LINGUIST List 24.4359

Mon Nov 04 2013

Review: Semantics; Syntax: Shibagaki (2013)

Editor for this issue: Joseph Salmons <jsalmonslinguistlist.org>



Date: 30-Aug-2013
From: Benjamin Brosig <benjaminling.su.se>
Subject: Analysing Secondary Predication in East Asian Languages
E-mail this message to a friend

Discuss this message

Book announced at http://linguistlist.org/issues/24/24-484.html

AUTHOR: Ryosuke ShibagakiTITLE: Analysing Secondary Predication in East Asian LanguagesPUBLISHER: Cambridge Scholars PublishingYEAR: 2013

REVIEWER: Benjamin Brosig, Stockholm University

SUMMARYIn “Analysing Secondary Predication in East Asian Languages”, RyosukeShibagaki compares the syntactic structure of resultatives and adjectivedepictives in the East and Central Asian languages Mandarin Chinese, Japanese,Khalkha Mongolian and Korean within the framework of Government Binding byapplying a number of syntactic and semantic tests to constructed sentencesjudged by a small number of informants.

For Mandarin (pp. 7-54), the main syntactic types under discussion areconsequence depictives like John chi[1]-ni[4] le man[2]tou J. eat-bored PFVbun ‘John ate the bun and became bored with doing so’ that exhibit a sequenceof events with a weak causal relationship and always refer to the subject, andcanonical resultatives like John da[3]-po[4] le bo[1]-li J. hit-broken PFVglasses ‘John hit the glass broken’ that are always object-oriented. These aretested with time adverbials and clefting, indicating properties of consequencedepictives somewhere in-between depictives and resultatives. This is followedby a critique of the previous literature about orientation patterns, andculminates into an analysis along patterns of causation and aspect such as[event [event x ACT ON-y] CAUS-ind [event BECOME [state x BE AT-y]]].

The discussion of Japanese (pp. 55-174) is about as long as the other threemain chapters combined. Both intransitive resultatives (there are notransitive subject-oriented resultatives in Japanese) and object-orientedresultatives can always be marked with -ni, while the additional suffix -kucan be applied to the latter only if the adjective is a term denoting colouror shape. Depictives are marked by -de. The test battery for resultativesconsists of:

(i) insertion of 10 pun-kan/-de ‘for/in ten minutes’ to detect telicity (onlytelic ...-de possible)(ii) paraphrasing with onaziyoo ‘in the same way’ (impossible, as it wouldindicate adverbials instead)(iii) VP nominalization (-ni can be replaced by -e for goal, but not forresultative phrases)(iv) being the target of wh-question, pseudo-clefting and do-so-replacement(e.g. the resultative must be subsumed under soo si-ta ‘did so’, i.e., it iswithin the VP) (v) insertion of overt notional subjects (impossible) (vi) combination of two resultatives (impossible) (vii) replacement with antonyms (impossible)(viii) morphological properties of lexical items as an additional heuristics.

Japanese resultatives are thus analyzed as part of the VP, namely, [VP [NPi][V’ [XP [[pro-i] [resultative predicate]] V]. Those canonical resultatives arethen contrasted to a number of predicates claimed to be resultatives in someof the previous literature, which can be excluded on the basis of the testsmentioned above. The section concludes with a discussion of lexical propertiesof resultative adjectives based on the previous literature. A mostly identicaltest battery is applied to depictives (non-telic, not paraphrasable byonaziyoo, coordinatable, not targetable by wh-questions, can, but needn’t beclefted (for subject-depictive), notional subject impossible, no particularsemantics required (cf. antonym test)), placing subject depictives under T’ orvP, while object depictives occupy a somewhat lower position. Discussionsabout fake depictives, the lexical properties of depictives, and aspect(stative vs. inchoative, lexical vs. syntactical) follow.

The sections on Khalkha Mongolian (pp. 175-224) and Korean (pp. 225-254)basically follow the pattern established in the chapter on Japanese, though nosources (of any kind) have been used for Mongolian, and the discussion ofprevious research on Korean is considerably less extensive than thediscussions of Mandarin and Japanese.

Mongolian resultatives, marked with the converb -tal, only combine with thetelic time adverbials, cannot be clefted, have to be included duringdo-so-replacement, only work with one item of an antonym pair, allow fornotional subjects that can take accusative in case of differential subjectmarking and cannot be scrambled to the position behind the resultative, andcan occur more than once in a sentence iff non-identical notional subjects arepresent. Ultimately, Mongolian resultatives are identified as not belonging tothe complement type and, as pseudo-resultatives, are posited at no less thanthree different positions: subject resultatives as TP under T’ and TP under vPand object resultatives as TP under V’. Potential depictives are divided intothe following types: Dorj Tuyaa-g a. nücgen-eer / b. nücgen-eer=n’ / c.nücgen-eer=ee / d. nücgen bai-h-ad / e. nücgen bai-h-ad=n’ / f. nücgenbai-h-d=aa / g. nücgen shalga-san ‘Dorj examined Tuyaa naked’ (pp. 194-195).Here, -aar is the instrumental case, =aa is reflexive-possessive (i.e.indicates that the possessor is the subject of the clause in which it occurs),=n’ indicates a third person non-subject possessor, and bai-h-ad isCOP-NPST.PTCP-DAT. Of these, a and g are said to be ambiguous, while b, d, eare subject-oriented and c and f non-subject-oriented. The test batteryconsists of adilhan (hopefully the same as onaziyoo), the temporal adverbials(only atelic), pseudo-cleft (needs the depictive to stay within the VP forobject depictives, while both OK for subject depictives), do-so-replacement(OK for ambiguous types, * for baihad, ? for the rest), insertion of anotional subject (OK with baihad(=n’/=aa)), scrambling (OK except for baihad;the ambiguous forms stay ambiguous), twosubject-depictives/object-depictives/mixed depictives with notional subject(OK with baihad(=n’/=aa)), two depictives with identical orientation withoutnotional subject (all *), wh-extraction (all ?/??), and orientation towardsobliques (mixed results). As a result, a, c, f and g with theirsubject-orientation can either occur under T´ or vP, while the object-orienteddepictives a, b, d and g occur under VP and d under V’. d, e and f, due to thepossible insertion of a notional subject, are not small clauses and thuspseudo-depictives.

The analysis of Korean is peculiar in that secondary predicates with -lo areclaimed to exhibit the properties of canonical resultatives and depictives,only to be ignored in the following discussion based on test batteries solelyapplied to predications in -key that are found to lack the properties ofeither canonical resultatives or depictives (e.g., both allowing for anotional subject to be added).

EVALUATIONLinguistic evidence and a lack of concern for it is the book’s majorshortcoming. Shibagaki relied on a single linguist native speaker for Khalkha,and on just two linguist native speakers of Korean. Single native speakerswill regularly fail to recognize some situations in which certain sentencescould be felicitously uttered, and linguist informants might perceivelinguistic data according to the schemes they were trained to believe, suchas, e.g., an extreme gradability of grammaticality judgments. A belief inuniversal linguistic structure may worsen matters. For Korean and in somecases even for Chinese, Shibagaki was comfortable enough to take sentencesmarked by “?” and “OK/??” as proof of grammatically acceptable sentenceswithout further discussion. (I don’t understand why “OK/??” is used instead of“OK/?”, as two question marks indicate something worse than just one questionmark at least on p. 76.) The situation is somewhat better for Japanese wherethe author often does discuss sentences marked with “?” before accepting them.But even here he is not consistent: on p. 76, the grammaticality of 49b (“?”)is argued for on the basis of gut feeling rather than reasoning, and on p. 77Shibagaki states that an example marked as “?” is lexically problematic butgrammatically fine, which would require such a division to be cognitively realin the first place. Handling of evidence might also be fair for Chinese wherehe relied on “a number of native Mandarin (non-)linguists from Beijing,Nanjing, Hong Kong and Taiwan” (p. 7), though Hong Kong is not particularlyknown for its Mandarin speakers.

A similar indifference toward linguistic evidence is probably behindShibagaki's approach to phonemes. Mandarin was transcribed without tones,leading to drastic underdifferentiation. For Mongolian, /tsʰ/ and /ts/ areboth written as < z >, /ʉ/ and /ʊ/ both as < u >, and non-initial /i/ and someof the instances of palatalization as < i >. Similarly, he doesn’tdifferentiate between functions and word classes, speaking of adverbs insteadof adverbials (passim) and adjectival forms instead of attributive forms (p.74).

As both the discussion of Chinese and Japanese intensively engage with and arewell integrated into discussions from previous literature, there is littledoubt that they actually contribute to our understanding of these forms,whether one is inclined to side with Shibagaki or, as a non-generativist wouldhave to do, to put forward an alternative analysis. Given the rather thinevidence employed in the chapter on Korean, I am not confident that as muchmay be said here. The data for Mongolian cannot be readily used as it is. Idiscussed a number of Shibagaki’s examples with four informants. While toosmall to create unproblematic evidence, this limited sample should besufficient to show how questionable grammaticality judgments on Mongolian arein this book.

One of the most problematic examples is the use of resultatives with atelicand depictives with telic time adverbials, which might have been consideredungrammatical on the basis of linguistic prejudice based on grammaticaltraining. In all examples below, the orthography was modified to be phonemicand English names were replaced with Mongolian ones:

(1)*[sic!] Dorj ene metal-iig 10 minut-iin tursh havtgai bol-tol davt-san. (p.179)D. DEM.PROX metal-ACC 10 minute-GEN during flat become-CVB.until hammer-PST‘Dorj hammered the [this!] metal flat for 10 minutes.’

(2)*[sic!] Dorj Tuyaa-g nücgen-eer=n’ 10 minut-iin dotor shalga-san. (p. 199)D. T.-ACC naked-INS=3POSS 10 minute-GEN within examine-PST‘Dorj examined Tuyaa-i naked-i in ten minutes.’

“tursh” ‘during/for’ in (1) was immediately accepted by all four informants,as was “dotor” ‘within’ in (2). In both cases, the event was understood to becompleted (as is implied by the use of the perfect to perfective past form-san, the only difference being that the event with “tursh” took the entire 10minutes for completion, while the event with “dotor” most likely lasted forless than 10 minutes. The reason this test doesn’t work might be that theEnglish adverbials Shibagaki had in mind have no one-to-one correspondence inMongolian. Similarly, Shibagaki assumed from previous studies that “anothertypical characteristic of a real resultative construction” is “that only oneof the antonym pairs qualifies as a resultative predicate in a resultativesentence” (p. 181):

(3)Dorj ene conh-iig ceverhen/ *[sic!] bohir bol-tol arch-san. (p. 181)D. DEM.PROX window-ACC clean / dirty become-CVB.until wipe-PST‘Dorj wiped this window clean / *dirty.’

Informants are fine with the variant with “bohir” ‘dirty’ if provided with aplausible context, e.g. some condition that prevents Dorj from noticing thathe is wiping a clean window with a dirty cloth. Two informants suggested analternative interpretation in which it was Dorj himself who got dirty, though,as this point differs over my informants, it would require furtherinvestigation.

In one set of cases, a mismatch between (possibly) acceptable Mongoliansentences and their actual meaning was caused by presupposing a universallinguistic structure, while ignoring well-documented features of Mongoliangrammar. Here, Mongolian sentences with a direct object and a dative-markedrecipient were equated with the English (“UG”?) Double Object Construction,while the examples “with the preposition ruu ‘to’” “represent the dativecounterpart of the double object construction” (p. 215):

(4)Dorj ene zahia-g Tuyaa-d / Tuyaa[-]ruu nücgen-eer=n’ ög-sön. (p. 215, 216)‘Dorj gave Tuyaa this letter naked [=without envelope].’/ ‘Dorj gave this letter to Tuyaa naked [=without envelope].’

(5)Dorj Tuyaa-ruu nücgen [D/T] / nücgen-eer=ni [?T] / nücgen-eer [D/T] yar`-san.‘Dorj spoke to Tuyaa naked.’

Virtually any English or Japanese language grammar of Khalkha Mongolianexplains that “ruu” is subject to vowel harmony and irregular initialconsonant dissimilation. It must thus be regarded as a case suffix (and ofcourse neither as a prefix nor preposition). More crucially, -ruu expresses amore specific meaning, namely, directionality: ‘down towards’ > ‘towards’. Itis thus not applicable to recipients. My informants are divided about whether(4) with -ruu is acceptable, with two rejecting and two basically acceptingit. Whether considered well-formed or not, it does manage to convey themeaning that Dorj gave the letter to somebody else who subsequently handed itto Tuyaa. (5) rather means that Dorj PHONED Tuyaa naked. The semantic oddityand subsequent overall problematic acceptability of (4) does affect thesuitability of these examples to determine which syntactic phrases depictivescan relate to, even though they still represent oblique arguments, as intendedby Shibagaki.

Other instances of questionable grammaticality judgments include the examplesof the aspectlessness of small clauses in Mongolian (both variants of thesentence in p. 183, footnote 1 are acceptable, and neither contains a smallclause), do-so replacement (which works just fine, at least for 36a-c on p.204), coordination of depictive adjuncts (which appears not to work forexamples like 39a on p. 207 containing bai-had=n’ without a significant breakbetween the phrases that probably indicates reconsideration, while it seems towork with two instrumentals without reflexive-possessive marking in sentenceslike 45f. on p. 213 -- unconditionally accepted by 2 informants, while theother 2 were slightly unsure), and wh-extraction (which works fine as well).

I conclude that Shibagaki’s data on Mongolian is not reliable enough to beused for typological work without additional research. A more serious attemptto come to terms with problematic data might start with a corpus as the sourceof sentences presented to informants, and work with a larger number ofinformants, either both qualitatively and quantitatively or justqualitatively. Shibagaki could also have used visual stimuli, so thatinformants produce the sentences in question rather than just confirm them.

This being said, parts of Shibagaki’s analysis can certainly be defended, buthow much of it is new? In Brosig 2009, found in Shibagaki’s bibliography butnot quoted, I discussed bare adjectives and adjectives in -aar as depictives,while ADJ+baihad was a priori excluded (as they obviously don’t belong to thesame prosodic unit as the main predicate, thus failing Schultze-Bernd andHimmelmann’s (2001: 77-78) definition of depictives). And they also take adifferent scope from actual depictives:

(6)?[sic!] Dorj Tuyaa-g nücgen bai-h-d=aa / nücgen-eer=ee shalga-san, harinBaatar sogtuu bai-h-d=aa / sogtuu-gaar=aa teg-sen.‘Dorj-i examined Tuyaa naked-i, but Baatar-j did so drunk-j.’ (p. 204)

Both variants were translated identically. However, with baihdaa, Dorj andBaatar examined Tuyaa within a longer period during which they were naked anddrunk, respectively, while the use of the instrumental -aar implies that thesestates are exactly simultaneous with the main predication, that is, Dorj andBaatar got naked and drunk on purpose expressly to examine Tuyaa. (As thisrarely coincides with real world experience, informants indeed considered thesecond variant a bit strange without context.) As in the case of the directive“ruu” discussed above, knowing something about the meaning of the sentencesinvestigated (that, given different forms, is expected to be non-identical)would have furthered the investigation.

There are other instances where Brosig 2009 contradicts or supplementsShibagaki 2012, and discussing those might have been helpful. For instance,while it is not controversial that the notional subjects with bare adjectivedepictives that Shibagaki constructed ad hoc are ungrammatical, that doesn’tnecessarily mean that such notional subjects are altogether impossible, e.g.“tergüün ihemseg yav-” head haughty go- ‘walk about haughtily’ was attested inmy corpus. I am not certain how such an example would have influencedShibagaki’s analysis, and I would have liked to know. Moreover, whileShibagaki would probably have been able to refute most of my rather brief andclumsy discussion on resultatives (quite parallel to what he did include intohis pseudo-resultative section on Japanese), I still wonder whether someinstances of the word class that has been termed “descriptive adverbs” inMongolian studies (e.g. Sechenbaatar 2003: 166-167) might qualify asresultatives (and, in that case, as real ones, not being subordinate converbalclauses as those marked by -tal):

(7)Hödölmör-iin baatr-iin gar-iig huga coh’-jee.movement-GEN hero-GEN hand-ACC into.pieces hit-2HAND.PST‘[Somebody] hit the hand of the hero of labour very severely / so that itbroke.’

Finally, some editorial issues have to be mentioned: cross-referencing betweenchapters is poor, with a number of dead links (reference in 4.1 to 4.1 on p.26, cross-reference to 3.3 instead of 3.4 on p. 69 and p. 89, reference to thenon-existing footnote 62 on p. 221 instead of to chapter 4 footnote 1, etc.),and pages 83-86 are missing (at least in my copy). “Int[ended meaning]” isused inconsistently, a few examples (e.g., p. 208) are aligned withcut-and-paste translations from other examples, the Mongolian copula iswritten -bai instead of bai- (p. 209) and the verb teg ‘do so’ should be teg-(p. 203). There are slight translation mistakes such as demonstratives missingin the translations on p. 179 cited above and 187. The English overall israther poor, though perfectly intelligible. While such mistakes are almostunavoidable, they are too numerous to have made it into print.

In summary, Shibagaki’s work sheds some new light on resultatives anddepictives in Chinese, Japanese and Mongolian. However, for Korean and to alesser degree Mongolian, it is regrettable that only part of the relevant datawas discussed, leaving the picture incomplete. Moreover, the sections onMongolian and Korean rely on problematic evidence, and a certain degree ofcarelessness with fitting evidence into his model can be observed in otherparts as well. As the grammatical judgments in the chapters on Mongolian andKorean cannot be taken at face value, syntactic typologists and generativisttheoreticians, surely among the target audience of this book, won’t be able tomake much direct use of such data. Given that Shibagaki’s discussion doeshave the potential to inspire further research into any of the four languagesanalyzed, a very thoroughly revised version of this book might still make avaluable contribution to the syntactic typology of secondary predicates.

REFERENCESBrosig, Benjamin. 2009. Depictives and resultatives in modern KhalkhMongolian. Hokkaidō gengo bunka kenkyū 7. 71-101.

Schultze-Bernd, Eva & Nikolaus Himmelmann. 2001. Depictive secondarypredicates in cross-linguistic perspective. Linguistic typology 8. 59-131.

Sechenbaatar, Borjigin. 2003. The Chakhar dialect of Mongol: a morphologicaldescription. Helsinki: Finno-Ugrian society.

ABOUT THE REVIEWERBenjamin Brosig is a doctoral candidate in general linguistics at StockholmUniversity with a master's degree in Mongolian studies and generallinguistics, plus Japanese studies as a minor. His research interests rangefrom aspectuality, evidentiality and other morpho-syntactic verbal categoriesover historical linguistics, dialect grammar and field linguistics topoliteness, with main expertise in Mongolic. Along with a dissertation onaspect and evidentiality in Middle Mongol, Khalkha Mongolian and KhorchinMongolian, current or recent research activities concern negation in Mongolic,the semantic field of temperature in Khalkha, and a documentation of DurvudOirat (led by Yu. Tsendee).

Page Updated: 04-Nov-2013