Woodard R.D. (editor) The Ancient Languages of Europe

Подождите немного. Документ загружается.

ancient nordic

227

5.3.2 Auxiliary verbs and passive voice

In the two occurrences of a complex verb form, the auxiliary follows the main verb (sup-

porting an OV analysis of the language; see §5.4): ﬂagdafaikinaz ist “is threatened by deceit”

(Vetteland); haitinaz was “was called” (Kalleby).

These two sentences must be interpreted as passives. The passive auxiliary may be omitted,

however, as in ana hanhai slaginaz “slain on the horse” (M

ojbro), and . . . iz hlaiwidaz

“. . . i buried here” (Amla).

There are no attested occurrences of the inﬂectional passive which is found in Gothic and

in non-Germanic Indo-European languages (the only trace of the Indo-European middle

voice is perhaps the verb hait¯e “I am called”).

5.4 Word order

5.4.1 Verb position

The examples above having a single complement – be it a predicate complement, an ac-

cusative object, or a prepositional phrase – may be taken as evidence that Ancient Nordic

is a verb-ﬁnal (OV) language (there are, however, no postpositions in the corpus, only

prepositions; note also the predominant head-dependent order in NPs [see 5.1.1]). This is

by no means surprising, since this is the order which can be reconstructed for Proto-Indo-

European, and since there are traces of an underlying verb-ﬁnal pattern in Old Norse.

In contrast, the two sentences above with double complements (ek Hrazaz satido staina

ana ...r ...“I H.setstone(acc.) on...”;and[falh] Woduride staina “dedicated the stone

to W.”) appear to suggest a VO order (as do several other sentences in the corpus). It is

worth noting, however, that in all the examples with a nonﬁnal verb, the verb is ﬁnite, and

it is in ﬁrst or second position. This is consistent with a rule of verb movement, shifting

the ﬁnite verb into second position, as in later stages of Germanic and in all of the modern

Germanic languages (except English): ek Hagustadaz hlaiwido magu minino “I H. buried my

son” (Kjølevik).

The sentences with the verb in ﬁrst position are subjectless sentences (cf. §5.4.2), except

wate halli hino horna “wet this stone, horn!” (Strøm), where the verb is in the optative mood

and perhaps fronted for emphasis.

Since we ﬁnd no verb in any other position than ﬁrst, second, or last, and since we ﬁnd

no nonﬁnite verb preceding its complement, it can be concluded that Ancient Nordic is V2

(verb-second) and OV (verb-ﬁnal) at the same time, just like Modern German.

5.4.2 Subject position

There are eighteen sentences in the corpus having a ﬁnite verb and a nominative subject. In

ﬁfteen of these the subject is in ﬁrst position, as in Bidawarijaz talgide “B. carved” (Nøvling)

and ek Hlewagastiz holtijaz horna tawido “I H. of Holt made the horn” (Gallehus). More

examples are provided by sentences cited above. In wate halli hino horna “wet this stone,

horn!” (Strøm), the verb is in the optative and in ﬁrst position. In wurte runoz an walhakurne

Heldaz Kunimundiu “wrought runes on the Celtic corn, H. for K.” (Tjurk

o), the subject has

been focused and moved to the right. In Hariuha hait-eka farawisa “H. I am called, the

travel-wise” (Sjælland), the subject is expressed as an enclitic on the verb. And in ek erilaz

Sawilagaz hait-eka “I, the erila, am called S.” (Lindholm), the clitic repeats the subject in

ﬁrst position.

228 The Ancient Languages of Europe

There is no doubt that the apparent regularity with regard to the position of the subject

must be due to the homogeneous nature of the material, consisting solely of epigraphic

texts. Ancient Nordic must have a rather free word order, like its relatives in Germanic and

other Indo-European language groups.

5.5 Pro-drop?

On the basis of epigraphic material alone it is impossible to determine securely whether the

language has pro-drop or not – that is, whether the subject can be omitted and represented

by verbal inﬂection alone, even when it is not recoverable from the context. It is true that

there is not one single occurrence of a pronoun as a subject in the corpus; all of the subject

pronouns attested occur as constituents of appositional constructions (cf. §5.1.2). Moreover,

we do ﬁnd ﬁve occurrences of a missing subject. Four of these, however, follow immediately

after other lexical material in which the subject referent is mentioned: Hariuha haiteka

farawisa gibu auja “H. I am called, the travel-wise, give (1st per.) luck” (Sjælland). In this

sentence, gibu is ﬁrst-person present, and the subject is the same as that of hait-, namely

-eka “I.” In haitinaz was “was called” (Kalleby), the subject can be inferred from a preceding

genitive noun,

rawijan (PNm). In the case of Uha haite “I am called Uha” (Kragehul),

preceding is ek erilaz Asugisalas. The sentence wurte runoz an walhakurne “wrought runes

on the Celtic corn” (Tjurk

o) occurs together with the two names Heldaz Kunimundiu,in

the nominative and dative, respectively, on the same stone (Grønvik 1987:151); the subject

is therefore recoverable (Heldaz). This leaves us with one short inscription with two words:

tawo la

odu “make (1st per.) the invitation” (Trollh

attan). Bearing in mind that this is

epigraphic material, we certainly have no evidence to conclude that Ancient Nordic is a

language in which subject pronouns can be freely omitted.

5.6 Nonverbal sentences?

Examples have already been given of deleted auxiliaries. The question is whether this is due

to the epigraphic style (comparable to modern newspaper headlines – cf. “Ten killed in car

crash”), or part of the regular grammar of the language (as in, e.g., Modern Russian). The

question is further complicated by apparent appositional constructions consisting of two

nominative NPs (cf. §5.1.2). When these stand by themselves in an inscription, they may

also be read as a copular sentence with an omitted copula: ek Unwodiz [em] “I am Unwodi”;

ek gudija ungandiz [em] “I am Ungandi’s priest”; and so forth.

6. LEXICON

The vocabulary in the Ancient Nordic inscriptions consists almost exclusively of inherited

Germanic items. In the extant material there is no certain example of a word with a distinctly

non-Germanic form, or a loanword from a non-Germanic language, although we know from

later attestations that, for example, Celtic words had been adopted during the early Iron Age.

Acknowledgments

I am grateful to Ottar Grønvik, Jan Ragnar Hagland, KathyHolman, Brit Mæhlum, and Arne Torp, who

have read a previous version of this chapter and given me many valuable comments and suggestions.

ancient nordic

229

Bibliography

Antonsen, E. H. 1975. A Concise Grammar of the Older Runic Inscriptions. Sprachstrukturen. Reihe A.

Historische Sprachstrukturen 3. Tübingen: Niemeyer.

Bammesberger, A. 1990. Die Morphologie des urgermanischen Nomens. Heidelberg: Carl Winter.

Faarlund, J. T. 1990. “Syntactic and pragmatic principles as arguments in the interpretation of runic

inscriptions.” In J. Fisiak (ed.), Historical Linguistics and Philology, pp. 165–186. Berlin/New

York:MoutondeGruyter.

Grønvik, Ottar. 1981. Runene på Tunesteinen. Oslo: Universitetsforlaget.

———

.1987. Fra

Agedal til Setre. Oslo: Universitetsforlaget.

———

.1996. Fra Vimose til Ødemotland. Nye studier over runeinnskrifter fra førkristen tid i Norden.

Oslo: Universitetsforlaget.

Haugen, E. 1976. The Scandinavian Languages. An Introduction to their History. London: Faber and

Faber.

Knirk, J. E. 1977. Review of A Concise Grammar of the Older Runic Inscriptions,byE.H.Antonsen.

Maal og Minne, pp. 172–184.

Krause, W. 1971. Die Sprache der urnordischen Runeninschriften. Heidelberg: Carl Winter.

Krause, W. and H. Jankuhn. 1966. Die runeninschriften im ¨alteren Futhark.G

ottingen: Vandenhoeck

and Ruprecht.

Moltke, E. 1985. Runes and their Origin. Denmark and Elsewhere. Copenhagen: Nationalmuseets

Forlag.

Morris, R. L. 1988. Runic and Mediterranean Epigraphy. NOWELE Supplement 4. Odense: Odense

University Press.

Østmo, E. 1996. The Indo-European Question in a Norwegian Perspective. A View from the Wrong End

of the Stick. Journal of Indo-European Studies Monograph 17.

Smith, J. R. 1971. Word Order in the Older Germanic Dialects. Ann Arbor: University Microﬁlms.

Syrett, M. 1994. The Unaccented Vowels of Proto-Norse. NOWELE Supplement 11. Odense: Odense

University Press.

appendix 1

Indo-European

henry m. hoenigswald and

roger d. woodard

with a discussion of syntax by

james p. t. clackson

1. THEORETICAL AND HISTORICAL CONTEXTS

1.1 The comparative method

The parent language of the Indo-European linguistic family is an “ancient” language in a

special sense: it is a protolanguage, not attested but reconstructed. Since a protolanguage

is, broadly speaking, the collection of all retentions in the daughter languages, the ability

to segregate innovation from retention in the latter is crucial for the reconstruction of

the former. The “comparative” method (in the narrow, phonological sense of the term)

accomplishes that segregation to a large extent (on the comparative method of historical

linguistics, see also WAL Ch. 45). Those innovations which we classify as sound-changes are

capable of producing homophony among morphs; they are phonemic mergers, with the

algebraic form /a/ > /m/, /b/ > /m/ (further elaboration is needed for conditioned sound-

changes). Owing to the “Polivanov” property of sound changes (“no split without merger”),

which follows from their deﬁnition as replacements statable in purely phonological terms

(without reference, that is, to particular morphs), it is the case that if one phoneme, or

one phonemic component (distinctive feature speciﬁcation), or one phoneme combination

(diphthong, cluster, syllable, etc.) in language A corresponds to one phoneme or phonemic

component or phoneme combination in a related language B in one set of morphs, and to

some other phoneme (etc.) in another set of morphs, then language A has in this detail

innovated. As regards other details the converse may be the case, and language B may be the

innovator. If A is found to have innovated in all details and B in none, A is a descendant (or

later stage) of B and B the ancestor (or earlier stage) of A. In this case, language B may be

predicted to have occurred in time before language A.

The comparative method aims at the recovery of the phonological shape of morphs.

When it comes to morphemics – obsolescence, neologism, semantic change, borrowing,

analogic change, and so forth – what is sometimes also called the comparative method is in

reality something quite different (hence the preponderance of phonological subject matter

in comparative work). The methods available for morphemic retrieval are much more akin

to “comparison” in the everyday meaning of the word. They tend to rely on grammatical and

lexical consensus and on resemblances and differences that do not by themselves, typological

considerations aside, carry any clear-cut chronological implications. Extensive use, however,

is made of internal reconstruction which operates not only with phonological alternations

which result from conditioned sound-changes, but also with semantic isolation of forms in

morphological and syntactic paradigms and the like.

230

Indo-European

231

1.2 Scholarly tradition

The conceit of related languages having their descent from a no longer spoken “parent

language” is old (Metcalf 1974:251). For the Indo-European languages it was memorably

voiced in 1786 by Sir William Jones (1746–1794), the justly admired and inﬂuential British

jurist and scholar who served in India. Though without a marked intellectual interest in

language as such, Jones was riding the crest of the new-found wave of enthusiasm (an

enthusiasm in the creation of which he was himself a leading spirit) about things Indic.

In matters of language he argued in traditional fashion from the “perfection” of Sanskrit,

Greek, and Latin. Of the sober efforts directed at the Finno-Ugric languages by Strahlenberg

(1676–1647) in 1730 and Sajnovics (1733–1785) in 1770 he was unaware.

The nineteenth and twentieth centuries witnessed the unfolding of the great work of ﬁlling

old metaphors with a new technical content, not necessarily acknowledgedin the abstract but

abundantly clear from substantive, especially polemical, endeavor. Sincethe days of Wilhelm

von Humboldt (1767–1835) proof of “relationship” in the form of carrying out convinc-

ing reconstructions has been provided for language families as diverse as Austronesian,

Afro-Asiatic (including Semitic), Dravidian, Algonquian, among others.

1.3 Internal and external relations of Indo-European

Proto-Indo-European (PIE) may well have been spoken somewhere in the Black Sea area

before the middle of the ﬁfth millennium BC. At that time the speech community began

to break up in a complex, long-drawn-out, and only partly recoverable process. The

main branches which survived into historical times are (to list them in the chronological

order of their ﬁrst documentation): Anatolian (now extinct; see WAL Chs. 18–23), Indo-

Iranian (WAL Chs. 26–30), Greek (Chs. 2–3), Italic (Chs. 4–5), Celtic (Ch. 8), Germanic

(Chs. 9–10), Armenian (WAL Ch. 38), Tocharian (extinct), Balto-Slavic, and Albanian (the

three last-named being too recently attested for inclusion in the present volume). Additional

Indo-European languages are attested in antiquity which do not clearly belong to any of

these ten subfamilies, or whose membership is debated, such as Phrygian (WAL Ch. 31),

Venetic (Ch. 6) and Messapic (Ch. 1,

§7).

Once severed from one another, each branch went through changes that were largely but

not entirely independent. Subgroupings based on the principle of shared innovation in the

manner of the well-known family tree (German Stammbaum), or some other topological or

geometrical scheme, will in general be discussed in the later chapters (noted above) which

deal with the comparative evidence, that is, with the changes that deﬁne the descendant

languages.

Proto-Indo-European is certain to have had outside connections of two kinds: (i) common

descent from an anterior pre-protolanguage, and (ii) contacts recognizable from member-

ship in areal typologies. Efforts to identify either kind have remained inconclusive.

2. PHONOLOGY

2.1 Consonants

The reconstructed consonantal inventory of Proto-Indo-European is comprised of obstru-

ents (stops and fricatives), nasals, and sonorants (liquids and glides), as well as the so-called

laryngeal consonants.

232 Appendix 1

2.1.1 Obstruents

The stop phonemes of Proto-Indo-European, identiﬁed following established practice, are

produced at ﬁve articulatory positions: (i) bilabial; (ii) dental; (iii) palatal; (iv) (pure) velar;

(v) labiovelar. For each position, a (i) voiceless, (ii) voiced, and (iii) voiced aspirated stop is

reconstructed:

(1) bilabial dental palatal velar labiovelar

voiceless ptk



voiced bdg



voiced asp. b



The voiced bialabial

∗

b occurs only rarely. In the recently advocated “glottalic” view, the

values of traditional

∗

b, and

∗

(etc.) are

∗

(h)

(aspirated, with unaspirated allophones),

∗

p’ (voiceless glottalized), and

∗

(h)

(etc.) respectively; see Gamkrelidze and Ivanov 1995 and,

for an evaluation, Watkins 1998:38.

The labiovelar phonemes

∗

, and

∗

are distinct from the sequences

∗



∗



∗

(palatal stop + labiovelar glide) as well as perhaps from the sequences

∗

kw,

∗

gw,

∗

w (velar

stop + labiovelar glide). Still, it is remarkable that something like the geminate prohibition

(see §2.3) neutralizes labiovelars and velars before [u], with outcomes that are those of the

velars. This is especially visible in post-Mycenaean Greek where

∗

a gives Attic [pa] but

∗

u yields [ku] (see Ch. 2, §3.7.1).

On the basis of the evolutionary outcome of the Proto-Indo-European palatal, velar, and

labiovelar stops, Indo-Europeanists have traditionally divided the Indo-European daughter

languages into two major groups, labeled centum (Latin for “100”) and satem (after Avestan

sat

m “100”; both forms from PIE

∗



mtom). The general case is that western Indo-European

(centum) languages merge the palatal and velar stops, whereas in the eastern (satem) dialects,

the palatal stops exhibit distinct reﬂexes while the velars and labiovelars fall together (see

Melchert 1987). The conspicuous exception to this distributional pattern is provided by

Tocharian. Spoken far to the east in antiquity (with documentary remains surviving in the

deserts of Chinese Turkestan or Xinjiang Uygur), Tocharian shows the centum treatment of

back consonants.

Proto-Indo-European possessed the dental sibilant

∗

s , presumably with allophones [s] and

[z], the latter occurring before plain voiced and voiced aspirated obstruents. The occurrence

of an interdental fricative /

þ/ has long been proposed to account for that stop/fricative

correspondence seen in cognates such as, for example, Greek  (´arktos) and Sanskrit

´r

ks

a-, “bear,” but this remains problematic as another, more sophisticated solution has been

proposed.

2.1.2 Sonorants

The Proto-Indo-European sonorant phonemes occur as both nonsyllabic and syllabic allo-

phonic variants (see §2.1.4):

(2) nasals liquids glides

n/n r/r y/i

m/m

 l/l w/u

Indo-European

233

2.1.3 Laryngeals

Those consonantalsounds identiﬁed as “laryngeal” likewise occur in nonsyllabic and syllabic

forms:

(3) h

/ə

(for other notations and other views, see Watkins 1998:40). Phonetically, these are, to judge

from their comportment in conditioned sound change in the descendant languages, neutral

), a-colored (h

), and o-colored (h

), respectively. The nonsyllabic allophones of the ﬁrst

two laryngeals seem to be voiceless; that of the third, voiced.

2.1.4 Vocalic versus consonantal

The full-grade vowels (see §2.2), the long vowels, the syllabic allophones of glides and

laryngeals, and the diphthongs will henceforth be referred to, when convenient, as vocalics;

nonvocalics are consonantals.

2.1.5 Nonsyllabic versus syllabic

In certain respects the three laryngeals resemble the sonorants. The resemblance is weakened

and tends to disappear in the descendants. Very roughly, the following holds:

1. After a full-grade vowel (see §2.2) and preceding a consonant, both the sonorants and

the laryngeals appear in their nonsyllabic shapes, the sonorant combinations forming

diphthongs and the laryngeal combinations merging in the descendants (if not earlier),

for the most part, with the long vowels. Similarly, syllabic

∗

i and

∗

u with a following

laryngeal generate

∗

¯ı and

∗

¯u. The syllabic allophones of the liquids and nasals lead to

different results in the descendants.

2. Unless following a full-grade vowel, sonorants and laryngeals preceding a consonantal

appear in their syllabic shapes. However, special provisions require certain sonorants

and certain laryngeals to appear word-initially in nonsyllabic form when followed by

certain nonsyllabic sonorants which are followed in turn by vowels, so as to form an

initial sonorant cluster (e.g.,

∗

#[wr-]). In the descendant languages, laryngeals in their

syllabic shapes end up merged with the full-grade vowels and their outcomes (

∗

= e,

∗

= a,

∗

= o) – once again a process that may have commenced in Proto-Indo-

European.

3. Word-medially when occurring after the sequence short vowel + one consonant and

before a vocalic (VC

[+ vocalic]), sonorants appear in their nonsyllabic shape

(algebraically, ..et[y]e..). When occurring after the sequence vowel + two consonants,

or long vowel + one consonant, and before a vocalic ({VCC or VVC}

[+ vocalic]),

sonorants appear in their syllabic shape (i.e., ..ekt[i]e.., Sievers’ Law). After a single

word-initial consonant, syllabic and nonsyllabic shapes both occur – generalized from

occurrences after a preceding word-ﬁnal vowel or word-ﬁnal consonant respectively

(i.e., ..#t[y]e.., ..#t[i]e..).

2.2 Vowels

The Proto-Indo-European vowel inventory consisted of the “full-grade” short vowels

∗

and

∗

a,aswellas

∗

i and

∗

u, the syllabic allophones of the glides

∗

y and

∗

w (see §2.1.2); and

234 Appendix 1

the “lengthened-grade” long vowels

∗

¯e,

∗

¯o, and

∗

¯a,pluslong

∗



and

∗

¯u. The resulting vowel

systems, short and long, were thus:

(4) short long

front back front back

high iu

mid eo

low a

Moreover, there occurred the automatic syllabic outcropping [

] between obstruents,

known as schwa secundum (see §2.4, 3.1).

2.3 Phonotaxis

Various phonotactic constraints limit the permissible sequences of sounds in Proto-Indo-

European (see also §3.3):

1. There are no geminates. Geminate clusters arising across morpheme boundaries were

simpliﬁed: for example,

∗

´es-si “you (sg.) are” yields

∗

´esi, as in Sanskrit ´asi (though

a marginal process of gemination creates hypocoristic by-forms of personal names

and the like; see Watkins 1998:40). The sequence

∗

..t-t.. was, however, analogically

restored.

2. There are no clusters (hiatus) of full-grade vowels, both like and unlike. Where such

sequences arise at morpheme boundaries, the vowels are contracted into long vowels

bearing a distinctive accent in some descendant languages (the “long diphthongs,”

where they are not contraction products [as in, e.g., the thematic dative singular

ending, see §3.5.3], pose difﬁcult problems).

3. Obstruent (and s ) sequences are entirely voiced or entirely voiceless. If a voiceless

and a voiced or voiced aspirated obstruent abut at a morpheme boundary, regressive

dissimilation will take place. It is likely, by the same token, that the distinction between

the three manners of articulation was neutralized, phonetically in favor of voiceless-

ness, before a word boundary (see §2.5). The word-ﬁnal sequences

∗

-ms# and

∗

-ns# are

likewise neutralized (Leumann 1977:415); this is relevant for the animate accusative

plural ending; see n. 36.

4. Bartholomae’s Law speciﬁes that “if the ﬁrst member of an obstruent cluster

is . . . aspirated, the assimilation is progressive” (see Watkins 1998:40–41).

2.4 Syllabicity

There are hints of an overarching principle governing syllabicity. This principle is accessible

only in a schematically simpliﬁed and chronologically ﬂat form which fails to convey the

sliding nature of the scale along which developments took place, and which stretches from

a remote past well into the era of the descendant languages. While most of the evidence

is Indo-Iranian and Greek, it testiﬁes nevertheless to a state of affairs that is essentially

Indo-European. It is likely that syllabicity largely falls out in such a way as to preclude the

accumulation of more than two consonantals in the ﬂow of speech (with a word boundary

as well as the sibilant s playing an uncertain role; see Beekes 1982:110) – hence, before

vowels, Sievers’ Law (..et[y]e, ..et[r]e but ..ekt[i]e, ..ekt[r

]e;see§2.1.5) as modiﬁed by

Lindeman’s Law (which regulates word-initial obstruent + sonorant clusters; see Lindeman

1965).

Indo-European

235

There could well have existed an Extension of Sievers’ Law before consonantals and before a

pause if –as the surviving difference betweenGreek(Attic-Ionic)  (´op

ra) “in order that”

(with one short vowel and one consonant preceding) and  (hˆepar) “liver” (with one

long vowel and one consonant preceding) suggests – the allophonic notation

∗

[r] stands for

two quite distinguishable allophonic entities:

∗

], with (it may be imagined) increasingly

prominent syllabicity (in ); and

∗

[

r], with syllabicity decreasing to the rightward

(in ).

In word-initial syllables where the determining environment is not built in, one would

expect vacillation between [ra] and [ar], with a potential for mutual analogic exchanges and

generalizations. This is indeed what one ﬁnds: for example, in Homeric  (krad´ı¯e)

beside  (kard´ıa) “heart.”

In Greek, as in other descendant languages, this

∗

[

] adjacent to liquids and nasals became

phonemic by merging with some existing vowel (in Attic-Ionic with [a]). In the case of

Indo-European

∗

[y/i] and

∗

[w/u] (these from the oldest period), and (much later) Indo-

Iranian

∗

[r/r] (Sanskrit . . . /r [but

∗

[r] > Sanskrit ir before vocalics (..aktira..) under Sievers’

Law proper; i.e., not the Sievers’ Law “Extension”], Avestan . . . /

ərə), the three pairs of two

positional variants are transformed into one segment each, perhaps of steady (i.e., neither

increasing nor decreasing) vowel-like quality. Under similar circumstances [

] in the vicinity

of obstruents can end up phonemic in the descendant languages by merging with one of

the existing vowels, though here the data remain shadowy. As a result of all of this, overlong

syllables (short vowels with more than two consonants, or long vowels or diphthongs with

two consonants before the next vowel) are rare, for example in Vedic and in Greek, until

sound changes create new overlengths (see Hoenigswald 1994 for the details; lengthened

grade [see §3.2] in certain formations is [still?] extremely rare in Sanskrit before consonant

clusters; see Debrunner 1954:61).

Thephenomena treatedabovemilitateintheirown typological wayin favorof theretentive

nature of pitch accent and quantitative meter; see §§2.6, 2.7.

2.5 Word boundaries

Word boundaries (i.e., seams between so-called minimum free forms; see Hoenigswald

1992) loom large as conditioning factors in sound changes. So far from indicating, however,

that all word boundaries are phonologically marked and contrast with Ø in word-interior

position (note §2.4 on phonetic conditioning across a word boundary), word boundary is

best considered an analogical development made possible by the circumstance that pause

(the absence of sound which contrasts with the presence of sound, a universal condition) is

an option at word boundaries. Post-pausal and ante-pausal allophony was generalized and

turned into apparent word-initial and word-ﬁnal phonology, each contrasting with word-

interior phonology. The descendant languages differ somewhat in the extent to which this

analogic change is carried through. Where analogic generalization is complete, utterances

may indeed be treated as “composed of ” (rather than “analyzed into”) words in external

sandhi (some of the sandhi phenomena of Insular Celtic may be relevant survivals – see

Russell 1995; sandhi phenomena were, however, created again and again in the separate

branches).

2.6 Accent

The fragmentary character of the scripts in which the texts of the descendant languages

are recorded, combined with the neglect of relevant phenomena despite their syntactic

236 Appendix 1

centrality, have prevented deciding whether phonemic stresses forming stressed morphs

existed, let alone reconstructing them. For some daughter languages metrical indications

are available but have scarcely been exploited. Such a determination would be of paramount

importance for syntax. Much of syntax is customarily discussed, faute de mieux,interms

of word order. In many languages, however, word-order phenomena (recognizable in the

texts) are correlated with, or even dependent on, stress phenomena (ignored in the texts);

see Hoenigswald 1980.

A lexical word accent (/´/) – likely a pitch accent – contrasted with the absence of ac-

centuation. Such an accent may be reconstructed from Vedic Sanskrit, Greek, Anatolian,

Balto-Slavic, and from the effect it had in Germanic (Ve r ne r’s L aw ;seeCh.9,§3.6.2). Clitics

were unaccented, enclitics occupying the second place in a clause (Wackernagel’s Law;see

Szemer

enyi 1996:81–82, with references).

Little is known about sentence intonations. It is possible, though unlikely, that the ﬁxed

high pitch of the question pronoun in Greek ,  (t´ıs, t´ı) represents the survival of an

Indo-European interrogative intonation.

2.7 Meter

It is uncertain whether Proto-Indo-European meter is quantitative in nature and based on

the characteristics of syllables, as it is in Sanskrit and in Greek, or whether these two daughter

languages have innovated (so Watkins 1995:21). The absence of any metrical function for

word accent in these two branches is often associated with quantitative meter, whether

retained or innovated. Verner’s Law in Germanic (see §2.6) as well as the dependence of

the ablaut zero-grade (see §3.2) upon lack of accent seem to point to an original strongly

“dynamic” character for word accent; see Lehmann 1952:109.

3. MORPHOLOGY

3.1 Word formation

The morphology of nouns/adjectives (including pronouns) and verbs, comprises derivation,

inﬂection, and compounding. A single root, minimal or extended (see §3.3), precedes a

derivational sufﬁx or sufﬁx sequence (or accommodates the

∗

-n/ne- inﬁx) which, in tandem

with syntactic function, deﬁne the resulting “word” (marked, as often as not, by the incidence

of accent) as a noun or verb. The resulting root + afﬁx complex is a stem, though in some

instances the root alone can function as a stem. In compounding (always binary), noun

stems combine to form more complex noun stems. Verbs are not in that sense capable

of compounding. Stems in turn are followed by a single nominal or verbal inﬂectional

ending which likewise contributes to syntactic identiﬁcation. The paradigms that result in

this synthetic structure are close-knit and, especially insofar as the endings are concerned,

characterized by well-recognizable and clear-cut allomorphies.

3.2 Ablaut

Proto-Indo-European ablaut, or apophony, originally depended on word accent (see §2.6)

in ways which are only in part transparent. The phenomenon is a pervasive, nonautomatic,

morphologically conditioned alternation of the vowels of (5):