Etymological Dictionary of Old Frisian Online

Etymological Dictionary of Old Frisian Online

Introduction

1.1 Origin and design of the Dictionary

The department of IE Comparative Linguistics at Leiden University has been working on a new Indo-European Etymological dictionary (IEED)^[1] for more than a decade. As a first step, etymological research was conducted into the areas of less known languages for which the expertise was at hand. Jörundur Hilmarsson undertook the compilation of a Tocharian etymological dictionary. Bardhyl Demiraj has recorded the complete IE heritage of Albanian during his stay at the Leiden University 1997.

In 1994, Dirk Boutkan started to work on an Old Frisian etymological database for the IEED project. The work on Old Frisian had as a positive side-effect that the database could also function as a rudimentary etymological dictionary. In 1998, the database was published on the internet.^[2]

Before the project could get underway in 1994, two preliminary questions had to be answered.

1.     How could the number of lemmata be restricted (there was only one editor available for a period of three years).

2.     Which information was going to be included and in which order.

ad 1) Basing the material on existing collections would not be expedient. The dictionaries by von Richthofen (1840) or Holthausen and Hofmann (1985) and even Köbler (1983) use arbitrary parts of the OFris. corpus and are too elaborate for the purpose of a reconstruction database. A discussion of the OFris. evidence that Pokorny (1959) lists is even less useful. Pokorny’s choice of OFris. cognates seems completely arbitrary.

The best option seemed to be the restriction to the lexical material of one reasonably sized manuscript. The Riustring 1 manuscript was selected, because linguistically it is the most archaic Old Frisian manuscript. The mediaeval inhabitants of the land between the rivers Wezer and Jade in Northern Germany called Riustringaland have given us two legal texts R₁ (approximately 1300 AD) and R₂ (1327 AD).^[3] The dialect of R₁ belongs to Old East Frisian, which has an older written tradition than Old West Frisian. In the following centuries Old East Frisian was replaced by Low German. Modern Frisian has descended from Old West Frisian. The codex has a diverse range of topics; apart from the more classical legal texts such as the ‘seventeen privileges and twenty-four statutes’, it also contains some religious prose (‘the fifteen signs before Doomsday’ and ‘The Riustring Sendriucht [synodical law]’) and a number of compensation tariffs (formerly referred to as ‘fine registers’). The codex is extensive enough to supply a core lexicon of Old Frisian. Finally, there is an excellent edition of the text available with an elaborate glossary, making the work optimally accessible (Buma 1961).

ad 2) A structure of tightly formalized lemmata was chosen (see section 1.3). These entries are relevant for the IEED project and also preserve the possibility of using the database for an etymological dictionary. Besides the obvious advantages of transparency of the material, the technical realisation of the overall Indo-European database would benefit from it as well. The design of this dictionary has some consequences for its etymological nature. There are two sides to etymology: semantic and phonological. The choice of reworking the database of Old Frisian etymologies and basing it solely on the Riustring 1 manuscript limits the semantic use of the dictionary. Since the discussion of a form does not look beyond the Riustring 1 dialect for semantic differences, the lemmata are mainly concerned with formal (phonological) reconstructions. This makes it less useful for answering any questions on semantic change within (Old) Frisian. Thus, for example, the dictionary does not discuss the semantic development of a legal term (or any word) from PIE or PGmc. to Old Frisian (or modern Frisian). It must be noted, however, that the rudimentary building blocks for a semantic study are already present in the respective lemmata.

The explanatory focus of the discussion of lemmata thus is linguistic rather than literary. The (lemma-specific) semantic information comes straight from the wordlist that accompanies the R₁ text-edition (Buma 1961:157-270) and, although a great many Old Frisian forms from other manuscripts are cited, their semantic uses are not individually treated.

The lemmata contain sparse grammatical information. The concise grammar on Riustring 1 (Boutkan 1996) is designed to function as a grammatical companion. Originally, the work on the grammar was meant to become the grammatical part of the introduction of this dictionary, but it soon became apparent that its 203 pages were far too much to be incorporated into an introduction.

Hence, it was published separately.

1.2 Language change: from Proto-Indo-European to Old Frisian

Old Frisian is a member of the Germanic language family and Germanic is in turn a subgroup of a larger Indo-European family. The relationship is based on the virtue of the fact that Old Frisian (or any other Old Germanic language) is the product of unbroken transmission from a remote, unattested ancestor called Proto-Germanic, which in turn is a descendant of an even more remote ancestor called Proto-Indo-European. The Indo-Europeans probably lived to the north of the Black Sea in the fourth millenium BC^[4], by the time of the beginning of the Frisian written tradition around 1250, the speaksers of IE spread in several migration waves as far east as India, as far west as Greenland, as far north as northern Russia and as far south as the northern part of Africa.

The major IE subgroups are listed in the table of contents of Appendix “indices”.^[5] A very good introduction to language change, comparative linguistics and Proto-Indo-European and a list of basic terminology can be found in Beekes (1995).

1.2.1 Substratum influence of non-IE languages on IE

The languages of the non-IE tribes native to mainland Europe before the arrival of the Indo-Europeans are called pre-IE, whereas Proto-Indo-European is the earliest language stage of Indo-European language. The technologically and military superior Indo-Europeans became the dominant (or superstrate) language and culture in the areas they migrated to. Although there must have been some kind of mixing over a period of time between these two cultures, there is little tangible evidence — both historic and linguistic — of these preIE languages in Europe. Through archaeological evidence we can more or less separate the Pre-IE and IE eras. Marija Gimbutas was one of the first anthropological archaeologists to combine comparative linguistics and archaeological findings to arrive at a new hypothesis about history, culture and religion of pre-historic peoples. She extensively published on the Kurgan culture and the subsequent Indo-Europeanization of Europe. These works form an important clue to the pre-IE situation in Europe. The languages of this — as Gimbutas (1997) calls it — Old Europe are the likeliest donor languages — or substrate languages — to IE.

Early loanwords are a matter of some controversy in Indo-European linguistics, especially when substratum origin is suspected. In other disciplines, like Uralic linguistics, the influence of loanwords is more accepted because the source of the loans is often known. Because the donor language is usually unknown there is some reluctance to accept loanwords in IE. However, it is necessary to identify non-IE loanwords for a proper understanding of IE and because the “study of loanwords can be a powerful tool for determining prehistoric cultural contacts and migrations” (Lubotsky 2001).

In recent years, a methodology of dealing with borrowings from an unknown source has been developed by Kuiper (1995), and applied by Beekes (1996), Schrijver (1997) and Boutkan (1998a). As these scholars have pointed out, an etymon is likely to be a non-IE loanword if it is characterised by some of the following features: 1) limited geographical distribution; 2) phonological or morphological irregularity; 3) unusual phonology; 4) unusual word-formation; 5) specific semantics, i.e. a word belongs to a semantic category which is particularly liable to borrowing.

Polomé (1986: 661-3) identified several non-IE words in Germanic and distilled the following semantic categories as likely candidates for non-IE loanwords: animals; animal products; plant names (and thus probably also tree-names); simple (domestic) implements; features of the environment; human feelings and perceptions; and human activities.

In the following paragraphs, a survey will be presented of identified substratum languages and features, but first two remarks.

The term “substratum” is used for any donor language, without implying sociological differences in its status, so that “substratum” may refer to an adstratum or even superstratum. It is possible that Proto-Indo-European borrowed words from more than one language and thus from more than one substratum. Another point concerns dialect differentiation. In general, language unity exists as long as the language is capable of carrying out common innovations, but this does not preclude profound differences among dialects. In the case of the European branches of IE, there probably was a relative language unity between Celtic, Italic, Germanic and Balto-Slavic; after these tribes had migrated to Europe, they settled close to each other north of the Alps.

For the moment, it seems safe to say that at least four substratum layers to European IE can be identified:

1) The first substratum layer is the so-called “Old European hydronymy” or the A3 layer (Kuiper 1995). Krahe (1954, 1964) was the first to draw attention to striking similarities between a list of relatively homogeneous hydronyms, and concluded that they derived from a common ancestor. Krahe thought that it was an IE language, which he called Illyrian. However, weighty arguments have been produced to show that this cannot be the case (Kuhn 1967, Vennemann 1994 and Kuiper 1995), such as the nearly complete lack of lexical items in IE languages corresponding to elements in the Old European hydronymy. Furthermore, the language of the Old European hydronymy has remarkable phonotactics: a) the vowel a is by far the most frequent vowel, whereas it did not exist, or was at best very rare in PIE (Lubotsky 1989); b) there is an inordinately high frequency of resonants and s as opposed to stops, which, again, is unlike Indo-European phonemic system.^[6] Other characteristics are: c) an unusual word structure of the type (C)V-CV-CV-, at the end of a word also -VC (Arar, Varar) or VV (-ia);^[7] d) vowel elision (which may have operated in the source language or in any of the later languages); e) the use of interchangeable suffixes on a large scale (Krahe 1954: 53-58), which are now identified as non-IE (Kuhn 1967). On these grounds, it can be stated that the “Old European hydronymy” is non-IE.

Judging from the distribution of Old European hydronyms, the language reflected in them was spoken over large parts of western and northern Europe, including the British Isles, Scandinavia, Germany, Poland and the Baltic region. According to Vennemann (1994), the language is related to Basque, but this has not been generally accepted. Although some (Vennemann 1994 and Beekes 1998) have claimed that this language donated some loanwords to IE languages, the suggestion that this substratum played an important part in donating lexical material to IE is generally rejected.

2) The second substratum layer has many labels: Atlantic (Vennemann 1995); A1 (Kuiper 1995); North Balkan Substrate; European (Beekes 1996); and the language of bird-names (Schrijver 1997, 2001). The overlap between certain elements of these labels is not clear, but there is no reason to treat them as completely separate entities.

The bird-names provide evidence on some significant features of the structure of this substratum language (Schrijver 1997). The most important evidence is that it had a prefix a-, which was probably stressed and accompanied by syncope of vowels in the rest of the word. The language had fricatives such as X, θ, and a diphthong alien to Germanic and Celtic (something like [aə]), which was rendered as a in British Celtic and as ai in Germanic. The language of bird-names is attested through Germanic, Celtic, Italic, which would make up a sizeable territory in Middle Europe.

The A1 substratum language (Kuiper 1995: 65-68) was first identified by Kury lowicz (1956). There is one characteristic that most of the words Schrijver listed have in common, namely the absence of voiced stops: in antevocalic position we exclusively find aspirated voiced stops, cf. Kuiper (1995: 67) for an explanation. The ‘European’ substratum language (Beekes 1996) was characterised by: a) a frequent vowel *a; b) a stop system different from IE, resulting in a variation of stops (which is unknown to IE), mainly between labial and velar stops (Kuiper 1995: 81-84); c) one frequent category of stops which probably was perceived as aspirated; d) initial PIE *b- was not uncommon; e) a vowel-system or ablaut system different from PIE, resulting in vowel ‘variations’ unknown to IE, e.g. a/au, a/ai, u/ū,^[8] , a/ā, æ/a, i/ī;^[9] so vowel length seems to be variable, which may have meant that ‘European’ had no opposition in vowel length; f) an initial *k-, if this is a plain velar and not a palatal.

The suffixes -ut- and -it- probably belong here too. There is no evidence for a PIE suffix -ut- or -it-, except in the perfect participle. Germanic has instances of suffixes of the form -a/i/uC-. These may partly have arisen from PIE forms in -a, -i or -u followed by the zero grade of suffix -eC/C-, but the large-scale existence of these alternations must be partly due to the existence of non-IE suffixes with this alternation. The suffix -ont did not occur in PIE, and cannot be explained as an ablaut variant besides IE -ent (Beekes 1996: 231).

3) The third substratum layer is heavily present in Germanic (see e.g. Boutkan 1998a), but there is also some material in Celtic and Balto-Slavic. As a consequence, the territory of this language may be found somewhere in Northern Europe. Kuiper (1995), who may be credited with the identification of this substratum, called it A2. In a recent article, Schrijver (2001) investigated the nature and origin of non-Indo-European words in northern IE languages (Ger-manic, Celtic) with cognates in Lappish and/or Finnish (Schrijver 2001). It appears that these IE and Uralic languages individually borrowed words from the A2 language, which he calls “the language of geminates”.^[10]

A highly characteristic feature of words deriving from this language is the variation of the final root consonant, which may be single or double, voiced or voiceless, and pre-nasalised, cf. Kuiper (1995: 68-72). The reason for this phenomenon is unknown, but it seems likely that the alternation found in Germanic reflects a similar alternation in the substratum language. The only regularity of this alternation is that after a long vowel apparently no geminate consonants occur. The A2 substratum cannot be Uralic, as another of its characteristics is the frequent occurrence of word initial *kn- and *kl-, whereas the Uralic languages do not allow consonant clusters at the beginning of the word.

The vowels are primarily a, i and u, apparently without preponderance of a. It can also be demonstrated that both disyllabic *CVCVC formations and vocalic variation (i ∼ a ∼ u) in Germanic reflect features of this substratum language (Boutkan 1998a). A promising explanation for this variation is the assumption that the vowel system of the substratum language and that of Germanic did not entirely match, cf. Boutkan (1998a: 128-9). At any rate, when the central vowels were integrated into their own vocalic system, Italic, Celtic, Baltic and Slavic apparently used these vowels differently from Germanic. Continental Germanic also shows *a ∼ *e variation, which cannot be explained by an ablaut pattern (*o >) Gmc. a ∼ (*e >) Gmc. e.

4) The fourth substratum language is present mainly in the Mediterranean. This language, which furnished many loanwords in Greek, is called ‘Pre-Greek’ by Furnée (1972) and ‘Helladic’ by Beekes (1996). Helladic is characterised by: a) prothetic vowels, mainly a- (Furnée 1972: 368-378); b) vowel alternations o/ou, a/au, i/ī. Schrijver (1991) suspects that many Latin words are loans from this substratum language.

The relation between these four substrata is not entirely clear. It may be the case that they belong to the same substratum language family, because in several instances, apparent non-IE words could be assigned to more than one substratum language. A common denominator among these substrata is the presence of the vowel *a, which can be used as a heuristic principle.

1.3 The structure of the lemmata

The lemmata are diachronically structured: from contemporary Old Frisian (ca. 1300 AD) to the Proto-Indo-European reconstruction (ca. 3500 BC), spanning a period of nearly fivethousand years. With three stepping stones — the PFris., PGmc. and the PIE reconstruction — the Old Frisian etymon is taken back to its oldest reconstructable root. The information contained within a lemma has been divided up into as many meaningful fields as possible, in order to facilitate the database structure. The fields with reconstructions and cognate forms are especially important for the compatibility with the other IEED databases.

The Old Frisian etyma can be separated into four etymologically distinct types of lemmata: i. words with unclear etymology; ii. loanwords; iii. inherited Indo-European words; iv. substratum words. Below, a schematized form of the lemma structure is given with horizontal rules that define up to which point the general lemma structure is relevant for the different types of lemmata. The optional lines are marked with an asterisk.

1) lemma entry grammatical remarks; ‘meaning of the word’; etymological label

i. Lemma with unclear etymology

2)* (miscellaneous preliminary remarks)

3) Old Frisian cognates

4) Discussion of Old Frisian forms

5) PFRIS: Proto-Frisian reconstruction

ii. Loanword

6a) Germanic cognates

6b)* related Germanic reconstruction: Germanic cognates

7) •PGMC: Proto-Germanic reconstruction

8)* Discussion of Germanic forms

iii. PIE inherited

9a) •PIE: PIE reconstruction POKORNY: Page of reconstruction in Pokorny

iv. Substratum etymology

9b) • No PIE reconstruction

10) PIE cognate forms and discussion of the possible PIE reconstruction

11)* Lit: References to relevant literature

12)* see also: References to other relevant lemmata

13)* notes: Numbered notes

The first line contains the lemma entry, the meaning of the word, its grammatical category and its gender. These have been taken over from the glossary by Buma (1961: 158-270), and are rarely problematic. There are three etymological labels: LW (loanword), Gmc. (only cognates in [a part of] Germanic), and PIE (a word inherited from Proto-Indo-European with cognates outside Germanic). In most cases, the label Gmc. denotes substratum words and PIE words that are inherited from the common Proto-Indo-European language stage.^[11] When there is doubt about the origin a question mark is added. Line (2) is optional and may contain information about the distribution of the word in R₁ or other relevant information, such as syntagmata in which the word can be found. This is the case with e.g. bli, kona and ābēl (cf. section “i. Lemmata with unclear etymologies”). Line (3) lists the Old Frisian cognates^[12] and line (4) contains a synchronic interpretation of these cognates that leads up to the Proto-Frisian reconstruction.

The following line is marked with a bullet (). It signals that a line contains a reconstruction. The reconstructed forms themselves are always marked with an asterisk (*). For the purpose of line (5), the Proto-Frisian reconstruction, the archaic nature of R₁ is an advantage. In most cases, the R₁ form contains enough information for the reconstruction. Sometimes, the reconstruction of Proto-Frisian variants is necessary, when, for example, two variants are in fact competing reconstructed formations or when, as in the case of kuic (cf. s.v.), there seems to have been an umlaut variant quek in Proto-Frisian besides kuic. After the uniform common structure (lines 1-5), the four types of lemmata start to diverge.

i. Lemmata with unclear etymologies

The first category, the lemmata with unclear etymology, is slightly different from the other categories, because it basically consists of everything that does not belong in any of the other categories. Here, words are grouped that are already problematic apart from the discussion about their etymology. For example, Old Frisian has a number of words, of which the meaning and thus the etymology are unclear. An example is the well-known Riustringer currency term kona. Sometimes the discussion breaks down even earlier, at the level of philological interpretation. The lemma bli is an example of this, since the interpretation of the word is subject to dispute. Every lemma of this category demands an exhaustive discussion of the precise problems and possible solutions. If the problem can be overcome, the word can be placed in one of the other categories. Otherwise the etymological label in line 1 becomes a question mark.

ii. Loanwords

Loanwords are discussed up until line 6a, where the Gmc. cognates are discussed. There are some inventories of Old Frisian loanwords. The oldest survey is by Holthausen (1921: 34-39). Wollmann (1990) is of a more recent date. For every loanword, it is mentioned whether the word is listed in either of these inventories, and if so, where the word is dealt with by Wollmann. Following the Proto-Frisian reconstruction (line 5), the origin of the word is discussed. To illustrate the spread of a loanword, examples are given from other Germanic languages that have also borrowed this word. An elaborate example is abbit.

iii. Inherited words

The category of inherited words contains words that have an Indo-European etymology. Following lines 1-5, the Germanic cognates are listed (line 6a). Sometimes when cognates with different root suffixation are relevant for the interpretation of the PGmc. reconstruction (line 7) these are listed with a reconstruction in line 6b. This line is usually introduced by ‘cf. also’. When the PGmc. reconstruction (line 7) is problematic it can be followed by line 8, which contains a discussion of these problems.

The Old Germanic languages are relatively well accessible through the usual dictionaries and secondary literature. However, some problems may have remained untreated (and could even been incorporated into this dictionary) due to the limited amount of time and the enormous quantity of literature. See, for example, the remark on ON ábóti in the ‘abbit’ lemma. Here, detailed research has led to a revision of the ON dictionary form listed by de Vries (1977). This kind of detailed research was not always possible within a reasonable time frame, nor was its necessity always eminent. Thus, most of the time, the standard reference works had to be relied on. The PIE section of the lemma starts with line 9a, which gives the PIE reconstruction and the page in Pokorny (1959) on which this root can be found. Line (10) contains the IE cognates and a discussion of the IE etymology. This section incorporates many new insights of modern Indo-European comparative linguistics. The Laryngeal Theory plays an important role in the reconstructions, because it has shed new light on the interpretations of IE vocalism and ablaut. Also, whenever necessary, the Balto-Slavic accent has been incorporated, in the application of Winter’s Law^[13] or in order to assign an acute accent to an original laryngeal in the root. Often, these new insights lead to the rejection of an Indo-European etymology, cf. s.v. flarde.

There have been some significant changes to the traditional ‘Pokorny’ notation, as for example, the reconstruction of the laryngeal. On the other hand, it is deemed unnecessary to explicitly use the glottalized consonants as long as one realizes that the spellings <b>, <d>, <g>, <g^w> correspond to what in fact were glottalized consonants.

There are two complications, which might have led to the addition of a question-mark to the etymological label ‘PIE’ in line 1:

1.     When it seems likely that the etymon is an inherited word, but there are still different possibilities for an IE etymology. Lemma a- 2 is an example of such a complication.

2.     When there are some doubts about the IE origin. For example, when some formal details are different from what is to be expected, but a rejection of the etymology seems unnecessary. See, for example, mār.

iv. Substratum words

The lemma structure of substratum words differs, in comparison with the inherited lemmata, in the last part of the lemma, the PIE section (line 10) and often in the inclusion of line 6b with related Gmc. reconstructed roots. Instead of a full discussion of the Indo-European etymology or possible solutions, the Proto-Germanic etymology (line 7) is followed by line 9b, which generally states that there is no certain IE etymology. The next line, (10), contains a discussion which tries to disprove the alleged IE etymology and/or deals with the elements that argue for a substratum origin. These elements can refer to:

1.     Formal characteristics. This is the case when a word does not show the usual IE root structure, contains a radical *a, or has doublets that cannot be ascribed to IE ablaut, etc.

2.     Limited geographic spread. When a word is, for example, only attested in Germanic and Baltic.

3.     Semantics. Germanic has certain semantic fields which for cultural historical reasons do not have many IE counterparts. These include for example terminology of social order (landownership, marriage, parts of the laws, etc.), but also anatomy, as can be seen in lippa.

Although the substratum origin of the body part ‘lip’ may seem doubtful, the Old Frisian corpus does contain the words ‘beard’ (OFris. berd), ‘mustache’ (kenep, see below), ‘head’ (hāued, Lat. caput, etc.), ‘gristle’ (gristel), which are all substratum words. Often taboo words lead to borrowing. An example of a word that meets all three criteria is flarde.

It has a suspicious root structure, a limited geographical spread, and belongs to the field of anatomy. It is not always possible to assign a word to a specific substratum layer. This is the case when an obvious substratum word has no distinctive features or no feature that excludes one of the layers.

The influence of substratum forms is far more pervasive than was thought a couple of decades ago. Pokorny’s dictionary, for example, contains an unexpected wealth of substratum material — even etyma that are recorded only in a single PIE language are listed under a reconstructed root. Most of these etyma belong, suspiciously enough, to the Germanic branch. Sometimes even Pokorny had doubts about the IE origin of an etymon, as can be seen in the lemma kletsie.

References

Finally, all lemma types may contain references to the secondary literature (line 11) or references to other relevant lemmata (line 12). The literary references to the secondary literature mainly refer to literature on specific fields of research. For example, when dealing with strong verbs, there will be a reference to the place where this particular verb is dealt with in Seebold (1970), cf. brēda. These can be followed (line 12) by possible internal references to relevant lemmata. The semicolon in the list of references indicates different ranges of related lemmata. The final line, (13), is optional and may contain numbered notes to the lemma under discussion.

1.4 From manuscript to dictionary

The sole purpose of this section is to give insight into the choices made while finalising the dictionary. The only tangible evidence on how Boutkan designed the work on the dictionary can be found in Boutkan (1998c), which is an article based on a lecture. Most of the explanations (given in section 1.3) on how the dictionary works has been distilled from the manuscript itself. Revising the manuscript took from May 2002 to July 2003. The work was complicated by the fact that Boutkan had worked on several versions of the dictionary^[14] simultaneously, which meant that all differences had to be carefully judged so as to determine which one would end up in the final manuscript.

In order to create one definitive text from the four different versions, a PDF (Portable Document Format) version was made from the WordPerfect 5.1 files in Boutkan’s legacy. After the index markers were added automatically with a Python script, the PDF and the two paper versions were manually collated against the database version. During this phase of edition, all citations were replaced by ConTEXt bibliographical markers, which made it possible to create one coherent citing regime throughout. The bibliography itself is now generated from a bibliographical database. The bibliographical markers also structured the verification of citations and their representation in the bibliographical database.

After deliberation with the editor of the Series, Alexander Lubotsky, it was decided to use the Starling database version as a basis for the final edition. The versions up until the autumn of 2003 were typeset with L^AT_EX2e. During the final stage of typesetting ConT_EXt was used, because it had better multiple index and unicode support, and was overall more user-friendly.

During the collation phase, it appeared that there was some sort of caesura between the letters M and N. Halfway the dictionary, Boutkan had apparently adopted a more elaborate lemma design for the PGmc. database and probably planned to rework the previous letters later. In general, these structural differences are still present in the final manuscript, because they are enrichments of the original design. Lemmata in the first section (i.e. A-M) on which Boutkan’s view had changed over the years have been updated accordingly. Most of these are substratum words on which he had written in various articles. Generally, these can be recognized by the Gmc. level indicator in the A-M lemmata.

Another problem was that the macros used in the WordPerfect 5.1 files were not always rendered properly, which meant that all cognates with diacritics had to be checked manually. In most cases, common Old Germanic and etymological dictionaries were used to determine the correct form. Finally, the letters N (in Dutch) and R, which had been written by Jim Holders and Han Nijdam under the supervision of Boutkan in 1997 and 1998 respectively, differed structurally from the other lemmata and needed special attention. Even though Boutkan had made some corrections on paper, the N and R lemmata needed major additional revisions with respect to both content and structure.

In October 2003, the additional lemmata were written. Although at an earlier stage a non-intervention strategy had been adopted, the manner in which Boutkan had split up and referred to the lemmata demanded that at least 65 core lemmata had to be written to further the coherence. The remaining 300 lemmata, which were either compounds with references, enclitica or proper nouns (such as religious days, proper names, etc.), could theoretically have been left out, but the reworking of the reference system (as described in section 1.3 “The structure of the lemmata”) would have taken a lot more time than adding them

Below are lists of all the missing lemmata that were added to the dictionary. The first one contains the core lemmata and the second one contains the compound and reference lemmata.

1. core lemmata

2. compound and reference lemmata

Early January 2004, when the first versions of the new lemmata were ready, the corrections of the review commission were incorporated into the manuscript. The review commission of PIE specialists (see “preface”) had as a working premise that only obvious errors were to be corrected. Thus, for example, even though they might in certain cases disagree with Boutkan’s views, the original explanation would be retained, provided it was persuasively supported by arguments.

In May and June 2004, the final tasks were concluded: the preface and the introduction were written, the last corrections on the indices and the bibliography were implemented, and a list of all abbreviations was generated and subdivided in meaningful categories (see “conventions, symbols and abbreviations”). Finally, the manuscript was typeset and the proofs were corrected.

Conventions, Symbols and Abbreviations

Citation of orthographic forms

If a language has a conventional roman orthography, this will be used except where phonetic or phonemic representation is appropriate. If the original alphabet is non-roman (as in Sanskrit, Greek, Old Church Slavic, Germanic runic inscriptions), the normal transliterations are used, with diacritics for length and (where relevant) accent. I will comment below on special conventions for some of the more exotic languages.

Length

Long vowels are marked with a macron <¯> in all languages except Old Icelandic and Old Irish, where length is conventionally indicated by an acute: OE gōd, OIc. góðr ‘good’. In general, where length is marked (as in OE, OFris., Latin, Greek, etc.) short vowels will be left unmarked: except when shortness itself is of importance, in which case a breve <˘> will be used: Lat. stă-tus vs. stā-re.

Special conventions for specific languages

i.              Proto-Indo-European: *[m, n, l, r] are syllabic resonants; */H/, with or without subscript numerals, represents a laryngeal.

ii.             Sanskrit: <ṭ, ḍ, ṇ, ṣ, ḷ, ṛ> are retroflex; <ś> = /ʃ/; <c> = [tš]; <j> = [dž]; <ñ> is a palatal nasal; <ḥ> = [h]; <h> = voiced glottal fricative [ɦ]; <bh, dh, gh> are breathy-voiced stops; <ṁ> = nasal with the same place of articulation as a following consonant; <r l> are syllabic; <e, o> are long vowels; <ai, au> are long diphthongs;<´> marks an accented high-tone vowel.

iii.            Avestan: <ą> = nasalized a; <x> = [x]; ϑ = [ϑ]; <β,δ,γ> are voiced fricatives; <c, j> (formerly <č, ǰ>) as in Sanskrit; <ń, ŋ́, x́> are palatalized; <-t> is an unreleased (i.e., the first part of) [t]; <y>, <v> are also written <ii> and <uu> respectively in a non initial position.

iv.            Old Persian: <c, j>, see Sanskrit; <ϑ> = [ϑ]; <x>; <ç> = [ř].

v.             Hittite: <s> also written <š>; <z> = [ts]; <h> (also ḫ) = [x].

vi.            Tocharian: <a> = [ə]; <ā> = [a]; <ä > = [I]; further as in Sanskrit.

vii.           Armenian: <p^c> etc. = [p^h]; <c> = [t^s]; <j> = [d^z]; <c^c> (formerly) = [t^sh]; <č> = [t^š]; <ǰ> = [d^ž]; <č^c> (formerly c̣^c) = [t^šh]; <r> (formerly ṙ) = /rr/.

viii.        Lithuanian: <˜> = rising pitch; <´> = falling pitch; <`> = accented short vowel; <ė> = long closed [e]; <e, a> (originally nasalised) = long [œ, a]; <y> = /ī/; <ie> formerly <ë>; <uo> formerly <ů>; <š> formerly <sz>; <č> formerly <cz>.

ix.           Old Church Slavonic: <ě> = [ɛ]; a hook under a vowel symbol as in <ę> = nasalization; <ь, ъ> as in pit, cut; <š> = [ʃ]; <c> = [t^s]; <č> = [t^š]; <č> = [t^š]; <y> see Avestan.

x.            Old Irish: <c> = [k]; <´> marks a long vowel.

xi.           Welsh: <ll> = a voiceless fricative /l/ whereby air escapes on one side.

xii.          Old Norse: <ǫ> = [ɔ].

xiii.          Gothic: <ei> = /ī/; <ai, au> (also aí, aú) = [ɛ, ɔ] before <r, h, ƕ>; <ƕ> = [x^w]; <gg> = [ŋg]; <q> = [k^w].

Other symbols

Abbreviations of languages

Abbreviations of grammatical terms, sound changes, etc.

Abbreviations of OFris. mss. and dialects