Etymological Dictionary of the Slavic Inherited Lexicon Online

Etymological Dictionary of the Slavic Inherited Lexicon Online

Introduction

1. Origin of the dictionary

The dictionary is based on a database that was created within the context of the Indo-European Etymological Dictionary project (IED). The circumstance that the dictionary originates from a database is still apparent from the way the lemmata are structured. This type of rigid structure is at times a burden for the author, but it also increases the value of the dictionary as a book of reference.

The main objective of the dictionary is to present an up-to-date etymological account of the Slavic inherited lexical stock. Since there is no consensus on neither the reconstruction of the Indo-European proto-language nor on the reconstruction of Proto-Slavic, the etymological dictionary is bound to represent my personal views, which obviously reflect my academic background (see section 2). Few will deny, however, that especially from the Indo-Europeanist’s point of view the dictionary is more up-to-date than, for instance, the Ėtimologičeskij slovar’ slavjanskix jazykov (ESSJa) or Pokorny’s treatment of the Slavic material in his Indogermanisches etymologisches Wörterbuch (IEW).

As to the scope of the dictionary, it must be said that it has not been easy to define the set of etyma that were to be discussed. It was completely out of the question that I should be able to cover as many lemmata as the ESSJa, which at the moment contains about 20,000 entries (many of them marginally attested and of unknown origin), neither would it have been desirable for a dictionary concentrating on the inherited lexical stock. It was therefore decided to focus on etyma that have been part of the scholarly discussion in the field of Indo-European linguistics, e.g. etyma that occur in Pokorny’s dictionary or LIV. Obviously, this selection includes a number of etyma that may be considered borrowings from a non-Indo-European substratum language (see 2.2). Excluding such etyma would not be very wise, as the classification of a word as, for instance, “North Indo-European” is merely provisional. Furthermore, even the possibly non-Indo-European elements of the Proto-Slavic lexicon usually meet the IED’s criterion that an etymon must be attested in at least two branches of Indo-European in order to be included.

An important difference between the present dictionary and etymological dictionaries such as the ESSJa or the Słownik prasłowiański (Sławski SP) is the fact that an attempt is made to reconstruct the prosodic characteristics of the Proto-Slavic etyma. I regard this as a justifiable goal in itself, but Kortlandt’s theory about the origins of the Balto-Slavic acute and circumflex intonations (see 2.2.3.3), to which I subscribe, adds significantly increases the importance of Baltic and Slavic accentology for the reconstruction of Proto-Indo-European. The accentuation of an etymon may reveal the former presence of a laryngeal or be decisive in cases where it is unclear whether or an aspirated or an unaspirated voiced stop must be reconstructed.

At an earlier stage of the project I planned to present the reconstruction of the accent paradigms more or less as given facts. Then I decided to provide more information on the accentuation of the forms attested in the individual languages, for instance by adding information on the accentuation in Old Russian or presenting a number of case-forms. While this dictionary does not aim at completeness, I have strived for an accurate representation of reliable data. Since the database hopefully remains available on-line, the set of data may be expanded.

Unlike many other etymological dictionaries, the present dictionary does not abound in references. The Etymologický slovník jazyka staroslovenského (ESJS), for example, painstakingly lists the most prominent etymologies, which are subsequently evaluated. Though I highly value this approach, I felt that my dictionary had to be set up in a different manner because otherwise too much time would be taken up by reproducing and scrutinizing the scholarly literature. I realize that by limiting the number of references I am at risk of ignoring valuable contributions to Slavic etymology. This I regret, but I think that it is a concession I had to make.

2. Theoretical framework

2.1 General considerations and Proto-Indo-European

Not surprisingly, the ideas about the structure of the Indo-European proto-language underlying this book conform with the theories propagated by Leiden based Indo-Europeanists such as Beekes, Kortlandt, and Lubotsky. The best introduction to this theoretical framework is Beekes 1995, which is the English edition of Beekes 1990. The existence of this introductory work largely relieves me of the obligation to present an outline of the theories on which my Proto-Indo-European recon­structions are based. Thus, I shall confine myself to representing some of the main aspects.

It is a well-known fact that Leiden Indo-Europeanists tend to deny that there was a Proto-Indo-European phoneme *a (see especially Lubotsky 1989). Nevertheless, it is quite possible that a Proto-Slavic etymon derives from a form containing *a. The Proto-Slavic lexical stock contains numerous elements that do not have an Indo-European origin. These may have been borrowed from a substratum language, possibly at an early stage. The *a that these words may contain is sometimes called “European *a” because the substratum language was located on European soil.

I subscribe to the hypothesis that Proto-Indo-European did not have an opposition between palatalized and plain velars (cf. Meillet 1894, Steensland 1973). The latter arose from depalatalization in certain constellations, in particular after *s (though not before *i) and after *u, where the opposition between the palatovelar and labiovelar series was neutralized. Depalatalization before resonants unless followed by a front vowel occurred in Balto-Slavic and Albanian (cf. Kortlandt 1978a: 240-242). The latter development is to a considerable extent responsible for the variation between velar stops and sibilants that we observe in both Baltic and Slavic.

The traditional Proto-Indo-European system of voiceless, voiced, and aspirated voiced stops has repeatedly been challenged on typological grounds. As an alternative, it was proposed that the unaspirated voiced stops were actually glottalic (e.g. Gamkrelidze and Ivanov 1984: 5-84). The glottalic hypothesis was applied to great effect by Kortlandt, who employed it to tackle a diversity of issues in various branches of Indo-European (cf. Kortlandt 1985a). Crucial to the present publication is Kortlandt’s interpretation of the Balto-Slavic development known as Winter’s law as the merger of the glottalic element of the traditional mediae with the reflex of the Proto-Indo-European laryngeals (see 2.2.2.3).

In Proto-Indo-European, the lengthened grade vowels *ē and *ō occurred in a limited number of categories, which can ultimately be reduced to monosyllables and forms ending in a resonant (cf. Beekes 1990: 204, 1995: 167, Kortlandt 1986: 154-155). Contrarily to what is generally assumed, lengthened grade vowels are regularly circumflex in Balto-Slavic. Forms presented as counter-examples contain either a laryngeal or can be regarded as examples of Winter’s law. Another source of circumflex long vowels is contraction. A special case is the constellation *ēH (*ōH), where according to Kortlandt the laryngeal was lost (Kortlandt 1985b: 115, 118-120).

As can be gathered from the preceding paragraphs, I adhere to the view that there once existed a Balto-Slavic linguistic unity. It can be demonstrated that Baltic and Slavic underwent a sequence of common developments, a number of which relate to the place of the stress. In view of its complexity as well as the important role it plays in this dictionary, Balto-Slavic accentology will be the subject of a separate section.

2.2 Balto-Slavic accentology

2.2.1 Introduction

As stated above, Kortlandt’s theory about the origins of the Balto-Slavic acute and circumflex intonations significantly increases the relevance of Balto-Slavic accentology for etymological studies. This is reflected in numerous publications by Indo-Europeanists from Leiden, for instance in several monographs that appeared in the Leiden Studies in Indo-European series. A good example is Schrijver 1991, where a conscious effort is made to take the Balto-Slavic accentual evidence into account.[1] My own book in the series, Derksen 1996, is a slightly different matter, as it deals with an accentological subject, the problem of metatony in Baltic. It contains a brief survey of Balto-Slavic accentology, which partly coincides with Derksen 1991. I shall here repeat some of the points I tried to make in these two publications, while shifting the emphasis onto Slavic phenomena.

2.2.2 Balto-Slavic developments

Over the years Frederik Kortlandt has devised a detailed relative chronology of developments ranging from Proto-Indo-European to disintegrating Slavic. It was first published in a Serbo-Croation translation in 1989. The English version, which was marred by many misprints, appeared in 1994, to be followed by a corrected version on the World Wide Web (2002). Articles reproducing and discussing large parts of the relative chronology are Kortlandt 2005 and 2007. Kortlandt’s chronology might be called the backbone of my investigations in the field of Balto-Slavic historical linguistics, which is not to say that it will be treated as if it were carved in stone.

The Balto-Slavic section of Kortlandt’s relative chronology contains a number of developments that concern accentology. These are conveniently listed in Kortlandt 2006a (349):

“1. Loss of PIE accentual mobility, of which there is no trace outside the nominal flexion of the consonant stems, e.g. Lith. duktė̃ ‘daughter’, piemuõ ‘shepherd’, and the flexion of the athematic verbs, e.g. duodą̃s ‘giving’ (cf. Kortlandt 1985b on the latter).

2. Pedersen’s law: the stress was retracted from medial syllables in mobile accent paradigms, e.g. acc.sg. dùkterį, píemenį, Greek thugatéra, poiména.

3. Barytonesis: the retraction of the stress spread analogically to vocalic stems in the case forms where Pedersen’s law applied, e.g. acc.sg. ãvį ‘sheep’, sū́nų ‘son’, diẽvą ‘god’, žiẽmą ‘winter’.

4. Oxytonesis: the stress is shifted from a medial syllable to the end of the word in paradigms with end-stressed forms, e.g. inst.sg. sūnumì, inst.pl. žiemomìs.

5. Hirt’s law: the stress was retracted if the vowel of the pretonic syllable was im-mediately followed by a laryngeal, e.g. dúona ‘bread’, výras ‘man’, dū́mai ‘smoke’, Vedic dhānā́s, vīrás, dhūmás.

6. Winter’s law: the PIE glottalic stops dissolved into a laryngeal and a buccal part. The former merged with the reflex of the PIE laryngeals and the latter with the re-flex of the lenes stops, e.g. Latvian pȩ̂ds ‘footstep’ < *pedóm, nuôgs ‘naked’ < *nogʷós, duômu ‘I give’ < *dodH₃mí.

7. Retraction of the stress from final open syllables of disyllabic word forms unless the preceding syllable was closed by an obstruent, e.g. Lith. gen.sg. vil̃ko ‘wolf’, dat.sg. vil̃kui, gálvai ‘head’, nẽša ‘carries’, Serbo-Croatian vȗka, vȗku, glȃvi, nȅse ‘carried’, neuter pȋlo ‘drank’, but Lith. gen.sg. aviẽs, gen.pl. vilkų̃ < *-òm, nom.sg. galvà < *-àH, Russian pilá ‘she drank’ < *-àH, neuter nesló, infinitive nestí, where syllable-final consonants (including word-final laryngeals) prevented the retraction of the stress.”

2.2.2.1 The rise of the mobile paradigm

The developments 1-4 and 7 are intended to account for the accentual curve of the Balto-Slavic mobile paradigm. Here a few words on the historical background of the problem are in order. According to de Saussure (1896), the Lithuanian opposition between barytona and mobilia continues an Indo-European opposition between barytona and oxytona. As the identity of the Lithuanian and Proto-Slavic mobile paradigms is beyond doubt, this implies that the Balto-Slavic mobile paradigm arose from an oxytone paradigm. De Saussure’s explanation for the origin of the Lithuanian mobile paradigm started from consonant stems of which the number of syllables did not remain constant within the paradigm. In Kortlandt’s chronology it appears under the name Pedersen’s law in view of Pedersen’s effort (1933: 25) to reformulate the law proposed by de Saussure.

The prehistory of the Balto-Slavic accentual system has been the subject of much debate (see especially Olander 2006 for an overview). Kortlandt’s development (1) places him at the same starting-point as de Saussure: the early Balto-Slavic system mainly had an opposition between barytona and oxytona. The few traces of original accentual mobility that survived only played a modest role. The question is if it is possible to strengthen the link between the Balto-Slavic accentual mobile paradigm and Proto-Indo-European accentual mobility. We may note that Illič-Svityč, when he set out to provide comparative proof for de Saussure’s hypothesis on the Indo-European background of the Balto-Slavic barytone and mobile paradigms (1963, English translation 1979), tried to gain a better understanding of Pedersen’s law by suggesting a link with the survival of mobile root nouns. With respect to the Indo-European situation he preferred the term “mobile-oxytone” to “oxytone”.

It so happens that Kortlandt himself (2006b) has recently modified his account of the rise of the Balto-Slavic mobile paradigm. Instead of assuming an early loss of Proto-Indo-European accentual mobility, he now starts from the Late Indo-European hysterodynamic and proterodynamic paradigms. The first step is the replacement of the isolated root stress of the Nsg. of the hysterodynamic paradigm by final stress. The sigmatic Nsg. of the proterodynamic paradigm then underwent the same development, creating a clear accentual opposition between end-stressed masculines and feminines on the one hand and root-stressed neuters on the other. The medially stressed case-forms, viz. the Asg., Lsg., Apl., and Npl. forms of the hysterodynamic paradigm and the Dsg. and Npl. forms of the proterodynamic paradigm, subsequently retracted the stress to the initial syllable. Kortlandt’s objective here is to reformulate Pedersen’s law as a phonetic development. Furthermore, the scope of the barytonesis has been limited:

“The barytonesis did not affect acc.sg. ãvį ‘sheep’, sū́nų ‘son’, which had preserved Indo-European radical stress, nor žiẽmą ‘winter’, which was built on the original nom.sg. form *ǵheim (cf. Beekes 1985: 44), but did yield the retraction in diẽvą ‘god’, cf. Vedic devám, because the o-stems had fixed stress from the outset.” (Kortlandt 2006b: 3)

Finally, the existence of oxytonesis as a Balto-Slavic development is put into doubt. The accentuation of Lith. sūnumì Isg. and žiemomìs Ipl. may be old because it replaces the end-stressed instrumental in *-bʰi. On the whole, the new scenario links the Balto-Slavic state of affairs more closely to Late Indo-European nominal accentuation.

The retraction listed as number 7 was first formulated by Kortlandt in 1975 (5-7). Since it was inspired on a retraction formulated by Ebeling, it was baptized Ebeling’s law, but in recent publications by Kortlandt this designation tends to be avoided. An interesting consequence of the condition that the stress was not retracted to syllables ending in an obstruent is the rise of a class of oxytone neuters. In Slavic, these oxytona ended up in AP (b). If the root contained the reflex of a laryngeal or the laryngeal part of a PIE glottalic stop, it was lost in pretonic position. In Baltic, the oxytone neuters became barytone when the stress was retracted from final *-à, yielding metatony (Derksen 1996: 96-128, 229-232, see also 2004: 87-89). Words belonging to this class were occasionally misinterpreted by Illič-Svityč (1963). The effects of Ebeling’s law may also be observed in masculine o-stems (Derksen forthc. a).

2.2.2.2 Hirt’s law[2]

Hirt’s law, which is listed above as development 5, was proposed in order to account for the large number of correspondences between Baltic and Slavic barytona and nouns which considering the Sanskrit, Greek and Germanic evidence had mobile/oxytone accentuation in PIE. In Hirt’s original formulation (1895) the stress was retracted to long root syllables. Since then the law has been reformulated a remarkable number of times, among others by Hirt himself (1899). An important observation was made by Bonfante (1935, 1937), who showed that the stress was only retracted to non-apophonic long vowels, i.e. to sequences of a short vowel and a laryngeal.

After a thorough investigation of the “Hirt-Bonfante hypothesis”, Illič-Svityč (1963: 80 = 1979: 63) concluded that the retraction was indeed limited to non-apophonic long syllabic elements, i.e. to non-apophonic long vowels, long resonants and long diphthongs. He contrasts syllables containing “new length” from laryngeal loss after a syllabic element with syllables containing apophonic length or an original sequence of a short vowel followed by a resonant and a vocalized laryngeal, e.g. *tenəu̯ós < *tenHu̯ós, cf. Gk. ταναός ‘outstretched, tall’, Latv. tiêvs ‘thin’. This invites the conclusion that at the time of the retraction the laryngeals were still present, as has indeed been argued by some scholars (cf. Kortlandt 1975a: 2, Rasmussen 1985: passim). In that case one might simply say that the stress was retracted to an immediately preceding syllable containing a vocalic element followed by a laryngeal. That the position of the laryngeal plays an essential role was demonstrated by Kortlandt in connection with the accentuation of the Slavic l-participle (1975: 2-4). He suggested that in instances such as Ru. pilá f.sg. ‘drank’ the laryngeal must originally have preceded the i of the root. His reconstruction *pHiláH is supported by full grades of the type *pe/oh₃i- or *pe/oh₃- e.g. Skt. pāyáyati. The opposition between *-HI- and *-IH- has a parallel in Greek and Italo-Celtic, where we find indications that pretonic *-HI- yielded a short reflex (Schrijver 1991: 512-536). It may be clear that Hirt’s law is a strong argument for a Balto-Slavic linguistic unity.

2.2.2.3 Winter’s law

Winter’s law, which in its original formulation is vowel lengthening before PIE unaspirated voiced stops (Winter 1978), is without doubt a sound law of major importance. So far, however, it has not received the recognition it deserves. The main reason for this is probably the fact that a number of appealing examples seem to violate the law. Since a survey of the evidence clearly indicates that the law is essentially correct (cf. Young 1990, Rasmussen 1992, and especially Dybo 2002), the next logical step is to look for special circumstances which might provide an explanation for the apparent exceptions. For ‘water’ (cf. OCS voda vs. Lith. vanduõ 3^a) and ‘fire’ (cf. Lith. ugnìs, OCS ogn’ь), Kortlandt has proposed that the law did not affect the clusters ndn and ngn (1979: 61, 1988: 388-389). The nasal infix which may be reconstructed for Balto-Slavic must have developed from a nasal suffix in PIE times already (cf. Thurneysen 1883). Another major exception is Slavic *xodъ ‘going, course’. Here the absence of Winter’s law may originate from a reduplicated present stem *sizd-, where the law was blocked by an intervening z (Kortlandt 1988: 394).

This is not the place to discuss the various attempts to modify the formulation of Winter’s law, for which I refer to Derksen 2003a, 2004, and forthc. b. I would like to elaborate, however, on Kortlandt’s interpretation of Winter’s law and its relationship to the Balto-Slavic prosodic system. According to the traditional doctrine, the Balto-Slavic acute intonation, which is usually reconstructed as a rising tone, reflects length, i.e. original length or length resulting from the loss of a laryngeal. If Winter’s law is interpreted as vowel lengthening, the fact that the law yields acute long vowels and diphthong is only to be expected. The regular reflex of a lengthened grade, however, is circumflex in Balto-Slavic, as Kortlandt has argued on several occasions (e.g. 1985b, 1997a). The main reason why this is not generally recognized is the ease with which some Indo-Europeanists postulate lengthened grades, thereby obscuring the original situation. Since both the presence of a laryngeal and Winter’s law generate acute syllables, one may try to link this observation to the hypothesis that the PIE voiced unaspirated stops were actually (pre)glottalized (Kortlandt 1978b). In Kortlandt’s interpretation, Winter’s law is the merger of the laryngeal element of the glottalic stop with the reflex of the Indo-European laryngeals, which had become a glottal stop in Balto-Slavic.

An advantage of Kortlandt’s interpretation of Winter’s law is the possibility to regard the Latvian and Žemaitian broken tones as direct continuations of a Balto-Slavic glottal element (Derksen 1995, Kortlandt 1998). This does not imply that already in Balto-Slavic glottalization existed as a vocalic feature. There are no indications that the Balto-Slavic glottal stop lost the status of a segmental phoneme which it must still have had when Hirt’s law operated (see 2.2.2.2). The Balto-Slavic distinction between acute and circumflex syllables, which was clearly independent of the place of the (free and mobile) ictus, was originally the distinction between the presence and absence of a glottal stop. It most certainly was not a tonal distinction originating from PIE (cf. Kortlandt 1985b, Nassivera 2000). The rise of tonal distinctions must probably be dated to the separate branches of Balto-Slavic. One of the subjects of the next section is the rise of distinctive tone in Proto-Slavic.

2.2.3 Slavic accentology

2.2.3.1 Introduction: Stang 1957

The starting-point of modern Slavic accentology is the publication of Stang’s Slavonic accentuation (1957).[3] In this study Stang effectively did away with a number of concepts of what is often called “classical accentology”, though in particular the interbellum witnessed many unrealistic theories. Stang ends his book with a list of conclusions (1957: 179), which I shall now try to rephrase and provide with comments (cf. Derksen 1991: 53-55).

Stang established three (Late) Proto-Slavic accent paradigms, each with its own prosodic characteristics:

(a)    Fixed stress on the stem. If the stem is monosyllabic, the stressed syllable is “acute”, i.e. we find a short rising tone on a historically “long” syllable, which is equivalent to saying that the nucleus of the root syllable is not constituted by monophthongal *e, *o, *ъ, or *ь. A special class is formed by nouns of the so-called *vòļa type, where the root has neo-acute intonation. Stang discusses these nouns within the context of (j)ā-stems belonging to AP (a), but there are good arguments to classify them as belonging to AP (b), which is what will be done in this dictionary.

If the stress is on a medial syllabe, there is a greater number of possibilities regarding the prosodic qualities of the root. Since these cases are not crucial for understanding the three basic types, I shall not go into the matter here.

(b)    The stress alternates between the last syllable of the stem and the first syllable of the ending. Stem-stressed forms have a rising tone with retention of the old quantity of the syllable. This tone is called “neo-acute”. Forms with stress on the ending have a short rising tone.

(c)    The stress alternates between the first syllable of the stem and the ending. Stem-stressed forms have a falling tone and lose the stress to a clitic. End–stressed forms usually have a short rising accent, but in some cases a long rising accent. The falling tone is sometimes referred to as “circumflex”. This is potentially confusing because the same designation has been used to refer to any non-acute long syllable or even to any non-acute syllable (cf. Derksen 1991: 55).

In classical accentology, an acute or a circumflex tone changed under certain conditions into a neo-circumflex and a neo-acute tone, respectively. This process, called metatony, yielded four distinctive tones (Kortlandt 1978c: 271). In Stang’s system there are three tones, which are all connected with a specific accent paradigm. Stang now showed that the neo-acute originated from a retraction of the stress.[4] He also showed that the neo-circumflex is an innovation of Slovene and the Kajkavian dialects of Serbo-Croatian rather than a Proto-Slavic tone.[5] The next question that we must address is the relationship between the Slavic and the Baltic accent paradigms.

2.2.3.2 Progressive shifts

Lithuanian nouns belong to one of four accent paradigms, of which (1) is barytone, while (2), (3), and (4) are mobile. Monosyllabic stems are acute in (1) and (3), while they are circumflex or short in (2) and (4).[6] If the stem is polysyllabic, the situation is slightly more complicated, but that need not concern us here. The four accent paradigms can be reduced to a barytone and a mobile paradigm if one takes into account the progressive shift which is commonly referred to as de Saussure’s law. Employing the method of internal reconstruction, de Saussure (1896) demonstrated that at a certain point in the history of Lithuanian accentuation the stress shifted from a circumflex or short syllable to an immediately following acute syllable. This development was independently discovered by Fortunatov (1897). Hence, de Saussure’s law, when applied to both Baltic and Slavic, is sometimes referred to as Fortunatov’s law. Propagated by none other than Meillet,[7] de Saussure’s law came to occupy an important place in classical accentology. The law was often considered a Balto-Slavic innovation, though Meillet regarded the progressive shifts in Baltic and Slavic as parallel developments (1900: 350-351, 1924: 145).

The decline of de Saussure’s law as a Balto-Slavic development may be said to have started with Kuryłowicz (1931: 75ff, 1952), who denied its operation in Slavic, albeit basically without addressing the facts. A much heavier blow, one might argue, was delivered by Stang (1957: 15-20), who by presenting a series of factual arguments undermined the at the time prevailing view that de Saussure’s law had also operated in Slavic. Now as we have seen, Stang reconstructed three Proto-Slavic accent paradigms, whereas the Lithuanian situation points to a system with two paradigms, one of them barytone and the other mobile. Since AP (a) corresponds to Lithuanian AP (1), while AP (c) corresponds to (3) and (4) (see the next section), the core of the problem is the relationship between AP (b) and AP (2). As shown by Stang, the neo-acute tone originates from a retraction of the stress, a development now generally referred to as Stang’s law. This means that AP (b) was originally oxytone. AP (2), however, is a mobile paradigm originating from a barytone paradigm as a result of de Saussure’s law. Stang did not have an answer for this discrepancy, but he made it clear that the answer most certainly was not de Saussure’s law.

A solution was proposed by Dybo and Illič-Svityč, who argued that the oxytone paradigm which must have existed prior to Stang’s retraction had been generated by a progressive stress shift that cannot be identified with de Saussure’s law (see especially Dybo 1962, Illič-Svityč 1963: 157-161 = 1979: 140-144). According to Dybo’s law, also known as Illič-Svityč’s law,[8] a syllable which was neither acute nor falling lost the stress to the following syllable, causing a split of the Proto-Slavic immobile paradigm. The syllable which received the stress became falling, which provided the input for Stang’s law, the retraction of the stress from long falling vowels in final syllables.[9]

The scenario proposed by Dybo and Illič-Svityč allows us to derive the Baltic and Slavic accentual systems from a stage when there were only an immobile barytone and a mobile or oxytone paradigm. As later publications from the Moscow accentological school have shown (see especially Dybo 1968a), it is possible to distinguish between dominant (“strong”) and recessive (“weak”) morphemes at this stage. The place of the stress is governed by the valency of the morphemes that constitute a given form (cf. Dybo 1981: 260-262, 2000a: 10-14, Lehfeldt 2001: 67-69). Whether a morpheme is dominant or recessive cannot be predicted on the basis of its phonological structure: the distribution of morphemes over the two classes is “traditional” (Dybo 2000a: 10).[10]

2.2.3.3 Illič-Svityč’s law and the neuter o-stems[11]

In his monograph on nominal accentuation in Baltic and Slavic, Illič-Svityč tried to explain why so many PIE neuter o-stems appear to have become masculine in Slavic, an observation which was first made by Hirt. A comparison with accentual data from Baltic, Greek, Sanskrit and Germanic led Illič-Svityč to conclude that PIE barytone neuter o-stems correspond with Slavic masculine o-stems belonging to the barytone class in the case of “long” roots and to the oxytone class in the case of “short” roots (in Stang’s terminology to accent paradigms a and b, respectively). As we have seen, AP (a) and (b) continue a single barytone paradigm, which allows the conclusion that PIE barytone neuter o-stems became barytone masculine o-stems in Slavic. This shift of gender must be rooted in Balto-Slavic (see below). In originally masculine mobile o-stems with a non-acute root, accentual mobility has been generalized (Illič-Svityč 1963: 109-119 = 1979: 94-104), a development that is sometimes called Illič-Svityč’s law. Thus, Slavic masculine o-stems belonging to AP (b) in principle continue old neuters. I consider it possible, however, that masculine o-stems that were oxytone in Late Balto-Slavic, i.e. after Ebeling’s law, escaped the transfer to the mobile class (cf. Derksen forthc. b).

Whereas the barytone neuter o-stems became masculine, PIE oxytone neuter o-stems remain neuter in Slavic. According to Illič-Svityč, the majority of the Slavic neuter o-stems belong to the oxytone class, Stang’s AP (b). Mobile neuter o-stems (c) contain, as a rule, a historically long root or have a i̯o-suffix. In my opinion, the distribution between AP (b) and (c) is not completely clear. We can say with a high degree of certainty, however, that originally oxytone neuters of the structure CVC₁C₂-ó (where C₁ is an obstruent) belong to (b), in conformity with Ebeling’s law (see 2.2.2.1). Proto-Slavic neuter o-stems belonging to AP (a) originate from the retraction generally known as Hirt’s law, which generated a new class of neuter o-stems with fixed root stress in Balto-Slavic times already.

It is remarkable that Illič-Svityč, who reaches the conclusion that the Baltic and Slavic accentual paradigms were identical, does not make an attempt to connect the Slavic NAsg. -o with the Lithuanian ending -a, which now only occurs in adjectives, participles and pronouns but must have been the East Baltic NAsg. ending of neuter o-stem nouns, as is evident from Baltic borrowings in Finnic. While he follows Nieminen (1922) in deriving the East Baltic ending from pronominal *-od, Illič-Svityč assumes that Slavic -o continues stressed *-om, a delevopment advocated by Hirt (1893). In my opinion, it would be natural to look for a common origin. Since I do not believe that *-om ever yields Balto-Slavic *-o, the best option would be to assume that in Balto-Slavic the ending *-om was replaced by *-od in oxytone neuters. In that case one would expect Old Prussian neuter o-stems to correspond to Slavic neuter o-stems and end-stressed neuters in other Indo-European languages. The evidence seems indeed to point in that direction, e.g. (cf. Kortlandt 1983: 183).

Illič-Svityč’s law implies that barytone neuter o-stems were still distinct from masculine o-stems. Though the above-mentioned bifurcation of neuter o-stems seems to be Balto-Slavic, suppletive neuter plurals may have existed both in Baltic and Slavic. The existence of suppletive neuter plurals may also explain why we find so much vacillation between neuter and masculine o-stems belonging to (a) and (b). Illič-Svityč’s law must have preceded the rise of distinctive tone in mobile paradigms because the transfer to the mobile class was based on the identity of the barytone case forms. For the same reason, Illič-Svityč’s law must have preceded Dybo’s law.

The above-mentioned developments may be illustrated with the following examples:

PSl. *tỳlъ (a) ‘back of the head, back’ (e.g. Ru. tyl, Cz. týl) < *túHlom, cf. Skt. tū́la- n. ‘tuft, reed, panicle’. Secondary *tỳlo in Slk. tylo.

PSl. *dvòrъ (b) ‘courtyard, door’ (e.g. Čak. dvõr, Cz. dvůr) < *dʰuórom, cf. Skt. dvā́ra- n.‘door, gate, passage’.

PSl. *zǫ̑bъ (c) ‘tooth’ (e.g. Čak. zȗb, Sln. zǫ́b) < *ǵómbʰos, cf. Skt. jámbha- m. ‘tooth’, Gk. γόμφος ‘pin, nail’, Lith. žam̃bas ‘sharp edge’ 2/4.

PSl. *jàto (a) ‘flock, herd’ (e.g. SCr. jȁto) < *i̯áHto < *i̯eh₂tód << *i̯eh₂tóm, cf. Skt. yātá- n. ‘course, motion’. Secondary *jàtъ in Ru. jat (dial.) ‘shoal of fish’.

PSl. *però (b) ‘feather’ (e.g. Ru. peró, SCr. pèro) < *pèro < *peró < *peród << *perHóm (*tperóm?), cf. Gk. πτερόν ‘feather, wing’.

PSl. *mę̑so (c) ‘meat, flesh’ (e.g. SCr. mȇso, Pl. mięso) < *mēmsó < *mēmsód << *mēmsóm, cf. Skt. māṃsá- n. ‘id.’.

2.2.3.4 The fate of the Balto-Slavic acute and circumflex

As I explained in the section on Winter’s law, the Balto-Slavic opposition between acute and circumflex syllables is in Kortlandt’s framework equivalent with the respective presence or absence of a glottal stop. Before discussing the fate of the glottal stop in Slavic, I would like to present a concise account of the rise of the East Baltic tones (cf. Kortlandt 1977, Derksen 1995).[12] The crucial point is that the broken tone is an archaism.

In East Baltic, the glottal stop became a feature of the neighbouring vowel, yielding the laryngeal pitch that in Baltic linguistics is known as “broken tone”. Tonal oppositions arose when the stress was retracted from prevocalic *i and word-final *‑a. In the Aukštaitian varieties of Lithuanian, retraction onto glottalized syllables yielded a rising tone and loss of the laryngeal feature, e.g. ė̃dis ‘food, fodder’, cf. sti ‘eat’ (of animals). In originally stressed syllables, the glottalic pitch changed into a falling tone, e.g. sti ‘sow’, whereas the non-glottalic pitch merged with the new rising tone, e.g. duktė̃ ‘daughter’. Retraction of the stress onto non-glottalized syllables yielded a middle tone, which later merged with the falling tone, e.g. vìlkė ‘she-wolf’, cf. vil̃kas ‘wolf’. In unstressed syllables, glottalization was eventually lost. In Žemaitian, the broken tone was preserved under the old ictus, e.g. ộmž́iọs (Kretinga) ‘age, century’ = ámžius.

In Latvian, the retractions of the stress yielded a rising tone on both plain and glottalized vowels. The other stressed vowels became falling per oppositionem. Subsequently, glottalization was lost under the falling tone. The result was a stretched tone, which later merged with the rising tone, e.g. sẽt ‘sow’ with the same tone as sniẽdze ‘snow-bunting’, cf. snìegs ‘snow’. The remaining glottalized stressed vowels, which had lost their distinctive tone when the glottalic feature was lost under the falling tone, lost their glottalization as well and became falling, e.g. dȩ̀sts ‘plant’, cf. dẽstît ‘plant’, dêt ‘lay (eggs)’. In originally unstressed syllables, glottalization was preserved as a broken tone, e.g. gal̂va ‘head’, Lpl. gal̂vâs, cf. Lith. galvà, Lpl. galvosè. This scenario is in conflict with the widespread view according to which the broken tone results from retraction of the ictus. The system with a threefold tonal opposition only survives in certain Central Latvian dialect areas.

Apart from the fact that it is not always easy to tell if the tone of a given syllable is metatonical, the way in which the Balto-Slavic acute and circumflex are reflected in East Baltic is fairly straightforward.[13] It is often insufficiently realized that this is not the case in Slavic. A common misapprehension, for instance, is the idea that the “Serbo-Croatian” short falling tone indicates that the syllable was originally acute. In reality, the situation is much more complex. If the form belongs to the neo-Štokavian variant of Serbo-Croatian, the short falling tone indicates that the syllable was already stressed before the neo-Štokavian retraction of the ictus and that it is short. The quantity may be related to the fact the syllable was originally acute, but it may also have been originally short or originate from a comparatively late shortening, for instance the shortening of long falling vowels in forms counting more than two syllables. The fact is that the history of Slavic quantity is immensely complicated. Both the vowels that on qualitative grounds are considered “historically long” and the ones considered “historically short” may be reflected as either long or short. In order to establish the origin of a morpheme in terms of acute and circumflex, one must evaluate the information offered by the individual Slavic languages regarding stress, tone and quantity within the context of the Proto-Slavic accent paradigms, which is by no means simple (cf. Vermeer 1992, Kortlandt 2005).

In the classical view, sequences of vowel plus laryngeal merged with lengthened grade vowels. Subsequently, long vowels acquired an “acute” tone movement, probably a rising tone.[14] Thus, the Balto-Slavic acute is about vowel length. As one might expect, Winter’s law, insofar as the law is accepted, is interpreted as vowel lengthening. The difficulties raised by the classical scenario are numerous (cf. Vermeer 125-126). In Kortlandt’s theory, sequences of vowel plus laryngeal (including the glottal stop that arose from Winter’s law) remain essentially distinct from lengthened grade vowels up to the end of the Proto-Slavic period. With the exception of certain positions where the distinction was lost (see below), the original contrast is reflected by a quantitative difference. I shall now give an overview of the fate of the laryngeals in Slavic (cf. Kortlandt 1975: 21-37, Vermeer 1992: 127-130):

(1)    The laryngeals were lost in pretonic and postpostonic syllables with compensatory lengthening of the adjacent vowel. In mobile paradigms the loss of the laryngeals gave rise to an alternation between long vowels and sequences of vowel plus laryngeal. In root syllables the long vowel was generalized. This is Kortlandt’s explanation of Meillet’s law, according to which mobilia with an acute root underwent metatony (Meillet 1902).

(2)    The laryngeals were lost in the first posttonic syllable without compensatory lengthening. In stressed syllables the glottal stop became a feature of the adjacent vowel. Since the new short vowels had the same timbre as the long vowels and the glottalized vowels and therefore did not merge with the old short vowels, the timbre distinction became phonemically relevant. In pretonic syllables, where the laryngeals had been eliminated at stage (1), quantitative oppositions were rephonemicized as qualitative oppositions. In other words: pretonic long vowels were shortened. At a later stage, Dybo’s law reintroduced phonemic length in pretonic syllables.

(3)    Glottalized vowels lost their glottalic feature and became distinctively short rising. This development must have been posterior to Dybo’s law because the progressive shift only applied to non-acute non-falling syllables.

Summarizing, we could say that originally acute syllables are reflected as short vowels in syllables which prior to Dybo’s law were stressed or immediately followed the stressed syllable. In originally pretonic or postposttonic position acute and non-acute long syllables merged. Non-acute long vowels and original diphthongs are long in AP (b), whereas in AP (c) they often fell victim to the widespread shortening of falling vowels. Length was preserved in monosyllabic and disyllabic word-forms in Serbo-Croatian and in Slovene monosyllables.

2.3 Substratum borrowings

The Indo-Europeans who populated Europe must have come into contact with speakers of non-Indo-European languages, who to a certain extent were assimilated to the invading tribes. Consequently, we expect to find traces of non-Indo-European substrata in the attested Indo-European languages. In Leiden, the study of substratum influences received an impetus from Kuiper 1995, where mainly on formal grounds three substratum layers were distinguished (see also Beekes 1996: passim, Boutkan and Siebinga 2005: xiii-xvii). What these layers have in common, is the frequent occurrence of the vowel *a, which did not belong to the Proto-Indo-European phoneme inventory (see 2.1). One of the layers, labelled A3, is the language of Krahe’s hydronymy and is usually called “Old European”. It is, among other things, characterized by the vocalism *a and the high frequency of continuants and *s. Substratum A2, qualified as “European”, only had aspirated voiced stops in antevocalic position, or rather the antevocalic stops were identified with the traditional mediae aspiratae. Furthermore, there seems to have been variation between labial and velar stops. The vowel *a was frequent and there probably was no distinctive vowel length. Another characteristic feature are vowel alternations of the type *a :*ai and *a : *au. Substratum A1, which is mainly reflected in Germanic, but also left traces in Italo-Celtic and Balto-Slavic, is claimed to have had *a : *i : *u vocalism, prenasalization, initial consonant clusters *Kn- and *Kl-, as well as a remarkable alternation of root-final stops, including geminates.

While A3 did not prove to be a fruitful subject of investigation, one might say that as far as Kuiper’s substratum layers A1 and A2 were concerned, the hunt was on, e.g. Beekes 1996, Schrijver 1997, Boutkan 1996, 1998, 2003, Derksen 1999, 2000. Kuiper’s criteria for identifying substratum borrowings were applied to various Indo-European languages and attempts were made to establish more phenomena indicative of non-Indo-European origin, of which Schrijvers prefix *a- (1997: 307-312) is among the most spectacular.[15] At the same time, Kuiper’s distinction between A2 and A3 was called into question by Beekes (1996: 217), who proposed to group these two together under the name “European”. I am inclined to agree with him that in this respect Kuiper’s classification seems premature. Beekes (ibidem) also suggested the designation “Helladic” for the non-IE substratum language that left so many traces in Greek, but he himself now seems to prefer “Pre-Greek”. Without question, Beekes deserves great credit for his attempt to reconstruct Pre-Greek through a careful analysis of the Greek material (1997, cf. Furnée 1972). A different approach is applied by Schrijver (2007), who tries to link the Pre-Greek substratum to “Minoic” (attested in Linear A) and Hattic. In his view, we are dealing here with the language of the first agriculturalists, who migrated from Asia Minor to Central Europe through Greece and the Balkans and whose language left traces in Anatolian, Greek, Albanian, Germanic, Balto-Slavic and Italo-Celtic.

Undeniably, it makes perfect sense to look for formal characteristics of non-Indo-European elements in the vocabulary of the individual Indo-European languages. Within a couple of years after the publication of Kuiper’s article, however, the limitations and inherent dangers of this new line of research became apparent, in particular with regard to the Pre-Germanic substratum, where there was a tendency to pay insufficient attention to the role played by specifically Germanic developments, such as Kluge’s law (see Kroonen forthc.). Though it seems to me that the Indo-European origins of Balto-Slavic etyma which in view of their Germanic cognates seem to violate Winter’s law are suspiciously often unclear, it cannot be denied that the typically Germanic alternation of root-final consonants may at least partly result from Proto-Germanic sound laws. Nevertheless, the question why Germanic underwent these specific developments (the rise of geminates, for instance) is still legitimate.

So far I have only discussed borrowings from a non-Indo-European substratum, but we must also reckon with the possibility that etyma were borrowed from an unknown Indo-European language. An interesting attempt to demonstrate a layer of borrowings of this type was made by Holzer (1989, cf. Kortlandt 2003). Holzer reconstructs an Indo-European language whose main feature is the fact that it underwent a consonant shift: the traditional tenues became mediae, while the mediae aspiratae became tenues (hence the name “Temematic”). This shift enables Holzer to propose alternative etymologies for words that often belong to classes where substratum borrowings are expected. Holzer’s etymologies will occasionally come up in this dictionary.

3. Structure of the entries

3.1 The reconstructed etymon

The reconstructed etyma represent a late stage of Proto-Slavic, posterior to the loss of glottalization under the stress and Stang’s law. Quantity, tone and stressed are marked accordingly (see 3.5) The most recent development that I have taken into account is the shortening of the falling tone in word-forms of more than two syllables, e.g. *sь̏rdьce. With respect to the metathesis of liquids (and the East Slavic polnoglasie), which shows dialectal differentiation, I had no option but to let the forms reflect the stage where the syllable was still closed, even though the metathesis preceded the above-mentioned developments.

The alphabetical order observed in this dictionary is: a b c č d e ě ę g x i j k l m n o ǫ p r s š ś t u v ъ y ь z ž. Please note the following:

·            The results of the second and third palatalizations of velars (*k, *g, *x) are written *c, *dz and *ś. This may seem inconsistent, but I considered that there was nothing to gain by using *ć and *dź or *ʒ́. The introduction of *ś, on the other hand, could not be avoided, cf. *vьśь ‘all’ vs. *vьsь ‘village’.

·            I have employed the signs *ļ, *ņ and *ŗ to render sequences of resonant + *j. In the alphabetical order these signs are equivalent to *lj, *nj and *rj, respectively.

·            I follow the ESSJa in writing anachronistic *tj, *dj rather than *tć, dʒ́ vel sim.

·            Word-initially, I do not distinguish between *e- and *je-. I simply write *e- because the *j- was automatic before front vowels from a certain stage onwards. I also write *ě- for etymological *ě- and *ja-, which merged after the rise of prothetic *j-. I do distinguish between *u- and *ju-. Initial *jь- < *i- is more complicated. I have argued that we basically had stressed *ji- vs. unstressed *jь- with generalization of the latter in mobile paradigms (Derksen 2003b). Nevertheless, I have decided in favour of a uniform spelling *jь-, which is more conventional.

3.2 Grammatical information

Following the reconstruction, there is an indication of the word class the etymon belongs to. In the case of substantives, the stem class is preceded by an indication of gender, e.g. “m. n” for “masculine n-stem”. The flexion types to which OCS mlьn’i(i) ‘lightning’ and svekry ‘church’ belong are designated with ī and ū, respectively. If there is more than one lemma, the grammatical information runs parallel to the lemmata.

For the sake of readability, the attested Slavic forms are generally speaking only provided with grammatical information if they belong to a different word class than the etymon.[16] This practice extends to non-Slavic forms insofar as it does not cause confusion.

3.3 Accent paradigm

In those cases where I deemed it justified to reconstruct the accent paradigm of a noun or verb, the paradigm is indicated by Stang’s (a), (b) or (c). I am not convinced that there ever was a Proto-Slavic paradigm (d) (Bulatov, Dybo, and Nikolaev 1988, cf. Vermeer 2001). I have occasionally resorted to designations such as (b/c) in those cases where there is strong evidence for two accent paradigms.

Tone and quantity are indicated in the same way as is conventional for literary Serbo-Croatian:

short rising: *màti (a) ‘mother’, *mǫ̀ka (a) ‘torment, torture’, *nòžь (b) ‘knife’

long rising: *bě́lъ (b) ‘white’, pǫ́tь (b) ‘way’

short falling: *sь̏rdьce (c) ‘heart’

long falling: *mę̑so (c) ‘flesh, meat’, *bȏgъ (c) ‘god’

long unstressed: *mǭkà ‘flour’ (b), *osnovā ‘base, foundation’ (a)

Thus, I have not adopted any of the special signs that are sometimes used to indicate the original acute, e.g. *ma̋ti or *maˀti.

3.4 Meaning

Unlike the ESSJa, I have attempted to provide a reconstruction of the Proto-Slavic meaning of an etymon. In principle, attested forms meaning the same as the reconstructed etymon have not been glossed, though occasionally the meaning has been retained for the sake of clarity. This holds good for both the Slavic and the non-Slavic forms.

3.5 ESSJa

The line beginning with the reconstructed etymon is concluded by a reference to the ESSJa unless, of course, the latter dictionary does not have a corresponding lemma. The most recent volume at my disposal was volume 32 (-*orzbotati). I have not included a reference to the Słownik prasłowiański because at this point the number of published volumes is too limited.

3.6 Church Slavic

The field “Church Slavic” contains forms that occur in texts belonging to the Old Church Slavic canon as well as forms that occur in Church Slavic texts whose language was influenced by the local vernacular. The latter varieties of Church Slavic are called recensions. With the aid of the Slovník jazyka staroslověnského, which, by the way, includes a number of texts that fall outside the canon, and the Staroslavjanskij slovar’, I have tried to keep Old Church Slavic and Church Slavic forms apart. Here I should also mention Birnbaum and Schaeken 1997, where attention is paid to the lexicon of the manuscripts that were discovered at the Monastery of St. Catherine at Mount Sinai in 1975.

It is not unusual to present Old Church Slavic forms in a normalized shape and this is the practice I have adopted here. If an etymon is only rarely attested, however, I usually present the form as it occurs in the manuscript(s). If an etymon occurs in a fairly limited number of manuscripts, the latter have been mentioned. I have used the following abbreviations:

Ass.: Codex Assemanianus	Ps. Dim.: Psalter of Dimitri
Boj.: Bojana Evangeliary	Ps. Sin.: Psalterium Sinaiticum[17]
Cloz.: Glagolita Clozianus	Ril.: Rila Folios
En.: Enina Apostol	Sav.: Sava’s Book
Euch.: Euchologium Sinaiticum	SPbOkt.: St. Petersburg Oktoich
Hil.: Hilandar Folios	Supr.: Codex Suprasliensis
Hilf.: Macedonian Folio[18]	Zogr.: Codex Zographensis
Mar.: Codex Marianus	Zogr.²: Codex Zographensis palimpsest

3.7 East Slavic

The East Slavic field comprises three living languages: Russian, Ukrainian, and Belorussian. Furthermore, an important place is occupied by Old Russian, which can sometimes hardly be distinguished from Russian Church Slavic. Here I generally follow the ESSJa. For Old Russian accentual data I have relied on Zaliznjak 1985. The accentuation of a noun or verb in Old Russian is chiefly mentioned in those cases where it deviates from the modern Russian standard language.

My main sources for Russian dialect material are the classic dictionary by Dal’ and the Slovar’ russkix narodnyx govorov (SRNG). To indicate the region where a form is attested I have used a number of abbreviations: Psk. (Pskov), Olon. (Olonec), Arx. (Arxangel’sk), Novg. (Novgorod), Smol. (Smolensk), Rjaz. (Rjazan’).

3.8 West Slavic

Czech forms may belong to the literary language, for which the SSJČ is an important source, or originate from dialects. Dialect material generally stems from the ESSJa or from Machek’s etymological dictionary. A special category is formed by the 18^th century dictionaries of Jungmann and Kott, which include archaic and dialectal forms, but also borrowings from other Slavic languages and neologisms. An analysis of the material from these dictionaries is beyond my competence, so I have confined myself to indicating Jungmann (Jg.) or Kott as the source. Slovak forms, insofar as they are not dialectal, are generally quoted according to the Slovník slovenského jazyka (SSJ). Since the juxtaposition of Czech and Slovak forms is interesting from an accentological point of view (cf. Verweij 1993), the collection of Slovak material is relatively comprehensive.

Upper Sorbian is one of the languages that are not heavily represented in this dictionary. Nevertheless, the language has retained a number of features which may provide additional information about Proto-Slavic prosody (cf. Dybo 1963, 1968b, Derksen forthc. c). In this respect Lower Sorbian has less to offer. Schuster-Šewc’s Sorbian etymological dictionary (HEW) deals with both languages and may also function as a source of older attestations and dialect forms.

A considerable portion of the Polish material in the ESSJa, which subsequently found its way to the present dictionary, is quoted from the voluminous dictionaries by Karłowicz et al. (1900-1927) and Doroszewski (1958-1969). For Old Polish the main source is, of course, the yet unfinished Słownik staropolski. The etymological dictionaries by Sławski (unfinished) and Bańkowski (in progress) provide a lot of information about the earliest attestations of an etymon. In this context the Słownik języka polskiego XVII i 1. połowy XVIII wieku (Karplukówna and Ambrożewicz 1999‑) is also worth mentioning.

Within West Slavic, Slovincian – now extinct – and the North Kashubian dialects are unique in having preserved accentual mobility, albeit with certain restrictions. Furthermore, old quantitative distinctions have been transformed into qualitative distinctions, like in Polish. In Slovincian and Kashubian, however, this phenomenon (called pochylenie in Polish) applies to a greater number of vowels. Our most important source for Pomeranian, as Slovincian and Kashubian are sometimes called,  are the works of Friedrich Lorentz (e.g. 1903, 1908-1912, 1958-1983).

The westernmost attested Lechitic language, Polabian, only plays a marginal role in this dictionary. Forms will be quoted according to Polański and Sehnert 1967.

3.9 South Slavic

The name “Serbo-Croatian” will occasionally be used as a generic designation for all varieties of the language spoken in Serbia, Croatia, Bosnia-Hercegovina and Montenegro. The abbreviation “SCr.”, however, refers in principle to neo-Štokavian, i.e. to those Štokavian dialects that underwent the neo-Štokavian retraction of the stress. A prominent example is the language that was codified by Vuk Karadžič and Đuro Daničić in the 19^th century and subsequently became the basis of normative grammars and dictionaries, for instance the Rječnik hrvatskoga ili srpskoga jezika (RJA). The Serbo-Croatian (neo-Štokavian) forms presented in this dictionary usually conform to aforementioned normative tradition. In some cases I have quoted directly from Vuk Karadžič’s dictionary (abbreviated as “Vuk”).[19]

The Čakavian dialects of Serbo-Croatian are mainly represented by Jurišić’s description of the Vrgada (Vrg.) dialect (1966-1973), Kalsbeek’s description of the dialect of Orbanići (Orb.) near Žminj (1998) and Belić’s description of the dialect of Novi (1909), which is the best-known description of a Čakavian dialect. In addition, I have occasionally added data from Hvar (Hraste 1937), Cres (Tentor 1909, 1950), and Orlec[20] (Houtzagers 1985). The Kajkavian dialects are respresented by Jedvaj’s description of the Bednja dialect.

The Slovene material originates almost exclusively from Pleteršnik’s dictionary (1894-1895), which is a compilation of data from a great variety of sources presented in a historical spelling, i.e. in a spelling that reflects dialectal differences that are absent from many dialects, including the ones underlying the modern standard language. The reader will therefore be confronted with the spectacular dialectal differentiation that is characteristic of Slovene (cf. Greenberg 2000).

Bulgarian and especially Macedonian provide little information on the prosody of Proto-Slavic.[21] Nevertheless, it is clear that this corner of the Slavic territory must not be neglected. The material presented here is chiefly modern Standard Bulgarian.

3.10 Balto-Slavic

Since I adhere to the view that there was a Balto-Slavic proto-language, I found it atttractive to include a field “Proto-Balto-Slavic” in order to obtain a modernized version of Trautmann 1923a (cf. Derksen forthc. d). As I explained above, I follow Kortlandt in not reconstructing any Balto-Slavic tones. The stage represented by my reconstructions is posterior to Winter’s law. The reflex of the laryngeals and the glottal element of the (pre)glottalized stops is indicated by ʔ, the IPA symbol for a glottal stop. The phonological system is as follows (cf. Kortlandt 1994):

p	b		m
t	d	s	n	l	r
ś	ź
k	g
ʔ				j	w
		i	ī			u	ū
		e	ē			o	ō
				a	ā

With respect to morphology, it is important to note that the barytone neuters have a Nsg. in *-um < *-om, while the originally oxytone neuters have *-o < *-od << *-om.

Notwithstanding my comparison with Trautmann’s Baltisch-slavisches Wörter­buch, it is my opinion that these reconstructions in themselves do not have independent evidential value for the Balto-Slavic hypothesis, as striking formal similarities between Proto-Baltic[22] and Proto-Slavic etyma usually relate to those common developments on the basis of which the Balto-Slavic linguistic unity was postulated. We must reckon with the possibility that seemingly Balto-Slavic etyma are actually old borrowings from Slavic into Baltic or vice versa. In particular, Baltic substratum words in Slavic are a factor that is to be taken into account (cf. Nepokupnyj 1976).

The assumption that there was a Proto-Baltic-Slavic stage raises the question how to act if an inherited Proto-Slavic etymon does not have a Proto-Baltic counterpart? Does it make sense to reconstruct a Balto-Slavic form anyway, thereby assuming that the etymon was lost in the other branch? In my opinion, it does not. The main point of the field Balto-Slavic is to get an impression of the number of lexical items shared by Baltic and Slavic. A Proto-Balto-Slavic reconstruction of an etymon that is exclusively found in Baltic or Slavic would just be one of many intermediate stages.

Another, quite common problem is the determination of the orginal stem class of a noun in those cases where Baltic and Slavic diverge. Unless there are clear indications which stem class is more archaic, I considerered it best to allow the variation to be reflected in the Balto-Slavic reconstructions.

3.11 Baltic

It stands to reason that Baltic is fairly well represented in this dictionary. On the other hand, it would hardly make sense if the Baltic material were so extensive as it will be in my Baltic etymological dictionary. For this reason I cut down on the Lithuanian and Latvian data by leaving out some of the morphological and accentual variants.

Where the modern Standard Lithuanian form serves our purpose, I have quoted from the fourth edition of the DLKŽ. The most important source for Lithuanian, however, is the LKŽ, which has incorporated data from older periods and dialect data. The most important Old Lithuanian text is Daukša’s Postilla (DP). Latvian data will be quoted from the dictionary by Mühlenbach and Endzelīns (ME) and the supplement by Endzelīns and Hausenberg (EH). The orthography used in these works deviates from modern Standard Latvian orthography but is still customary in the scholarly literature. Forms followed by ² originate from a dialect with two tones (instead of three), where the tone of that particular form is ambiguous from a historical point of view (cf. Derksen 1996: 11-14). We are dealing either with a conflation of the broken and the falling tone (West Latvian) or a conflation of the sustained and the falling tone (East Latvian). These tones are indicated by ^² and `², respectively, but it should be noted that there exists considerable variation regarding the realization of the conflated tones.

The Old Prussian evidence stems from the Enchiridion (or Third Catechism) unless indicated otherwise: I write “I” for the First Catechism, “II” for the Second Catechism, “EV” for the Elbing Vocabulary, and “Gr.” for Simon Grunau’s vocabulary. The forms are quoted from Trautmann 1910 and Mažiulis PKEŽ.

3.12 Proto-Indo-European

The field “PIE” basically contains forms that may have belonged to the Proto-Indo-European vocabulary. I do not wish to condemn the practice of reconstructing quasi-Indo-European forms, as the latter may contain useful information, even in the case of etyma that may have been borrowed from a non-Indo-European language, but I personally prefer to let this information be part of the discussion of the etymology.

3.13 Cognates

It is not my intention to present an exhaustive list of cognates from other branches of Indo-European. One may expect to find quite comprehensive lists of cognates in the dictionary that is the ultimate goal of the Indo-European Etymological Dictionary project. I have tried to mention forms that are in all respects close to the Proto-Slavic etymon, but occasionally I had to settle for forms that merely contain the same root.

3.14 Discussion of the etymology

As a rule, the etymology of a given root is discussed under a single lemma. If the eymology is perfectly clear, there may be no discussion at all. Instead, only a Proto-Indo-European reconstruction is given.

3.15 Notes

Any lemma may contain a number of notes pertaining to a variety of subjects. The notes do not directly concern the etymology of the lemma, which is discussed elsewhere (see 3.13).

3.16 Cross-references

The last field mentions all cognate lemmata except the ones that are referred to in the discussion of the etymology. If the discussion of the eymology contains a reference to a cognate lemma, the field is omitted.