modal auxiliary and has the potential to indicate intention or volition (Wybenga 1993), apart from predictions. On the other hand, gaan expresses futurity without intention, but rather in terms of prediction, and has been described as a ‘purer’ form of future reference (De Stadler 1992). This corresponds to the distinction made by Dahl (2000) between intention-based futures and prediction-based futures—sal comfortably expresses both, while gaan is mostly used for prediction-based futures.3 This difference in modal connotations is discussed in Kirsten (2019, p. 99). While Conradie (2018) groups sal and gaan together when discussing modal chains in Afrikaans, for instance, the discussion of the different modal uses focuses on sal, and the examples illustrate that sal cannot be replaced by gaan in all contexts (p. 265).

The Afrikaans gaan is directly related to the Dutch gaan, and it is useful to keep the parameters of use for Dutch gaan in future time reference in mind when investigating the use of the Afrikaans construction. The same goes for the Afrikaans sal and the Dutch zullen. In Dutch, the alternation between gaan and zullen patterns according to a number of factors, including the lexical verb, the proximity of the future referenced, transitivity, aspect, and animacy of the syntactic subject (Hilpert 2007).

Regarding the use of be going to/gonna for future reference in English, there are also several factors at play. Since English is also related to Afrikaans, it is worth considering as well. Tagliamonte (2012) lists the syntactic subject, the proximity of the future referenced, the lexical verb, and the clause type. Poplack and Tagliamonte (2000) also specifically include the grammatical person of the subject as a factor that could influence the choice between the different future references.

In previous research, it has been determined that some of these factors are relevant for the alternation between Afrikaans sal and gaan—the formality of the text (Kirsten 2019), clause type, transitivity, and subject animacy (Kirsten 2018a). However, as pointed out by Kirsten (2018a), these investigations only include written data, and furthermore leave certain other variables unexplored.

In this article, the focus is on more recent uses of sal and gaan in Afrikaans (since the 1970s), and it includes spoken data to fill the gap left by previous research. The main aim of this article is to determine what the alternation between gaan and sal in recent Afrikaans language use entails in both spoken and written data, and to briefly contextualize it in the West-Germanic language family. Relating Afrikaans to the West-Germanic language family is motivated by its absence in other studies focusing on future time reference in the broader Germanic language family, such as Hilpert (2008). Finally, this article could also further illuminate the ongoing grammaticalization of gaan, although it is not a main focus.

2. Materials and Methods

2.1. Materials

Variation and change in language use can often be linked to register, and how language is used differently in different registers. Register is about how a language user performs in specific contexts and settings (Tagliamonte 2012), and so refers to varieties based on different situations of use (Biber and Conrad 2009). Kruger and Van Rooy (2018) further point out that register represents a multidimensional constellation of features that are characterized by probabilistic co-occurrence relationships. Biber and Gray (2013) recommend that language change be investigated relative to particular registers, rather than, for instance, an average for a particular language.

Previous studies of register variation have indicated a number of important differences between registers, especially between written and spoken language use (Biber and Conrad 2009; Biber 2012). As Biber and Conrad (2009, p. 261) phrase it: “speech and writing differ in their potential for linguistic variation”. Writing permits a range of linguistic expressions while speech is more constrained in its typical linguistic characteristics. In a study of Spanish clitic climbing, for instance, Cacoullos (1999) found that register is a much stronger contributing factor to variation than social factors such as class and education of speakers.

The spoken data used in this article represent the 1970s and the 2000s—the 1970s are represented by the Ponelis Corpus of Spoken Afrikaans (PCSA) (Ponelis 1976), and the 2000s by the Spoken Corpus of Afrikaans from the early 2000s (SCA) (Van Rooy 2003). The former consists of more or less 555,000 words and the latter, 72,000. Because of the small size of the Van Rooy (2003) corpus, its results are interpreted with caution—no strong claims can be made based on this small corpus but adding it to the study can still add value and detail to the bigger picture.

The written data used in this study come from the Historical Corpus of Written Afrikaans (HCSA) (Kirsten 2018b). The corpus spans 1911 to 2010 and includes data from every third decade. For the purposes of this article, I include the last two decades that coincide with the spoken data—1971–1980 and 2001–2010. Each decade consists of more or less 260,000 words and represents a number of different written registers, including a small portion of manuscript data (i.e., letters and diary entries).

Adding further corpora was considered, especially because of the small size of the most recent spoken corpus. However, since there aren’t other recent or contemporary corpora of spoken Afrikaans available, that would entail adding a different register. One significant option would be internet language, such as the WatKykJy Corpus drawn from a blog for a specific target group, or a corpus of web comments, both available on VivA’s corpus portal (VivA 2022). However, there are not any equivalent data for the 1970s, and the comparative aspect between the two periods and the registers would be lost. For that reason, it was decided to only include the corpora already mentioned.

2.2. Methods

In the analysis of Afrikaans future time reference, there are two variable auxiliaries that are equivalent to some extent, sal and gaan, as previously discussed. In alignment with the principle of accountable analysis (Tagliamonte 2012), on the one hand all of the relevant forms should be included in the analysis, but on the other hand, those contexts that are not functionally parallel should be excluded. Regarding sal and gaan, this means that not all of the instances of either should be included in the analysis.

As far as gaan is concerned, it can be used in different grammatical functions—as a lexical verb (1), as a linking auxiliary verb (2), and as a future auxiliary (3).

• Nee kyk, dis ‘n man saam met wie jy oorlog toe kan gaan. (2000s written)
  no look it.s a man together with whom you.SG war to can go
  ‘Look, it’s a man with whom you can go to war.’
• So, so mense moet maar self gaan kyk. (1970s spoken)
  so so people must but self go look
  ‘So people should go and look for themselves.’
• Hulle gaan hom môre merk. (1970s written)
  they go him tomorrow mark
  ‘They are going to mark him tomorrow.’

For obvious reasons, only those in the category of future auxiliary are included in the analyses for this article. Distinguishing between the two auxiliary uses is not, however, straightforward, and is specifically discussed in more detail in the next subsection.

The verb sal is only ever used as an auxiliary, so there are no lexical verb uses to exclude. There are some modal uses that do not actually include a future implicature (4) though, which are excluded in accordance with the principles applied in Wagner and Sankoff (2011). There are, furthermore, future references with participant-internal modality (intention-based futures) (5) that are not interchangeable with gaan, and these are also excluded here. (See Kirsten 2019, pp. 92–94 for a discussion of the distinction between participant–internal and participant–external modality in the future reference uses of sal, which is replicated here.)

4.

Hoe sal ek nou weet, Shane. (2000s written)

how shall I now know Shane

‘How should I know, Shane.’

5.

Ek sal die storie maar van voor af begin. (1970s written)

I shall the story but from beginning of start

‘I will start the story from the beginning, then.’

After determining which instances of both sal and gaan to include in the analysis, the next step was to code every instance for several factors that could condition the variable. As Jankowski (2013) mentions, previous research can be a useful source for potential predictors (or factors, or independent variables) to code for in the data. Previous research on future time references in Germanic languages indicates a number of variables that could be relevant for the alternation between sal and gaan: lexical verb, proximity of the future, transitivity, clause type, and different aspects of the syntactic subject including animacy, number, and person (Hilpert 2007; Tagliamonte 2012). As mentioned before, previous research on Afrikaans future time reference suggests that, among others, formality, clause type, and subject animacy all influence the alternation between gaan and sal in Afrikaans (Kirsten 2018a, 2019). In this article, future proximity and sentence type are added for a more comprehensive analysis. Future proximity is included because of its significance in Dutch (zullen vs. gaan) and English (will vs. be going to/gonna) (Van Olmen and Mortelmans 2009). Sentence type is included because of its effect on the alternation of will and be going to/gonna in English, where be going to/gonna is favored with both interrogative and negative sentences (Tagliamonte et al. 2014). Regarding the subject, number is not included because Afrikaans has no verbal inflection for number requiring subject-verb-agreement, as is the case for be going to, for instance. Person is added, however, because its influence might be similar to Dutch and English.

The identification of all the relevant instances and coding each instance for the different factors prepare the data for the descriptive statistics and interpretation of the results. The basis of the descriptive statistics, apart from the normalized frequencies, is the relative frequencies—the frequency of a variant compared to the frequency of the other variant (in all interchangeable contexts), or as a proportion of the total occurrences of the variables take together. This provides an overall distributional picture of the data. It should be noted that the relative frequencies of specific variables in different corpora are not compared to each other, but to the overall relative frequency of gaan vs. sal in the specific corpus.

In two of the subsections, additional statistical analyses were performed in the form of distinctive collexeme analysis. In this analysis, alternating pairs of constructions, such as gaan and sal, are compared in terms of their relative preferences for words or categories that can occur with both of them (Gries and Stefanowitsch 2004). It is well suited to investigate near-synonymous constructions and the words that appear in particular slots in the constructions. The analysis indicates if specific words are distinctive for each construction, that is, if its use with specific words can significantly distinguish it from the other. The distinctiveness is the measure indicating the strength of the association between the word and the specific construction, and the higher it is, the more significant the distinction is. In the formula used for this paper (Gries 2022), a log likelihood ratio is used for significance testing, in which case a result of 6.63 or higher means the likelihood of the difference being due to chance is 1% or less. The threshold for a 5% or less likelihood of the difference being due to chance is 3.84. To do these analyses, the script by Gries (2022) was used in the program RStudio (RStudio Team 2020).

2.3. Distinguishing Future gaan

There are two auxiliary uses of gaan, and this section provides details about how the future gaan is distinguished from the linking verb gaan, since only the former is relevant for this article.

Regarding the non-lexical uses of gaan, some scholars such as De Villiers (1971), Ponelis (1979), and Conradie (2020) only refer to gaan as an auxiliary with different meanings or functions, without differentiating between the different meanings or functions on a syntactic level. Both Breed (2012, p. 71) and Wierenga (2022, p. 167), however, acknowledge that there is more than one auxiliary usage expressed by gaan, and that these are grammatically and functionally distinct.

The auxiliary gaan is often described as a (direct) linking verb (Cavirani-Pots 2020; Conradie 2020; De Vos 2005), and sometimes ascribed modal meanings (Conradie 2020; De Villiers 1971, p. 16; De Vos 2005). Instead of this focus, Wierenga (2022, p. 66) describes gaan in terms of three syntactic and semantic contexts. The first is a hybrid lexical-grammatical use, the one typically described as a linking verb, such as in (2) above. This usage includes some measure of aspectual meaning, and more specifically, the inchoative aspect (Wierenga 2022, p. 66). Then there is a more grammaticalized aspectual use of gaan, as illustrated in (6), where it is used as a marker of the phase-specifying inchoative aspect. Finally, the most grammaticalized use of gaan is exemplified in (3) above, where it is used for future reference.

6.

Fred se maag gaan aan die bewe. (1970s written)

Fred gen.ptcl stomach go on the shiver

‘Fred’s stomach starts to shiver.’

There are some important differences between the different auxiliary uses, apart from their meaning (aspectual and andative versus temporal). Afrikaans sentences typically have a V2 structure—if there is only one verb, it will take the V2 position, and if there are auxiliaries, one of them will take the V2 position, and any remaining auxiliaries will cluster around the main verb (usually at the end), as in example (2) (see also Conradie 2006, p. 86). When gaan is used as a linking verb, it does not take the V2 position if there are any other auxiliaries, such as modals (for example moet ‘must’ in 2) or tense markers (for example het ‘have’ in 7).

7.

… ons het daar op die damwal gaan sit… (1970s spoken)

we have there on the dam.wall go sit

‘… we went and sat there on the dam wall…’

Example (7) also illustrates the use of linking verb gaan in the past tense, something that obviously does not occur when it is used for future reference. Regarding the more grammaticalized aspectual use of gaan as shown in example (6), it is used in very specific fixed expressions (Wierenga 2022, p. 167) which were excluded from the data sets.

The use of gaan as a future marker defaults to the V2 position, even when there are modals (8) or linking auxiliaries (9). This is similar to the syntactic position of sal.

8.

…en daarom gaan die kulture ook nooit kan meng nie… (2000s spoken)

and therefore go the cultures also never can mix neg

‘…and for that reason the cultures will never be able to mix…’

9.

…ek gaan nou daai gaan doen… (1970s spoken)

I go now that go do

‘…I will go and do that now…’

As illustrated in (9), the future reference and linking verb uses of gaan can and do occur in the same sentence, although typically in informal and spoken-language use. This is further confirmation of the distinction between these two uses on a syntactic and semantic level.

The syntactic position of gaan is an important factor in determining whether it is an instance of a future reference or a linking verb, especially in relation to modals and other auxiliaries. If the clause is in the past tense, it would also indicate its use as a linking verb. In the very few instances that remained unclear, it was excluded from the data set rather than incorrectly included.

3. Results

3.1. Usage per Corpus

Before delving into more specific aspects of the data, it is useful to get an overall picture of the use of gaan and sal in Afrikaans.

There are two obvious observations to be made from

Table 1. Frequencies per 10,000 words for each corpus.

	GAAN	SAL
1970s written	5.4	29.7
1970s spoken	11.9	22.5
2000s written	8.5	27.7
2000s spoken	29.3	12.8

. Firstly, the use of gaan as the newer variant is much higher in spoken-language use than in written-language use. Secondly, the use of the newer variant increases with time, in both spoken and written usage. Neither of these observations are unexpected—it is well known that new variants are taken up and spread more quickly in spoken language, and previous research has already indicated that the grammaticalization of gaan is ongoing (Kirsten 2019). The data also show that the increase of gaan is at the expense of sal, where a higher frequency of the former accompanies a lower frequency of the latter in interchangeable contexts.

3.2. Time: 1970s vs. 2000s

The first division of the data is time, divided into 1970s and 2000s.

Table 2. Relative frequencies divided by time.

	GAAN		SAL
	per 10,000 Words	% of Future Ref	per 10,000 Words	% of Future Ref
1970s	9.8	28%	24.8	72%
2000s	13	35%	24.5	65%

shows that the use of gaan increases with time. This is expected, as previous studies focused on written language (Kirsten 2019) already indicated as much. This is simply a confirmation of the pattern with spoken data added to the analysis.

3.3. Register: Written vs. Spoken

The second factor to consider is register, with the broad categories of written and spoken. It also serves as a proxy for formality, since written language generally is more formal and spoken language is more informal.

Table 3. Relative frequencies in the different registers.

	GAAN		SAL
	per 10,000 Words	% of Future Ref	per 10,000 Words	% of Future Ref
written	7	20%	28.7	80%
spoken	13.9	39%	21.4	61%

shows the relative frequencies of the two variants in the data, divided according to register.

Again, the results are more or less what is expected. Since gaan is the newer variant undergoing grammaticalization, its higher relative frequency in the spoken data compared with the written data fits the general pattern that written data is more conservative in adopting new variants. The same pattern can be seen in Dutch, where gaan is more common in spoken language (Beheydt 2005).

3.4. Lexical Verb

The next factor is verb type, regarding the type of lexical verb the future auxiliary is used with. The most significant type of verb in this regard, according to studies on related languages, is motion verbs (6). It has been said to have had an effect in English (Poplack and Tagliamonte 2000) and in Dutch (Hilpert 2007) at some point in time. Here, in

Table 4. Raw and relative frequencies with different verb types according to corpus.

1970s Written	GAAN		SAL		1970s Spoken	GAAN		SAL
motion	7	17%	34	83%	motion	30	33%	60	67%
other	163	15%	747	85%	other	629	35%	1192	65%
overall		15%		85%	overall		34%		66%
2000s Written	GAAN		SAL		2000s Spoken	GAAN		SAL
motion	9	35%	17	65%	motion	13	72%	5	28%
other	216	23%	716	77%	other	198	69%	87	31%
overall		23%		77%	overall		70%		30%

, it is the only type distinguished, with the rest simply categorized as other.

10.

…hy gaan nou eers weer loop. (2000s spoken)

he go now first again walk

‘He’s gonna leave now.

Interestingly, only the most recent written data show a clearly higher relative frequency of gaan with motion verbs compared to other verbs (see Table 4). If it does influence the alternation between gaan and sal, then, it must be a more recent development, and not a longstanding factor. For Dutch, the frequent use of gaan with motion verbs is an earlier phenomenon, with the preference declining over time (Hilpert 2007). In English, be going to was initially averse to being used with motion verbs (Poplack and Tagliamonte 2000), and it never developed a preference for it (Hilpert 2007). While these trends are opposite in Dutch and English, there are currently no strong preferences in either. Chances are, thus, that the same goes for Afrikaans, and the preference in only one corpus is not meaningful.

Because the focus on motion verbs does not reveal much, an additional analysis was performed: distinctive collexeme analysis with the main verbs as a variable.

In the 1970s written corpus (

Table 5. Distinctive collexemes for gaan and sal in 1970s written corpus.

gaan (N = 143)		sal (N = 781)
Collexeme	Distinctiveness	Collexeme	Distinctiveness
maak	5.297955475590710	wees	6.064734654782740
luister	3.982654580283170	hê	5.115862841584680
kyk	3.982654580283170	hou	4.085376599174950
gebeur	3.898217581241480

), none of the lexical verbs distinguish gaan from sal with 99% or more certainty, but there are some with a 95% or higher certainty (indicated in italics). For gaan, the most distinctive verb is maak ‘make’, followed by luister ‘listen’, kyk ‘look’, and finally gebeur ‘happen’. Only the latter verb is not an action verb performed by an agent. In contrast, the most distinctive verb for sal is wees ‘to be’, followed by hê ‘to have’ and hou ‘keep’ or ‘like’, which are either stative verbs or not really an action, even if an agent is involved.

The results of the distinctive collexeme analysis performed on the 1970s spoken corpus is in

Table 6. Distinctive collexemes for gaan and sal in 1970s spoken corpus.

gaan (N = 659)		sal (N = 1252)
Collexeme	Distinctiveness	Collexeme	Distinctiveness
maak	21.217557727287200	wees	20.287611253982700
probeer	10.667269014440200	vergeet	17.035097608655100
speel	9.759001989956970	regkom	11.864231709894100
antwoord	8.529827125107510	weet	8.116486028486250
saamvat	8.529827125107510	sien	6.311338636189710
gesels	7.092320708276080	glo	5.087171577991120
word	7.782845180271590	ken	5.087171577991120

. The most distinctive verb for gaan is again maak ‘make’, followed by: probeer ‘try’, speel ‘play’, antwoord ‘answer’, saamvat ‘take along’, gesels ‘chat’, and word ‘become’. Apart from the last verb, word, all of the others are actions performed by agents again. Both antwoord and gesels are communicative verbs, and saamvat is a motion verb. Regarding sal, the most distinctive verb is wees ‘to be’ again, followed by: vergeet ‘forget’, regkom ‘get better/right’, and weet ‘know’. While vergeet and weet require an agent, they are mental verbs without action. The remaining less significant verbs for sal are sien ‘see’, glo ‘believe’ and ken ‘know’, which are also mental verbs. They contrast with luister and kyk from Table 5, which are deliberate actions, and are more distinctive of gaan in that corpus. It shows, then, that the distinctive verbs in both corpora from the 1970s are similar in nature—active and deliberate actions attract the use of gaan for future reference, while stative and (passive) mental verbs attract the use of sal.

In the written corpus of the 2000s, the picture is somewhat different (

Table 7. Distinctive collexemes for gaan and sal in 2000s written corpus.

gaan (N = 225)		sal (N = 733)
Collexeme	Distinctiveness	Collexeme	Distinctiveness
gebeur	9.731059905018040	neem	4.821569803202440
aanbied	5.825325993229690	[help]	3.745609523500110
lag	5.825325993229690	[leer]	3.745609523500110
nooi	5.825325993229690	[voortgaan]	3.208598393634110
opstel	5.825325993229690	[weet]	3.208598393634110
ry	4.776537572826410
toelaat	4.776537572826410

). The only highly distinctive verb is the collexeme gebeur ‘happen’ with gaan. The rest of gaan’s collexemes have a lower level of distinctiveness: aanbied ‘offer’, lag ‘laugh’, nooi ‘invite’, opstel ‘set up’, ry ‘ride’, and toelaat ‘allow’. The most distinctive verb with gaan is, in contrast with the pattern from the 1970s, not an action requiring an agent. There are communicative verbs still, such as aanbied and nooi, and the movement verb ry. In terms of sal, only one verb is distinctive at all, with a low level of distinctiveness: neem ‘take’. Interestingly, it is an action verb, and one that would often include movement when it is used in terms of taking something or someone to a destination. The other collexemes in the table (in square brackets) are included for interest’s sake but are not significant in their distinctiveness: help ‘help’, leer ‘learn’, voortgaan ‘go ahead’, and weet ‘know’.

The 2000s spoken corpus is particularly small, so it is not surprising that one of the variants do not have any significant collexemes (see

Table 8. Distinctive collexemes for gaan and sal in 2000s spoken corpus.

gaan (N = 211)		sal (N = 92)
Collexeme	Distinctiveness	Collexeme	Distinctiveness
[behandel]	3.640585518281850	praat	12.146500036226400
[gee]	3.441222810613520	hou	7.240638711457050
[doen]	2.931224452613840	lees	7.240638711457050
[uitwerk]	2.906618356077150	sien	7.240638711457050
[staak]	2.906618356077150	weet	7.240638711457050

)—the top collexemes of gaan are behandel ‘treat’, gee ‘give’, doen ‘do’, uitwerk ‘work out’, and staak ‘stop’, although none of these are significant. For sal, the most distinctive collexeme is praat ‘talk’, followed by hou ‘keep’ or ‘like’, lees ‘read’, sien ‘see’, and weet ‘know’. While several of these are mental verbs, the top distinctive verb is the communicative praat.

The distinctive collexeme analysis paints a fairly consistent picture of the 1970s apart from gebeur in the spoken corpus, but not so much in the 2000s. This could be an indication of ongoing change, where the semantic restrictions on the use of the future gaan declines as its frequency increases. The clear preferences of the 1970s do not feature in the 2000s—gaan is used more with action verbs (including motion verbs) and communication verbs, apart from gebeur ‘happen’; sal is used more with stative and passive mental verbs. In the 2000s, the only verb that distinguishes gaan with high significance from sal is gebeur, although the low-level distinctive verbs still include communication and motion verbs. For sal, it is more readily used with motion verbs and communication verbs, even though mental verbs still feature strongly. These patterns could link with the next variable, which regards the syntactic subject.

3.5. Subject Animacy and Person

In terms of the syntactic subject, I included two variables. The first variable is subject animacy—I distinguish between animate subjects and inanimate subjects (see Tagliamonte et al. 2014; Hilpert 2007). Furthermore, I looked at subject person as a variable, distinguishing between first and non-first person subjects (see Poplack and Tagliamonte 2000; Tagliamonte et al. 2014). Since first-person subjects are necessarily human and animate, it made sense to collate the data for animacy and person, as was done in Tagliamonte et al. (2014). In

Table 9. Raw and relative frequencies with different subjects according to corpus.

1970s Written	GAAN		SAL		1970s Spoken	GAAN		SAL
first person	29	35%	54	65%	first person	217	47%	243	53%
non-1st animate	61	17%	306	83%	non-1st animate	260	32%	563	68%
inanimate	53	11%	421	89%	inanimate	182	29%	447	71%
overall		15%		85%	overall		34%		66%
2000s Written	GAAN		SAL		2000s Spoken	GAAN		SAL
first person	61	45%	75	55%	first person	69	68%	32	32%
non-1st animate	77	22%	276	78%	non-1st animate	81	73%	30	27%
inanimate	87	19%	382	81%	inanimate	61	67%	30	33%
overall		23%		77%	overall		70%		30%

, the distinction is made between first person (all animate), non-first person animate, and inanimate (all non-first person).

The results from Table 9 show two trends. Firstly, the relative frequency of gaan is higher with first-person subjects than the overall relative frequency in three of the four corpora. Secondly, the relative frequency of sal is higher with inanimate subjects. These patterns are visible in all of the corpora except the 2000s spoken corpus, where all three categories are fairly close to the overall relative frequencies.

While the trends are clear, proportional frequencies do not reveal significance as such. To test the strength of these trends, distinctive collexeme analyses were performed for each of the corpora to determine the distinctiveness of each of the types of subjects with the two future auxiliaries (see

Table 10. Distinctiveness of gaan with first person and sal with inanimate subjects per corpus.

	First Person	Inanimate
	gaan	sal
1970s spoken	42.5141333427973	12.9809002576663
1970s written	21.4230268570147	13.829026735698
2000s spoken	0.931148905455146	0.413545291095626
2000s written	35.7439306518278	12.5549547230611

).

The analyses confirmed that non-first-person animate subjects are not distinctive in any of the corpora. The trends from the proportional data are corroborated regarding the first-person and inanimate subjects—the former is distinctive of gaan, and the latter distinctive of sal. The results from the 2000s spoken corpus are not significant still, probably because of the small size of the corpus.

The Afrikaans differs from the Dutch, where gaan prefers inanimate subjects (Hilpert 2007). Regarding English, be going to favored animate subjects initially, but the preference has levelled out with time, and the same goes for first-person subjects (Poplack and Tagliamonte 2000; Tagliamonte et al. 2014). Afrikaans, then, shows a similar pattern to earlier English use.

3.6. Clause Type

The next factor to consider from the data is clause type. In studies on English, the distinction is between main clauses and subordinate clauses (Tagliamonte et al. 2014; Poplack and Tagliamonte 2000). Distinguishing these clause types in Afrikaans may seem straightforward, as the word order is typically different in subordinate clauses—in subordinate clauses, the V2 position falls away and the main verb and auxiliaries cluster toward the end of the clause. However, complement clauses, such as the object clause of a speech-act verb, are often used without a complementizer, and the word order then remains unmarked. Furthermore, in spoken language, the word order would sometimes remain unmarked even when a subordinator is used, and the spoken data used here do have a few such cases. There are different ways to approach this analysis, then. I first made the distinction based on word order, reported in

Table 11. Raw and relative frequencies with different clause types according to corpus (1).

1970s Written	GAAN		SAL		1970s Spoken	GAAN		SAL
main	79	13%	535	87%	main	495	34%	948	66%
subordinate	64	21%	246	79%	subordinate	166	35%	304	65%
overall		15%		85%	overall		34%		66%
2000s Written	GAAN		SAL		2000s Spoken	GAAN		SAL
main	137	20%	542	80%	main	145	73%	53	27%
subordinate	88	32%	191	68%	subordinate	66	63%	39	37%
overall		23%		77%	overall		70%		30%

.

Secondly, the distinction was conceptualized in terms of embeddedness: simplex clauses, coordinated clauses, and matrix clauses (with embedded clauses) in one category, and embedded clauses (regardless of word order) in the other category. This resulted in

Table 12. Raw and relative frequencies with different clause types according to corpus (2).

1970s Written	GAAN		SAL		1970s Spoken	GAAN		SAL
main/matrix	71	12%	510	88%	main/matrix	491	32%	844	68%
embedded	72	21%	271	79%	embedded	270	40%	408	60%
overall		15%		85%	overall		34%		66%
2000s Written	GAAN		SAL		2000s Spoken	GAAN		SAL
main/matrix	124	19%	522	81%	main/matrix	113	70%	43	30%
embedded	101	32%	211	68%	embedded	98	70%	49	30%
overall		23%		77%	overall		70%		30%

.

Table 11 and Table 12 show similar patterns: gaan is more likely to be used in subordinate/embedded clauses than in main clauses, but not consistently. The written data show this preference consistently, but not the spoken data. In Table 11, the distinction between the clause types is very small for the 1970s spoken data, and in Table 12 it is non-existent for the 2000s spoken data. In Table 12, gaan’s preference for embedded clauses is clear in the 1970s spoken data, but in Table 11 this preference is reversed for the 2000s spoken data. What can be concluded from this is that gaan’s attraction to subordinate/embedded clauses is clear in the more conservative written data, but not in the spoken data.

Subordinate/embedded clauses favoring gaan is similar to the usage pattern of be going to in English (Tagliamonte et al. 2014); however, Poplack and Tagliamonte (2000) indicate that this preference decreased with time. It is possible that the same is happening in Afrikaans, where this preference is still present in the more conservative written language and to an extent in the older spoken language, but not in the more recent spoken language any longer.

3.7. Sentence Type

Related to clause type is the factor of sentence type. In line with Tagliamonte et al. (2014), I distinguish between affirmative (11), interrogative (12), and negative (13) sentences. Sentences are marked negative only when the negation is in the clause where gaan or sal occurs. Interrogatives include all sentences formulated as questions.

11.

Hulle sal gelukkig wees hier. (1970s written)

they shall happy be here

‘They will be happy here.’

12.

OK wat gaan jy vir my vertel die laaste tien minute… (2000s spoken)

okay what go you.SG for me tell the last ten minutes

‘Okay what are you going to tell me in the last ten minutes?’

13.

Maar ook dit, weet Arboreta, sal nie werk nie. (2000s written)

but also it know Arboreta shall not work not

‘But also this, Arboreta knows, will not work.’

The majority of sentences are affirmative, and

Table 13. Raw and relative frequencies with different sentence types according to corpus.

1970s Written	GAAN		SAL		1970s Spoken	GAAN		SAL
affirmative	109	15%	661	85%	affirmative	504	34%	995	66%
interrogative	25	51%	26	49%	interrogative	90	57%	69	43%
negative	9	9%	94	91%	negative	65	26%	188	74%
overall		15%		85%	overall		34%		66%
2000s Written	GAAN		SAL		2000s Spoken	GAAN		SAL
affirmative	168	21%	626	79%	affirmative	145	64%	83	36%
interrogative	23	50%	23	50%	interrogative	37	100%	0	0%
negative	34	29%	84	71%	negative	29	76%	9	24%
overall		23%		77%	overall		70%		30%

shows that the spread of sal and gaan remain fairly close to the overall distribution of each corpus in affirmative sentences. The other sentence types show clear patterns throughout all of the corpora. The relative frequency of gaan is much higher in interrogative sentences in all of the corpora. Regarding the negative, there is an interesting development. In the 1970s data, negative sentences with sal are more frequent than its proportional frequency, whereas this changes to gaan in the 2000s data. In English, negative sentences favor be going to, so it seems like Afrikaans is developing in the same direction in recent decades.

3.8. Future Proximity

The last factor considered here, shown to be relevant for future alternation in related languages, is that of future proximity. The distinction is made between proximate (14), which is known to affect the alternation in Dutch and English, and other (15).

14.

Hier sal hy my nou vannag kry. (1970s spoken)

here shall he me now tonight get

‘He will get me here tonight.’

15.

Dit gaan nog belangrik word. (2000s written)

it go still important become

‘It is going to become important.’

Table 14. Raw and relative frequencies with different future proximities according to corpus.

1970s Written	GAAN		SAL		1970s Spoken	GAAN		SAL
proximate	43	30%	102	70%	proximate	111	63%	66	37%
other	100	13%	679	87%	other	548	32%	1186	68%
overall		15%		85%	overall		34%		66%
2000s Written	GAAN		SAL		2000s Spoken	GAAN		SAL
proximate	37	45%	46	55%	proximate	38	79%	10	21%
other	188	21%	687	79%	other	172	68%	82	32%
overall		23%		77%	overall		70%		30%

indicates that gaan is used more frequently with reference to proximate futures than in general, which is not a surprising trend given that similar trends can be found in Dutch and English (Van Olmen and Mortelmans 2009). This pattern is present in all four corpora, indicating that it is a long-term, persistent trend.

Interestingly, the effect of future proximity differs across studies, according to Tagliamonte et al. (2014), possibly due to different classifications used in different studies. However, as mentioned before, both Dutch gaan and English be going to are typically associated with proximity.

4. Discussion

In this discussion, the overall trends will be pointed out, after which signs of possible ongoing language change will be discussed. The differences between the registers will then receive attention, and the factors that remain stable. The following subsection will compare the findings for Afrikaans with what we know about future construction alternations in both Dutch and English.

The overall picture painted by the data in this study highlights the ongoing grammaticalization of gaan. One of the effects of grammaticalization is an increase in frequency—the token frequency of future gaan increases with time, and it is more frequent in spoken language than in written language. This pattern is accompanied by a mirror image in the use of future sal, which is lower the more frequent gaan is.

Another indicator of grammaticalization is an expansion of usage contexts, or the relaxing of certain restrictions or preferences. With the use of gaan, there are two possible manifestations of this phenomenon. Firstly, the lexical verbs used with the different future auxiliaries are more distinct in the 1970s than in the 2000s. Initially, the preference for gaan with concrete and communication verbs, and for sal with stative and mental verbs, is clear. In the 2000s, however, these distinctions do not hold, with the agentless gebeur ‘happen’ being the most distinctive for gaan, and the communicative praat ‘talk’ being the most distinctive for sal.

There is another indication of ongoing change, which regards the sentence type. While interrogatives remain favorable for the use of gaan throughout, negatives show possible change. In the 1970s, negatives favor the use of sal, but this flips to gaan in the 2000s. This pattern in the more recent data is similar to that in English.

There is one clear difference between the registers: subordinate clauses favor the use of gaan only in the written data. While this could be a case of ongoing change progressing quicker in the spoken language, it patterns differently than the other indicators of possible change. For that reason, it could also be register variation, where clause type simply does not contribute to the choice in the same way in spoken and written language.

The contribution of the remaining factor to the alternation between gaan and sal is fairly consistent and so does not show signs of change or differences between registers: gaan is more likely to be used with reference to proximate futures, and this remains consistent through all of the corpora.

Comparison with Dutch and English

In this section, the two variants will be compared to their counterparts in Dutch and English, starting with sal.

In Dutch, zullen occurs freely with stative verbs, does not occur with willen as main verb, and is associated with more distant future events (Hilpert 2008). Regarding the English will, it is the more formal future reference, and is more preferred in main clauses (Szmrecsanyi 2003). The negative won’t is the preferred negative form in British English, but not used as much in American English.

Afrikaans sal shows some parallels with its counterparts, as well as some differences. Its preference for stative verbs (similar to the Dutch zullen) is clear in the 1970s, although it is not in the more recent data. While specifically distant futures are not investigated here, gaan is more attracted to proximate futures than sal, similarly to English be going to and will. Differently from English will, it is not more preferred in main clauses, which in Afrikaans shows no clear preference for either of the variants. In terms of the negative, there is not a common negative contraction like English won’t, but it is used more readily in negative sentences in the 1970s.

The English counterpart for gaan is be going to and its shorter form gonna, and in Dutch it is also gaan. In English usage, be going to/gonna tends to combine with movement and speech-act verbs (Hilpert 2008)—with gaan, the same can be seen in the 1970s, although not in the 2000s. In the later period, it is used more similarly to the Dutch gaan, which is often used with verbs with a low degree of agentivity (Hilpert 2008). In terms of the proximity of the future that it references, gaan tends toward proximate rather than unspecified or distant futures, which corresponds with both the English and the Dutch variants (Hilpert 2008). The same preference for be going to/gonna in subordinate clauses (Szmrecsanyi 2003) is seen with gaan, but interestingly only in the written data. Finally, both be going to/gonna and Dutch gaan is more frequent in informal and spoken language (Beheydt 2005), which is similar to Afrikaans gaan.

The clear similarities in usage patterns between the movement futures, on the one hand, and the modal futures, on the other hand, between Dutch and English are also visible for Afrikaans, along with some differences that are to be expected. Where Dutch and English differ, Afrikaans does not tend to pattern similarly to one or the other consistently—this shows that while the Afrikaans future auxiliaries originated from Dutch, and while Afrikaans is in continuous contact with English, there were and are developments in usage which are unique to Afrikaans.

5. Conclusions

The main aim of this article is to determine what the alternation between gaan and sal in future time reference in Afrikaans entails, and to relate it to similar constructions in other Germanic languages. The alternation is described and discussed, and brief mention is made of similar alternations in Dutch and English—the two closely related languages with similar alternating future constructions.

The alternation between gaan and sal in future reference has been investigated before, especially in terms of recent and ongoing changes. However, the addition of spoken data brings another dimension to these analyses and reveals some new insights where usage patterns are different between written and spoken data. The same goes for the distinctive collexeme analyses.