Four symbols are used as annotations of the Old English entries in the TOE. These four flags relate only to word forms, not to meanings. No phrase is flagged. The flags point to aspects of word frequency that should always be held in mind, given that the extant corpus of Old English is small and probably skewed in its representation of Anglo-Saxon vocabulary. Many Old English words occur once only. Yet quite how to identify a word as truly a hapax legomenon is problematic if its appearance in multiple manuscripts of some one work be held to constitute evidence of its greater currency. The flag o should be viewed as a warning that a particular word form is very infrequent. Conversely, not all hapax legomena may have been identified. Occasionally the reader may find the tag q attached to a word, to indicate that a form is highly dubious. Forms flagged q are of varied origins: they may reflect the inventiveness of generations of ingenious editors; or plausible forms have been abstracted from place-name evidence or from words first recorded in later English; and some are dictionary ghosts of long standing. Queried forms are avoided as far as possible. The flag p points to a form that seems to occur only in poetry. This flag assumes no value judgement but deals solely with distribution. Similarly, g points to word forms that appear generally in glossed texts or glossaries. It would have been both easier and safer not to supply these four distribution flags, but their presence should serve as a salutary reminder of the difficulties of interpretation presented by the small but diverse corpus of Anglo-Saxon words that haphazardly remain.
The decision whether or not to flag entries was long discussed. That we decided to flag may be thought foolhardy but, because of the nature of a thesaurus, in which words and word-meanings do not, as in a dictionary, have one obvious preordained place, it seemed best to give some rudimentary indication of word distribution, if only as a constant reminder that many Old English words occur infrequently. In lexicographical work generally the incidence of hapax legomena can be very high, if strict differentiation of forms is observed. For Old English the number is certainly high, even when criteria approaching those of Waldorf are observed. In the TOE the adjectival ending -lic and the adverbial -līce are taken as supporting one another, and all adverbial forms that are inflected substantival forms are generally regarded as supported and hence not hapax legomena. Adjectival participial forms, whether used as adjectives or nouns, are not flagged if related to an attested verb form. Again, where adjectives are used as nouns, these are generally taken as mutually supportive. Conceptually the forms are similar, and it would therefore be misleading, and often impracticable, to label an occasional one-off substantival use of an adjective as a hapax legomenon.
Unfortunately, many of the word senses listed in Old English dictionaries have the support of a single context, and that too often in a glossed text. These senses have been included, but where they occur for well-attested word forms, they go unflagged. Tempting though it was to discard such items, we decided that they should remain in the TOE, which is, after all, a pilot thesaurus for the HT. To have left out the evidence from glossed texts and glossaries would have meant disregarding much valuable material for the history of English lexis. In any case, until the better sifting of evidence being provided by the DOE team is complete, there can be little secure information about the frequency of the senses in which words are used. For now, therefore, we have contented ourselves with presenting a comprehensive listing of Old English word senses, and the four flags must be regarded as tentative, as minimal signposting of some of the most striking distributional features of the corpus. Inevitably there are grey areas in the application of the four flags. Some examples follow, collected together in illustration of particular difficulties thrown up by the four flags.
Forms that for one reason or another seem familiar can prove surprisingly infrequent. Thus, because the seemingly unremarkable word hāmlēas (Rid 39 7) occurs once only in Old English it is flagged as found once only, in poetry: it comes as a surprise to find that the earliest OED example of 'homeless' is dated 1615. The solitary glossary instance of wēodung (AntGl 2 (Kindschi)) predates by some 300 years the next evidence for this word, also from a glossary. Within the whole recorded history of English these are not noncewords, although they so infrequent in Old English. Attempts were made, during early checking of the categories 02.03 Mankind and 02.08 Health, to leave unflagged any forms found in later English, but it soon became apparent that the experiment was premature. To look beyond the Anglo-Saxon period, however loosely defined, to the whole later history of English would confer apparent currency on even more words than was already the case. Further, the omission of the o flag would conceal evidence that could be of use in the editing of the HT.
Inconsistent word division within edited texts often makes it difficult to be sure that a word really does occur once only. Although consultation of standard dictionaries suggests op as an appropriate label for hēahland (Exo 384), the clause Hē sette hī ofer hēah land (Deut 32 13) indicates otherwise. Two elements occurring side by side are interpreted as compound by some editors, but not by others, as is evident from hēahhleoþu (Christ A,B,C 744) and hēah hlioþo (Gen 1438). Consideration of both led to the decision that hēahhliþ should be flagged p, not op. Again, the dictionaries list three nonce adjectival compounds from the translation of the Orosius: westnorþlang (Or 1. 18 19), ēastsūþlang (Or 1. 18 19) and norþēastlang (Or 1. 19 11). Doubtless these elements are tacked together by analogy with the somewhat simpler common words westlang and ēastlang, but essentially they are as insecurely welded as many poetic compounds. In all three contexts lang can be interpreted as a separate word, with the function of adjective complement, and three hapax legomena disappear. Because these forms have long been in dictionaries, they are retained in the classification, but they are flagged q.
The ge- problem is largely circumvented by following Waldorf's lead. Thus, if a form appears both with ge- and without, it is not flagged as a nonceword. For example, the prefixed form geswin is found once only (Phoen 134), but is conflated with the simplex attested from glosses (Ald V 13.1 (Nap) 26110 swinne, 47260 swinn, Ald V 1 (Goossens) 2534 swinne, 4606 swinn and Ald V 1 (Page) 105 swinne/) under (ge)swin, which is consequently without flags. Even so, some anomalies remain. An infixed ge may similarly appear as (ge) where appropriate, although sometimes separate forms are shown, to allow through flagging extra information that would otherwise be concealed. Thus stāngefōg (Elene 1017) is flagged op, whereas stānfāh, a form that occurs twice in poetry (And 1232 stānfāge and Beo 320 stānfāh), is flagged p. Although superficially similar in form, they are after all two separate words in the dictionaries. Any speculation that stānfāh, which is generally taken as colour related, might also refer to construction in stone, is a matter for separate examination.
Words beginning with the prefix ā- are, so far as the evidence was available, included as separate words. It could be argued that in verbs the ā- prefix as often as not stands for ge-, but because of its multiple origins it has been recorded when found. Once forms assembled separately from different areas of the dictionaries come together within a group or even more narrowly under a single heading it is tempting to link them and to cast aside their flags. For example, the verbs āfāran and fāran both appear in 02.04 Body, and āmæstan and (ge)mæstan in 04.04 Farm. However, conflating them under one or other form would conceal the evidence for linking the forms and so, much against inclination, the ā- forms are not reconciled to ge- forms. When fuller evidence is available it may even turn out that some verbs altogether eschew ge- in favour of ā-, for example, fūlian. For the present, it seems best to be as informative as possible. Thus the one instance of solian (Rim 6 solaþ) is flagged as a poetic nonceword, whereas āsolian is unflagged because found once in poetry (LEProv 3 āsolaþ) and twice in prose (Prov 2 (Zupker) 2 āsolaþ and Prov 3 (Roeder) 3 āsolaþ). It is interesting to note that in all four passages the verb has (-)hwīt as subject and collocates with phrases that contain the words hāt and -cōlaþ.
The be- prefix also presents difficulties. Generally it is presented as be-, whether realized as bi-, big- or in other less common shapes. In the late Old English period it seems to have assumed some of the roles of ge-, and it too may therefore be found alongside forms in ge- and ā-. It would be interesting to know how far verbs in be- parallel simplex verbs followed by be- in the same meanings. Again, did the poets of the Exeter Book have recourse disproportionately often to be- verb forms, perhaps in preference to perfectly adequate forms in ge-? Certainly, this prefix gives rise to many noncewords in Old English. For one of these, bewīcian, because it may reflect Wheloc's mistranscription of Nowell's transcript of a Chronicle manuscript now lost, the nebulous and safer q flag is invoked.
Lexicographical convention is more apt to take together as a single word a preposed element and verb than a verb and following element. Here too word division can vary from one editor to another. Sometimes inflexions help in the assessment of forms. Thus, inflected participial forms may be flagged as infrequent, despite being well supported within the corpus, as for example uppeornendre (Bede 5. 21 476 10) of the sun. Certain prefixes alternate so frequently as to be at least equivalents if not alternates and, although so far as possible treated as distinct, they should probably be taken together or at least examined together in detailed work. Obvious pairs are in- and on- or æt- and oþ-.
Some singletons can acquire a spurious validity through accepted reconstruction. The actual forms, which are not found here because dictionary practice is on the whole followed, are less convincing. Reconstructed Anglo-Saxon spellings give us cwabba instead of the qu- of its appearance in a charter copied in the fifteenth century (Ch 762 (Birch) 1218 quabben) or weardgerēfa (not in MCOE, but see Bosworth-Toller II for L1 Th i 479, 25-28 Et guardereve, id est prepositus custodum).
Overall the identification of hapax legomena is a perilous business, and what at one time seems an isolated form may with increased knowledge be identified as an instance of some better-attested word. A moment's reflection on the alternate forms thrown up by metathesis is enough for the reconciliation of the manuscript variants ealdorbold and ealdorbotl (Bede 2. 8 122 13 aldorbold), with one form flagged o standing for both. And luckily where senses may seem to have diverged with alternate forms, there are often sufficient examples of each spelling type, as with bearhtm/breahtm, for there to be no need to flag either. Forms that for various reasons proved hard to categorize are nevertheless numerous, and some examples follow. For crypelas 'lattice', flagged og, we follow the DOE description, although tempted to identify it with the better-attested forms for 'drain', which would support the flag g. Similarly sceacga is flagged og in its use of 'hair', although it may be no more than a separate sense of a word of similar shape for 'copse'. For Waldorf unlǣde 'stray' (Ch 1445 (HarmD 18) 49 þā unlǣdan oxan) was a nonceword, but deeming it merely an extension of the adjective 'unhappy' we gave it no flag. Time and time again the problem lay in glimpsing connections and in being convinced of their likelihood: so agen 'ear' ægnan 'chaff' and egenu 'blades', apparently infrequent g forms, are mutually supportive.
Given this litany of difficulties a final random selection of words flagged o should confirm the virtual impossibility of devising firm guidelines for the categorization of Old English vocabulary ahead of the completion of the DOE. Maybe gurgullione (Lit 4.3.5 (Logeman) 46) should be regarded as a Latin word rather than a loanword? There are, after all, glosses that provide an Old English word ymel alongside the Latin lemma 'gurgulione'. Should hnor 'sneezing' (Occ Gl 36 (Gough) 8880) be og or conflated with fnora? Would it not be better to leave out styficlēah (Ch 371 (Birch 606) 3 Stificleie) altogether because it looks like a place-name? Conversely, should not place-name evidence allow the removal of the flag o from stoc (GD 1 (C) 1 12 92)? The o flag as often as not should serve to give rise to speculation and inquiry.
This flag is used of words found only in Old English poetry, which is roughly defined as the texts contained in the Anglo-Saxon Poetic Record (ASPR) volumes. Even so, some words are not easily categorized. The three āsolian forms (LEProv 3, Prov 2 (Zupker) 2 and Prov 3 (Roeder) 3) mentioned above in comparison with solian (Rim 61 solaþ, come from both verse and prose according to MCOE categorization, but the close relationship of these three Distichs of Cato collections might be held to show that āsolian is essentially a 'poetic' word. Again, weallfæsten, found once in charter bounds (Ch 339 (Birch 392) and four times in ASPR texts (Gen 1056, Ex 282, Ex 482 and PPs 79 12), may be a form that should not be flagged as restricted in usage. However, the charter bounds within which it occurs look to be sound Old English metre, and wealIfæsten is given a p flag, with some trepidation. Despite such inconsistency, the attempt has been made to use the p flag strictly, and it is therefore indicative only of distribution. Any urge to label particular senses of words as p was resisted as inappropriate in the current state of Old English lexicography. Thus words regarded traditionally as 'poetic' will be so flagged only if restricted to poetry and conversely some unremarkable words that just happen to occur in poetry only will also be so flagged. Although the Clark Hall use of the dagger to indicate poetic words is an invaluable guide, given the stricter observance of distribution attempted, quite a few words traditionally marked as poetic are without flags in the classification. The p flag is therefore an indication of distribution only, and not of any 'poetic' quality within the word flagged. Differentiation among the words restricted to poetry must wait for more detailed studies.
Many of the compounds found in poetry have p flags, and it is interesting to see them cluster, much as might be expected, in certain areas of the classification. Examples of groups especially heavily studded with p and op are: 01 Earth, world; 01.01.03.01.02 Sea/ocean; 01.02.01.01.03 Sun; 02.02.05.01.01 A grave, burial place, sepulchre; 02.03.03.03 A nation, people; 05.12.01.09.03 A voyage; 05.12.01.09.03.01 A ship, boat; 06.01.05.02.01.01 Sagacity; 06.01.08.06.02 Great fear, terror, horror; 06.02.07.06 Courage, boldness, valour; 08.01.01.03 Good feeling, joy, happiness; 08.01.03 Bad feeling, sadness; 09.03.05.01.01 Poetry; 12.01.01.04 A leader, ruler; 13.02.08 Military equipment; 15.01.03 Treasure, riches, wealth; 16.01.01.01.01 The Almighty; 16.01.02.01 Heavenly dwelling place. Less marked concentrations are found in: 01.01.02.01.01.03 Shore, bank; 02.03 Humankind; 02.03.01.01 Male person, man; 03.01.12 Brightness, light; 04.04.10.01 A ring (for finger, arm, neck); 04.05.03 A dwelling-place, abode, habitation; 04.05.03.02.02.06.01 A hall, building; 05.11.04.02.01 Sunset; 06.01.01 Thought, the faculty of thinking; 06.01.08.05.01.01.01 An exercise of power, mighty work, miracle; 07.06 Pride; 08.01.03.01 Despondency; 08.01.03.09.04 Hatred; 08.01.03.09.05 Enmity; 12 Power; 12.05.04 Strife, hostility; 13.02 War; 13.02.02 Battle; 16.01.05 Hell, lower world, abode of the dead; 18.02.07.01.01 A song, a poem to be sung or recited. Again more vague but still discernible scatterings of forms restricted to poetry are found in such groups as: 02.03.01 People; 02.06.10 Monster, strange creature; 03.01.12.02 A beam of light; 05.08 Strength; 05.10.01 Amplitude, spaciousness, vastness; 05.12.03.01 Swiftness, velocity; 08.01.02.02 Love, affection, care; 08.01.03.05.02 Anger; 08.01.03.09.03 Malevolence, malice; 12.01.01.12.04 Treachery; 13.05.01 Victory; 16.01.04 Sorcery, magic, witchcraft; 16.02.05.11 The cross (as Christian image); 17.02.05 A work/product of art. Other areas are less the concern of Anglo-Saxon poets. Category 04.01 Eat is, for example, relatively free of p flags, at least so far as food is concerned: exceptions include the group of words for 'banqueting hall' in 04.01.02, words which on the whole place more emphasis on liquid than solid intake. Category 04.01.03 Drink has clusters of poetic words. Categories 10 Possession, 11 Action and Utility and 14 Law and Order are almost entirely without poetic vocabulary. Because the vocabulary of Old English poetry has been more extensively analysed than the rest of the extant vocabulary, it is useful to see such long-recognized groups of terms arrayed among words in more general use. Certain concepts were apparently likely to spawn compounds, not all of them restricted to the Anglo-Saxon poetic corpus. It comes as a shock, on reading through 13 Peace and War, to find randbeah unflagged, whether as 'shield' or 'shield boss’: all twenty-four instances occur outside the poetic corpus.
Some words that occur outside the poetic corpus only as proper names are flagged p, for example wīnburh (Dan 57, Dan 620, And 1636, 1672, Jul 78, Wid 75; Vain 9, PPs 79.12, Instr 44; compare Ch 911 (Kem 714) 3, the etymon for the place name Wimbourne) or heaþolāc (Beo 583, 1970; compare Bede 5. 22 478 12, a bishop's proper name). So too æscstede (Vain 9) is flagged op, even though formally it has possible support in the charters (Ch 274 (Birch 392) 2, Ch 619 (Birch 982) 15, 18, Ch 766 (Searle) 2 282) with to-day's place-name Ashstead. In such cases it seems more helpful than not to supply flags.
Again, there are words that appear mutually supportive, but are nevertheless flagged op: gemǣl (And 1330), for example, looks like gemǣled (Jul 590; and compare ClGI 1 (Stryker) 3723 and ClGl 3 (Quinn) 20630). The close similarity of certain compounds that occur only once suggests a dedicated seeking for variety on the part of the poets: heorucumbol (El 105) and herecombol (El 24); heresyrc (Beo 1506) and heorusyrc (Beo 2538 hiorosercean); heoruwǣpen (Jud 261) and herewǣpen (PsFr 34 3). When such forms are drawn together conceptually according to their long accustomed meanings, it is only too tempting to redefine some of them. On the whole it is a temptation firmly resisted. The classification is a summary of past work; and it would be wrong to tamper unnecessarily with the evidence for meaning provided in the standard dictionaries.
Meaning is especially a problem with poetic compounds. Because of their nature and infrequency they are more like phrases than single words. We have endeavoured to place poetic compounds in an appropriate lexical set, although they may not be placed to the taste of all readers of Old English poetry. The meanings drawn from the standard dictionaries must serve as a starting-point for reconsideration. Thus, campwudu (El 46) is entered under 'shield' when perhaps it would be better under 'spear'. Both hlence (Ex 211) and wælhlence (Ex 172 and E1 23) have been placed under 'spear' instead of 'coat of mail'. Notoriously, poetic compounds defeat exact analysis; yet the positioning of, say, ealuscerwen (Beo 767) and meoduscerwen (And 1526) according to their overall meaning rather than their constituent parts may seem to some a bold step. To have done otherwise would have led to an unwieldy enlargement of the TOE and possibly also to the incursion of poetic words into unlikely areas of the whole. These compounds add spice to the words grouped under 06.01.08.06.02 Great fear, terror, horror within Category 06 Mental Faculties as well as to those under 08.01.03.06.02 Adversity, affliction within Category 08 Emotion; other possible homes might be: 'pouring out', 'provision of drink', 'generosity' (ironic); or 'deprivation', 'deprivation of drink'; or under 'beer' or 'mead' as single compounds dealing with specific deprivation (ironic). The classification could become a 'black hole', exhaustively drawing in components of meaning, a classification of parts of words rather than word senses. The meaning of a poetic nonceword is inherently insecure, no matter how unanimous the translators and dictionary editors. What, for example, is a wælsteng (Beo 1634)? Four men carry Grendel's head on this piercing implement, readers visualizing it usually as a spear or spike. It could equally well refer to a sword. Its appearance preceded by the definite article seems to indicate that the implement has already been mentioned, so does Beowulf after all have a use for Hrunting on the return from the mere?
Lengthy accounts could be written on the placing within the classification of word after word. During editing we kept open a special notes line in the database, to indicate reasons for decision, worries, indecisions: information that could be useful for more detailed handling of the materials of the database. But where traditionally a thesaurus-user has been expected to know the meanings of words scanned, users of this thesaurus are expected to be aware of the sorts of problems Old English words present and to speculate generously as to the choices made by the editors. Metaphorical compounds must be regarded either as a source of enjoyment or as grounds for despair. Tempting though it would be to place such compounds for their separate elements as well as for any recognized derived meaning, a mixture of daring and caution led to their being placed according to derived meanings as far as possible. Thus, words as merehengest and sundhengest will be found only under the heading 'ship', and not within 02.06 Animal or even 01.01.03 Sea, whereas sǣhengest appears both as a ship and a hippopotamus. Yet, the more exotic the image found within a poetic compound, the more likely it is as a candidate for two or three, or even more positions within the classification. Equally, words found outside the poetic corpus can be similarly charged: for example the image contained in scÿrmǣlum (Bo 20 47 25) relates to sword decoration, but the word is properly found at 01.03.01.03.01 Storm, tempest.
Were we certain of having achieved consistency, we would assert that this flag is used of words restricted to glossed texts and glossaries. Two areas of particular doubt must be noted. First, Latin/English word equivalents within Ælfric's Grammar are anomalous, being a part of prose usage according to Cameron numbers. If the form in question is infrequent or indeed a nonceword the decision not to invoke g is a hard one to take. The form æcermann (ÆGl 303 19), for example, occurs only in Aelfric's Grammar, but is flagged o rather than og. So too ǣfyrmþa (ÆGr 85 5) is given non-gloss status. At least its o flag indicates its rarity; the supporting form ǣfyrm (AldV 13.1 (Nap) 39180) differs from it in meaning.
Counting g forms is so fraught with difficulty that an accompanying o is invoked only after considerable thought. Once the decision had been taken to regard use in the Lindisfarne Gospels Gloss and the Rushworth Gospels Gloss as separate instances of a form, it became the harder not to be similarly generous in respect of other g words. It would be fair to say, therefore, that multiple copies are not generally reconciled. This is comparable with the practice adopted in respect of the Anglo-Saxon Chronicle, although not for all prose texts that occur in more than one manuscript. Thus g on its own may relate to a rare infrequent form, but any word flagged both o and g is very probably found once only, in g texts. Further doubts linger: there is great need for differentiation among glossed texts. Some contain free-running materials and can present English as good as or even better than patches of texts categorized as Old English prose texts. Influence from 'the practice of interlinear glossing' has been advanced for the Bede translation, which is not alone among prose texts in providing forms every bit as questionable as many a word flagged g. Future work will require a better sense of the quality of glossed texts, each of which should be examined carefully with the distinctions made by Gneuss among loan translations in mind.
Every effort has been made to use this flag as little as possible. It should be thought of as indicating a word that does not occur, indeed a word whose putative existence is gravely to be doubted. Its use is best explained as indicating words that it might have been wiser to exclude. The retention of such words is a matter almost of book-keeping, an entry showing that the form was thought worthy of inclusion when the dictionaries were excerpted and that it was retained throughout the editing of the TOE materials. Word senses and words altogether rejected from the thesaurus are not recorded. There are three sorts of remainders: some absolutely intractable g forms; some words rejected because rejected by Meritt and other authorities; and some words from the language's closed systems, for example æt, and, hiere, his, mīn, sīn.
The q flag is found disproportionately often in certain areas of the classification. The place-name elements included in Campbell's Addenda cluster markedly in Category 01 The Physical World and in 04.03 Farm. Sometimes quite a number of the items under a heading are flagged q, as for example roþ, roþer, ryd, ryding/-en (postulated cognates for rod) in 04.02.03.01.02 Cleared land. These are listed alongside (ge)ryþer, another of the place-name elements included in Campbell's supplement to Bosworth-Toller, but (ge)ryþer is unflagged because supported by use within Old English (Ch 468 (Birch 756) 5, Ch 619 (Birch 982) 6, Ch 1028 (HarmApp 2) 10). In 01.01.02.01.02.01 Rising ground, eminence the form hylloc is another of these place-name additions, placed coincidentally under the heading Hillock; and it is comparable with the unflagged hylc, under the heading Hummock within the same group. Here we might have omitted hylloc altogether, but it remains and is separately flagged, therefore giving more evidence than would its absence. Had it been mapped on to hylc (that is, as hylc/hylloc in both entries), it would not have been possible to show it as apparently unexemplified within the extant corpus of Old English.
Some words have a dictionary life but are not to be found anywhere in Old English as separate words. Where these have been recognized, they are flagged q. Examples include forms not found as independent words, whether fairly common elements such as gelegu 'tract of land' and rād 'road' or less frequent forms such as pēac 'peak' (supported by the single occurrence of two proper names only, Chron A (Plummer) 924 42 Pēaclond, Rec 26.2 (Birch 297) 35 Pēcsǣtna and by a charter phrase Ch 1503 (Whitelock 20) 35 æt Pēacesdēle). Otherwise the q flag is invoked for words in some way untrustworthy. For example, a solitary occurrence in Old English of swān 'warrior' gets into dictionaries, reflecting the emendation swānas from swanoc, the form found in Hickes's printed text of the Finnsburh Fragment (Finn 37 swētne/). The form appears in the TOE, but the accompanying q flag ensures that its inclusion cannot be misread as acceptance of the word's existence in Old English.
 See Waldorf (note 6 in the source dictionaries section) for discussion of this problem.
 Citations appear in the form recorded by the MCOE and are identified by MCOE short titles and references.
 Support is perhaps to be found in Ch 412 13; OED3 cites the Old English passage and gives 1613 as the next attestation.
 The DOE editors record eighteen forms, no one with the ge- prefix.
 It is not recorded in the DOE.
 As evidently was the decision of the DOE editors.
 E. V. K. Dobbie, ed., The Anglo-Saxon Minor Poems, ASPR, 6, New York, 1942; ibid., Beowulf and Judith, ASPR, 4, New York, 1953; G. P. Krapp, ed., The Junius Manuscript, ASPR, 1, New York, 1931; ibid., The Vercelli Book, ASPR, 2, New York, 1932; ibid., The Paris Psalter and the Meters of Boethius, ASPR, 5, New York, 1932; ibid. and E. V. K. Dobbie, ed., The Exeter Book, ASPR, 3, New York, 1936.
 For discussion of these bounds, see Peter Kitson, 'Some unrecognized Old English and Anglo-Latin verse', Notes and Queries, 232 (1987), 147-51.
 For discussion of the difficulties of categorizing Old English vocabulary, see E. G. Stanley, 'Studies in the Prosaic Vocabulary of Old English Verse', Neuphilologische Mitteilungen, 72 (1971), 385-418; M. S. Griffith, 'Poetic Language and the Paris Psalter: the decay of the Old English tradition', Anglo-Saxon England, 20 (1991), 167-86; Roberta Frank, 'Poetic Words in Late Old English Prose', From Anglo-Saxon to Early Middle English: Studies presented to E. G. Stanley, ed. Malcolm Godden, Douglas Gray and Terry Hoad, Oxford, 1994.
 Such concentrations were signalled as long ago as J. W. Rankin, 'A Study of the Kennings in Anglo-Saxon Poetry', Journal of English and Germanic Philology, 8 (1909), 357-422; 9 (1910), 49-84.
 For examples, reference is made to the head of each group.
 G. Storms, in his review of P. O. E. Gradon's Elene: English Studies, 46 (1965), 54, points out that warriors shake spears rather than coats of mail.
 Multiple copies of the Old English Bede are reconciled.
 See Dorothy Whitelock, 'The Old English Bede', Proceedings of the British Academy, 48 (1962), 76, for the mixture of 'stiff and clumsy' and 'vigorous and idiomatic' passages in this translation.
 See Helmut Gneuss, Lehnbildungen und Lehnbedeutungen im Altenglischen, Berlin, 1955, and 'Some Problems and Principles in the Lexicography of Old English', Festschrift für Karl Schneider, ed. Kurt R. Jankowsky and Ernst S. Dick, Amsterdam/Philadelphia, 1982, pp. 153-68.