Identifying, Categorising and Exploring ‘Ælfrician’ Vocabulary Using the Dictionary of Old English, A Thesaurus of Old English and Evoke


Ælfric of Eynsham (c.955×957–c.1010) is one of the most prominent authors of the Anglo-Saxon period. Despite this fact, there has not yet been an exhaustive study into his typical vocabulary. This article employs the Dictionary of Old English and prior scholarship in order to collect and categorise the lexis that is characteristic for his works. This vocabulary is then analysed using the web application Evoke together with A Thesaurus of Old English, which provides insights into the semantic domains that predominate in Ælfric’s vocabulary, as well as the degrees of ambiguity, synonymy and specificity of his typical lexis.


Introduction
AElfric of Eynsham (c.955×957-c.1010) is arguably the best-known and most prolific writer of Anglo-Saxon England (Hill, 2009: 36-37). AElfric's significance for the history of the English language stretches beyond the Norman Conquest, since his works were copied until the early thirteenth century (Treharne, 2009: 400). Aside from his own works being copied in the centuries after the Conquest, AElfric's influence on new compositions made in this period is also occasionally cited. For instance, one study by Elaine Treharne, which focuses on the twelfth-century English translation of Ralph d'Escures' homily on the Assumption of the Virgin Mary, notes that " [m]uch of the vocabulary [of the text] is AElfrician in nature, so that relatively rare words like 'wiðmeten,' 'bearneac[n]inde,'1 and 'earplaettigen' appear to be based on a thorough knowledge of AElfrician prose" (Treharne, 2006: 185, n. 26). The three lexemes that Treharne provides as examples of 'AElfrician' vocabulary differ in some important aspects. According to the Dictionary of Old English (DOE), the adjective bearnēacniende 'big with child' and the verb ēarplaett(i)an 'to strike on the ear' are quite rare in the Dictionary of Old English Corpus (DOEC): bearnēacniende occurs only three times in the corpus, twice in the works of AElfric and once in Ralph d'Escures' homily (DOE, s.v. bearn-ēacniende), while ēarplaett(i)an occurs twice in the DOEC, once in the works of AElfric and once in d'Escures' homily (DOE, s.v. ēar-plaettan, ēar-plaettian).2 By contrast, a search in the DOEC for all instances of the verb wiðmetan 'to compare' reveals that 20 out of its 56 occurrences are found in the works of AElfric, and that this lemma is found in more than 20 distinct texts in the corpus. The discrepancies between these lemmata, which have all been termed 'AElfrician' by Treharne, raise some important questions. First of all, if a lemma is found only in the works of AElfric and one other text, but occurs rarely, can this lemma really be considered characteristic of AElfric's lexis? Similarly, can a lexical item that is found in more than twenty distinct texts also be labelled as 'AElfrician'? Indeed, how can 'AElfrician' lexemes such as bearnēacniende and wiðmetan be compared to each other? Is it possible to make a classification system that can differentiate between lexemes which are either more or less typical for AElfric's vocabulary?
In order to answer these questions, a general overview of AElfric's characteristic vocabulary would be helpful. To my knowledge, such a large-scale study has not yet appeared, although there are some smaller studies in which some tendencies in his lexis are highlighted (e.g., Pope, 1967: 99-103;. In order to fill this lacuna, this article demonstrates how this prior scholarship and the DOE can be used to collect and categorise vocabulary that 1 The lack of the medial <n> in Treharne's quotation is most likely a typo, since the word appears as "bearneacninde" in Rubie D.-N. Warner's edition of the manuscript, London, British Library, Cotton Vespasian D. xiv (Warner, 1917: 138, l. 6), and also appears as such on fol. 156r of the manuscript: see http://www.bl.uk/manuscripts/Viewer.aspx? ref=cotton_ms_vespasian_d_xiv_f156r. 2 The related verb ge·ēarplaettan has one occurrence in the corpus, in the works of AElfric (DOE, s.v. ge·ēar-plaettan).
has been identified as being characteristic of AElfric's writings. Next, the article shows how the web application Evoke (Stolk, 2018) may be used to further explore this AElfrician vocabulary. Section 2 will address the collection and the categorisation of the AElfrician lemmata, as well as discuss a number of issues relating to the use of the DOE for studies of this type. Subsequently, section 3 will focus on using A Thesaurus of Old English (TOE) and Evoke to explore AElfrician lexis. I will discuss the process of tagging the AElfrician vocabulary in Evoke and the issues that were encountered during this process, the tendencies which characterise AElfric's typical lexis, and a number of categories in TOE in which AElfric's vocabulary is over-and underrepresented. In the conclusion, some further possible avenues of research into AElfric's vocabulary will be pointed out. A full overview of the AElfrician vocabulary established on the basis of the DOE and prior scholarship is provided in Appendices A and B.

Identifying and Categorising AElfrician Vocabulary
The label 'AElfrician' is not one which was used in AElfric's own day. Rather, it is a term that will be employed in this article to refer to vocabulary which prior scholarship and the DOE have identified as being restricted or predominantly found in AElfric's works, or lexical items that were preferred by him over synonymous lexemes. When the term 'AElfrician' is used to refer to vocabulary that is primarily found in or restricted to the works of AElfric, it is quite likely that his contemporaries, whose works have simply not come down to us, may have used the same words. Since the corpus of Old English texts is incomplete and AElfric's works are overrepresented in this corpus, especially so in particular text genres, such as grammars, the label 'AElfrician' is simply used in relation to the texts that we have left (see also section 2.4).

Sources: DOE and Prior Small-Scale Studies
The DOE is the most important source for any study dealing with Old English lexis. In addition to listing senses of lemmata, the DOE also provides citations for these senses, and occasionally provides information on the usage of particular lemmata, for instance, when they are found frequently in the works of AElfric. For this reason, the DOE was consulted first in order to find lemmata which have been labelled by this dictionary as AElfrician. These words can be identified in the DOE by the information that the entries provide following the number of occurrences of a lemma. For instance, the entry for the lemma antimber 'material, substance' mentions the following about the occurrence of the lexeme in the DOEC: "ca. 45 occ. (freq. in AElfric)" (DOE, s.v. an-timber). The DOE uses a number of different labels for AElfrician vocabulary. For instance, the lemma anmōdlīce 'resolutely' has six occurrences "in AElfric", meaning that it is wholly restricted to the works of AElfric (DOE, s.v. an-mōdlīce). Two other labels which are often encountered are "mainly in AElfric" (see, e.g., DOE, s.v. ǣfnung) and "freq. in AElfric" (see, e.g., DOE, s.v. ed-wist), which are applied to lemmata that also occur outside of AElfric's works.3 Searching for 'aelfric' in the "Occurrence" field identifies all of the lemmata which have been labelled as AElfrician in the DOE. In addition, searches were performed for the short titles of AElfrician works, such as 'AECHom' , 'AELS' , and 'AEGram' in the same field, since lexemes that primarily or exclusively occur in these works can also be seen as part of AElfric's vocabulary as a whole.4 These lexemes employ labels similar to those mentioned above, such as "in AEGram" (see, e.g., DOE, s.v. āxiendlic) or "mainly in AEGram" (see, e.g., DOE, s.v. dǣdlic).
The current edition of the DOE only goes up to the letter I. In order to complement the data from the DOE with information about lemmata beyond the letter I, a literature review was also conducted, which has aimed to include as many sources as possible that mention lexemes seen as characteristic for AElfric.5 Through a number of small-scale studies, previous scholarship has established that AElfric exhibits a consistent lexical usage which is characteristic of his works. The first to note AElfric's preferred usage of certain lexical items over synonymous lemmata was Dietrich (1855: 544-545, fn. 140). Since Dietrich's article, there have been many studies which have mentioned similar preferences, as well as the restriction of particular lemmata to the works of AElfric; prominent studies include those by , Pope (1967: 99-103), Godden (1980), .6 Another important facet of research into AElfric's vocabulary relates to his usage of the 'Winchester vocabulary'a particular lexical usage associated with the school of AElfric's teacher AEthelwold -of which "AElfric is considered [the] most prominent and most consistent proponent" (Gretsch, 2009: 125).7 3 It is not immediately clear what the difference is between these labels; see section 2.2. 4 I also searched for the abbreviations of the biblical books that were translated, either wholly or in part, by AElfric, which resulted in four lemmata that primarily occur in the prose translation of the book of Genesis (search term: 'gen'; I also searched for 'num' , 'josh' and 'judg'). In order to identify which parts of Genesis (and the other books of the Heptateuch) were translated by AElfric, I relied on Kleist (2019: 132-135) and the citations given there. 5 Due to the scope of this endeavour, these claims were not verified in the DOEC. The grammatical terms that occur in AElfric's Grammar form an exception to this rule, since their relevancy has been determined by checking their occurrences in the DOEC. See section 2.2 below. 6 For a summary of the earliest research into AElfric's vocabulary, see Ono (1988: 75-84). 7 Cf. Gretsch (2001: 47) and Hofstetter (1987: 58). For an analysis of AElfric's usage of the Winchester vocabulary, see Hofstetter (1987: 38-66) and see also Gretsch (2001: 47-54) for possible additions to the vocabulary items mentioned by Hofstetter.
An attempt has been made to include as many studies as seemed relevant, but sources that have been shown to be problematic in later literature have been avoided.8 In consulting the sources, the focus was solely on AElfric's lexical usage, which includes lexemes that are restricted to his works, preferred lexemes, and the use of meanings which are particular to AElfric. In other words, data such as AElfric's preferential use of the verb bedǣlan with an object in the genitive, rather than the dative (Jost, 1950: 122), have not been included. Features of AElfric's vocabulary that either I or the relevant source deem of questionable relevance, such as AElfric's preference of swā swā over a single swā (Pope, 1967: 102-103), have also not been cited. It should be stressed that there is a vast amount of literature on the peculiarities and tendencies in the vocabulary of AElfric of Eynsham. Although my study is not exhaustive, I believe that I have gathered the most important sources on AElfric's vocabulary.

2.2
Categorisation Although the DOE employs a number of different labels for AElfrician vocabulary, it is not immediately clear how labels such as "mainly in AElfric" and "freq. in AElfric" differ from each other, nor how these labels differ from other, less frequently used labels, such as "disproportionately freq. in AElfric" (see, e.g., DOE, s.v. cyre). For this reason, it was deemed necessary to create a categorisation which could be used to create a distinction between the lexemes which are more strongly associated with AElfric and the lexemes which may be less characteristic of his works. In this categorisation system, an AElfrician lemma is assigned to one of four categories, A-D, based on the number of non-AElfrician texts in which the lemma occurs. The reasoning behind this system is that a higher number of non-AElfrician texts implies that a lemma is less exclusive to the works of AElfric and, for this reason, may be less characteristic of his lexis. The four categories are given below: -Category A contains lexemes which exclusively occur in the works of AElfric, e.g., bedūfan 'to sink' (DOE, s.v. be-dūfan). -Category B contains lexemes which occur in the works of AElfric and one other text, e.g., hremman 'to hinder' (DOE, s.v. hremman). -Category C contains lexemes which occur in the works of AElfric and between two and four other texts, e.g., flǣsclicnes 'incarnate condition; incarnation (of Christ)' (DOE, s.v. flǣsclicnes). -Category D contains lexemes which are frequently found in the works of AElfric and occur in five or more other texts, e.g., aeþelboren 'of noble birth' (DOE, s.v. aeþel-boren).
In order to make a more detailed distinction between more and less characteristically AElfrician vocabulary, categories A-C each have two subcategories, which relate to a lemma's total number of occurrences in the DOEC:9 -Category 1 contains lexemes which occur five or more times in the DOEC.
-Category 2 contains lexemes which occur fewer than five times in the DOEC.10 If a lemma is rare even in the works of AElfric, it may be argued that this lemma is less characteristic of his vocabulary, and of limited relevance for the identification of typically AElfrician lexis. The four categories listed above, in combination with the subcategories used for categories A-C, facilitate the use of a convenient shorthand. A lemma such as dydrung 'delusion' may be referred to as a 'B1 lemma' , which indicates that it occurs in the works of AElfric and only one other text, and has at least five occurrences in total in the DOEC (DOE, s.v. dydrung).
In order to categorise the lemmata retrieved from the DOE using the categorisation given above, it is necessary to be able to identify AElfrician and non-AElfrician texts. The works of AElfric are indicated as such in the DOE by either their Cameron number (B1) or otherwise the prefix 'AE' (e.g., 'AELS' , which refers to AElfric's Lives of Saints).11 In addition, I have relied on the work of Aaron J. Kleist in order to determine which parts of the Heptateuch were translated by AElfric, and to identify any other texts which are believed to have been written by AElfric, but which have not been categorised as such by the DOE (Kleist, 2019: 66-206).12 Counting the number of non-AElfrician texts in which a lemma occurs was carried out based on the texts cited in the entry of a lemma in the DOE. Whenever these texts numbered fewer than five and not all occurrences of the lemma were given in the entry, I also consulted the DOEC in order to check for any other texts, whenever this was reasonably possible. Determining whether similar texts, such as manuscript variants, should be considered different texts is always a complicated task. Whenever possible, the DOE entries have been followed. For instance, if different texts are given 9 This information is provided in every DOE entry. 10 This subcategorisation is not possible for category D, since the lemmata in this category occur in at least five non-AElfrician texts, which means that they have at least five occurrences (not counting their occurrences in the works of AElfric).

11
Notable exceptions are 'AEColl' (the gloss to AElfric's Colloquy), and 'AEGl 1' , 'AEGl 2' and 'AEGl 3' (the additions to his Glossary) which were not written by AElfric. 12 All of the categorised lemmata may be found in Appendix A (AElfrician vocabulary found in the DOE) and Appendix B (AElfrician vocabulary found in prior scholarship) at the end of this article. In the appendices, references to Kleist (2019) are provided in the appropriate footnotes whenever this is relevant, for instance, if the consideration of a text as AElfrician has led directly to a particular categorisation of a lemma (e.g., as C1, rather than D); in other cases, references are not provided.
in a single quote, these texts are usually counted as one single text. However, other texts have been counted separately despite their similarities; charters, for instance, may use similar formulas, but are nevertheless different texts. Composite homilies which make use of AElfrician material presented a difficult case. Sometimes the AElfrician text in the composite homily may be virtually identical to the edition of the AElfrician base text in the DOEC; at other times, the composite homily may differ from the base text in terms of word order, omissions, etc. In order to be consistent, all composite homilies containing AElfrician material have been counted as non-AElfrician texts. Works that have been identified by the DOE as having been written by the same author have been counted as a single text, such as the combined works of Wulfstan. These texts are arguably all examples of the same, idiosyncratic lexical usage of their author. Other texts that have been taken to constitute a single unit include the various versions of the glosses to Aldhelm's De laude virginitatis, such as 'AldV 1 (Goossens) C31.1' and 'AldV 13.1 (Nap) C31.13.1' , due to their similarity and the fact that they gloss the same text, and three versions of the Benedictine Rule in 'BenR B10.3.1.1' , 'BenRW B10.3.4' and 'BenRWells B10.3.3' , for the same reason. However, individual glosses to the psalter and canticles have been counted separately, since the gloss to the Vespasian Psalter is obviously not the same as the one to the Royal Psalter. If a lexeme occurs in a psalter gloss and a canticle gloss in the same manuscript, both glosses have been counted as one text, even though the two have been assigned different Cameron numbers. With respect to this policy, I believe my results would not be significantly different if I had made different choices.
The categorisation of the lemmata in prior scholarship is based, for the most part, on the secondary sources themselves; the claims made in the sources have not been checked against the DOE or DOEC. Nevertheless, the DOE and DOEC have been used for the categorisation of a number of lexemes about which very little information, e.g., in terms of their frequency, was given in the sources.13 I limited myself to those lemmata which are found in the DOE. Lemmata that do not begin with the letters A-I, which are not found in the DOE, have been placed in a separate category. Lastly, whenever sources have indicated that a specific lemma or specific lexical usage is part of the Winchester vocabulary, this has been indicated in the relevant footnotes in Appendix B.
In contrast to the other lexemes found in prior scholarship, the words found in previous research on AElfric's Grammar have been checked in the DOEC as far 13 See, for instance, gramlic and gramlīce in categories B1 and C1, respectively, in Appendix B. Consequently, some of the lemmata that are part of Appendix A also appear in Appendix B.
Identifying, Categorising and Exploring ' AElfrician ' Vocabulary Amsterdamer Beiträge zur älteren Germanistik 81 (2021) 384-441 as possible.14 It seemed preferable to categorise only those grammatical terms which had a significant number of occurrences in the works of AElfric, and reject such lexemes mentioned in the literature as nama 'noun' and word 'verb' (Chapman, 2010: 423), which are arguably quite general. My rule of thumb is as follows: if a lexeme has been determined to belong to category D and fewer than 50% of its occurrences are found in the works of AElfric (not necessarily AElfric's Grammar), then this word is not categorised. If, however, a lexeme has been determined to belong to categories A, B or C, it is always categorised, even if, for instance, only one of nine occurrences of this word is found in the works of AElfric.15 A number of words which were quite difficult to check in the DOEC (because their forms were similar to other lemmata and these forms could not easily be distinguished from each other) have been discounted. One guiding principle of the categorisation is that the DOE is followed whenever this is possible. This principle has led to some inconsistencies in the categorisation of the words in AElfric's Grammar. For instance, if a word mentioned in a secondary source is a present participle such as faestnigende 'affirmative' (Chapman, 2010: 441), and this word can only be found as part of the DOE entry for the whole verb ( faestnian), which is not an AElfric word according to the rule above, then it is not categorised. However, for words beyond the letter I, i.e., those which could not be checked in the DOE, present participles are taken as separate from their main verbs if these present participles are specifically mentioned in prior scholarship, e.g., ofcumende 'derivative' (Chapman, 2010: 443), since the DOE is not always consistent in categorising present participles or lemmata derived from present participles. 16 The secondary sources that I consulted often featured various types of information about AElfric's lexical usage, which could not easily be compared to each other. Information such as AElfric's preference of one lemma over another 14 For these words, I primarily made use of Appendix 2 in Chapman (2010: 438-445), ignoring the multiword terms on pp. 443-445; I also consulted Sauer (2009: 171) and Williams (1958: 461-462). 15 I applied the same rule to the four lemmata which are labelled by the DOE as frequently or exclusively occurring in the prose translation of Genesis. 16 For instance, the lemma healfclipiende 'semivocalic' can be found in this form in the DOE (s.v. healf-clipiende), rather than being subsumed under a reconstructed infinitive *healfclipian. Similarly, some citations in the entry for the agentive noun dǣlnimend 'participant; participle' feature forms ending in -ende in the nominative singular (DOE, s.v. dǣl-nimend); this ending indicates that they are present participles, rather than agentive nouns (which would end in -end in the nominative singular). Hence, instances of the present participle dǣlnimende seem to have been subsumed under the entry for the agentive noun dǣlnimend, rather than having been assigned to a reconstructed infinitive *dǣlniman in the DOE.
for the expression of a certain concept seemed relevant to record, but could not be categorised in categories A-D due to the lack of information pertaining to the number of non-AElfrician texts in which these preferred lemmata occurred, and their total number of occurrences in the DOEC. In order to ensure the accurate categorisation of the AElfrician vocabulary identified by prior scholarship, it was necessary to add an additional four categories to the categorisation outlined above. The following four categories were created: -Category E contains particular lexemes that AElfric prefers over synonymous lexemes. This category is further subdivided into categories E1 and E2, which relate to whether these preferences are constrained by semantic, contextual or other factors: -Category E1 features preferences which are, generally, unrelated to specific semantic or contextual usages. -Category E2 features preferences which are, generally, related to specific semantic or contextual usages, or certain other factors. One example of an entry in category E1 is AElfric's preferred usage of gefrēdan, rather than fēlan, to express the verb 'to feel' .17 This preference is independent of contextual or semantic factors. Conversely, an example of an entry in category E2 is AElfric's preference of the verb (ge)rihtlǣcan over (ge) rihtan when expressing the verb 'to correct' in figurative senses (and, conversely, the verb (ge)rihtan over (ge)rihtlǣcan in literal senses) (Hofstetter, 1987: 51).18 -Category F contains particular morphological forms of lexemes that AElfric prefers over other morphological forms. The root is the same for both preferred and dispreferred equivalents; the synonyms merely differ in terms of the other morphemes that they may contain, such as prefixes. For instance, AElfric prefers the form bebod 'command' , with the prefix be-, over gebod, with the prefix ge- (Sato, 2011: 308).19 17 Pope (1967: 99), who cites Dietrich (1855: 544-545, fn. 140). Klein (2012: 491) notes that AElfric never uses fēlan or related forms. 18 While the verb (ge)rihtlǣcan is part of the Winchester vocabulary according to Hofstetter (1987: 38), it would seem that the difference in usage between (ge)rihtlǣcan and (ge)rihtan is particular to AElfric (51). Jost (1950: 137-138) labels gerihtlǣcan as an 'AElfricwort' . 19 According to Sato (2011: 308), there are no instances of gebod in both series of AElfric's Catholic Homilies and in his Lives of Saints.
-Category G contains widely used lemmata that AElfric uses in particular contexts or with specific meanings. For example, the sense 'to bury' for the verb bestandan is primarily attested in the works of AElfric (Jost, 1950: 144).20 -Category H contains lemmata that do not fit in the preceding categories. This is where claims have been placed such as 'most instances of þwȳrlic can be found in AElfric' (Jost, 1950: 130). Since there is no information pertaining to the number of occurrences of this lemma in non-AElfrician texts, it is not possible to place it in categories A-D. At the same time, it is impossible to place þwȳrlic in categories E-G, since the source does not mention if AElfric prefers this lemma over an equivalent lemma, or if he uses it in a specific sense. Taken together, categories A-H allow for the creation of an overall characterisation of AElfric's lexical usage, featuring lemmata that are primarily or exclusively restricted to his works, preferences of particular lemmata over others, and lemmata that have semantic or contextual usages which are specifically AElfrician.

2.3
Results The results of the categorisation are presented in Tables 1 and 2.21   table 1 Results of the categorisation of AElfrician vocabulary in the DOE and prior scholarship for categories A-D

Categories Total
E1 (preferences unrelated to semantic/contextual usages) 26 E2 (preferences related to semantic/contextual usages) 8 F (preferences related to morphological forms of lexemes) 12 G (lemmata used in particular contexts or with specific meanings) 3 H (lemmata that do not fit in the other categories) 6 Total E-H 55 Out of the eight categories A-H, which contain a total of 465 items, the two largest categories are those which, respectively, contain words which are the most (category A) and the least (category D) restricted to the works of AElfric.
The vast majority of the 152 lexical items in category A can be found in category A2, which features 120 lemmata (78.95%). These AElfrician words are quite rare, occurring between one and four times in the works of AElfric. For categories B and C, this tendency is reversed: category B2 contains fewer items than B1, and there are no lexical items at all in category C2. This outcome is not surprising, since a higher number of non-AElfrician texts in which a word occurs directly correlates with a higher overall frequency of that lemma. In categories E-H, there are 55 items, of which the majority can be found in category E: 34 items (61.82%).22 Within category E, the best represented category is category E1, which contains 26 items (76.47%). This result implies that AElfric's preferences for particular lemmata over other, synonymous lemmata that are unrelated to specific semantic or contextual constraints have received the most attention in prior scholarship. Note that there is an important difference in the way the items in categories A-D and those in categories E-H have been counted. Whereas in categories A-D each lemma is counted individually, this is not the case for categories E-H; in the latter categories, the entire entry, regardless of the fact that it may contain more than one AElfrician lemma, is counted as a single unit. 22 Note that a search in the DOE for 'aelfric' in the "Definition" field yields 84 entries with senses primarily or exclusively found in the works of AElfric, which could have been placed in category G. However, in order to keep the data to be categorised at a manageable level, I decided to consult only prior scholarship for categories E-H.
For instance, the entry "ǣlc/gehwā/gehwilc 'every' preferred to ǣghwilc"23 in category E1 is counted as a single unit, despite the fact that there are three preferred lemmata. One reason for counting in this way is that the logic of having 'preferred' lemmata versus 'dispreferred' lemmata breaks down when counting entries in category E2. Recall the aforementioned example of (ge)rihtlǣcan being preferred over (ge)rihtan when expressing the verb 'to correct' in figurative senses: this fact does not entail that (ge)rihtan is a dispreferred lemma, since AElfric, conversely, prefers to use (ge)rihtan over (ge)rihtlǣcan in literal senses. In other words, the preference goes both ways. Furthermore, all lemmata in an entry, whether preferred or dispreferred, have identical or strongly related senses, which also implies that it is sensible to count them as a single unit.

Reflection on the Use of the DOE for the Collection of AElfrician Vocabulary
The collection of data on the lexis of a particular author from a dictionary such as the DOE is perhaps somewhat unorthodox when compared to such methods as consulting secondary sources or analysing a corpus of the author's works. In this section, I will briefly reflect on some of the issues that were encountered during this study.
The choices made by the DOE with respect to lemmatisation directly influence which lemmata are considered to be AElfrician in this study. Some of these lemmata are of questionable relevance. For instance, due to the policy of the DOE to create two separate entries for lemmata with and without the prefix ge-, the lemmata edcennan and ge·edcennan are lemmatised separately in the DOE. Although the A1 lemma edcennan occurs six times, only in the works of AElfric (DOE, s.v. ed-cennan), the longer form ge·edcennan occurs twice, both times in non-AElfrician texts (DOE, s.v. ge·ed-cennan). Lastly, the past participle ge·edcenned also receives an entry of its own, because it cannot be determined if this past participle belongs to edcennan or ge·edcennan. Although the past participle, which has 21 occurrences in total, does appear in AElfrician texts, it is also found in at least six non-AElfrician texts (DOE, s.v. ge·ed-cenned). It is very likely that, if all these forms had been subsumed under a single entry, e.g., (ge·) edcennan, this entry would not have been labelled as AElfrician in the DOE. In addition, the lemma edcennan is, according to its label in the DOE, primarily 23 Jost (1950: 162-166), reiterated by Pope (1967: 100). According to Jost (1950: 162-166), the first lemma, ǣlc, is used most frequently by AElfric; he also mentions that there are only three occurrences of ǣghwilc in the works of AElfric; see also Jost (1950: 162-166) for more detailed information on the usage of these lemmata.
found in a late twelfth-century manuscript (DOE, s.v. ed-cennan).24 It may be argued that a lemma which is mainly restricted to a copy written almost two centuries after AElfric's lifetime cannot be seen as characteristic of his lexical usage. Both of these factors -the lemmatisation policies of the DOE and the restriction of certain lemmata to late copies of AElfric's works -affect the way the DOE might be used as a source for AElfrician vocabulary.25 Manuscript-specific readings such as edcennan can be problematic in other ways. For instance, if two authoritative copies use different lemmata, it may be difficult to determine the 'true' AElfrician reading. A relevant example is the lemma flocc 'flock' , which has been placed in category C1, based on the fact that it occurs in four non-AElfrician texts.26 However, one of its occurrences, 'floccum' , occurs in manuscript P of AElfric's translation of the book of Judges, while manuscript Z, which is the base manuscript used by the DOE, employs a form that is based on the lemma folc, namely, 'folcum' .27 If 'folcum' is the original AElfrician reading, the occurrence of 'floccum' in manuscript P should be counted as non-AElfrician, which brings the total number of non-AElfrician texts to five, and requires this lemma to be placed in category D. Although both readings make sense in the context, a case can be made for 'floccum' being the original AElfrician reading, due to the fact that it is the more plausible variant: according to the DOE, the two instances of folc with the sense "band of men, company, division of an army" (one of which occurs in the quotation found in Judges, and the other in the D version of the Anglo-Saxon Chronicle) were "perhaps intended for flocc q.v." (DOE, s.v. folc, sense 13).28 The issue of counting an instance of a particular lemma as AElfrician if it only occurs in one or two manuscript copies is especially relevant to the works of AElfric, which often exist in multiple manuscripts.29 This factor may, therefore, also influence the use of DOE data in studies on AElfrician vocabulary. 24 The label is "in AElfric, mainly in MS of s.xii2". According to the DOE (s.v. ed-cennan), four out of its six occurrences are found in manuscript B, which refers to Oxford, Bodleian Library, Bodley 343 (see Kleist, 2019: 208). 25 For similar examples, see, for instance, ge·bōgian (A2), bōgian2 (C1); cwēmednes (A2), ge·cwēmednes (B1); edcwician, edcucian (B1), ge·edcwician, ge·edcucian (C1), ge·edcwicod, ge·edcucod (D); ge·hrepian, ge·hreppan (A2), hrepian, hreppan (D) in Appendix A.
As TOE does normally subsume variants with the prefix ge-under a single form (e.g., (ge)cwēmednes), this poses a problem for tagging the DOE entries in Evoke (see section 3.1). 26 According to the DOE (s.v. flocc), and a search in the DOEC for 'floc' . 27 Manuscript P refers to Oxford, Bodleian Library, Hatton 115; manuscript Z refers to Oxford, Bodleian Library, Laud Misc. 509 (see Kleist, 2019: 226, 236). 28 See also the entry for flocc in Appendix A, category C1. 29 See, for instance, the DOE (s.v. for-scyldig); this adjective occurs in some copies of AElfrician texts, whereas other copies instead have the adjective scyldig or the verb forscyldigian.
Lastly, there were a number of more general issues with this study. As has been mentioned above, AElfric was a prolific writer whose works have been well-preserved. This fact is borne out by his presence in the DOEC, in which his works may be said to be overrepresented. The works identified as having been written by AElfric constitute 22.66% of the prose corpus (B) and 15.91% of the entire DOEC -these percentages would be even higher if the word counts of the AElfrician parts of the Heptateuch were included.30 In other words, there may be said to be a higher-than-average chance of a lexeme being found exclusively in the works of AElfric.
An issue that is related to this overrepresentation is that some lemmata which, according to the DOEC, are found exclusively in the works of AElfric were possibly used by other writers as well. For instance, the A1 adverb cēnlīce 'boldly' , which has eight occurrences (DOE, s.v. cēnlīce), is derived from the more common adjective cēne 'bold' , which occurs around fifty times in a number of different, mainly poetic, Old English texts (DOE, s.v. cēne). The higher frequency of occurrence of cēne, coupled with the transparent derivation of cēnlīce, makes it plausible to believe that this adverb must also have been used by other authors, whose texts have now simply been lost to us.31 A similar example is provided by the A1 adjective hārwenge 'grey-haired' , which occurs six times in AElfrician texts (DOE, s.v. hār-wenge). The existence of a derived noun hārwengnes 'greybeardedness' , which occurs only once in a non-AElfrician glossary (DOE, s.v. hārwengnes), seems to imply that the adjective must have been more common than the corpus shows. If exclusively AElfrician lemmata, such as cēnlīce and hārwenge, have strongly related lemmata which are not restricted to the works of AElfric, then this factor may reduce the significance of these AElfrician lemmata for studies into AElfric's vocabulary.
One final point is that the DOE has not consistently labelled words that primarily or exclusively occur in the works of AElfric. The noun alēfednes 'infirmity' , for instance, has only one occurrence in the corpus, in AElfric's works, but it does not receive a specific label in the DOE. This lack of labelling implies that there are still more AElfrician lemmata to be found in the DOE, which may perhaps be labelled in future editions of the dictionary.32 30 Calculated based on the word counts of 'OE words' given by the DOEC; see https://tapor .library.utoronto.ca/doecorpus/wordcount.html. 31 Cf. the equivalent adverb in Present Day English, keenly, which is, of course, not restricted to any particular author. 32 Lemmata that are in some way restricted to or frequently occur in the works of AElfric, as indicated by prior scholarship or by analysing their occurrences in the DOE or DOEC myself, but which have not been labelled as such by the DOE, are preceded by a plus symbol in the appendices. I have not systematically searched the DOE for these non-labelled, AElfrician lemmata, but I did think it was relevant to include those I had found.

Analysing AElfrician Vocabulary in Evoke
3.1 Methodology In order to discover the characteristics of the AElfrician vocabulary that was categorised in the previous section, the lemmata were tagged in Evoke. Since Evoke uses a Linguistic Linked Data version of TOE,33 which lemmatises differently from the DOE, a number of choices had to be made in order to tag the words found in the DOE and prior scholarship in Evoke. These choices will be outlined in this section. More specific information about the tagging of individual lemmata can be found in the footnotes in the appendices. For the purpose of tagging the AElfrician vocabulary in Evoke, only categories A-D have been taken into account, since the lemmata in these categories form a cohesive unit in that they are either restricted to or occur frequently in the works of AElfric. The lemmata in categories E-H are more difficult to quantify in this sense. For instance, AElfric may prefer the verb gefrēdan 'to feel' to its synonym fēlan (see section 2.2), but this fact does not imply that the verb gefrēdan is in some way restricted to the works of AElfric; this entry in category E1 simply indicates a preference.
Each lexical entry (i.e., not the individual lexical senses) for an AElfrician lemma in Evoke receives three tags: -#AElfrician: All AElfrician lemmata receive this tag, which allows for the immediate selection of all AElfrician vocabulary in Evoke. -#AElfrician_A/#AElfrician_B/#AElfrician_C/#AElfrician_D: These tags indicate the category (A-D) to which a lemma belongs. -#freq5plus/#freq1to4: These tags indicate the subcategory (1 or 2) to which a lemma belongs. Subcategory 1 is tagged as #freq5plus (since the lemma which receives this tag has five or more occurrences in the DOEC) and subcategory 2 is tagged as #freq1to4 (the lemma has between one and four occurrences in the DOEC). In addition, a number of lemmata also receive the tag #comment, which is accompanied by a brief explanation outlining the discrepancy between the ways in which these lemmata are treated in the DOE and TOE (see below).
There were a number of issues with tagging the AElfrician vocabulary in Evoke. One issue is that some DOE lemmata do not have equivalent lemmata 33 Henceforth simply referred to as 'TOE'; cf. in Evoke, which means that they could not be tagged.34 Other issues relate to the different lemmatisation choices made by the DOE and TOE. For instance, the DOE considers words which occur with and without the prefix ge-, such as the verbs gehūslian and hūslian 'to administer the Eucharist' , as separate lemmata (DOE, s.vv. ge·hūslian, hūslian). Since the past participle gehūslod could theoretically belong to either of these verbs, it too receives its own entry (DOE, s.v. ge·hūslod). In TOE, however, these three entries correspond to a single entry: (ge)hūslian, which creates problems for categorisation, since the three DOE entries each have their own category (gehūslian is A1, hūslian is B1, and gehūslod is not in the appendices). Conversely, the opposite may be true: a single entry in the DOE may correspond to two or more entries in TOE. For instance, a search for the verb bedydrian 'to delude' in Evoke gives two results: bedydrian and bedydrian … wiþ. The second of these entries, with the sense 'to conceal' , is listed as sense 2 in the DOE (s.v. be-dydrian).
In order to employ a consistent strategy for dealing with these discrepancies, it was once again established as a main principle that the DOE is followed whenever possible (see section 2.2), since the categorisation of the AElfrician vocabulary is primarily based on the DOE. This principle led to the following solutions to the problems mentioned above: when the DOE has multiple lemmata which correspond to a single lemma in TOE, the labels for these lemmata are consolidated. In other words, the lemmata gehūslian (A1), hūslian (B1) and gehūslod (not part of the appendices) are taken as a single lemma, and their occurrences in non-AElfrician texts are combined. Therefore, the equivalent lemma (ge)hūslian has been tagged in Evoke as C1. Comments have been added to Evoke entries which subsume multiple DOE lemmata for the purposes of clarity; e.g., in the case of (ge)hūslian: "#comment Conflation of three DOE entries: gehūslian (A1), hūslian (B1) and gehūslod (not in appendices)." Conversely, when a single DOE entry corresponds to multiple TOE entries, all relevant TOE entries in Evoke are assigned the same category as the DOE entry; the DOE lemma has not been 'split up' into TOE lemmata which are then recategorised. In other words, both bedydrian and bedydrian … wiþ are tagged as C1 in Evoke; the fact that bedydrian … wiþ with the sense 'to conceal' only occurs once in total, in the works of AElfric (DOE, s.v. be-dydrian, sense 2), has not been taken into account.
Note that for entries in TOE in which a preposition is part of the lemma for a verb (such as bedydrian … wiþ), only those entries have been tagged which have demonstrably been used by AElfric, i.e., there is an AElfrician quote for this 34 E.g., fornēan 'almost' , which was not yet available in the Linguistic Linked Data version of TOE used for this article. See footnote 33. particular verb + preposition combination in the equivalent DOE entry. This decision was also made in order to reduce the number of errors with respect to tagging senses of AElfrician lemmata which do not actually occur in AElfric's texts. For instance, with respect to the D lemma abūgan, one of the four results in TOE, ābūgan fram, has not been tagged, since the sense that is attested for it in TOE, 'to move from' , seems to correspond to sense 2.b in the DOE (s.v. a-būgan), which does not list any AElfrician quotes. The same principle has been applied to other lemmata: if a TOE entry is solely associated with senses which are not found in the works of AElfric for that particular lemma, then this entry is not tagged in Evoke.35

Results
The AElfrician vocabulary which has been tagged in Evoke can be subjected to a number of statistical analyses, which highlight the similarities and differences between AElfric's vocabulary and all words in TOE as a whole. Therefore, these analyses provide insights into the characteristics of AElfric's vocabulary. Due to the discrepancies between the DOE and TOE (see section 3.1), the number of tagged entries per category in Evoke differs from the number of entries which have been categorised based on the DOE and prior scholarship, as found in Appendices A and B (see Table 3 below and cf. Table 1 above). For reasons of space and since this is an exploratory study, AElfric's lexis will be analysed as a whole in this section, without taking into account the differences between categories A-D. 35 Discrepancies between the DOE and TOE and the subsequent choices with respect to tagging in Evoke have been indicated in the footnotes in the appendices (and the corresponding entries are preceded by asterisks), but, for reasons of space, no information is given about the reasoning which underlies these choices (e.g., the sense of a TOE entry does not occur in the works of AElfric). Occasionally, tagging multiple DOE entries as one TOE entry in Evoke leads to category D words with fewer than 50% of their occurrences in the works of AElfric. These TOE entries have nevertheless been tagged, since they subsume at least one lemma which the DOE has labelled as AElfrician (the only exception being ge·ered in category A2, Appendix A). However, if a lemma found in scholarship on AElfric's Grammar corresponds to a TOE entry which subsumes another lemma (e.g., a variant with the prefix ge-) and these lemmata taken together are categorised as D with fewer than 50% of their occurrences in the works of AElfric, then this word is not categorised at all (i.e., it is not part of the appendices either). A slightly stricter approach towards the categorisation of the words found in AElfric's Grammar is warranted, since the claim in the literature (Williams, 1958;Sauer, 2009; is not that all of these words are frequently found in the works of AElfric, but rather that they simply occur in AElfric's Grammar as translations of particular Latin lemmata. First of all, Evoke can be used to determine the degree of ambiguity of AElfric's lexis. The degree of ambiguity is related to the number of senses that a lemma may have. For instance, if AElfrician words generally have a low number of different possible senses, this result would imply that AElfric's lexical usage can be characterised as unambiguous, and could mean that he is particularly concerned about writing as precisely as possible. A high degree of ambiguity would indicate the opposite: a lack of a particular concern for precision in lexical usage, and perhaps a deliberate effort to allow for multiple interpretations of his words. As Figure 1 shows, AElfric's vocabulary is somewhat more ambiguous than the vocabulary in TOE. Around two-thirds -65.95% -of the lexical entries tagged as AElfrician have only one sense associated with them, as opposed to 78.40% of the entries in TOE as a whole. Conversely, AElfric's vocabulary contains relatively more lemmata with two, three, four or five senses than TOE does. According to Evoke, an AElfrician lemma has, on average, 1.66 senses associated with it, while a lemma in TOE has 1.45 senses. While this is perhaps not a significant difference, it seems that, on average, AElfric's vocabulary is somewhat more ambiguous than Old English vocabulary in general. Nevertheless, this difference is not great enough to allow for the conclusion that AElfric was deliberately ambiguous or unconcerned with lexical precision in his works.
Evoke can also determine the degree of synonymy of AElfric's vocabulary. One crucial difference between this analysis and the previous one is that the degree of synonymy relates to lexical senses, rather than lexical entries. Since only lexical entries have been tagged in Evoke, not lexical senses, the analysis in Evoke takes into account all 693 senses that are associated with the 417 AElfrician lemmata, including those senses which have not been attested in the works of AElfric.36 The degree of synonymy is related to the number of synonyms that are available for a lexical sense. If AElfric mainly uses lexical senses with a high number of synonyms, the implication would be that AElfric, in choosing one particular synonym over other available equivalents, often made deliberate lexical choices in his writings. On the other hand, if the works of AElfric generally feature lexical senses with few synonyms, this fact would make it more difficult to argue that AElfric frequently made particular conscious lexical choices.
The graph in Figure 2 shows that a quarter of the senses associated with AElfric's vocabulary (25.69%) have zero synonyms available for them (i.e., there is only one lemma associated with these senses). This percentage is almost the same for the senses in TOE: 24.63% of all senses in TOE have zero synonyms 36 An attempt has been made to mitigate this effect somewhat: if a DOE lemma corresponds to multiple TOE entries, the TOE entries are only tagged if they are associated with senses that actually occur in the works of AElfric (see section 3.1).
figure 2 The degree of synonymy of AElfric's vocabulary (orange) and TOE (blue) available for them. The graphs for the senses of the AElfrician lemmata and the senses for the TOE lemmata are roughly equivalent. This fact is borne out by Evoke's statistical analysis, which shows that, on average, a sense of a given AElfrician lemma has 4.88 synonyms available for it, while a sense of a lemma in TOE has 4.92 synonyms associated with it. In other words, it is very likely that AElfric made deliberate lexical choices in his writings, something which is also borne out by categories E and F, which feature lemmata that AElfric prefers over their synonyms. Nevertheless, AElfric probably did not make particular lexical choices to a greater extent than was normal in Old English. Another analysis which can be carried out in Evoke relates to the degree of specificity of particular vocabulary. This analysis also uses the total number of senses associated with lexical entries to show the distribution of these senses in the taxonomy of TOE (which Evoke refers to as the 'tree depth' of these senses), with 1 being the most abstract level and 11 the most specific level in meaning. For instance, if the senses associated with AElfric's vocabulary are found, overall, at higher taxonomy levels, then AElfric's vocabulary may be said to have a high degree of specificity. This result might imply that AElfric created some of his vocabulary in order to fill gaps in the Old English lexicon. On the other hand, if many senses are found at lower taxonomy levels, this tendency would point to a preference on AElfric's part for using specific terms for more general concepts. This vocabulary is less likely to have been created by AElfric, as the chances are higher that AElfric simply made use of particular pre-existing lexemes. Figure 3 shows that the senses associated with AElfric's vocabulary follow much the same distribution in TOE's taxonomy as the distribution of all senses found in TOE. On average, a sense found in AElfric's vocabulary has a taxonomy level of 5.14, which is virtually identical to the average for all TOE senses: figure 3 The degree of specificity of AElfric's vocabulary (orange) and TOE (blue) van Baalen Amsterdamer Beiträge zur älteren Germanistik 81 (2021) 384-441 5.11. This result is to be expected, since words are seldom extremely general or extremely specific: most words fall somewhere in between these two extremes. Nevertheless, the graph does show that senses associated with AElfric's vocabulary are found relatively less often at a taxonomy level of 6 (AElfric: 20.35%; TOE: 23.81%) and relatively more often at a taxonomy level of 8 (AElfric: 5.92%; TOE: 2.95%). Although it is true that the degrees of specificity of AElfric's vocabulary and all of the vocabulary in TOE are, on average, virtually the same, it does seem that AElfric's vocabulary is somewhat more specific than Old English vocabulary in general, since it features a higher-than-expected percentage of senses at taxonomy level 8 and a lower-than-expected percentage of senses at taxonomy level 6.
Lastly, Evoke can show how the senses associated with AElfric's vocabulary are distributed across the categories in TOE, in order to establish the semantic domains in which AElfrician vocabulary predominates and the domains in which it is underrepresented.
One category that immediately stands out in Figure 4 as containing a particularly high proportion of the senses associated with AElfric's vocabulary is "09 Speech, vocal utterance": 15.73% of all the senses associated with AElfric's vocabulary are found in this category, whereas the same category contains only 3.67% of all senses in TOE. If the bars belonging to this category are clicked, Evoke shows a new graph, featuring the distribution of senses across the subcategories of "09 Speech, vocal utterance". This graph shows that the vast majority of the senses associated with AElfric's lexis in "09 Speech, vocal utterance" are found in the category "09.03 A language" (AElfric: 12.41%; TOE: 1.34%), and, more specifically, in a subcategory of "09.03 A language": "09.03.02 Art of grammar" (AElfric: 10.25%; TOE: 0.29%). This result is not surprising. AElfric's Grammar teaches Latin grammar, but was written in Old English. For the purposes of this grammar, "AElfric rendered into English practically every Latin term at least once […]. Thus AElfric used over two hundred English grammatical terms, most of which he presumably coined" (Chapman, 2010: 422). In other words, the restriction of a great amount of Old English grammatical terminology to AElfric's Grammar means that grammatical terminology is overrepresented within AElfric's characteristic vocabulary. Another category in which AElfric's vocabulary is overrepresented, albeit to a lesser extent than in the category "09 Speech, vocal utterance", is the category "16 The extrasensorial world". In this category, 9.81% of the senses associated with AElfric's vocabulary can be found, compared to 6.68% of all senses in TOE.
When clicked, the graph shows that 3.75% of the senses linked with AElfric's lexis occur in the subcategory "16.01 A divine being" (TOE: 1.98%), while 6.06% of the senses of AElfric's vocabulary are found in the subcategory "16.02 Religion" (TOE: 4.71%). The slight overrepresentation of AElfric's vocabulary in these categories is most likely related to his literary output, which, among other texts, consists of homilies, saints' lives, Bible translations and other religious texts.
Some of the categories in which AElfric's vocabulary is underrepresented include "04 Consumption of food/drink" (AElfric: 3.61%; TOE: 7.79%), "02 Creation" (AElfric: 11.26%; TOE: 14.94%), and "01 Earth, world" (AElfric: 1.01%; TOE: 3.50%). The lack of AElfrician vocabulary in the first category is perhaps not unexpected, but "02 Creation" and "01 Earth, world" seem like categories in which AElfric could have used his own characteristic vocabulary, since words relating to these domains may be found in texts such as homilies. When analysing the subcategories of "02 Creation", AElfrician vocabulary is especially lacking in the categories "02.07 A plant" (AElfric: 0.00%; TOE: 2.45%) and "02.06 Animal" (AElfric: 0.58%; TOE: 1.72%). Moreover, within the subcategories of "01 Earth, world", AElfric's vocabulary is conspicuously absent in the category "01.01 Surface of the earth" (AElfric: 0.43%; TOE: 2.60%). One possible reason for the lack of characteristically AElfrician vocabulary in these categories is that AElfric did not feel the need to choose or create his own terminology when referring to concepts in these semantic domains, but was simply happy to use lemmata which enjoyed a wider usage.

3.3
Discussion Based on the statistical analyses of AElfric's vocabulary in Evoke, it would seem that the characteristics of AElfric's lexis are not vastly different from those of Old English vocabulary in general, as found in TOE. The averages of the degrees of ambiguity (AElfric: 1.66 senses per lemma; TOE: 1.45 senses per lemma), synonymy (AElfric: 4.88 synonyms per sense; TOE: 4.92 synonyms per sense) and specificity (AElfric: average sense has taxonomy level of 5.14; TOE: average sense has taxonomy level of 5.11) show that AElfric's vocabulary exhibits the same tendencies as Old English vocabulary as a whole. Nevertheless, there are differences: the very similar averages for AElfric's vocabulary and all vocabulary in TOE with respect to their degrees of specificity obscure the fact that AElfric's vocabulary has a higher chance of having a sense at taxonomy level 8 (AElfric: 5.92%; TOE: 2.95%), and a lower chance of having a sense at taxonomy level 6 (AElfric: 20.35%; TOE: 23.81%). Furthermore, a comparison of the distributions of the lexical senses associated with AElfric's vocabulary and the senses associated with all vocabulary in TOE shows that AElfric's vocabulary is overrepresented in the category "09 Speech, vocal utterance" (AElfric: 15.73%; TOE: 3.67%) -especially in its sub-sub-category "09.03.02 Art of grammar" (AElfric: 10.25%; TOE: 0.29%) -and in the category "16 The extrasensorial world" (AElfric: 9.81%; TOE: 6.68%). The relatively higher percentage of AElfrician vocabulary with a taxonomy level of 8 is most likely due to the overrepresentation of his vocabulary in the category "09.03.02 Art of grammar". A number of the typically AElfrician words in this category, such as names of grammatical cases (e.g., wrēgendlic 'accusative' in TOE category "09.03.02.03.01.01.01|03 Case: Accusative") and names for verbal moods (e.g., āsciendlic 'interrogative' in TOE category "09.03.02.03.01.02.01|02 Mood: Interrogative"), can be found at a taxonomy level of 8. All in all, AElfric's vocabulary does exhibit a number of features of its own.
The present analysis of AElfric's vocabulary in Evoke must be seen as an exploratory study: a first step towards understanding the characteristics of the typical lexis of AElfric of Eynsham. Future refinements with respect to the data set and methodology of this study are necessary.37 Firstly, during the process of tagging the AElfrician lemmata in Evoke, a number of unlabelled lemmata were encountered in the DOE which may be added to a subsequent version of the appendices.38 In addition, the present analysis of AElfric's vocabulary 37 Note also that the comparison between AElfric's vocabulary and all Old English vocabulary in TOE involves data sets which are vastly different in terms of their sizes: AElfric's vocabulary in Evoke consists of 417 lexical entries associated with 693 senses, while TOE features 35422 lexical entries associated with 51480 senses. 38 Many of these lemmata have, nevertheless, been taken into account in the analysis in Evoke, since they have indirectly been tagged as part of TOE entries which subsume them. See, for instance, the DOE entries for gehūslod (tagged as part of (ge)hūslian), gebyrþere (tagged as part of (ge)byrþre), bytlung (tagged as part of (ge)bytlung), ? gegyht, ? gegyhte (tagged as part of (ge)gyht), ānlǣcan (tagged as part of (ge)ānlǣcan), gehremman (tagged as part of (ge)hremman), efenlǣcing, efenlǣcung (tagged as part of (ge)efenlǣcung), ge·egesod, ge·egsod (tagged as part of (ge)egesian), elþēodlīce (tagged as part of elþēod(ig) līce), ge·andwyrd (tagged as part of (ge)andwyrdan), cneordnes (tagged as part of (ge)cneordnes) and efenlǣcan (tagged as part of (ge)efenlǣcan). In addition, the adverbs stōwlīce in Evoke has, for reasons of space, not individually analysed categories A-D, which show the degree to which the AElfrician lemmata are restricted to his works. Future analyses could focus on the four categories (and subcategories 1 and 2) separately and contrast them with each other, in order to find out how they differ. For instance, almost a third (32.36%) of the senses of the lemmata in category A can be found in "09.03.02 Art of grammar", which means that much of the vocabulary that exclusively occurs in the works of AElfric is related to grammar.39 Indeed, much of AElfric's grammatical terminology is virtually restricted to his works: of the 71 senses which are associated with AElfric's vocabulary in the category "09.03.02 Art of grammar", almost two thirds (46 senses = 64.79%) are found in category A (cf. Chapman, 2010: 422). Another improvement to this study would be to tag in Evoke only those lexical senses which occur in AElfrician texts, rather than tagging lexical entries which incorporate lexical senses that do not necessarily occur in the works of AElfric. This process is time-consuming, but would improve the accuracy of the statistical analyses in Evoke, since these primarily work with lexical senses, rather than lexical entries. Lastly, future analyses should take the remaining categories (E-H) of AElfric's vocabulary into account. Contrasting AElfric's preferred lemmata in categories E and F with their dispreferred equivalents, for instance, would provide insights into the semantic differences between these lemmata.

Conclusion
This study has collected, categorised and characterised the vocabulary of AElfric of Eynsham using the DOE, secondary literature, TOE and Evoke. Although the present study does not claim to be exhaustive, it is hoped that it has demonstrated the usefulness of resources such as the DOE, TOE and Evoke for researching the particular vocabulary of Anglo-Saxon authors. While it is true that the use of the DOE for studies of this kind is not entirely without its flaws, the DOE's corpus-based approach allows for the quantification of lemmata in the works of particular authors such as AElfric, and its usefulness will increase with each new fascicle that is published. The results of the analysis of AElfric's vocabulary in Evoke are perhaps not entirely surprising: the predominance of vocabulary relating to grammar and religion may be expected for an author who is primarily known for his Grammar and religious texts, such as homilies.
There are a number of further avenues of research which may employ the corpus of AElfrician vocabulary compiled for this study. For instance, the overview of AElfrician lexis can be useful for the establishment of AElfrician authorship of anonymous texts. In addition, the corpus can be used to establish the influence of AElfric's vocabulary on later texts, such as post-Conquest works.40 Lastly, once the typical lexis of other known authors, such as Wulfstan, has been compiled, these vocabularies may be compared to each other, in order to discover the differences and similarities between AElfric's lexical usage and that of other Old English authors. For all of these avenues of research, it is hoped that this study may be a useful starting point for finding þā cǣge, ðe ðǣra worda andgit unlīcð41 of the most prolific writer of Anglo-Saxon England: AElfric of Eynsham.

Acknowledgement
I would like to thank Thijs Porck for his help during the process of collecting and categorising the data and Sander Stolk for his help with Evoke, and I am also grateful to them both for their excellent suggestions in the early stages of writing this article. I would also like to thank Lucas Gahrmann for his help with tagging the AElfrician lemmata in Evoke.

Baalen, A. van. After AElfric: Tracing the Lexical Influence of AElfric of Eynsham in Two
Twelfth-Century English Texts, MA Thesis (Leiden University Centre for Linguistics, Leiden University, 2021). Breeze, A. "AElfric's truð 'Buffoon': Old Irish druth 'Buffoon' ." Notes and Queries 42 (2) (1995), 155-157. 40 For an example of this type of research, see Van Baalen (2021), which uses an earlier version of the present article's corpus of AElfrician vocabulary to analyse two early twelfthcentury English translations -one of which is Ralph d'Escures' homily on the Assumption of the Virgin Mary, discussed by Treharne (2006) -and their possible use of AElfrician vocabulary. 41 Original quote: "staefcraeft is seo caeg, ðe ðaera boca andgit unlicð" [grammar is the key which unlocks the meaning of the books], cited from the first line of the DOEC version of AElfric's Grammar. Since andgit also means 'understanding' , the intended meaning of the modified quote is 'the key which unlocks the understanding of the words' .
This appendix features vocabulary that has been labelled by the DOE as frequently, primarily or exclusively occurring in the works of AElfric of Eynsham. In addition, the appendix also features lemmata (starting with letters A-I) which have been found in secondary sources on the vocabulary of AElfric's Grammar (Williams, 1958;Sauer, 2009;. As all of these lemmata can be found in the DOE, no direct references to these sources are given in Appendix A. The spelling of the lemmata generally follows the DOE, but hyphens in compounds have been removed for the purposes of presentation, and in order to save some space. Minor spelling differences between entries in the DOE and their corresponding entries in the Linguistic Linked Data version of TOE (henceforth: TOE) are not mentioned, but major differences are given in the footnotes. The information regarding the frequency of the lemmata in the DOEC and the labels that have been assigned to them have all been copied directly from the entries in the DOE. Whenever there are discrepancies between the DOE and TOE, e.g., when one DOE entry corresponds to multiple TOE entries, or vice versa, the relevant entries in this appendix and Appendix B are preceded by an asterisk and feature a footnote explaining the difference. Whenever relevant, these footnotes also mention the new category assigned to the TOE entry or entries (see section 3.1). Entries which could not be tagged in TOE are also preceded by an asterisk.42 Plus symbols in this appendix and Appendix B indicate lemmata which are restricted to or frequently occur in the works of AElfric -either based on my own counts or (in Appendix B) on claims made in prior scholarship -but which have not been labelled by the DOE as such; i.e., they do not have a label featuring the name 'AElfric' or an AElfrician text such as 'AEGram' .43 42 See footnotes 33 and 34. 43 I have also added plus symbols to words labelled 'gram.' (since there is no indication that this label specifically refers to AElfric's Grammar) and words with the label 'in Gen' (since the book of Genesis was only partially translated by AElfric). Note that the plus symbols do not necessarily indicate lemmata which would have been labelled by the DOE as AElfrician (see, e.g., the B2 lemma edlesende which has two occurrences in a non-AElfrician text, and only one occurrence in the works of AElfric), but they do indicate lemmata which are restricted to AElfric in some way, and which cannot be found by searching the DOE for 'aelfric' or a specific AElfrician text. According to the DOE (s.v. ge·bitt) this noun only seems to occur as part of the collocation "toþa gebitt 'gnashing of teeth'", and this is the entry which has been tagged in Evoke. 47 According to the DOE (s.v. ed-cennan), this lemma only occurs as a past participle in the collocation "beon edcenned 'to be regenerated, born again' (mainly ref. to rebirth through baptism)". Hence, only the past participle has been taken into account, which corresponds to the TOE entry (ge)edcenned. This entry subsumes two DOE entries: edcennan (A1) and ge·edcenned (not in appendices) (the verb ge·edcennan constitutes a separate lemma in both the DOE and TOE). For this reason, the entry has been tagged as D in Evoke. 48 The occurrence in the translation of Judges ( This lemma could not be found in TOE and has not been tagged. 51 This lemma corresponds to the TOE entry (ge)hūslian, which subsumes three DOE entries: ge·hūslian (A1), hūslian (B1) and ge·hūslod (not in appendices). For this reason, the entry has been tagged as C1 in Evoke. 52 The hyphen in the entry in this appendix has been retained for clarity. 53 The one occurrence which is not AElfrician can be found in the thirteenth-century Lambeth Homilies (DOE, s.v. āgen-slaga), but this seems to occur in an AElfrician quote: "Nan seolf cwale, þet is, aʒen-sclaʒa, ne cumeð to godes riche"; cf. AElfric, De duodecim abusivis: "nan sylfcwala, þaet is agenslaga, ne becymð to Godes rice"; quotes taken from MED (s.v. āʒen-slaʒa n.) and DOE (s.v. āgen-slaga), respectively. In any case, since the DOEC does not contain the Lambeth Homilies, it would be more accurate to say that this lexeme occurs only three times, all in works by AElfric. 54 This lemma could not be found in TOE and has not been tagged. 55 Tagged in Evoke as ālecgendlic word. 56 This lemma could not be found in TOE and has not been tagged. The DOE (s.v. andwyrdnes) mentions that these forms may be variants of the A1 lemma awyrdnes, and, for this reason, a comment has been added to the entry āwierdnes in Evoke. 57 Tagged in Evoke as of āscrēadian. The DOE (s.v. eall-gōd) notes that "[t]his compound may alternatively be taken as two words" and this is how it appears in TOE. 69 This lemma corresponds to the TOE entry (ge)ēarplaett(i)gan, which subsumes two DOE entries: ge·ēarplaettan (A2) and ēarplaettan, ēarplaettian (Appendix B, category B2). For this reason, the entry has been tagged as B2 in Evoke. 70 This lemma corresponds to the TOE entry (ge)erian, which subsumes three DOE entries: ge·ered (A2), erian (not in appendices) and ge·erian (not in appendices). Since neither of the main verb entries have been labelled as AElfrician by the DOE, but only the past participle ge·ered, the TOE entry has not been tagged in Evoke. 71 This lemma corresponds to three entries in TOE, fāgettan (as a verb), fāgettan (as an intransitive verb) and fāgettan mid wordum, all three of which have been tagged. 72 This lemma corresponds to the TOE entry feþerhama, which subsumes two DOE entries: fiþerhama (A2) and feþerhama (not in appendices). For this reason, the entry has been tagged as D in Evoke. 73 This lemma corresponds to the TOE entry (ge)fiþerhamod, which subsumes two DOE entries: ge·fiþerhamod (A2) and feþerhamode (not in appendices). For this reason, the entry has been tagged as B2 in Evoke. This lemma could not be found in TOE and has not been tagged. 76 This lemma corresponds to two entries in TOE, forgifendlic and forgifenlic, but only the first of these has been tagged. 77 This lemma corresponds to two entries for forscrencend in TOE, a noun and an adjective, of which only the noun has been tagged. The DOE (s.v. for-screncend) notes that, in all instances, this lemma is given as "an interpretation of the name Jacob". 78 Tagged in Evoke as gefrēdendlic (on). 79 The occurrence in the translation of Joshua