The Position of Albanian

Eric P. Hamp, University og Chigaco
(Ancient IE dialects, Proceedings of the Conference on IE linguistics held at the University of California,
Los Angeles, April 25-27, 1963, ed. By Henrik Birnbaum and Jaan Puhvel)

(Part   _1_2_3_4_5_6_7_8_9_10_11_12_13_14_15_16)

1. It is fashionable at conferences to come out with no clear affirmative assertion, but rather with a statement of all the many difficulties. While I have no taste for fashion, this should prove a fashionable paper, on this ground if on no other. It is often hard enough to say conclusively where the IE features of Albanian lie, let alone to identify them unambiguously and assign them to a restricted relationship of shared innovation. This is not to say that things are as G. Meyer is all too well known to have put it; there is plenty of good Indo-European material in Albanian but it is often ambiguous and represented by small numbers of examples for each feature and combination.

Furthermore, I am not yet in a position to say what I hope will be possible when the dialect materials from most enclaves have been sifted and compared. This applies particularly to the verb.

There are also relative unknowns that are important in the total question on which I do not feel adequately informed to hold a worthwhile opinion: Thracian, with Dečev's bewildering material, is the notable example here.

2. There are ways in which our subject has been synthesized in the past that lighten our task somewhat: N. Jokl (Eberts Reallexikon der Vorgeschichte; articles "Albaner," "Illyrier," and "Thraker") gives a very just review; but he does nonetheless have his point of view. W. Porzig (Die Gliederung des indogermanischen Sprachgebiets [Heidelberg, 1954]) gives a fair and fairly complete summary, but he has no incisive point of view. Moreover, there has been a good bit of activity recently, for such a small field, and I have tried to sift through the output as fully as I could. Thus I hope to reach a fair degree of completeness in reporting, although, I suppose, at the same time some of my prejudices will show through.

3. When one looks over the ground to be covered, it seems that our subject falls naturally into three parts: the geographic position of Albanian in the Balkans; the corpus, location, and relations of Illyrian, Thracian, and their congeners; and the genetic ties of Albanian to its sister IE subgroups. These, in fair part, match three rather separate fields of expertise: "Balkan linguistics"; Classical linguistics, philology, and epigraphy; and Indo-European studies in the traditional sense. No one can be equally competent in all.

4. On the question of the earlier location of the Albanians, there is a good summary and batch of references in A. Rosetti, Istoria limbii romîne. II. Limbile bakanice3 41-44 (Bucureşti, 1962). Rosetti, however, mistakenly repeats the myth that some Tosk dialects show Geg characteristics, thus pointing, allegedly, to a more recent dialect split. The isogloss is clear in all dialects I have studied, which embrace nearly all types possible. It must be relatively old, that is, dating back into the post-Roman first millennium. As a guess, it seems possible that this isogloss reflects a spread of the speech area, after the settlement of the Albanians in roughly their present location, so that the speech area straddled the Jireček line.

In this context it is possible to find almost every opinion. Many agree that Albanian lacks an old maritime terminology, yet D. Dečev (Charakteristik der thrakischen Sprache 113 [Sofia, 1952]) thinks they have had it and lost it!

More positively, one may say that the mere absence of inherited maritime vocabulary can prove nothing. Recently, however, E. Çabej (VII Congresso intemacionale di scienze onomastiche, 4-8 Aprile 1961, 248-249) has argued for the actual presence, insufficiently noticed heretofore, of certain preserved old terms. But it seems to me that these are for the most part inconclusive in themselves: dēt 'sea' (related to 'deep') could refer to any deep water; 'ford, anchorage', mat 'beach' valë 'wave' could be applied to various bodies of water. A word like grykë 'narrows' is an easy metaphor ('throat'); aní ( : an(ë) 'vessel') and some names of parts of boats (ballë 'forehead', pëlhurë 'sail', shul 'mast', lugatë 'rudder', belonging with lugë 'spoon') are also easily understood as metaphors; likurishtë 'polyp' (cf. likurë 'skin') and many other names, often fairly transparent compounds (p. 249), are descriptive and could presumably have arisen in their attested uses at almost any time. The word ngjalë 'eel' < *engella, even if related correctly to Illyrian 'EggelaneV, does not necessarily presuppose the sea. Thus, we still lack a demonstrated body of native marine morphemes, with no other morphological or semantic connections in the language to make transfer possible. For such an argument, only isolated forms will be convincing.

Even recent history is checkered: Shqiptar first appears in the fourteenth century. Albi occurs in an Angevin document of 1330; according to Ptolemy, in the second century the Albanoí lived around Albanópolis (Kruja), where the ethnic has been recorded in modem times. The enclaves of Italy and Greece, to the extent that they use a traditional name, use this term: arbrésh (e.g., Vaccarizzo Albanese), arbëríshte (Greece). The earlier data are rehearsed, with references, in H. Barić's Lingvističke studije (Sarajevo, 1954; abbr. LS), and Hŷmje në historín e gjuhës shqipe 7 (Prishtinë, 1955; abbr. Hymje; = trans. Istorija arbanaškog jesika 30 [Sarajevo, 1959]).

The question of the names Elbasan, Arbëni, Albanopolis, Shqipëtar, Shqipëri is discussed at length, but somewhat inaccessibly, in Dh. S. Shuteriqi, Buletin për shkencat shoqërore 1956:3.189-224 (abbr. BShkSh) and Buletin i Universiteiit Shtetëror të Tiranës 1958:3.45-70 (abbr. BUShT) .

It is clear that in the Middle Ages the Albanians extended farther north (Jokl, Albaner §2); that there are persuasive arguments which have been advanced against their having extended as far as the Adriatic coast — the fact that Scodra 'Scutari' (Shkodër) shows un-Albanian development (see §6 below), that there is no demonstrated old maritime vocabulary (see above), and that there are few ancient Greek loans (Jokl, Albaner §5; but see §5 below); and that there are arguments in favor of old Dardania: Niš < Naíssos, with development as in pyll 'forest' < *pëýll < *padūle(m) : palūdem (Jokl, Albaner § 5). Admittedly, many of the arguments are negative; they are dealt with further below.

In a series of studies, G. Reichenkron has recently elaborated on Albanian-Rumanian correspondences, and has even brought in Armenian. This latter argument is not new, having been first forcefully set forth by H. Pedersen (KZ 1900:36.340-341). Pertinent aspects of Reichenkron will be discussed below, but his work does not essentially alter the borrowing situation as it has been understood. S. Puşcariu (trans. Die rumanische Sprache [Leipzig, 1943], from which citation is here made) reviews these matters under "Das autochthone Element" (pp. 203—210) and in his discussion of common Latin inheritances (pp. 326-336).

Although he deals with other views (pp. 336-338), he sees (p. 205) the Abanian-Rumanian elements as derived from Thracian, and thinks them inherited (as substratum) in Rumanian but loans into the Illyrian ancestor of Albanian. The richest account of this subject now is Rosetti Istoria II3, which commendably treats the Balkans as a historic unit. For Albanian-Rumanian the phonological correspondences are set out (pp. 103-106), as well as the lexical (pp. 106-121); many of these are too well known to need exemplification here — in the gross, they are obviously true, and largely well understood. They point solidly to (1) a local native language, and (2) a special dialect of Latin.

G. Reichenkron (Romanistisches Jahrbuch 1960:11.19-22) rehearses succinctly a number of hypotheses, which I summarize here:

a) Not all Albanian-Rumanian correspondences are loans from Albanian into Rumanian; they may be from Illyrian and Daco-Thracian as sources.

b) "Autochthonous" elements of Rumanian show only in part Illyrian-Thracian-Albanian regularities; in part proto-Romance developments appear.

c) Most Albanian-Rumanian correspondences come from borrowings by Vulgar Latin (as precursor of Rumanian) in Dardania from an Illyrian substrate. Then, we suppose, pre-Rumanian moved north of the Danube and merged with a Daco-Romance dialect, which contained Thracian elements showing correspondences with Armenian (allegedly a sound shift, and certain affixes dealt with in Rom. Jb. 9; for details, see below).

d) Daco-Thracian yields Rumanian č < IE *q before eu; ţ < IE *s + front V, and IE *k; -f- < IE *p ( > p').

e) Of the residue of unexplained words, loans from Slavic and Magyar account for many.

f) Some ancient Greek loans are to be reckoned with, even though one would not expect Rumanian to borrow wholesale in areas where other Romance did not.

g) There are also some Germanic loans. Therefore, we must reckon with five IE components: Germanic, Latin, Greek, Dacian, Slavic.

h) We must be prepared for the situation where two unrelated etyma fall phonologically together but continue two meanings, such as OFr. mont 'world, mountain' < mundum, montem; this possibility has too often been overlooked.

Reichenkron's reasoning (Rom. Jb. 1958:9.59-105, esp. 59-62) on the Albanian-Rumanian sound correspondences runs as follows: Such correspondences might reflect either (1) Daco-Thracian to Rumanian, and to Illyrian, which later becomes Albanian; or (2) Illyrian, which later becomes Albanian, to Getian Thracian to Rumanian. On the basis of the assumption of a Thracian sound shift from IE, similar to that in Armenian, Reichenkron follows Gamillscheg's theory that the West Rumanian dialects (i.e., Dardanian and South Danubian) go with Albanian in their loan reflexes, while East Rumanian dialects go with Thracian and show sound-shifted reflexes. Thus

Hence, the main diagnostic reflexes are: IE d, g, g, ĝ > East Rumanian t, k, k, ṣ.

On the basis of this Daco-Thracian theory, Reichenkron tries to explain various difficult Rumanian words involving z, some of which may be related to some Albanian words. He tries to elucidate certain Rumanian words in zg- as being originally borrowed from Thracian forms with a prefixed *ghō-, comparing certain Armenian developments. His attempt, which I consider unsuccessful or at best dubious, I criticize elsewhere, at least so far as the Albanian evidence goes. In any event, his main argument, whether right or wrong, would not need to affect our conclusions on Albanian, as it really has to do with the nature of Daco-Thracian and its putative reflexes in Rumanian.

Reichenkron argues repeatedly on the supposed direction of borrowing in a way that assumes that linguistic borrowing always moves from a higher sociological structure to a lower one. Without entering into the probably improvable factuality of these aspects of the cultures in question (the Dacians, Getes, and pre-Albanians), nor into the anthropologically unclear concept of equality and superiority of cultures, it is worth noting that in the case of cultures we know much about we could scarcely hypothesize in advance which way many categories of loans would move.

In the course of discussing shtrungë 'enclosure for milking animals', Rumanian strungǎ, Reichenkron (Rom. Jb. 11.51-52) has an excursus on Baltic and Slavic pa/po- 'Art, After-, Nach-'. This argument loses force when we consider E. Westh Neuhard's article in Scando-Slavica 1959:5.52-63, showing that these Slavic compounds are calques on German, built on a very slender inherited Slavic base; moreover, they seem to reflect a rather literary (or literate) cosmopolitan intrusion of German culture rather than contact on the folk level. Therefore, it would be all the less likely to see such an origin in this item of Rumanian folk culture. As for the interesting Baltic forms adduced by Reichenkron, two types of explanation seem to suffice to dispose of them as calques, too. The step terms of kinship seem clear calques on the long established Slavic use of this prefix (pásynokъ). The other compounds of "approximation" seem again traceable to German diffusion, particularly when one considers how strong this influence has been, specifically in Lettish and Old Prussian. Thus the restricted size of the Old Prussian corpus, emphasized by Reichenkron in connection with the relatively large attestation of this feature, loses its probative value. Reichenkron goes on to urge a special relationship embracing Thracian, Slavic, and Baltic, based on a po/pa- prefix, in turn associated with dialect variants comprising the lone Rumanian postrungǎ (beside strungǎ) and the obscure and otherwise unelucidated pociump and pozmóc. With the above considerations, the assumption of such a special relationship dissolves into thin air. It should be noted, in fairness, that Reichenkron (p. 53) allows the possibility that the dialect form po-strungǎ may arise from early Serbian contacts.

Reichenkron's further argument (pp. 52-53), giving an alternative to the conventional (i.e., Jokl's) accounting for pârîu 'brook', is, independently of the above question, susceptible of a different solution. Jokl had pârîu < pre-Albanian *per-rēn- (> Albanian përrua, përroni; cf. Latin frēnum > Rumanian frîu); Reichenkron suggests Thracian pa-(assimilated to -) + Latin rivus > rîu 'river'. Equally possible, if one insists on an alternative to *per-rēn-, is *per-rīvus.

5. Before continuing with the dimmer Balkan past, there are two sets of old loans in Albanian which lead us to a slender, but valuable, conclusion. It has long been recognized (since A. Thumb's basic article, IF 1910:26.1-20) that the ancient Greek loans are rare. Pre-Albanian was scarcely in close contact with Greek in antiquity. This places the Albanians north of the Jireček line.

However, Çabej has recently argued (VII Congresso intemazianale di scienze onomastiche 250-251) that these Greek loans do not necessarily remove the pre-Albanians far from Greek territory; that is, that they fit well with a location in present-day Albania, in contact either with Doric Greek colonists or with the Northwest Dorians. His points on the Doric character of the loans certainly look persuasive: drapën, Tosk drapër 'sickle' < *drapanon rather than drepanon; kumbull 'plum' < kokkumhlon, brukë 'Tamariske' < murikh, trumzë 'thyme' < qumbra ~ qrumbh. The last three (and, for that matter, reflexes of the first) occur in parallel forms in the Greek enclaves of southern Italy (though the Doric nature of these dialects is another famous debate!). But this still does not tell us precisely where the Dorians in question were at the time of contact.

There are a few ancient Germanic loans: fat 'spouse', shkum 'foam', tirq 'trousers' (Goth. þiubrokis) look best. Barić (LS 73-91) has up-to-date pertinent detail. These are supporting evidence, but do not place things any closer geographically. Presumably the farther north and east the Albanians were, the better were their chances of contacts at this time with Goths, but the whole question is uncertain in the extreme.

6. W. Cimochowski (BUShT 1958:3.37-48) displaces the Albanians much less than others: to the mountains near the Mati, north to Niš. Çabej (BUShT 1958:2.54-62) is even less willing to see them moved: on the basis of toponyms, he argues for a coastal region.

Particularly because of the relative inaccessibility of these articles, and because their theses have tended to be out of favor, it is worthwhile discussing them at some length. Cimochowski starts by reviewing, briefly and critically, Weigand's arguments (Balkan-Archiv 3.227-251) for a Thracian background for Albanian, and for an earlier home east of its present location:

a) Toponyms of Latin origin in Albania show Dalmatian, not Albanian, phonological development.

b) Inherited nautical and fishing terms are absent in Albanian. These facts are easily understood, says Cimochowski, since Albanian must have continued in remoter areas where Romance would not absorb it completely — hence not in areas where such place names of Latin origin continued strongest. The Albanians would have lived inland from the seacoast, in the mountains, but not necessarily beyond the border of Albania.

c) Certain words, such as man 'mulberry, blackberry' are shared with Thracian (manteia). But this could merely show that there were contacts; besides, Thrace-Phrygian BrigeV are known to have lived near Durrës. Moreover, Çabej thinks that even these words can be shown to be Illyrian. Cimochowski goes on to point out (p. 48) that karpë and mën are shared in the Italian pre-Romance area; hence this alleged Thracian correspondence is vitiated.

d) Certain Thracian names are supposedly explained with the help of Albanian. Of these, only Dacia Maluensis ( : mal) is well explained in this way; Decebalus ( : ballë) and Burebista (burre + bisht) are surely wrong.

e) Albanian toponyms known from antiquity do not show Albanian phonological development. That should not be surprising; from the end of the tenth century the whole of southern Albania was overrun by Bulgarians. But that does not necessarily mean that there were no Albanians anywhere in Albania.

f) Old loans in Rumanian from Albanian and shared Albanian-Rumanian developments from Latin point to an eastern origin. But the nomadic habits of the Vlachs and the herding culture of the Albanians would have brought them into contact for perhaps long periods in the past. Moreover, granting that the Albanians may well have had eastern contacts, we still do not know exactly where the Illyrian-Thracian line was, and NaissoV (Niš) is regarded by many as Illyrian territory.

From these observations Cimochowski concludes only that the south of Albania, the north around Shkodër, and the Adriatic seacoast are excluded as earlier Albanian territory; but this does not prove a Thracian relationship. There then follows a long discussion of the evidence for an Illyrian relationship, which will be taken up in part below, after which Cimochowski concludes, with Stadtmüller, that the home of the Albanians was somewhere in the vicinity of the Mat, stretching to Niš.

Çabej's claim is even stronger than Cimochowski's. He first runs through the history of views on the early Albanian habitat in a convenient way: The Albanians continue the habitat of Illyrian (claimed by Thunmann, Hahn, Kretschmer, Ribezzo, La Piana, Sufflay, and Erdeljanović). Half-Romanized Illyrians spilled south from the mountains between Dalmatia and the Danube (the view of Jirecek). In the third through sixth centuries, as nomads, they moved from the Carpathians south (Parvan, Puscariu, Capidan). They came from Pannonia (Procopovici, Philippide). Albanians and Rumanians were in Thracian territory between Niš, Sofija, and Skopje (thus Weigand). Albanians were in Dardania, where Illyria and Thrace meet, and moved to Albania in the late Roman period, so that the Slavs found them in the Bojana basin (Jokl, Durham, Skok). From the Balkan and Rhodope mountains they moved to Albania before the Slavs (Barić). They were in the Mati basin in Northern Albania, and expanded south in the Middle Ages (Stadtmiiller). This last location is too restrictive, according to Çabej. However, in VII Congresso internazionale 245, Çabej relates Mathis fluvius (Vibius Sequester) to mat 'river bank'.

Çabej points out that villages in the Balkans are generally of recent date and changeable settlement. Hence for the study of toponyms city names and rivers are best. If we inspect such names attested by ancient sources, we find that many follow Albanian phonological development: Scardus > Shar, with no metathesis, as in Scardona > Skradin. Scodra > Shkodër; Çabej remarks that sk- > h- belonged to the pre-Balkan period, and compares (VII Congresso internazionale 244), for phonology, shkamb < scamnum and kulshedër < chersydrus. (Rogame is a recent suffixation in -ame of rëge, and therefore no problem because of the medial -g-.) Barbanna > Buenë is regular, as shown by Jokl (IF 1932: 50.33 ff.), Slavia (1934-1935:13.286 ff.), Glotta (1936:25.121 B.). Lissus > Lesh (cf. missa > meshë, etc.); Çabej points out (VII Congresso intemazionale 245) that Latin ĭ + CC is regular, a statement I can neither affirm nor control at the moment. Dyrrachium > Durrës, Isamnus > Ishm, Drivastum > Drisht show, as Krahe claims, the Illyrian initial accent. Shkum(b)î < Scampinus is regular in the Central Albanian dialect, where pretonic ë > u and mb > m are expectable (VII Congresso internazionale 246). Aulwn > Vlorë may perhaps involve a Slavic intermediary. Thyamis > Çamëria, as Leake saw in 1814, is accepted by Çabej; however, one might expect s < t (cf. pus 'well' < Lat. puteus). Arachthos > Arta is supposedly better explained by Albanian than by Greek; but, apart from the surprising syncope, kt should yield ft or jt, and not t, from that time level. Ragusium (Ragusa) is Rush in Bogdan (1685).

Thus, says Çabej, the seacoast has remained Albanian since antiquity.

The foreign names represent several layers of later intrusions, which Weigand failed to weed out, and treated indiscriminately, according to Çabej (VII Congresso internazionale 243).

Barić (LS 25 ff.) gives an account that is as plausible on the other side of the debate, based on the careful work done by Skok on Balkan toponyms in relation to Romance. He sees Albanian as sharing with Thracian *kt > t (p. 26), but it should be noted that, as we shall see, V. Georgiev's "Thracian" has this, but that excludes his Daco-Mysian. Using the known symbiosis with the pre-Rumanians and the place names Niš, Škup, and Štip (p. 26), Barić places the Albanians in the Dardanian-Peonian region (p. 27). He then goes on to discuss (pp. 30-34) the problem of the location of the pre-Rumanians; whether they were spread out and far north of the Danube at that time need not concern us here.

It has long been recognized that there are two treatments of Latin loans in Albanian. Barić sets forth (LS 27-28, and Godišnjak, Balkanološki Institut, Sarajevo 1.1-16 [1957], esp. 7-11) a very convincing looking solution for this duality. Latin ct, cs gives Albanian ft, (luftë 'war', kofshë 'thigh'), which matches Rumanian luptǎ, coapsǎ; these would easily represent sound substitutions after IE *kt had become *t. (One problem I see in this is ftua 'quince' < cotónum, which would have to have become *ct- almost immediately to avoid falling in with këta 'this [n.], these [m.]'.) This group also includes Albanian traftār < tract-. On the other hand, we have in derjt 'straight' < d(i)rectus and trajtonj a different outcome, which matches Old Dalmatian traita < tract-. Similarly, there are both Albanian a and e as reflexes of Latin a, which match Rumanian and Dalmatian developments. These, then, would look back to two chronological and geographical layers, one an "inner Balkan" and the other a "coastal Adriatic." Barić (Godišnjak 13) considers that since Rumanian has loans from Albanian, but Albanian has practically none in the opposite direction, these Rumanian shapes must all be "Restwörter," not "Lehnwörter"; but, as Reichenkron (above) takes into account, the loan situation may easily be more complex than this.

7. There is, then, the question of where the Albanians were when the Slavs arrived. Barić discusses this (LS 28-29). Seliščev thought that the Slavs met only Romans in Albania. He showed clearly that most Albanian territory was at least exposed to Slavs in the Middle Ages; only the central region is thin on Slavic toponyms, perhaps pointing to early concentration there by the Albanians. In my opinion, the chronology of the Slavs and Albanians in Albania is uncertain in the extreme. Barić (Hymje 77) considers the loss of intervocalic voiced C in Albanian as post-Slavic, after Jokl (IF 1926:44.37 ff.). This would explain Shkinikë 'Bulgaria' < Sclavinica; the etymon recurs clearly in the Greek enclavee as škeríšte 'in the other [Greek] language'. But these could well have had a Latin etymon in the first place. Labërija in the south has Tosk -r-from intervocalic -w- and the Slavic metathesized la-, but we could posit either order for the occurrence of these. Skok has Durrës 'Durazzo' < Dračь < Dyrrachium (but note */dú-/ is required!). Yet pre-Serbian must have accented Drač on the second syllable. Moreover, to make matters more vexed, Cimochowski (Ling. Posn. 1960:8.133-145) posits Durracion [dur:akhion], taken into Illyrian as dúraku- (after *o > a) > *dúrrać(An) > *dúrrëc(ë) > Dúrës; this enlarges on and sharpens the account referred to above in Çabej's treatment of these names.

Perhaps it is naïve to look for neat, unbroken settlement areas, and doubly so for those familiar with the prenational state of the Balkans. On the present evidence, I cannot accept as a whole any one of the above vexed solutions; nor can I reject totally any one as clearly wrong.

An improvement of Barić's presentation of the name of the Bojana river (LS 29) might be to posit from Livy's Barbanna a form *baranna (note that Berat lost its Slavic -g-) = /baranna/ > *borjan(n)a (by Slavic adoption) > *bojana (in earlier Albanian; cf. ujë 'water' < *udrjā). Here we would have all changes explained by known rules but no clear chronology.

Of course, in any event we could only prove the Albanians did, and never that they did not, precede the Slavs.

On the question of the erstwhile spread of the Albanian speech area, I. Popovic (Istorija srpskohrvatskogjezika 23 [NoviSad, 1955]) points out clear evidence of earlier remains in Crna Gora. But no argument can be raised on this, however well it may fit in with our general picture of the percolation south and west of the Albanians, for a similar argument could then be constructed for the older spread of the Tosk area to the south.

8. We must turn now to the troublesome and inconclusive question of Illyrian and Thracian, and their possible relation to Albanian.

Without entering into his arguments in detail — for I find their longer range aspects unconvincing, and his safer observations of concordances no advance over those of earlier workers — Barić (LS 24 and elsewhere) plumps for an Albanian-Armenian relationship, with Thracian as intermediate. More precisely, he would posit an Albano-Thracian and Phrygo-Armenian continuum. Note that this is quite a different relationship from that assumed by Reichenkron, above.

I.I. Russu (Cercetǎri de lingvisticǎ 1958:3.89-107) finds Illyrian to be a satəm language, and Thracian likewise; but since they have a clearly different toponymic and onomastic lexicon, they are not one and the same language. Illyrian would have been Romanized at an early date, and Albanian, since it survived as an independent, would more likely be from Thracian. But, Russu declares, the problem of Albanian is still not solved.

Rosetti (Istoria II3 51-63) reviews the question generally. The two areas of Illyrian and Thracian were divided by the Morava-Vardar river line. While asserting what I take to be his considered conclusion that Albanian is a Thracian dialect, Rosetti mentions Georgiev (p. 53) and Barić (p. 54), citing V. V. Ivanov and Hamp to the effect that Albanian is neither satəm nor centum typologically (see more on this below in relation to Illyrian), and mentioning Russu and Cimochowski as defending a satəm character for Illyrian (see below also), while C. de Simone (IF 1960:65.33) doubts the latter. A good list, of the proposed lexical equations with Illyrian and Thracian, follows (pp. 56-62). A proper consideration of this list would easily generate a good-sized essay, for there are problems on all sides, and Rosetti is essentially reporting the state of scholarship as he sees it.

While opinion may differ on the above matters, none of the positions differs essentially from positions long held by one or another worker in the field. When we turn to the recent work of Georgiev, a new ingredient is added. In his La toponymie ancienne de la péninsule balkanique et la thèse méditerranéenne (Sofia, 1961; = Linguistique balkanique 3.1), he sets up seven regions, which number among them the three groups Daco-Mysian, Thracian, and Phrygian. (Roughly, the first two match the "Thracian" of many others.) The first of these groups is evidenced by toponyms in -deva/-dava/-dova (the variants are explained by chronology) < *dhēwā, is the ancestor of Albanian, and illustrates its relation by the sound changes in the above form. Georgiev posits a whole set of phonological changes for this language, which match known developments in Albanian phonology: *é, o, ā, ē, ō, ū, au, ei (> e), eu (> e),  (> a),  (> ri),  (> s, þ), ĝ(h) (> z, đ, d), tt (> s), s (> š). This is discussed in his Toponymie 7-8, as well as in Issledovanija po sravitel'no-istoriceskomu jasykosnaniju 145 (Moscow, 1958). I am not sure that I understand what is posited for *ei and *tt in the light of what I understand for Albanian. This prelanguage would have arisen in Dacia and spread to Dardania and Eastern Macedonia, and thence down the Axios (Vardar). Georgiev mentions chronologies, but I do not know how he arrives at them.

Georgiev's Thracian is defined (p. 9) by para 'river', bria 'town', diza 'fortress'; as is customary in such matters, there are etymologies for all these. The Thracian area occupied the region bounded by the Black Sea, the Propontis, the Aegean, and the Timachus, Strymon, and Danube rivers. If Georgiev's phonological rules were to turn out to be correct, we are still faced with a formidable lexical job, in view of the sparsity of manageable items: Darda- appears as both Daco-Mysian and Thracian. The following, which we could try to fit into the Albanian schema, are declared Thracian (Issledovanija 119-121): -bistas (Boure-bista) 'pistoV', b(o')ur- (to burrë?), zeiz-, zis- (i-zi?), mal- (mal?), and the gloss skiárē 'Kardendistel' (sh-qer?).

Again, Illyrian (pp. 32-34) occupies Illyria, Dalmatia, and southern Pannonia. Here we find Delm- (delmë 'mouton'; the ordinary form is dele, and we may wonder where other such forms are found), Ulc- (ulk, ujk 'loup'). Daco-Mysian supposedly penetrated Illyria and Dalmatia by the first millennium B.C. Also, Venetic and Keltic came in from the northwest, thus giving the analyst a wide range of possible alternatives. This would allegedly explicate the two traditional conceptions of "Illyrian": Hirt, Krahe, Barić, Pokorny, Popović (centum), versus Kretschmer, Jokl, Ribezzo, Pisani, Mayer (satəm, with an ingredient of centum).

In Issledovanija (pp. 133-137) Georgiev goes on to elaborate his Daco-Mysian/Albanian/Thracian relationship; there are two theories, which he elaborates but which we can pass over here. In the "Mysian" of Asia Minor, the solitary well-known inscription yields patrizi = Greek patraVi; this would show Albanian ri < *. An inscription in Bulgaria comes up with diernēs, which is derived from *kwersna = Černā; here, supposedly, the labiovelar is palatalized and spirantized, as in Albanian. Thus Albanian and the relevant elements of Rumanian come from Daco-Mysian; Athrus > Jantra and Utus > Vit in northern Bulgaria show that that region was never Thracian, but rather Daco-Mysian.

The concrete evidence for the above claims is wholly beyond my control.

V. Pisani is well known to be against simple "Stammbaum" connections, yet he has from time to time pointed out apparent parallels in Albanian and Illyrian. In Paideia (1958:12.271) he draws an isogloss for "Macedonia-Tracia" with the words for 'name': Alb. emen, Slavic imę, Baltic emnes/emmens, Keltic ainmN, etc. Doric would also show Illyrian relics in EnumakratidaV, EnumantiadaV (both Laconian); and to these Pisani adds Laconian diza 'capra' = Albanian dhi. In Paideia (12.298) he adduces Laconian grifasqai = grafein, with "Illyrian" * > ri and Hellenized phi; and deisa 'sterco', first attested in deisozos in Leonidas of Tarentum, which he equates with Albanian dhjes 'defecate'. In his review of Volume I of A. Mayor's Die Sprache der alten Illyrier (Paideia 1958: 13.319-320) Pisani lists various Illyrian glosses, most of which show no hopeful connection with Albanian, but do show considerable philological difficulty: pelioV, pelia 'vecchio, -a' might conceivably be put in relation with plak 'old man'; we could guess at tritw 'testa' alongside trû 'brain'; medoV 'hydromel' does not occur in Albanian (see below); perhaps the most interesting is dibriV 'qalassa' ("senza etnico"), which has been suggested in connection with Albanian déet, but which Pisani thinks probably Phrygian.

Dečev thinks that Albanian is from Thracian, not from Illyrian. R. Gusmani (Paideia 1957:12.164-165) remarks: "Ora qui il D. non ha tenuto calcolo del fatto che ogni lingua è la confluenza di diverse e molteplici tradizioni linguistiche, non di un filone unico, com’egli implicitamente pensa." Thus, Albanian would possibly be from an ancient Balkan kóine linguistica, but this evades the central quesiton of how the "mixture" came about.

Jokl's Illyrian-Albanian correspondences (Albaner §3a) are probably the best known. Certain of these require comment: Strabo (7.314) eloV Lougeon : lëgatë 'swamp'. This could be *lug-, but there is also *lag- 'wet', which might of course also represent *loug-.

Ludrum : Tosk lum 'muck', Geg lym, Tosk ler, but there are also Latin and Greek cognates.

Aquae Balizae : baltë 'mud'. But Krahe (IF 1962:67.151-158) thinks Balissae is from Bal-is(i)a : *Bal-sa in Balsenz < *Bal-s-antia (: *Ap-s-antia > Absentia) : Lith. balà 'swamp' : OCS blato, Alb. baltë. Therefore, for Krahe Balissae/Balizae is "Alteuropaisch" (see below).

Metu-barbis ~ -barris is ambiguous.

Malo/untum, etc., involve root etymologies and are dubious.

Place names in -V-ste/a/o : kopshtë 'orchard', vresht 'vineyard' : (Illyrier §4) Lith. -ysta 'membership'. But even this seemingly solid item has been challenged by J. Hubschmid ("Substratprobleme," Vox Romanica 1960:19): "Letzten Endes sind sie aber vorindogermanischen Ursprungs. Sie drücken die Zugehörigkeit aus, haben ferner kollektive oder frequentative Bedeutung" (p. 177). Hubschmid claims the suffix occurs from Basque and Western Romania to Asia Minor, against Georgiev's Pelasgic -s(s)- (pp. 298-299).

Schulze's -is- in names is now Krahe's "Alteuropäisch."

That Alb. -ínj is a plural-collective is clear, but what about the meaning of Delminium?

Jokl's fragile Thracian correspondences need a thorough overhauling in the light of recent work, on more than one count.

While we must exercise due caution in the use of supposedly Illyrian forms (see below), Cimochowski (BUShT 1958:2.41-46) has some important discussion to contribute to the lasting debate on the reflexes of Indo-European "gutturals" in Illyrian. He points (pp. 41-42) to evidence for both satəm and centum character for Illyrian (-Messapic). Doubtless, he says, some proposed etymologies have been wrong: Volturex, Regontius, Regius, Rega, Genthius; yet many good examples of palatals > velars remain. Likewise, Barzidihi could be < *Barzes < *Bard-jo-s (cf. Alb. mjekrrosh 'bearded'); yet there remain many presumably original palatals written s, z, s, z, q in classical sources. Also, in his view Aquilis, Aquincum indubitably show labiovelars. Cimochowski further argues (pp. 42-44) that all satəm languages show some erratics with velar reflex for original palatal, which many scholars have tried to explain away as loans of ancient date. Jokl (Eberts Reallexikon der Vorgeschichte 1.89-94, 6.38-45, 13.29) tried to show that this occurred in the presence of r and n. Against this, Cimochowski adduces Gentius, Genusus, Epicadus, Magaplinus (the last supposedly belonging with Skt. mahant-, Alb. i madh 'big'), Bersumno beside Berginium and Bargulum, Barzidihi beside Bargilius and Bargulis. Cimochowski thus believes that Illyrian (-Messapic) shows velars where uncontested satəm languages do, and that therefore these reflexes fail to make Illyrian a centum dialect. I agree provisionally with Cimochowski's conclusion here, but on other grounds. True, the facts speak against a centum status for Illyrian; but Cimochowski has too simple a formula for the centum-satəm dichotomy. In all of his examples, the following environment always involves a resonant, while the other cognates adduced are sometimes weak or dubious or susceptible of other explanations: Vescleves, Can-davia (for which *- is gratuitously reconstructed, but which points only to *- at most), Acra-banis, Bargulis/Bargilius, Skerdis, ''AggroV. This environment matches exactly that posited by me for the merger of palatals and velars in Albanian (KZ 1960:76.275-280), and on no account depends on erratic matches in the satəm languages as conventionally understood.

A special feature of Illyrian claimed by Cimochowski is its separate reflex of the labiovelars (pp. 44-46). Before front vowels, as Pedersen and Jokl showed, Albanian distinguishes the labiovelars. Jokl correctly saw that Illyrian distinguished them, too, but tried wrongly to prove that Thracian did also. Jokl's argument rested on Akuenision (which is simply Latin Aquensium), Kouimedaba/Koumoudeba (of uncertain first element), Gouolhta (of uncertain segmentation), Zououath(r) (unclear even for the Thracian values of the letters), and Koadama (whose analysis rested circularly on the first two). Thus, according to Cimochowski, the evidence for Thracian labiovelars crumbles away. The distinct reflexes of labiovelars in Albanian and Illyrian form, then, a capital proof of the Illyrian ancestry of Albanian.

But, in the face of all this, I feel we must bear in mind that the positive Illyrian labiovelar evidence is sparse and conjectural in the extreme. Moreover, as a retention it would be, strictly speaking, only weakly diagnostic.

Cimochowski also claims that Albanian shares with Messapic au > a and with Illyrian IE *ē > ā (then *ā > Albanian o); the last would be seen in Spalatum : Spolhtion in Italy. But O. Haas (Messapische Studien 173-174 [Heidelberg, 1962]) states that au > a occurs in Vulgar Latin adaptations (Ascoli : Ausculum; Basta : Bausta), and not in Messapic itself, which had au > ao > o.

On the loss of C's before *s in Illyrian and Albanian, see Hamp (IF 1961:66.51-52).

Furthermore, Krahe's "Alteuropäisch" has added a new ingredient. For example, in his "Baltico-Illyrica" (Festschrift für Max Vasmer 245-252 [Berlin, 1956]), we see various equations that for Jokl might have been marks of kinship between Illyrian-Albanian and Baltic. With such sparse evidence, too, there is a self-defeating aspect to this scholarship; consider Krahe's equations in BzN (1956:7.1-8); Nette, Netze would match Skt. nad, nadyh, while na would match nadá-; so far so good. But then *ned- would also appear in Neta (Norway), Greek Neda, Nedwn. Which language do we have now, and how do we know when we meet a new language on this level, much less who its kin are? More recently (BzN 14.1-19 and 113-124 [1963]) Krahe has screened "Die Gewässernamen im alten Illyrien" and sorted them into "Alteuropäisch," Baltic-Adriatic, Northwest (Germanic) connections, and a newly defined "Illyrian." In the last category we find only ArdaxanoV, Artatus, DizhroV, Drīnus, Drilwn (are these related?), Genusus, Kat-arbathV, Clausala, Pizwn.

Clearly, one must be very circumspect before assigning any form definitely to Illyrian.

9. We will deal separately with the Messapic problem, partly because I have dealt with it before (Studies Presented to Joshua Whatmough 73-89 [The Hague, 1957]) and wish here to revise my statement of the problem, partly because we should not too lightly lump Illyrian and Messapic together. On this latter point, see now Haas’ eagerly awaited Messapische Studien: On his page 11, Haas states that he intends to discuss elsewhere his views on the insufficiently grounded assumption of Messapic and Illyrian unity; here he simply illustrates the flimsiness of some grounds that depend on quite arbitrary segmentations of words. On his page 12 he says that the Illyrian thesis for Messapic belongs to the past, and hopefully soon to oblivion. This is not to exclude a fresh proof, when Illyrian may in the future become better specified; it is only that no such demonstration has been made up to now. On his page 13 he states that it is possible that Illyrian names may be clarified on the basis of our knowledge of Messapic; but the reverse is methodologically unsound.

After the recent painstaking philological work on the texts by O. Parlangèli (Studi messapici [Milano, 1960]), Haas (op. cit.), and de Simone (largely in IF), no forms should be used without being freshly checked.[This article had already been sent to press when I received, thanks to the courtesy of the authors, the marvelously meticulous joint work of C. de Simone (Die messapischen Inschrifien) and J. Untermann (Die messapischen Personennamen [Wiesbaden, 1964]), continuing Krahe's Die Sprache der Illyrier. Likewise, I had not seen de Simone's article on the Messapic diphthongs (IF 1964:69.20-37), nor Parlangèli's review of Haas (Kratylos 1963:8.179-186).]

Apparently without having seen my above-mentioned article (abbr. A&M), (Çabej has dealt with some Messapic words in his "Unele probleme ale istoriei limbii albaneze" (Studii şi cercetǎri lingvistice 1959: 10.527-560, esp. 555), and some of our treatments overlap.

Although the question of Phrygian takes us beyond the scope of this paper, those interested should now consult further (ad A&M 76) the recent papers of Haas in Die Sprache and Linguistique balkanique.

Taking up specific points in A&M: (§3.1) Çabej (p. 555) has likewise remarked this. (§3.3) Çabej adduces balias, balakriaihi, bálakros, Pliny's balisca vitis, and Apulian dialect bálaku, all beside Albanian balosh (term for horses and cattle with a white forehead). On the other hand, de Simone (IF 1962:67.36-52) lucidly reads baleyias as baleas (= Illyr. Diteius, Poteius, Ateia, etc., in form); connects Bales (< *balas) with balásh; and says that Messapic balasiiri[hi] is not to be equated with Bálakros. But the connection of these forms directly with balásh is inexact, for this -l- comes from *-lC- and not from *VlV or *-l-.

(§3.4) Haas (Mess. Stud. 144) posits for bijë *bhūlā, but that cannot be, for it would give *byjë. If bolles and bili(v)a really reflect *bhūl(i)-, as Haas assumes (pp. 28, 41, 131, 142-144), then we must abandon the Albanian equation. On bir, Pisani (IF 1959:64.170 n. 1), after E. Risch, has fīlia primary to the secondary fīlius and *putlo- remodeled to puer after gener, socer; here might be a parallel to bolster *bi- > bir. (§3.5) If the suggestion of delme 'sheep' to the name of Dalmatia is sound, then my suggestion falls away. (§3.10) If Alb. mëz really joins Basque mando 'mule', as Barić (Hymje 57) has it, then these go with the -st- suffix above. Barić also includes here (h)ardhí 'grapevine' : Basque ardao 'wine' and bisht 'tail' : Basque buztan. (§3.13) I hope to refine the account of mjegullë on another occasion.

(§3.14) Çabej also adduces ndë 'in', but not the others. (§3.16) Pertinent to the comparative aspect of the discussion of atavetes and sivjet now is Mycenaean za-we-te (opposed to pe-ru-si-nwa PY Ma 225) = kjawetes 'this year' according to Palmer and Killen (Nestor 240 [March, 1963]), and 85-u-te, which would not be *sjawetes, as Palmer wants, according to Killen (Nestor 258). In Mycenaean *kj and *tj would perhaps give the same result in this instance. (§3.19) Krahe (IF 1959:64.248) sees here the Messapic suffix -ido, also seen in alzanaidihi (gen.). This could then be compared to the Albanian plural and diminutive -z-. (§3.24) Çabej, too, adduces this equation. (§4.3) Çabej wonders whether veinan is not to be equated with Lith. víenas. Note that Haas (Mess. Stud. 37 and 221) continues the unacceptable reconstruction of Albanian vetë as *se-ti- by suggesting a comparison with Messapic vetai 'ihr selbst'.

In passing, it is worth observing that Haas (p. 95) makes an identification and Messapic reconstruction that is suggestive of a new line of thought. He translates aran as 'illam' (contrast A&M §3.1) and compares Umbrian oro-; this may or may not be so. Here (and again on p. 177) he translates ennan also as 'illam', reconstructing *enām and comparing Greek enh 'jenen Tag', OCS onъ, Latin enim. If so, this same reconstructed shape would also accommodate Albanian një 'one', and the sense is not too far off.

Also (Haas, 46 ff.), graiva, graibia (-f-) (a feast in Tarentum), derivative of an old u-stem, allegedly seen in B.1.43. grahis damatria *grāūs = grauV (Haas, op. cit. 142 on ū > i), suggests Albanian grā plural of grua 'woman'. On this last word, see Hamp (KZ 1960:76.276).

(§4.6) I am glad to see that Haas agrees (pp. 184-185) with me (save for a few details on which I am unclear) on the developments of the "gutturals." The separate reflexes shown for the labiovelars (pp. 185-187) and the dentals derived from palatals (p. 188) are highly suggestive of Albanian, but the supporting examples are as yet insufficiently certain. On this matter, see also the discussion of Cimochowski’s ideas above. Haas' discussion of the labiovelars in Ling. Posn. (1953:4.78-80) seemed to me inconclusive by comparison with his later work. Further, Barić (LS 9-14) discusses the satəm question; all his examples for Illyrian centum features (p. 11) are dubious, as is his treatment of kr. On palatal-before-resonant in Albanian, see Hamp (KZ 1960:76.275-280); and cf. the above observations on the Illyrian question. Popovic (pp. 21-22) essentially follows Barić for Illyrian and satəm-centum. The matter of supposed palatalization of labiovelars in Thracian goes beyond our discussion. S. Josifović "Nova mišljenja o indoevropskim jezičkim supstratima na Balkanu," Godišnjak Filozofskog Fakulteta u Novom Sadu 1959:4.97-115) comments on Budimir's works of 1950 and 1956 on "Pelasti," and on Pavlović's of 1957 and 1958 on the Mediterranean substratum in the Balkans. Most of this need not concern us here, but Budimir allegedly (p. 99) separates kohë from časъ < *kso-, which he relates to česati, česno, kosa, xainw, xew, xoanon. "Thus Albanian preserved the explosive character of the palatal gutturals for a long time, which is not the case with the other satəm dialects, and besides it distinguished velars and labiovelars in contradistinction to satəm languages." It may be so, but I do not see how this statement follows naturally from the context.

(§4.7) On haivayias (p. 89), see de Simone, IF 1960:65.31-34. The ghost word ana now falls, away; see L. Ognenova, Studia in honorem D. Dečev 333-341 (Sofia, 1958) K(uri)e bohqh ''Anna, and E. Çabej (BShkSh 1957:2.122-126, conveniently reported by M. Lambertz, Südostforschungen 1959:18.402-403) ana[b]ohqh 'I(h)s(ou) K(u)r(ie).

In addition to the forms I have discussed, Çabej (p. 555) has also proposed the following equations: Meduma (place name) = Albanian i-mje(t)më 'middle' (which, however, it should be pointed out, is simply a productive derivative in -me of the particle mjet); tabaras, tabarra 'priest, -ess (?)' = Albanian preverb (fossilized) - plus the root bar- 'carry'; ma = Albanian mos 'modal negative'; (ma)kos 'ne quis (?)', (ai min)kos 'si quis (?)' = Albanian kush 'who?'. Most of these Messapic forms are as yet of highly uncertain interpretation; consult Haas for more detail, Çabej also suggests Calabrese dialect menna minna = Albanian ménd 'suck, nurse'.

For Salentine Greek, G. Rohlfs (Die Sprache 1959:5.173-175) has proposed a Messapic etymon *squèros for the word sero, and puts this in relation with Albanian hirrë "Käsewasser."

W. P. Schmid (IF 1960:65.26-30) reads Messapic genitive + no and equates this with Letto-Lithuanian nuo; since Lettish shows gen. sg. and dat. pl. here, Schmid posits original ablative syntax, which Messapic would have lost.

10. We come now to the proposed relations between Albanian and other Indo-European groups. The material will be quickly passed in review.

M. Durante ("Etrusco e lingue balcaniche," Annali... Napoli 1961:3.59-77) has some hazardous implications tied to a few observations on Albanian which do not convince me.

It is convenient here to reproduce Georgiev's subgrouping of Indo-European (Issledovanija 282-283):

North: Baltic-Slavic-Germanic, perhaps Tocharian
West: Italic-Keltic, Venetic, Illyrian
Central: Greek, Daco-Mysian (including Albanian), Indo-Iranian, Phrygian-Armenian, Thracian, Pelasgic
South: Hittite-Luwian, Etruscan
Apart from the many debatable points that fall outside the scope of this paper, since Albanian belongs to the largest group, there is little to say about crucial problems. If one thing is clear to me, it is that no special relations have as yet been proposed for Italic, Keltic, or Anatolian. But in this field perhaps anything can happen.

11. By way of orienting ourselves, summarizing open issues, and correcting some unevennesses in past scholarship, we will now consider matters dealt with in Porzig's Gliederung.

First, some corrections, dorë 'hand' (p. 187), -ceir, etc. are taken back to *ĝher- 'greifen', i.e., *ĝher-s-; of course, in the light of Hittite the preform is something like *ĝhsr (fem. in Albanian < old neut pl. ?). The apparent reconstructions for Albanian are sg. dorë < *ĝhē(s)rā, pl. duar < *ĝhē(s)res (C-stem). An alternative, preserving the gender considerations along with formal shape, is dorë fem. < (by form-class analogy) *neuter < *ĝhē(s)r-An (in Jokl's symbols) = *ĝhē(s)r-om, thematized from *ĝhsr; duar < *dor < *gera < *ghē(s)ra.

zjarr (p. 163) 'fire' is derived from an n-suffix form, and is equated with Skt. ghá m. 'Glut, Hitze'; but as I have demonstrated in a recent oral presentation, this is not a separate lexeme, but rather an old n-plural with suffix in suppletive relation to the sg. zjarm.

mjal-të (p. 203) is an interesting case where careful dialect study pays off. In a few villages of Greece that show the contrast, and in reflexes in some enclaves of Italy, we find that we have mjáltë 'honey', in contrast to 'mud'. Thus the first is not an original *lt cluster, but has lost a vowel by syncope; on the other hand, the l (not orthographic ll) must come from an old cluster, and *ll is the only plausible one. The etymon is, then, the Latin word, and not Indo-European. Thus, Albanian here goes with Balto-Slavic, Tocharian, and Aryan, after all. Culturally, this gains in interest when we recall that Jokl (Linguistisch-kulturhistorische Untersuchungen aus dem Bereiche des Albanischen 289-296 [Berlin, 1923]) has traced bletë 'bee' to Latin *mellētum.

For 'hit' < 'split' (pp. 204-205) alongside Lat. feriō, ON berja(sk), Lith. bar(i)ù, Lett. baŕu, OCS borj, it is likely that we should posit Alb. bie < *b(h)erjō, homophonous with bie 'carry' < *b(h)erō.

Now to some more general matters. The fate of the syllabic resonants (pp. 66-68) is a vexed problem. It is difficult merely to establish the facts. Unfortunately, S. E. Mann's article on the subject (Lg. 1941:17.23) is correct or cogent only where the same solution has been proposed many years before him. The question must be entirely rediscussed, but we would do well to start from Jokl (Die Sprache 1963:9.120-122).

For the voiced aspirates (pp. 68-72), Albanian fits in with Baltic, Slavic, Macedonian, Illyrian, and Keltic; but this is not diagnostic.

On the matter of gutturals (pp. 72-76), I have already stated my position in A&M (see above); see also Jokl, Die Sprache 1963:9.123-127. On the kl (p. 75) of so-called Grenzdialekte, see Hamp, KZ 1960:76.275-280 and above.

On dental-plus-dental (pp. 76-78), Indic tt and Iranian st point to *tst (which we see in Hittite); see also A. Meillet, Dialectes indoeuropéens 60. Greek st and Balto-Slavic st point to *tst, according to Meillet, op. cit. 61. Italic, Keltic, and Germanic, however, share ss (which could conceivably come from a mediate *ts). Porzig refers (p. 77) to "die Lücke unserer Kenntnis beim Armenischen und Albanischen." Meillet (p. 57), however, has st for Albanian, Illyrian, Thracian, and Phrygian. The truth is that Albanian shows a present-day s (pasë 'had [participle]', besë 'faith, loyalty'); see Hamp, KZ 1961:77.252-253. This must go back to a groove affricate, perhaps *ts.

Albanian preserves many interesting old suppletions in the verb-stem system, among them 'sit', 'stand', 'lie' (these latter are poorly distinguished in the Balkans), and 'see'; here the situation is unlike Balto-Slavic, Germanic, and Latin (Porzig, 91-92).

Porzig makes a number of Greek equations (pp. 177-179) which require comment: Hom. odmh Att. osmh : amë 'odor' is too ambiguous to be certain. kapnoV : kem < *kepnos would be of unclear relationship. koiloV : thellë (not thelë). What is special about the *-nō of ufainw : venj 'weave'? marh : marr 'take' is dubious. Why should alfi : elb 'barley' be < *'white'? dhmoV : dhjamë 'fat' is not a correct match in vocalism. xenfoV : huaj 'stranger' (not huai). dorpon : darkë 'supper' also involves drekë 'noon meal' (in ablaut), and I think this is to be equated with the otherwise unexplained Breton dibri ~ dribi 'eat', which I deal with elsewhere. The (Greek-Armenian-)Albanian ândërrë 'dream' is an archaism, and thus nondiagnostic, as are many other items above.

Porzig (p. 179) notes that when some other subgroup is involved in an equation, Greek or Balto-Slavic always is. 'voice' (p. 180) matches Balto-Slavic and Armenian, djathë 'cheese' (dithë is a ghost, IF 1962:67.144), with Indic and Balto-Slavic matches, is a good comparison, yet a survival; but the others on pages 180-181 are trivial. (h)yll 'star' cannot be used in evidence, since it is still unclarified. mjekrë 'beard' has a perfectly regular k, and thus means nothing (see KZ 1960:76.275-280).

For Porzig (p. 181) Albanian is Eastern Indo-European and goes with Greek, and especially with Balto-Slavic. In Western Indo-European it is supposed to be connected with Illyrian only.

12. Meillet (pp. 109-113) summarizes the distributions of -ye- presents: Greek and Indo-Iranian have -ye/o-; Balto-Slavic and Armenian havefor "state" and -ye/o- for derived verbs; Germanic and his Italo-Keltic have -yo/, almost leveled out for both verb types. In the avant-propos to the reprinted edition (p. 14), Meillet adds Albanian to the last set.

He states erroneously (p. 17) that *prəwo- is limited to Indo-Iranian and Slavic; Albanian has i-parë 'first'. Another set in agreement (Skt. dáhati, Lith. degù, OCS žeg, Alb. djeg 'burn') is probably too routine to be important.

To Meillet's "northwest" vocabulary (which includes Balto-Slavic) we could add grurë 'wheat' (p. 18), shat 'hoe'(p. 21), and mos (modal negative; p. 23). There is not much positive evidence here; these are, in the main, retentions.

Perhaps of considerable importance are the following two traits. Baltic and Slavic lack perfect reduplication, the archaic state of affairs (Meillet 104-107); Albanian agrees. Balto-Slavic also shows an old indicative aorist, but an active participle from the perfect; Germanic, Keltic, and Italic have pooled the old aorist and perfect. Here Albanian seems to agree on a slender base with Balto-Slavic. In obscure ways we may see agreement in mora 'took' : marrë, lashë 'left' : lënë, erdha 'came' : ardhurë, hëngerë 'ate' : ngrënë.

Albanian is particularly rich in *ē preterits, and this belongs to a lengthy discussion — too long for this paper — of the role of the long ablaut grade in Albanian, a discussion that Jokl (IF 1916:37.90-122) only opened and failed to see in its far-reaching main issues. Such a discussion would find a new base in the studies of Kuryłowicz.

13. F. B. J. Kuiper (Annali... Napoli 1960:2.159-164) has shown that OP qātiy 'says' is a root present qā- in suppletion with qah-. I think we must abandon the old equation with Latin cēnseo (on other grounds, too) and place Albanian thom, thotë (3 sg.), aor. tha, alongside these. What has never been adequately brought out is the fact that if thom 'I say' is *kē(n)smi (which it could be), thotë cannot be *kē(n)st..., which should give *thoshtë. Since *kēn(s)t... would give *thând (or perhaps *thân), we must posit *kēti for the 3 sg.

14. A very important item has recently appeared, which relieves us to a large extent of the task of exposition that would otherwise be called for to bring together the very scattered earlier literature; that is the posthumous long article of Jokl, "Die Verwandtschaftsverhaltnisse des Albanischen zu den ubrigen indogermanischen Sprachen" (Die Sprache 1963:9.113-156), which was located only a few years ago among Jokl's Nachlass, now housed in the Vienna National Library, in the original draft in Gabelsberger shorthand. This article sets forth the distinctive phonological developments of Albanian from Indo-European from the point of view of what is shared with other groups (pp. 116-129); the similar morphological characteristics (pp. 129-148); and the lexical correspondences (pp. 148-156). There is not space here to criticize Jokl's points in detail, and I intend to consider the more interesting points on another occasion. Moreover, some have been implicitly dealt with in the foregoing discussion; finally, since Jokl's article is about forty years old, there should be no wonder that even in so neglected a field as Albanian many items now simply fall away, overtaken by more recent scholarship, some of which is reflected in the preceding pages.

Fundamentally, Jokl's presentation suffers from two technical defects by current standards: an atomistic approach to the data, and a failure to distinguish conservative features from innovations. I shall deal with Jokl's results as concisely as is consistent with clarity.

15. Jokl summarizes his phonological results (p. 129) by stating that they fit well with a North European affiliation, specifically with Balto-Slavic. Jokl's considered opinion that Albanian goes most closely with Balto-Slavic, if often on the basis of far-from-obvious evidence, is well known from various of his publications; on this, see the relevant sections of Porzig's Gliederung. I have also expressed in print my agreement with Jokl on this to the extent that a clear-cut opinion can be held at this stage of our knowledge. However, the phonological evidence now in question scarcely manages to support this, no matter how correct the view may be. The Albanian merger of *o and *a, of *ē with earlier *ā, the phonetic drift of *[ā] to o, the treatment of *ə, and of *ouo, and the change of *s to [š], are all too isolated structurally (as presented) and too nonunique as events to associate Albanian clearly with any one group of dialects; they merely make certain trivial exclusions likely. The "helle Färbung" associated with *- > ri does not really match Balto-Slavic and Keltic either in allophonic distribution or in phonetic detail without a great many more supporting considerations. Perhaps we may ultimately be able to sharpen these claims; at present I see no clinching phonological link, in the form of a structured shared innovation, with any other Indo-European group.

16. The conclusions presented (pp. 147-148) for the morphology make similar claims. We will inspect here only the correspondences that include Balto-Slavic; that is not to say that on a complete reexamination of the problem we might not find further fundamental shared features involving other groups more intimately. One immediate disadvantage is noted in the nature of the features identified: they are almost all derivational morphs and processes of some sort; thus, they belong to a less structured, more open-ended, part of the grammar.

The development of the type foroV (shared with Greek, too) was of old formation, and is not diagnostic. Long-grade nouns (also shared with Germanic) and preterits (more broadly shared) seem ultimately important to me, but cannot be assessed on this brief survey basis. Participles in -mo- (shared with Armenian) and in -eno- (also Aryan and Germanic) belong to an old layer of adjectival derivation. Nouns in -imo-, if related to similar feminines (e.g., in Keltic), are also difficult to assign a precise innovational status to. Collectives in -īno- (also claimed for Latin and Germanic) are of ambiguous standing. Verbal nouns in - are claimed also for Armenian, but surely Tocharian and Hittite further complicate this picture.

The spread of -m to the neuter of interrogatives, if true, and the formation of the accusative of 1 sg. and 2 sg. personal pronouns, are features that are most difficult to evaluate at such reconstructive distance. The use of active plus reflexive for a passive is not quite parallel in Balto-Slavic and Albanian (where it is restricted by tense); but, if pertinent, that and the syntax of the 'teens of the numerals and perhaps the combination of interrogative and demonstrative pronouns seem good candidates for diffusional origin.

This leaves as possible uniquely shared items: diminutives in l + -o, -to in ethnica, alternation of -ti and - (a weak possibility), verbal nouns in -g- and -esā, and secondary adjectives in -usto. Considering that many of these call for searching discussion, and might not stand up very well in all cases, it is scarcely an impressive list. Moreover, our views of Indo-European have altered greatly since Jokl was writing, and only a full reexamination ab initio could take proper account of this.

Clearly, the whole question remains completely open. But perhaps we have been able to clear a little ground.

17. For convenience, and by way of summary of the main fields covered, a selective Bibliography is appended.