Towards a Morph-syntactic Typology of Split Intransitivity

This paper analyses the split intransitivity by introducing data from Japanese and Mongolian. The finding reveals that Japanese split intransitivity links to postposition selection, i.e. unergative motion verbs describe processes with a durative motion event and thus are likely to yield directional postpositions or a route with an endpoint. Unaccusative verbs, on the other hand, indicate a punctual motion event and therefore often occur with locative postpositions. Intransitives further split in lexicalisation, i.e. Japanese unergative verbs tend to convey the MANNER of motion while unaccusative verbs appear to favour the PATH of motion. Mongolian seems to render the path in the main verb, leaving manner to be encoded in an optional constituent, i.e. a converbal construction. The combination of converbal construction is restricted to [non-scale change morphemes + totally closed-scale change morphemes] and [non-scale change morphemes + lower closed-scale morphemes]. Essentially, unergative verbs can be non-scale change morpheme or totally open-scale morpheme, contributing to the manner of motion. Unaccusative verbs can be totally closed-scale, or upper closed-scale, or lower closed-scale morpheme, denoting the path of motion.


Introduction
Split intransitivity refers to a phenomenon that different intransitive verbs in certain language may display different linguistic behaviours. Intransitives could split in syntax or semantics.
Following the Unaccusativity Hypothesis (Perlmutter [22]), intransitive verbs can be subcategorised into two types: unaccusatives and unergatives. An unaccusative verb has a syntactic external argument that is not the semantic agent. Semantically, the subject of an unaccusative verb is similar to the direct object of a transitive verb. For instance, fall, rise. An unergative verb semantically has an agent argument, such as run, walk.
The syntactic distinctions of the two lies in that, unaccusatives have an internal argument (1a) whilst unergatives have an external argument (1b).
There is another line of research that bears significance to the study of lexicalisation, arguing in favour of constructional factors. Remarkable works in this field are Arad [1], Borer [5], van Hout [26,27], McClure [21]. The constructional approach considers the unaccusativity and unergativity of verbs are not solely lexical properties, but are derived from the syntactic configurations of where verbs appear or how they are combined. The challenge remains, however, cross-linguistically, some verbs displays inconsistent behaviours, e.g. 'blush' is unaccusative in Italian whilst unergative in Dutch. Intra-linguistically, continuare 'continue' can take both auxiliary essere/E 'be' and auxiliary avere/A 'have' in Italian (Legendre [16]). It appears that split intransitivity within or across languages is far from being a clear-cut case.
Anticausativisation: The object changes in terms of the property of the object itself. The transitive verb comes to bear an intransitive function via identifying the object and causer.
e.g. waru/wareru ; yaburu/yabureru Decausativisation: The object changes in terms of external factors. The intransitivisation affix '-ar-' suppresses the causer in semantic structure, without projection to the syntactical structure. Through this manipulation, transitive verbs come to have an intransitive function.
x CONTROL [ y BECOME [ y BE AT -z] ] ↓ Ø e.g. kimeru / kimaru; atumaru / atumeru (Based upon Kageyama [9]) Split intransitivity in Japanese further extends to motion construction. It occurs that unergative verbs and unaccusative verbs show a very different preference: In (7a), the verb iku (past tense is 'itta') is unaccusative. It renders the path of the motion: go to the station. In (7b), aruku denotes the manner of 'walk'; the manner verb aruku alone cannot render a motion expression. Unergative verbs tend to yield the MANNER of motion while unaccusative verbs tend to convey the PATH of motion.
Turning to Mongolian, unaccusative and unergatives verbs are illustrated in (8) and (9) 2 : (8) Mongolian unaccusative verbs хагар ах 'br ea k i n t r . ' тас ара х 'cut i n t r . ' эвд эр эх 'br ea k i n t r . ' шат ах 'b ur n i n t r . ' хү р эх 'arrive ' The equivalent constructions of (5) are given in (10): The manner of motion/action in Mongolian is usually realised via a converbal construction: (11) a. с эмэр хэ н ойртох 'crawl towards' b. avaad gargax 'take out' In (11), the second constituent, which renders the PATH, is the core of the motion/action expression. The manner/means is conveyed by a converb (the first constituent).
The above are preliminary illustrations that inspire us to investigate the phenomenon of Altaic split intransitivity in more depth. The analysis will follow the 'scale structure' framework to account for the similarities as well as distinctions of the two languages. The data for Modern Japanese is from the corpus of Balanced Corpus of Modern Written Japanese by National Institute for Japanese language and linguistics. The data for Modern Mongolian use 'hand-made examples'. Two Mongolian native speakers from different parts of Mongolia judge all data.
The paper is mapped out as follows: Section 2 introduces the framework that is adopted in this study: scale structure. Section 3 discusses the semantic representations in terms of motion event, delving into the split regarding postposition selections. Section 4 turns to Mongolian, exploring the split regarding motion expression. Section 5 draws some general conclusions.

Framework: Scalar Structure
This paper takes the scalar structure theory as the framework. The adoption of scale in linguistic track can be traced back to 70 years ago (Sapir [24]). Later, it was adopted in formal semantics (Bolinger [4]). It was not until 1990s that scale was started being introduced to syntax (Hay et al. [8]). Recently, lexical semantians apply it to adjectives, verbs, etc. (e.g. Kennedy and McNally [14]; Kennedy and Levin [13]). A scale is constituted of a set of degrees on a particular dimension (e.g. cost, depth, height, temperature), with an ordering relation. The dimension represents an attribute of an entity, with the degrees indicating the possible values of this attribute (Kennedy and McNally [14]).

Split Intransitivity Regarding Motion Event
With the framework of scale structure highlighted, this section proceeds by posing the questions of how motion constructions might reflect Japanese split intransitivity.
To begin with, motion verbs in Japanese can be divided into five groups based upon their scale strcture:  (12) The distribution of unergatives and unaccusatives is given in (13): Totally closed-scale morphemes as well as upper closed-scale morpheme have specific goals. Therefore they can take ni 'to', e 'toward' case particles. On the other hand, non-scalar change morpheme, totally open-scale morphemes and lower closed-scale morpheme do not have inherent endpoints. When expressing a motion, they would have to employ a boundary marker made 'until' to indicate the endpoint of motion, as in (14): (14) Taroo wa eki made aruita. Taroo TOP station till walk PAST 'Taroo went to the station.' The addition of the goal phrase makes the aspectual properties of those morphemes available for an endpoint reading (Aske [2]; Beavers et al. [3]). In fact, this applies to change-of-state events. The goal phrase functions as an accomplishment, indicating the result of an action, e.g. (15).

Split Intransitivity Regarding the Postpositions
The split further links the selection of postpositions (PPs). Below is a list of Japanese postpositions. Their scale property is given in (16). Japanese unergative verbs fail to combine with the PP ni. This is due to the following reasons: ni is locative. Unergative verbs convey the manner of motion and therefore is unable to predicate a result location with the postpostion ni. ni may perfectly well co-occur with an unaccusative verb, because an unaccusative verb denotes the path, therefore the PP can bear a directional interpretation.
The ungrammaticality of (17) can be improved by replacing ni with e, which itself is a totally open-scale 3 e-to is between direction e and delimitation made.
postposition. With this e in place, the motion expression changes the aspect from telic to atelic, c.f. (18): (18) Taroo ga kooen e aruita. Taroo NOM park toward walk PAST 'Taroo walked to the park.' (18) can also be improved by replacing ni with an allative case marker made 'until', which is delimited and denotes an endpoint of motion. The scope of made includes the route and endpoint. On the other hand, ni denotes a scope limited to the goal, as furnished by its appearance in locational motion events. Consequently, ni tends to favour path verbs solely, such as tsuku 'arrive' and chakuriku suru 'land' (e.g. eki ni *aruita/tsuita), made is likely to occur with unergative verbs, such as aruku 'walk' as well as unaccusative verbs.
Furthermore, e-to, an upper-closed postposition, also appears compatible with unergative verbs, as in (19). Perhaps we can pause and draw a preliminary conclusion: unergative motion verbs such as aruku 'walk' and hashiru 'run' tend to describe processes with a durative motion event and are likely to yield directional PPs, e.g. e-to, e; or a route with an endpoint, e.g. made 'till'. On the other hand, unaccusative vers such as tsuku 'reach' or hairu 'enter' tend to indicate a punctual motion event and therefore often occur with locative PPs, e.g. ni. The foregoing discussion is summarised in Table 2:

Mongolian Case System Regarding Motion
Having drawn a picture of Japanese split intransitivity; we are in a better position to engage in the analysis of Mongolian data. Our starting point is the case selection in terms of motion event. Then, we will move on to the scale property of Mongolian motion verbs. The Mongolian case system regarding motion event is as follows: Nominative Case, Accusative Case, Genitive Case, Dative-Locative Case, Ablative Case, Instrumental Case, Illative Case and Comitative Case. Table 3 provides a list of case particles with regard to motion in Mongolian and Japanese. Table 3. Case particles in terms of motion in Mongolian and Japanese Case particle Japanese Mongolian Comitative と to -тай 3

Split Intransitivity Regarding Motion Expression in Mongolian
Mongolian case system distinguishes from Japanese case system in many respects, for instance, the manner of Mongolian motion verb уна takes an accusative case particle, e.g. -ыг/-ийг/-г, while the corresponding manner verb in Japanese, i.e. noru takes the dative particle ni. The path of Mongolian motion verb чиглэ takes an accusative case particle, e.g. -ыг/-ийг/-г; while the corresponding path verb in Japanese mukau takes the dative particle ni. Some path verbs, such as явах 'go', cалах 'leave' take the instrumental particle -аар 4 . The path verb явах 'go toward' takes an illative case particle, -руу 2 . Table 4 is a comparison of case selection in terms of motion in Japanese and Mongolian. Mongolian tends to lexicalises the path in the main verb, leaving manner to be encoded in an optional constituent, as in (24), where the core of the motion expression is rendered by verb гарах 'climb'; the manner of motion is expressed by a converbal construction. b. Accompanying Manner + Path сэмэрхэ н -ойртох 'crawl towards' бүд эр эн -унах 'tumble-fall' Essentially, the combination of converbal construction is not arbitrary, but restricted to the order of [Non-scale change morphemes + totally closed-scale change morphemes] or [Non-scale change morphemes + Lower closed-scale morphemes]. That is, the second constituent has to be a closed-scale change morpheme. This rule applies not only to motion event but also change-of-state event.

Summary
There are variations in split intransitivity even within the same language family, i.e. Japanese and Mongolian. This paper conducted a comparison of the two languages based upon the mereological framework: scale structure. The finding brings us to the point that intransitives split in lexicalisation of motion events as well as postposition selections. Mongolian tends to lexicalises the path in the main verb, leaving manner to be encoded in an optional constituent, i.e. a converbal construction. Japanese unergative verbs tend to yield the MANNER of motion while unaccusative verbs tend to convey the PATH of motion. Furthermore, unergative motion verbs tend to describe processes with a durative motion event and are likely to yield directional PPs. On the other hand, unaccusative verbs tend to indicate a punctual motion event and therefore often occur with locative PPs. This is a preliminary work that inspires us to explore the phenomenon of split intransitivity regarding Altaic language in more depth. thank the anonymous reviewers for their comments, which have helped me sharpen and develop the manuscript a great deal. All remaining errors and shortcomings are entirely mine.