Proto-Indo-European Syntax

6. Lexical Entries

6.1. Comprehensive Treatments of the PIE Lexicon.

During the century and a half of IE investigations, the lexicon has been thoroughly studied. A comprehensive summary of these studies was made by Alois Walde and Julius Pokorny (1927-1932; usually referred to as WP); it was based on earlier dictionaries, notably that inaugurated by August Fick (1890-1909). Pokorny's later dictionary (1959-1969) is essentially a rearrangement of the data in WP in accordance with the order of the Latin alphabet. Dictionaries for individual branches of IE and for individual dialects contain more recent findings than does WP. But like it they concentrate on etymologies and lexical meanings. They do not discuss the roles of individual lexical elements in syntactic units; WP, for example, does not indicate the cases accompanying verbs. While these dictionaries are indispensable for IE syntactic studies, they lack information which a comprehensive lexicon should include and which a subsequent compilation will presumably contain.

Information concerning some lexical features of direct importance to syntactic studies is, on the other hand, included in grammatical treatments, notably Delbrück's second volume of 1897; as its bibliography indicates, this volume in turn is based on earlier studies. Subsequent studies, such as Meillet's of the root *men- (1897) and Emmanuel Laroche's of the root nem- in Greek (1950), have investigated the uses of specific lexical elements in further detail. These too need to be pursued with similar precision. Lexical analyses must be carried out in accordance with a specific theory of grammar; and lexical entries must be so described that their role in sentences is specified as fully as are the characteristic syntactic patterns of PIE. Here only a highly preliminary treatment can be included, with selected lexical items presented only as examples. In view of the central importance of the verb, a selected verbal root will be examined first in the following section, then this root accompanied by preverbs, and then other elements of the PIE lexicon.

6.2. Lexical Analysis of a Selected PIE Root: γer- ‘move’.

WP includes a long entry for this root, written as er- (I,136-142; see also Persson 1912:767-773). In Sanskrit grammars it is written ṛ-, or ar-. For it Walde and Pokorny give the meaning: ‘sich in Bewegung setzen, erregen (auch seelisch, ärgern, reizen); in die Höhe bringen (Erhebung, hochwachsen), z. T. aber auch von Bewegung nach abwärts’ (‘move, arouse, raise, in part however also of downward motion’). This attempt to give the meaning of a root must be interpreted carefully, for, as examples of its occurrences indicate, the root differs in meaning in accordance with the formations it adopts. If we would set out to characterize the meaning of the unextended root, we would describe it as intransitive and punctual, in accordance with the occurrence of the third singular present in the Hittite Laws (J. Friedrich 1959:74, § 50):

Further examples will permit us to describe its meanings in greater detail.

When listing the root WP gives as its “base-forms”: er-, ere- (“thematic”), erē(ʔ), erei-, ereu-, eras. In accordance with the lexical analyses carried out particularly by Meillet and Benveniste, we would propose a root consisting of two consonants, which might be followed by suffixes of specific shapes (Lehmann 1952:17-18). Because of the initial vowel in Greek forms like ȏrto ‘has arisen’ (Delbrück 1897:101), I assume that the initial consonant was the o-coloring laryngeal γ. The root and suffix combinations must be written in accordance with the patterns determined in IE lexical research; thus the PIE suffixed form written ereu- in WP would be cited as γr-ew-.

With its intransitive, punctual meaning, the simple root γer- would be expected to take the endings of the h- conjugation, as indeed its Hittite reflex does; arhi ‘I arrive (at a place)’. Moreover, for transitive and durative uses we would expect affixed forms. These are found in Vedic and also in other dialects, including Hittite. Virtually all the present system forms of this root in the Rigveda are made in accordance with the third reduplicating class, íyarti, or with the fifth nasal infixing class, ṛṇóti. In addition there are three occurrences of present forms with the -sk- suffix, all in the late tenth book, and two with the suffix -e/o-, both in near-contiguous hymns of the seventh book. It is likely therefore that the unextended forms of the root were inflected in the aorist or the perfect; although the Rigveda contains only constructions of perfect forms with preverbs, Delbrück, on the basis of Skt. ā́rat and Gk. ṓreto reconstructs the “proethnic aorists” *ṓret, ṓreto (1897:101). The “similarly proethnic” Skt. ā́rta, Gk. ȏrto, which he also cites, permit us to reconstruct PIE é-γr-to ‘he moves, rises’, as in the following line from an Archaic hymn addressed to Agni:

Rather than the analysis of Delbrück (1897:94-95), according to which such aorists might be derived from perfects, the assumption that PIE roots might be inflected either in the h- conjugation or the m- conjugation clarifies the attested forms of roots like γer-. As a root with a lexical meaning which is intransitive and punctual, γer- would be expected to inflect in accordance with the h- conjugation.

The lexical entry of the root would then be as follows:

This entry would require an agent noun, as in Example 2. If the agent generated the first- or second-person pronoun, this would be indicated in the verb form, as in the following passage from a Strophic hymn to Agni:

Number, whether plural, dual, or the unmarked singular, would also be introduced as a property of nouns lexically inserted in sentence strings through the K nodes. For early PIE, number was not significant, as the reconstructible forms indicate (Lehmann 1958). In late PIE and the dialects, number was indicated in the verbal phrase as well as the nominal phrase, by a congruence rule requiring number and person agreement of verbs with subjects. The well-known use of singular verb forms with neuter plural subjects provides insight into this development.

The definition ‘move, rise’ should be analyzed for its lexical features. Such analysis is a task of the future. At present, in IE studies as well as in general linguistic studies, it is being carried out for selected nominal sets, such as the kinship terms and the names of trees (Wordick 1970; P. Friedrich 1970). In the future, similar studies should be carried out for verbs. For the time being we can indicate specific features associated with verbs, such as the requirement for γer- that its agent be active, usually animate; in this way it contrasts with g^wem-, which commonly has inanimate subjects, such as wagons or prayers. A set such as verbs indicating motion might well be investigated much as Paul Friedrich has investigated the set of names for trees.

Here a tentative set of features is given for the roots γer- and g^wem-:

These roots accordingly are highly similar. Both differ from wegh- ‘move’, which is +transitive and -punctual. They also differ from their affixed forms and their combinations with preverbs. When used in sentences, the actual forms, such as *eγrto > Gk. ȏrto, would be generated on the basis of categories introduced in the Q component, as well as those like -transitive, +punctual, which determine its inflection according to the h- conjugation.

In an OV language, roots are commonly followed by suffixes. Examples of such possibilities may be found in dictionaries like Grassmann's for the Rigveda (1872) or Ananthanarayana's for the Taittirīya Brāhmaṇa (1970a). The forms for the root ṛ listed by Ananthanarayana (1970a: 7-8) are as follows:

In addition to the “indicative”, three imperative forms are attested: ṛṇvatu, ṛchatu, arpayatu. As noted in § 4.1.6, the imperative was marked with affixes, such as -u in these forms.

Moreover, four optative forms are attested: ṛchét, ṛchéran, archeyuḥ, arpayet. The optative affix, like the imperative affix and other affixes which survived into the inflectional system of late PIE and the dialects, was placed after those affixes which indicated Q categories, like the causative; the difference in position and in frequency led to the subsequent distinction between derivational categories and inflectional categories.

In addition, two perfect forms of ṛ are attested: ārimá and an “imperfect” ārchat. The form ārchat illustrates the development from a verbal system with forms made from the root to a system in which forms were based on the “present” stem. This development is even more notable in dialects with late materials, such as Latin; its reflexes of the root γer may be illustrated with the principal parts: orior, ortus sum, orīrī ‘arise’. The deponent inflection of orior recalls the intransitive lexical meaning of the PIE root, though its structure of inflections differs greatly from that of PIE; as its principal parts indicate, it is inflected as a deponent throughout the six tense forms of the Latin verb.

Affixed forms of roots illustrate the PIE verbal system, such as forms with -ew-, which are maintained in Lat. ruō, ruere ‘move down, tear down’. Because the IE verbal system has changed considerably, it is difficult to determine the meanings of the affixes like -ew-, which came to be derivational. In general such affixed forms must be described as separate lexical items, with distinct lexical characteristics. Thus Lat. ruō < γr-ew- is transitive in contrast with the simple root. For its PIE etymon the lexical entry would be:

γr-ew- ‘move/tear down’ ___ Agent Target (K)

plus its lexical features. The extended form is most widely maintained with an n infix, as in Skt. ṛṇóti, Gk. órnumi, Hitt. arnuzi, as in the following passage from the Hittite Laws, in which the object is implicit (J. Friedrich 1959:20, § 19):

The n forms add to the extended root a factitive or causative notion; but the transitive characteristic of the form is conveyed by the -ew- extension, as may be noted by examining the various “meanings” of ṛṇóti in the Rigveda, all of which are accompanied by an accusative.

Reduplicated forms of the root have an iterative and often an intensive meaning (Delbrück 1893:16-26). Thus íyarti means ‘to move back and forth’, whether or not it is transitive, as it probably is in all of its occurrences in the Rigveda, such as the marked sentence pattern in the Strophic hymn of dialogue between Varuna and Indra, who says:

The same meaning, though in a transferred sense and intransitive, is found with an alternate reduplicated form: álar-, as in the following Strophic hymn:

Reduplicated forms of the root, however remodeled phonologically, would then have the lexical entry:

γir-γer- ‘move back and forth’ ___ Agent Target (K)

One further suffixed form of the root will be noted here, with -sk-, as in Skt. ṛcháti, Gk. érkhomai, Hitt. aršk-, an iterative. Although the forms of Skt. ṛch- are consistently transitive in the Rigveda, with the intensive meaning ‘reach, strike, harm’, the forms in Greek are commonly intransitive, except in passages like Andromache's lament for Hector:

The lexical entry of the root with this suffix would then be:

γer-sk- ‘move steadily to’ ___ Agent (Target) (K)

To such extended forms, as well as roots, the affixes could be added which expressed the meanings introduced through the Q component. Examples of such forms were cited above from the Taittirīya Brāhmaṇa. One further example is cited here, a middle form from Hittite, where the reflexive meaning ‘move oneself, place oneself’ often simply corresponds to ‘stand’, as in the following Old Hittite sentence from Otten and Souček (1969:18-19):

The combination of Q categories with verb forms and their meanings in combination with lexical items have also been discussed in Chapter 5. The Q categories simply added nuances of meaning to the lexical meanings, which could however be modified by other lexical elements.

6.3. Lexical Items Consisting of Verbal Roots and Preverbs.

Already in PIE, combinations of roots and preverbs have special lexical characteristics and are to be treated as individual lexical items. For example, in the Rigveda twelve preverbs may be used with the root ṛ-. The combined lexical unit made with one of these, Skt. prá ‘forward, ahead’, illustrates the constructions and senses as they differ from those of the simple root.

The combination is attested with the straightforward meaning of both constituents, and with intransitive force, as in the following Strophic passage:

But a passage in a neighboring Strophic hymn illustrates a distinct “idiomatic” meaning for the combination and its particular syntactic construction, accompanied by an accusative and a dative:

In both of these passages the preverb occupies initial position in the clause, in accordance with marked order. But in Example 10 it must be taken as a member of the compound lexical item, which would be stated as follows, in addition to its lexical features:

pra ṛ ‘send’ ___ Agent Target Receptor (K)

It is noteworthy that such compound verbal expressions are accompanied by more than one case node, in contrast with the syntactic patterning of most simple roots, as pointed out in Chapter 2 (see also Gaedicke 1880:35-36).

The history of such lexical combinations is well known. In the course of time they came to be units, as the dialects became SVO in structure. Thus Greek has a separate lexical item proérkhomai ‘advance’, attested in post-Homeric materials, though even in Homer many such combinations were already units. And among the compounds of Goth. rinnan ‘run’, presumably a reflex of PIE γr-n-ew-, is faurrinnan ‘proceed’. Yet the length of time that elapsed before such combinations became units is indicated by the difference in accent and vocalism in contemporary German between verbs like erteílen and Úrteil, and also by earlier evidence, such as the Gothic sequences in which clitics were placed between the preverb and the verb, as in the form of atgiban ‘give to’ in Mark 14:44:

See also § 4.3.1 b. Accordingly, in PIE, combinations of preverbs and verbs must still be viewed as morphologically and phonologically distinct; but they must be treated as lexical units, much as is pra ṛ- in Vedic.

6.4. Nominal Lexical Units.

Among the lexical elements which may fill the K categorial positions are nouns. As noted above, Indo-Europeanists have long insisted on assuming a strict distinction between nouns and verbs. The distinction is expressed most clearly in derivation; as Benveniste pointed out (1935), nouns may have more suffixes added to roots than may verbs.

Whether or not Gk. órnis, órnīthos ‘bird’ and cognates such as OE earn ‘eagle’ are derived from the root γer-, as tentatively proposed by WP (I, 135) and assumed here, the forms may be used to illustrate the greater morphological complexity of nominal elements. Thus the Greek stem ornīth-o- has an n suffix, as in Germanic, and a -th- suffix. The dialectal Lith. arẽlis has an l suffix. Germanic has a w suffix, in Old Norse (ON) ǫrn < *arnuz. Hitt. haran- ‘eagle’, which must be related to the forms in the other dialects, has an n suffix; by the etymology adopted here, the Hittite word for ‘eagle’ provides direct evidence for the initial γ- of the PIE root.

Another noun derived from the root γer- is Skt. árṇas ‘streaming flood, sea’, as well as arṇavá ‘surging, sea’. These may be related to Hitt. arunaš ‘sea’, And Germanic forms, such as Goth. runs, OE ryne, ON run ‘stream’ provide further examples of nouns with the meaning ‘stream, sea’ derived from this root. Like ON run, Skt. árṇas is a neuter noun.

Assuming PIE γorn- ‘eagle’ and γern- ‘sea’, their entries in the lexicon would have the following features, among others:

Nouns thus would be specified for gender, as common or proper, for concreteness or abstractness, as animate or inanimate, as count or mass nouns, and for specific lexical features.

One set of lexical items which has been thoroughly described is that of the kinship terms. The most recent description is given by Frank Wordick in his doctoral dissertation (1970). When semantic study is further developed, additional sets of nominal elements will be described with thoroughness similar to that applied by Wordick, whose work may be used as a pattern for such study. As did Wordick in accordance with his lengthy bibliography, such lexical studies will make use of many admirable treatments of the PIE vocabulary, such as Benveniste's “Vocabulary of IE institutions” (1969). Like Delbrück's monograph of 1889 on the PIE kinship terms, these treatments contain excellent information on the IE lexicon, though it is not analyzed in the detail found in the studies of Paul Friedrich (1970) and Wordick.

6.5. Pronominal Lexical Units.

The so-called personal pronouns provide us with insights into the earlier nominal system. The separate lexical elements for singular and plural forms, as well as for the oblique cases, given above, § 5.4.3, indicate that the sets of pronouns were made up of individual lexical items rather than members of a close-knit paradigm. Such a situation is also found in the r/n nominal stems, with their distinction between r forms in the nominative/accusative and n forms in the oblique cases, as in Hitt. uttar ‘word, thing’, Gen. uddanaš; Nom. Acc. pl. uddār, Gen. uddanaš We assume that personal pronouns, like nouns, were introduced through the K nodes.

Presumably, at an early stage of PIE, person was not a syntactic category. When a person was to be specified, a lexical element was used, e.g., *eg-/me- ‘I’. At a subsequent period person was indicated in the verb, first with no distinction for number, but by late PIE with the inflectional markers for person and number that are listed in IE grammars (Brugmann 1911:378-427; 1916:583-666). The complicated morphological problems and the reconstructions for the various stages of PIE cannot be pursued here. Brugmann and his predecessors solved some of the problems, and others have been treated subsequently (see Seebold 1971). In a syntactic treatment the primary interest lies in the introduction of the category for person and markers for person into the sentence. This was effected through K nodes, as we have noted, both for personal pronouns and for demonstratives and interrogatives (for these, see Brugmann 1904b; 1911:302-377).

The lexical entries for pronouns would then be similar to those for nouns, though they would contain very few features. The lexical entry for *eg- would only have the features personal pronoun and first person or in Brugmann's term I-deixis (1911:312). That for a demonstrative pronoun would have the feature this-deixis or that-deixis. Except when emphasized, these features would be associated with the verb form in late PIE, and expressed through the first, second, and third endings of finite forms. When emphatic, they would be marked with special lexical items, as described in morphological treatments.

6.6. Adjectival Lexical Units.

Adjectival modifiers in early PIE were simply nominal elements preposed to nouns, as noted with the example of Hittite kurur above (§ 5.3.2; see also J. Friedrich 1960:116-117). The lexical entries of the PIE etyma of works like Hitt. mekki- ‘much, many’, kurur- ‘inimical, enemy, enmity’, takšul ‘friendly, friendship, peace’ would then be given as are those of the nouns in § 6.3. Their function as modifiers or nouns would have been indicated by their position in the sentence.

The lack of distinction between adjectival substantives and nominal substantives in PIE is reflected in further characteristics of the early dialects. One such characteristic is the late introduction of gender inflection in the adjectives, as noted above (§ 5.3.2). Another is the possibility of affixes indicating comparison on nouns as well as adjectives in Sanskrit, as in the following Strophic passage:

Further evidence for the earlier lack of distinction between nominals used attributively and those used as nouns is provided by the cardinal numerals, which for the most part are uninflected. Accordingly, lexical entries would not distinguish between PIE nouns and adjectives; these two differing functions would be indicated in the early period by order and only later by special inflections and thus distinct selectional characteristics.

6.7. Adverbial Lexical Units.

As Brugmann has indicated (1911:667-758), adverbs are words that are used in the first instance to modify verbs, though they may also be used to modify adjectives and nouns. Moreover, while they may have suffixes comparable to those of substantives, they are not inflected in paradigms; even though a frozen genitive autoȗ is used in Attic and a frozen locative auteȋ in Doric Greek to indicate the adverb ‘here’, the two forms and other similar forms cannot be associated in a productive paradigmatic set. Adverbs then are frozen forms, some of which came to be disassociated from paradigms.

Adverbs are constantly being “renewed” from other elements in the language, and accordingly it is difficult to reconstruct specific forms for PIE. Hitt. parā ‘forward, away’ and Skt. parā ‘away’, however, as well as Lat. prō ‘forward’, permit the reconstruction of PIE *prō, with alternate forms *pṛrō, and *perō. As generally assumed (WP:II, 29-40), the adverbial forms are derived from a word *per, homonymous with the root *per- ‘cross over, penetrate’. Besides *perō, other extended forms were made from this root, and used as function words: *peri, *prai, *pru, *pres, *preti, each with possibilities of further extension, as may be observed in the long entry for the root in WP, such as Gk. próka ‘at once’ and OCS proče ‘therefore’ (see also Brugmann 1911:864-888). Which of these extended forms one should assume only for the dialects and which one should reconstruct for PIE is a difficult problem. But on the basis of Hitt. parā and Skt. parā, we may assume a PIE adverb *perō. Its precise etymology is not of primary concern here, that is, whether it should be regarded as a case form of a root noun *per or an extended form of a particle *per (Delbrück 1893:536-643, esp. 641-643). The reconstructed adverb *perō would have had uses such as the following in an Old Hittite text (Otten and Souček 1969:20, § 20):

The Hittite adverb parā is similar in form to other prominent adverbs in Old Hittite: anda ‘in’, appa ‘back’, šara ‘upward’, and also (a)šta ‘away’ (see the excellent discussion in Otten and Souček 1969:82, 86-88). Its lexical entry would be parallel with that of other adverbs such as Hitt. anda, which is related to Skt. ánti, Gk. antí ‘towards’ (WP:I, 58-59, 65-67).

Further lexical characteristics are facultative, like the last given here. Selection of them would vary with the verb of the clause and accompanying nominal forms; these modified considerably the meanings of adverbs as the discussions of Delbrück (1893) illustrate.

6.8. Preverbs, Postpositions, and Particles.

In their discussion of parā and other adverbs, Otten and Souček point out the difficulty of distinguishing such words from preverbs and postpositions (1969:88). Actually the occurrences of parā in the Rigveda are all interpreted as preverbal by Grassmann (1872:782). The following example may illustrate the problems involved in distinguishing whether such words are to be interpreted as adverbs or preverbs.

When the adverbial element generally combined with the verbal to form a unit with distinct meaning and syntactic construction, it would be considered a preverb, and the combination would be treated as are the items in § 6.3.

As a further development, preverbs were used as postpositions. Another passage containing parā and the root gā- may illustrate such a development:

Indo-Europeanists have long noted this development and also the further development of such postpositions to prepositions (Delbrück 1893:643-774). The use of preverbs as postpositions can be understood on the basis of recurrent patterning in the OV period of PIE; their further development to prepositions resulted as the dialects came to be VO.

In a similar way, preverbs could be used to introduce clauses, and they then developed into conjunctions (Brugmann 1904a:666). The Greek particle per may provide an example (Denniston 1966b:481-490); the various uses of per cannot be discussed here, but an illustration of its concessive force as a postposed particle is found in the following line:

Although Denniston accepts the etymology of Brugmann, by which per developed an intensive meaning, ‘completely’, from the meaning ‘around’, and further a contrastive force, as illustrated in Example 16, the steps in such developments are difficult to construct in the absence of earlier records. The particles reconstructed for PIE (Brugmann 1916:969-1009) must be described to the extent permitted by our evidence and our understanding of the uses of particles; see above, § 5.5. They were ancillary elements indicating relationships between the primary words in sentences or between sentences, and in this way they developed a wide variety of uses.

Their role in the sentence structure of PIE presumably was similar to that in early Indic prose, as illustrated in Example 64 of Chapter 3. Particles, such as ha and vai of the first sentence of that example, and preverbs, such as prá and sám in Sentences 6 and 7, specified the relationships between nouns and verbs. Still other particles introduced sentences, such as átha in Sentence 9 and sá in Sentence 7, to which the sentence-introductory pronominal elements of sentences 2, 3, 4, 5, 6, 8, 10, and 12 are related. In Vedic, the particles were placed in accordance with their role in the clause, initially when modifying the entire clause, medially when modifying nouns and verbs. In Hittite, on the other hand, they came to be grouped initially, as indicated in § 5.5.3. The difference in arrangement resulted from different syntactic principles, presumably introduced into Hittite by the influence of neighboring languages in Asia Minor, as suggested in § 5.5.3. Though ultimately distinct in their uses, many of the particles, preverbs, and adverbs developed from the same lexical elements, as the variety of their uses in the early dialects suggests.

6.9. Extended Forms of Verbal and Nominal Roots; Further Development of the Lexicon.

The PIE roots are to be regarded like the roots of OV languages, such as Turkish, e.g., Turkish gel- ‘come’, git- ‘go’, and so on (see § 5.4). To these roots various affixes can be added, e.g., gel-me-di-m ‘come-not past-I = I didn't come’, gel-e-bil-eceǧ-im ‘come-gerund-can-will-I = I will be able to come’. Moreover, by means of derivational suffixes, nouns can be derived from these roots, such as the action noun geliş ‘the act of coming’. These in turn may be the bases for denominal verbs, such as gelişmek, infinitive, ‘grow up, develop’ in contrast with gelmek, infinitive, ‘come’. Similar processes can be assumed for PIE. As a result of these processes, not merely root nouns and simple verbs but also extended forms of roots and various formations of verbs and nouns must be treated as basic forms in late PIE. The bases of the lexicon were accordingly undergoing change.

As a result of such changes, forms that are listed as roots do not necessarily have the canonical shape of PIE roots: CeC-. For example, the widely attested root *weyd- is always found with a -d- extension. Such extensions may be expected to specify in greater detail the meanings of roots. To ascertain their effect, Per Persson carried out his monumental studies (1912). Traditionally in IE investigations such extensions are called determinatives when they are consonantal, suffixes when accompanied by a vowel. Another notable study of a determinative, -dh-, was carried out by Benveniste (1935; see also Lehmann 1942). Because of the time depth involved, the meanings of the determinatives and suffixes may now be difficult to ascertain. W. S. Allen, in a detailed study of the affix -bh- (1950), likened such elements to the affixes marking gender classes in Bantu. Whether or not their meanings can be determined, such affixes had a marked effect on the IE lexicon.

Their effect may be best determined by thorough study of the forms in the early dialects. Thus Elmar Seebold has made a detailed investigation of the root *weyd- (1973), finding various meanings for it in Aryan. If active and accompanied by the genitive, it meant ‘come to know’; if middle and accompanied by the genitive it meant tentatively ‘become aware of’. If active and accompanied by the accusative, it meant ‘decide’; the comparable middle meant ‘make a decision’. Further, a middle form in absolute use meant ‘turn up, come’. To understand these various meanings one might examine simple roots of the shape wey- in PIE. WP lists three such roots: wey- (I, 223) ‘turn, bend’; (I, 227) ‘wither’; (I, 228) ‘go after something’. The extended form *weyd- might well be based on one of these; ‘decide’ for example might result from a modified form of wey- ‘bend’. Or *weyd- may be based on a root which is not attested. It would be important to account for the diverse meanings of roots like weyd-. Here the example is cited as an illustration of difficulties involved in exploring earlier forms of the IE lexicon.

Other difficulties result from phonological losses. Thus roots with initial laryngeal, like *γer- have many reflexes in which only the last consonant of the original root is maintained. By means of a rigorous formal analysis, such as that of Benveniste, the PIE lexicon can, however, be determined with some precision. The various nominal and verbal affixes have been extensively investigated, as by Brugmann (1906:126-581; 1913:86-390) and in many subsequent studies, such as those cited here. Such investigations must be based on an understanding of the formal characteristics of the PIE lexicon and its possibilities for modification.

To be sure, borrowings are included in the earliest reconstructible corpus of the lexicon. Some of these may be indistinguishable from the native vocabulary and accordingly subject to false analysis. The nominal borrowings are most readily recognized. Thus the apparent archaic word for ‘axe’, Skt. paraśús, Gk. pélekus, may be a borrowing from Akkadian pilaqqu; though since this apparently means ‘spindle’, the borrowing would have been made in some sort of ceremonial context (Mayrhofer 1953:213-214). Another word which may have been borrowed through cultural contact with the Mesopotamian area is that for cattle, *g^wōus, cf. Sumerian gu (Pokorny 1959-1969:482-483).

Whatever the source of such words, they must have contributed to changes in the rules of derivation for the IE lexicon. Like that of any language, this was undergoing modification. Even the lexicons of early dialects, like Vedic and Greek, contain words which differ considerably in their structure from those that must be assumed for PIE.