A CASE OF WAS-W IN HESSIAN: NEW EVIDENCE AGAINST SCOPE-MARKING Chis Bodenbende Depatment of Linguistics Univesity ofvictoia 1.0 INTRODUCTION This study investigates the was-w constuction in the Geman dialect Hessian (HE). It builds on pevious eseach done on was-w constuctions in High Geman by Kathol (1999) and Hinichs and Nakazawa (2000) in the famewok ofhead-diven Phase Stuctue Gamma (HPSG). Extactions out of embedded clauses occu fequently in High Geman (HG). The example in (1) shows the contast between a declaative sentence and a sentence with the extaction of the object-pp (mit jemand - Eng!.: 'with someone') of the embedded clause. The pupose of the extaction is to fom a question asking fo the object, as in (1 c). To ask fo the object, the object-pp is substituted with the coesponding wh-expession (mit wem 'with who'). (l b) shows the non-wh question with the statement of (1 a) as the embedded clause. (l c) is the extaction constuction, i.e. the matix question contains the wh-extaction fom the embedded clause. (1) a. statement Jakob hat mit iemand geedet. Jakob has with someone talked "Jakob has talked to someone." b. question with embedded clause Glaubst du [class Jakob mit iemand geedet hat]? believe you that Jakob with someone talked has "Do you believe that Jakob has talked to someone?" c. matix question which includes extaction out ofembedded clause Mit wem glaubst du [class Jakob geedet hat]? (KatholI999: 357) with who believe you that Jakob talked has "Who do you believe Jakob talked to?" Howeve, Geman speakes give pefeence to (2) ove (Ic). In (2) the wh-phase mit wem is only fonted within the embedded clause, with the accompanying inteogative ponoun was ('what') in the immediately supeodinate clause. Hence, (2) is an instance ofwas-w. (2) Was glaubst du [mit wem Jakob geedet hat]? OCathoI1999:358) what believe you with who Jakob talked has "Who do you think Jakob talked with?" The constuction in (2) is often efeed to as patial wh-movement as the mit wem is moved to the font in the fom of was, while at the same time also emaining at its oiginal position. The inteogative ponoun was functions like a "dummy" wh-wod and in the patial extaction can stand fo any othe wh-expession, such as wem ('whot-dat), wen ('who(m)'-acc), we ('who'-nom), wie ('how'), wo ('whee'), wann ('when'), was ('what'), wiesolweshalblwaum ('why'). Hence, the tem WS8-W efes to a question constuction in which an embedded whexpession OCathol 1999) o the whole inteogative phase (Hinichs and Nakazawa (H&N) 2000) is patially extacted to the font of the matix question in the fom of was. This study shows that HE allows was-w constuctions to be licensed by the pedicate wisse ('to know'), which, accoding to H&N, HG does not. Futhemoe, it shows that it is not just the embedded wh-expession but the entie inteogative phase that is patially extacted to the font, thus poviding futhe evidence against a scopemaking appoach and suppoting H&N's indiect analysis.
2 Bodenbende In 2 the pape pesents the HE data, which is discussed in 3 in light of Kathol's scope-making analysis. Evidence against a scope-making appoach leads to an H&N-based indiect analysis of the data in 4. In 5 I summaize the main fmdings of this pape. 2.0 WAS-W IN HESSIAN Hessian is the name ofa vaiety ofthe Fanconian dialects spoken in the Gennan state of Hessen but also in the aea ofrheinhessen, which fonnely belonged to Hessen but afte 1945 was amalgamated with othe lands 'to fonn the new state ofrheinland Pfalz. In HE, as in HG, wh-expessions (undelined) occu in non-embedded questions (3a), embedded questions (3b), and embedded inteogative clauses (3c). (3) a. non-embedded question Wo mache me hie? (Udezo 1999: 7) whee make we to "Whee do we go?" b. embedded question Koennte Sie uns vielleicht sage, ~ me was zum anziehe fue den Klaane da could you us maybe tell whee we what to wea fo the little-one thee fmne koenne? (Udezo 1999: 22) find can "Could you maybe tell us whee we can fmd something to wea fo the little one?" c. embedded inteogative clause...un wisse ga net, wo es hingehe tut. (Udezo 1999: 7) and know totally not whee it go-thee does "...and don't even know whee we ae heading to." HE also allows was-w constuctions, like the one in (4c) which is based on (4b). The question in (4b) is elated to the coesponding statement in (4a), in which the becoming of the wine is indicated as guud ('good'). To fom a question out of (4a), the adveb guud is eplaced by the wh-wod fo manne, wie, which is fonted. To etain veb-second stuctue, the auxiliay veb duht ('does') is fonted to the position behind wie. (4) a. statement De Wein, de neue, dubt sich guud ate. the wine the new-one does itself good gow "The new wine is coming along well." b. non-embedded question Wie dubt sich de Wein, de neue, ate? how does itselfthe wine the new-one gow "How is the new wine coming?" c. was-w Was kamme wisse, wie de Wein, de neue, sich dubt ate? (Witte 1974: 119) what can-we know how the wine the new-one itselfdoes gow "Can we know how the new wine will become?" (implying that we cannot) It is difficult to tanslate the sentence in (4c) into English, because its meaning is not eally a question but - a statement: 'We cannot know how the new wine will become.' The suface stuctue of the sentence is a question, but a question that implies that thee is no answe, theefoe expessing the impossibility of knowing the futue. shows. This sentence is gammatical in HE. Howeve, a simila constuction in HG is ungammatical, as (5) -
A case ofwas-w in Hessian: New evidence against scope-making 3 (5) * Was hat Hans gewusst, we kommen soli. (H&N 2000: 152) what has Hans known who come should "Could Hans have known who should come?" Hence, the veb wisse(n) I ('to know') behaves diffeently in HE and HG with espect to licensing was-w. High Geman does not allow the intepetation of the sentence as a statement of impossibility. Rathe, the mismatch between the pedicate that equies something to be known and the fact that this something is asked fo, and theefoe not known, makes the sentence ungammatical. How the gammaticality ofthe HE sentence in (4c) can be accounted fo is shown in the emaining pats of this pape. The analysis ofthis HE was-w sentence as a scope-making constuction is poblematic as the following section shows. Howeve, HE was-w also challenges the indiect analysis, in which cetain classes of pedicates license was-w constuctions, but wissen is not included in these classes. 3.0 WAS-W AS SCOPE-MARKING Accoding to Kathol (1999), it is only the wh-expession of the embedded inteogative clause that is patially extacted to the matix question, such as we ('who') in (6b). In compaison, (6a) shows a full extaction. (6) a. full extaction We denkst du hat das getunken? who think you has that dunk "Who do you think dank that?" b. was-w Was denkst du we das getunken hat? what think you who that dunk has "Who do you think dank that?" Kathol calls these constuctions (wh) scope-making constuctions, since the was-element is viewed as a scope make. The was indicates the syntactic domain that eceives a question intepetation in the semantics, i.e. it flags eveything that follows as pat of the question. Hence, Kathol assumes that the semantic intepetations of both sentences in (6) ae identical. Howeve, the same is not tue fo the HE was-w data, which is listed again in (7b). In compaison, (7a) shows the coesponding full extaction ofthe wh-expession wie. (7) a. full extaction Wie kamme wisse, dass de Wein, de neue, sich dubt ate? how can-we know that the wine the new-one itself does gow "How can we know that the new wine will come along?" b. was-w Was kamme wisse, wie de Wein. de neue. sich dubt ate? (Witte 1974: 119) what can-we know how the wine the new-one itself does gow "Can we know how the new wine will become?" (implying that we cannot) The meaning of the two sentences is quite diffeent. In (7a), wie efes to the knowing but not the gowing, i.e. asking 'how can you know this,' while in (7b) wie only has scope ove the gowing (scope is indicated by undelining). Theefoe, unlike the sentences in (6), the sentences in (7) do not esult in the same semantic intepetation. This povides evidence against a scope-making analysis ofwas-w in HE, as was is not a copy of wie but athe stands fo the answe ofthe embedded question. Was kamme wisse asks IF we know the answe to the embedded question that was stands fo but does not ask fo the answe itself. Simila evidence against the scope-making idea comes fom Dayal (1996) and (H&N 2000). Dayal also agues that the was of the was-w constuction is associated not just with the embedded wh-wod (wen) but with the embedded clause as a whole. The HG data in (8) illustates this point. In (8a) to (8c) the undelined phase is the 1 In HE, many infmitive veb foms dop the veb-final "n" that is so chaacteistic fo infmitive veb foms in HG. Thus, HG wissen becomes HE wisse.
4 BOOenbende object ofbehauptet ('claimed'), but in (8c) the undelined phase is the object of wissenlgewusst ('to knowlknown') as well. (8) a. statement E hat es behauptet, ohne ~ wiklich zu wissen. he has it claimed without it eally to know "He has claimed it without eally knowing it." b. was-w with 'behaupten' Was hat e behauptet, wen sie liebt? what has he claimed who she loves "Who did he claim that she loves?" c. WQS-W with 'behaupten' and 'wissen' Was hat e,ohne wiklich zu wissen, behauptet, wen sie liebt? (Dayal1996) what has he without eally to know claimed who she loves "Who did he claim that she loves without him eally knowing it?" The inteesting thing about (8c) is that both behauptet and wissen take the same complement was. While it can be assumed that in (8b) was stands fo the loved one, i.e. the peson wen, the same cannot be said about (8c), since wissen cannot take a diect object that efes to a peson. Hence, was stands not just fo the wh-wod wen but fo the whole agument wen sie liebt. This speaks against the idea of was as patial extaction of wen with scopemaking function. H&N bing foth futhe evidence, which is povided in (9) (H&N 2000: 150). (9) a. Was Hans sagt, Wen e vedaechtigt, das habe ich uebepueft. b. Was Hans sagt, lyen e vedaechtigt, *den habe ich uebepueft. what Hans says who he suspects (a) that/(b) him have I evaluated "I evaluated what Hans says about the peson whom he suspects." In (9a), was stands fo das, which stands fo the whole agument wen e vedaechtigt. In (9b), was stands fo den, which efes to wen, i.e. the suspicious peson. In the sentences, this diffeence has been indicated though undelining the agument that is efeed to. Following fom the gammaticality, in this sentence, again, was stands fo the whole agument and not just fo the wh-expession. This leads to the conclusion that the was-w constuction is not a scope-making constuction as assumed by Kathol. The new HE evidence discussed in this pape and the evidence fom Dayal (1996) and H&N (2000) seem to suggest an indiect dependency (Dayal 1994) between was and the embedded wh-expession. This leads to an indiect, i.e. non-scope-making, analysis ofthe was-w constuction. 4.0 INDIRECT ANALYSIS 4.1 Vebs that license was..w constuctions Tuning away fom the scope-making focus of pevious analyses, H&N (2000) investigate the pedicates that license was-w constuctions in HG and the type coecion necessay to account fo the appaent mismatch between the syntactic fom ofthe embedded question and its semantic function. Following Ginzbug and Sag (2001), H&N distinguish fou classes of pedicates that take eithe whsentential complements (+WH) o non-wh-sentential complements (-WH). The table in (10) is a modification of a table povided by H&N (2000: 154). It gives sample vebs of English fo each, states whethe the pedicates take +WH o -WH complements, lists ifthey license was-w constuctions in HG, and indicates examples.
A case ofwas-w in Hessian: New evidence against scope-making 5 (10) Pedicate classification Resolutive Pedicates (RP) tell guess oedict Factive Pedicates (FP) eveal know discove Question Pedicates (QEP) ask wonde investie:ate Tue/False Pedicates (TFP) believe deny Dove +WH&-WH +WH&-WH +WH was-w do not license was-w do not license was-w example: (11) example: (12) example: (14) -WH was-w example: (13), In HG, the set ofwas-w licensing pedicates includes some (RP, see (11» but not all (FP, see (12» vebs that allow both +WH and -WH complements outside ofthis constuction. (11) RP (H&N 2000: 151) a. + WH complements Hans hat entschiedenlbeichtetlsich vogestellt, we kommen soli. Hans has decided/epoted/imagined who come should b. -WH complements Hans hat entschiedenlbeichtetlsich vogestellt, dass Pete kommen soil. Hans has decided/epoted/imagined that Pete come should c. was-w Was hat Hans entschiedenlbeichtetlsich vogestellt, we kommen soli? What has Hans decided/epoted/imagined who come should (12) FP (H&N 2000: 152) a. + WH complements Hans hat gewusstlvegessen/sich einnet/eaten, we kommen soli. Hans has known/fogotten/emembeed/guessed who come should b. -WH complements Hans hat gewusstlvegessen/sich einnetjeaten, dass Pete kommen soil. Hans has known/fogotten/emembeed/guessed that Pete come should c. was-w *Was hat Hans gewusstlvegessen/sich einnet/eaten, we kommen soli? What has Hans known/fogotten/emembeed/guessed who come should TFP ae the class of vebs that appea most fequently in' was-w constuctions, but outside of these constuctions they only allow -WH complements as shown in (13). (13) TFP (H&N 2000: 151) a. + WH complements *Hans hat gesagt/geglaubt, we kommt. Hans has saidlbelieved, who comes b. -WH complements Hans hat gesagt/geglaubt, class Pete kommt. Hans has saidlbelieved that Pete comes c. was-w Was hat Hans gesagtigeglaubt, we kommt? What has Hans saidlbelieved who comes In contast, vebs that only license +WH complements but not -WH complements do not license was-w. These ae QEP and they patten as illustated in (14). (14) QEP (H&N 2000: 151) a. + WH complements Hans hat nachgefoschtlgefagt, we kommt. Hans has investigated/asked, who comes
- 6 Bodenbende b. -WH complements *Hans hat nachgefoschtlgefagt, dass Pete kommt. Hans has investigated/asked that Pete comes c. was-w *Was hat Hans nachgefoscht/gefagt, we kommt? What has Hans investigated/asked who comes H&N investigated the diffeence between these fou classes to establish why RP and TFP allow was-w constuctions, while FP and QEP do not. They point out that RP and FP have fact-denoting aguments that ae taken to povide the answe to the question of the embedded +WH complement. The diffeence between (15) and (16) illustates this. (15) shows that in conjunction with an FP, Jean only discoveed the question, not the answe, while in (16), in conjunction with a QEP, Jean's action aims towads the content of the question, Le. the answe. (15) FP (H&N 2000: 153) Jean discoveed an inteesting question. The question was who left yesteday. It does not follow that: Jean discoveed who left yesteday. (16) QEP (H&N 2000: 153) Jean asked an inteesting question. The question was who left yesteday. Hence: Jean asked who left yesteday. Although the substitution test in (15) shows that FP and RP involve fact-denoting aguments, the gammaticality of (12 a) and (15) shows that they do allow +WH complements nonetheless. Thus, to link answe and question in (12 a), the question we kommen soil is foced into a fact that esolves the question. This allows the wh-inteogative we to appea in the fact-denoting agument position of the FP. This focing is called "type coecion" and its implementation in HPSG is discussed in section 4.2. What follows fom this is that RP and FP behave diffeently fom QEP. In thei pape, H&N then go on to discuss what distinguishes RP fom FP, to establish RP and TFP as a natual was-w licensing class. Howeve, as the following discussion of the HE data shows, this is unnecessay fo HE. In this dialect, RP, TFP and FP license was-w constuctions. Hence, excluding QEP, as shown, is sufficient. (17) Was kamme wisse, wie de Wein. del neue, sich dubt ate? (Witte 1974: 119) what can-we know how the wine the new-one itself does gow "Can we know how the new wine will become?" (implying that we cannot) The gammaticality of was-w in (17) with the FP wisse ('to know') in HE indicates that thee is evidence in this dialect that FP license was-w. As in (15), but unlike (16), it does not follow fom the sentence in (17) that we know how the wine will become. The challenge that this data poses fo H&N's indiect analysis is to include FP as was-w licensing fo HE. Howeve, that is not much of a challenge, since HE behaves moe geneally than HG by allowing all vebs that license declaative complements to also license was-w. Hence, thee is evidence that in this dialect RP, TFP and FP fom a natual was-w licensing class. Fo both RP and FP to take +WH coml'-lements it is necessay to apply type coecion of questions to facts, i.e. facts esolved (fo FP) and unesolved 2 (fo RP). The implementation of type coecion in HPSG is discussed in the following section. 2 H&N call these 'popositions.'
A case ofwas-w in Hessian: New evidence against scope-making 7 4.2 Type-coecing the denotation of the embedded inteogative The sentence in (17) contains the embedded inteogative wie de Wein, de neue, sich duht ate. The embedded inteogative can be deived syntactically as shown in (18). The semantic epesentation of the inteogative phase is given in (19).. In (18), STORE seves as the scope indicato of the wh-index. It diffes fom SLASH in that SLASH allows the wh-expession to be extacted and bound offat a highe level, while STORE is only a placeholde fo the index to indicate scope. In (19) it can be seen that "embedded inteogatives ae assigned denotations of type question" (H&N 2000: 157), containing the featues INDICES and FACT. This is diffeent fom H&N as in thei case the embedded inteogative is assumed to be a poposition. Hence, they use the featue PROP instead of FACT. Futhemoe, in H&N the question was aiming at a peson, while hee it is aiming at an advebial of manne. Accodingly, the semantic epesentation in (19) incopoates a estiction on the veb ate ('to gow') in the fom of the advebial of manne. Following Kaspe (1994), soa stands fo "esticted state of affais" and has the featues QFSOA, which stands fo "quantifie-fee state ofaffais," and RESTR. (18) LOC[5] CONT =x WH <[1]> NP STORE {[I]} l [1] I~ {manne-el([1])} - - ONTS [~] ~SH{} ~ STORE{} S ONT [8~ ~ SLASH {[5]} WH <> STORE {[I]} wie 'how' de Wein, de neue, sich duht ate 'the wine, the new one, itself does gow' (19) question INDICES {[I]} CONT [2] fact SIT s soa gow-el ] FACT [8] QFSOA [9] LgROWER wine J SOA NUCL RESTR manne-el ~ISSUE [9] MANNER [1] In the was-w constuctions the embedded complement syntactically has all the popeties of an embedded inteogative, while semantically it seems to have the popeties of a fact, as discussed in section 4.1 above. This is whee type coecion comes into play.
8 Bodenbende (20) shows the lexical enty fo a wisse ('to know') that licenses was-w. Fom H&N's discussion it follows that thee ae at least two lexical enties fo these vebs, one fo when they occu in non-was-w constuctions and one fo was-w. (20) PHON wisse SYNSEM CAT~UBJ <NP[3]> J LOC CONT [6] STORE {[4]} SLASH {np_was[4]} [OMPS <S[CONT [2] question]> soa NUCL know-ei ] KNOWER [3] KNOWLEDGE ~WER {[4~ IQUESTION [2]] This lexical enty fo wisse shows that the KNOWLEDGE ofwisse in a was-w constuction is of type qafact (shot fo question-answefact) instead of question as in (20). This is a deviation fom H&N's HG analysis, since in thei analysis they intoduced the type qa-poposition. Howeve, as a distinction between esolved facts and unesolved facts, Le. popositions, poves to be unnecessay in HE, I intoduce qa-fact as a subtype of a factive supetype. In (20), the qa-fact contains the link to the question wie de Wein, de neue, sich duht ate as well as to the answe was. Thus, QUESTION connects to the embedded inteogative (19), which is of type question. This is the type coecion, the question is esolved into a fact, because wisse takes the qa-fact that contains the question. The lexical enty fo was of the was-w constuction is given in (21/. Accoding to H&N, the was of a was-w constuction has a special type of local value np_was and its CaNT I INDEX value is esticted to being an answe (H&N 2000: 158). (21) PHON was SYNSEM np_was CAT np LaC CaNT How (20) and (21) combine fo the question Was kamme wisse, wie de Wein, de neue, sich duht ate? is shown in (22). 3 Taken fom H&N (2000: 158), but eplacing the tag [1] with the moe appopiate tag [4].
A case ofwas-w in Hessian: New evidence against scope-making 9 (22) s NP uestion ~ CONT INDICES {[4]} FACT [11] ~ SLASH {} WH <> STORE{} STORE {[4]} fact SIT s ao LOC[5] CONTF [4~ CONT [11] know-el SOA [6] KNOWER [3] WH <[4]> SLASH {[5]} WH <> was STORE {[4]} what NUCL s ~1act KNOWL ANSW {[4]} QUE [2] S [CONT [2]] kam me wisse wie de Wein, de neue, sich duhtate can we know how the wine, the new one, itself does gow This section shows that with a few adaptations, H&N's indiect analysis can popely account fo the HE was-w data. 5.0 CONCLUSION This pape used data fom the Geman dialect Hessian to the discussion ofwas-w constuctions in Gennan. Since the data could not be analyzed with a scope-making appoach, but wee unpoblematic fo the indiect analysis, this study povided futhe evidence fo the indiect analysis of was-w constuctions (H&N 2000), as opposed to scope-making (Kathol1999).,.. Additionally, this study discusses a diffeence between HG and HE in the classes of pedicates that license was-w. In HE, factive pedicates, such as wisse, ae also was-w licensing, while in HG they ae not. f""" I
- - 10 BOOenbende REFERENCES Dayal, V. S. 1994. Scope making: In defence of indiect dependency. In Papes on wh-scope making. Abeitspapiee des SFB 340 N. 76, pp. 107-130. Ginzbug, J. and I. A. Sag. 2001. Inteogative investigations: The fom, meaning and use of English inteogatives. Stanfod: CSLI Publications. Hinichs, E. W. and T. Nakazawa. 2001. The was-w constuction in Geman: A case study in type coecion. In D. Flickinge and A. Kathol (eds.) Poceedings of the 7th intenational HPSG confeence. Stanfod: CSLI Publications, pp. 147-165. Kaspe, R. 1994. Adjuncts in the Mittelfeld. In J. Nebonne, K. Nette and C. Pollad (eels.) Geman in Head-Diven Phase Stuctue Gamma. Stanfod: CSLI Publications, pp. 39-69. Kathol, A. 1999. The scope-making constuction in Geman. In G. Webelhuth, J-P. Koenig and A. Kathol (eds.) Lexical and constuctional aspects of linguistic explanation. Stanfod: CSLI Publications, pp 357-371. Udezo, Albet. 1999. Asteix babbelt Hessisch 2 - RufJun Runne. Tanslated into Hessian by Juegen Lebe. Stuttgat: Les Editions Albet Rene/Goscinny-UdezolEhapa Velag Stuttgat. Witte, Hedwig. 1974. Hessisch, wie es nicht im Wotebuch steht. Fankfut: Societits-Velag.