References
Transcription
References
Going Romance 2003, Nijmegen 20-22 November 2003 A look at the Gallo-Romance trouble with muta cum liquida through the positional prism Philippe Ségéral Tobias Scheer Université de Nice Université Paris 7 [email protected] [email protected] this handout and more at www.unice.fr/dsl/tobias.htm Purpose - to show that there is no universal or cross-linguistically stable syllabification of obstruent-liquid clusters. Gallo-Romance is particularly well suited to demonstrate the eternal trouble with TR sequences. - while liquid-obstruent clusters (= RT) always represent Coda-Onset clusters, surface TR may be either syllabified as a branching Onset or as a Coda-Onset sequence. We hold that this typology of TR-identities is not sufficient in order to cover the Gallo-Romance situation. A third object is called, for, i.e. a mono-positional "affricate". - all three possible identities of TR are present in the lexicon, they are not derived by some Coda-capture or the like. The classical Maximal Onset approach encodes a tendency/ markedness, but is irrelevant in the on-line functioning of the grammar. A. General framework [1] Evolution of consonants from Latin to French. Overview : a. b. c. # __ C. __ __ .C "word initial" "post-consonantal" d. __ # "Coda" a. examples p porta porte talpa taupe rupta b bene bien herba herbe cub(i)tu t teela d dente k cor ceera *capu g gula gente gamba f fame s sorte r reege toile dent cœur cire chef gueule gent jambe faim sort roi cantare ardoore rancoore merceede arca angustia argentu virga infernu versaare terra cam(e)ra mer(u)la cum(u)lu arma cornu malva chanter ardeur rancœur merci arche angoisse argent verge enfer verser terre chambre merle comble arme corne mauve plat(a)nu adveniire facta lup(u) *cap(u) coude ub(i) bib(o) OFr plane mariit(u) avenir nuud(u) faite amiic(u) rig(i)da sagma roide somme steph(a)nu musca barba Etienne mouche barbe alba aube l luuna lune m mare mer n naasu nez B viinu vin route gamba jambe cantaare chanter naav(i)gaare nager w werra guerre *skarwahta échauguette rabja rage j jocu jeu 1 leu chef où OFr boif mari nu ami noos cor amaar(e) fiil(u) cuul(u) fam(e) noon bov(e) nous cœur aimer fil cul faim non bœuf maaj(u) mai e. V __ V "intervocalic" *sapuutu riipa *nuuba faba viita coda locaare liceere paacaare *agustu paagaanu su rive nue fève vie queue louer loisir payer août païen deeforiis dehors causa chose poire pira veela voile amaare luuna paavoone lavaare *cawa raja aimer lune paon laver OFr choue raie b. summary thereof p p b b t t d d k k/s/$ g g/J f f s s r r l l m m n n B v w g y J p b t d k/s/$ g/J f s r / ÉPENTH. l / ÉPENTH. m n v g J ø ø ø ø I I/U ø ø r U ~ø ~ø ø ø/f ø/f ø ø ø ø r/ø l /ø ~ø ~ø ø/f ø/v ø/v ø ø ø / I / Iz ø/I ø z r l m n ø/v I I [2] Two different causalities : a. "phonetic" : produces a segmental effect because of an interaction with neighbours (typically with adjacent vowels) where melodic primes are exchanged : e.g. intervocalic p / b / B —> v, but —> ø if either the preceding or the following vowel is o or u (yields more than one result in a given cell in [1]). b. "positional" : produces a segmental effect because the consonant holds a particular position in the syllabic structure. c. crucially : "positional" > "phonetic" : no "phonetic" process can produce a result which is incompatible with the "positional" regularity. [3] The positional regularity : a. b. # __ C. __ result ≥ original segment : integrity or [max.] strengthening (fortition) c. d. e. __ # V __ V result ≤ original segment : weakening (lenition) or [max.] integrity __ .C STRENGTH WEAKNESS [4] Framework : a. syllabic structure : Government Phonology, in particular "CVCV" option (Lowenstamm 1996) C | p V | o C | r V \ t e C | t V | a C | l V C \ V / u C | n V | a / r a b. initial CV (Lowenstamm 1999) : the "beginning of the word", i.e. what phonologists use to refer to as "#", is an empty CV unit. c. lateral relations : Government (G), Licensing (L), both right-to-left d. Empty Category Principle (ECP) : a Nucleus may remain phonetically unexpressed if it is 1) governed by a following (full) Nucleus or 2) word-final (domain-final). [5] Interpretation of consonantal "strength" / "weakness" (Ségéral & Scheer 2001a) : a. three possible configurations of the two lateral forces Government and Licensing (the fourth logical possibility * ø __ ø is ruled out independently, cf. 4d) : i. ii. iii. G G G ø __ V Lic V __ ø V __ V Lic Lic 2 b. correspond to : i. initial ii. Coda + postconsonantal __{#,C} = {#,C}__ "Coda-miroir" (Ségéral & Scheer 2001a) iii. intervocalic V__V c. strength = "preceded by a governed empty nucleus" : only the word-initial and the post-consonantal position correspond to this description G ø ___ B. "Branching onsets" [6] C+yod (Ségéral & Scheer 2001b) : a. are possible "branching onsets" (cf. modern French : rabiot ra.bjo not *rab.jo). b. but always heterosyllabic in Gallo-Romance : rab.ja rage, not *ra.bja i. in all cases (but tj) the fate of C+yod clusters can be interpreted as resulting from • the regular weakening of C in Coda position • the strengthening of yod in post-consonantal position, the result being either gemination (modjolu > mojjolu > moyeu), or segmental fortition of yod to t$ / dJ (rubju > rouge [ruJ) ii. the preceding tonic vowel develops as a checked vowel : rabja > rage, not *rège (compare with : arca > arche but faba > fève). [7] the trouble with mutae cum liquida (henceforth TR), type colubra: a. b. i. colubra cathedra tenebras tonitru *taretra (?) [cl. terebra] taratru (Is.S) ii. podagru alacre i. pullitra, -tru feretru ii. integru palpebras palpetra (Varr.) iii. *presbytru [cl. presbyter, presbyteru] c. *colubra *catedra *teneblas *tonitru *taretra *podagru *alecru ---*integru -*palpetra -*presbytru ad retro / de retro couleuvre OFr. chaiere OFr teniebles OFr tonoire tarière OFr pouacre OFr (h)aliegre poutre OFr fiertre OFr entre entier OFr paupres paupière prêtre OFr prevoire, proarrière / derrière it. puledro Nota : retro > OFr riere [8] The problem : a. the fundamental trouble i. stress moves one vowel right in ...vCvTRv# words (type colubra) : Fr. couleuvre, Sp. culebra. ii. the tonic vowel is unchecked : e.g. diphtongization of short (open) e > ie in *catedra > Ofr chaiere, *teneblas > Ofr teniebles, petra > pierre as in pede > pied (compare with herba > herbe). b. syllabic translation i. stress shift ii. evolution of the tonic vowel supposes supposes ... C V T . R V # ... C V . T R V # c. further related facts : i. evolution of T : supposes ... C V . T R V # b > v in colubra > couleuvre as in faba fève, vs subtiile > OFr sotil, cub(i)tu > OFr cote, code ii. final V > @ (= schwa) even when ≠ a supposes ... C V . T R V # *colubra > couleuvre, fabru > Ofr fevre, patre > Ofr pere vs arcu > arc iii. the evolution of velars supposes ... C V T . R V # *integru > entier, sacramentu > Ofr sairement as facta > faite 3 [9] Numerous different proposals, of two main types : a. stress placement depends on syllable structure i. anaptyctic vowel (e. g. De Groot 1921), followed by syncope : colubra > colubera > colub(e)ra ii. gemination followed by degemination : • Fouché 1969: 152 (colubra > colubbra > colubra) • Bourciez 1971 : 27 (colubra ("sermo cotidianus") X *collubra ("sermo rusticus") -> colubra "moyen terme") iii. successive or concomitant contrasting syllabification of TR : • Timparano 1965 : 1093 "Piuttosto che ad una netta separazione tra una pronunzia popolare in-tégrum, mantenutasi dall'epoca preistorica fino al sorgere delle lingue romanze, e una pronunzia dotta ín-te-grum, mantenutasi con altrettanta costanza almeno per tutta l'età classica, io credo a une prevalere ora dell'una ora dell'altra accentazione (in conseguenza di sillabazioni diverse) in diverse epoche", and 1095-96, • Loporcaro to app. : §3 "Se si ammette dunque per il latino arcaico e tardo l'eterosillabicità, si deve ricostruire un'evoluzione in quattro fasi : a. lat. arcaico -VC.RV-, b. lat. class. -V.CRV-, c. lat. tardo/proto-rom. -VC.RV, d. lingue rom. -V.CRV-« b. evolution makes stress placement independent from syllable structure. The algorithm that assigns stress changes under analogical pressure of various kinds. For example : • Pope 1952 : 100 “...the rule for the position of the tonic stress is simplified for in Late Latin it may be said that the penultimate syllable is stressed whenever it contains a long vowel, a diphtong,or a vowel of any kind followed by any two consonants or a double consonant”, • Ward 1951 : 484 “And our accent rule should read : the Latin accent falls on the penult not only when that syllable is long, but also when it is short, provided that it consists of a short vowel followed by stop + l”, • Steriade 1988 : 399 “I suggest then that the stress shift in integra was due to the reinterpretation of the Latin Stress rule as : Stress the penult if it is followed by a consonant cluster”, • Lahiri, Riad & Jacobs 1999 : 395 "trend towards paroxitony", • Bullock 2001. [10] Not only a Romance problem : a. Latin vowel reduction in internal unstressed open syllables ("internal apophony") i. open syllable facio / conficio = tonitru, pullitra, etc. ii. checked syllable factus / confectus = tenebrae (*tenibrae), integru (*intigru) b. metrical ambiguity in Latin (positio debilis). See Timparano 1965, Loporcaro to appear [11] Our position : the problem is of syllabic nature, i.e. [9a] is correct. The new accent rule advocated by Pope, Ward, Steriade... cf.[9b] is nothing but a linear way of saying that any CC cluster has become heterosyllabic... C. What is / can be TR ? [12] Two known syllabic representations for TR clusters : a. Coda + Onset b. branching Onset with consonantal interaction "<==" whereby the sonorant establishes a lateral relation with the obstruent, i.e. roughly a "branching Onset" (Scheer 1996,1999) G G T ø R v T ø R | | t <= r v N.B.: these are the representations that we actually assume. For the sake of exposition, though, we will be using the familiar lingua franca "Coda-Onset and branching Onset" below. [13] colubra -> colubra supposes that TR was [12b] in Classical Latin, and became [12a] in Late Latin. This is because a. velar + r : soc(e)ra > OFr suire as cocta > cuite, opposed to locaare > louer cf [8.c.iii] b. labial + l : tab(u)la > tôle [and Picard regular evolution of bl > ul] ; even in table < tab(u)la, the evolution of the stressed vowel shows that it is in closed syllable (*tèble). c. C + jod : rabia > rage supposes a fortition of j to dJ (analogous to the one of j in intial position : jocu > jeu) which cannot be explained unless j is in "strong position", that is rabia is syllabified rab.ja d. other Romance languages : evolution of tr/dr : creed(e)re > creire (Occ.), creure (Cat.), evolution of C+j in italian > CCj, etc. 4 [14] Our claim : in subsequent evolution, i.e. in Gallo Romance, TR could not be a [12a] structure (Coda-Onset) anymore. Its syllabic interpretation had to change : back to [12b] (branching Onset), or towards a third structure: mono-positional TR, i.e. an "affricate". MONO-positional BI-positional a. b. c. Coda+Onset branching Onset "affricate" with "<==" G G G C | T henceforth : V C | R T•R V C V C | | T <= R T=R V C V /| T R TR [15] French epenthesis in CR clusters emerging from internal post-/pretonic vowel syncope : in CR, C = i. a nasal (m/n) or l, ii. s/z i. cam(e)ra chambre ii. *ess(e)re être sim(u)laare sembler laz(a)ru ladre cin(e)re cendre spiin(u)la épingle moudre mol(e)re [16] In the case of epenthesis, T=R [12b] is excluded, because a CV unit would have to "fall from heaven" : C V C V C V =//=> * C V C V C V C V | | | | | | | | | | | k a m r a k a m b r a [17] Principled view on the origin of syllabic material: a the insertion of syllabic space (i.e. x-slots) cannot be provoked by purely melodic reasons, i.e. the existence of melodic information cannot trigger the appearance of a CV unit by itself i. example of floating melody: French "consonne de liaison" : C V C V =//=> * C V C V C V | | | | | | | | | p @ t i t p @ t i t ii. floating tonal information cannot entail the appearance of a vowel iii. morphemes without prosodic support / templatic effects : for an example in Romance languages (the case of the Portuguese verbal conjugation, see Carvalho (in press). b. CV-insertion can only be commanded by morphology i. Latin : venit (pres.) vs ve:nit (pf) ii. Greek : voc. sg. daîmon / nomin. sg. daímo:n iii. Classical Arabic iv. Muskogee (data from Haas 1940) basic form (I) katabcompletive tenseless (basic) nisintensive (II) kattabcompletive (immediate past tense) ní-h-sfrequentative (III) kaatabcompletive (remote pas tense) nî:sh k t b n s | | | | | C V C V C V C V C V C V C V | | | a a i c. or by prosodic events such as stress : e.g. Italian Tonic Lengthening /kasa/ -> [ka:sa] C V C V C V | | | | k a s a 5 [18] => Epenthesis : C | k V | a C | m V C | r ==> V | a C | k V | a C | m V C V /| | br a [19] Representing TR as TR ("affricate") is not an original proposal : see Hirst 1985, Steriade 1994, Rennison 1998, Lowenstamm 2003 for affricate analyses of TR clusters in various languages. [20] A first generalization : internal C1 C2 C3 -> C1 C3 (except when C2 = s) jaune - l k t - kulk(i)ta OFr coute a. - l b g - galb(i)nu - m p t - komp(u)taare conter - n d t - vend(i)ta vente - r b t - derb(i)ta OFr dert[r]e - r m t - dorm(i)tooriu dortoir - s t m - test(i)mooniu témoin - r k b - ark(u)ballista arbalète charme - r p k - erp(i)ke herse - r p m - *karp(i)mu (See Fouché 1969 : 829 sqq ; Steriade 1988 : 392 ; Jacobs 1992 : 64 ; Lahiri, Riad & Jacobs 1999 : 307) b. -rdr-rtr-ndr- c. C | g perd(e)re turt(u)rilla vend(e)re V | a C | l V -rkr-rgr-ngl- perdre OFr tortrele vendre C V C | (b) (i) n V | u C | p V | e kark(e)re surg(e)re ung(u)la C | r V C OFr chartre sourdre ongle V C V /| | (d) (e) d r e d. CC is the maximal internal cluster that is tolerated by the grammar. [20] On this analysis, TR is the strengthened version of [r]. This is in line with the general positional regularity [3] : the strengthening at hand takes place in post-consonantal position (strong position), i.e. after a governed empty Nucleus. [21] naturalness of TR : the case of Madagascan a. TR is necessarily TR ("affricate", monopositional) i. Madagascan is CV on the surface : no word may end in a consonant *C#, and any sequence of two consonants is prohibited *CC. Except for homorganic NC clusters and four other consonantal sequence s: [ts], [dZ], [tr] and [dr]. consonantal inventory p b t d k g f v s z ç h m n N r, l ts, dZ tr, dr ii. examples of [ts], [dZ], [tr] and [dr] #__ C__ V__V ts tsahu muntsana katsaka rumeur, broyé, maïs bougie, pistache, lascif dZ dZiru vuandZu dZedZu tr tranu antra tratra maison, compatissant, poitrine dr dredru sendra sedra flagornerie, par hasard, épreuve iii. on static grounds alone, [tr] and [dr] could hardly be interpreted as branching Onsets because this would suppose the existence of labial and velar clusters as well: [pr], [br], [pl], [bl], [kr], [gr], [kl], [gl]. These, however, are absent. 6 b. [dr] is 1) monopositional and 2) the strong version of [r] i. composition: [word 1 + word 2] whereby the final -a of word 1 is lost, and the initial consonant of word 2 strengthens (see Ali 2003 for details). word 1 word 2 composition gloss word 1 gloss word 2 gloss composition f > p fufuna fati fufumpati odeur cadavre odeur fétide v > b fufuna vadi fufumbadi odeur époux fiancé l>d fufuna luza fufunduza odeur malheur pressentir le mal r > dr manana reni manadreni avoir mère avoir une mère s > ts fufuna savuni fufuntsavuni odeur savon odeur du savon avoir droit avoir le droit z > dZ manana zu manandZ Zu h > k uluna hala personne détesté ennemi uluNkala ii. strengthening occurs in post-consonantal position. [dr] thus represents one single consonant, and moreover is the strong version of [r]. D. TR and T=R [22] An interpretation of occasional French r-Metathesis in the light of [20] : a. data (from Pope 1952 : §124 ; Fouché 1969 : 751-3 ; Bourciez 1971 : §178,180) i. from post-cons TR --> #T__ iii. from R in Coda --> #T__ temp(e)raare tremper torc(u)lu treuil fimbria frange *berbiice brebis *formaat(i)cu fromage *fim(o)riaare Ofr frambaier fund(u)la fronde *turb(u)laare troubler *torsaare trousser ii. from intervoc. TR --> #T__ *bertjolu Ofr breçuel germ. *thorp troupe / trop *bib(e)raat(i)cu breuvage *tursum (< thyrsum) Ofr tros néerl. versch friche furloone (< frk *hurslo frelon b. two regularities : i. migrating R always docks on a word-initial T. ii. in an overwhelmingly majority of cases, the origin of migrating R is a Coda position. (lists a.i. and a.ii. are probably not very far from being exhaustive, list a.iii. could be continued (cf. frelater < Neerl. verlaten), cravache (< Germ. Karbatsche), cramoisi (< Sp. carmesí), calembredaine (/ calembourdaine), fripe (OFr frippe, ferpe, felpe from *faluppa), fredaine (< fard-), etc.) c. if the initial TRs created are TR, metathesis makes sense : i. unlinking (= lenition) of r in weak position (Coda) : in line with the tendency of r to weaken in wordfinal Codas (e.g. -er infinitive > [-e]) and even sporadically in internal Codas (OFr berfroi, mod. beffroi [befrwa], pop. [mEkr@di] for mercredi] (cf. Bourciez 1971 : §180), ii. tendency to preserve the original segmental information. In the case of metathesis, this is achieved by the creation of TR clusters. Numerous parallels in French phonology. For instance : • preservation of nasality : n in (final and internal) Codas is unlinked (= lenites), but the segmental information is partially preserved : nasality docks on the preceding vowel. bonne bon bonté [bOn@] [bÕ] [bÕte] N N b O n @ b O n | | | | | | | C V C V C V C V • conservation of word-final unlinked (= lenited) "consonnes de liaison". b O n t e | | | | | C V C V C V obstruents as floating segments = the d. no migration of R from intervocalic TR : TR is not TR in intervocalic position ? (The lenition of intervocalic TR leads to weakening of T, not to R loss). 7 [23] Parasitic r in French [all periods] a. data (from Fouché 1969 : 756-760 ; Bourciez 1971 : §178) i. #T__ V iii. C.T__V vitiicula vrille perdiice trésor *cardoone (< carduus) thesauru Neerl. tingel tringle celt. *derb(i)ta *term(i)te ii. V.T__V *can(a)pu patte X dial. gadrouiller patrouille encaustu Flemish pleute (??) pleutre spelta pol(y)pu arcuballista calendaariu regesta rusticu mite (< neerl.) + -aille Mod. Fr. tarte perdrix OFr chardon / chardron dartre tertre chanvre encre épeautre pieuvre arbalète, but arbalétrier calendrier registre rustre OFr mistraille (mod. mitraille) pop. -> "l'Entartreur" b. docking-site : overwhelmingly post-consonantal (or initial). c. interpretation : creation of a TR cluster. [24] One more observation regarding [22] and [23] : a. the observation : R never migrates or appears next to an intervocalic T (that is, all intervocalic TRs seem to be either inherited or resulting from syncope). This is understandable when T is lenited to ø or to a vocalic element (case of t, d or k, g), but not when T is a labial, in which case lenition produces the obstruent v) : i. no migrating r landing on an intervocalic v : travée -/-> *tavrée, brève -/-> *bèvre, trêve -/-> *tèvre, *gavrement for gravement... ii. no parasitic r ever lands on an intervocalic v : *rivre for rive, *fèvre for fève, *touvre for trouve,... b. interpretation : TR is prohibited in intervocalic position. c. as a consequence of the general prohibition of T•R [14], TR in French is either TR or T=R, but i. TR is the only possibility in post-consonantal position (cf. epenthesis), and probably more generally speaking. ii. T=R seems the only possibility in intervocalic position : cf. metathesis / parasitic r [25] Evidence for bi-positional T=R in French, alongside with TR : the evolution of intervocalic tr/dr a. examples i. r < postonicTR ii. rr < pretonic TR araatru OFr. arere atramentu OFr. arrement *catedra OFr. chaiere *atrapica OFr. arrache buut(y)ru OFr. bure / burre <= buut(y)raariu OFr. burrier OFr. pere / perre *patriinu OFr. parrin patre maatre OFr. mere / merre *matriina OFr. marrine matriice OFr. marris matriculaariu OFr. marreglier OFr. veire / veirre <= vitraariu OFr. verrier vitru latro OFr. lere / lerre <= latroone OFr. larron latrociiniu OFr. larrecin petra OFr. pierre *petraariu OFr. perrier OFr. perron *petroone *petrosiiliu OFr. perresil foodr (frk) OFr. fuerre / fuere, foare *fodraaticu OFr. forrage retro OFr. riere ad+retro arrière OFr. toneire / toneirre quadraatu OFr. carré *tonitru *quadrifurcu OFr. carreforc ed(e)ra OFr. ierre / iere, eire, ere Theod(e)riicu OFr. Tierri lootr (frk) OFr. loire / luerre *Bit(u)riigu OFr. Berri rire / => rirra *riid(e)rat OFr. rirra riid(e)re claud(e)re OFr. clore / => clorra *claud(e)rat OFr. clorra creed(e)re OFr. creire / => crera *creed(e)rat OFr. crerra *nutriire OFr. norrir OFr. porrir *putriire 8 / arement / arace / parin / peresil / OFr ariere b. graphic ambiguity : Pope 1952 : §372, Fouché 1966 : 721-2 (rq.1), Bourciez 1971 §144-2 + H. c. despite wavering orthography due to further levelling between r and rr under analogical pressure and / or merging of r and rr into [r], it seems that i. tr/dr has become either r or rr ii. both outputs are in complementary distribution : TR > r in post-tonic position TR > rr in pre-tonic position latro > lere [Nom. sg] latroone > larron d. if this is correct, rr is to be analyzed as a compensatory lengthening following the normal loss of t. Such a compensatory lengthening implies that TR is bi-positional : l | C a | V t | C V r | C o | V n | C e | V e. remaining question : why no compensatory lengthening in latro, that is in postonic position ? *larro > laro : * l | C Lic a | V C V [stress] larrone : t | C Lic V r | C o | V Lic l | C a | V t | C V r | C => l | C a | V t C V C [stress] V r | C o | V Lic o | V C V [stress] n | C e | V f. => French : both TR and T=R available for TR E. Tentative Romance typology [26] Different configurations ? (1) : Italian a. intervocalic gemination of TR i. C+j (and C+l > Cj) : => CCj : rabia > rabbia, oc(u)lu > occhio, etc. ii. C+r : gemination of br : fabru > fabbro, labru > labbro, etc., (but not of other TRs — with or without voicing : capra, vetro, quadro, lacrima, padre, magro...) b. r-metathesis (cf. Rohlfs 1966 : I.§322) : in dialects, one landing site, i.e. word-initial T as in French, cf. [22.b.i] i. --> #T__ Rohlfs' comments • from post-consonantal TR comprare -> crompare (e. g. Elba) "tanto raro in testi letterari • from intervocalic TR capra -> crapa (e. g. Elba) quanto è frequente nei dialetti" • from R in Coda dormire -> dromire (Marche) "con una certa facilità" but : ii --> Coda • from word-initial TR fratello -> fardelo (Rovig.) "piú raro" • from intervocalic TR aprire -> arví (Lig, Romagn.) "neppure comune" iii. --> intervocalic T • from word-initial TR pratica -> pàtriga (OPadov.) "ancora piú raro" • from R in Coda erba -> evra (Nap. "dei secoli passati") "rarissimo" [Note : intervocalic R to word-initial T, e.g. craíno / carino (Elbano di Marciana) : "niente affatto comune".] 9 c. la r parassita (Rohlfs 1966 : I.§333) i. #T__ taverna traverna (Velletri) tesoro trisoru (Sicil.) tuono tròno (Umbr.) stella strela (Emil.) ii. C.T__ genista balista encaustu -mente ginestra balestra inchiostro -mentre (OLomb...) iii. V.T__ *anata vitice siepe *arbitu anatra vetrice scepre(Lucch.) árbatro (Pis.) d. thus, intervocalic T is a possible landing site for parasitic R. Compare with French [24b] where it is not. [27] Different configurations ? (2) : Gascon a. r-metathesis (Rohlfs 1935 : §399, Baldinger 1958 : 246, Dumenil 1987) i. from post-cons TR --> #T__ ii. from intervoc. TR --> #T__ compraare croumpa paup(e)re praube (ex)combraare escrumba capra crabo ventre brente cup(e)raare croba furunc(u)lu hloronc coop(e)riire crubi proubo *tuuf(e)ra truho pulv(e)re fenestra hrieste *wiip(e)ra gripo capistru crabeste febre hrebe heure castroone creston fabru haure brespes tab(u)la taulo vesp(e)ras ep ten(e)ru trende patriinu payri ep cam(e)ra crambo latraare layra ep spiin(u)la esplingo ep mem(o)raare bremba ep pess(u)lu plesc ep cing(u)la sinclo ep gen(e)ru jendre ep num(e)roosi numbrubi hirund(u)la hurounglo iii. from R in Coda --> #T__ dormiire droumi firmu hrém forma arroumo formiica arroumigo [shadowed cells : no R-metathesis] b. converging movement towards #T__ c. note that TR clusters resulting from T-epenthesis (indicated by ep in i.) undergo R-metathesis as well. d. intervocalic TR : either R-metathesis or (when initial C ≠ T) T•R maintained with consecutive lenition of T (shadowed cells in a.ii) [28] Tentative interpretation : a. Early Romance : * T•R => TR is either TR or T=R b. different distributions according to the position STRONG #__ C.__ TR T=R TR T=R French Italian Gascon ? ? *? ? ? ? (TR) * * * WEAK V__V TR T=R * * * * i. French : TR in strong position, T=R in weak position ii. Italian : only TR in weak and strong (post-consonantal) positions iii. Gascon : neither TR not T=R in weak position, avoidance of TR in strong post-consonantal position. Tendancy to "no-TR" ? Note frequent vowel epenthesis in initial TR : branc > baranc (Rohlfs 1935 : §401) : strong evidence for initial TR being T=R and for a tendency to eliminate this structure... 10 [29] F. Conclusion a. there is no way out of the Gallo-Romance trouble without acknowledging a mono-positional "affricate" TR. It goes without saying that the affricate TR does not solve all problems. However, it allows for a fresh look at known data through a new prism. b. for example, epenthesis (cam(e)ra > chambre) and metathesis can be interpreted as positional phenomena that imply lenition and fortition. c. "affricate" TR call for a principled management of syllabic space : nothing falls from heaven. d. the early generative conception whereby syllabification is dynamic and operates over a lexically unsyllabified string does not make sense : the identical phonetic object [TR] enjoys two different syllabic interpretations. The back-bone of this approach, the Maximal Onset principle, is but the expression of a tendency and/ or markedness: most, but not all TR clusters are branching Onsets. When they are Coda-Onset sequences, Codacapture is applied. One could surely think of some transformation that makes them mono-positional structures as well, but what it this good for? TR clusters are not anything per se or in Universal Grammar. They enjoy different syllabic interpretations in various languages, and these are recorded in the lexicon. References Ali, F. (2003), Les positions fortes et faibles des consonnes en malgache, DEA thesis, Université de Nice. Baldinger, K. (1958), "La position du Gascon entre la Galloromania et l'Ibéroromnia", Revue de Linguistique Romane 22 : 241-289. Bullock, B. E. (2001), "Double prosody and stress shift in Proto-Romance", Probus 13 : 173-192. Carvalho, J. Brandão de (in press), " Templatic morphology in the Portuguese verb", in Nouveaux départs en phonologie. Actes du XVII. Deutscher Romanistentag (Munich, 7-10 octobre 2001) (T. Meisenburg & M. Selig éds), Tübingen (Gunter Narr). Dumenil, A. (1987), "A rule-accent of metathesis in Gascon", Linguisticae Investigationes 11,1 : 81-113. Groot, A. W. de (1921), Die Anaptyxe im Lateinischen, Göttingen (Vandenhoeck & Ruprecht). Haas, M. R. (1940), "Ablaut and its function in Muskogee", Language 16,2 : 141-150. Hirst, D. (1985), "Linearisation and the Single-Segment Hypothesis", in Grammatical Representation (Jacqueline Guéron, Hans Obenauer & Jean-Yves Pollock eds) : 87-100, Dordrecht (Foris). Jacobs, H. (1992), "Evolution of French foot and syllable structure in the evolution from Classical Latin to Old French", in Theoretical Analyses in Romance Linguistics (Ch. Laeufer & T. A. Morgan eds) : 55-79, Amsterdam / Philadelphia (Benjamins). Jacobs, H. (1993). La palatalisation gallo-romane et la représentation des traits distinctifs, in Architecture des représentations phonologiques (Bernard Laks et Annie Rialland éds), Paris : CNRS Editions. Lahiri, A., T. Riad & H. Jacobs (1999), "Diachronic Prosody", in Word Prosodic Systems in the Languages of Europe (H. van der Hulst ed.) : 335-401, Berlin / New York (Mouton de Gruyter). Loporcaro, M. (to appear), "La sillabazione di Muta cum liquida dal latino al romanzo". Lowenstamm, J. (1996), "CV as the only syllable type", in Current Trends in Phonology (Durand, J. & B. Laks, eds), vol. 2: 419-441, Salford / Manchester (ESRI). Lowenstamm, J. (1999), "The beginning of the word", in Syllables ?! (Rennison, J. & K.Kühnhammer eds) : 153-166, La Haye (Holland Academic Graphics). Lowenstamm, J. (2003), Remarks on Mutae cum Liquida and Branching Onset, [ms] University Paris 7. Rennison, J. (1998), "Contour segments without subsegmental structure", in Structure and Interpretation. Studies in Phonology (E. Cyran ed.), Lublin (Wydawnictwo Folium). Rohlfs, G. (1935), Le Gascon. Etudes de philologie pyrénéenne, Halle (Niemayer). Rohlfs, G. (1966), Grammatica Storica della lingua italiana e dei suoi dialetti, vol. 1 Fonetica, Turin (Einaudi). Scheer, T. & Ph. Ségéral (2001), "Les séquences consonne + yod en gallo-roman", Recherches Linguistiques de Vincennes 30 : 87-120. Scheer, T. (1998), "Governing domains are head-final. Structure and Interpretation", in Studies in Phonology (E. Cyran ed.) : 261-285, Lublin (Folium). Downloadable at http://www.unice.fr/dsl/tobias.htm Scheer, T. (1999), "A theory of consonantal interaction", Folia Linguistica 32, 201:237. Downloadable at http://www.unice.fr/dsl/tobias.htm Ségéral, Ph. & T. Scheer (2001), "La Coda-Miroir", Bulletin de la Société de Linguistique de Paris 96,1 : 107-152. [Older English version downloadable at http://www.unice.fr/dsl/tobias.htm Steriade, D. (1988), "Gemination and the Proto-Romance Syllable Schift", in Advances in Romance Linguistic (David Birdsong & Jean-Pierre Montreuil eds) : 371-409, Dordrecht (Foris). Steriade, D. (1994), "Complex Onsets as Single Segments : The Mazateco Pattern", in Perspectives in Phonology (Jennifer Cole & Charles Kisseberth eds) : 203-291, Stanford (CSLI). Timparano, S. (1965), "Muta cum liquida in poesia latina e nel latino volgare", Rivista di cultura classica e mediovale 7 : 1075-1104. Ward, R. L. (1951), "Stop plus liquid and the position of Latin accent", Language 27 : 477-484. 11