This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
P > t) ko-to-(i)-na /ktoyni/ "plot of land" (KTolval kto nai Hesychius) tekotone Itektonesl TEKTOVEC; "carpenters" (H. 6.3 15+) ekoto ( tuwp Htflcl ar H. 1.242+) ponikipi Iphoynik n) aminiso lamnis6s1 ' A� YlO Or;
nasal + cont. + stop
teqade Itbegwc1ns+de/ "to 'J'hegWai" cf. teqaja It�gWayyail "Theban" (NOMlg F) nasal + nasal (n > m)
"Amnisos"
IIIQ1IQSiweko Imnasfwergos/ MJI'I)U{EPYor; "Mnesiergos" nasal + liquid o-mi-ri-jo-i lomrf(y)oyhil "to the
Rain-spirits" (=' Op.�p(otc; Ombrfois; cf. ZEUc: OI1�Ploc; ai4s ombrios "Zeus sender of rain") ; note that anaptyctic b is not present, or the fonn would be *o-bi-ri-jo-i lomhrfyoyhi/, since br- is a possible onset by the SH.
stop + liquid tarlJllJl ltbIinusl "beam; footstool " (loo. 6pi'iVIK thr � Od. 19.57+) d1ll1l1mno Idrut6moyl BPUT6jl Ol "wood-cutters" (11. 1 1.86+)
aktrese lagresel "took" or lagr�yl "will take" (dypElI) agre6 "take" preserved in AeoIic)
qirijalo /kwrl(y)atol "bought" (1fPlaTOPrfato Od. 1.430)
liquid + nasal
a-TnO lc1rhmol "wheel" (apl1a
hdrma "chariot" 11. 5.23 1+) e-ma-a2 lhermihAy/ "to Hermes" ( Epjli'jc; Hermes Od. 5. 54+-; Epic NOM ' EPI1Elac;' Herme(� Od. 1.42+) keniqa IkhemigW-1 (ACC?) "vessel for washing hands" (cf. XEpy� ov /fftemibon 11. 24.304) '
liquid + stop
larthm6s1 "fellowship, league" (h. Mere. 524+) wodowe Iword6+wenl "rose-scent ed" (of oil; POSOEYTl... l).a(lI} rhodOenti...elaifi "(with) rose scented olive-oil" 11. 23. 186) wekata Iwergatay/ "workers" (of oxen; = �oijc; lpyd'TIlc;' bo m ergtit� Archil., Soph.) aIOmo
21
UNBAR B SYlLABARY etewok.ereweijo /etewoklew�hiosl "son of Eteocles" (cf. �(1) ' ETEO" ICA1)El1) hf� Eteoldeefe n. 4.386) dereuko /dl�wkos/ "sweet new wine" ( y�elilCo&; gle iI«Js Arist Meteor. 3801>:32+) erutara /eruthri/ "red" (NOMsgF) (epv9p6t; eruth� n. 19.38+)
akomawo /Alkmawos/; cf.' A�JCl.l(:iCl) Y AJlcmjfiJ (*AlkmlMloo) n. 12.394 koko /khaIk6s/ xaAICcX "bronze" akuro /argdroy/ apyUp� "w. silver" atemito /wmitos/ ApTEll lToc; "of Artemis" poUpi lp6rti+phi/ "with heifers" (nopTK ptJrtis "calf' n. 5. 162+)
stop + stop + liquid
liquid + stop + obstruent poroek.eterija Ipro=helkterf(Y)i/ "instrument for drawing forth" akosano /alksinor/ (cf AAX01}VWP AJkhsiniir DOE 761: Naxos, VI)
arekuturuwo /alektru(w) on! AAEK. TPUtilv Alektru 6n n. 17.602 reukotoro /I�wktron/ AelinpoY (capital of Further Province) pekitira2 /pektOay/ "woolcarders" ra-pi-ti-ra2 !Jliptriay/ "sewing women" (fIC:i1TTPW rhdptria) '
•
. •
nasal + glide
glide + nasal
pe-ru-si-nu-wa, pe-ru-si-nwa /perusinwli! " last year's" (Npl) (1TepoolvOc:; perusirWs Aristophanes+) k.eseniwijo /ks�nwion! "guest-type" (of oil) ; alternate spelling kesenuwija /ksc!nwial (Npl) (of textiles); rare variant spelling k.esenewija /ks�nwial (ee(vw ksefnia Hom.+)
a-pu k.e-lca-u-me-no /apu= kekawm�nos! "bumt away" (PPP of ano�(CI) apo1caf6) (This exception vanishes with the consideration that the u diphthong is always written; also [aw.m] is not the same as [awm]; see below.)
stop + gUde
glide + stop ka-ra-u-ko /gl'wkos/ rMiJlCoc;
tetukowoa(2) /t(h) et(h)uki1w6ha1 "finished" (PP of TEUXW teUkh� ;cf. Te TEuxwc; teteukhtis Od. 12.423) widowoijo, wi-dwo-i-jo /widw6hios/ Name: cf. partic. el8ti1c; eidtr. (Epic fern. l8uta idu il) "knowing" ( *widw�)
GIa flws 11. 2.876+ (Recall that u-diphthongs are always written and that [aw.k] is not the same as [awkJ.)
22
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWlEDGE
glide + glide (w+y) diwijaldi-u-ja Idiwylsl "of Diwia" (fern. deity) mewijolmeujo Imeywy6s1 "smaller; younger" ( Il dlll v me(fJn Hom.+) a-ra-ru-wo-jala-ra-ru-ja laranlwyal "fitted" (fern. of PAP · [Epic apllPIllc; arc}"�] apapula araruia n. 15.737+
glide + glide (y+w)
glide + continuant
continuant + glide wisowopana Iwiswo-I "equal" (Cret., Arc. F(oFoc; wiSwos, Epic to oc: 1;os (Hom.+), Alt. "100(' !sos) asiwijo laswi(y)osl ( A010C; ...\sios); cf. asiwija lasw{iyl "Lydian" (DAT sg) (11. 1 1 .461 Ao(� k(oi ; cf. Hilt. ASSUwa-) "
•
allele + liquid wirino Iwnnoyl "ox-hides" (plvOc; rh�s Od. 1. 1 n)
memanamenoi ImemnAm�noyl "re
nasal + continuant weretase probably Iw�tasl and not Iwretansl (ACCpiF)4 "pacts" [217. 28, 29] ; cf. Att. (nlTpa rh 4,.jJ, Ion. P�TPTl rh .ir�(Od. 14.393+)
nasal + nasal (n > m)
membering" (pf. medpass. part , NOMplM = 11 Ell VllI1 EVOl) [261]
nasal + liquid
liquid + nasal kumerenai Ikum�mahil "guide" (3pl pres.) [264.4: Neumann 1974] (formerly read kumer iiUll)
4 Arcado-Cyprian apparently did not have the Sec:ond CompensalOry Lengthening (-Vns- -+ -V:s-); cf. Arc. lI'cXJIO'UI; pdnsas [DGE 66S A12] "all", o+��JIUt ophll/omi [DOE 665 A23] "(they) owe" (see Wyalt 1973:41); -ns- remains except in anal syllables: Axe. T� r; tds « *tom: Att.-Ion. TOUr; toUs) "the" (ACCpIM); wcivO'ar; pdn.r "all" (ACCplF *pan(t)san.s: Alt.-Ion. lI'CiO'ext; plsftr); see Dubois (1986 §38ff). Cyprian i ambiguous, but, based on the Axcadian evidence, it is aeneraUy agreed that to-se is Itosl than IOOns/, ItiJsl, or �y other of the theoretical possibilities (cf. Bechtel GD 1 .416; Thumb-Scherer, p.161; Scbmilt 1977:94). Viredaz (1983: 186, 206) signals Paphian spellings iike -(JUS (*-ans) as evidence of nasalization or nasaI Cs without closure. Whatever the details, retained - Vns is most unlikely.
CYPRIAN SYU.ABARY
31
paramenone "Pannenon" [154.2] aramaneuse larmc1news/ "(son) of Armanes" (see ad ICS 217.21 ) autarami lautar mi/ "but me" [235]
stop + liquid pilokuporone Iphilolo1pr6n1 "of Philokypros" [217. 1 ] tamatiri ldimitril "to Demeter" [182] turumione Idrumionl [217.19] .6.pUp. LOC: D,.,Jmios ( ACC) "Copse Stream" etewatoro letewc1ndrOl "of Etewan dros" [ 176] nikokelewese Inikoklewesl "Niko kl(ew)es" [6. 1 , 7.2+] [exc.: ekerato lekhratol "used" (1) [306.2] (other readings possible)) continuant + liquid
liquid + stop talalone ldaltonl "(writing-)tablet" [217.26] (SEATOC: diltos Batr. 3+) arakuro largdrO/ "of silver" [217.6+] kateworokone lkateworgonl "be sieged" (3pl . aor. ) [217.1] (see Masson 1983:265-266) sunorokoise Isun (h)6rkoysl "with oaths" [217.28] (OPICOc;' hOrkos Hom.+)
esolo leslo-/ "Eslo-" [1Sa, 327+ va riants in Viredaz 1983: 193]
liquid + continuant pereseutai lpersewtiyl "Perseutas" (DAT) [ 181.3] (hero: 2 alph. texts)
nasal + glide
glide + nasal
[no examples?]
glide + Uquid [like linear B)
weretase Iwretas/ "agreements" (Accpl) [217.28. 29] e(u)weretasatu le+wretAsatol "agreed, contracted" [217.4, 14] (with variable resyllabi fication [ew.re.ti.sa.tu]6) zowara[- /dlOwn1[1ios]/ Name (ZwF paALoc: ZfMJrdlios) [327.4 (Bulwer Tablet)]
liquid + glide [cf. Linear B) alawo- /fJlwo-/ "vineyard" (Acc?) S arawasatu /arwasatol "prayed" [3438 4] (dpaOll a L ar hnai Hom/poet) [ambiguous as to onset or -xxla treatment, but -rw- codas are to be expected] puruwoso /pl1rwo-so/ "Pyrwos" [ 1 98 (2x)] (but probably Eteocypriot)
5 The problems with Ibis word (3x in Idal. Bronze: lines 9, 18, 21) involve its Case (GFN or ACC) and its derivational morphology (see Masson, ad loc.); the rust three segments of the word, which are all that concern us here, are Dot in dispute (pace Beekes 1971:350-352). 6 Compare Buck 1955 §§ 55, 70.3, and Cypr. a-TO-U-Ta (e.g. , ldal. Bronze 20) "land"; note the consistent spelling from Mycenaean (§2.4, last entry) to the Classical period (clpoupa:).
ANCIFNT SCRIPrS AND PHONOLOGICAL KNOWLEDGE
32 stop + glide
[no examples?]
glide + stop
glide + glide
[no examples?]
glide + gUde
continuant + glide
[no examples?]
glide + continuant
So far, precisely the same strings are treated as onset or coda as in the Unear B syllabary. Especially noteworthy is the consistent exception to the SH involving glide and liquid sequences, confirming that something general (non-idiosyncratic) is at issue here.
3.4 Countercoll)lentioll' : The Writing o/Stop.conJillualll String. As noted in §3. 1 (Ca), there are two ways of writing continuant-stop clusters. The same is true of certain strings involving stop plus continuant. Consider first the status of [ks]. (3) [ks] as Complex Sign (a) e-J:e leksl "from" (te( EK ) [217.6. 7. 24 (2x). etc.] (b) e-J:e l o-ru-xe leks=oniksel "banishes" (aor. subj. 3 sg. of le Opt'W ex-or(z 0); cf. All. ttoptCJ'IJ exor{sei (Thumb-Scherer, 135, 157, 160. 170; Cowgill l964:358; Masson, ad Idal. Bronze 12) (c) wa-na-;re /wWiaks/ "lord" (ava� dnax) [211.1. 220.2. 264. 1+] (d) e-we-re-xa /ewerksa/ "I did" (aor. act. 1 sg. of F E P Y- = bd fJ Hom.+ Ion., poet) [261] (=Pfohl 1966. # 1 1: Golgoi. -V) (e) ka-ru-:xe-e-mi = "ap� 1'Jj.l ( karux emf "I am Karyx" (digraphic stele [260]: Golgoi, mid -VI) = Alt.-Ion. td,\put kfTux "herald"? (4) Normal Onset Treatment of [ks] (a) e-u-ka-sa-me-no-se lew.ksa.me.nosl E\Jtaj.lEvoc:; "praying" (aor. mid. part NOMSgM of EUXOJ,laL eJikhomai) [ 181.2] (b) to-ka-sa-to-ro Idoksandrol "of Doxandros" [l68a] (c) ta-pi-te-ki-si-o-i It'amphideksfoy/ "Amphidexios (DATSg) [335.2]
� Onl�
Since complex cluster signs in Mycenaean (§2. 1) are for onsets and alternate with regular conventions for spelling onsets, assuming con� n�it)'.. (3) and (4) are both onset treatments. It is then clear from these core examples that [ks] is treated as an onset, as the SH predicts. Why then are two different
7 The authority OIl the accentuation is Herodian (Gr. 1.44. etc.). On clitics with such words. SCXDJDeIBtein (1973: 176ff) ; Scbrijver (1991 : 95. 1 12. 128. 219) cites the form as tUrI«.
see
CYPRIAN SYlLABARY
33
graphic representations used? xV and kVsV should be equivalent alternative spellings. but the xV signs are normally reservc;d for word-final position. Why should that be the case? I submit that this was originally another ingenious way of resolving the dilemma of what to do with theoretically possible onsets that happen to be in coda position (cf. §2.6): those that are in coda position were written with the xV series, which simultaneously insists on their onset value; those in onset positions are written by the usual convention for onsets. Another stop + sibilant string involves [ps] in o-pi-si-si-ke f6psis kef "whoever" (= Alt. oone; av Mslis an) in the ldalium Bronze [2 17.29].8 By the SH. [ps] is a possible onset, and the [sk] in the same construct is a coda. Both are correctly written. 3.5 Morphological Input and Boundary Phenomena The one (verifiable) exception is e-ke-so-si /(h)eksonsif "will have" (3pl) in the ldalium Bronze [2 1 7.3 1], where a coda spelling is found. Since the root is (h)ekh- "have", morphological spelling (cf. Guion 1994) is more likely than a difference in syllabification proposed by Viredaz ( 1983: 188), unless such a resyllabification could itself be motivated by the presence of the morpheme boundary. A similar case of compositional spelling is directly verifiable in (3b), where a word divider separates the particle and the verb. Given the * normal Greek treatment of preverbs. one might expect ( )e-Ico-so-ru-xe 8 For the variety of interpretations of this construct. see Masson (ad ICS 217.29). The reading adopted hen: is that of Thumb-Scherer (1959: lOO, 160, 168. 174). The meaning is clear. Like all curses. it must begin "whenever someone or whoever breaks this law, on him [.. .]". The phrase therefore contains Cypr. sis (= Att-Ion. TtI;' lis ) "someone" and modal ke. But what is op(ip. Even if it were unequivocally established that Myc. opi can mean "when" (d. TheS8. inlE( KE 0 I [K ]cnpOl;' KClTEvtKEl opeC ke ho I [kJairos katenikei [Buck 33.26-27: Larissa. 11] "when the time arrives"), the opi- adopted by Wathelet (1970:84 w. lit.) and Viredaz (1983: 191) is not without problems. since the Idalium Bronze uses (h)ole (STE) for "when", as does Mycenaean (and most of the rest of Greek). If the curse were to begin "whenever someone [... ]", we might then expect *(h)Ote-sis-ke (or the like). One more-or less expects the curse to begin "whoever [. ..)" (Att. lSaTlI;" MSlis). Various dialects have a -It form (e.g., Lesbian �TTLC; dltis) generalized from the neuter �TTl dlli (Sappho. etc.). Hom. (hTt hOtli < *yod-kWitJ (Wackemagel l885:89ff; Jacobsohn 1910: 1 14-124; cf. Bechtel l .78); d. Myc. jo-qi (PY 318) ly6kWJtwil "which". This is the formation posited by Thumb-Scberer (p. 160) to underlie Cypr. 16p-sis-kel « *h6d kWis ke). comparable to Att. �aTll;" l£v MS-lis dn (cf. Beekes 1971 :341). Masson (ICS. ad loc.) disputes this on the grounds that it is "phonetically implausible". lt is true that *1u5d-kwis > *1u5kw./cwis > *(h)OkW -ISis > (h)dp-sis is difficult to molivate. On the other band, a phonelic development *Mkw-kwis > *(h)op-pis. with replacement of isolated *pis by regular (productive) sis is completely plausible. Also, paradigmatic crossing may have played a role, viz. common gender (*ho-kwis » *(h)d-sis beside neuter (*M/cW ./cWi(d) » (h)dppi provided more motivation for generalization of (h)op- and si(s) . whence c. (h)dp-sis , neut. (h)dp-si.
34
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWlEDGE
leksorUksel, or, more accurately,
(*) e-ko-so-ru-ke-se ,
since the final
e
is the
subjunctive marker rather than a dummy vowel. But even with clitics the signs served to mark word boundaries, as in where the choice of
xe
ka-ru-xe-e-mi [Karux�em{]
xV
(3e),
indicates word boundary and lack of resyllabification.
As for the second [ks] of e-xe I
o-ru-xe
(3b), the aorist stem is oruk-s-, and the
spelling may be morphological or by generalization of the frequent sign
-xe
with dummy vowel. Both doubtless operated together motivating spellings like
xV
e-we-re-xa
(3d), in which the aorist stem is e+werk-s-. In summary, the
signs were originally used to spell a theoretical onset cluster [ks] in an
absolute coda position, i.e., word-finally, whence they were generalized, first
for morphological coding, then as a graphic convention, as word-final signs, regardless of whether they represented onsets or codas (ef. Masson, ad I es
§44. 1 , citing one exception: me-te-xe-i [ Ub.6] I1 EeUEL "will share"?). 3.6. ContinlUlnt.Stop Sequences Recall that [sC] appears to be treated indifferently as onset or coda. (5) [sC] as Onset (a)
se-pe-o-se
Ispeosl "cavern" ( GENsg) [2.2. 3.2] (01TEOc; speos
primarily Homer and Epic; on the paradigm, cf. Miller 1982, §83)
(b) se-pe-re-ma Isperma/ CJ1TEPl1a "seed" [231] (c) sa-ta-si-wo-i-ko( -ne) Istisiwoyko(n)1 "of Stasiwoikos" [ 165.2-3+] (d) sa-ta-si-ku-po-ro-se /staslkuprosl "Stasikypros" [217.2] (e) (to-)sa-ta-si-wo-se l(t6) st isiwosl "Stasis" (GENsg) [ 15.1-2, 165. 1] (t) so-to-ro-pi-ki /str6p!liogil "in the pivot" (?) [229-231] = oTP04>t}f, O TPOq:,l'Y'YOc; strophigx, GEN strop higgos (Euripid.+), unless to be read as storphitBi (see Viredaz 1983: 191); in either event, the st remains treated as an onset. � � as � (a) a-ri-si-to-se laristos/ "Aristos" [ 102] (b) a-ri-si-to-ko-ne laristOkhon/ "Aristokhos" (GENsg) [ 181.1] (c)
)
a-ri-si-to-ke-le-we-i laristokl6weyl "Aristoklewes" (DAT) [352.4] (d) mi-si-to-ne Imisthonl "payment" (GEN) [3x in ldal. B ronze: 217.4. 5, 15] (= I1tOe6< mist hos "recompense" n. 10.304+) (e) ko-ra-sa-to-se Ikorastas/ "richly" [264.2] (hapax: Neumann 1974) (I) e-pi-si-ta-i-se lepfstahisl "care, attention" (h(o Taotc; ep(stasis) [264.3] (Neumann 1974)
(g) e-se-ta-se
16stase/ "stood ; erected" [1 18. 163.2] (= Att.-Ion. EOTllOE
estese) and related forms:
CYPRIAN SYlLABARY
35
1) e-se-ta-sa-ne I�stasanl "set up; erected" (3pl) [261] ( = Att. Ion. E(JTl\(Jav estesan) 2) e-pe-se-ta-se lepestfAsel "erected" [ 103+] (freq.) 3) ka-te-(e-)se-ta-se lka�stasel "set up; erected" (to ..:aOl(J·f1UJ.l kathfstem) [6.2+] (freq.) (h) a-ku-we-u-su-ti-ri-yo "Alruweustn(y)o"? [327.1 1 (Bulwer Tablet) ; see Viredaz 1983: 188, 194 w. lit]
It should be clear from (5) and (6) that the treatment of [sC] as onset or coda is in no way haphazard. Word-internally, the coda treatment is regular, as predicted by the SH.9 The onset treatment is regular only in word-initial position. Mirror-image to the problem of theoretical onsets in coda position (above), these cases involve SH codas in onset position (cf. Ouion 1994, with 'extrasyllabic' s). What does one do with a coda cluster in onset position? It could be treated as a coda and ignored, as in Pylian Linear B tu-ru-pte-ri-ja Istrupten(y)is/ "of alum" (alphabetic oT(ProuTllPw st(r)uptiT{j), but Cyprian writes codas. That poses a conflict since coda segments copy the (preceding) nucleus vowel, but word-initially there is no preceding vowel. The conflict was resolved in the only feasible manner. The dummy vowel (barring the somewhat bizarre alternative of generalizing the sign with dUmmy e from final position) can only come from the syllable nucleus, which aligns such examples with onsets. This does n'ot mean that they were regarded as onsets. Syllable adjuncts (Levin 1985) pose a serious graphic problem for strictly SH-based syllabic representations. That the Cyprian writers were (at least implicitly) aware of the problem is indicated by the fact that in non-initial position they maintained the SH-sanctioned treabnent. That they were able to go beyond the confines of a SH-based syllabic representation and write the adjunct Isl at all illustrates a clear conception of the segments involved and their relation to the SH. 3.7 Exceptions and Other Conditioning Factors Rare exceptions occur in both directions. A rare (regional?) variant of the (e-p)e-se-ta-se l(ep)estfAsel class (6g) is e-sa-ta-se [92.2] and e-pe-sa-ta-se [93.1] (Salamiou). Since the root is st. "stand", the spelling is most likely motivated morphologically (cf. Beekes 1971 :341 ; Viredaz 1983: 191). 9 In this context, observe the spelling of inaIaIisimena linalilismenanl "engraved" in §3.3 (continuant + nasal). This is consistent with the adjunct representation of s in se clusters, except that 151 should be lower in sonority than ImJ, wherefore Guion (1994) posits morpho logical spelling, viz. [[-lis]menan).
36
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWLEOOE
The major (verifiable) exception in the other direction involves a word internal onset treatment in Iea-ra-si-li Ignisthil "eat" (ICS 264. 1 Neumann 1 974). The form i s imperative to (the rare verb) 'Ypaw grd6 "gnaw, eat", (probably colloquial; cf. 'YpaoTlOpoc;) teles-(1'6ros) "fulfillment-(bringing)" and 'subtracted' in names like TEh€-$opoc; TeW-phoros (details in Miller 1982:78). The spelling of grdsthi represents the awareness of a problem: Ip+thil can be spelled compositional ly in the syllabary (po-li); Igra+s+thi/ cannot. The choice of spelling indicates the synchronic analysis of the -s- as suffixal rather than part of the root. 3.8 Conclusion It seems evident that the linguistic knowledge of those employing the Mycenaean and Cyprian scripts was far more acute than scholars prejudiced by biases about the alleged effects of literacy and especially alphabets have been willing to admit. In fact, the sophistication of developing a script based on the SH, consistently performing exhaustive SH analyses of each s able in each word and spelling individual segments according to their positi n on the SH, devising solutions to problems like SH onsets in coda posi . n or codas in onset position, handling problems of SH violations in the lan uage, and occasionally attempting to represent compositional information as well, go lightyears beyond anything predicted by Daniels ( 1992), Faber ( 1992), and others.
�
1 0 The imperative formative -Il _,hj is characteristic of athematic and monosyllabic verb forms. It is especially common in Homer and Arcado-Cypriot; cf. Hom. 5i&'1k dftJ iith; (= S(60u didou) "give", etc. (see Smyth-Messing 1963, §466; Schwyzer. Or. 1.800; Buck 1955. § 140; Thumb-Scherer 1959: 132-133, 168).
CYPRIAN SYILABARY
37
All of this gives additional substance to the claim of Morpurgo Davies (1987:97) that there was a continuity of at least folk-linguistic analysis from the Mycenaean to the late Hellenistic period. Ouion ( 1994) objects, but only because she follows the tradition of those who believe that Cyprian spelling reflected actual syllabification (e.g., Meister 1894: 177; Thumb-Scherer 1959; Masson 1961; Beekes 1971 ; Lejeune 1972:285). Those who do not believe the spelling reflected syllable structure include Viredaz ( 1983) , Morpurgo Davies (1987), and Woodard (1993, 1994). To my way of thinking, the latter is a more interesting hypothesis because it entails that Cyprian scribes were following something more abstract than their own syllabification. Unfortu nately, the metrical evidence seems to support the traditional hypothesis. All of the Arcadian and Cyprian poetic texts (of any antiquity) are investigated by Ouion (1994). While not unequivocal, the rare, atypical Epic scansions in which onsets (by the SH) do not constitute heavy syllables, e.g., [ka]s(gntYoi "siblings" at the end of a dactylic hexameter line (rCS 261: Golgoi), suggest (with Guion) that Arcadian and Cyprian syllabify by maximizing onsets according to the SH (cr. Wathelet 1970). One must be careful to distinguish the different kinds of disagreements. Nearly everyone now agrees on the 'fact' that Cyprian spelled according to the SH (or something very much like it). This is the leading idea that can be gleaned from Hermann, Lejeune, Viredaz, Steriade, Woodard, and Guion, all of whom differ on details like the universality of the SH and the implication of how far back this syllable structure is to be projected. For our purposes here, the latter concern is totally irrelevant. Nor does it matter whether or not the Cyprian scribes were following their own syllabification. Our point is that, even if they were, they were spelling out all the segments that people without alphabets are claimed not to be aware of. If they were writing at variance with their own syllabification, then our case for the (at least implicit) knowledge of the organization of segments according to the SH is even stronger. Unfortunately, the evidence is not clear and leans slightly in the more concrete direction. The degree to which the Linear B and Cyprian scripts were SH-based has only recently begun to be fully appreciated. Incorporating parameters and the notion of syllable adjunct (Levin 1985) has enabled a finer understanding of the SH principle in Linear B and Cyprian spelling.
4. THE GREEK ALPHABET 4.0 IntroductWn The idea espoused by Powell ( 1991), that there was one alphabet adapter who created the Greek alphabet, insofar as it is sufficiently testable and subject to empirical verification or falsification to merit being treated as a lHEORY , makes a radically different set of predictions from the hypothesis of Naveh and others, that the Greek alphabet developed over a period of time (hereafter, 'Evolution Theory/ET') with some differences in different areas, especially regarding the 'supplementals' (cl> a X /p h th kh/). Powell necessarily predicts that no inscription could turn up prior to Homer, since he believes that the alphabet was invented to record Homer. The ET predicts that inscriptions prior to Homer could eventually turn up. Powell 's hypothesis is
in fact circular. Since, by assumption, there could be no inscriptions prior to Homer, he would presumably be forced 10 move the date of Homer back, should some inscription turn up that dated to, say, -900 . The real question, then, is just how far back Powell would be willing to finess Homer's dates to make his hypothesis work. -1000? Earlier? My point is, simply, there is no way to falsify Powell's claim short of finding an alphabetic inscription of Mycenaean date. The issue then becomes, is there any other way to ' formulate evidence for or against such a claim? This chapter will attempt 10 establish the greater antiquity of the Greek alphabet based on the existence of no less than 33 derived Greek alphabets, the formal properties of the letters and their values in the context of some phonological changes in the history of Greek, and evidence for influence from the syllabary tradition. The history of the debate is collected in pfohl ( 1968), and a list of scholars with the dates they assume for the invention of the Greek alphabet is found in Heubeck (1979: 75fO. Extreme positions in favor of the recent (--8th cent.) date are defended in Wachter (1989), Jeffery & Johnston (1990), and Powell ( 1991). We will reargue the case for an earlier date. 1 1 Readily available technical accounts of the history of the Greek alphabet include Nilsson ( 1918), Guarducci (1967-1978), Heubeck (1979), Moms (1988). Wachter (1989). Jeffery & Jobnston (1990), Powell (1991). See also the overview in Strond (1989).
40
ANCIENT SCRlPfS AND PHONOLOGICAL KNOWLEDGE
4. 1 Transmhsion oftIN Northwelt Semitic Script According to Naveh ( 1 982: 9ff, 42), the Proto-Canaanite script, invented ca. -1800/ 1700 by Canaanites with some knowledge of Egyptian writing (details in Sass 1988: 86-87, 135-166; cf. Cross 1989:84-85), evolved into the Phoenician consonantal script, attested as early as the Byblos Matrix (ca. -10(0). Prior to that, the Ugaritic cuneifonn script broke off ca. -1400, and the Greek alphabet ca. -1200/ 1 100. Sass ( 1988: 167) demonstrates that a -10th century borrowing is also possible (cr. Diringer 1968:359). The Proto-Canaanite script contained 27 consonant signs, reduced to 22 by the -13th century. The signs were pictographs and most had acrophonic values (Sass 1 988: 106- 134; Cross 1989:80). These signs evolved into linear letters. The pictographic conception pennined writing in any direction: right to left, left to right, vertical (columnar), and vertical or horizontal boustro phedon. In the course of the evolution to the Phoenician script (ca. -mid 1 1th century), the stances of the 22 linear letters became stabilized, and the letters came to be written only horizontally from right to left. That is one way in which the Greek alphabet is more archaic than the Phoenician. Also, some Proto-Canaanite inscriptions (ca. -12(0) use the same word dividers ( : and D found in archaic Greek writing (Naveh 1982:36). As to location, there is epigraphic evidence of Phoenicians in the area of Knossos in the -1 1 th century, suggesting a possible avenue of transmission of the consonantal script to the Greeks (other possibilities and discussion in Heubeck 1979:80-87; LSAG 1990:425ff w. lit.). I personally favor Cyprus because of the evidence for (mutual) influence between the syllabary and alphabet traditions (§4. 13). Because of the mixed populations of Eteo cypriots, Greeks, and, after the -10th cent., Phoenicians (Heubeck 1979:85), Cyprus would have been an ideal location for the type of contact and interaction necessary to facilitate the development of the alphabet. . Since a major portion of the discussion in this and other chap1ers is predicated on the early fonns of the letters and their evolution, the sential development of the alphabet is outlined in Appendix A. The next sec ' on will concentrate on the inventory of letters inherited by Greek and the egree to which they could accommodate the Greek phonological system.
i;
4.2 Northwest Semitic Scripts and tIN Adaptation to Greek The phonological system in ( 1) could be accommodated by a Canaanite script. On the sibilants, see Faber (1981, 1993). Since there is no phonolo gical system without vowels (at some level), I will assume that the Northwest Semitic alphabet ignored vowels for root recognition and their morphological predictability (discussion in §4. 1 1).
41
1HE GREEK ALPHABEf
(1) Phonological Inventory of the Source Canaanite Script
LABIAL DENIAL VELAR lNULAR PHARYNG GLOTIAL
STOP [-VOICE] [+VOICE] [+PHARYNG)
t d �
P
b
AFFRICAlE
q
ts (semk) dZ (zai)
[-VOICE) [+VOICE] fRICATIVE
[-VOICE] [+VOICE] NASAL SONORANT
'1
k g
s
� �
§
n
m w
h
I r y
A glance at the Ancient Greek phonological system in (2) reveals that the borrowed script has both more and fewer characters than needed. (2) Ancient Greek Phonological System (a) CONSONANfS LABIAL DENTAL VELAR GLOITAL
STOP [-VOICE) [+VOICE) [+ASPlRATED] AFFRICAlE [+VOICE]
t d th
P
b
ph
[zd]
fRICATIVE
[-VOICE) NASAL SONORANf
k g kh
s n
m (w)
(b) VOWELS
I
0
e a
u
r [y]
h
[ u] i
�
f
i
�
Q
U
For the basic voiced and voiceless stops there was no problem matching letter with phoneme. The problem areas involved (i) the plethora of laryngeal and pharyngeal fricatives, which, by the acrophonic principle, could supply letters for the basic vowel system in (2b) ; and (ii) the plethora of sibilant characters in the Canaanite script, for which Greek had very little use. All four sibilant letters occur in the Greek alphabet in the same place as in the
42
ANCIENf SCRIPTS AND PHONOLOGICAL KNOWLEOOE
Semitic prototype (see Appendix A). Nevertheless, the I-shaped zai and the Byblos :2 (entry in Appendix A) continued in use as the [e e � ], [0 ij Q] [a a]. Additionally, it had long been realized that the signs representing Iyl and Iwl could be used for lil, lul (cf. already the Ugaritic script of -1400) . Greek no longer had Iy/; which had been lost by this time; in Mycenaean already, Iyl variably alternated with Ihl or 0 (see Wyatt 1 968; Miller 1 982: 144ff). Therefore First, I [y] served for lil, tt/. Second, Y (Originally [w)) presumably, like Latin V (Nilsson 1 918: 186fO, also served for luI, IQI (cf. (?) rare Cretan spellings like aFTov aWtOn "him(seJf)", etc.) . The fonn was stylized to F ; it defies common sense to imagine that the resemblance to Linear B (+ luI is accidental (cf. Cypr. � lu/). Because of the stylization of waw to F , the contemporary Phoenician fonn Y could be reborrowed for lul, IQI (cf. Diringer 1968:36 1 ; Heubeck 1 979:89-90). As a late borrowing, it was placed at the end of the alphabet -
TIIE GREEK ALPHABET
47
the first of the supplementals.6 Incidentally, older and younger forms of the same letter in an alphabet presuppose prehistoric evolution - a problem for Powell 's hypothesis. Reborrowing Phoenician Y provided for a unique letter for each of the five cardinal points of vowel articulation (quantity and tenseness (referred to by some theoreticians as advanced tongue root) were not distinguished): ii e£ �
at
u 1i 0 5 . evidenced by the inconsistency in their values and positioning. Wachter cannot explain the invariant association of . with Iph/. To explain that requires that • be more established than the rest. On the 'primitive' alphabets see the main text below. =
50
ANCIENT SCRlPfS AND PHONOLOGICAL KNOWLEDGE
whether or not there is any empirical evidence for either of these positions. The most empirical evidence adduced so far has been by Powell ( 1991:57), who shows that it is likely that the earliest Cretan abecedarium had at least cp because it occurs in Eteocretan inscriptions from the -6th cent., e.g., in the name of the town �paLao- "Praisos". One cannot, however. accept Powell's idea (p.56) that the supplementals were lost in Cretan because the aspiration was not sufficiently salient If it was salient enough to be represented later, it makes no sense to argue that i t was imperceptible earlier. Some areas, such as Thera, Melos, represent the (supplemental) aspirates as 1Th IJIl/, ICh Ikhl (see Heubeck 1979:98; Powell 1991:49-53). Why then did Crete use theta and tau correctly (given some deaspirations; Bile 1988: 140141) but disregard the supplementals and represent Ik/ and {khl simply as le, or /p/ and Iph/ as 1T? The absence of stop + h spellings has been explained by way of Cretan psilosis (on which see Bile 1988: 10lfO. Quite simply, the lack of IhI (heta was eta) precluded spellings with h (cf. Lejeune 1972:59; Wachter 1989:35). The absence of the supplementals is more difficult, given the distribution of theta and tau, and that has led some scholars (e.g., Lejeune 1 972 §46; Wachter 1 989:35) to the conc1usion that that area received no supplementals. Again, there are two possibilities: ( 1 ) that area never received them because there was no displacement; that is, 'modern' kappa simply replaced the older form which regionally dropped out of existence; or (2) the supplementals were lost in that area. The presence of cp in Eteocretan, if the evidence is to be taken at face value, would seem to tilt the scale slightly in favor of position (2).9 Nil sson ( 1918) showed that the distribution can be explained by a simple proportion, altered slightly here to accommodate the present theory. On our account, there was an early representation for It! : Ith/, and a somewhat later one for IpI : Iphl (see below), leaving one opposition (/kl : Ikh /) without representation. Different areas could fill in the one remaining gap by using the more recently displaced "'" (see above), or 9 Based on a newly found bronze tablet (one of four). on which an archaic form of the Greek
alphabet is repeated from alpha through tau twenty-four times. Heubeck (1986) speculates that this is the earliest form of the Greek alpbabet. before any of the supplementals were added. Since this is consistent with the invention theory. but not the ET. it must be explained (if the tablets are genuine). So far questions have been raised as to precise location. date. and even whether they are Greek (see Wachter 1989:40; Powell l991:3 1). What they do show. however. is the force of tradition. and that may have been an important factor in the local decision to abandon the supplementals. except for the syllabary-sanctioned 1' luf. In that context. it is interesting to note that the syllabary tradition did not represent the aspirates distinct from the non-aspirates. showing that salience was not the issue but rather economy and tradition. In some areas, the force of the syUabary tradition was strong enough to allow only the supplemental for lul to remain.
TIIE GREEK ALPHABEf
51
ignore all of the supplemental consonant letters, as i n the 'southern' area where the force of the syUabary tradition of no separate characters for (non dental) aspirates remained strong. What about the Western ('red') scripts in which X had the value [ks]? This is generally agreed to derive from the typical regional manner of writing [ps], [ks], as l:, Xl: (as in Attica; cf. Euboean I), followed by simplifica tion of Xl: to X, permitted because r represented Ikhl in Euboean (see Powell 1991 :60), analogous to the simplification of FH (i.e., w + h) for IfI to F in Latin (Diringer 1968:418-419; Gordon 1969: 1 60- 161, 169- 170, w. lit. ; cf. Gessman 1975:84; Sampson 1985: 109). where V functioned for Iwl and lu/, tu/, alike (Wallace 1989: 126ff). For additional discussion, see Morris ( 1988, chap.3). On Xl:. X, and r in Euboea, see Powell ( 1 991 :49-63), where a different interpretation is offered. The ubiquitous value of cl> as IrPl (a problem signalled by Praetorius 1902:676; cf. Powell 1991 :48-53) including Eteocretan! - must be explained, but PoweII's account of an invention (pp.58ff) does not explain why cl> should have been the first of the supplemental consonant letters, nor does it explain the form, nor the fact that it alone of the supplementals has a constant value. One remote possibility is that Iphl may have been given a letter before Ikhl because Iphl was more aspirated than Ikhl and therefore more perceptually prominent, but that seems unlikely. 1 0 It is also conceivable that the reason is historical; that the invention of preceded the split of "" Ik, khl (--. /kh/) and K lkI. How could that be? -
4.8 The Evolution ofQoppa
Some very archaic (early -7th cent.) fonns of qoppa, namely q>, � , etc. (see EG 4.265-266, 328-329, etc.) bear a strong resemblance to cl> (references 1 0 The evidence is not unequivocal but it seems to be the case that IkJ was more aspirated than Ip/ . I base this conclusion on the evidence presented by Teodorsson (1974: 131-137) and
1breatte ( 1980:449-469), that there is more confusion between K k and X K/I than between tr K and 12 K for x. Contrarily. there are 13 cases of tr for $, but only context-sensitive $ for tr . Threatte presents extensive discussion but no statistical summary. Both indicate more cases of unaspirated stop for aspirated than the other way around, suggesting an increase in aspiration among the unaspirates. One is reminded of the large number of languages in which. of the unaspirated series. IkJ is the most aspirated, followed by ItJ or Ipl in either order (see Lehiste 1970:22; Devine I 974b: 130-13 1). The evidence from gemination (Teodorsson 1974: 148-153, 23 1 23 5) reveals that IkJ and Ikhl are the most frequently doubled. three times more frequently than ItI. and nearly twice as often as Ip/. This suggests that there was a positive correlation between aspiration and tenseness, as in Hindi (Miller 1986), and forces the conclusion that, as in Hindi. IkJ was the most aspirated and perceptually prominent stop.
p and $ ph. Teodorsson. for instance. records 15 X for
52
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWLEDGE
in Jensen 1 969:462� Heubeck 1 979:92-93). Wachter ( 1989) rejects any con nection between «( and cl> on the grounds of phonological incompatibility. At face value this is true. However, there are additional considerations. Recall that the value of W. Semitic «( was a pharyngealized uvular stop [q] (marked value), which could be used for a marked value (aspirate) in Greek. So why was it used for Iphl and not for /kh/? If it is true that Euboean, etc. (i.e., 'red') r Ikhl derives from f , as is sometimes assumed, then it would regionally have a derived shape with the superficially expected value. However, the truly interesting possibility that the alphabet was borrowed when Greek still had the labiovelars has not been discussed anywhere, to my knowledge. Powell (p.41) and others miss the point of the early use of «( (qoppa) before 0 and u as reflecting lip-rounding, as a retention from the period when Greek still had labiovelars - quite late, on the evidence of Mycenaean and the very different dialectal reflexes (Lejeune 1972:43-53 ; Wyatt 1975b� Miller 1981, discussed in Stephens & Woodard 1986). One could appeal to the fact that Latin Ipl was borrowed into Old Irish as /kwl (Lewis & Pedersen 1961 :62) in order to account for the use of f Iq/ or Iq(w)1 for Iph/, but another possibility suggests itself. It is hardly accidental that tantalizingly close versions of the sign «( (qoppa) are found in Linear B : Jr' qo , ..qo qa, and 'i1 ko. One and the same sign was, in one form or another, associated with lip-rounding and/or round vowels from Mycenaean to the --6th century. l l Let us suppose that when the Greeks first borrowed the alphabet, 'P represented Ikwl and Ikwh/. The normal phonological development of a string [kw(h)o/a] to [p(h)oIa] would explain the consistent association of cl> with labials. Subsequently, for the non-aspirate, n would naturally be used, leaving the letter cl> exclusively for the labial aspirate Iph/. The implication is that the Greek alphabet was in fact developed somewhere around the -10th/-9th century, which accords well with recent observations by Semitic epigraphers, especially Sass ( 1988: 167). The reborrowing of contemporary (qoppa) from Phoenician 12 would entail a displacement of the sign phi with the marked value to the end of the 1 1 This is not to suggest, as Rex Wallace (p.c.) formulates my claim, that the similarities to the Mycenaean signs provided "secondary support" for the use of qoppa to represent the labialized velars (although the intemction between the syUabary and alphabetic traditions (§4. 13) does not rule out such a scenario), but mther that it hardly seems fortuitous that the same W. Semitic sign was used in both traditions for the labialized velars. 1 2 Rex Wallace inquires hy would the Greeks go the trouble to reborrow qoppa eveD w though it doesn't adequately represent any new phonological information? IkI is already adequately represented by kappa." Phonologically, this is unequivocal, but the evidence is that the Greeks were quite fond of the use of qoppa in the context of lip-rounding, again illustmting the force of tradition. The model abecedaria, to which the Greeks kept referring "
TIIE GREEK ALPHABET
53
alphabet, accounting both for its form (similar to qoppa) and its (motivated) position, adjacent to Y (also associated with lip-rounding), itself displaced by the reborrowing of F , and before ""'IX , displaced by the later reborrowing of �/k from Phoenician. Now it makes sense why k was reborrowed. Greek initially received ® for Ith/; it then developed (phonologically) et> for Iph/, leaving one gap - no paired set of signs for Ik/ : Ikh/. The reborrowing of the contemporary velar stop signs <j> and le, restored the inherited inventory of letters in their conventional order, plus the supplementals, already in place by the time of the earliest inscriptions from Lefkandi, Pithekoussai, and Attica (cf. Wachter 1989:68). 4.9 SupplelMntaU and I'M SibilDnt Letters One problematical supplemental remains, (jJ [ps]. Making the reasonable assumption that the problem of (jJ is bound up with the problem of the sibilant letters, we can treat affricate and affricate-like strings together. The si bilant letters, for which Greek had no use, were distributed for use in clusters: 1) -=F- [tS ] (�-) gave its name to s(gma, but sigma has the shape and position of rann-IBn . In the original position of samek is a letter with the same shape and the usual (Eastern) value of [ks] named x(e) i . But Praetorius ( 1902:679) already noted that there are also areas where E retains the affricate value of Phoenician samek, viz. IdZf (or the like), rarely, in Lyttos (Crete), Corinth (Heubeck 1979:91 : Wachter 1989:56), and Thera (LSAG 3 17, pI. 61 Ib, i; Powell l991: 130-131):
Whatever the details of the name changes among the sibilant letters (some specUlations in Nilsson 1918; LSAG 25ff; cf. Powell 1991:47-48), this letter was most assuredly regarded as an 'extra' and used, like 1fUle (below), for affricates or the affricate-like [ks] (cf. Praetorius 1902:679; Faber 1992: 1300.23). Wachter ( 1989:49fO is clearly right that the prototype contained E , M, and 'I, and that san users who kept E: used the original dental affricate value, while sigma users (like the Samos type) reassigned to E the value [ks] and got rid of san. 2) 'V I� I bears a conspicuously close resemblance to q,. Beyond the problem that this supposedly yields san (M) (though Naveh thinks M is just right into the historical period, had a slot for er with a given name and value, from which the contemporary q, differed too drastically, motivating the reborrowing of 9 to maintain the traditional conception of this letter.
54
ANCIENT SCRlPfS AND PHONOLOGICAL KNOWLEDGE
a rotation of sigma (1:), which makes better sense), there are two problems to motivate: (i) the value Ikhl in Western ('red') scripts, and (ii) the order at the end of the supplementals. The incredible confusion (pp; within but across alphabetic traditionsl3) about the fonn '¥, )K. - and value ( [ps], [kh]) with regard to k;h(e)1 - W, A. - might be explained by the assumption of two very similar letters, initially distinct, one an affricate, the other a velar displaced by the reborrowing of kappa. Suppose the former was an older borrowing but an unneeded letter. Then, as the quasi-affricates (clusters of stop + s) received letters,1 4 it was used for that purpose. What about the order? Why did it not just go in the place of san? There was general agreement among the Greeks that san had the shape M (whatever its source - a recent reborrowing causing displacement of older '+'?). Whatever the details, it could not occupy the position of the current conception of san, and fit fonnally and functionally with the other supplementals at the end of the alphabet. l S For a radically different proposal, see Wachter ( 1989:49-61).
'¥.
�,
4. 1 0 The Antiquity ofSegmental Writing The judicious survey of the history of literacy in India by Patel ( 1993) concludes that writing may have existed in India since the Indus civilization and that it was practiced in the Vedic period and during the time of PiI)ini. (See § 1 . 1 , for devan agari as a segmentally coded script.) The absence of ex tant writing from before the -5th century is explained by the fact that "early Brahmanical literature was written on frail and perishable leaves, birchbark. 13 I owe this formulation to Rex Wallace. 14 Why a language should have cluster-symbols for [ks] and [ps] has occasioned considerable speculation. For instance, Sampson (1985: 103) attributes their existence to the
are
fact that those the only clusters that could occur at syllable and word end. More to the point. they are the only onset clusters that could occur in coda position. The unitary letter reflects awareness of the old convention of providing complex signs only for onset clusters. 1 5 One other possibility (from Praetorius 1902:677ff, 1908:287-288) is mentioned by Jensen (1969:463 w. lil) and Faber (1992: 125, 13 1n.27), who discuss early South Arabian letters that closely resemble phi and psi (variant form and function of khi). Faber speculates that the old Phoenician alphabet from which both descended (Sass 1988: 166-167, dates the South Arabian planned creation of the alphabet to the -1 11 10th century) must have had those letters (cf. Praetorius 1908:288). The letter that ends up as psi, as noted above, does have a correlate that could easily be confused with and/or split off from the old form of IkJ, which agrees with the S. Arab. value as a pharyngeal fricative. The variant of phi with the v8Iue Iwl or Ivl is again not too difficult to motivate if also derived from qoppa, which had always been associated with lip-rounding. Already Nilsson (1918: 183) emphasized that Praetorius' hypothesis of Greek letter borrowing from S. Arab. was "sicher unrichlig", emphasizing internal parallels.
1HE GREEK ALPHABEf
ss
and later on hand-made paper" (Patel 1993:202). Similarly, Old Canaanite written on papyrus did not survive (Segert 1993:87). Mycenaean documents were probably transferred to perishable materials (Olivier 1986). Continuity of the Cypro-Minoan script in the Cyprian syllabary (§3.0-3. 1) presupposes interim documents on perishable materials (cr. Heubeck 1979:73 ; Olivier 1986). As noted above and in chap.S, similar arguments have been made for Greek and Germanic. Nevertheless, it must be emphasized that this has no bearing on the literacy of Homer (against Bellamy 1989, see Miller 1990) or of the public in general. William Harris (1989, e.g., 10lff) finds that literacy in Ancient Greece was restricted largely to a privileged minority and co existed with an oral culture (cf. Andersen 1989; Thomas 1989, 1992). The absence of documents, then, is not a major obstacle to the general idea that alphabetic writing in Greece can antedate the earliest epigraphic monuments. Another parallel with India can be adduced. Just as the oral mode of transmitting Vedic literature prompted development of grammatical analysis (Scharfe 1 9n, Patel 1993), so the orality of the very popular Homeric texts may have underlain the shift from a 'folk' grammatical tradition (Morpurgo Davies 1 987) to a (more) professional guild of grammarians, whose main function seems to have been textual exegesis, as stated fairly explicitly in the introduction to Dionysius Thrax 's TEXVll 'Ypa j.lj.la TtKll "Art of Grammar" (see Kemp 1 987: 172-173), and directly manifested in the copious Homeric scholia (e.g., Erbse 1969-) and other ancient commentators. 4. 1 1 Adopllllion and Development'Plu"e ofthe Greek Alphabet Another argument for greater antiquity is that the Brahmi script of the -3rd cent. Ashokan inscriptions represents the Sanskrit phonological system so well that it must have had a long history of development (Basham 1967:394; Patel l993:203). The Greek alphabet, on the other hand, appears to be less well adapted to the phonological system of Greek, but the opposite has also been argued; cf. Coulmas ( 1989: 162): The Semitic alphabet applied to a non-Semitic language could not be used to represent the sounds of that language without significant adaptations. The lack of signs for vowels was crucial here since, in contrast to the Semitic languages, vowels in Greek occupy a position on a par with consonants. By finding a solution for the problem of vowel indication the Greeks overcame this obstacle, thus making the alphabet more suitable for both their language and other non-Semitic languages.
I t has been argued (e. g., Harris 1986: 1 20) that the North Semitic alphabet may have ignored vocalic differences in reducing the earlier
.56
ANCIENT SCRIPTS AND PHONOLOGICAL KNOWlEDGE
syllabary, thereby adapting better to the word structure of the Semi tic languages. So, by (re)developing symbols for the vowels, the Greek alphabet was undergoi ng adaptation to the Greek morpho-phonological system. Sampson ( 1985: 101) emphasizes that Greek had many lexical contrasts with vowels, which were "important for communication" (cf. Diringer 1968:263, 435). Moreover, words frequently begin with vowels in Greek but not in Semitic, and sequences of vowels are virtually unknown in Semitic but frequent in Greek (e.g., Alaia " Aeaea", Circe's island). Daniels ( 1992:97) sees consonant clustering in Greek as a factor. Other changes (development of separate aspirate symbols, addition of other vowel letters, etc.) were all attempts to better adapt the alphabet to the Greek phonological system. Although the adaptation was never complete(d), the process of change and adaptation at the dawn of documentation seems to be continuing a prehistoric process, from which it is reasonable to conclude that, like the Indic scripts, the Greek had a long history of development Coulmas (1989: 164) affirms: In principle the Greek alphabet was suitable for representing all the phooemes of the Greek language [.. ). Systematic vowel indication is attested in the earliest Greek documents; no developmental state with defective vowel writing is known. .
This suggests a period of evolution and development of orthographic norms and conventions (cf. Gelb 1963: l8Of0. There were also choices made. In the Indic linear scripts, syllable structure was partially coded by different representations of onset vs. coda resonants (Mahulkar 1981:49); cf. the Greek syllabic scripts (chaps. 2, 3 above). But the Greek linear script was more strictly segmental in that little reference to higher levels of organization was made. While the Greek tradition was well aware of organizational units beyond the segment, the decision was made to represent little more than linear segmental units. Therefore, it is pointless to argue that the Greek alphabet was poorly adapted to the phonological system. All of this presupposes at least implicit knowledge of segments and their organization into higher linguistic units. That is, of course, another area of major controversy, which will be treated in Chapter 6.
4. 12 Sumiruuy and Conclusion Powell ( 199 1) unfortunately ignores completely the evidence for an early borrowing of the Greek alphabet: (i) the letter-forms and their Proto Canaanite prototypes; (ii) internal evolutionary evidence, such as Y to F; (iii) evidence for reborrowings by comparison of 'duplicate ' . letters with their earlier forms and contemporary Phoenician counterparts. Moreover, (iv), he
TIIE GREEK ALPHABET
57
fails to motivate the order of the supplementals and the constant association of cl> with Iph/, and (v) his theory of an adapter (as opposed to a bilingual, literate environment) fails to motivate the matching of unmarked and marked values across the languages. The stages in the development of the Greek alphabet reconstructed here are the following: ( 1) assignment of old 'pharyngeal ' letters to vowels; (2) evolution of Y to F ; (3) reborrowing of Y at the end of the alphabet (first of the supplementals) as a vowel sign sanctioned by the corresponding syllabary traditions; (4) ., evolved to IrPI via the normal phonological development of [kw(h)o/a] to [p(h)o/a] ; (5) for the non-aspirate Ip/, 1f was naturally used, leaving the letter cl> exclusively for the labial aspirate Iph/, all of which alone explains the consistent association of cl> with labials; (6) the reborrowing of contemporary « (qoppa) from Phoenician entailed a displacement of the sign phi with the marked value to the end of the alphabet; (7) evolution of W to provided a possible separate sign for the marked value Ikh/, which X and was exploited by the reborrowing of 'modem' kappa for the unmarked value IkI, as provided for in the model abecedaria; (8) regional differentiation of signs for affricates and affricate-like clusters; (9) rejection by conservative 'southern' areas of supplementaIs not sanctioned by the syllabary tradition (thus l' was allowed to remain, but not the supplemental consonant letters) ; ( 10) the East Ionic . evolution of eta and (within the historical period) creation of omega. Powell may be right that the spread of the alphabet in Euboea had to do with recording Epic, but Thomas (1992:56-65) rightly assails this view as "romantic", emphasizing that most of the earliest Greek a1ph�betic writings are not poetic at all. More likely, it was the high level of cultural activity at that place and time that prompted a renaissance of interest in the Greek alpha bet, as a consequence of which (at least short) examples of epic verse were recorded, along with a wide array of other things. In any event, Powell is surely wrong that the Greek alphabet is the product of a single adapter at that time and place, though a single adapter may have ultimately been responsible for the initial creation of the alphabet. The two views, of course, make entirely different predictions. Powell predicts that no very early inscriptions, say, using ., for Ikw(h)1 should ever turn up and that, if earlier inscriptions and abecedaria do turn up (and the date of Homer would - circularly - have to be moved back), regardless of where they are found, they should have the full set of supplementals he reconstructs. My prediction is the opposite, of course. To the extent that it is plausible for inscriptions a century or two earlier than the current corpus to
'¥
.58
ANCIENT SCRIPTS AND PHONOLOGICAL KNOWLEDGE
ever turn up, neither of these hypotheses is properly testable (empirically verifiable or falsifiable). Hopefully time will tell. 4.13 Interaction Between the Syllabary and Alphabet Tradition. Another issue involves the assumption of some continuity between the syHabary and alphabet traditions. Such continuity is typically denied (e.g., Stroud 1989:110). Faber writes (letter of 18 Jan. 1993): I'll grant you your demonstration that Mycenaean spelling conventions probably reflect sub-syllabic awareness, if you will grant me that there could not have been complete cultural continuity between users of Linear B and of early True-Greek orthography. If there had been such continuity, one would expect to see some reflexes of it in early True-Greek orthographic convention [ ]. ...
Of course there was no complete continuity, or there would have been no change. Yet there was continuity, and it has been documented by Hermann (1923), Morpurgo Oavies (1987), and others. A dditionally, everyone grants that there had to be continuity of the syllabary tradition from Cypro-Minoan to the Cyprian syllabary of the historical era, despite the absence of Greek documents in any related script during the 'Oark Age' (cf. Heubeck 1979: 65ff, 85ff; see §3.0-3.1 above). There was continuity of the conventions of syllable division based on the Sonority Hierarchy (and feature geometry) from Mycenaean on in both the syllabary and alphabetic traditions. This was carried down to the detail that s plus stop received special treatment. The alphabetic tradition continued the recognition of the special problem of onset clusters in coda position. Meister (1894: 185) noted the agreement of the syllabary and the alphabet in having a unitary sign for [ks], and Nilsson (1918: 184) further noticed that the form of X = [ks] is paralleled by the form of the Cyprian sign for [ksa]. This was generally taken to imply influence of the alphabet on the syllabary (cf. Masson, ad ICS § 28.7a), but it is actually the other way around, since the only clusters that traditionally (Mycenaean+) received special letters were onsets, and that is inherently a syllabic notion. Another syllabic coding in the Greek alphabet was the non linear (and nonsegmental) representation of aspiration (see Steriade 1982). Oaniels (1992: 197) makes the astute point that one of the reasons (if not the main one) for the shift to the alphabet had to do with the extremely awkward representation of Greek consonant clusters. For example, str6phigx "pivot" in some tradition might have been (*) so-to-ro-pi-ni-xe (cf. § §3.4, 3.6). To that one can add the problem of identifying where the vowels were.
1HE GREEK ALPHABEI'
59
For instance, se-pe-re, as in se-pe-re-ma IspermaJ "seed", could also be read [sepre], [spre], [sep-e], [spre], etc. These motivations also presuppose inter action between the syllabary and alphabetic traditions. This will be elaborated in Chapter 7. The representation of the same five syllabary vowels was continued. This is important in response to Gelb's point (1963: 182) that it is improbable that one person developed the exceptionless use of vowel letters using as a model the li near Semitic scripts with their highly irregular vowel notation. Since syllabaries invariably represent vowels (by definition!), which is natural, that being head of the syllable (chap. 1), this feature of the alphabet was another point of contact between the two traditions. Moreover, at least one of the vowel symbols exhibits identity across the systems - conspicuously the only one of the supplementals to be permitted in the 'south' where the syllabaty tradition of no separate signs for the aspirates prevailed (except for e which was sanctioned by the Phoenician script). Evidence has been presented that at least one symbol (with amazingly close forms of similar function in Linear 8) evolved over time by means of changes in the phonological system of Greek. In the earliest inscriptions, the same assumptions about words, clitics, and word-divisions were maintained in the alphabetic tradi tion that prevailed in the syllabary tradition (cf. §6.4). Finally, Heubeck (1979: 67-68, 86) makes the interesting point that a very archaic Cyprian inscription (lCS 174: Paphos, second half of the -Sth cent.) is atypically retrograde (to-ro-to-so-si[- i.e., [L Ii]si-st6rt o "of Lusi-stortos" [Neumann's restoration, accepted by Viredaz 1983: 19 1]), and claims that this can most easily be explained by interaction with the Semitic script. Moreover, the Cyprian syllabary and the alphabet both begin to thrive around the -Sth century, suggesting a renewal of interest in both forms of writing. Any of these points in isolation is subject to challenge, but the composite picture that emerges is one of considerably more interaction between the syllabary and alphabet traditions than is typically granted (except by Nilsson 19 18). The composite evolutionary theory points to a compromise between the vowelless Semitic script and the superfluous vowels of the syUabaries to create a script with the advantage of indicating where and what the vowels were (chap.?). Influence from the syllabary tradition came both in the form of particular symbols and orthographic conventions. Given the continuity from the Linear 8 to the Cyprian syllabary, there is nothing inherently bizarre about the idea that the alphabetic tradition developed concurrently and that scribes familiar with both scripts transferred syllable-based conventions to the alphabet, or, in the case of consonant clusters, deliberately distanced them selves from the awkward vowel repetition. Sometimes, as noted, the influence
60
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWLEDGE
went the other way. This Greek-internal evidence accords better with the findings of Semitic epigraphers who recently opt for a -10th (or -9th) century date for the origin of the Greek alphabet
5. THE RUNIC ALPHABET 5.0 Introduction The Gennanic peoples used a 'runic' alphabet from around the first century to the Middle Ages. The word rune has occasioned much speculation. Gothic r ana can translate Lat. mysterium "mystery; secret". Old English run means "mystery; counsel; discussion; word" (Fell 1991). There is one mention of runic letters (run-sw/as Accpl) in Beowulf 1695. in a description of inscribed golden hiltplates on the captured sword that Beowulf gives to Hrothgar (cf. Elliott 1989: 17). Runic letters (runstajum DATpl) are them selves equated with magic (drfrrteft) in ..Elfric. Homilies 2.358 (cf. Elliott 1989:81). The 'secret' was important enough to be borrowed by the Celts, e.g., O.lr. run "secret" (cf. Eliott 1989: 1ff). Just why runes were surrounded by so much mystery and secrecy is itself a mystery. Antonsen (198Oa; 1988 ; 1989: l4Off) takes a very strong position against magical theories of the runes. It is true that the development of the runes need not be further obfuscated by the fantastic, but nothing precludes associations with ritual and magic (DUwel 1983: 1 1 1ff w. Iit.), as in Ancient Greece (Thomas 1992:78-88). In any event, as Antonsen insists. that is just one USE of the script that has no bearing on its creation, original function(s), or the original meaning of the word, which may have had to do rather with scratching (Morris 1985). but see Fell ( 199 1) ; early runic r iin was "message; text" (Antonsen 1990:3 14). Since runes were the stock-in-trade of the writers in runes, 1 the very knowledge of the letters in a largely illiterate society could have prompted the interpretation as "mystery; secret", more-or-Iess as in ancient Babylonia the 'supreme secret', the key to the universe, that the god Ea taught his son was the concept of the number 1 The word eriJoz is sometimes translated "rune-master", which Elmer Antonsen (letter of 24 Feb. 1993) deBaibes as "a stab in the dark. All we know about it is that it is used in parallel with terms like gudija 'priest' and pewar. 'servant' ." Antonsen (e.g., 1981:56-57) translates eri/az simply "eril" and suggests to me the phrase writers in runes for the present context. It has also been suggested that the term originated as a tribal name, Heruli, of people skilled in nmecraft (cf. Elliott 1989: 1 1-12 w. lit.), but there is no evidence for that (Antonsen 1990: 314). I wish to take this opportuni ty to thank Professor Antonsen for several sets of extensive comments on this chapter. Marie Nelson and W. C. Wall also read an earlier version, and Jay Jasanoff sent me detailed comments on the Germanic Vowel System (Appendix).
62
ANCIENT SCRIPIS AND PHONOLOOICAL KNOWLEDGE
(Hopper 1969: 12). More directly relevant is the point made by Watt ( 1989: 92n.3) that from at least the'-6th century 10 the Middle Ages mystical power was associated with the correct recitation of the abecedarium. There are in the neighborhood of 5000 runic inscriptions, some 3000 in Sweden alone, ca. 1 100 in Norway, some 700 in Denmark, around 60 in England, and so on. In general, the farther south one goes, the more rare they become (DUwel 1983:3). This creates a problem for Italic theories of the origin of the runic alphabet This chapter presents additional evidence for the archaic Mediterranean theory of the origin of the runic alphabet, and shows that it was created along phonetic parameters analogous to those underlying the ancient scripts of Byblos, Ugarit, and the Phoenician script. 5. 1 The Older Runic Fu park
There are extant some 250 early Germanic inscriptions in the older runic alphabet (DUwel 1983: 123), though only a little over 50 have more than two identifiable words (Antonsen 1980b: 1). An idealized version of the 'older fu]xuk' (named from the first six letters), which had 24 letters, is presented in ( 1), following Antonsen (1989:142). (1) Older Runic Alphabet (Idealized)
f " t>
�
N of.
� J'
f u
h
t»
a
j
n
� r
ae
t � M f1 rt
b
e
m
I
< k
f> X w g
r: r p
z
0
r>4
ng
d
s s
�0
There are not many archaic variants of the runic letters, and they are difficult to identify in a non-circular manner (Antonsen 1982; Williams 1992). Most notable is A for luI, the second letter, and the twelfth, or j-rune, has 'horizontal ' variants prior to 400 (Odenstedt 1990). The runic letters can be assigned a number for reference: (2) follows the arrangement in ( 1). (2) Numbers Assigned to Runic Letters (a) u2 fl P3 8.4 iu n lO j l2 (b) h9 m20 el9 big (c) t1 7
r5 il 13
hi
kt; P l4 U22
g7
Z15
d23
Wg S l6 OM
TIIE RUNIC ALPHABEf
63
Transliteration values (especially of < 13» follow Antonsen ( 1 975: 1- 1 0). Discussion of the problematical values follows below. Some of the orderings will be modified in §5.8. 5.2 An Idealized Runic Abecedtuium and its Divisions The order of letters in ( 1) and (2) is that of the Kylver stone (3a). The runic alphabet is written out completely in three abecedaria, listed in (3).2 (3) The Oldest Complete Runic Abecedaria (a) Kylver stone (ORI 30: Gotland, Sweden, 0-400 [Williams 1992: 196 w. lit.]). For the possible ritualistic use, see EIliott ( 1989:82). (b) Vadstena and Motola bracteates (ORI 90: bstergBtland, Sweden, 500-550).3 (c) Grumpan bracteate (ORI 91: VastergBtland, Sweden, 500-550). The idealized alphabet in ( 1 ) and (2) differs slightly from the extant abecedaria. In (3a), IpI precedes lel The bracteates (3b/c) present a sJightly different coda: 101 and Id! are reversed. The tripartite division, in three groups of eight (ON �iltir "families; rows of eight"), is presented by means of interpunct dividers (vertical dots) on the bracteates (3b/c). The reason for this arrangement is unclear. Antonsen ( 1989: 142- 143) denies any magical significance, but see Elliott ( 1989: 13- 14) and DOwel ( 1983:9). A feasible rationale would be rhythmic subdivisions in which the alphabet is to be uttered. The 8-8-8 is reminiscent of the 7-9-7 division of the English alphabet jingle: "A B C D E F G 11 H I J K L M N 0 P 11 Q R S T U & V [... r' (cf. Watt 1 989:83-84, noting similar subdivisions in other alphabets). 5.3 Origin ofthe Runic Alphabet: Medite"anean Theory Of all the speculations about the origin of the runic alphabet, the one that has claimed the most proponents is the North Italic hypothesis, richly discussed along with alternative proposals in Moltke ( 1985:49-73) and Elliott ( 1989:6- 1 2). Nevertheless, the problems with even that hypothesis have been long known. For instance, Etruscan had no 10/, and did not distinguish voiced and voiceless stops (Devine 1974a1b).4 The main arguments in favor of the
2 Somewhat incomplete abecedaria are also known (OR! 89, 99, 104, 105, 106). For dis cussion, see Moltke (1985:24£1). Schematized drawings appear in Page (1987: 18). 3 A bracteate is a stamped gold medallion, warn as an ornament or amulet around the neck, perhaps as a lucky charm (Elliolt 1989:83; photos and discussion in Moltke 1985: 108-121). 4 See the extensive references in DUwe1 (19681 1983:9lf1) and the more recent and detailed discussions in Morris ( 1988) and Odenstedt ( 1990, 1991). Rex Wallace objects (p.c.) that the
64
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWlEDGE
Mediterranean hypothesis are summarized by Antonsen ( 1 982; 1 989: 145146): (i) there is not a single early inscription from middle or southern Germany; (H) sporadic agreement in shapes is not sufficient to establish a relationship of origin; and (Hi) the most striking overall correspondences are to the archaic scripts of the Mediterranean. A fourth argument, that "older and older [runic] inscriptions are found" (Morris 1988: 1 57; cf. 150) has been claimed not to be true (Odenstedt 1989; 1991:363-367), but Antonsen (letter of 24 Feb. 1993) upholds the claim, accepting the validity of the 1st century Meldorf fibula and other recently discovered (2nd century) inscriptions. The most important epigraphic correspondences with the archaic scripts include the following (Antonsen 1982; Morris 1988 §§2.4-2. lO, 3 .6-3. 12): 1) The direction of writing is not fixed, but there are no violations of linearity; inscriptions read left-to-right, retrograde, vertically, or boustrophe don (Antonsen 1983; Morris, pp.69ff, 99). Odenstedt ( 1991 :383) counters that after the 6th century there are few examples of writing that were not left to right, but that is irrelevant for the early period and the origin of the script. 2) Nasals are frequently not written before certain consonants (Anton sen 1 972: 127; 1975 §4.2; Morris, pp 68-69) : Widuhudaz (ORl 5: SjmUand, Denmark, 200) = Widu-hundaz "wood-dog, forest-hound" (possibly "fox"; cf. Moltke 1985: 128; Insley 1991:320) ; asugisa/as (ORI 15: Kragehul spear shaft, Fyn, Denmark, 3(0) tansU-giSl-ast "of Ansugisl". A simpler explan ation is suggested by Morris himself (pp. 126- 127), that the absence of nasal letters could indicate nasalized vowels and/or the absence of complete nasal closure, as in Eng. hunt, hump, etc. (see §6. 10 below). 3) Double consonants did not have to be written in archaic Greek (cf. Morris, p. 155); in runic, they are "virtually never used" (Odenstedt 1 99 1 : 384); cr. ginu tginn-ut "mighty" (ORI 15: Kragehul spearshaft, Fyn, Den mark, 3(0). 4) Occasional interpuncts with a variable number of dots appear (Mor ris, pp.75ff, lOO, 137ff, 155) ; see, for instance, the picture of the Tune stone (ORI 27: 0stfold, Norway, 4(0) in Antonsen ( 1989: 146), and the discussion in Antonsen ( 1983:30-39). However, the runic use of interpuncts is too rare and unsystematic to draw any firm conclusions (cf. Odenstedt 1991:384). .
"North Italic hypothesis does not necessarily entail 'Etruscan' per se." This is certainly true,
and he goes on to argue that the absence of 0 in later Etruscan abecedaria is no problem since the Veneti had the letter (see Lejeune 1974). Moreover. Wallace maintains that. since the runic letters for Igl and Id! do not seem to be gamma and delta, it is likely that they were taken from some script in which "they were at home." l completely agree. The question is not only the source of the letters but also the other writing conventions, and those point to an esdiCl' source of the runic alphabel
TIlE RUNIC ALPHABEI'
65
Based on the above evidence, Monis ( 1988) derives the runic alphabet from a preclassical, epichoric Greek alphabet, ca. -SOO, prior to the loss of F Iwl and <j> Iql, which he claims (pp.59, 152) served as the models for runic IfI and l,y . This seems a little strange in light of Monis' claim, with which I am in complete agreement (see below), that the inventor(s) of the fupark had to be in a position to analyze the phonological systems of both the source and target languages. As Monis (pp.93, 95) acknowledges, since the classical Roman script was the only Mediterranean script that used F for IfI (but see §4.7), it hardly seems likely that F would get reassigned to IfI and not keep its proper value Iwl, to which a different letter got assigned.
5.4 Critique o/the LaJin Origin Theory Odenstedt ( 1990; 1991: 367-368, 376-383) objects to the antiquity of the runic alphabet on several grounds: 1) One does not find different runic alphabets as one finds different preclassical Greek ones; there are minor variants but surprising uniformity over the half millennium of its use. However, all that means is that the runic
alphabet was consciously designed, like the Indic (Alien 1953:20; Diringer 1968: 262; Jensen 1969:362; Watt 1989:7 1 ; Coulmas 1989: 185), the Korean
Han'gul script, also built on phonetic features (Sampson 1985: 120- 144), the Ugaritic and South Arabian scripts (Diringer 1968: 178- 179; Sass 1988: 166167), etc. (Gelb 1963: 144, 206ff). 2) There are thousands of extant Iron Age objects on which a runic inscription could appear, but there is not a single instance, which is difficult to explain if the runic script dates to ca. -500, as Morris ( 1988) claims. On counterpoint, A ntonsen (letter of 24 Feb. 1993) affirms that "it is quite reasonable to assume that runic writing was not used for inscribing on metal until a considerable time after its invention and then probably in imitation of the Roman practice of putting 'makers' marks' on weapons. See the most recent discussions by Marie Stoklund and K1aus Dtiwel." 3) All of the letters can be derived from the classical Roman script. This is especially true of IfI from F and IrY from Q, or possibly from O. Odenstedt has no problem with the arbitrary reassignment of values. He claims that Germanic did not 'need' Q Ikwl, which is not true (it did not, of COurse, need qoppa Iq/). He does not address the issue of why [0] should be so much more important in Germanic than in Latin as to merit its own letter,. when in fact it was allophonic in both. Clearly, the creator(s) of the fupark must have had good reason to invent a letter for [0] (see §5.5 below). The problem has been adequately described by Antonsen ( 1982: 5) : 8 letters are identical in the Latin and runic alphabets (B F H I L R T V); 6 are
66
ANCIENT SCRIffS AND PHONOLOGICAL KNOWLEDGE
related in form and function (A elK D M 0 S); 3 correspond in form but not function ( � = Iw/; M = le/; X = Ig/) ; and 7 (5 of which are in the middle row and two in the coda) have no fonnal or functional correspondent (1nl , /yl < 12>, lel < 13>, IUI , Ipl . 171 , Id! I J » , that is, ii combined in various artistic arrangements (see Lejeune 1974). Compare also the early OE use of uu (whence w 'double u ' ) for Iwl and the treatment by the First Grammatical Treatise (below) of Iy/, Iwl as phonological variants of lil, Iu! (cf. Haugen 1972 §3. 1O). 5.5 Runic as an Invented Script While I do not care to enter into the source dispute, which frequently has more of a religious than a scientific aura, it should be observed that the requirement of a single-source alphabet seems naive and obsessive. Many scripts have letters of different sources, among them the Cyrillic and, within Germanic itself, the Gothic, which possibly added thorn to the repertory of mixed Greek and Latin letters, plus a very runic-looking IfI, lul, and /o/,
TIIE RUNIC ALPHABET
67
reserving qoppa and sampi for "90" and "900" (discussion w. lit. in Braune Ebbinghaus 1961: IOff; Jensen 1969:484ff; Gessman 1975:70). Vennemann ( 1 97 1 : 129) makes the point that the Gothic script utilized borrowings to remain "phonetically accurate". Known modern script creators freely borrow from different sources (Daniels 1992). Nor must the element of creativity on the part of script inventors be denied. Exemplaria featuring creative additions include the Cyrillic script (Diringer 1968:374ff; Jensen 1969: 494-495. 502), the Indic scripts (Patel 1993), and numerous others (Gelb 1963: 143 - 144. 206ff; §5.4. 1 above). Within the Germanic tradition. the anonymous author(s) of the early thirteenth-century Icelandic First Grammatical Treatise (FGT), on the philological problems and interpretation of which see Koerner ( 1993: 122fO. devised a phonemically detailed script (Haugen 1972). To avoid the ambiguities of adapting the Latin alphabet to Old Norse, FGT establishes nine vowel letters, each distinguished for length (by an apex) and nasality (by a dot over the vowel), illustrating each by means of (over 50!) minimal pairs. The creativity can be exemplified by the genesis of some of the letters, e.g. pp. 13- 14 (Haugen's translation) -
Q gets its loop from a and its circle from 0, since it is a blending of their two sounds, spoken with the mouth less open than for a, but more open than foro.[... ] (J is made up [...] with the cross-bar of e and the circle of o. Y is a single sound made up from the sounds of i and u.5 On the consonants, it becomes clear that FGT also knows English, Latin, Greek, and possibly Hebrew (but see Haugen), and freely uses letters from a variety of sources, including the runic thorn. Haugen ( 1950:43) formulates the underlying principle (explicitly stated in FGT 84. 16): "If the symbols could not be given their Latin values, they must not arbitrarily be assigned new ones ; instead. new symbols must be found or created to fill the gaps." It is probably safe to assume that FGT did not simply concoct this tenet. and that some tradition was being followed. Since it is not clear how old that tradition was, we cannot be sure that the fupark inventor(s) made the same assumption, but there is at the very least a message of caution for scholars like Odenstedt who would freely let arbitrary values be assigned. There is no reason to accord the jupork inventor(s) any less creativity or prerogative than known script designers. For instance, why not admit the 5 The description is
to make the y-rune
reminiscent of the modification of the OE u-rune n with the i-rune I tn- /U/. On the form of the FGT Y letter, see Haugen (pp.51-54).
68
ANCIENT SCRIPTS AND PHONOLOOICAL KNOWLEDGE
obvious in the case of the runic character P? Scholars generally accept the Norse name of this letter as more authentic than OE porn "thorn" (e.g., Page 1987: 15; Polom� 1991 :430), but nothing forces that conclusion. Suppose its Germanic name was in fact thorn. It certainly 'looks like' a thorn. Depending on one's ASSUMPTIONS about the date of the runic alphabet, neither standard contemporary Greek nor Latin had the sound 1'pI (Alien 1988: 22-32). When confronted with a sound for which no available script had a letter, the inventor(s) of the Jupark selected (by the acrophonic principle) a salient word containing the sound rpl and represented it by a quasi-pictograph. On the creation of a sign for 10/, it need only be commented that there was a long tradition for such a letter. The Medieval Icelandic FGT created a ligature out of N and G (Haugen 1 972 14.9), and called it eng , reminiscent both of the Old English name of rune , ing (the disputed source of which is discussed in Polom� 1991: 43 lff), and of the Greco-Roman aYlla dgma h1gmal. Alien ( 1988:35-39) reports that Nigidius Figulus already described " the sound as "inter li tteram n et g and discusses at Jength the spelling convention of using y Igl before another velar to represent l,y, a practice continued by the Goths, e.g., laggs [laugs] "long" (Braune-Fbbinghaus 1961 § §50, 67-68) . Given this tradition of using two Greek gammas for IrI, I suggest that the inventor(s) of the runic alphabet likewise combined a 'regular' and a retrograde r to invent the box-shaped sign 0 for 10/.6 The counter-facing, juxtaposed I -variants, < and >, to make Ijl , provide a fitting analogue (see above, §5.4 end). Based on their sophisticated phonological knowledge (below), it defies common sense to think that the creator(s) of the runic alphabet did not know several languages AND THEIR SCRIPfS , at the very least, Latin and Greek, and probably also some Northwest Semitic language as well. This broader range of Classical knowledge is supported by Bremmer ( 1991), who demonstrates a parallel between Woden-Odin and Hermes-Mercury as inventors of scripts. 5.6 Ger1fUlllic Vowels and Runic Letter. Phonological evidence for the archaic nature of the runic script has been adduced by Antonsen (e.g., 1 975:3-6; 1982; 1987; 1 989: 149ff). Given that ancient scripts could use the same vowel signs for long and short vowels, it is strange that, if there were 10 vowels (five short lieuoal and five long lieuoil) to be represented, the runic alphabet should have 6 vowel letters ( u , a , i , ;1; , e , 0 OF mede, OHG mete, OE me(o)du "mead" vs. *widhu- (Gall. Vidu - , O.lr. fo1 "tree") > OS widu, OHG witu, OE (widu » Iwudu "tree, spear, etc.", Eng. wood. Occasionally there are doublets, like OHG skiJIskef (OE scip > slUp), OS, OF, OE wulf beside OHF wolf "woIr', which, according to Hock ( 1973), are due to the coalescence of, e.g., NOM *wulfaz > *wolfaz (> wolf) and Voc *wulfe (> wul/). This leaves the Germanic languages with doublets to (re) distribute. Significantly, in neuters like OHG fel "skin", berg "mountain", horn horn", joh "yoke", etc. , where there was never a distinction between nominative and vocative, no doublets exist ( *fil, *birg, *hurn, *juh), despite other places where an e/i- or o/u-alternation occurs in these words, as infillen "to skin", gibirgi "mountain range" (Gebirge), etc. 15 reflexes, cf.
"
5.16 Short Vowel.r in Unstressed Syllables The fate of lel and lil in unstressed syllables is less certain. OHG
2pl. *iJher+e+te) "you (p.) bear, carry" may be crucial evidence, if it is not leveled. Antonsen (1972: 123, 138- 139) takes it for original and con trasts it with 3sg. birit, from * iJher+e+ti, Gmc. *peri «i). 1 6 Antonsen claims pres.
beret «
that unstressed */el did not affect the vocalism of root syllables. The idea that le! became [i] in unstressed positions has been signalled as the explanation of
ek/ik, OS eclik, ON ek, OE ic , etc.) and *seg hes (Skt. sdhas "power") > *se yiz > *si },iz > Oath. sigis, ON sigr, OS, OHG sigi, Germ. Sieg "victory". Meid ( 1967 § 1 1 1) derives the final -i- from the *-is- of Skt. arc-($- "ray, beam", etc., and Antonsen (1972: 139) discusses e-raising before *lzI. Hollifield ( 1 980:34) alternations like
ik/ek
"I" (runic
putative derivations like
reformulates as e-raising to lil in unaccented syllables except before Ir/. In the case of OHG (etc.)
sigi,
there is another possibility - a Caland compound
15 1bis is dialect-internal. Cross-dialectally. there are exceptions. pointed out to me by Antonsen (FAX of 30 Nov. 1993): Germ. Gold, Eng. gold vs. Dan. guld; Eng. word « IwurdI) beside Germ. Wort, and even OHG skijlskef, Dan. skib, Swed. skepp "ship", etc. 16 Jay Jasanoff (p.c.) thinks that 2pl. beret is most likely leveled from the already leveled fOOll bera!. Hollifield (1980:34-35) claims that unstressed le! (except before Ir/) became lil in Germanic, wherefore he believes that the Monsee-Vienna fragments preserve the Gmc. 2 pI. in quidit "speak, say", etc., forcing the conclusion that forms like beret are leveled from leveled bera!. Antonsen (FAX of 30 Nov. 1993) finds it unlikely that bera! would have been leveled to beret in light of 1 pI. beram es and 3 pI. berant. and reiterates the probfem of explaining the absence of umlaut in the 2 pI. pres. ind. versus its presence in 2 sg. biris. 3 sg. birit. He challenges the lengths to which scholars go to maintain a traditional rule ("all unstressed PIE lels become lil in Gmc.") Suffice it to say, there is little agreement on whether or not all cases of unstressed lel became li/. Nothing here depends crucially on that detail.
TIJE RUNIC ALPHABEf
79
form *segh-i- (extensive discussion of such formations in Bader 1962, chap. 1 ; Nussbaum 1976; see al so Szemer�nyi 1990:204-205 w . lit.), beside the neut. -s-stem *segh-es-, in names like Sige-ricUs (cf. Ooth. reiki, OHG r1r.hi, Germ. Reich "kingdom"), Germ. Siegreich; Germanic-Latin Segi-m lrus, etc., and these names show contamination with sigis; cf. Seligis-mundus (6th cent.), Sigi(s)-bertus (6th cent.), Sigis-meres (5th cent.) , etc. Since there is independent evidence for a later preservation of final lil than lel, forms like *se y-i- and *peretf suggest a possible solution. Suppose le! became [i] before lil of the following syllable (cr. Streitberg 1896 §63). Such a rule has a considerable amount of support from early attestations. To begin with, the early evidence for ist (ca. 350) from *esti "is"P and Sigi- in runic Ssigaduz « *Sigi-ha]Juz) [K 47: Svarteborg, Sw., ca. 450] , agrees with the onomastic evidence, pointing to an e-raising rule as early as the 1 st century; cf. (ca. 100) Segi-m lrus (Tacitus, Annals 1.71 [2x]), Segi-mundus (Tacitus, Annals 1.57), but Sigi-m�us (Velleius Paterculus 2. 1 18.2 [ft ca. 30]). Velleius Paterculus' Sigi-m erus is particularly interesting because it appears beside Segestes (2. 1 18.4) ; cf. (-1st cent.) Segestes in Strabo (7. 1.4 ) and in Tacitus (Annals 1.55 [3x], 57 [4x] , 58, 59 [3x], 60, 7 1). Given older Seges- and Segi-, there is no reason for later Seges- beside Sigi- unless Sigi reflects a change in Germanic about that time. IS By everything known about Germanic compounding, there is no way the radical *e of Segi- could not have been stressed, viz. Segi-merus (cf. Streitberg 1896: 53, 55, 121, 142; Bennett 1972: 104). The change of Segi- to Sigi- around the 1st century agrees with that in the name of the Finns: Latin (ca. 100) Fenm (Tacitus, Germans 46x) but Finni, 4>( VVOL (2nd cent.: Ptolemaeus, Geographia 2. 1 1 . 16, 3.5.8); cf. ON Finn(a)r, OE, OS, OHG Finn, and runic Fin(n)o "Finnish woman" (K 86; ORI 74, Berga stone, SOdermanland, Sweden, ca. 500) . Another reasonably early example of radical e- raising is found in the word for "friend": runic uiniz (K 135: S0nder Rind, Denmark, 500; cf. earlier ekwinai "I for a friend": 17 A number of factors presumably played a role here, e.g., the normal clitic status, plus the fa Lat. mare "sea", Gmc. •mar; (Goth. mari-saiws "sea", OS , OHG meri, OE merilmere) "lake; sea", but .eti > Lat. et "and", Goth. ;p "yet, but", ambiguous because Gothic lost final fif in absolute fmal position (cf. Streitberg 1896:54f1). Hollifield (1980: 175) also accepts the Idea that .-i. "though generally retained in Proto-Germanic, was lost in the personal endings of the verb at least as early as common Germanic." 18 Not everyone agrees on the validity of the loanword evidence, as Antonsen (FAX of 30 Nov. 1993) points out to me, citing Marchand ( 1959). My point is, simply, that when: there is no conflict between the internal and externaI evidence, there is no reason to doubt their mutual corroboration.
80
ANCIENT SCRIPrS AND PHONOLOGICAL KNOWLEDGE
ORI 12: Rogaland. Norway. 3(0), which yields ON vinr, OS, OHG wini, OE wine , Eng. (Good)-win. Since the Indo-European root is *wen- (Lat venus "love", Skt. vdnas "lust", van(- " desire''), the Proto-Oennanic fOmi would have been something like *wen+i+s, and the change to Iwin-i-z/ obviously occurred well before 300, since there is no trace of the inherited radical *e anywhere in Gennanic. 5. 17 New Long Vowell Following yet another e-raising rule, this one in nasal clusters (cf. *wendaz [Lal ventus "wind"] > Oath. winds, OE, Eng. wind), a sequence of the type *-Vm-- Ioses the nasal with compensatory lengthening. The historical sequence was something like ( 13) for Goth. peiJum "to thrive, prosper" (cf. O.lr. con-tecim "coagulate") and Goth., OS, OHG INuln "to catch" (cf. Lat. pang- lI "fasten, fix, settle"). For discussion, see Vennemann ( 197 1: 102ft) ; Hollifield ( 1980:32) ; Voyles ( 1 992:� i ). ( 13) Some Early Gennanic Phonological Changes 1) Pre-Gennanic *teuk+n+m 2) Grimm's Law *�x+onom 3) Gmc. vowels (etc.) *l!eox+an(an) 4) e-Raising before Nas. *}>iux+an *J>ix +an 5) Nas. Deletion 6) Other }>ih+an
*pagk+n+m * fag x+nom *fagx+an(an) *fiX+an fih+an
Nasal Deletion ( 13-5), despite much support in the literature (some references above), ultimately brought new IAI into the system, but the precise dates of that change are unclear. As Jay Jasanoff insists (p.c.), the reflexes of Nasal Deletion remained distinct fonn the lal reflex of * eJ (OE siUon "we sowed") into Old English, where it fell together with the reflex of *-ans- (etc.); cf. *gans (OHG gans etc.) > OE gas "goose", like pohte « *}>au x-ton [Hollifield 1980: 150ff, 1 60fl] ; cf. OHG d,;;;hta) "I thought" (cf. Streitberg 1896:76; Antonsen 1972: 127). This suggests that Nasal Deletion was in fact rather a nasalization process and that its output was a nasal(ized) vowel. Moreover, this new vowel did not fall together with new lif in loanwords in some dialects; cf. OHG suochjri (= Oath. sokareis) "seeker" vs. OE (Beowulf 253) (leas)-sctfaweras "(deceitful) observers; spys" (NOMpl) with shortening of - Ii (from Latin - arius). Adducing shortening i n the extreme northwest corner of Gennanic (Lowe 1 972:2 14) does not explain why that never affected the reflex from the nasal, again pointing to a distinctive contrast in NW Germanic between long lif and long nasalized It/. What lends this hypothesis some
1lffi RUNIC AlPHABEf
81
credence i s that the 3 sg. pres. of ON fd "to get, grasp" (= Goth. filtan, OE fon "to seize") is given by the First Grammatical Treatise (§5.5 above) as f�r. i.e., If�/, showing that in the 13th century, the vowel of f�h)- was still nasalized. Therefore, Nasal Deletion should be reformulated as a Nasalization process (cf. Streitberg 1896 §93). 5. 18 TM Stattu of e Another potential problem for the segment-letter match in §5.6 involves the status of the peculiarly Germanic * iZ. There is no problem with *� which bears the reflexes of lE ·leI ( *s e-ti-s [cf. Lat s e-vi"l sowed", semen " seed"] > Goth. se ps, OE s iiJd > seetl) Throughout Northwest Germanic the reflexes of this vowel were kept distinct from a new higher vowel. traditional *t#. It is fair to say that there is extremely little agreement among scholars on the origin and development of this vowel in Germanic, and this is not the place for a lengthy digression on this topic, so I will concentrate on areas of general agreement. It has long been realized that this high, tense, close aIel was a Germanic innovation (cr. Streitberg 1896 §79). As noted by Streitberg, the new vowel was categorially limited. It occurs in only two words with any claim to antiquity within Germanic (Streitberg 1896:65): Ooth. fera, OHO fera, feara. fiara "side", of unknown origin, and ·hir "here" > Goth., ON, OE, OS her, OHG hiarlhear. Kurylowicz (1952) has explained the latter essentially as a new lengthened grade based on alternations like OE seIse "this; he", helhe "he", etc., whence the vowel of hit' was held in place by the alternation he : M(r) ,l9 The second category is the NW Germanic preterit of the 7th class, e.g., OE llitan : let "let", replacing an older reduplication pattern, in Ooth. letan : lailot. One possibility here (so already Streitberg 1896 §79(3); er. Kurylowicz 1952) involves contraction andlor compensatory lengthening, viz. *he-hait (Ooth. hai-hait) "named" > OE hehtlhtit. etc. The details of this formation remain murky despite much recent attention (e.g Fulk 1987; Kortlandt 1 99 1 ) . Nevertheless. one of the sources of the new vowel is unequivocally compensatory lengthening: *mizd () (Ooth. mizdo) "reward" > .
.•
19 The traditional account deriving .Mr from .ktir (cf. Streitberg 1896 §79. 1) is reiterated in Voyles (1992:72-74, and §2.33). Jay Jasanoff (E-mail message of 22 Det 1993) supports a variant of the morphological solution, comparing ORG th ... OE par "there" vs. (with short vowel) Ooth. (1), ON par "there", ORG thara "thither". "The long .. then, must have origiDally been an expressive variant [. . .]." ORG hera "hither" (> Germ. her) has a short le! vs. the .e of her/hiar/hier "here". "The conclusion naturally suggests itself that [ ...1 Gmc.. • h � is a deictically lengthened vatriant of .her." To that I would only add that, quite naturally, the alternation with helM, etc., could have been a contributing factor.
82
ANCIENT SCRIPTS AND PHONOLOGICAL KNOWLEDGE
*merdu > OE meord ( Ix) > med, Eng. meed; *kizn- "fir, pine tree" > *kern > OB cen "torch". As also noted by Streitberg, the status of the new vowel was reinforced by loanwords, e. g., Vulgar Lat. m e;a « Lat. m msa) "table" > Goth. mes, OE me;e, etc. Another contributing factor may involve *ei before a low vowel, in which case OHG stiaga "stairs" (vs. stigan "to climb") would result from different leveling processes (see Antonsen 1972: 1 3 1 w. lit.). Voyles ( 1992, §3.4.2) derives this form also from a long diphthong. Whatever the details, it seems reasonably clear that in Gothic the new vowel fell together with inherited */�; cf. Goth. her "here" and loanwords like Vulgar Lat. Gr«:us "Greek" > Goth. Kreks, OE Crec, etc. If we assume that spellings like haihail "called" represent [hehe:t], with I £:1 from *ai (see discussion in Vennemann 197 1 : 1 1 1-126), then this was even lower than the inherited *leI, which evidently had a higher realization in Gothic than in the rest of Germanic (cf. §5. 14).
5. 19 Proto-Ger"",nic Vowels and the Runic Alphabet To conclude this discussion, the evidence is substantial that the Proto Germanic long vowel system reconstructed in § §5.6, 5 . 1 4 is correct. The Gothic-specific monophthongization of *ai, *au to Is /, 131 is irrelevant, and there is no evidence in Gothic for the new *ez, i.e., for a vowel any different from the inherited front mid vowel. The evidence is good that the new higher vowel originated through contraction, compensatory lengthening. and a derivational process involving a new lengthened grade that affected several morphological categories, especially preterits of the seventh class of strong verbs. Given the new contrast in front vowels, loanwords with higher front vowels could now be accommodated with the new vowel. As to the matching of runic letters and Proto-Germanic vowels, even if one assumes that the new vowel *e2 was post-Gothic, there remains the potential problem that creation of a new *laJ (§5. 17) may have been pre Gothic, i.e., within Proto-Germanic. At the same time, as emphasized by Vennemann ( 1971 : 104), it is a peculiar phoneme in Gothic in that, in native words, "it occurs only before fhl and very infrequently." In other words, it is sufficiently new in Germanic that its phonemic status has not yet been reinforced (cf. Antonsen 1972: 127). This suggests that, if Nasal Deletion ( 135) was indeed Proto-Germanic, it could not have been more than a century before the first Gothic attestations. While it does not prove anything, being from a different dialect, runic hilzai "on a steed" (ORI 1 1: Moj bro stone, Uppland, Sweden, 3(0), from *konk-oy (cf. Lith. Sanku.r "fast"), lends some credence to the suggesion that Nasal Deletion was Proto-Germanic, albeit ,
83
1HE RUNIC ALPHABEf
somewhat late.20 On the other hand, if the testimony of the First Grammatical Treatise
( §5. 17) is to be trusted, it seems preferable to formulate Nasal
Deletion as Nasalization and to accept that long nasal(ized) vowels subsisted into the dialects, in which case the same vowel letters were used to represent non-nasal and nasal vowels, as likely in hahai "on a steed" on the Mojbro stone.2 1 If that is the correct interpretation, it continues to follow that the runic alphabet was created sometime between Stage I of new lil
( 1 1) and the innovation
[± nasal].
This overview of the early development of the Germanic long vowels
confirms a safe period of several centuries between Stage 1
( 1 1) and the
changes that brought in first new -'iI, then new -leI. Turning the argument around, given the inventory of runic vowel letters in
§§5. 1 , 5.6, it is clear that
a matching set of those letters with vowel phonemes could only have been made during the time when Germanic had four short and four long vowel phonemes, i.e., in Proto-Germanic, after the changes in before Nasal Deletion with compensatory lengthening
( 1 1) and ( 12), and ( 13-5). At any time
later than that, some important vowel phonemes would not have been represented by that inventory of letters.
20 Significantly. Voyles ( 1992:60-61) gives Nasal Deletion as the last in a set of changes between -400 and 200. Antonsen (FAX of 30 Nov. 1993). while confuming that hahai pr bably contains a long nasalized vowel. simultaneously claims that "The lack of designation of 1nl is simply a consequence of the runic tradition of not designating nasals before obstruents" (§5.3.2). This seems to imply that he does not believe Nasal Deletion ( 13-5) has applied. Technically. of course, the form is ambiguous. In an E-Mail message (14 Feb. 1994). Antonsen clarifies that by non-phonemic. he means all cases of smface long nasalized vowels continued to be derived from underlying nasal consonant plus fricative. Throughout this study I have used phonemic in the sense of Schane (1971), i.e., involving a surface contrast. whether the segments continue to be derived or are lexicalized. In other words. Antonsen accepts Nasal Deletion (13-5). but claims that its output yielded few. if any. lexical contrasts. 2 1 I accept. with Hollifield (1980: 150). that -ai was still a diphthong until after ca. 400 (cf. Bammesberger 1991).
6. LITERACY AND LINGUISTIC KNOWLEDGE 6.0 Introduction This chapter presents evidence for the implicit knowledge of words, affixes, syllables, and segments, independent of any kind of script. It is argued that experiments that suggest the contrary are deficient in their design and results because they are in fact testing explicit knowledge, or even meta knowledge, and consequently do not begin to broach what native speakers do with their language spontaneously, much less what they know about it implicitly. There can be little doubt that literacy and knowledge of a script and its conventions influence judgments about linguistic units, sometimes in a manner that is contrary to speakers' implicit linguistic knowledge. But what is generally not taken into consideration in discussions of scripts and literacy is the implicit linguistic knowledge that underlies the development and use of scripts to begin with, in particular, the knowledge of segments that was coded in the ancient Western scripts. 6. 1 The Word What is a WORD? There is a body of conventional thought that the word is difficult, if not impossible, to define. See, for instance, the discussion and references in Coulmas ( 1989:39-40). Before discussing some of the mooem research on this topic, let us advance the hypothesis that part of the reason for the difficulty is that word can be defined in different ways, by reference to different parts of the grammar. The thorough discussion by Di Sciullo & Williams (1987) notes the following definitions of WORD that have appeared in the technical linguistic literature: 1. MORPHOLOGICAL OBJECf . The word has traditionally been defined by a set of atoms ( 'morphemes' ) plus rules of combination (affixation, com pounding, etc.). Halle (1973) pointed out that every native speaker knows that un-drink-ahle is a possible word, but *un-able-drink, *drink-un-able are not (cr. Scalise 1984:24). The central task of morphology is then to ascertain the laws of fonn that determine membership in this set. As the plethora of recent
86
ANCIENT SCRIPTS AND PHONOLOGICAL KNOWlEDGE
theories (see, e.g., Spencer 1991, Anderson 1992, Lieber 1 992, Miller 1993, Stonham 1994) shows, this is not easy. 2. SYNTACTIC ATOM. The word, as a syntactic element, is, for instance, the head of XP, i.e., the head of some phrase (the head noun of a noun phrase, the head verb of a verb phrase, etc.), insertable into XO slots in syntactic structure (see Chomsky 1986). Syntactic word and morphological word are not necessarily coterminous. Not all items insertable into syntactic structure have morphological substance (e.g., PRO), and some syntactic heads consist of more than one morphological word (New York, Humpty Dumpty), a compound (China syndrome), frozen phrase (jack-in-the-box), or sentence (a how-they-do-it book). See the discussion in Miller ( 1993, chaps. 1 , 3, 4). 3. LISTEME (a word coined by Di Sciullo & WiIliams), the listed units of language. If Iistedness is the same as idiosyncratic, listemes are of no interest The problem is, what does Iistedness entail? Different scholars define listed in different ways, entailing different implications (see Miller 1993). Since various grammatical constructs are subject to listing, Iisteme is not the same as morphological object or syntactic atom. 4. PHONOLOGICAL UNIT (defined by stress and other phonological pro perties, such as restrictions on word-final consonants and clusters; see Levin 1 985; Inkelas 1990) includes things like c1itics and their contractions, e.g., I'll, should've. These phonological units are not the same as morphological objects because ( l ) they cannot be formed by morphological rules/principles (I'll is not just 1 + 11, like seedy is seed + y), and (2) contractions like I'll are produced by a phonological 'welding' of I will. With metrical phonology (see Goldsmith 1990; cf. Spencer 1991, chap.5; Lieber 1992, chap.5), there is agreement that c1itics (Spencer 1991:350-394; Anderson 1992, chap.S; 19(3) are prosodicaIly deficient metrical fragments that must be incorporated into the metrical structure of an adjacent host (cf. Anderson 1988: 165ff; Inkelas 1990). Finally, such phonological units are not equivalent to syntactic atoms because they do not undergo movement as a unit, as shown in ( 1) . ( 1) Phonological Unit ;t Syntactic Atom (a) Dana thinks (that) who will win (b) Dana thinks (that) who' ll win (c) who does Dana think ! will win (d) *who'lI does Dana think ! win The who will of ( la) appears contracted in ( l b) and with WH-movement in ( l c) , where the ! (= 'trace' ) marks the position from which who moved. What is important is that the contracted form who 'll cannot move as a single
llTERACY AND LINGUISTIC KNOWLEDGE
87
syntactic unit, as shown by (Id). That implies that syntax only sees the string who will and that contraction to who'll occurs later, as a kind of phonological operation. When such contractions become lexicalized (listed), they can be moved as a unit, as shown in (2). (2) Listed Phonological Units and Syntactic Atoms (a) you should have done that (b) you should've done that (c) should you have done that (d) should' ve you done that [regional] The should have of (2a) appears contracted in (2b) and with what has traditionally bee n called Aux-inversion in (2c). The inversion of the total contracted unit (2d), which is still unacceptable to many speakers, was not possible until should've evolved (regionally) into a separately listed fonn that could be selected as a syntactic atom (cf. Joseph 1992: 135ff, on let's and have to).
6.2 Word Bou.ndtzrie. Confusion about what a word is stems from the different perspectives (above), according to which it must be considered linguistically. That is, a word in the lexicon (defined here for simplicity as the repository of the idiosyncratic) is obviously different from the morphological word with all of its derivations d inflections, and that is different from the output word with all that plus de ved phonological properties. That in turn differs from what aspect of a wo or combination of words (compound etc.) is relevant as a syntactic head, and so on. This does not mean that the word is difficult or impossible to define, unless one insists on a monolithic view from a single perspective. To the contrary, the fact that the word has different properties in different parts of the grammar facilitates an explanation of the apparent confusion in people' s minds. All of this is part of a speaker' s implicit knowledge. The confusion stems from attempts to make this knowledge explicit, resulting in a focus on one or another view of the word, giving the impression of inconsistency or inability to 'identify' a word. Bearing in mind the problem of what aspect of a word one might make explicit when asked for 'a (unique) definition' , some modem research can now be discussed. One problem is that the experiments often make assumptions based on English orthography that are not necessarily linguistically sound. Scholes ( l 993b:85-86) reports a finding that little over 2/3 of second graders agree that the is a word. Given the possible use of word to mean phonological
�
88
ANCIFNT SCRIPTS AND PHONOLOGICAL KNOWLEDGE
word, and the status of clitics as part of the larger phonological (metrical) unit which includes the head (noun, in this case), in stating that the is not a 'word' children may be either separating out functional elements from lexical (Abney 1987; Miller 1993, chap.4) and/or recognizing the normal clitic status of articles (cf. SPE 366ff; Miller 1977 §3). Either way, they are stating an
implicit awareness of the problem of treating the on a par with a major lexical category item. It is interesting that the Ancient Greek syllabic scripts also wrote clitics together with the head (§3 . 1 etc., above), and the accent-marking tradition used the so-called grave accent to indicate metrical subordination (Wackemagel 1893 ; Jakobson 1937:264-265; Sommerstein 1973: 161 ; Miller 1976c: 16fO. Languages differ on what function words can be clitics or hosts
(Kaisse 1985; Inkelas 1990, esp. chap.S).
Tasks experimenters assign to children are not only ambiguous (Mann 1991:55-56) but presuppose a knowledge of English orthographic conven tions, which are notoriously haphazard. One need only consider the different ways of representing compounds: wordword ( bookstore), word-word (stage manager), word word (China syndrome). Since adults frequently hesitate and have to look up the specific 'spelling' of particular words, it seems ridiculous to imagine that children would/should find the conventions any less arbitrary. 6.3
Knowkdge of Word Con.tituency Scholes ( l993b) perpetuates the age-old prejudice that people have no
notion of word without the aid of writing. Just how/why people suddenly realize they can write words separately is not addressed. He also misses the point that by the lexical and/or syntactic definition, it is impossible not to know what a word is. He even cites counterevidence from Saenger ( 1991), who rightly insists, based on ancient scripts, that the word precedes writing. Nor does he address the issue of how people know what modifies what if there is no concept of the word, or, for that matter, how speakers know which words to put which derivational and inflectional affixes on, if they do not know what a word is to begin with. Based on prior experiments by himself and Brenda Willis, Scholes maintains that illiterate speakers cannot segment words like dishonesty into their constituent parts. It is true that speakers have difficulties with non productive morphology, but Scholes does not test productive morphology (discussion in Miller 1993:4ff etc., w. lit.). Since all speakers freely coin and understand new words involving productive derivation and inflection, they must have the implicit knowledge. Experimenters have a threefold problem: ( 1) separating productive and non-productive morphology; (2) getting the child to understand the nature of the task; and (3) designing the experiment in
UTERACY AND LINGUISTIC KNOWLEDGE
89
such a way as not to require metaiinguistic transfer from implicit to explicit knowledge (i.e., test implicit knowledge). Scholes in fact borders dangerously on claiming that all forms of all words are memorized when he says (p.84), "If, then, word cannot be defined, the idea that some words have (stems and) affixes becomes at best suspect if not patently vacuous." How, then, do speakers recognize a new word, a word they have never heard before? How do they know hqw to inflect a neologism? Why would anyone hearing the verb we for the first time (recently backformed1 from laser) automatically know that it can have a past tense and that it is wed? For that matter, how, if speakers do not know what a word is, do they create a verb like lase to underlie laser to begin with? . Daniels ( 1 992:89) rightly insists on knowledge of the word, based, for a change, not on experimental results, but on direct observation of speech situations, with self-corrections, and requests for repetition. One might also inquire how, if people do not know what a word is, they could ever question the meaning of a word they have never heard before. A preliterate four-year old was overheard to ask, "What does [stu] mean?" (asking about eschew). If people cannot recognize (as sucll 0rds they know, how could they ever detect the presence of an unknown w in a continuous stream of speech? Since speakers clearly identify old, w, known, and unknown words, it follows that the experiments of Scholes (and others) are faulty in their design and results. They do not begin to get at what native speakers necessarily know implicitly in order to create, understand, and correctly derive and inflect new words. They do a major disservice to the linguistics community in conveying to the population at large misleading information about linguistic knowledge, thereby pretending to validate under pseudoscientific guise age old stereotypes and prejudices.
::E
6.4 Knowledg. ofWonb in Antiquity Scholes' claim ( l993b) that word cannot be defined apart from writing systems is also misleading. Again. how do speakers of languages that have never had a script derive, understand. and properly inflect new words? It is also misleading to assert (p.84) that "languages without literary traditions [ .] have no words for word." In reality, languages with literary traditions that clearly recognize words frequently do not have an unambiguous word for word. Greek AOYoc; lOgos only has "word" for one of a very broad range of meanings. including "reckoning. account(ing) ; explanation; principle; reason; narrative. speech" etc. (nearly six columns in Liddell-Scott-Jones, A Greek..
1 For a m:ent discussion of backformalioo , see Miller (1993: 1 10- 1 15).
90
ANCIENT SCRlPfS AND PHONOLOOICAL KNOWLEDGE
English Lexicon). Compare the more technical tenns OYOlla onoma ' cramsay), I told her to put thing into Pig Latin. Her output was ingthay, by her implicit knowledge, not hingtay by her orthographic contamination. All of this shows that literacy/illiteracy has little to do with the problem of teaching speakers how to consciously access their implicit knowledge. All one can say at this point is that some speakers are better at it than others, and that it has nothing to do with literacy, but what determines it is not yet clear. Explicit knowledge differs crucially from implicit in that 10 Maranzana (1993) emphasizes that the results typically attained do, however, indicate true
phonological awareness. A typical response is that cal contains two parts, [ke] and [et],
which illustrates the syllabic organization of the segments.
100
ANCIENT SCRIPTS AND PHONOLOOICAL KNOWLEDGE
implicit is not subject to such differences in individual competence. Raising implicit knowledge to the level of explicit seems to require extra talents. Gombert ( 1992: 8-13 et pass.) distinguishes metalinguistic activity (cognition about language) from epilinguistic, which is not consciously monitored. These correspond roughly to Chomsky' s explicit and implicit knowledge, respectively (see Chomsky 1986). Gombert plots a continuum from epilinguistic to metalinguistic processes, illustrating how fluid the transition can be from implicit to explicit knowledge (cf. Birdsong 1989; Maranzana 1993). To avoid confusion, it must also be emphasized, with Maranzana ( 1993), that explicit (Gombert' s metaiinguistic) knowledge is not the same thing as the linguist' s metaknowledge. Although native speakers necessarily know the phonological system of their language (or they could not say/understand anything), and some of that knowledge may be explicit! conscious, "they may not be able to articulate how the system works, in which case their metaiinguistic knowledge would not be equivalent to their linguistic knowledge" (Maranzana 1993: 18). There is a three-way distinction between implicit linguistic knowledge, conscious knowledge, and metaknow ledge, that is not covered by ei ther Chomsky' s or Gombert' s binary theories.
6. 13 Implkit SegmenflJl Awarenell The following evidence has been adduced for segments as part of every one' s implicit linguistic knowledge (cf. Birdsong 1989; Miller 1990: 175): 1) Toddlers (age 1-2) perform segment substitutions (/pal, /baI, /dal, Igal, etc.), show clear signs of word recognition, and make segment insertions, harmonic processes, and other operations within words (Stampe 1980; Slobin 1985a; Gleitman et al. 1988 ; Fowler 1991 ; Stemberger 1993). 2) Preliterate children have many kinds of assimilation and substitution processes, which would be impossible without implicit knowledge of what and where segments are (Kiparsky & Menn 1977; Stampe 1980; Gleitman et al. 1988 ; Stemberger 1993; pace Fowler 1991). 3) Operations of all sorts presuppose (implici t) knowledge of segments or they could never be applied to words never heard before. Specifically relevant to the problem of initial clusters, many languages (including PIE and Greek) have reduplicative processes that 'pluck out' one or more consonants of the cluster (see Steriade 1982, 1988). In order to do this in a systematic manner, speakers must necessarily know (implicitly, at least) what segments are present and be able to manipulate them linguistically. Again, this cannot simply be memorized, or it could never be applied to new words. 4) In all cultures, literate and illiterate, there are word games and 'secret languages' that feature the coinage or substitution of 'code' words, and/or
UTERACY AND LINGUISTIC KNOWLEDGE
101
games of the ' Pig Latin' variety which metathesize consonants, vowels, or syllables, insert segments between consonants, and so on, within individual words (see Kirshenblatt-Gimblett 1976; V ip 1982; Mann 1991). Especially interesting is the dialect of Pig Latin that extracts the first onset segment, viz. scram .... [k�msey]. How do illiterates and preliterate children do this if they cannot identify words, cluster constituents, or segments? Such word games presuppose an ability to play with individual segments on a nearly conscious (explicit) level. Before we could read/spell/write, my friends and I played a word game that inserted a copy of the root vowel after every consonant, viz. strap .... /sretre�pre/, presupposing a knowledge of the individual segments in clusters. Bagemihl ( 1987:36) claims that segment-based games presuppose alphabetic knowledge (cf. Faber 1992: 1 16), and Bellamy ( 1989) asserts that segments in clusters cannot be heard. How do children learn them in the first place? 1 1 Segment-substitution games in numerous languages are played by Iiterates and illiterates alike (Mann 1991:59f1). McCarthy ( 1 985) documents an intricate segment-based game played by prostitutes in Addis Ababa. 5) Speech errors that anticipate/switch segments show that words/fonns are stored in our brain and articulated by some sort of segmental represen tation; cf. shried frimp for fried shrimp, trong and slong for long and strong , Skan Tenton for Stan Kenton, frish gotto forfish grono, etc. (Fromkin 1988) . Moreover, these exhibit onset-rhyme coding (Treiman & Zukowski 1991:70). 6) Part of our knowledge of English as native speakers is that strup, spUm, blark, stalm are POSSIBLE (but nonexisting) words, but that sbUsh, sknap, bnik, tsaml are not possible words. Since these are precisely not ' real' (existing) words of English, it is not memorization of a list that detennines this, but (implicit) knowledge of segments and their organization (cf. Halle 1962; SPE 3f1>ff etpass . ; § 1.3-1.4 above). That speakers are puzzled when asked about such facts is not surprising, given that language knowledge is implicit rather than explicit. Very young (preliterate) children know far more about linguistic representations than the experimenters would predict. 12 Speakers learn all sorts of facts that are not coded anywhere in the writing system. With or without writing, then, the facts of a language are acquired by every native speaker. 1 1 There is evidence that onsets may be acquired initially as such before they are decomposed segmentally (freiman & Zukowski 1991:71) but, again, there is the question of explaining the task to a child that young, whence the problem of eliciting unequivocal experimental results. 1 2 See Kiparsky & Menn (1977) ; Stampe ( 1980); Gleitman et al. ( 1988) ; Roeper ( 1988); Birdsong (1989); Treiman & Zukowski. (1991); Carlisle (199 1) ; Stemberger ( 1993); and various sdections in Bloom (1994)..
1 02
ANCIENT S CRIPfS AND PHONOLOGICAL KNOWLEDGE
6. 14 Conclurion This chapter has reviewed the major evidence for implicit knowledge of linguistic units and demonstrated that experimental results do not come close to accessing that knowledge. In the words of Maranzana ( 1993: 1-2): Although the literacy debate is cross-disciplinary. the arguments can be shown to be founded upon presumably linguistic givens which are unsupported by linguistic theory and lack empirical validation. [...] the implicit underpinnings of the literacy debate are Nor linguistic. psychological or even pedagogical issues as they are most often presented but instead are issues of cultural dominance and exporta tion. technological determinism. and intellectual superiority.
Quite simply, Maranzana ( 1 993:3) states, "metalinguistic performance is a capacity which CANNOT be proven to reflect grammatical competence". Nevertheless, misled by the recent pseudoscientific experimental 'evidence' , some researchers have actually proclaimed that all linguistic knowledge is imparted by the script. Bugarski ( 1993: 15) asserts that "Basic linguistic constructs - phoneme, morpheme, word, sentence, grammar, language, and so on - are all [ . ] determined by the written mode of linguistic expression." Similarly, Patel ( 1 993: 203) succumbs to the bias that "The metalinguistic ability to segment continuous speech into sentences, phrases, words, sylla bles, and phonemic units is considered to be part of literate cognition." Such an extreme position is untenable, as noted throughout. Even the less extreme position that segmental knowledge results from alphabetic is untenable. The most one can say is that alphabetic knowledge helps to transfer implicit segmental knowledge to a level of conscious awareness, but even that is not the same as explicit metaknowledge (§6. 1Off). Moreover, the less extreme position is belied in preceding chapters by the varied knowledge represented in ancient syllabic scripts and the sophisticated knowledge of features that went into the construction of the syllabic and alphabetic scripts. In a discussion of the phonemic principle Roy Harris ( 1986: 104) adds the disclaimer that "it is far from clear [ ] that the inventors of the alphabet were inspired by anything like modem phoneme theory." In fact, this is a bias, given ( 1) the psychological and linguistic reality of contrasting segments, (2) the fact that segments were encoded in all of the (non-pictographic) Western scripts from antiquity to the present day, and (3) there is ample evidence in support of a strong version of ARTICULATORY ICONlOlY , the idea that letters may be devised iconic to distinctive/contrastive articulator (lip, tongue) positions (Harris 1986:93). Matrices from Ras Shamra and Byblos ( §S. IOff) to the runic fupark (§S.8-S. 12) and the First Grammatical Treatise (§S.5) , support articulatory iconicity as the origin of segmentally coded scripts. ..
...
7. IMPLICATIONS: AN IDEAL SCRIPT? 7.0 Introduction It has been observed that syllables are always part of conscious (explicit) knowledge while segmental knowledge, by contrast, is implicit but not necessarily explicit. This goes a long way toward explaining the ubiquity of syllabaries vis-A-vis the rarity of segmental scripts, but simultaneously raises several problems. The first involves our observations in earlier chapters regarding the clear representation of segments in syllabaries. If only syllables are salient, why is it that syllabaries normally code segments? Secondly, is there such a thing as an ideal script, and what might it be? Naturally, function is a major consideration. If the intent is to create a readily learnable script, it is suggested that a syllabary is the closest to ideal, but given the limitations imposed by the form of the script, it is not easy to design one that avoids the awkwardness of the ancient syllabaries. Consequently, it is suggested that those disadvantages constituted the primary motivation for the shift to the alphabet, a compromise between the vowelless scripts and the syllabaries, but that there were more efficient ways the goal could have been accomplished that would have been more consistent with what is currently known about human phonological knowledge. 7. 1 Realities to be DeaIJ WUh As emphasized by Daniels (1992) , there must be some reason syllabaries were created in numerous places but the alphabet was not. This study has emphasized that there is something 'unnatural ' about alphabets in that they attempt to represent phonological knowledge in a strictly linear manner, in apparent violation of our (at least implicit) knowledge of higher levels of organization. Not fortuitously, this correlates with the modern research which shows that syllables are always 'accessible' (in the sense of Maranzana), l i.e., part of explicit (conscious) awareness, while segments belong to implicit knowledge and may or may not be consciously accessible.
1
Maranzaoa (1993: 13-14, 60, etc.) defIDes
absence of specific training.
accessible as explicit COI1SCious awareness in the
104
ANCIFNf SCRIPTS AND PHONOLOGICAL KNOWLEDGE
What about all the evidence for the antiquity of segmental knowledge, including the segmental coding of syllabaries? To state an obvious fact at the outset, since syllabaries from all over the world are segmentally coded, that proves that the script inventor(s) in each case had segmental knowledge as well as syllabic knowledge. However, the fact remains, those scripts were invariably syllable-based, implying a conceptualization in which segments do not exist apart from their syllabic organization. The analysis of the Sonority Hierarchy as a ' syllable-structure template' (chap. 1) supports this conclusion. 7.2 Script Abstractness and Phonological Cues The discussion so far suggests that the early desyllabarizing scripts (those that removed all vowels), such as the Egyptian hieroglyphic (Davies 1987:30ff; Healey 1990: 16; Powell 1991 :76-88) and the Northwest Semitic, or Phoenician, script (§§ 4.2, 4.5), were in fact more abstract than any other (phonologically coded) system to date. Not only was there total abandonment of the syllabary principle that segments do not exist apart from their function as syllable onset or coda, but also, an antithetical countersyllabic principle was adopted to replace the syllabary principle. The result was a degree of abstraction that is still unrivalled, and which had the added advantage of being a useful heuristic for representing roots in Semitic languages (§4. 1 1). While adopting a countersyllabic approach, the vowel-removers in fact substituted another level of programming, the word, since the only way such a script can be read is logographically: the configuration klb is read as (some form of) "write", qtl as (some form of) "kill", and so on. Such a system is advantageous in that there are (some) phonological cues to the recognition of the word, as English dove (bird) and dove (= dived) must be read holistically rather than phonemically (cf. §O.2), but with phonological cues for root/word recognition. 2 It is of course disadvantageous in that it can only be read by people who already know the language (but scripts are designed for native speakers, not for foreigners to learn the language; cr. SPE, p.49). While segments were abstracted in the Egyptian and Northwest Semitic/Phoenician scripts, they were not strictly segmental in the sense that there is no way texts could be read purely segmentally (phonemically). 7.3 The Alphabet Compromise VI. More [Mal Scripts By combining the syllabary principle with the Northwest Semitic script, a number of traditions, essentially from Greece to India, created a script that 2 Such systems are not uncommon. The Mayan glyphs are morphemic with a phonological (syllabic) component (Lo\DlSbury 1989; Coe 1992).
IMPIlCATIONS
105
was better adapted to the morpholphonological structure of those languages, with the extra advantage that clusters could be represented in a simpler fashion (§4.13), and non-native speakers could more easily learn to read the language (given the mixed populations of Crete and Cyprus, for instance, that could have been one realistic issue). Another way to view the syllabaries versus the consonantal scripts is as an 'all-or-nothing' situation, which resulted in an advantageous compromise. The compromise, unfortunately, was not very good because it represented syllable constituents, not as constituents of syllables, but rather as independent segments. The very success of the alphabet was simultaneously its maximal failure. Maranzana ( l993:6lff) collects a large amount of modem research which affirms that it is easier for children initially to learn a logographic script than a 'phonemic' script. possibly suggesting3 that explicit (conscious) segmental awareness, if it develops at all, is a function (at least in part) of the maturation process. Since, on the other hand, syllables are readily accessible (explicit/conscious) to very young children, the maximally natural, advantageous, non-pictographic/non-Iogographic script would be a syllabary without the disadvantages of a syllabary, i.e., one that can represent [stra] in some more efficient way than sa-Ia-ra. In other words, what the studies suggest is that a good writing system would allow for onset and rime (nucleus plus coda) representations of the syllable structure of words (plus a way of representing phrases). One of the closest to the ideal of representing at least onset clusters as such is the Indic devan�arl script (§ 1 . 1 ). 7.4 Devising an EJJicienl Sy/Jabary Another consistent inconsistency of the ancient syllabaries (chaps. 2, 3) was their insistence on representing initial vowels or vowels in isolation differently from the same vowels with onsets. In other words, the signs for [a], [tal, [lea], etc., had nothing in common (of a non-fortuitous nature). More over, in [tal, [te], [ti], [to], [tu], there was no constant element that would represent [t] . Thus, the scripts themselves reveal less segmental knowledge than do the conventions for spelling syllables with onset and/or coda clusters. The ideal , then, would combine the syllabary principle with the knowledge of syllable constituency, again, like the devaniigarI script, among others. In reality, the experiments prove nothing since they are based primarily on English, with one of the least phonemic orthographies. What is needed is more research on scripts that are more phonemic. On the other hand, modem research shows that "Readers learn associations between visual forms of words and their semantic referents. In 'sight reading', the words are stored in memory as visual gestalts" (Maranzana 1993:64 w. lit.). That would explain the observation that (limited) logographic systems are among the easiest to acquire.
3
106
ANCIENT SCRIPfS AND PHONOLOGICAL KNOWLEDGE
In the ideal situation, as emphasized by Maranzana ( 1993), what she has called the alphabet paradigm (alphabetocentrism), should be replaced by something that is more generally accepted by linguists, psychologists, and reading specialists. To begin with, the alphabet should be portrayed honestly as the ' mixed bag' it is. Then, for a variety of reasons, noted especially in Chapters 1 and 6, everyone agrees on the salience of syllables. We have noted throughout a number of advantages of syllabaries, especially the closer representation of what native speakers know implicitly (though they may not be able to articulate that knowledge) about the hierarchical organization of speech sounds. The ideal, phonologically-based orthography, then, would be a syllabary, but one without the disadvantage of a plethora of symbols. That would mean one in which vowels are consistently segmented out to the extent that vowels in isolation and with onsets would have the same representation. Consonants likewise would have a consistent representation with deference to onset and coda positions. There is no evidence as yet for how coda clusters should be represented. More studies are needed to concentrate on the aware ness of consonantal segments specifically as onsets and codas of syllables, rather than as independent elements. 7.5 Reprise and Conclusion Some recent experimental work (cf. the papers in Scholes 1993) suggests that nonliterate speakers do not have segmental knowledge and that only syllabic knowledge is ' real' or accessible, whence the ubiquity of syllabaries (cf. Daniels 1992, Faber 1992). The present work emphasizes that: 1) There is a difference between implicit, explicit, and metalinguistic knowledge. Maranzana ( 1993) shows that explicit (Gombert' s metalinguistic) knowledge is not the same as the linguist' s metaknowledge because speakers are typically unable to articulate how the system works. 2) Experiments have so far tested metalinguistic knowledge. Maranzana ( 1993) emphasizes that the results of experiments do, however, support true phonological awareness. A typical response to a Scholes-type test is that cat contains two parts, [kre] and [ret], which illustrates the syllabic organization of the segments. The conclusions generally drawn from such experiments are faulty because experimenters have failed to comprehend the true implications of the results and to distinguish implicit from explicit knowledge. 3) There is empirical evidence from language acquisition, use, and change for the psychological and (implicit) linguistic reality of contrasting segments. Preliterate children have all sorts of substitution processes, which would be impossible without implicit knowledge of what and where segments are. There are also language games (in literate' and nonliterate societies) that
IMPUCATIONS
107
manipulate segments, speech errors that switch segments, and evidence from language change for the reversal of a process in the inducing environment in which contrast is lost, providing evidence for segmental contrast and there fore for segments. But that only means they are implicit. They may or may not be part of explicit (conscious) knowledge. 4) Segments obeying the Sonority Hierarchy (SH) are coded in all of the (non-pictographic) Western scripts from antiquity to the present day. This study supports the arguments of Steriade ( 1982) and others that Linear B spelling reflects (at least implicit) knowledge of the arrangement of segments according to the SH, and goes on to demonstrate that the knowledge was quite sophisticated in devising an ingenious solution to the dilemma of what to do about onset clusters in coda position or the converse. The Cyprian syllabary, likewise, was based directly on the SH. Solutions to problems involving coda clusters in onset position, SH violations in the language, occasional attempts to represent compositional information, etc., go I ightyears beyond anything predicted by Daniels, Faber, and others, and reveal that syllabaries typically represent more linguistic knowledge than alphabets do. 5) There is ample evidence consisting of a series of phonetic-based matrices from Ras Shamra and Byblos to the runic fupork and the 13th century Icelandic First Grammatical Treatise that support a strong version of articulatory iconicity, the principle that letters may be devised iconic to distinctive/contrastive articulator (lip, tongue) positions (Hams 1986:93). The Byblos Matrix is arranged: laryngeals > bilabials > alveolars > velars > dentals, and has 22 segments and 8 open slots (gaps in the phonological system). The Ras Shamra Matrix has 27 letters with 21 empty cells (cr. the Indic and Korean Han'glil scripts, for arrangement according to phonetic features). The Runic Matrix is arranged labial > dental > a1veopalatal > velar, and has 24 letters with 1 1 empty cells, five double occupancies, and one slot with three phonemes. This can be reduced by recognizing, within a given place feature, independent projections of consonants and vowels, a non problem for the Proto-Canaanite scripts. It can hardly be accidental that, of the 6 multiple occupancies, four are paired C and V sets, viz. flu, air, nli, yl a The triply-filled slot has the two strident sibilants /s zI along with the coronal stop It!. The suggestion is that the inventor(s) of the runic alphabet viewed class and place features as independent planes. 6) Different scripts have different advantages and disadvantages. The best evidence suggests that a script must contain lexical information, but how that is to be accomplished is not clear. If we opt for a script that also codes some phonological knowledge, the ideal appears to be one based on the
1�
ANCIENT SCRIPTS AND PHONOLOGICAL KNOWLEDGE
syllabic principle (arrangement of segments according to onset and coda position) , but without the customary disadvantages of syllabaries. A truly complete script, representing all of our phonological knowledge, even if theoretically possible, would be too cumbersomely inefficient and confusing to read. Since reading is primarily done with long-term memory, in which a word (or some other unit) is treated as a visual gestalt (Henderson 1992; Taft 1992; Maranzana 1993; articles in Willows, Kruk & Corcos 1993a), all that is needed is a quick phonological c(l)ue to the recognition of the gestalt. 7) The alphabet was a compromise between the vowelless West Semitic (Phoenician) script and the very cumbersome syllabaries that duplicated the nucleus for each clustering consonant and failed to factor out identical segments. But it was not a very good compromise. One problem with the alphabet is that there is more phonological information than is needed and it appears to inhibit the speed of gestalt recognition. (For the largely irrelevant details it contains, note the frequent reaction on writing a word and deciding that it does not ' look right' .) Another problem with the alphabet, signalled throughout, is that it attempts to represent segments on a strictly linear plane in violation of our (to some extent explicit) knowledge of their organization into syllables.
APPENDIX A From Proto-Sinaitic to Greek earlv name
Proto-Sinaitic
1>
1llp- "ox-head"
(;y �
2>
bet- "house"
0
3>
gaml- "throwstick"
�
4:-
digg- "fish" 1
e>-
110- "man calliDlz" 1
1:
)
6>
wo ( waw) "mace"
7>
ze(n-) 11
8>
Qi(I-) "fence" 1
9>
fi(I-) "spindle" 1
NW SemitidCanaanite
tJ £7
IJ
IIT
?
1
f\
=I y
D.
Y
1 1 ::>
Et)
A
A
aNpa
d a 7
1 B
B
�liTa
r
yalllla
l:J.
8EhTa
E
E
F
8(yallll a
Z
'liTa
H
(h)ijTa
a
9ijTa
D. ark and Old English Personal Names". Bammesberger 1991.309-334. . Iverson, Gregory. 1987. "On Glottal Width Features". Lingua 60.33 1-339. Jacobsen, Thorkild. 1974. "Very Ancient Texts: Babylonian grammatical texts". Studies in the History 0/ Linguistics: Traditions and paradigms ed. by Dell Hymes, 41-62. The Hague: Mouton.
REFERENCES
125
Jacobsohn, Hermann. 1910. "Aeolische Doppe)consonanz: Zur Sprache und Verstechnik der homerischen Epos". Hermes 45.67- 124, 161 -219. Jakobson, Roman. 1930. "Principes de phonologie historique". Jakobson 1962.202-220. ----------. 1 937. "On Ancient Greek Prosody". Jakobson 1962.262-27 1. ----------. 1962. Selected Writings. Vol.I. Phonological Studies. The Hague: Mouton. (2nd extended ed., 1971.) Jeffery, Lilian Hamilton. 1961. The Local Scripts of Archaic Greece. Oxford: Clarendon. ---------- & Alan W. Johnston. 1990. The Local Scripts of Archaic Greece. 2nd ed. Ibid. Jenness, Diamond. 1927. "Notes on the Phonology of the Eskimo Dialect of Cape Prince of Wales, Alaska". International Journal of American Lin guistics 4. 168-180. Jensen, Hans. 1969. Sign, Symbol and Script. Transl. by George Unwin. 3rd ed. New York: G. P. Putnam's Sons. Jespersen, Otto. 1904. Lehrbuch der Phonetik. Berlin & Leipzig: B . G. Teubner. ----------. 1948. A Modern English Grammar: On historical principles. Vol.I. 2nd ed. Copenhagen: Ejnar Munksgaard. (Repr. London: Alien & Unwin, 1965.) Joseph, Brian D. 1992. "Diachronic Explanation: Putting speakers back into the picture". Davis & I verson 1992. 123- 144. Justeson, John S. 1988. Review of Sampson (1985). Language 64.421-425. Justus, Carol F. 1993. "Implications of the Evolution of Writing". Dia chronica 10.97-1 10. K = Krause ( 1966). Kachru, Braj B., Robert B. Lees, Yakov Malkiel , Angelina Pietrangeli & Sol Saporta, eds. 1973. Issues in Linguistics: Papers in honor of Henry and Renee Kahane . Urbana: Univ. of Illinois Press. Kaisse, Elaine. 1985. Connected Speech: The interaction qf syntax and phonology. Orlando, Aa.: Academic Press. Kawasaki, Haruko. 1986. "Phonetic Explanation for Phonological Universals: The case of distinctive vowel nasalization". Ohala & Jaeger 1986.81- 103. Kemp, Alan. 1987. "The Tekhne Grammatike of Dionysius Thrax. English translation with introduction and notes". Taylor 1987. 169-189. Kenstowicz, Michael. 1 994. Phonology in Generative Grammar. Oxford: Blackwell. Kim, Chin-Wu. 1970. "A Theory of Aspiration". Phonetica 2 1 . 107- 1 16. Kiparsky, Paul. 1973. "Phonological Representations". Three Dimensions of Linguistic Theory, ed. by Osamu Fujimura, 1-126. Tokyo: Institute for Ad vanced Study of Language. ----------. 1 979. "Metrical Structure Assignment is Cyclic". Linguistic Inquiry 10.42 1-441. ---------- & Lise Menn. 1977. "On the Acquisition of Phonology". Mac namara 1 977.47-78.
126
REFERENCES
Kirshenblatt-Gimblett. Barbara, ed 1976. Speech Play. Philadelphia: Vniv. of Pennsylvania Press. Koemer, E. F. Konrad. 1993. ''The Problem of Metalanguage in Linguistic Historiography". Studies in Language 17. 1 1 1- 134. Kortlandt. Frederik. 1991. 'The Germanic Seventh Class of Strong Verbs". North-Western European Language Evolution (NOWELE) 18.97-100. KnUnsky. Jitf. 1976. Papers in General linguistics . The Hague: Mouton. Krause. Wolfgang. 1966. Die Runeninschrijten im dlteren Futhark. Mit Bei trigen von Herbert Jankuhn. 2 vols. OOttingen: Vandenhoeck & Ruprecht. Aclll Kurylowicz, Jerzy. 1947. "La nature des prores dits «:analogiques» linguistica 5. 15-37 (Copenhagen). (Repr. in Kurylowicz 1960:66-86.) ---------- 1952. "The Germanic Vowel System". Biuletyn Polskiego Towa rzystwa J �koznawczego 1 1.50-54. ----------. 1960. Erquisses linguistiques. Wroclaw & Krak6w: Nauka. ----------. 1964. The Inflectional Categories of Indo-European. Heidelberg: Winter. ----------. 1968. Indogermanische Grammatik. VoI.II. Aklent-Ablaut. Heidel berg: Carl Winter. Larsen, Mogens Trolle. 1989. "What They Wrote on Clay". Schousboe & Larsen 1989. 121- 148. Laziczius, Gyula 1966. Selected Writings. Ed. by Thomas A. Sebeok. The Hague: Mouton. Leffel, Katherine. 1985. "X-bar Theory and Phonology". Manuscript, Univ. of Aorida, Gainesville, Aa Lehiste, Use. 1970. Suprasegmenlllls . Cambridge, Mass.: MIT Press. l..ejeune, Michel. 1972. Phonitique historique du mycenien et du grec ancien. Paris: Klincksieck. ----------. 1974. Manuel de la langue venete. Heidelberg: Carl Winter. Levin, Juliette. 1985. A Metrical Theory of Syl/abicity . Ph.D. dissertation, MIT, Cambridge, Mass. Lewis, Henry & Holger Pedersen. 1961. A Concise Comparative Celtic Grammar. Gottingen: Vandenhoeck & Ruprecht. Lieber, Rochelle. 1992. Deconstructing Morphology: Word formation in syntactic theory. Chicago: Univ. of Chicago Press. Lounsbury, Aoyd G. 1 989. ''The Ancient Writing of Middle America". Sen ner 1989.203-237. Lowe, Pardee, Jr. 1972. "Germanic Word Formation". Van Coetsem & Kufner 1972.2 1 1 -237. LSAG = Jeffery ( 1961), Jeffery & Johnston ( 1990). Lundberg, Ingvar. 1991. "Phonemic Awareness Can Be Developed without Reading Instruction". Brady & Shankweiler 1991 .47-53. Macnamara, John, ed. Im. LAnguage Learning and Thought. New York: Academic Press. MacWhinney, Brian. 1985. "Hungarian Language Acquisition as an Exempli fication of a General Model of Grammatical Development". Slobin 1985. 1069- 1 155. ".
REFERENCES
127
Mahulkar, D. D. 198 1 . The PratiJ8khya Tradition and Modern Linguistics. Barcxla: Maharaja Sayajirao Univ. of Baroda Press. Mann, Virginia A. 1991. "Are We Taking Too Narrow a View of the Condi tions for Development of Phonological Awareness?". Brady & Shankweiler 1991 .55-64. Maranzana, Eisa. 1993. The Cognitive Consequences of Literacy: A linguis tic mythology. Manuscript, Univ. of Florida, Gainesville, Fla Marchand, James W. 1 959. "Names of Germanic Origin in Latin and Romance Sources in the Study of Germanic Philology". Names 7. 167-181. Masson, Olivier. 1961. us inscriptions chypriotes syllabiques: Recueil critique et commente. Paris: FA. de Boccard. (Rev. ed., Athens: Ecole Fran�se, 1983.) ----------. 1983. "Remarques sur quelques passages de la tablette chypriote d'ldalion (ICS 217)". Bulletin de la Societe de Linguistique de Paris 78.261-281. McCarthy, John J. 1985, "Speech Disguise and Phonological Representation in Amharic". V an der Hulst & Smith 1985:305-3 12. Meid, Wolfgang. 1 967. Germanische SprachwissenschaJt. Vol.III. Wort bildungslehre. Sammlung Gtlschen 1218. Berlin: Waiter de Gruyter. Meillet, Antoine. 1903. "De la diff�renciation des phonemes". Memoires de la Societe de Linguistique de Paris 12. 14-34. Meister, Richard. 1894. "Zu den Regeln der kyprischen Silbenschrift". Indo germanische Forschungen 4. 175- 186. Millard, A. R. 1986. ''The Infancy of the Alphabet". World Archaeology 17. 390-398. Miller, D. Gary. 1974. "Vocalization of Resonants in Indo-European". Manuscript, Univ. of Florida, Gainesville. ----------. 1976a. "Glide Deletion, Contraction, Attic Reversion, and Related Problems of Ancient Greek Phonology". Die Sprache 22. 137- 156. ----------. 1 976b. "Liquids plus s in Ancient Greek". Glotta 54. 159- 172. ----------. 1976c. "The Transformation of a Natural Accent System: The case of the Ancient Greek enclitics". Glotta 54. 1 1-24. ---------- . 1977. "Language Change and Poetic Options". Language 53.21-38. ----------. 1978. ''Time and Word-Length as Meta-Determiners of Phonological Change". Paper presented at the Annual Meeting of the Linguistic Society of America, Boston. ----------. 1981. ''The History of the Labiovelars in Greek". Paper presented at the A merican Philological Association meeting, San Francisco (28 Dec.). --- 1982. Homer and the Ionian Epic Tradition: Some phonic and phonological evidence against an Aeolic 'phase '. Innsbruck: Innsbrucker Beitrage zur Sprachwissenschaft. ----------. 1986. Review of Bubenfk (1983). Phoenix 40. 104- 109. ----------. 1990. "Homer and Writing: Use and misuse of epigraphic and linguistic evidence". Classical 10urnal 85. 171- 179. ----------. 1 993. Complex Verb Formation. Amsterdam & Philadelphia: John Benjamins. -----
-.
128
REFERENCES
Miller, Patricia. 1972. "Some Context-Free Processes Affecting Vowels". Ohio State University Working Papers in linguistics 1 1 . Columbus, Ohio. Mithun, Marianne. 1989. "The Acquisition of Polysynthesis". Journal 0/ Child lAnguage 16.285-3 12. Moltke, Erik. 1985. Runes and Their Origin. Denmark and Elsewhere. Copenhagen: The National Museum of Denmark. Morais, Jos�. 1991. "Constraints on the Development of Phonemic Aware ness". Brady & Shankweiler 1991 .5-27. Morpurgo Davies, Anna. 1972. "Greek and Indo-European Semi-Consonants: Mycenaean u and w". Acta Mycenaea, 80-121. - ---------. 1986. "R:>rms of Writing in the Ancient Mediterranean World". The Written Word ed. by G. Baumann, 51-77. Oxford: Oxford Univ. Press. ----------. 1987. "Mycenaean and Greek Syllabification". Ilievski & Crepajac 1987.91- 104. ----------. 1988. "Mycenaean and Greek Language". linear B: A 1984 survey, 75- 125. Louvain-La-Neuve: Peelers. ---------- & Yves Duhoux. 1984. "Linear B: A survey". Proceedings 0/ the Mycenaean Colloquium 0/ the Vlllth Congress 0/ FIEC (Dublin 1 984). Louvain-Ia-Neuve: Cabay (1985). Morris, Richard L 1985. "The Etymology of NwG run 0-". Beitrtige zur Geschichte der deutschen Sprache und literatur 107.344-358. ---------. 1988. Runic and Mediterranean Epigraphy. North-West European lAnguage Evolution (NOWELE) Suppl. Vol. 4. Odense: Odense Univ. Press. Murray, Robert W. 1982. "Consonant Cluster Developments in Pili". Folia linguistics Historica 3:2. 163- 184. ----------. 1988 . Phonological Strength and Early Germanic Syllable Struc ture. Munich: Wilhelm Fink. Naveh, Joseph. 1982. Early History o/the AJphabet: An introduction to West Semitic epigraphy and palaeography. !..eiden: E. J. Brill. Neumann. Gtinter (with Klaus Stiewe). 1974. "Zu den Hexametern der kyprischen Inschrift ICS 264". Kadmos 13. 146- 155. Newmeyer, Frederick J., ed. 1988. linguistics: The Cambridge survey. 4 vols. Cambridge: Cambridge Univ. Press. Niang, Mamadou. 1993. "Syllable 'Sonority' Hierarchy and Stress Patterns in Pulaar: A metrical approach". Manuscript, Univ. of Illinois, Urbana, Ill. Nilsson. Martin P. 1915. "Die Obernahme und Entwickelung des Alphabets durch die Griechen". Danske Videnskabernes Selskab, Historisk-filologiske Meddelelser 1 .3ff (= Opuscula Selecta, 1029ff. Lund: Gleerup, 1952). (Repr. in Pfohl l968. 172- 196. whence our citations.) Nussbaum, Alan J. 1976. Caland's lAw and the Caland System. Ph.D. disser tation, Harvard Univ., Cambridge, Mass. Odenstedt, Bengt. 1989. "Further Reflections on the Meldorf Inscription". Zeitschriftflir deutsches Altertum und deutsche literatur 1 1S.77-SS. ----------. 1990. On the Origin and Early History 0/ the Runic Script: Typology and graphic variation in the older jutluz,k. Uppsala: Acta Academiae Regiae Gustavi Adolphi.
REFERENCES
129
---------. 1991. "A New Theory of the Origin of the Runic Script: Richard L. Monis's book Runic and Mediterranean Epigraphy". Bammesberger 1991. 359-387. Ohala, John J. & Jeri J. Jaeger, eds. 1986. Experimental Phonology. Orlando, Fla: Academic Press. Olivier, Jean-Pierre. 1986. "Cretan Writing in the Second Millennium S.C. ". World Archaeology 17.377-389. ----------. 1993. "KN 1 15 = KH 1 15. Un m�me scribe a Knossos et a La Ca n�: Au MR IIIB: Du sou�n a la certitude". Bulletin de Correspondance Hellinique 1 17. 19-33. Olson, David R. 1994. The World on Paper: The conceptual and cognitive implications o/writing and reading . Cambridge: Cambridge Univ. Press. - --- ------ & Nancy Torrance, eds. 1991. Literacy and Orality. Cambridge: Cambridge Univ. Press. ORI = Antonsen (1975). Osthoff, Hermann. 1881. Review of Gustav Meyer, Griechische Grammatik (Leipzig: Breitkopf & Hartel, 1880). Philologische Rundschau 1 . 1 5871598. Page, R. L 1987. Runes. Berkeley & Los Angeles:Univ. of California Press. Palmer, Leonard R. 1 980. The Greek Language. Atlantic Highlands: Humanities Press. Patel, P. G. 1993. "Ancient India and the Orality-Literacy Divide Theory". Scholes 1993.199-208. Pedersen, Holger. 1939. "Zur Tbeorie der altgriechischen Palatalisierung". Reed 1964.289-291. Pepperberg, Irene M. 1990a. "Referential Mapping: Attaching functional significance to the innovative utterances of an African grey parrot (psittacus erithacus)". Applied Psycholinguistics 1 1.23-44. � - 1990b. "Cognition in an African Grey Parrot (Psittacus erithacus): Further evidence for comprehension of categories and labels". Journal 0/ Comparative Psychology 104.41-52. ----------. 1991. ""Referential Communication with an African Grey Parrot". Harvard Graduate Society Newsletter (Spring. 1991), 1-4. Peters, Martin. 1980. Untersuchungen zur Vertretung der indogermanischen Laryngale im Griechischen. Vienna: bsterreichische Akademie der Wissen schaften. Pfohl, Gerhard. 1966. Griechische Inschriften als Zeugnisse des privaten und (j.ffentlichen Lebens. Munich: Emst Heimeran. ----------, ed. 1968. Das Alphabet: Entstehung und Entwicklung der griechi schen Schrift. Darmstadt: Wissenschaftliche Buchgesellschaft. ----------. 1969. ""Die wtesten Inschriften der Griechen". Quaderni Urbinati di Cultura Classica 7.7-25. ----------, ed. 1972. Inschriften der Griechen: Grab-. Weih-. und Ehrenin schriften. Darmstadt: Wissenschaftliche B uchgesellschaft. ----------, ed. Im. Das Studium der griechischen EpigraphiJc. Ibid. ------
--
.
130
REFERENCES
Polom6, Edgar. 1991. ''The Names of the Runes". Bammesberger 1991 .42 1 438. Poser, William J. 1992. ''The Structural Typology of Phonological Writing". Paper presented at the Annual Meeting of the Unguistic Society of America (Philadelphia). Powell, Bmy B. 1988. The Dipylon Oinochoe and the Spread of Literacy in Eighth-Century Athens". Kadmos 27.65-86. ----------. 1 99 1 . Homer and the Origin of the Greek Alphabet. Cambridge: Cambridge Univ. Press. Powell, Marvin A. 1981. ''Three Problems in the History of Cuneiform Wri ting: Origins, direction of script, literacy". Visible lAnguage 15.419-440. Praetorius, Franz. 1902. "Zur Geschichte des griechischen Alphabets". Zeit schrift der deutschen morgenlandischen Gesellschaft 56.676-680. ----------. 1908. "Zum semitisch-griechischen Alphabet". Zeitschrift der deu tschen morgenllindischen Gesellschaft 62.283-288. Prakash, P. , D. Rekha, R. Nigam & P. Karanth. 1993. "Phonological A wareness, Orthography, and Literacy". Scholes 1993.55-70. Raghavendra, Parimala & Laurence B. Leonard. 1989. ''The Acquisition of Agglutinating Languages: Converging evidence from Tamil". Journal of Child lAnguage 16.3 13-322. Read, Charles. 1991. "Access to Syllable Structure in Language and Learn ing". Brady & Shankweiler 1991 . 1 19- 124. Reed, Carroll E., ed. 1964. Etudes phonolgiques didiees a la memoire de M. le Prince N. S. TrubetzJcoy. University, Ala.: Univ. of Alabama Press. (Ori ginally Travaux du Cercle linguistique de Prague 8, 1939.) Rice, Keren D. 1989. "On Eliminating Resyllabification into Onsets". West Coast Conference on Foreign lAnguages (WCCFL) 8.33 1-346. ----------. 1 992. "On Deriving Sonority: A structural account of sonority relationships". Phonology 9.61-99. Ringe, Donald A., Jr. 1984. The Perfect Tenses in Greek InscriptWns. Ph.D� dissertation. Yale Univ. , New Haven, Conn. Risch, Emst. 1973 . Wortbildung der homerischen Sprache. 2nd ed. Berlin: Walter de Gruyter ( 1974). Rix. Helmut. 1976. Historische Grammatik des Griechischen: lAut- und Formenlehre . Darmstadt: Wissenschaftliche Buchgesellschaft. . Roeper, Thomas. 1 988. "Grammatical Principles of First Language Ac quisition: Theory and evidenge". Newmeyer 1988 #11,35-52. Ruijgh. Comelis J. 1 957. L'Eliment acheen dam la langue epique. Assen: Van Gorcum.
----------. 1967. Etudes sur la grammaire et le vocabulaire du grec mycinien. Amsterdam: Hakkert.
----------. 1985. "Problemes de philologie mycenienne". Minos 1 9. 1O5- 167.
Saenger. Paul. 199 1 . "The Separation of Words and the Physiology of Reading". Literacy and Orality ed. by D. R. Olson & N. Torrance, 198-214. Cambridge: Cambridge Univ. Press.
REFERENCES
131
Sampson, Geoffrey. 1 985. Writing Systems: A linguistic introduction. Stanford, Calif.: Stanford Univ. Press. Sapir, Edward. 1949 [ 1 933]. "The Psychological Reality of Phonemes". Selected Writings of Edward Sapir in lAnguage, Culture and Personality ed. by David G. Mandelbaum, 46-60. B erkeley & Los Angeles: Univ. of California Press. Sass, Benjamin. 1988. The Genesis of the Alphabet and its Development in the Second MiUennium B.C. Wiesbaden: Hanassowitz. Saussure. Ferdinand de. 1878. Memoire sur le systeme primitif des voyel/es dans les langues indo-europiennes. Leipzig: B. G. Teubner. (Repr. Hildes heim: Georg Olms. 1968.) Scalise. Sergio. 1984. Generative Morphology. Dordrecht: Foris. Schane, Sanford A. 1971. "The Phoneme Revisited". lAnguage 47.503-52 1 Schindler, Jochem. 1969. "Die idg. Warter ftir 'Vogel ' und 'Ei'''. Die Spra che 15. 144- 167. ----------. Im. "Notizen zum Sieversschen Gesetz". Die Sprache 23.56-65. Schmandt-Besserat, Denise. 1 992. Before Writing. VoU: From counting to cuneiform. Austin: Univ. of Texas Press. Schmitt. Rtidiger. Im. Einftihrung in die griechischen Dialekte. Darmstadt Wissenschaftliche Buchgesellschaft. .
Scholes, Robert l , ed. 1993.
lituacy and Language Analysis. Hillsdale, N ol.:
Lawrence Erlbaum. ----------. 1 993a. "In Search of Phonemic Consciousness: A follow-up on Ehri". Scholes 1993.45-53. ----------. 1993b. "On the Orthographic Basis of Morphology". Scholes 1993. 73-95. ---------- & Brenda J. Willis. 1990. "On the Orthographic Basis of Phonemic Segments in Linguistic Competence and Performance". lAnguage Sciences 12.33 1-343. Schousboe, Karen & Mogens Trolle Larsen, eds. 1 989. literacy and Society. Copenhagen: Akademisk. Forlag. Schrijver, Peter. 1991. The Reflexes of the Proto·lndo-European lAyrngeals in lAtin. Amsterdam & Atlanta: Rodopi. Schwyzer. Eduard, ed. 1923. Dialectorum Graecarum Exempla Epigraphica Potiora. Leipzig: Hirzel. (Repr. Hildesheim: Georg Olms, 1 960.) ----------. 1939-7 1 . Griechische Grammatik. 4 vols. Munich: Beck.. (I: 4th ed 1968� II-IlI: 3rd ed., 1966-68.) SEG = Supplementum Epigraphicum Graecwn. Leiden. Segert. Stanislav. 1993. "Cuneiform Alphabets from Syria and Palestine". Journal ofthe American Oriental Society 1 13.82-9 1 . Senner, Wayne M., 00 . 1989. The Origins of Writing. Lincoln. Neb.: Univ. of Nebraska. Sherman, Donald. 1975. "Stop and Fricative Systems: A discussion of paradigmatic gaps and the question of language sampling". Working Papers on lAnguage Universals ( WPLU) 17. 1-3 1. Stanford, Calif. : Stanford Univ. .•
132
REFERENCES
Sievers, Eduard. 1893. GrundzQge der Plwnetik. 4th ed. Leipzig: Breitkopf & Hartel. Slobin, Dan I., ed. 1985. The Crosslinguistic Study of lAnguage Acquisition. 2 vols. Hillsdale, N.l. : Lawrence Erlbaum. ----------. 1985a. "Crosslinguistic Evidence for the the Language-Making Capacity". Slobin 1985 #11,1 157-1256. Smyth. Herbert Weir. 1894. The Sounds and Inflections of the Greek Dialects: Ionic. Oxford: Clarendon. ----------. 1956. Greek Grammar. Rev. ed. by Gordon Messing. Cambridge, Mass.: Harvard Univ. Press. Sommerstein, Alan H. 1973. The Sound Pattern of Ancient Greek. Oxford: Basil Blackwell. SPE = Chomsky & Halle (1968). Spencer, Andrew. 1991. Morplwlogical Theory. Oxford: Basil Blackwell . Stampe. David L . 1 973a. A Dissertation on Natural Phonology . Ph.D. dissertation. Univ. of Chicago. ----------. 1973b. "Speech as Music: Toward an understanding of the prosodic characteristics of language". Paper presented at the Annual Meeting of the Linguistic Society of America. ----------. 1980. A Dissertation on Natural Plwnology. New York: Garland. Stemberger. Joseph Paul. 1993. "Glottal Transparency". Plwnology 10. 107138. Stephens. Laurence D. & John S. Justeson. 1978. "Reconstructing 'Minoan' Phonology: The approach from universals of language and universals of writing systems". Transactions and Proceedings of the American Philo
logical Association 108.271-284.
Stephens, Laurence D. & Roger D. Woodard. 1986. "The Palatalization of the Labiovelars in Greek: A reassessment in typological perspective". Indo
germanische Forschungen 91. 129- 154.
Steriade. Donca. 1982. Greek Prosodies and the Nature of Syllabification. 2 vols. Ph.D. dissertation, MIT, Cambridge, Mass. ----------. 1988. "Reduplication and Syllable Transfer in Sanskrit and Else where". Plwnology 5.73-155. Stonham, John T. 1990. Current Issues in Morphological Theory. Ph.D. dissertation, Stanford Univ., Stanford, Calif. ----------. 1994. Studies in Combinatorial Morphology. Amsterdam & Phila delphia: John Benjamins. Streitberg. Wilhelm. 1896. Urgermanische Grammatik. Heidelberg: Carl Winter. Stroud. Ronald S . 1989. "The Art of Writing in Ancient Greece". Senner
1989. 103-1 19.
Svenbro, Jasper. 1989. "Phrasikleia - An A rchaic Greek theory of writing". Schousboe & Larsen 1989.229-246. Szemerenyi, Oswald. 1990. EinfiJhrung in die vergleichende Sprachwissen schaft. 4th ed. Dannstadt: Wissenschaftliche Buchgesellschaft.
13 3
REFERENCES
Taft, Marcus. 1992. Reading and the Mental Lexicon. Hillsdale, N.J.: Law rence Erlbaum. Taylor, Daniel J. ed. 1987. The History oJ Iinguistics in the Classical Period. Amsterdam & Philadelphia: John Benjamins. Templeton, Shane & Donald R. Bear, eds. 1992. Development oJ Ortho
graphic Knowledge arid the Foundations oJIiteracy: A memorialJestschriJt Jor Edmund H. Henderson. Hillsdale. N.J.: Lawrence Erlbaum. Teodorsson. Sven-Tage. 1974. The Phonemic System oJ the Attic Dialect 4()()340 B.C. Lund: Berlingska Boktryckeriet.
----------. 1 978. The Phonology oJ Attic in the Hellenistic Period.
Goteburg:
Acta Universitatis Gothoburghensis. Thausing, Moritz. 1 863. Das natiirliche
lAutsystem des menschlichen Spra che, mit Bezug auJ BriJcke's 'Physiologie und Systematik der Sprachlaute '.
Leipzig: W. Engelmann. Thomas, Rosalind. 1 989. Oral Tradition and Athens. Cambridge: Cambridge Univ. Press.
Written Record in Classical
----------. 1992. Literacy and Orality in Ancient Greece.
Cambridge: Cam bridge Univ. Press. Threatte, Leslie. 1980. The Grammar of Attic Inscriptions. VoU. PhOlwlogy. New York: Waiter de Gruyter. Thumb, Albert & Ernst Kieckers. 1932. Handbuch der griechischen Dialekte. 2nd ed. Vol.l. Heidelberg: Carl Winter. Thumb, Albert & Anton Scherer. 1959. Handbuch der griechischen Dialekte. 2nd ed. VoU!' Ibid. Treiman, Rebecca. 1993. Beginning to Spell. Oxford: Oxford Univ. Press. ---------- & Andrea Zukowski . 1991. "Levels of Phonological A wareness". Brady & Shankweiler 1991.67-83. Tmka, B [ohumil]. 1939. "Phonological Remarks Concerning the Scandina vian Runic Writing". Reed 1964.292-296. Ultan, Russell. 1971. "A Typological View of Metathesis". Working Papers on lAnguage Universals ( WPLU) 7. 1-44. Stanford, Calif.: Stanford Univ. van Coetsem, Franz & Herbert L. Kufner, eds. 1972. Toward a Grammar of Proto-Germanic. TUbingen: Max Niemeyer. van der Hulst, Harry & Norval Smith, eds. 1985. Advances in Nonlinear Phonology. Dordrecht: Fons. Vennemann, Theo. 1 97 1 . "The Phonology of Gothic Vowels". lAnguage 47.
90-132. ----------. 1988. Preference lAws for Syllable Structure and the Explanation oJ Sound Change: With special reference to German, Germanic, Italian, and lAtin. Berlin & New York: Mouton de Gruyter. Vine, Brent. 1993. Studies in Archaic lAtin Inscriptions. Innsbruck: Institut
fUr Sprachwissenschaft, Univ. Innsbruck. Viredaz, Remy. 1983 "La graphie des groupes de consonnes en mycenien et en chypriote". Minas n.s. 18. 125-207. Voyles, Joseph B . 1992. Early Germanic Grammar: Pre-, proto-, and post Germanic languages. San Diego: Academic Press. .
134
REFERENCES
Wackemagel , Jacob. 1885. "Miszellen zur griechischen Grammatik: Die indefiniten Relativa". ZeitschriftfUr vergleichende Sprachforschung 27.8992. (Repr. in Wackernagel 1953.569-572.) ----------. 1893. "Beitrage zur Lehre vom griechischen Akzent". Programm zur Rektoratsfeier der UniversiUit Basel, 3-38. Reprinted in Wackernagel
1953.1072- 1 107. ----------. 1896. Altindische Grammatik. Vol.!. Lautlehre. Gottingen: Vanden hoeck & Ruprecht (2nd ed. With supplements by Albert Debrunner, 1957.) ----------. 1905. Altindische Grammatik. VoI.II. 1. Einleitung zur Wortlehre: Nominalkomposition. Ibid. (With supplements by Albert Debrunner, 1957.) ----------. 1953. Kleine Schriften. 2 vols. GoUingen: Vandenhoeck & Ru precht. (Repr. , 1969.) Wachter, Rudolf. 1989. "Zur Vorgeschichte des griechischen Alphabets". Kadmos 28. 19-78. ----------. 1991. "Abbreviated Writing". Kadmos 30.49-80. Wallace, Rex. 1989. "The Origins and Development of the Latin Alphabet". Senner 1989. 121- 135'Wathelet, Paul. 1970. Les traits eoliens dans la langue de l'epopee grecque.
Rome: Ateneo. Watkins, Calvert. 1976. "Observations on the ' Nestor's Cup' Inscription". Harvard Studies in Classical Philology 80.25-40. Watt, W[illiam] C. 1987. "The Byblos Matrix". Journal of Near Eastern
Studies 46. 1- 14. -----:-----. 1989. ''The Ras Shamra Matrix". Semiotica 74.61-108. ----------, ed. 1993. Writing Systems and Cognition. Dordrecht: K1uwer. Wetzels, Leo. 1 986. "Phonological Timing in Ancient Greek". Studies in Compensatory Lengthening ed. by Leo Wetzels & Engin Sezer, 297-344.
Dordrecht Foris. WilIetts, Ronald F., ed. 1967. The Law Code of Gortyn . (=Kadmos Suppl . 1.) Berlin: Walter de Gruyter. WilIiams, Henrik. 1992. "Which Came First, M or n ?". Review of Odenstedt
(1990). Arkivfor nordiskfilologi W7. 192-205.
Willows, Dale M., Richard S. Kruk & Evelyne Corcos. 1993. "Are there Differences Between Disabled and Normal Readers in Their Processing of Visual Information?". Willows, Kruk & Corcos 1993a. ----------, eds. 1 993 a. Visual Processes in Reading and Reading Disabilities . Hillsdale, N.J.: Lawrence Erlbaum. Woodard, Roger D. 1993. "On the Interaction of Greek Orthography and Phonology: Consonant clusters in the syllabic scripts". Watt 1993. ----------. Forthcoming. Elementa. Manuscript, Univ. of Southern California, Los Angeles. Wyau, William F., Jr. 1964. "Arcado-Cypriote (cit; Iuis". Glotta 42. 170-182. ----------. 1 968. "Early Greek Iy/". Glotta 46.229-237. ----------. 1970. ''The Prehistory of the Greek Dialects". Transactions and
Proceedings of the American Philological Association 101.557-632.
REFERENCES
135
----------. 1973. ''The Aeolic Substrate in the Peloponnese". American Journal ofPhilology 94.37-46. ----------. 1975a. "Homer's Linguistic Ancestors". 'E7TloTTJJ.l 01'lK7J 'E7TET11P k $€ooaAo1'{ny; 14. 133-147. ThessaIonike. ----------. 1975b. "Aeolic Reflexes of Labiovelars in Homer". Greek, Roman, and Byzantine Studies 16.251-262. Yip, Moira. 1982. "Reduplication and C-V Skeleta in Chinese Secret Lan guages". linguistic Inquiry 13.637-661. Zec, Draga. 1988. Sonority Constraints on Prosodic Structure. Ph.D. disser tation, Stanford Univ., Stanford, Calif.
GENERAL INDEX A.
Aaonym: 3,2 AcrophoDic priDci.ple: 0. 1 . 4. 1. 4.2. 4.5. 4.6. 5.5 Altides: see