4.1.1 Source markers

Contemporary Thai has a few source markers (ablatives) denoting ‘from’, consisting of either a single word or two words, the most frequent one being càak. Examples of these ablatives taken from various sources are shown below in (1) (note that in the examples, complex forms, as in (1e), are shown as a single form with a period between the components):

A detailed discussion of the ways that each of these ablatives was grammaticalized is beyond the immediate scope of the present study, but a brief description helps understand how multiple forms of (near-)synonymous functions arise. (Some relevant discussions are found in Park, Reference Park2017a, Reference Park2017b; Park & Rhee, Reference Park and Rhee2018; and Rhee, Reference Rhee2021.) The ablative càak in (1a) is grammaticalized from the heterosemous verb denoting ‘leave, depart’. As indicated above, grammaticalization that began in verb serialization, i.e., ‘fly out leave’ has been reinterpreted as ‘fly out from’, along with morphosyntactic reanalysis of [V-V] into [V-preposition]. A similar pattern is found with tâŋ in (1b), whose lexical origin is the verb denoting ‘start’, as well as with (1e), which involves càak ‘leave’. The ablative tɛ̀ɛ in (1c) is unique in that, in its development, the ‘departure’ notion, central in ablative, seems to have come from the adversative semantics of its source tɛ̀ɛ ‘but’. The development of ablative troŋ in (1d) (also found in (1e)) is likely to have been motivated by the conceptual association of ‘straightness’, ‘straightforwardness’, and ‘definiteness’. When it modifies a spatial location (troŋ thîi, lit. ‘straight place’), it would acquire the emphasis meaning (‘the exact place’), which, in turn, is subjected to the reanalysis triggered by the cooccurring verb ‘bring out’; thus, the string eventually became ‘bring out from’. Schematically, the conceptual progression may be [bring out straight place] > [bring out this exact place] > [bring out exactly from]. The ablative càak.troŋ merits special attention. Thai ablatives and allatives include a large number of polylexemic forms involving multiple lexemes, each of which is, in many cases, already ablative or allative (note that càak and troŋ are both ablatives). This type of lexicalization and grammaticalization has been noted in previous studies (Khammee & Rhee, Reference Rhee and Khammee2023, Reference Khammee and Rhee2024).Footnote ³

A list of the inventory of Thai ablatives exemplified above, totaling six, along with their lexical sources, is shown below in (2):Footnote ⁴

The list in (2) clearly shows that Thai ablatives are predominantly V-adpositions, most of which have their origin in verb serialization. The N-adposition is not strongly represented since the only candidate troŋ is also (and, in fact, primarily) an adjective. Another notable factor is that in some of these, an identical form surfaces in polymorphemic forms, giving the impression that the form is ‘changing partners’ to create periphrastic forms of equal or similar function. This characteristic is more pronounced with allatives, as shall be clear in the following exposition.

4.1.2 Goal markers

The goal markers (allatives) in Thai, most of which denote ‘to’ or ‘for’ (if benefaction is implied), are among the grammatical forms that occur at high frequency. The extensive list is given in (5) below, and some uses of such allatives are exemplified in (3) for monolexemic and (4) polylexemic forms:

The allatives in the examples above have their transparent lexical origins in contemporary Thai: hây from ‘give’, phʉ̂a from ‘do for’, thʉ̌ŋ from ‘arrive’, and thîi from ‘place’. The conceptual motivations behind these grammaticalization scenarios are relatively obvious, i.e., transfer of an object from ‘give’, transfer of benefit from ‘do for’, and reaching the goal from ‘arrive’. The relatively ambiguous one is from ‘place’ to ‘to’. As was the case with some ablatives illustrated in (1), this is due to the contextual effect, i.e., the cooccurring verb ‘send’. This type of reinterpretation is a common phenomenon in Thai, as elaborated in Kölver (Reference Kölver1984, p. 14), who states that a cooccurring motion verb (e.g., ‘go’) transforms a static meaning ‘at’ into a dynamic, directional meaning ‘into’.

The next set of examples illustrates four polylexemic forms of a large number of polylexemic allatives (and, more generally, polylexemic prepositions (see Rhee Reference Rhee2021, pp. 307–360)) in Thai.

The examples in (4) all involve allatives consisting of two lexemes: hai kae ‘give mature’, pay thîi ‘go place’, pay yaŋ ‘go result in’, and maa yaŋ ‘come result in’. In (4a) and (4c), ‘mature’ and ‘result in’ seem to add the notion of culmination of a process, i.e., arrival at the goal. Notably, these lexemes constitute the second of the two compound lexemes, and the first lexemes (hây and pay) are well established allatives. Thus, the morphosyntactic configuration, in which the second lexeme occurs between an allative and a noun phrase, must have contributed to the semantic bleaching of these lexemes and their attraction to the preceding allative, e.g., [to-mature me] reanalyzed as [to me (completely reaching me)]. The development of the allative pay thîi ‘to’ from ‘go place’ is also likely to be the reanalysis and reinterpretation of ‘go’ as a directional and ‘place’ as the focused ground, which aligns with the observations in Post (Reference Post2007, p. 121), who points out the tendency of pay ‘go’ becoming a ‘pure functor’ in certain configurations. The same applies to maa ‘come’, another common motion verb in Thai, the only difference being the reversed directionality of motion.

The multiplicity of Thai ablatives illustrated in (1) and (2) above is far exceeded by the extensive inventory of allatives shown in Table 2 with their sources.

Table 2. Allatives in Thai

The comprehensive list of allatives and their lexical sources shown in Table 2 reveals the patterns observed with ablatives, specifically that allatives are predominantly V-adpositions, and even those involving nouns also tend to involve verbs, as in (b), (c), (x), and (y). Grammaticalization of allatives from source lexemes constitutes a meaningful research topic, but is beyond the immediate interest of the present research (Rhee and Khammee (Reference Rhee and Khammee2023) offer one such analysis). Also notable, as is the case with ablatives, is that polymorphemic forms are periphrases consisting of components that are mostly allatives themselves, a state of affairs carrying important implications, as shall be discussed below.

4.2.1 Syncretic forms

One peculiarity of Korean is that a number of forms show allative-ablative (-locative) syncretism, a cross-linguistically rare phenomenon (cf., however, −ni ‘to/from’ in Japanese, Peter Sells, p.c.; Rhee, Reference Rhee2010, and ‘motatives’ -şa, −le, −de in Ardeşen-Laz, Kutscher Reference Kutscher2010). These forms are listed below:

The polyfunctional postpositions -ey and -kkey are exemplified in the following constructed examples:

In example (6a), the postposition -ey occurs three times: as an allative (‘to’), locative (‘at’), and ablative (‘from’), respectively. The postposition -kkey in (6b) functions as both allative/dative and ablative. A noteworthy aspect of these examples (and other allative-ablative syncretic forms) is that the source encoded by them involves a noun phrase denoting an entity as a source of force, and the movement denoted or implied by the event tends to be abstract rather than physical. For instance, the ‘violent wind’ in (6a) is not a departure point in a strict sense and the ‘scolding’ in (6b) does not involve physical movement of ‘scolding words’ to the speaker. The fact that allative-turned-ablatives do not fully encode a spatial departure point suggests that the functional extension of allatives into the domain of ablatives has not (yet) progressed to a great extent.

4.2.2 Source markers

Contemporary Korean has many ablative markers, including those of syncretism. The most frequently used forms are -ey and -eyse. It is noteworthy that -ey occurs at a high frequency both as an ablative and allative, a point we will return to later. The uses of ablatives -eyse and -pwuthe are exemplified below:

Historically, the ablative -eyse ‘from’ is a coalesced form of -ey ‘at’, is- ‘exist’, and -e ‘and’, compositionally denoting ‘exist at x and’, and the ablative -pwuthe ‘from’ is from puth- ‘adhere’ and -e ‘and’ (note that -e and its variant -a are linkers glossed as connectives [cn] in the list). It is clear that ‘existing’ at a place and ‘adhering’ to an entity have contributed to the development of the notion of contact with the referenced ground. For instance, ‘exist at Thailand and’ becomes ‘from Thailand’ and ‘adhere to the house and’ becomes ‘from house’ (cf. Ahn, Reference Ahn1967; Park, Reference Park2015; Rhee, Reference Rhee2021; Ryu, Reference Ryu1962). Ablatives and their lexical sources are listed below:Footnote ⁵

The list of Korean ablatives above shows that some of them (such as -ey and -lo) involve unidentified lexical sources, while others grammaticalized mostly from constructions involving a verb, notably the verb of existence, is- ‘exist’. Some ablatives also involve nouns denoting ‘place’. As was noted with the Thai ablatives and allatives, Korean polymorphemic ablatives involve forms that are ablatives themselves. Of further importance is that many Korean allatives involve the morpheme -e at the final slot, which is a connective, largely denoting ‘and’. This is significant because the notion of departure point, the core semantic element of ablative, is largely derivable from this connective -e. For instance, as for -pwuthe in (8f), literally denoting ‘adhere to x and’, can be illustrated with thaykwuk-pwuthe [Thailand-abl] ‘from Thailand’ in (7), which developed from ‘adhere to Thailand and’. The verbal semantics of ‘adhere’ simply designate the entity to which the sentential subject adheres, and the notion of departure is signaled by the connective -e ‘and’.

4.2.3 Goal markers

Contemporary Korean also has a large inventory of allatives, among the most frequently used of which are -ey, −ulo, −kkaci, and -hanthey. Some are exemplified below:

Two allatives illustrated above, −ey in (9a) and -ulo in (9b), are incidentally syncretic forms with ablatives. As has been previously noted, the lexical sources of these allatives have not yet been identified (see, however, footnote 5), which suggests that they grammaticalized long ago and their lexical meanings have become completely lost. The allative -eytaka is a composite form, consisting of the allative -ey, taku- ‘approach’, and -a ‘and’. Thus, swul-eytaka ‘to wine’ is from the source construction of ‘approach wine and’. Similarly, the allative -hantheytaka is also from a complex construction, i.e., han ‘one’, tey ‘place’, taku- ‘approach’, and -a ‘and’, suggesting in (9d) a change from ‘approaching to the same place as the child is and’ to ‘to the child’. The allative -tele is relatively simpler in composition, i.e., tAli- ‘accompany’ and -e ‘and’, thus ‘accompany the older sister and’ has become ‘to the older sister’. For grammaticalization analysis of some of these allatives, see Heo (Reference Heo2007), Kim (Reference Kim1982), Rhee (Reference Rhee2010), Ryu (Reference Ryu1962), among others. The allatives in contemporary Korean and their lexical sources are listed in Table 3.

Table 3. Allatives in Korean

The extensive list of allatives presented here shows many notable aspects of grammaticalization of allatives in Korean. A few lexical items, e.g., the verbs ‘approach’, ‘touch’, and ‘face’, and the nouns ‘edge’ and ‘place’, recur in the creation of polymorphemic forms, a state of affairs similar to the developmental scenarios of ablatives. Compared with ablatives, allatives are particularly numerous, a point we will return to later.

4.3.1 Conceptual granularity

In the description of the source and goal markers above, the labels ‘ablatives’ and ‘allatives’ have been used. However, source and goal across languages are not unitary notions but can be hypernyms comprising a number of hyponymic concepts. For instance, source may consist of the ablative ‘from’ and elative/separative ‘out of, off’, and goal may consist of allative ‘to, onto’, directional ‘toward’, lative/illative ‘into’, terminative ‘up to, until’, dative ‘to (a person)’, and translative ‘(changing) into’, and so forth It is evident that in relative terms source is a conceptually (nearly) uniform category denoting the basic notion ‘departure’, whereas goal exhibits, in addition to the core semantics of ‘arrival’, more fine-grained specialization, modulated by various semantic features such as the directionality of the movement (e.g., whether or not the trajectory involves the orientation of the trajector), the nature of the target’s affectedness by the trajectory (e.g., whether the trajector lands on the target or penetrates it), scalarity of the movement (e.g., whether the trajectory involves movement along a gradient continuum until it reaches termination), humanness or animacy of the target (e.g., whether the entity is a person or an animate being or otherwise), contrastive property of the resultant movement, real or imagined (e.g., whether the end-point is contrasted with the starting point, thus highlighting the change), among other things. In other words, goal is conceptually more fine-grained than source, and this difference is reflected in adpositions in both Thai and Korean.

Apart from the conceptual granularity of the two polar concepts source and goal, Thai and Korean generally make fine distinctions between the semantic features of the target in goal, whereas in source they do not, or they do so more loosely. For instance, Thai source prepositions distinguish between temporal and nontemporal departure points (e.g., the ablative tâŋtɛ̀ɛ is specialized in time as in ‘from 5 o’clock’ and cannot mark a strictly spatial departure point as in *tâŋtɛ̀ɛ Bangkok ‘from Bangkok’). In Korean, on the other hand, the distinction between temporal and nontemporal is not prominent in ablatives. The distinction made in ablative specialization is largely between human (marked by -eykey, −kkey, −kkeyse, −hanthey, or -hantheyse) and nonhuman, and between honorific (−kkey or -kkeyse) and nonhonorific. Differences among ablatives, both in Thai and Korean, include the general tendency to multiply strung forms that are more emphatic in meaning than the monomorphemic forms.

In contrast to these rather simple specializations of ablatives in Thai and Korean, allatives exhibit much more elaborate distinctions carrying finer shades of meaning. First of all, as was the case with ablatives, multiply strung forms in both languages tend to be emphatic in meaning. Furthermore, certain allatives possess additional semantic features at varying degrees as a result of carry-over from the source semantics (an instance of ‘persistence’ in grammaticalization, Hopper, Reference Hopper, Traugott and Hopper1991). For instance, the allatives related to the verb hây ‘give’ (e.g., hây, hây kɛ̀ɛ, hây kàp, etc.) tend to imply benefaction. In Korean, allatives involving the verbs ‘approach’ and ‘touch’ (e.g., −eyta, −eytaka, −eytatayko, etc.) are largely pejorative in addition to being emphatic, a situation related to the fact that those verbs tend to suggest ‘recklessness’ as part of their semantic nuances, which is also an occasion of ‘persistence’. A more in-depth exploration of the complementation pattern of the allatives in Thai reveals a much more elaborate specialization with respect to the semantic features of the landmark NP and speech register, e.g., living (nonhuman and nonanimal), animacy (nonhuman), human, event, time, honorific, formal, emphatic, as shown in Table 4. A caveat is that the positive or negative value indicates that the form usually can or cannot be used for an entity with that semantic property, and that the pattern is a strong tendency but not an inviolable rule.

Table 4. Specialization of Thai allatives

As for Korean allatives, specialization involves similar semantic features, i.e., living (nonhuman and nonanimal), animate (nonhuman), human, honorific, pejorative, terminative, directional, formal, and emphatic. Their specialization pattern is shown in Table 5.

Table 5. Specialization of Korean allatives

Table 5 shows the elaborate system of Korean allatives comparable to those of Thai. The compatibility of individual allatives with the semantic features of the entity is of variable degree, often influenced by idiolectal idiosyncrasy or genre, but the overall organization clearly shows a high level of conceptual granularity and division of labor among the members of the paradigm. For this reason, allatives are more numerous than ablatives, as shown by the contrasts of 29 versus 6 in Thai and 36 versus 13 in Korean. This leads to a natural conclusion that there exists a plainly evident asymmetry between ablatives and allatives; in other words, our cognition is biased toward allatives. This conclusion is in line with findings from crosslinguistic studies by Bourdin (Reference Bourdin, Palek, Fujimura and Neustupný1997) and Zanchi (Reference Zanchi, Luraghi, Nikitina and Zanchi2017).

4.3.2 Pragmatic salience

In addition to the differences in conceptual granularity, ablatives and allatives in Thai and Korean also exhibit a significant difference at the level of pragmatic salience. In other words, allatives surface more frequently than ablatives, confirming the pattern that has been observed in a number of languages. The asymmetric salience in Korean and Thai is confirmed by the fact that the aggregate token frequency of allatives in the referenced corpus exceeds that of ablatives. The token frequency per million words (pmw) of individual ablatives and allatives in Thai is shown in Tables 6 and 7, respectively. Note that, as indicated earlier, many of these ablatives and allatives in Thai have homophonous lexical items (e.g., càak p. ‘from’ and v. ‘leave, depart’; hây p. ‘to’ and v. ‘give’). The numbers in the tables are only for those in adpositional uses, despite the distinction rarely being straightforward. Large numbers in the tables are proportional projections from random sampling, i.e., using proportions of adpositions in the random samples of 200 from the total hits for estimation of the total number of adpositions.

Table 6. Token frequency (pmw) of Thai ablatives

Table 7. Token frequency (pmw) of Thai allatives

A comparison of Tables 6 and 7 clearly shows the GOS asymmetry as follows: allatives are 3.3 times as frequent as ablatives (18,703 versus 5,614). Also notable is that one form in each category, i.e., càak for ablative and hây for allative, is used at a particularly high frequency compared to other members in the category, accounting for 85.6 percent of ablative, and 35.5 percent of allative. There is only one ablative with a token frequency of over 1,000 pmw, whereas there are five allatives occurring at such frequency. This usage pattern suggests that ablative is ‘predominantly’ marked by càak, whereas the conceptual space of allative is divided into multiple subareas, each of which is indicated by different allative markers.

Korean shows a similar pattern. The token frequency (pmw) of the ablatives and allatives in the corpus is shown in Tables 8 and 9, respectively. It is to be noted that the frequency marked as ‘0’ does not necessarily mean ‘no attestation’; it simply indicates that the frequency is below 1 token per million words. In the case of syncretic forms, the frequency count was separated by each function of ablative and allative. Large numbers are proportional projections from random sampling, as was the case with Thai.

Table 8. Token frequency (pmw) of Korean ablatives

Table 9. Token frequency (pmw) of Korean allatives

A comparison of the ablatives and allatives shown in Tables 8 and 9 reveals a state of affairs comparable to Thai. Allatives occur three times as frequently as ablatives (38,730 versus 12,907), and there are more high-frequency allatives than ablatives (defined here as those with a token frequency of 1,000 pmw or higher).

The GOS asymmetry is also observable with syncretic forms. We have already noted that Korean has six postpositions indicating allative-ablative (−Locative) syncretism. Four of these involve allative-ablative syncretism. Their token frequency for individual functions, the numbers of which are included in Tables 8 and 9, is summarized in Table 10.

Table 10. Token frequency (pmw) of Korean syncretic forms

As shown above, among the Korean syncretic forms, the proportion of use for allative marking far exceeds that for ablative marking. This is no surprise because historically these syncretic forms are allatives and their extension into the domain of ablatives is a rather recent development in history. How this extension is being actualized constitutes a meaningful topic of investigation but is beyond our immediate interest in the present research. The point here is that the usage pattern in contemporary Korean upholds the GOS asymmetry hypothesis.

In conclusion, an exploration into the conceptual granularity and pragmatic salience of ablatives and allatives in Thai and Korean strongly supports the GOS asymmetry reported in other languages, e.g., Japanese (Miyajima, Reference Miyajima and Miyaji1986), English (Koenig et al., Reference Koenig, Mauner and Bienvenue2003; Stefanowitsch & Rohde, Reference Stefanowitsch, Rohde, Radden and Panther2004), and multiple European languages (Verkerk, Reference Verkerk, Luraghi, Nikitina and Zanchi2017).

Incidentally, a comparison of the Thai and Korean statistics shown above presents an interesting issue which merits brief mention. Apart from the clear asymmetry between ablatives and allatives within each language, the use frequencies between the two languages are remarkably different: ablatives occur at 5,614 pmw in Thai versus 12,907 pmw in Korean, and allatives occur at 18,703 pmw in Thai versus 38,730 pmw in Korean. The overall frequency shows that Korean ablatives and allatives are more than twice as frequent as those in Thai. The reasons for the disparity between the two languages are not immediately clear. It can be suspected, however, that Thai, an isolating and analytic language, does not use grammatical markers in general as often as Korean, an agglutinating and synthetic language. A cursory look at the examples in the preceding exposition clearly shows that Thai sentences involve far fewer grammatical forms than Korean sentences. Therefore, while both languages nearly freely omit sentential arguments along with their nominal trappings including adpositions, the extent of use of grammatical markers is significantly different between the two languages. This echoes the characterization that Thai is a particularly ‘pragmatically oriented’ or ‘discourse-oriented’ language (Bamgroongraks, Reference Bamgroongraks1987; see also Huang, Reference Huang1994 for Chinese). A more conclusive answer to this puzzle, however, should await further research.

4.3.3 Innovation

The final topic for discussion involves innovation of linguistic forms. In both Korean and Thai, the multiplicity of forms in the two polar categories is largely due to the stacking of multiple markers of (near-)synonymous adpositions as a strategy to reinforce meaning or to add finer shades of meaning. For instance, the Thai polymorphemic allative con kràthâŋ thʉ̌ŋ ‘to’ involves both independent and combined allatives, i.e., con ‘to’, kràthâŋ thʉ̌ŋ ‘to’, and thʉ̌ŋ ‘to’. A look into the formal construction of Thai allatives, illustrated in (5) above, has made evident that such a pattern is a common strategy for creating polymorphemic adpositions. This type of doubling or tripling of forms in allatives is a strategy also used, albeit to a much lesser degree, in creating ablatives. In fact, this strategy, named the ‘polysemy strategy’ by Khammee and Rhee (Reference Khammee and Rhee2022) and Rhee (Reference Rhee2022), is commonly observed in lexicalization and grammaticalization in Thai. Similarly, the Korean polymorphemic allative -eytakataykose ‘to’ involves -ey, −eyta, −eytaka, and -eytatayko, all of which are full-fledged allatives in their own right. Naturally, the multiplicity of forms is due at least in part to frequent innovation of new forms by way of building up a new form based on an existing one. Thus, some discussion on the motivation for building up new forms, although speculative in nature, is in order.

The diachronic investigation of the Thai language suffers a regrettable paucity of historical documentation, while Korean, on the other hand, offers rich historical data, especially from Late Middle Korean (LMiK; 1443–1600). A brief historical survey of the forms used for allatives reveals the following chronology. In LMiK, a few forms were already used as full-fledged allatives, notably -ey and -eykey. During a later time of LMiK, −hanthey (from ‘one place’), −eyta (from ‘approach, draw near’), −tele (from ‘be accompanied by’), and -poko (from ‘see’) joined the paradigm of allatives. In the 19th century (Early Modern Korean; EMoK), a new form -eytatayko (from ‘approach’ and ‘touch’) is first attested, followed by numerous forms in the 20th century (Modern Korean; MoK), such as -hantheyta, −hantheytaka, −hantheytayko, etc., all of which involve ‘one place’ and ‘approach’ or ‘touch’. This development can be framed as ‘expressivity’ and ‘creativity’ as generally observed in languages.

Since language users are in constant search of meanings and patterns in language use, meanings are continually extended through reinterpretation, units are accordingly reanalyzed, and constructions are applied to novel structures for experiment and the desire for novelty or creativity (Rhee, Reference Rhee2021, Reference Rhee2023). Heine and Stolz (Reference Heine and Stolz2008: 332), echoing Croce (Reference Croce1912[1902: 172–174), stated that language is an essentially creative activity. Similarly, Lehmann (Reference Lehmann, Boretzky, Enninger and Stolz1987) asserted that the desire to be original, to say something that has not been uttered before, and to give one’s thoughts an imposing expression (creativity) are an essential part of linguistic activity (as cited in Heine & Stolz, Reference Heine and Stolz2008: 335). Recurrent innovations in language, leading to multiple layers within a single functional domain, can be accounted for by resorting to the language users’ desire for creative language use. Through creativity, language users innovate new forms despite the forms already existing for the same function. Innovated forms tend to involve more specific meanings, which the language innovator desires to exploit. Thus, certain innovative forms tend to carry the speaker’s strong emotional or evaluative stance.

If these notions of ‘expressivity’ and ‘creativity’ are applied to the Korean allatives, the developmental paths can be interpreted in the following way. In LMiK, there are already well-established allatives, and language users innovate -hanthey (from ‘one place’) to emphasize the co-presence of the trajector and the landmark at a place (i.e., ‘arrival’); −eyta (from ‘approach, draw near’) to highlight the trajectory extending toward the landmark (i.e., ‘directional movement’); −tele (from ‘be accompanied by’) to emphasize the togetherness of the trajector and the landmark (i.e., ‘arrival’); and -poko (from ‘see’) to highlight the trajector’s visual attention to the landmark (i.e., ‘directionality’). In the 19th and 20th centuries, forms involving ‘touch’ (in combination with ‘approach’) were innovated and popularized, all highlighting the trajector’s contact or movement intending to make a contact with the landmark. All these added components strengthen the general notion of allative. The novel semantics come from the semantics of the newly recruited lexical items. Evidently, the forms being innovated tend to carry more lexical content than the older, fully grammaticalized forms, and thus carry more expressive potential. As grammaticalized forms tend to be semantically bleached, recruiting forms with richer semantics may well be motivated by the desire for increased expressivity.

Since ablatives are considerably fewer in number than allatives, the innovation can be said to be more frequent with allatives, which, in turn, suggests that language users are more frequently driven to express goal more creatively and expressively, which is another form of the GOS asymmetry. This frequent innovation raises the paradox of the emergence of more complex forms at a higher rate running counter to the previous claim that goals are expressed ‘more simply’ (Stolz et al., Reference Stolz, Lestrade and Stolz2014). The resolution of this paradox does not seem to be readily available, but it may involve the more recently observed phenomenon of grammaticalization through ‘expansive’ motivation (Rhee, Reference Rhee2021, Reference Rhee2022) that brings forth formal complexity to otherwise reductive change in grammaticalization. A more conclusive answer should await further research.