OLD ENGLISH GE- AND THE DESCRIPTIVE POWER OF NERTHUS

The main purpose of this article is to carry out an extensive examination of the evidence available of the Old English affix gein general, and, in particular, of the ge-alternation, which holds between predicates that bear a form-and-function contrast such as hlid ‘lid’~gehlid ‘roof’. The gealternation is examined as a pure and a mixed alternation, co-occurring with gradation, zero derivation and affixation. The conclusion is reached that geis the most frequent and the most widely distributed affix in Old English. Moreover, it partakes in inter-categorial and recursive derivation. As a secondary goal, this article aims at assessing the descriptive adequacy of the database of Old English derivational morphology Nerthus both on qualitative and quantitative grounds. In this respect, the conclusion is reached that the amount of evidence of gethat can be drawn from Nerthus allows for an extensive analysis of the affix. On the other hand, Nerthus, in its present state, requires more accurate and formalised meaning definitions.


INTRODUCTION
The purpose of this article is to bring, by means of an extensive examination of the evidence available for Old English ge-, a new way of looking at this affix. Additionally, the evidence of the affix gedrawn from Nerthus (Martín Arista, Caballero González, González Torres, Ibáñez Moreno and Torre Alonso fc.) will be used to assess the descriptive power of this lexical database both on qualitative and on quantitative grounds.
Although this piece research develops along quite a different path, I would not do justice to Lindemann's work (1953;1970) if I did not acknowledge that their publication represents a milestone in the long tradition of studies in ge-. Indeed, this author proved empirically inadequate two ideas central to philological studies, namely that geis often meaningless (Benson 1701;Adelung 1796;Krapp and Kennedy 1929;Hollmann 1936;Samuels 1949) and that this affix is monosemic, expressing transitivity (Lenz 1886;Lorz 1908), intensification (Bernhardt 1870;Wackernagel 1878;Lorz 1908;Weman 1933), completion (Grimm 1878;Dorfeld 1885;Wustmann 1894;Swaay 1901) or perfective aspect (Martens 1863;Streitberg 1943). 2 In spite of his contribution to clearing the traditional underbrush, Lindemann's work is fraught with a fundamental problem: focusing on verbal ge-, Lindemann overlooks the transcategorial dimension of derivational morphology. On this question, Lindemann follows the long and crooked path of philological studies published from the 18th century onwards, most of which I have just quoted after Lindemann himself (1953). In this paper I harbour the conviction that the lack of a clear morphological perspective, based on a comprehensive theory of morphology, has weakened the approach to Old English derivation in general and to gein particular. I can cast two arguments in favour of this view. To begin with, if the distinction between inflectional morphology (always intra-categorial) and derivational morphology (often inter-categorial) had been drawn before in the history of linguistics, the approach to gewould not have been exclusively verbal, as in Samuels (1949) and Niwa (1966), to quote just two authors from the 20th century; and the inflectional or derivational nature of this affix had not been the spine of the discussion. And, in the second place, if derivational morphology had been understood as a phenomenon of its own, independent of inflection, the studies in the derivation of lexical categories, including, for instance, Schön (1905), Schuldt (1905) and Nicolai (1907), would have gone into the heart of the debate, instead of taking issue with the intricacies of inflectional categories in the diachronic perspective.
Given this state of the art, I offer an analysis of gewhich touches upon questions of morphological relatedness and contrast and focuses on morphological processes (Ablaut, zero derivation and affixation) and all four major lexical categories: Noun, Adjective, Verb and Adverb. With this aim, this journal article is organised as follows. Section 2 raises the methodological question of the interaction between synchrony and diachrony. Section 3 offers a quantitative description of geaffixation, as well as a qualitative approach to categorial distribution and change.
Section 4 deals with the pure ge-alternation, both intra-categorially and intercategorially. Section 5 engages in the overlapping of gewith other derivational principles, namely Ablaut, zero derivation and affixation. Section 6 offers a summary of the results of the research undertaken in this article.

DRAWING EVIDENCE FROM NERTHUS
The lexical database of Old English Nerthus can provide extensive empirical evidence on the affix gebecause it contains around 30,000 entries (headwords) with a full description of their derivational morphology, along with the inflectional morphology relevant for derivation. Nerthus analyses each headword in terms of more than sixty variables, which are grouped in three blocks of information: predicate (including category, form, variants, translation, inflectional morphology, type of predicate and morphological process) derivation and compounding (both including canonical, non-canonical, inflective, and phonologically-modified base and adjunct). The organisation of the database is categorial (distinguishing between nouns, adjectives, verbs, adverbs, adpositions and affixes, which are treated as predicates, or full lexical items). Terminal and non-terminal derivational chains constitute the main descriptive device of Nerthus (terminal chains represent non-recursive derivation whereas non-terminal chains represent the pre-final stage of recursive derivation).
As I have already remarked, it is a secondary aim of this research to test the descriptive adequacy of Nerthus against ge-, one of the most widespread features of the language, and a question that has engendered one of the longest debates in English historical linguistics. For this reason (and for the reasons given in the state of the art), the main thrust of the research that follows is description. Sections 3, 4 and 5 gather a body of evidence that represents the main contribution of this journal article, given its scope and depth. Regarding scope, the evidence furnished is both intra-categorial and inter-categorial; as for depth, the quantification of the affix geis exhaustive and many aspects of the gealternation are nearly exhaustive.
Of all the possible aspects relating to the affix ge-, the evidence under scrutiny here is the gealternation. The gealternation holds between predicates (in the sense of lexemes) that bear a form-and-function contrast such as those in (1a), and excludes those predicates that do not bear a functional contrast, like the ones in (1b). 3 (1) a. hlid 'lid'~gehlid 'roof', lang 'long'~gelang 'dependent', sprae:dan 'to spread'~gesprae:dan 'to stretch forth', hwae:r 'where'~gehwae:r 'everywhere' b. eardung 'tabernacle'~geeardung 'tabernacle', gylden 'golden'~gegylden 'golden', waerian 'to pass by'~gewaerian 'to pass by', hle:owe 'in a sheltered manner'~gehle:owe 'in a sheltered manner' The analysis that is carried out in this journal article regards Ablaut, zero derivation and affixation as formative principles of derivational morphology. These phenomena are illustrated, respectively, by (2a), (2b) and (2c): (2) a. full 'full'~gefyllednes 'fulfilment', helm 'protection'~gehilmed 'helmeted', he:ah 'high'~gehe:han 'to raise' b. baec 'back'~gebaecu 'back part', clibbor 'clinging'~geclibs 'clamour', gleng 'ornament'~geglengan 'to set in order' c. fre:o 'freedom'~gefre:ogend 'liberator', glo:f 'glove'~geglo:fed 'gloved', eftgian 'to repeat'~geeftgian 'to restore' The concept of the interaction between synchrony and diachrony underlying the alternations in (1a) and (2a)-(2c) is necessarily dynamic, since Ablaut, zero derivation and affixation are not simultaneous processes in a strictly synchronic analysis. It is usually the case that in the diachronic axis, phenomena overlap throughout processes often blurred by the lack of evidence or even contradictory data. Diachronically, strong verbs turned out nouns, adjectives and other verbs by means of graded or ungraded zero derivation (Hinderling 1967;Kastovsky 1968), while nouns and adjectives produced weak verbs. Then, fully productive affixation was followed by less productive affixation and more productive compounding (Hiltunen 1983). Categorial extension (as in riht 'right'~right 'something right') must have been active throughout the three diachronic stages I have just sketched. Overall, two central distinctions arise: between lexical derivation as stem formation versus word formation on the one hand (Kastovsky 1986(Kastovsky , 1990(Kastovsky , 1992, and between more productive affixation as opposed to more productive compounding, on the other (de la Cruz 1975;Horgan 1980;Hiltunen 1983). 4 It is my contention that gecan throw more light on the evolution of word formation in Old English than any other affix, for three reasons: firstly, because geis the most frequent (in the sense of type-frequency) affix in Old English; secondly, because geis the only universal affix (given that it distributes over all lexical categories); and, thirdly, because geco-occurs with all major morphological processes throughout the history of Old English word-formation, including, at least, Ablaut (gradation), zero derivation and affixation. In the remainder of this paper I describe geas the most frequent and universal affix in Old English (section 3) and examine contrastive geas the only formative principle in an alternation (section 4) or as one of the formative principles that motivate a given alternation (section 5).

DESCRIPTION OF GE-
Recapitulating, the gealternation links two morphologically related predicates between which there exists a functional contrast (involving meaning and/or category). In identifying an alternation I am not making a claim of productivity in the synchrony, neither am I assuming a unidirectional derivation. I have just remarked that the morphological processes that motivate the alternation have not taken place simultaneously: Ablaut, for instance, takes place in Germanic already and clearly predates productive affixation, leave alone productive compounding in Old English. 5 Regarding unidirectionality, the notion of alternation implies a choice within a derivational paradigm, which can be unidirectional or multidirectional. A unidirectional derivation can be described by means of a single derivational chain, as in (3a) and (3b). A multidirectional derivation cannot be described in terms of a single derivational chain, as in hand 'hand'>gehendnes 'proximity'/handlung 'handling'in (3c) and (3d) Considering (3c) and (3d), it is worth remarking that Ablaut and ge-affixation represent the formative principles which motivate the two directions of the derivation, hand>gehendnes vs. hand>handlung. In more general terms, (3c) and (3d), being instances of multidirectional derivations, constitute substantial evidence against any claim of unidirectional derivations in Old English. 6 It is not surprising that gehas received so much attention in English historical linguistics, given that it is the most generalised affix, both in quantitative and qualitative terms. I focus on the quantitative aspects of the prefix first.
Geis prefixed to 1,270 predicates, of which 532 are nouns, 359 adjectives, 84 adverbs, 293 verbs, and 2 belong to minor grammatical categories. No adposition is preceded by ge. If we consider the category Noun, 92 predicates are masculine, 193 feminine and 185 neuter; the rest are ambiguous with respect to gender. Within the category Verb, 15 predicates belong to the strong classes, and 278 to the weak classes. An illustration of each predicate category with gefollows in (4) The number of predicates to which the affix gecan be prefixed, that is, those predicates that are preceded by (+/-) in Clark Hall's (1996) dictionary, is 1,346, out of which 345 are nouns, 109 adjectives, 47 adverbs, 843 verbs, and 2 belong to minor grammatical categories. 7 No adposition has been found in this group, either. As for the category Noun, 79 are masculine, 169 feminine and 60 neuter. The rest are ambiguous for gender. If we focus on the category Verb, it turns out that 3 are irregular verbs, 162 belong to the strong classes and 689 to the weak class. An illustration of each category is given in (5) These figures have no rival among Old English affixes: no other affix can be attached to such a high number of predicates (up to 2,616); neither can any affix distribute so freely across categories (all except adpositions). The affix acan be attached to 928 predicates, all of which are verbs (Martín Arista and González Torres 2005), that is, geoutnumbers ain frequency and distribution. These figures also tell us that geis mainly but not exclusively a verbal prefix: it is prefixed to 21.5% of verbs, but also to 8.1% of adjectives and adverbs, 5.3% of nouns and 3.4% of the members of minor grammatical categories. Variation also revolves around verbs: 15% of verbs can take the prefix gewithout change of meaning. This is probably the reason why gehas been considered inflective. The inflectional character of gemust be ruled out on account of the stability of the affix throughout the derivation and the lack of generality of ge-affixation. I deal with these questions in turn.
As an argument in favour of the derivative character of ge-, I advance the stability of this affix all the way down the derivation. Consider the derivatives of geno:g 'enough' in (6): (6) geno:g 'enough': genehe 'sufficiently', genugan 'to suffice', geno:g 'sufficiently', geno:gian 'to be abundant', genyht 'abundance', genyhtful 'abundant', genyhtli:ce 'abundantly', genyhtsum 'satisfied', genyhtsumian 'to suffice', genyhtsumnes 'abundance', genyhtsumung 'abundance' It is certainly an argument in favour of the derivative character of gethat it is kept throughout the derivation all the paradigm down. This does not mean that derivative paradigms like the ones in (7) do not exist. The variation of ge-, however, is the result of the alternation ±ge in the basic predicate, namely bindan/gebindan, dre:fan/gedrefan and hlystan/gehlystan: And as an argument against the inflectional character of ge-, I raise the fact that the affixation of geis not a generalised process, even in verbs. Geis not affixed to every member of the class Verb, neither does it cause the same meaning change, whereas inflectional affixes can combine with all the members of a given category conveying the same meaning (Bybee 1985). Consider example (8). The affix gechanges valency in pairs like (8a)-(8d). However, there follow several instances where the affix does not motivate a change in the verbal valency. In (8e)-(8h) the two members of the pair take one internal argument, whereas in (8i)-(8l) the two members of the pair take two internal arguments: a. ri:nan 'to rain'~geri:nan 'to wet with rain' b. sadian 'to be sated'~gesadian 'to satiate' c. stincan 'to emit a smell'~gestincan 'to smell' d. ∂earfian 'to be in need'~ge∂earfian 'to impose necessity' e. ce:lan 'to cool'~gece:lan 'to quench thirst' f. ceorran 'to creak'~geceorran 'to turn' g. cuman 'to come'~gecuman 'to come together' h. hrisian 'shake'~gehrisian 'shake together' i. biddan 'to ask'~gebiddan 'to beg' j. fricgan 'to ask'~gefricgan 'to learn' k. gryndan 'to set'~gegryndan 'to found' l. sle:an 'to strike'~gesle:an 'to strike down' To round off the question of derivation versus inflection in ge-, it is probably worth commenting on the fact that it is hard to find instances of gender change that can be attributed to gelike timbru 'building' (femimine)~getimbre 'building' (neuter). Instances like the ones given under (9), on the other hand, are more frequent: Although textual work is probably needed in this area, it does not seem completely out of place to consider gea marker of nominal aspect (in the sense of Van Valin and LaPolla 1997), thus rendering countable nouns as uncountable or collective. No further evidence of inflectional gehas been found throughout the research reported here.
For the reasons just given, I consider gea derivational affix and identify gealternations in pairs such as the ones that follow in (10): 8 (10) a. be:or 'beer'~gebe:or 'pot-companion' b. byrd 'burden'~gebyrd 'burdened' c. stincan 'stink'~gestincan 'to smell' As I have already pointed out, gecan be prefixed to all categories except adpositions. It is prefixed to basic predicates in instances like those in (11): (11) a. hlid 'lid'~gehlid 'roof' b. biddan 'to ask'~gebiddan 'to beg' c. lang 'long'~gelang 'dependent on' Gecan also be prefixed to predicates already derived by transparent derivative means such as the ones in (12) I will refer to geby itself as the pure alternation and to gein combination as the mixed alternation. I devote the next two sections to the two alternations of ge-I have just distinguished.

THE PURE ALTERNATION
The pure alternation holds when geis contrastive in form (presence versus absence) and in meaning (more or less specialised meaning) without overlapping with Ablaut or affixation. The pure alternation can relate members of the same or different categories. The meaning contrast most often dealt with in the literature (Lenz 1886;Lorz 1908 Regardless of whether these pairs are explained in terms of Aktionsart or syntactic transitivity, the affixation of geto one of the members of the alternation has the effect of changing the valency or the verb from one internal argument to two internal arguments. I have already focused on this question, as well as on the exceptions to this generalisation in example (8) The pure alternation of gecan also be identified in adjectival pairs like the ones that follow in (18) Less frequently, the patterns involved in inter-categorial pure gealternation are adjective~ge-noun, as is illustated by (21a)-(21e), and adjective~ge-adverb, as is illustrated by (21f)-(21g): (21) a. fe:re 'able to go'~gefe:re 'company' b. filde 'of the nature of a plain'~gefilde 'plain' c. ne:ah 'near'~gene:ah 'abundance' d. sme:ah 'sagacious'~gesme:ah 'intrigue' e. weald 'powerful'~geweald 'power' f. hra∂e 'quick'~gehra∂e 'hastily' g. hwae∂er 'whether'~gehwae∂er 'which of two'

THE MIXED ALTERNATION
I have devoted section 4 to the pure alternation of ge-. This section gathers evidence of gein combination with other formative principles, including Ablaut, zero derivation and affixation. Beginning with Ablaut, the literature has paid attention to gradation mainly in deverbal nouns (Palmgren 1904;Kastovsky 1968). 10 In (22), the ge-alternation and Ablaut overlap in denominal, deadjectival and deverbal nouns, respectively: Before producing the evidence of geand zero derivation, it is necessary to make a terminological remark. For some authors (including Kastovsky 1968), what I have termed the pure alternation would constitute incontestable evidence of zero derivation. In order to maximize the difference between the pure and the mixed alternation, though, I have preferred a more restricted view of zero derivation, in terms of which this morphological process implies the presence of an explicit inflectional ending, as in co:c 'cook'~geco:cnian 'to season food'. Under less strict approaches to zero derivation, instances of the pure alternation byrd 'burde'~gebyrd 'burdened', would be included under this heading. In (25) there follow some instances of zero derived nouns, adjectives and verbs, respectively (note that clibbor and clibs are graded forms): Whereas zero derivation and the ge-alternation co-occur in a significant number of verbs (certainly more than nouns and verbs), it is interesting that affixation and the gealternation in verbs is restricted to these three instances only: a. eftgian 'to repeat'~geeftgian 'to restore' b. el∂e:odgian 'to live abroad'~geel∂e:odgian 'to make strange' c. o:nettan 'to hurry'~geo:nettan 'to get quickly' The ge-alternation does not co-occur with affixation in nouns very often. The nominal affixes found in the mixed alternation, by order of frequency, are -nes,ung, -end, -en and -scipe, as is shown, respectively, by (27a) The adjective is the category in which the ge-alternation co-occurs with affixation most often. The suffix -ed clearly outnumbers the other four affixes (lic, -en, -ende, and -sum). The only graded forms are ni:edan~geny:denlic:  (26), (27) and (28) are also telling us that whereas the three instances of verbs in (26) are prefixal derivatives, the nouns and adjectives rendered, respectively, by examples (27) and (28) are suffixal derivatives. Another lesson that can be learned from examples (26)-(28) has to do with the affixes that do not co-occur with the ge-alternation. Considering verbal derivatives, the ge-alternation never co-occurs with the verbal prefixal quantifiers ofer-(as in oferspercan 'speak too much'); neither does it co-occur with the telic prefixes for-, ofand to-(as in forcuman 'destroy', ofni:htan 'to gore to death' and tobla:wan 'to blow to pieces'). The number of co-occurrences with other telic verbal affixes like be-, and on-(illustrated, respectively by gebe:agian 'to crown' and geonli:cian 'to make like') is extremely low. In this line, two-thirds of edprefixed verbs can take ge-, but the affix is associated with the telic meaning (as in geedbyrdan 'to regenerate') rather than with the iterative one (as in edreccan 'to chew'). Tentatively, the meaning of verbal gewill have to be found in areas near repeatedness and away from quantification. Regarding telicity, the evidence turned up by affix combination seems inconclusive. Although the aim of this paper is not to draw conclusions regarding the meaning of ge-, I concur with Lindemann's (1970: 63) statement that geconveys iterative Aktionsart. On the other hand, I diverge from Lindemann's (1970: 64) conclusions on the question of monosemy: even though the functions of verbal gecould be reduced to a single descriptive label, I have noted the expression of nominal aspect in pairs like ba:n 'bone'~geba:n 'bones'.

CONCLUSION
I have established two aims for this research: to offer a descriptive account of the ge-alternation that goes beyond the bounds of an only category and to test the descriptive adequacy of the lexical database of Old English derivational morphology Nerthus on qualitative and quantitative grounds. To conclude, I report the contributions of my analysis in these respects.
Both quantitatively and qualitatively, geis the most widespread affix in Old English. On quantitative grounds, it can be affixed to nearly nine per cent of the predicates of the language. On qualitative grounds, it distributes over all lexical categories, except the Adposition. Considering the morphological processes in which it takes place, gepartakes in recursive and non-recursive derivation, as well as in intra-categorial and inter-categorial derivation. Moreover, the gealternation co-occurs with Ablaut, zero derivation and affixation.
As regards the descriptive adequacy of Nerthus, several remarks are in point. By launching simple searches on Nerthus, I have been able to draw exhaustive evidence of ge-affixation and the mixed alternation, both with Ablaut and zero derivation. By launching combined searches on Nerthus, I have got exhaustive evidence of the mixed alternation with affixation, too. I have also gathered exhaustive evidence of the pure alternation by means of a simple search, although I have offered a sample only for reasons of space. The amount of qualitative and quantitative evidence of gedrawn from Nerthus has much to do with the organisation of its fields: being a database of derivation, Nerthus has paid special attention to the affix ge-, given its frequency and distribution. All predicates are marked +ge, -ge or ±ge, which contributes to maximize the results of combined searches. At the same time, Nerthus requires more formalised meaning definitions of Old English lexical items so that ±ge predicates can be properly broken down.