Between exonormative traditions and local acceptance: A corpus - linguistic study of modals of obligation and spatial prepositions in spoken Ugandan English

: Research into Ugandan English places it in the nativisation phase of the evolution of Englishes, amidst a nexus of local acceptance with ingredients of endonormativity and ingrained exonormative traditions. The current study shows how the use of modal verbs of obligation and spatial prepositions provides insights into how the nexus of the above phenomena has shaped Ugandan English. For example, although the preference of have to over must is a global trend, in Ugandan English, it is more prevalent in Bantu - speaking than in Nilotic - speaking areas because of substrate in ﬂ uence. Crucially, although the use of spatial prepositions is generally similar to how they are used in, for example, ( standard ) British English, the peculiar use of from to encode stative location in Ugandan English is, despite some regional variations, so widespread in the country that it tends towards endonormative stabilisation.


Introduction
In Englishes spoken in post-colonial contexts, exonormativity, nativisation, and eventually endonormativity constitute part of the trajectory in their development as proposed by Schneider's (2003Schneider's ( , 2007 Dynamic Model. During the nativisation stage, phonological innovations and structural distinctiveness emerge and heavy borrowing takes place because of linguistic forces such as substrate effects, language contact, second language (L2) acquisition processes, as well as sociocultural reasons (e.g. when English lacks expedient means to express local sociocultural realities; see Kachru and Smith 2009). The phase follows on from a phase of exonormative stabilisation (with which it slightly overlaps and in which the linguistic processes of nativisation kick off), during which English first language (L1) settlers, administrative officials, or social workers maintain the norms of, mostly, British English (BrE), whilst at the same time adopting a "British plus" identity, "seasoned with the additional flavor of the colonial experience" (Schneider 2007, 37). Nativisation is followed by an endonormative stabilisation phase, during which a positive attitude towards the features of the local variety develops, despite the fact that resistance to these features will also manifest itself among some sections of the population. break up "begin holidays" is seen as speaking broken English in Uganda, where break off is instead used. On the pragmatic plane, Meierkord (2016b, 245) shows that, contrary to British norms, directness in making requests is not seen as inappropriate or impolite in UgE. In fact, in Uganda, not using direct strategies when making requests is construed as though the requests are not urgent or even not serious, as a result of substrate influence (Meierkord 2016b). Although the above cases are most likely a result of the fact that Ugandans may not be aware of their existence in BrE, they nevertheless evince some aspect of incipient but strong endonormativity, because using the relevant BrE forms is stigmatised in Uganda. Thus, although UgE has reached the nativisation phase in the sense of Schneider (2003Schneider ( , 2007, some indicators of endonormativity are visible, despite the dictates of the historical exonormative orientation. It might be the case that there are intersectional elements that straddle Schneider's (2003Schneider's ( , 2007 model at the transitory level from nativisation to endonormative stabilisation, where aspects of the latter manifest themselves in the former, on a par with how aspects of nativisation begin manifesting themselves right in the phase of exonormative stabilisation, as Schneider (2003Schneider ( , 2007 himself points out (see also Ugorji 2015 for the context of Nigeria). Crucially, endonormativity involves situations where a new form that has emerged during the nativisation stage "has lost its former stigma and is positively evaluated" (Schneider 2007, 50), which is typically reflected in its growing and more widespread use.
The remainder of this article discusses two grammatical areas, modals of obligation and spatial prepositions, to identify potential nativisation processes that may have moved towards endonormative stabilisation versus uses that rather point towards an exonormative orientation, hypothesising that the three may coexist in the variety (namely in one area there is exonormative orientation but in another endonormative orientation as the nativisation process continues to take place). Modals of obligation and spatial prepositions in UgE provide grammatical instantiations that point to this coextensiveness.

Modals of obligation 3.1 Modals of obligation in Uganda's L1s
Modals of obligation are expressed in two main ways in the majority of Uganda's first languages (L1s), namely by means of grammaticalised verbs and by means of affixes (Bavin 1995, Kawalya et al. 2018. Evidently, there are other means of expressing obligation in these languages, that is, lexical means (e.g. using adverbs) and the imperative mood (Bavin 1995, Kawalya et al. 2018. In this study, we will pay attention to the use of grammaticalised verbs and affixes. Because the latter are prototypically appended to verbs, this limits our study to the use of verbs in expressing the obligation modality. Kawalya et al. (2019) analyse three Luganda modal verbs and one affix that encode obligation, namely, -teekw-, -lina, -etaag-, and -andi-, the latter being an affix. According to them, the modal -teekw-, which they gloss mainly as "must" (and sometimes as "have to") involves participant-imposed necessity and mainly encodes directive obligation, as in the following example (p. 365, 368): (1) [ …]a-teekw-a o -ku-awul-ibw-a mu kalantiini 3SG-PRES.must-FV IV-INF-separate-PASS-FV in quarantine "He must be put in quarantine." On the other hand, -lina, which they prototypically gloss as "have" (notably it comes from the Luganda equivalent of the lexical verb "have"), also involves participant-imposed necessity; however, unlike -teekw-, it not only encodes directive obligation, but also combines it with moral desirability as well, as shown in the example (2a) (p.372) below. Both -teekw-'must' and -lina 'have to' thus encode strong obligation, albeit with important differences in their semantics. Furthermore, to realise participant-inherent necessity, Luganda uses the modal verb -etaag-, which Kawalya et al. (2019, 375) gloss as "need (to)," whereas the affix -andi-, which Kawalya et al. (2018) gloss as "should," is said to encode deontic necessity. The examples in (2) illustrate the above phenomena (Kawalya et al. 2019, 377, Kawalya et al. 2014: "People who need to be operated on." (c) […]era singa kisoboka na-bo b-andiyanguye okwenenya and if possible and-them 3PL-shouldhurry to apologise "And if it is possible, they should also apologise quickly." According to the authors, (2a) expresses the fact that although the necessity is imposed by the speaker, it is also morally desirable for the subject referent to follow societal norms, whereas (2b) encodes participant-internal dynamic necessity, namely the necessity is determined by internal circumstances to the subject referent, although the use of the modal verb -etaag 'need' can also trigger a participant-imposed reading if there are contextual factors outside the control of the subject referent. As for (2c), it is said to express deontic necessity, not in the sense of directivity, but in terms of "ethical norms." as well as "personal opinions and principles" (Kawalya et al. 2018, 97)², which can be equated to what has been known as weak obligation (see, e.g., Wasserman and van Rooy 2014). Taylor (1985, 164) indicates that the conditional marker -kaain Runyankole (another Bantu language) also expresses the obligation modality but he glosses it as 'ought to,' which he says is usually encoded by means of -shemereire, literally meaning 'be-right' as in the following example: (3) O-shemereire kw-e-tega You-be-right to-self-shave "You ought to shave." Our Runyankole consultants have provided -shemereire as the equivalent of 'should.' They have also shown that Runyankole has -teekwand -ina, which they gloss as 'must' and 'have (to)', respectively. These two forms are the equivalents of Luganda -teekwand -lina, respectively. Kawalya et al. (2019, 268-372) show that in contemporary Luganda, -lina occurs more frequently than -teekw-.
In Ugandan Nilotic languages, particularly, in the Luo language cluster, the picture is different from what we have observed above. Specifically, in Lango and Acholi, different shades of obligation semantics are usually encoded using the modal verb (o)myero (Odonga 2012, 328 spells it as myerro), which encodes both strong and weak obligation, whereas participant-inherent necessity is typically encoded by the verb mito "want" (Bavin 1995, 128). The Acholi example in (4a) is adapted from Odonga (2012, 328), whereas (4b) is from Bavin (1995, 113):  1 An anonymous reviewer has pointed out that the word here should be okulongoosebwa. However, in Kawalya et al. (2019, 377), where the example was extracted, it is written as okulongoosa.
You should go "You should go." Myero a-ngweci Must 1SG-run "I must run." As can be seen from the idiomatic gloss, myero is rendered as "should" in (4a) and as "must" in (4b).
What is clearly discerned from the above are the crucial differences between the Nilotic languages and Bantu languages. Although the Bantu languages have several modals they use to encode obligation in its various semantic forms, the Nilotic languages quintessentially have one modal verb (myero), although they can also use mito when encoding participant-inherent necessity. Collins (2009) investigates the use of modals and quasi-modals in world Englishes. He mainly compares Inner Circle Englishes with Outer Circle Englishes. Among Inner Circles, he considers AmE, BrE, Australian English, and New Zealand English, whereas among Outer Circles, he uses Hong Kong English, Indian English, Kenyan English, Philippine English, and Singapore English. Collins (2009) observes that as far as the quasi-modal of obligation have to vs the modal verb must is concerned, the Outer Circle Englishes use have to more than must, with Hong Kong English taking the lead. But, overall, he observes that there is a sharp increase in the use of have to across all the varieties of English under consideration in his study. According to him, the exponential use of have to to the detriment of must is premised on the semantics of the two verbs. He states that although "the primary meaning of both have to and must in contemporary English is (strong) deontic necessity…," a crucial difference between the two is that "the former is more commonly objective, the source of the obligation being external to the speaker," whereas the latter is "more commonly subjective, the speaker typically being the source of the obligation" (p. 287). As a result, he states that deontic have to is more attractive than must to contemporary speakers of English because of its "greater objectivity and consequently less overtly authoritarian tone" (p. 287). At the same time, the preference for have to reflects a general trend of quasi-modals rising at the expense of modals in both BrE and AmE as discussed by Mair and Leech (2006). In this respect, Collins' (2009, 286) data reveal that Kenyan and Indian English are the most conservative varieties in that they have retained uses of modals more than other varieties. We should also note that, in all varieties of English, there are syntactic contexts where must cannot be used, e.g. in past tense, where a form of have to has to be used instead (cf. Kotze and van Rooy 2020). Collins (2009, 288) also reveals that should is as popular as have to in both Inner Circle and Outer Circle Englishes because of its "milder subjectivity and consequently less forceful tone." However, of the two blocs of varieties of English, Outer Circle Englishes have considerably higher frequencies than Inner Circle Englishes. Among the Outer Circle Englishes, Kenyan English shows the least tokens of usage. On the other hand, have got to is generally dispreferred in Outer Circle Englishes with just 150 tokens per million words compared with 268 tokens in Inner Circle Englishes (p. 286). Hong Kong English stands out among the Outer Circle varieties in the usage of have got to.

Modals of obligation in L2 Englishes
In his study of Philippine English, Imperial (2014) looks at how speakers of this variety of English use the modal of obligation must and the quasi-modal of obligation have to. Just like Collins (2009), Imperial (2014) also found out that have to was more popular in Philippine English than must. He attributes the decline of must and the rise of have to to the modality system of Filipino. However, Collins et al. (2014) find that uses of must with deontic meaning nevertheless continue to dominate over those with epistemic meaning in Philippine English as well as in BrE and AmE. On the other hand, Seidel (2004) found out that should was more prevalent in Kenyan English than must. Contrary to the trend taken by the L2 varieties of English listed in Collins (2009) andImperial (2014), van Rooy and Wasserman (2014) have established a different direction taken by Black South African English (BSAfE), where have to has only shown very little increase in usage and this is not statistically significant. However, should has substantially increased in usage but again "below the criterion level for statistical significance" (p. 58), albeit just barely below that level. As in other varieties, must has declined in BSAfE as well, but again below the level of statistical significance. A noticeable and statistically highly significant increase was observed with need to.
4 Spatial prepositions 4.1 Spatial prepositions in Uganda's L1s One property of many locatives (spatial prepositions) in Ugandan languages (both Bantu and non-Bantu) is their polysemous nature. For example Noonan (1992, 107) states that the Lango (a Nilotic language) locative i is used to express stative, allative, or ablative semantics, where English uses different prepositions, for example: at/in (stative), to/into (allative), from (ablative). Likewise, although Ssempuuma (2019, 127-8) shows that the Luganda (a Bantu language) locative e is used with both stative and allative semantics, Sternefeld and Sseguya (2017) state that it is also used with ablative semantics. Remarkably, a language such as Lango, for example, will use its polysemous locative i to express the prepositional relations expressed by the three English prepositions, as follows: A-tye a-wot i poto. 1SG-be PROG-go to garden "I am going to the garden." (c) A-tye a-dwogo i poto 1SG-be PROG-return from garden "I am returning from the garden." Rutooro would use the locative ha for all the above relations, whereas Luganda would use the locative ku-. Isingoma (2012, 115;2021, 4) observes that such locatives are clearly underspecified in terms of directional properties, namely [±dir] properties, such that the feature [±dir] and the activation of [+dir] or [-dir] are induced by the predicate in the clause. Thus, because the verb tye 'be' is not locomotional, the locative i in (5a) is [-dir], but since wot 'go' and dwogo 'return' are locomotional, the locative i in (5b) and (5c) is [ + dir]. In addition, the predicate (or the context) will determine whether the motion is allative or ablative (see also Noonan 1992). However, Acholi, which belongs to the same family as Lango, has a special expression it uses in addition to i to encode ablative motion, namely ki, which Blackings (2009, 56) renders as "from." Thus, (5c) would be rendered in Acholi as in (6):

(6)
A-tye ka-dwogo ki i poto 1SG-be PROG-return from in garden "I am returning from the garden." Observably, ki 'from' cannot be used alone. Rather, it must be used coextensively with the locative i. Noonan (1992, 171) states that in Lango, a language very close to Acholi, having contiguous locatives or prepositions is the norm and such locative expressions must involve the locative i. However, Lango does not have the locative phrase ki i 'from.' Thus, unlike in Lango, in Acholi, as far as the ablative motion is concerned, the predicate has no role in determining this relation, because even predicates that are not locomotional are used with the locative phrase ki i to encode ablative semantics, for example the equivalent of I saw him from the window in Acholi is realised with the locative ki i as the equivalent of "from." What is more is that, in Acholi, this locative phrase is also used before noun phrases in clauses where ablative semantics seems to be absent, such as in (7).
Gulu University Lit. "I studied from Gulu University" "I studied at Gulu University." As can be seen in the English idiomatic translations, the sentences above do not have ablative semantics, and in Lango (a very close language to Acholi), they would be rendered with the locative i 'at, in' alone. However, some sections of Acholi speakers say that the renditions with i 'at, in' are also possible (albeit not common) in Acholi. Whether this could be a result of dialectal differences or influence from Lango or gradual language change is not clear to us, and we cannot pursue that in this study.
As mentioned earlier, another important property of Uganda's indigenous languages is the fact that they also express spatial relations using morphologically complex predicates known as verb extensions or derivations. One such prominent verb extension mechanism is known as applicativisation. By way of definition, Isingoma (2020, 31-2) states, "A well-researched verbal extension in Bantu languages, applicativisation adds a participant to the argument structure of a verb (although in some cases it just realigns the arguments). In Rutooro, this is realised by affixing -ir-(or -er-, -ras phonologically conditioned variants) to the root of the verb" (see also locus classicus works on applicatives, e.g. Baker 1988, Mchombo (ed.) 1993. Applicatives license a wide range of thematic roles such as benefactive, goal, instrumental, locative (with, e.g. stative, allative, ablative semantics), among others (see also Jerro 2017 for similar observations in relation to Kinyarwanda). For example, in Rutooro, the three spatial relations above are encoded by means of applicatives in (8): In (8a), the applicative is obligatory so as to indicate the location where the playing took place. If it is omitted and the location is maintained, the sentence becomes ungrammatical. The interpretation is that the applicative encodes the meaning of what the English preposition "in" would encode. Likewise, in (8b), the applicative provides the meaning encoded by "to"(that is the allative or goal semantics), and in (8c), it is the ablative (source) semantics that is realised by the applicative. It should be noted that the choice of whether to use a locative expression or a locative applicative or both is contingent upon several semantic and syntactic factors, which we cannot pursue here.
What this section tells us is that spatial relations are encoded in Ugandan L1s differently from how they are encoded in English. In addition to the polysemous nature of locatives in Uganda's indigenous languages, the fact that there is also applicativisation, equally characterised by polysemy, as well as idiosyncratic uses of some locative phrases, for example ki i 'from' in Acholi, points to possible interlanguage and ultimately nativisation or even endonormativity. Already, Schmied (2004, 930) points out that "besides analogy, interference from African languages is possible, since their prepositional system is relatively simple and thus polysemous."

Spatial prepositions in L2 Englishes
Research into how L2 Englishes use prepositions (e.g. Jibril 1991, Mwangi 2004, Ssempuuma 2019, among others) reveals notable divergences between L1 and L2 Englishes (even though there are substantial convergences as well). For example Jibril (1991) highlights the pervasive non-distinction between the spatial prepositions on and in among Hausa speakers of Nigerian English and attributes this to the fact that Hausa uses one particle, that is -a, to mean both on and in. This echoes Schmied's (2004) observation, already mentioned above, as regards the polysemous nature of locatives/prepositions in African languages and the resultant interference from these languages in the Englishes spoken in Africa.
In fact, Schmied (2004, 934) also mentions the role of analogy in the departures observable among East Africans, where phrases such as at Nairobi are common, despite the fact that Nairobi is not "a point in a global perspective."³ Similarly, Schmied (2004Schmied ( , 2017 notes the use of in instead of into is also because of analogy (e.g. as in […]many people are just coming in the country [Schmied 2004, 935]). A similar observation is made by Mwangi (2004, 27-8) as regards Kenyan English, where a distinction between in and to as well as on and onto is sometimes not followed and Kenyans would use in instead of to and on instead of onto, thereby blurring the distinction between the locational and directional semantics encoded by locational in and on and directional to and onto.
In addition to analogy and substrate influence, Takan (2013) argues that there are other two routes that Cameroonians use in the adoption of idiosyncratic usage of spatial prepositions in English, which also involves similar departures to those observed in Kenyan English or East African English. Takan (2013) mentions "logicalisation" and linguistic nationalism. He defines "logicalisation" as "a strategy used by speakers of a language to unconsciously "rectify" features that could be considered irregular, odd or abnormal in terms of the general patterning of the language" (p. 27). Thus, Cameroonians would use in a bus instead of on a bus to "rectify" or simply regularise what they see as anomalousness in standardised BrE and AmE. But one could observe here that this "logicalisation" may not be detached from the different dimensions covered by analogy or regularisation in language development (see Ellis 1994 andMeierkord 2004 for discussions of regularisation). As for linguistic nationalism, Takan (2013, 28) explains that this manifests itself as a way of indirectly refusing linguistic "westernisation," where local norms are promoted. He states that "even Cameroonians who have attained a high level of education are often observed to resist speaking or, at times, writing StE." This, he says, borders on endonormative stabilisation as proposed by Schneider (2007). However, one could state that this resistance may not be a result of deliberate efforts; rather, it might be linked to regularisation, where local forms are seen as more "logical" than given "illogical" BrE/AmE forms.
Although there has been sporadic mention of the peculiar use of spatial prepositions in UgE (namely Isingoma 2014 and Nassenstein 2016), it is Ssempuuma (2019) who delineates this aspect of UgE. In the three studies above, there is a revelation about the peculiar use of from in UgE (9), a feature not reported in any other variety of English (to our knowledge): The Kenyans studying from the two universities are… (Isingoma 2014, 53) (b) Where will we watch the match from? (Nassenstein 2016, 409) (c) We used to pray from the chaplaincy of Mulago hospital. (Ssempuuma 2019, 193) As can be seen, BrE and AmE would use at or zero preposition where UgE uses from, namely UgE uses from to encode stative location, as opposed to ablative location. Both Isingoma (2014) and Ssempuuma (2019) do not explicitly account for this special use of from, whereas Nassenstein (2016, 408-9) attributes it to the polysemous locative applicative in Bantu languages described in Section 4.1. However, because of the different realisations of the locative systems of Ugandan L1s (see Section 4.1), it appears that there is more than the applicative phenomenon behind this usage, or there appears to be something completely different from applicativisation as the crucial factor that influences this usage, as will be discussed in Section 7.

Data collection and participants
The use of modals of obligation and of spatial prepositions in UgE has been investigated on the basis of authentic conversational data. The data set comprises a total of 120 transcripts of semi-structured discussions (resulting from a transcription of 20 min each of a total of 120 h of conversations, of which 40 were collected in each of the three locations below) totalling 40 h, which were subsequently analysed. All discussions involve at least three speakers of diverse L1s and were recorded with year six secondary school and university students as well as professionals (ranging from hair dressers and seamstresses to mobile money agents and sales representatives)⁴ in Gulu, Kampala, and Mbarara. These locations were chosen for the fact that the likelihood of an influence of the English of Luganda L1 speakers on the Englishes of speakers with other L1s differs across the three. In Kampala and Mbarara, the selected schools and universities have, because of their prestige, sizeable intakes of pupils/students with Nilotic (and Central Sudanic) as well as Bantu L1s. By contrast, Bantu-speaking students hardly ever enrol in boarding schools or universities in the Gulu area, because prestigious institutions exist closer to their home areas. Hence, direct influence from regular interaction with Bantu L1 speakers is unlikely here. Similarly, work migration tends to be into the more affluent (and for a long time less restive) Bantu-speaking, Central and Western regions of the country rather than into the Nilotic North (Gulu). The distribution of speakers across the various Ugandan L1s and the respective language families is summarised in Table 1.
One third of the speakers were secondary students attending year six, aged 18 and 19, the second third were university students, and the final third professionals. The data have a slight bias towards male speakers, who make up 101 versus only 82 female participants.

Data analysis
After the raw conversational data had been transcribed and annotated, following the International Corpus of English (ICE) procedures, it was investigated using WordSmith Tools. The data were subsequently searched for the individual modals of obligation (have to, have got to, should, may, might, must, need to,  4 The professionals had finished at least secondary education, whereas many had also ventured into or even completed tertiary education. The main difference between them and the secondary school and university students is that they are no longer under the constant influence of prescriptive attitudes of the formal education institutions. ought to, shall, will, and would)⁵ and spatial prepositions (at, by, from, in, into, on, onto, and to) in declarative sentences. Hits were then manually assessed and verified as relating to expressions of obligation (epistemic uses were excluded) and spatial relations and then entered into an Excel sheet. In a next step, the individual entries in the Excel sheet were marked for location of recording, speakers' L1 and speakers' occupational status (high school students, university students, professionals). Furthermore, in the case of modals of obligation, findings were coded as regards person and number of the grammatical English † † 2 4 2 No information available 1 † The figures for this/these language(s) include our Ugandan colleagues who were present in various recordings: one Rutooro, one Lusaamia, and two Acholi speakers. As interviewers, they were participants in the recordings at all three locations. Hence, the "Bantu per location" and "Nilotic per location" differ from the total Bantu and Acholi. † † Individual Ugandan families, particularly those where spouses speak mutually unintelligible L1s, for example Runyankole and Acholi, have shifted to using English as their home language and raise their children in English. This is particularly pronounced in the upper middle class.
 5 Although will, shall, and would are primarily used in L1 English to encode various semantics, they are also used to encode obligation (Huddlestone andPullum 2002, Quirk et al. 1985). At the same time, Mesthrie and Bhatt (2008, 65) mention may as a polite form of expressing obligation in Indian English, whereas the eWAVE (Kortmann et al. 2020) mentions, generally, the nonstandard use of modals for politeness reasons in L2 Englishes. subject, animacy of the direct object, and declarative sentence type (positive versus negative). Prepositions were coded as regards scope (point, movement away, and movement towards an object), animacy of the agent, and potential non-standard uses with standardised BrE serving as the exonormative reference variety. Where assessments could not be clearly made, the entry was coded as "unclear."

Results
This section reports on those results that yielded clear influence of one of the independent variables mentioned just above. In case no influence was visible from the results of our analyses, that is for the grammatical variable potentially affecting uses of modals of obligation and spatial prepositions, results are not discussed here.

Modals of obligation
In total, 813 tokens of modals of obligation were found in the data. These fell under the modals have to (n = 442; 54.4%), should (n = 197; 24.2%), need to (n = 102; 12.5%), must (n = 61; 7.5%), and shall (n = 11; 1.4%). The modals have got to, may, might, ought to, will, and would were not observed in the data with obligation semantics. Apparently, just like other L2 varieties of English, UgE speakers make use of a restricted set of those modals of obligation that English makes, generally, available. Examples (10) to (13) illustrate their uses in the data (note that in these examples, bold print has been added to highlight the individual modals and quasi-modals):

(10)
Because sometimes it's the rainy season and you have to go to the garden and then you also have to be at school.
I should make sure I have a public address system.
But for them, when an elder is talking to you, you need to at least humble yourself and kneel down, you know.
But same, we must work hard so that we improve lifestyles of people (…) yes.
Across the three data collection locations, their use was distributed as captured in Table 2 and Figure 1: Although the overall preference of have to to express obligation can be observed in all three locations of data collection, there is a considerably lower amount of this item in the Gulu data, where, at the same time, should and need to have higher frequencies than in the other two locations. As such, Uganda follows the trend of all other Englishes worldwide, in that the semi-modals are preferred over the modals. However, the differences between the three locations are not statistically significant (p = 0.288, χ 2 = 9.68, df = 8).
Although the three locations differ as regards the dominant L1 spoken in any of them (with the Nilotic language Acholi being the dominant one in Gulu and the Bantu languages Luganda and Runyankole dominating in Kampala and Mbarara, respectively), informants generally had mixed L1 background because of migration. Furthermore, the few speakers of Central Sudanic L1s cannot be adequately catered for in a discussion of the localities. Table 3 and Figure 2, therefore, differentiate between the language groups captured in our data. As Figure 2 reveals, there is a considerable and statistically significant (p = 0.021, χ 2 = 18.04, df = 8) difference between the Nilotic, the Central Sudanic, and the Bantu speakers, with Nilotic speakers having a considerably smaller share of have to and larger shares of both need to and must. The amount of raw data on which Figure 2 relies differs from that used for the locations captured in Figure 1, because a number of speakers had not clearly stated their L1 or were not able to single out one such language because of their bilingualism. Also, individuals who had stated English as their L1 were not included in these calculations, for obvious reasons. However, the behaviour of speakers with Nilotic L1s captured by the numbers and percentages in Figure 2 closely matches the one of the location in which they are typically encountered, Gulu, as visualised in Figure 1. Similarly, the Bantu speakers' behaviour seen in Figure 2 is similar to that in both Kampala and Mbarara, where Bantu speakers are the dominant ethnic group.
In addition to location and L1 background, speakers' uses of the various modals also differ depending on their occupational status as either secondary or tertiary students or professionals. As is visible from Table 4 and Figure 3, there is no clear overall trend but two noticeably, seemingly unrelated, differences. On the one hand, university students have considerably higher shares of must and lower shares of should, in   comparison to both the secondary school students and the professionals. Together with the professionals, their uses of need to considerably exceed that of the secondary school students. These differences, however, are not statistically significant (p = 0.099, χ 2 = 13.39, df = 8).

Spatial prepositions
The analyses of expressions of spatial relations focused on the prepositions at, by, from, in, into, on, onto, and to. In declining frequency, these were observed in the data as summarised in Table 5, mostly with a semantics that matches that of standardised BrE. As Table 5 shows, the most frequent preposition is in, followed by to, at, and from, whereas onto was not observed in the data at all. Overall, the table documents that the speakers overwhelmingly used the individual prepositions with their standard BrE semantics. Although to and in are the prepositions with the highest (>95%) shares of standard uses, at and into have shares between 95 and 90%, and the most notable non-standard uses can be observed with on and from.   This differs from the rank list of prepositions that Mindt and Weber (1989, 233) established for the frequencies in the London-Oslo-Bergen (LOB) corpus. Although their list does not differentiate between spatial and temporal prepositions, in and to occupy the first and second position in both their and our analysis. Mindt and Weber's (1989) list also includes of, for, with, and as, which in their full list occupy ranks one, four, five, and ten, respectively. The rankings here refer to those that emerge when of, for, with, and as are deleted from their list to arrive at a ranking that only includes prepositions that can have a spatial semantics. Crucially, there is a clear difference as regards at, which is third in our data but fifth in their list for the LOB. At the same time, on is fifth in our data but third in the LOB. Furthermore, from, which occupies the fourth rank in our data, is ranked at six in the LOB frequencies. In sum, these latter differences point towards nativisation as regards the preferences for individual prepositions, namely at, on, and from, in UgE.
In a next step, the non-standard uses of the prepositions were assessed as regards precisely which standard BrE uses the speakers had replaced. The comparison was made with BrE as this is the historical and still current reference variety in Uganda. Although the low overall amounts on non-standard uses of   into, on, and to did not lend these prepositions to a meaningful investigation in this respect, the following trends can be summarised for the other prepositions:⁶ at tends to replace zero prepositions (4), in (4) and to (3), for example: (14) but my wife when I reach at home, she's annoyed in some times (Mbarara, Runyankole) (15) and there the fruits they get at their villages. (Kampala, Luganda) (16) We always goes outside at the beach. (Gulu, Lango) from tends to replace zero prepositions (21), and at (5) (17) But I also pray from home. (Gulu, Acholi) (18) You grew up from somewhere else. I grew up from the Central region. (Gulu,Lusamia) in tends to replace at (12), into (5), on (4), and to (4) (19) And is she in Gulu University as well or somewhere else? (Gulu, Acholi) (20) even when you give birth to a child and bring the child in a certain environment, the child is likely to adopt. (Kampala, Lango) (21) But it was what was what matters in campus is now much intake they can actually handle.
(Mbarara, Luganda) (22) It has it has eroded away culture especially for people who just come in Kampala. (Kampala, Luganda) Consistent patterns can, thus, be observed for both from and in, bearing in mind that, still, most of their uses in the data are standard. A further investigation into from revealed that the non-standard uses were in the Gulu and Kampala data only, with none found in the Mbarara recordings. Also, the Gulu data have an extreme preference for uses of from where standard BrE would for example use the preposition at (as in Example 17), that is, 14 of the 18 non-standard uses, as Table 6 documents. Similarly, the majority of nonstandard uses of in, that is 15, were observed in the Gulu data (for instance, Example 19), as opposed to six and five in Kampala and Mbarara.
Interestingly, the shares of non-standard uses at the three locations differ from preposition to preposition, as further summarised in Table 6.
Although Gulu has the highest share of non-standard as opposed to standard uses when it comes to the preposition from, Mbarara has standard uses only. On the other hand, Kampala informants have the highest share of non-standard uses of on, whereas for this preposition there are no non-standard uses in the Gulu data.

Discussion
Our results in Section 6 show that UgE exhibits not only aspects of nativisation, some of which tend towards endonormative stabilisation, but also exonormative orientation. Nativised forms may be homogeneous across the country or more pronounced in one region than another because of specific substrate influence. As we have seen, the preference of the semi-modal have to in UgE over the modal must to encode obligation may be said to be a global pattern, as it is also preferred in other Englishes, for example Philippine English (Imperial 2014), including L1 Englishes (Collins 2009), possibly because of its semantics, namely, encoding "less overtly authoritarian tone" (Collins 2009, 288). In addition, a similar pattern exists in some Uganda's L1 languages, for example Luganda, where -lina 'have to' is more frequently used than -teekw-'must' (Kawalya et al. 2019), a situation that is extrapolable to Runyankole.
Crucially, it is important to note that nativisation becomes more apparent in the use of modal verbs in UgE when we consider the results we obtained in the region where Nilotic languages are spoken as L1s. This speaker group had fewer instantiations of the use of have to than L1 speakers of Luganda or Runyankole ( Table 3). This may be due to the fact that Nilotic languages such as Acholi and Lango use the same verb, (o) myero, to encode both weak and strong obligation (Bavin 1995), as shown in Section 4.1. Importantly, none of the available dictionaries of Acholi/Lango (e.g. the Fountain Lango-English and English-Lango Dictionary 2008, Odonga 2012 andBlackings 2009) states that myero means 'have to,' although Bavin (1995) very sporadically renders it as "have to" (e.g. p. 121). It is rather rendered as must or should. It is, thus, not surprising that L1 speakers of the Nilotic languages above do not use the form have to as frequently as speakers of Luganda or Runyankole, where specific lexical equivalents for have to exist, which are, moreover, also used more frequently than any other modal verb of obligation in these languages.
It might be right to posit that despite the fact that L1 speakers of Nilotic languages do not use the semimodal verb have to at the same level as L1 speakers of Luganda and Runyankole, the observed occurrences of have to among them could be attributed to exonormativity observable in the process of acquiring English at school.
Remarkably, because myero means 'must' (see Bavin 1995, 112-3;Fountain Lango-English and English-Lango Dictionary 2008, 129), the high incidence of the occurrence of must among the Nilotic speakers compared with the Bantu speakers also speaks to nativisation, because Nilotic languages do not have a specific lexical item that expresses the semantics of have to and may end up using have to and must more or less interchangeably, whereas Bantu (for instance Luganda) speakers appear to distinguish the two based on the same distinction made in their L1(s) (see Kawalya et al. 2019).
Nativisation is also at work in UgE when one considers the absence of many L1 English modal verbs of obligation in our data (see Section 6.1). Verbs such as have got to, ought to, will, and would were not observed in our results and this could be because Uganda's L1s have a very limited number of modal verbs of obligation, with Acholi and Lango having typically one verb (myero), although they also use mito 'want' to mean 'need to,' whereas Luganda and Runyankole have just four. This may mean that for Ugandans the different shades of meaning encoded by the above English modal verbs of obligation are usually subsumed under have to, must, should, and need to.
The use of modals of obligation, as well as that of spatial prepositions in UgE, generally, depicts exonormative orientation, because most of the uses do not deviate from standard BrE, despite the fact that the L1s have far fewer modals and prepositions than Standard English (StE) and that these cover the semantics of various StE modal verbs and prepositions. This is understandable, because Ugandans typically acquire English at school, where standard BrE is the norm. Grammar books, for example Songa and Songa (2010), provide prescriptions on the use of modal verbs and prepositions following BrE norms.
However, despite the observed general exonormative orientation, the dictates of substrate influence (and other second language acquisition processes) make Ugandans deviate from standard uses in some cases. In the case of spatial prepositions, prominent among the deviant uses is the use of the preposition from. In agreement with Isingoma (2014), Nassenstein (2016), and Ssempuuma (2017) (see Section 4.1), our findings (see Section 6) indicate that, in UgE, from is also used to encode stative location, a use that does not exist in L1 English or as far as we can tell in other L2 varieties of English. This is a clear case of nativisation given that its provenance seems to be associated with an indigenous language. Specifically, as shown in Section 4.1, Acholi (a Nilotic language) uses the locative phrase ki i 'from' with noun phrases where no ablative semantics is involved. The role of Acholi here becomes more apparent when we consider the fact that speakers in Gulu (where Acholi is spoken natively and as the dominant language) had more cases of from used to encode stative location than any other speaker group, as shown in Table 6 in Section 6.2. The above state of affairs reinforces Meierkord's (2016c) observation that in contradistinction to Fisher's (2000) and Isingoma's (2013) assertion, UgE is not influenced by Luganda only; rather, each Ugandan L1 influences UgE in a way. Adokorach and Isingoma (2020) have gone ahead to show that there are some phonological aspects of UgE used across the country that stem from substrate influence from Nilotic languages. This peculiar use of from seems to be the first grammatical aspect shared by different L1 speaker groups in Uganda to be shown as having stemmed from Acholi (a Nilotic language). The idea by Nassenstein (2016) that this use of from is a result of the polysemous nature of the Bantu applicative system is farfetched, as such a feature should have been reported to exist in the hitherto widely studied African varieties of English where Bantu influence is prominent, for example BSAfE, Kenyan English, and others. Perhaps, the polysemous nature of the Bantu applicative system could have contributed indirectly by virtue of using the same form to encode different spatial semantics, thereby facilitating the adoption of the use of from to encode both stative and ablative semantics stemming from "Acholi English" influence. The apparent spread of aspects of "Acholi English" to the rest of the country could be attributed to their hegemonic role in the military during the protectorate times, as well as politics, the military, and civil service after independence in 1962 up to 1986 (see Amone 2015). As stated by Meierkord (2016a, 66), the 1970s saw locally trained teachers take charge of English teaching in the country, thereby "allowing for nativisation to proceed quickly in the country." Although there are no official statistics, there is anecdotal evidence whereby in Bantu areas, there is a belief that the Northerners (where the Acholi are one of the dominant groups) are the best teachers of English. Gulu University (located in the north of Uganda) is frequently approached by headteachers from across the country to send them teachers of English who are Northerners. Hence, it may not be surprising to have a feature from this ethnolinguistic group spread to other parts of the country.
Although our data (see Section 6.2) show that the above usage was not observed in Mbarara, where Runyankole (a Bantu language) is the dominant language, there is some evidence of the use of this form by Ugandans hailing from this region. For example the President of Uganda, who hails from this region, said, in one of his messages, Let us continue praying from our homes (Museveni 2020). However, it is not clear whether this was uttered analogously with the idiomatic StE work from home, which implies ablative motion, as one working from home has to interact in some way with one's routine workplace either via the phone or email. What is evident is that in acceptability tests that were conducted alongside the interviews (see Meierkord 2019, 8-9), there was overwhelming acceptance of this usage of from in Mbarara as well (i.e. on a Likert scale ranging from 1 for "fully acceptable" to 4 for "fully unacceptable," the item involving the preposition from received a mean value of 1.5). Thus, the use of from to encode stative location could simply be said to be a feature that cuts across the country. It is so pervasive that we have seen it used in an idiomatic translation in a local bilingual dictionary, namely He/she was operated upon from outside Uganda (Rubongoya 2013, 360). Thus, the level of pervasiveness and overwhelming acceptance of from as a stative locative could be considered to depict endonormative stabilisation as is the case with other features mentioned in Section 2.
We should note that different from the predominant use of the modal of obligation must in the Northern part of Uganda, the pervasive use of what can be said to be originally Northern UgE from as a stative locative in the whole country might have more readily been embraced by the Bantu (from other regions of Uganda) due to the fact that, as mentioned earlier, Bantu languages use a polysemous device to encode the meaning of in, at, and from. Hence, although for obligation, Bantu languages have different lexical items to mean must, should and have to and can thus be said not to have been influenced by the Northern predominant use of must, the polysemous device used in Bantu languages to mean in, at, and from could be said to have facilitated the adoption of from for stative semantics.

Conclusion
In sum, because of historical reasons (that is the fact that Uganda is a former protectorate and not a colony), UgE still presents substantial ingredients of exonormativity, but strong waves of nativisation have taken shape and patterns of endonormative orientation are visible at least with respect to some aspects of the features of the variety. Thus, UgE is a good example of cases where there are aspects of overlap in the different phases of development proposed by Schneider (2003Schneider ( , 2007 in such a way that features of three phases (exonormativity, nativisation, and endonormativity) manifest themselves coextensively, although at the moment the first two phases appear to dominate in Uganda, because so far only a few forms (e.g. the use of from with stative semantics) could be said to have been proved to be positively evaluated (cf. Schneider 2007, Meierkord 2019 by Ugandan speakers of English with concomitant widespread usage. It might be interesting to find out from other African Englishes whether the features described in this article manifest themselves in a more or less similar way to how they do in UgE, i.e. in terms of the coextensiveness of exonormativity, nativisation, and endonormativity, bearing in mind the fact that Uganda is a former protectorate as opposed to countries such as Kenya and Zimbabwe, which are former colonies. As Meierkord (2016a) points out, Ugandans, as citizens of a protectorate, have had to rely mainly on formal loci for the acquisition of English, whereas citizens of colonies have had both formal and informal loci for their acquisition of English. Author contributions: Both authors have accepted responsibility for the entire content of this manuscript and approved its submission. Both BI and CM collected and sorted the data, with the assistance of research assistants. BI wrote Sections 1, 2, 4, 7, 8, whereas CM wrote Sections 3, 5, 6. These were later exchanged for harmonisation.