Document 1532

Synonym Clustering in Beowulf

Author(s): Prof Christian Kay

Copyright holder(s): Prof Christian Kay


The paper describes an analysis of the repetition of words from twelve synonym sets in the poem Beowulf. The poet is found to have a tendency to repeat individual words or clusters involving compounds with a common element, or simplex-compound clusters. It is concluded that this clustering tendency may be accounted for by the fact that the poet composed orally or was influenced by an oral formulaic tradition. It is finally suggested that such clustering may be a feature of oral style in general.

The starting point for this study was the casual observation that the Beowulf poet tends to repeat words within the space of relatively few lines. This seemed worthy of investigation because many of the repeated words belong to sets of synonyms which offered him a wide choice of expressions, and because the vocabulary of Old English verse is more noted for variety than for repetitiveness.

Twelve sets of synonyms were therefore selected for analysis:

“corselet” (40 words); “sword” (26); “shield” (11); “spear” (10); “boat” (18); “sea” (24); “hall” (27); “God” (19); “king, chieftain” (59); “warrior” (42); “comrade” (12); “troop” (17). (1) Using Klaeber's edition of the poem, (2) all occurrences of these words were recorded. In this way, 304 of the poem’s content words (excluding prepositions, auxiliaries, etc.), approximately eleven per cent, were considered. In addition, note was taken of a selection of compounds in which one of the elements was one of the simplices under consideration. Searohæbbend, for instance, was listed with searo in the “corselet” group as well as with the group of words meaning “warrior”.

Occurrences of two or more of these words in significant positions are referred to as “clustering”. The first type of clustering investigated was “large-scale clustering”, that is the distribution of synonyms in relation to the three major sections of the poem (ll. l-1250, the introduction and the fight with Grendel; ll.1251-2199, the fight with Grendel’s mother and the return to Gautland; and ll. 2200-3182, the dragon fight and the death of Beowulf). The second type, “intensive clustering”, shows how individual words within the synonym groups occur in relation to one another. It should be emphasised from the beginning that the study is in no way intended to be statistical: its purpose is simply to illustrate a feature of the poet’s style.

Large-scale Clustering. This part of the survey showed that there were quite considerable differences in the vocabulary of the three major sections. In the first place, 163 of the 304 words occurred only once in the poem (84 in section 1; 47 in section 2; and 32 in section 3). Of the remaining 141 words, 50 occurred in all three sections, 59 in two out of the three, and 32 in only one..

Intensive Clustering. The study of repeated words, or of groups of words with a common element, yielded far more interesting results. Altogether, 281 instances of such phenomena were recorded. These will be discussed according to the type of repetition involved.

1. There were 99 examples of repeated simplices (for example, cempa in ll.1948, 2044, 2078 or Wealdend, ll.1661, 1693, 1752), and 26 examples of repeated compounds (e.g., sΣgenga, ll.1882, 1908). Many more examples of this kind were discarded because the words involved occurred too frequently in the poem to make their recurrence significant. In the case of less common words, and particularly of words like sΣgenga, which occur only at one point in the poem, recurrence was felt to be significant. Another example of this kind is flota, which occurs in 210, 218, 294 and 301, and nowhere else. Even allowing for the few passages in which “ship” synonyms are needed, such recurrence is interesting.

A more extended analysis of the four commonest synonyms in the “hall” group – heal, sele, reced, flet – gives evidence of clustering. The occurrences felt to be significant are underlined:

sele 81/ heal 89/ reced 310/ sele 323/ reced 326/ sele 411/ reced 412/ heal 487/ heal 614/ heal 642/ heal 663/ sele 713/ reced 720/ reced 724/ reced 728/ reced 770/ se1e 826/ sele 919/ heal 925/ heal 1009/ sele 1016/ flet 1025/ flet 1036/ flet 1086/ heal 1087/ heal 1151/ heal 1214/ reced 1237/ heal 1288/ reced 1572/ sele 1640/ flet I647/ reced 1799/ heal 1926/ flet 1949/ flet 1976/ sele 1984/ flet 2017/ flet 2034/ flet 2054/ sele 2352/ reced 3088/ sele 3128. (3)

Thus there are at least five instances of synonym clustering in this group. The alternating use of sele and reced in ll.310-412 is interesting, as is the fact that heal, the commonest word in the group, does not occur after l.1926. Similar tendencies were discovered in other groups.

2. There were 22 examples of the repetition of two or more compounds with the same first element, and 45 examples of the repetition of compounds with the same second element. An instance of the first kind is herebyrne, l.1443, heresyrce, l.1511, and herenet, l.1553. The second kind is exemplified by hringmΣl, ll.1521, 1564, and brogdenmΣl, ll,1616, 1667. A cluster combining both kinds is g)ðsearo, ll. 215, 328, fyrdsearo, l. 232, and g)ðgewΣde, l.227. In all these examples, the words happen to be unique or to occur only at the point recorded. In fact, almost all the instances of this kind of clustering included at least one unique compound. The fact that these unique words should make their appearance in the company of related words makes the clusters doubly interesting.

3. The 89 occurrences of simplex-compound clusters, that is of compounds grouped round the simplex which is one of their elements, showed by far the most interesting results. In such cases it seems reasonable to deduce that the compounds were, consciously or unconsciously, associated with the simplex in the poet’s mind. If the compounds in turn are, as in so many cases, unique or rare, this association becomes the more noteworthy. An example of this kind is the group of 5 words containing bil as one element. The distribution of this group is; bil 40/ hildebil 557/ bil 583/ g)ðbil 803/ bil 1144/ hildebil 1520/ bil 1557/ bil 1567/ w{gbiI 1607/ hildebil 1666/ bi1 2060, 2359, 2485, 2508/ g)ðbil 2584/ bil 2621/ hildebil 2697/ bil 2777/

Apart from the repetition of bil, itself one of the less common “sword” synonyms, it will be seen that only g)ðbil in l.803 is entirely dissociated from the simplex. None of these words occurs elsewhere in the poem, nor are there any other synonymous compounds with bil as an element. A similar cluster involving wudu, ll.216, 298, 1919, sΣwudu l.226, and sundwudu, ll.208 and 1906, is strengthened by the appearance of the comparatively unusual “sea” synonym, sund, in ll.213 and 223.

A more extended analysis involved the four simplex “shield” synonyms, bord, lind, rond and scyld. Without going into all the details, it is worth noting that bord, and six of the eight occurrences of its compounds, appear only in the third section of the poem. Three of the compounds appear in close association with the simplex, and three thereafter. Similarly, only one of the six rond compounds is isolated from its simplex. Moreover, five of the seven occurrences of these compounds come in the first 1300 lines of the poem, where rond is the most common simplex. Comparing these with the bord compounds, it almost seems as if the latter took over once their simplex had been introduced.

Of course, not all the compound-simplex groups yielded such positive results. The compounds of lind and scyld, for instance, showed few clustering tendencies. Nevertheless, taking the study as a whole, it was` felt that sufficient examples of the various kinds of clustering existed to suggest that they were a definite feature of the poet’s style, and not merely the results of coincidence, or the demands of alliteration, to name two possible explanations.

Having established that clustering was a feature of the poet’s style, the next step was to determine whether it was a significant feature, that is whether it could be explained in literary terms or used to elucidate problems of the poem’s composition or style. Beowulf criticism has undergone something of an upheaval in recent years, largely because of research on the formulaic composition of ancient oral poetry, and the application of oral formulaic theory to Old English verse. (4) While there is no real agreement on whether Beowulf was composed orally, or, if not, how much the oral tradition influenced its style, the weight of the evidence suggests that we should “...explain structural and stylistic features of Beowulf in terms of oral composition or at least as originating in an oral tradition...” (5) That is to say, if one has the temerity to summarise these theories in a few bald words, that rather than composing slowly in writing with the word as his unit, the poet composed a different oral poem for each occasion, working swiftly and spontaneously in formulaic units.

An acceptance of the theory that oral tradition at least influenced the Beowulf poet goes some way towards explaining the existence of clustering, as does the less controversial fact that the poem was presented orally to an audience. The very fact that the poem, because of its length, must have been presented at several sittings, may help to explain the overall differences in vocabulary.

The method of presentation may also explain intensive clustering. Unlike the lettered poet, the oral poet cannot review his previous work before beginning a new section, and thereby re-establish its vocabulary in his memory. Nor, at a more conscious level, can he look through his work and decide that a particular word has been overused or used with particular appropriateness and so on. Because of the demands of composing orally and at speed, the poet would concentrate his attention on short sections of the poem rather than on the work as a whole. It is conceivable that within these short sections he might repeat a word simply because it was conveniently in the forefront of his mind. Thus, under the stress of composition, he might well tend to use several formulas in which the key word was, say, byrne, rather than search for totally different formulas with searo or syrce. Similarly, if, having used byrne, he had to find a formula containing “warrior”, it seems natural that one with byrnwiga should spring to mind rather than one based on an entirely different word. Byrnwiga might lead him to scyldwiga or wiga itself, or he might vary his reference to armour by using {renbyrne or {sernbvrne. His choice of words would be influenced by both semantic and phonetic association.

Such speculation perhaps assumes too much knowledge of the poet’s mental processes, especially when the poet was composing according to conventions so different from our own. Comparison with written literature is unlikely to reveal much; if we must make comparisons, it is perhaps better to compare oral poetry with -the speech or impromptu written material of our own period. While a modern writer is unlikely to repeat a word except on rare occasions where he is aiming at a deliberate literary effect, such repetition is, I think, quite common in conversation, or a letter, or a rough draft of a piece of writing. This is, in fact, the familiar experience of having a word “stick in our minds”, and is something we try to eliminate from carefully written work as it is not stylistically attractive to a modern reader. Words may well have stuck in the mind of the oral poet, who would not have the opportunity, nor, since his stylistic assumptions were different, the desire to vary his vocabulary. Thus, clustering may have arisen naturally from the conditions under which the oral poet worked.

There is no reason to suppose that the oral poet’s audience would be conscious of the clustering under discussion. Like the poet himself, they would be more concerned with getting the story told than with stylistic minutiae. Nevertheless, repetition may have aided them in their comprehension of the poem by establishing a train of verbal continuity in their minds. A repeated word, or a group of related words, would help them to grasp an essential idea. On the other hand, for all we know to the contrary, they may have enjoyed the effect of a repeated word, or the interweaving of a “family” of words.

It is, then, possible, that clustering in the various ways outlined above is a feature of oral style, helpful to both the poet and his audience. As the phenomenon is preserved in our version of Beowulf, it may represent either an oral poet at work or the stylistic legacy left by generations of such poets to a lettered man.

In any event, the limits of the present paper by no means indicate the limits of the study of clustering as a whole, If the results are felt to be significant, it would obviously be desirable to extend the analysis to the entire vocabulary of the poem, and to see how the clusters are related to the known formulas in it. One could also study the clusters in terms of the fitts as this might be one means of determining whether the fitts were one of the units of composition used by the poet. It might also be profitable to make a similar study of other Old English poems where oral composition is a strong possibility, or even to study clustering in known oral style of different genres and periods. If it could be shown that clustering was a constant feature of oral style, the existence of this phenomenon in works of dubious origin might prove to be one method of determining how such works were composed. Such possibilities suggest that the clustering of individual words, or of words similar in sound or meaning, is a significant feature of style, and one which would repay further investigation.

1 A problem here was how to define a synonym in Old English. After consulting many authorities, whose own estimates of the number of synonyms in each group varied considerably, it was decided to include within a synonym group all words which are interchangeable to the extent that they convey the same essential meaning. Thus helmberend was considered synonymous with scyldfreca, holm with sund, etc. Although this ignores etymological and connotational meaning, it was felt to be consistent with the character of the OE poetic vocabulary.

2 Frederick Klæber, ed., “Beowulf and the Fight at Finnsburg” (3rd ed., Boston, 1950). All references are to this edition, and Klaeber’s emendations are accepted throughout.

3 Note also that the only two occurrences of flet in the sense of “floor” come in 11.1540 and 1586.

4 Most notably, Francis P. Magoun, Jr., “Oral Formulaic Character of Anglo-Saxon Narrative Poetry”, Speculum, XXVIII (1953).

5 J.C. Van Meurs, “Beowulf and Literary Criticism”, Neophilologus, XXXIX (1955), 127.

This work is protected by copyright. All rights reserved.

The SCOTS Project and the University of Glasgow do not necessarily endorse, support or recommend the views expressed in this document.


Cite this Document

APA Style:

Synonym Clustering in Beowulf. 2021. In The Scottish Corpus of Texts & Speech. Glasgow: University of Glasgow. Retrieved January 2021, from

MLA Style:

"Synonym Clustering in Beowulf." The Scottish Corpus of Texts & Speech. Glasgow: University of Glasgow, 2021. Web. January 2021.

Chicago Style

The Scottish Corpus of Texts & Speech, s.v., "Synonym Clustering in Beowulf," accessed January 2021,

If your style guide prefers a single bibliography entry for this resource, we recommend:

The Scottish Corpus of Texts & Speech. 2021. Glasgow: University of Glasgow.


Information about Document 1532

Synonym Clustering in Beowulf


Text audience

Audience size N/A

Text details

Method of composition Handwritten
Year of composition 1966
Word count 2531

Text type



Author details

Author id 606
Title Prof
Forenames Christian
Surname Kay
Gender Female
Decade of birth 1940
Educational attainment University
Age left school 18
Upbringing/religious beliefs Protestantism
Occupation Academic
Place of birth Edinburgh
Region of birth Midlothian
Birthplace CSD dialect area midLoth
Country of birth Scotland
Place of residence Glasgow
Region of residence Glasgow
Residence CSD dialect area Gsw
Country of residence Scotland
Father's place of birth Leith
Father's region of birth Midlothian
Father's birthplace CSD dialect area midLoth
Father's country of birth Scotland
Mother's place of birth Edinburgh
Mother's region of birth Midlothian
Mother's birthplace CSD dialect area midLoth
Mother's country of birth Scotland


Language Speak Read Write Understand Circumstances
English Yes Yes Yes Yes All
Scots No Yes No Yes Work