Adjective Distribution in Mongolian and Japanese

This paper tackles how adjectives distribute in Mongolian and Japanese in light of the framework ‘scale structure’. It is explicated that Mongolian resultatives are of three types, i.e. adjective-post resultative; adjective-initial resultative; co-verb resultative. The acceptability of adjectival complements in inherent resultatives runs from ‘Totally open-scale AP’ down to ‘Lower closed-scale AP, Upper-closed scale AP, Totally closed-scale AP’. Mongolian welcomes all layers of adjectives in direct perception expressions. Japanese adjectives are re-categorised into two types, i.e. open-scale adjectives (corresponding to the traditional i-adjective) and closed-scale adjectives (corresponding to the traditional na-adjective). Though both are capable of rendering an inherent result, the resultatives rendered by open-scale adjectives and closed-scale adjectives present different lexicalisation patterns. Moreover, like Mongolian, derived resultatives are also missing in Japanese. The lack of derived resultatives in Altaic languages is down to the following reason, i.e. Altaic languages are likely to lexicalise the RESULT into the MAIN VERB. In inherent resultatives, there is a connection between CAUSE EVENT and RESULT EVENT, i.e. the verb carries an implication of result. In derived resultatives, a GAP arises between ACTION and RESULT. Regarding direct perception expressions, Japanese seems to welcome both open-scale and closed-scale APs. The perceptual verb 見る miru solely delivers the perceived event and is thus deemed objective. 見 え る mieru, on the other hand, cannot fulfil a metaphorical interpretation, and appears to be subjective.


Introduction
Mongolian, an Altaic language, is deemed an exclusively suffixing agglutinative SOP language. Seven cases are mostly used: nominative, accusative, genitive, dative-locative, ablative, instrumental, and comitative. A salient feature of Modern Mongolian lies in the fact that adjectival complements may directly precede the verbs, as seen in the resultative construction (1).

c. Co-verb type [NP 1 V V NP 2 ]
Тэр эрэгтэй үсээнбййрэн бичиж байгаад хугалсан. DEM 1 st .masc.sg pen write-break-PAST 'He writes with the pen (until it is) broken.' In Mongolian resultatives, one issue that is particularly worthy of discussion is that not all adjectives seem capable of indicating a RESULT. In (1b), the adjective улаан ('red') is tolerated; while the following adjective, хатуу ('solid') is ruled out.
(ill-formed) lake solid freeze-PAST 'The lake froze solid.' The different treatments of (1) and (3) are probably down to the resultative construction type as well as the scalar property of adjectives (i.e. adjective хатуу 'solid' is a closed-scale AP whilst улаан 'red' is open-scale).
Moving on to Japanese, another deemed Altaic language, RESULT can be rendered via three grammatical elements: (i) a PP (4); (ii) a verb compound, where the change of state is potentially conflated into the main verb (5); and (iii) an AP (6): (4 This study intends to explore the distribution of APs; therefore, only constructions like (6) will be tackled.
Regarding Japanese adjectives, traditional linguists consider them as falling into two groups, i.e. i-adjectives (7) and na-adjectives (8 1 and Japanese licenses open-scale APs, as in (9b): a. Open-scale APs in English resultatives: *Taroo stretched the rubber long.
It appears that the two languages have the undeniable similarities in favouring open-scale adjective as resultative complements. This is a preliminary illustration that inspires us to investigate adjective distribution in more depth.
This paper is mapped out as follows. Section 2 sheds light on the framework 'lexical conceptual structure' and 'scale structure'. Section 3 categorises Mongolian resultative construction types, then examines the acceptability of adjectival complement. Next, it tackles the distribution of adjectives in direct perception expressions. Section 4 delves into the question of how adjectives are treated in Japanese.
1 Example (12b) is taken from Tsujimura (2001). Section 5 highlights the results and concludes the paper.

Framework
To set the stage for resultative and direct perception construction, we sketch an overview of previous studies on resultatives. The discussions referred to below focus on resultative constructions, which have long been an important issue in linguistic typological work.
For the past half century, various frameworks have been proposed to achieve a thorough analysis on resultatives. In earlier times, analysis focuses upon the syntactic perspective, representative work includes Chomsky's (1965) 'Aspects Model', Levin and Rappaport Hovav's (1995) 'Projection Approach'. However, this approach appears unable to explain why the unergative verb laugh can appear in both 'Mary laughed herself sick' and 'The audience laughed the actors off the stage'. Therefore, later on, a different view, i.e. 'Construction Grammar Approach' was provided by Goldberg (1995). This pioneering attempt has had a significant influence on the study of resultatives and is welcomed among linguists, e.g. Jackendoff It is pointed out that in inherent resultatives; the result of the theme is implied by the main verbs. For instance, 'wipe' may possibly give rise to the result of 'clean'.
In fact, Kageyama was not the first to classify the type of resultative construction. Many scholars have made similar proposals. For instance, Kageyama (1996)  The data for Japanese is drawn from the corpus of Balanced Corpus of Modern Written Japanese by National Institute for Japanese language and linguistics.

Adjective Distribution in Mongolian
This section begins by looking into the scalar property of Mongolian APs, and their distributions in resultative and direct perception expressions.

Mongolian Resultative Constructions
As touched upon earlier, Mongolian resultatives are of three types. (i) the adjective-post type [NP 1 V NP 2 AP]; (ii) the adjective-initial type [NP 1 AP V NP 2 ]; and (iii) the co-verb type [NP 1 V V NP 2 ]. Before looking into how adjectives distribute in these types, it would be necessary to shed light on the syntactic features of the types.
First, adjective-post resultatives display the following variations: transitive resultatives [NP 1 V t NP 2 AP], as in (13); and intransitive resultatives [NP 1 V i NP 2 AP], as in (14). (13) Note that in intransitive resultatives, when an unaccusative verb denotes the action verb, the CHANGE-OF-STATE EVENT becomes spontaneous, which leads to an anticausativisation phenomenon. 2 When an unergative verb renders the action verb, the CHANGE-OF-STATE EVENT becomes intentional, giving rise to decausativisation.
We now turn to adjective-post resultatives［NP 1 V NP 2 AP], as illustrated in (15). Six Mongolians from different areas provided the judgement. 2 The terms anticausativisation and decausativisation were initially coined by Japanese linguist Kageyama Taro (1996). He defines the them as follows: a. Anticausativisation The CHANGE-OF-STATE is brought about by the patient itself, therefore the agent and the patient can be identified. b. Decausativisation The CHANGE-OF-STATE is brought about by external facts. Syntactically, the agent is suppressed and thus is not projected onto the syntactic structure.

152
Adjective Distribution in Mongolian and Japanese c. Тэр эрэгтэй лъ сийг сажилажъ сэргээб. DEM  (15a) and (15c) are judged natural, whereas (15b) and (15d) are deemed to be unnatural. If, however, the adverbial, i.e. жинхэнэ ('very'), is supplied to the adjectives хабтагай ('flat') and asчийг ('wet'), the odd expressions (15b) and (15d) can be improved: (16). a. Тэр эрэгтэй төмөр таяг дабдажъ жинхэнэ хабтагай болгаб. (natural) b. Тэр эрэгтэй өбөст газар усулжъ жинхэнэ чийг болгаб. (natural) Why, then, are the above adjective-post resultatives treated differently? We analysed the 'lexical conceptual structure' and found that each construction displays the following distinction (17 represents 15a and 15c; 18 represents 15b and 15d): (17). LCS of (15a) and (15c) [x act -on y ] cause [y become [be-at z]] (18). LCS of (15b) and (15d) The CAUSE EVENT of (15a) and (15c) may directly result in the RESULT EVENT. Moreover, the CAUSE and the RESULT are associated (i.e. CONTROL). (15a) can be described as follows: On the other hand, the CAUSE and the RESULT in (15b) and (15d) are not associated. To put it another way, there is a gap between the CAUSE and the RESULT. Intriguingly, by supplementing the adverb жинхэнэ ('very') to the ill-formed expressions (15b) and (15d), the oddness of the phrasing can be reduced. This is because the extra adverb жинхэнэ ('very') temporarily fills the GAP between the CAUSE and the RESULT at a syntactic level; and because the syntactic supplement semantically draws the connection between the CAUSE EVENT and the RESULT EVENT. The process is described in (19).
One way to solve the problem is to employ an adverbial, and thus temporarily fill the gap. For LCS, this manipulation would be: The failure of anticausativisation in adjective-post resultatives lies in the fact that the ACTION verbs (дабдажъ 'pound' (28a); сажилажъ, 'shake' (28b)) denote MANNER and imply a strong agency. The patient cannot therefore be identified by the agent.
Moving on to adjective-initial Resultatives [NP 1 AP V t NP 2 ], these can be understood as 'x ACT ON y, until y BECOME z', where the adjective behaves like an adverbial. 3 This type of construction is described in (29) The process of anticausativisation of 'Үсээ хар өнгөөр будсан' would be: (34).
Regarding co-verb resultatives, V1 indicates the ACTION; V2 conveys the RESULT. V1 can be an unergative verb, unaccusative verb, or transitive verb; V2 can be an unaccusative verb or an unergative verb. The argument structure has the following variations: As touched on earlier, one of the significant roles played by co-verb resultatives is to render derived resultatives, as shown in (36). It may also convey inherent resultatives, as in (37). So far, three types of Mongolian resultatives, (a) adjective-post types, (b) adjective-initial types, and (c) co-verb types, have been demonstrated in the light of the 'lexical conceptual structure'. It has been observed that decausativisation is welcome in adjective-post types, and that anticausativisation is realised in adjective-initial resultatives. The significant role of the co-verb type is to fill the gap of ACTION and RESULT in derived resultatives. 4 A typological issue is how to identify grammatical category cross-linguistically. Croft (1990) names this the problem of cross-linguistic comparability. He argues that semantic and functional criteria are the best solution to this problem. байгаад in (55a) and улаанаар in (55b) are unaccusative verbs.

Adjective Distribution in Mongolian Resultatives
Having drawn a picture of Mongolian resultatives, this section delves into adjective distribution. The discussion will focus on resultatives and direct perception expressions.
As a starting point, we drew a list of the most-used adjectives in Mongolian from the corpus: The concept of 'scale structure' is adopted to measure their scalar property. The modifiers жинхэнэ ('very') and хагас ('half') helped with the diagnosis. Three Mongolian native speakers provided the judgements. The results are summarized in (41)

Adjective Distribution in Direct Perception Expressions
Now we are in the position of engaging in the analysis of adjectives in direct perception expressions. Tests along with the various scalar properties of Mongolian adjectives were carried out and again data were assessed by the native speakers. The finding brings us to the point that Mongolian welcomes all layers of adjectival complements in direct perception expressions.

Summary
To summarise, Mongolian resultatives are of three types: (i) adjective-post resultative [NP 1 V NP 2 AP], (ii) adjective-initial resultative [NP 1 AP V NP 2 ], and (iii) co-verb resultative [NP 1 V V NP 2 ]. Adjective-initial resultatives can be understood as 'x ACT ON y, until y BECOME z', where the adjective behaves like an adverbial. Decausativisation is tolerated by adjective-post resultatives, while anticausativisation is ruled out. Anticausativisation can be realised in adjective-initial resultatives. The

Adjective Distribution in Japanese
Having drawn a picture of the sensitivity of APs in Mongolian, we are in a better position to engage in the analysis of Japanese data. Our starting point is the scalar property of adjectives.

Adjectives in Resultatives
Traditionally, Japanese adjectives are divided into two  Table 2. Another important issue to be addressed is that the above constructions are all inherent resultatives. Derived resultatives are missing in Japanese; neither an adjective nor a postpositional phrase is possible to render a derived result, as in (53)

Adjectives in Direct Perception Expression
Two perceptual verbs are often seen in Japanese direct perception expressions, i.e. 見える mieru and 見るmiru. mieru is an unaccusative verb and means that a certain view leaps to the eyes. The subject is often denoted by a scene or an inanimate lexicon. In fact, the subject of the unaccusative verb mieru is actually the object of the transitive verb 見るmiru. The subject of miru is the observer, often rendered by animate lexicons. Crucially, miru delivers a pure perceived event and is therefore incapable to fulfil a metaphorical interpretation. The perception verb mieru indicates an evaluation of the scene, i.e. the sub clause 'Hanako ga hosoi (Hanako is slim)'. Note that the proposition can be false. After all, it is an evaluation by the observer, not a truth. In (56), it can have the following interpretation: Hanako is actually fat, but from Taro's point of view, Hanako is slim. In this regard, mieru is somehow subjective. Its transitive partner, miru ('to see'), fails to appear in this expression, as miru simply renders a true scene.
Not only open-scale APs are licensed: closed-scale APs are also welcomed in Japanese direct perception expressions, as illustrated in (57) Bear in mind that ni in (57) functions as a copular. The AP-complement and the copula ni together form the perceptual complement. This is exactly the same as ni in resultative constructions.

Summary
This paper has explored the adjective distribution of two Altaic languages, Mongolian and Japanese. The findings can be summarised as follows.
Mongolian resultatives are of three types, i.e. adjective-post resultative ; adjective-initial resultative; co-verb resultative. Among them, decausativisation is welcome by adjective-post type; anticausativisation is realised in adjective-initial resultatives. The co-verb type plays the part of rendering derived resultatives. This is summarised in Table 3. Regarding direct perception expressions, it seems that Japanese welcomes both open-scale and closed-scale APs. Moreover, there are two perceptual verbs, 見るmiru (transitive) and 見える mieru (unaccusative). 見るmiru solely delivers the perceived event and is thus deemed objective. 見える mieru, on the other hand, cannot fulfil a metaphorical interpretation, and appears to be subjective.
Finally, the lack of derived resultatives in Altaic language family is explained by the fact that Altaic languages are likely to lexicalise the RESULT into the MAIN VERB. In inherent resultatives, there is a connection between the CAUSE EVENT [EVENT 1 x ACT ON y] and the RESULT EVENT [EVENT 2 y BECOME Z]. To put it another way, the CONTROL part of LCS carries an implication of z. In derived resultatives, however, the ACTION fails to lead directly to the RESULT; a GAP thus arises between the ACTION and the RESULT.