CAPÍTULO 5. DISEÑO DEL ESTUDIO EMPÍRICO
5.1. INTRODUCCIÓN
In section 3.3.4 we saw that monosyllabic lengthening provides us with an independent method of determining coda extrasyllabicity in languages that have a Minimal Word Requirement. Depending on one’s theoretical view of reduplication, there may also exist an independent test of onset extrasyllabicity: the analysis of a language’s
redupli-46Perhaps the presence of *u in these protoforms – being inherently shorter and less sonorous than
*a – is a factor here?
47Continued by Dardic, Pashai (dialectal) ¯us./ux ‘long’ (Turner 1966:74, Nr. 1627). Sanskrit *us.t.a-, the expected reflex of *h2uksto-, has been replaced by uks.ita- (NIL 354-5).
48OHG f¯ust, OE f¯yst, OBulg. pe˛stž (IEW 839).
49Lat. sextus (←*sektos < *sekstos), Gk. héktos, Ved. s.as.t.há-, Goth. saihsta, Lith. š˜eštas, TB s.kaste (Weiss 2009:293).
cation patterns.50 According to this particular theoretical view, if a certain consonant occurs in absolute root-initial position and participates normally in the reduplication process, we may say that that consonant is syllabifiable in that language. Conversely, if a particular root-initial consonant is not copied into the reduplicant then we may say that that consonant is extrasyllabic.
For example, Sanskrit roots with no extrasyllabic segments in the onset redupli-cate the initial consonant (pr¯a- ‘fill’ → pi-pra-) while those roots with extrasyllabic consonants in the onset reduplicate the second (sth¯a- ‘stand’ → tí-s.t.hati ‘stands’ not
*sí-s.t.hati).51 As Steriade (ibid.) shows, only those roots with an SSP violation in the onset behave differently from the normal pattern of reduplication, since those violating consonants are extrasyllabic. In (Attic) Greek, reduplication is blocked for onsets with extrasyllabic consonants: the perfect of gn¯o ‘know’ is é-gn¯oka, not *gé-gn¯oka, since
#gn- contains an extrasyllabic g that blocks reduplication.52
Thus analyzing how reduplication worked in PIE may be able to tell us in a non-circular fashion whether there were extrasyllabic consonants in word-initial position and exactly which consonants could function as extrasyllabic. Fortunately for our study, reduplication is a common morphological process in Proto-Indo-European. It is found in many verbal categories (reduplicated presents, aorists, desideratives and perfects) and very rarely in nominal formations, in such forms as *kwe-kwlós ‘wheel’
(Skt cakrá-, Gk. kúklos, OE hw¯eol, etc.).
50See Steriade 1982:312ff., Keydana 2006, Keydana, forthcoming a. An alternate view is given by Carlson 1997 (see also Morelli 1999, Cho & King 2003 and Vaux & Wolfe 2009) who points out that the choice of consonant in reduplicative templates is rather due to reduplicant-specific markedness constraints and therefore nothing may inferred about the underlying syllabic structure.
51Steriade 1982:312ff., Kobayashi 2004:43, Keydana 2006:95-7.
52Steriade (1982:221) argues that in Greek adjacent tautosyllabic consonants must be at least four intervals apart on the sonority scale (see top of p. 221). The reason g is extrasyllabic in the onset #gn-is because there #gn-is only a sonority d#gn-istance of three intervals in th#gn-is cluster.
For the most part reduplication is a well-understood morphological process in PIE.53 There are two reconstructable types of reduplication, partial and full, with the latter utilized only in intensive/iterative formations, such as Skt. dár-darti ‘pry open’
to dar- ‘split’ and Gk. por-ph´¯ur¯o ‘swell, surge (of the sea)’ to an earlier *phur-.54 All other morphological categories reduplicated partially. Some reduplicated with e-vocalism (perfect, aorist, certain presents, and nominal formations), while others showed i-vocalism (certain presents and desiderative).
Roots with a single consonant in the onset simply reduplicated that particular consonant, followed by a vowel: *gwem- ‘come’ → *gwe-gwóm-e ‘came’ (> Skt.
jag´¯ama); *bher- ‘bear’ →*bhi-bhérti ‘carries’ (> Skt. bibhárti). Roots with two con-sonants in the onset reduplicated with the first: *drem- ‘run’ → *de-drom- (> Hom.
Gk. ana-dédrome ‘towered’); *mnah2- ‘remember’→*me-mnóh2-/me-mn
˚h2- (> Gk.
mémn¯emai ‘I remember’); *smei
“- ‘smile’ → *se-sm(o)i
“- (> Ved. sis.miy¯an.a- ‘smil-ing’), *h2ne“k- ‘reach’ → *h2a-h2no“k- (> Skt. ¯anám.´sa, OIr. -ánaic), etc.55 Based on these facts, the reduplication template for PIE may be reconstructed as *C1V-C1 (C)V-for morphological categories with partial reduplication. Complications arise, however, in the reconstruction of reduplicants for roots whose onsets consisted of /s/ + stop and /H/ + stop, namely, those roots where the first consonant of a bipartite onset was a fricative and violated the SSP.56
Let us first address the reduplication template for roots of the shape *sP-. Here, the uncertainty for reconstruction lies in the many different types of reduplicants attested in the IE languages. As we saw, Sanskrit shows PVsP- (cf. tí-s.t.h¯ati ‘stands’ to sth¯a-)
53For excellent and thorough discussions of the process of reduplication in PIE, see Brugmann 1897-1916:20-41, Niepokuj 1997 and Keydana 2006.
54Cf. Hitt parah
ˇzi ‘chases’ < *bherh2- ‘move quickly’.
55Kümmel 2000:287.
56See Keydana 2006.
while its most closely related branch, Iranian, attests to sVsP- (Av. hi-štaiti, vi-ša-star@, OPers. a-hi-štat¯a, all to st¯a- ‘stand’), as does Celtic (cf. OIr. se-scaind ‘jumped’
< *se-skond-e). In non-productive reduplicated i-presents in Greek, we find sVsP-(cf. hí-st¯emi),57 while productive perfect formations block reduplication altogether (e-strató¯onto ‘was on the battlefield’, with simple prefixed e-). In Latin, the highly archaic and non-productive reduplicated present si-st¯o (= Umbr. se-stu) attests to sVsP-, while later productive reduplicants to roots of the shape sPV- exhibit a curious process of reduplicative infixation: st¯a- ‘stand’→ ste-t-; spond- ‘libate’→ spo-po-nd-, etc.58 In Gothic the reduplicative template is always sPVsP- (stai-stald to stald-, etc.),59 while in Tocharian, we find three reduplicant shapes attested: in A, sV- (sp¯artw(¯a)- ‘turn, behave, be’→s¯a-spärtwu), and in B, PV- (sp¯artt(¯a)- ‘turn, behave, be’→pa-sparttau) and sPV- (stäm(¯a)- ‘stand’→´sce-´scamos.; staukk(¯a)- ‘swell’→sta-staukkauwa).60
Many scholars61have viewed this disparity among reduplicants in the IE languages as proof that the reduplication template for roots of the shape *sP- was *sPVsP-, ex-actly as is attested in Gothic. The basic idea is that those languages whose reduplicants have a single consonant (*s or *P) have simplified the original, more complex redupli-cant beginning in *sP- through a process of dissimilation: PIE *sti-stah2-ti ‘stands’ >
*si-stah2-ti or *ti-stah2-ti.62 However, as early as Brugmann,63it has been recognized that since Lat. sist¯o and Gk. híst¯emi do not follow their respective synchronic pattern
57As well as the archaic hé-st¯eka ‘I stood’.
58Fleischhacker 2002:5. See Keydana, forthcoming a for an alternative analysis.
59I follow Keydana 2006 and forthcoming a, who assumes a subsegmental status of /s/ in /sP/ clusters in Gothic (suffricates), which easily accounts for the entire copy of the initial cluster of the root.
60Forms taken from Malzahn 2009:964, 963, 959-60 and 990, respectively.
61For instance, Szemerényi 1970:249 and LIV 590.
62Keydana 2006:81 convincingly dispels this notion, demonstrating that dissimilation cannot produce the attested reduplicants as long as reduplication is still an active morphological process.
631897-1916:40-1.
of reduplication, they must be archaisms derived from an older reduplicative template
*sVsP-. It is simplest to assume that both forms go back to a common archaism, and that the reduplication pattern found in Avestan and Old Irish continues the original state of affairs.
Those scholars who reconstruct a PIE *sti-stah2-ti ‘stands’ typically see confirma-tion in their reconstrucconfirma-tion of the reduplicant to *h1ger- ‘awaken’, the only root of the shape *HP- with an attested reduplicated form (Kümmel 2000:191-4). There are three attested reflexes of the perfect to *h1ger- ‘awaken’: Skt. j¯ag´¯ara ‘wakes, is awake’, Av.
jaG¯ara ‘is awake’ and Gk. egr´¯egore ‘is awake’ (LIV 245-6). The Avestan form must be secondary, having eliminated the lengthened grade in the reduplicant by analogy to other perfects. The Greek and Sanskrit forms, however, appear to derive from an original *h1ge-h1gor-e ‘woke up’, with the [r] in the Greek reduplicant secondary by analogy to the aorist égreto (LIV 245).
However, as Keydana (2006:104-5) demonstrates, the Sanskrit form is more easily explained as deriving from *g¯egor- < *ge-Hgor-, following the reduplication tem-plate *PVHP-, analogous to the *PVsP- reduplication temtem-plate discussed above. Just as PVsP- replaced *sVsP- as the reduplication template to roots of the shape sP- in the prehistory of Sanskrit, so may we posit that *PVHP- replaced an earlier template
*HVHP- to roots of the shape *HP-. Similarly, Greek egr¯egore may be viewed as a later formation, remade from original *¯egore (< *h1e-h1góre) and re-marked as a reduplicated form through the process of Attic reduplication (cf. Rix 1992:204-5), as has occurred, for example, in the perfect of *h1ne“k- ‘take’: *h1e-h1no“k- > *¯eno“k-64→ Att. en-¯enokha ‘has carried’ (LIV 250). Likewise, *h1e-h1góre > *¯egore→*eg-¯egore, with the -r- secondarily inserted into the reduplicant.
Note also that if the reduplicant to *h1ger- ‘awaken’ had consisted of two
conso-64cf. Skt. ¯anám.´sa, OIr. -ánaic < *h2a-h2
no“k-nants in the onset, it would be the sole example of such a reduplication type recon-structed for PIE, since, as we saw, roots of the shape *sP- reduplicated with a simple
*sV-. This makes it increasingly likely that a single reduplication template existed in PIE: *C1V-C1(C)V-. In short, PIE reduplication proceeded in the same way for all bipartite onsets: *pleh1-‘fill’ → *pi-pleh1- (Gk. pímpl¯an¯o ‘fill’), *stah2- ‘stand’
→ *si-stah2- (Lat. sist¯o, Gk. híst¯emi, etc.), *u