Old English literature is largely preserved in manuscripts of the late tenth, eleventh and twelfth centuries, and Beowulf is no exception, surviving in a late 10th- or early 11th-c. codex. Based on external evidence such as historical references or authorship, some poetry, like Cædmon's Hymn, can be dated as early as the 7th-c., whilst other poems, like The Death of Edgar, can be dated as late as the mid-eleventh century. But the main body of Old English literature, including Beowulf, cannot be so easily dated, except roughly to sometime during a period of two or three hundred years or more. The importance of dating such literature is not only historical, for, in interpreting literature, knowledge of the setting in which a work was composed is of great importance. As Amos puts it, '[it] is as if we knew that Sir Philip Sidney wrote in the late sixteenth century and Wallace Stevens in the mid-twentieth century, but could not assign even relative dates within that period to the work of Dickens and Shakespeare' (pg 1).

One of the more interesting, and empirically-sound, clues to the date of composition of Beowulf lies in a consideration of certain aspects of the metre of the poem. Specifically, the conformity of the metre of Beowulf to 'Kaluza's Law' (governing the resolution of syllables into a single position) would be unexpected if the poem had been composed after 825 AD.  What follows attempts to present a straightforward synopsis of the relevant facts, largely based on R.D. Fulk's excellent work on the dating of Old English poems based on Kaluza's Law (see Fulk, A History of Old English Meter, §§ 170-183, §§406-421), which suggests that Beowulf was composed between 685 AD - 725 AD (though one should be aware that there are other valid arguments for a later date of composition, as well as some difficulties with the evidence provided by Kaluza's Law:-- see postscript below).

Contents Basics of Old English Metre Resolution: two syllables as one Kaluza's Law & history of OE vowels The Date of Beowulf Postscript Appendix A: Kaluza's law in Beowulf Appendix B: Kaluza's law in OE poetry Appendix C: Kaluza's original formulation (deutsch & English) Appendix D: OE inflexion endings Bibliography

I. BASICS OF OLD ENGLISH METRE 

Metre can be roughly described as the rhythm used in recitation. More exactly, metre comprises the patterns of stressed (or emphasised) and unstressed syllables, which are inherent in spoken language, but take on a more regulated form in poetry. In English, both modern and ancient, stressed syllables are usually distinguished from unstressed syllables in being longer and/or having more amplitude (i.e. being 'louder').

Verses in Old English are bound together by alliteration, much as rhyme forms the linking structure for rhyming poetry. Alliteration is the repetition of the initial sounds of words (thus it is sometimes called 'head rhyme' or 'initial rhyme'), serving not only to link verses, but also to emphasise important words within the verse. In Old English (and Old Germanic in general), a vowels alliterates with any other vowel (more correctly, any onsetless syllable alliterates with any other onsetless syllable) -- otherwise strict identity is maintained. Usually nouns and adverbs are the alliterating words, and the on-verse has one or two alliterating words, but the off-verse only one.  

We may broadly classify the verse patterns used in Old English (and thus in Beowulf) into the following types (based largely on Sievers' classification):

Verse Type	Pattern (Bliss)	Pattern (Pope)	OE example	modern Eng. ex.
Type A [trochaic]	/ x | / x	lift, drop, lift, drop	121a grim ond graédig	grim and greedy
Type B [iambic]	x / | x /	drop, lift, drop, lift	510a né léof né láð	not friend nor foe
Type C [clashing]	x / | / x	drop, lift, lift, drop	4a oft Scyld Scéfing	oft Shield Shayfing
Type D1	/ | / \ x	lift, lift, half-lift, drop	1848a wís wordcwida	wise word-speaker
Type D4	/ | / x \	lift, lift, drop, half-lift	2885a eall éðelwyn	all native joy
Type E	/ \ x | /	lift, half-lift, drop, lift	131b þegnsorge dréah	thane-sorrow pain'd

Though a vast number of verses in Beowulf (and other Old English poems) do not conform exactly to any of these 'basic' types, we find that most verses are equivalent to one or other of the above types. One of the most common verse-forms in Beowulf is exemplified by wéox under wolcnum (8a) [ / x x / x ], in other words Type A with an 'extra' unstressed syllable in the first foot. In general one may observe that in all of the types we find four positions - in fact in the basic types, these four positions map perfectly onto four syllables. But not all verses have a direct equivalence between the four positions and four syllables : l. 8a above [ / x x / x ] shows an expansion of the first weak position (also called the first 'dip') into two syllables, some verses show an even greater expansion, like 1587b tó ðæs þe hé on ræste geseah [ x x x x x / | x x-/ ]. l. 1587b also shows another 'accomodation phenomenon' in the extrametricality of the prefix ge- of geseah : thus the notation x-/, which indicates that these two syllables are combined into one position (more specifically, the prefix is cliticised to the following stressed syllable). A phenomenon similar to extrametricality is anacrusis (called Auftakt by German metrists). Anacrusis consists of one or two extrametrical weak syllables preceding an initial falling foot in an on-verse (i.e. a- verse). An example of anacrusis (with two weak syllables: ne and ge- ) occurs in l. 109a ne gefeah hé þaére faéhðe [ x-x-/ x x x | / x ].

This brief discussion only grazes the surface of the complexities of Old English metre - for a fuller introduction, see Sievers ('Old Germanic Metrics'), Stockwell & Minkova and Bliss ( Introduction to Old English Metre) - the purpose herein being only to acquaint the reader with some basics of OE metre to make the following discussion more understandable.

II. METRICAL RESOLUTION & KALUZA'S LAW
The metrical 'accomodation phenomenon' of interest to us here is resolution (concept introducted into Germanic metrics by Eduard Sievers, 'Rhythmik', in 1885, as Auflösung) : by which two successive syllables are counted metrically as if they were a single syllable. We represent resolution notationally as [ /-x ] (resolution of a primary stress) or [ \-x ] (resolution of a secondary stress). For example, in brimclifu blícan (222a), the word clifu which has two syllables, is treated by the metre as though it were a single syllable, thus the pattern of 222a is [ / \-x | / x ]. Essentially, to occupy the ictus of a foot (i.e., to be stressed), either a heavy syllable, or a light syllable combined with the following syllable is required. A heavy syllable is a syllable containing a long vowel (e.g. bá ), a long diphthong (e.g. béo ) or a vowel plus one of more consonants in the same syllable (e.g. bat ); a light syllable is a syllable containing only a short vowel or short diphthong. There are some circumstances in which two such syllables do not resolve as described above, specifically: if the light stressed syllable is immediately preceded by another stressed syllable, and if the weak resolving syllable is heavy or--in the oldest OE poetry-- if the syllable was historically heavy, then resolution does not take place.

This last condition on non-resolution, i.e. a light resolving syllable which was historically heavy, is the crucial distinction which we shall exploit in the dating of Beowulf. In 1896 the German scholar Max Kaluza (in 'Zur Betonungs- und Verslehre des Altenglischen') published his crucial discovery that, in some metrical positions, resolution is governed in part by etymological considerations. Specifically, that resolution is blocked in certain cases not only if the resolving syllable is heavy but also if it is historically heavy. Historically heavy syllables in this case are those 'long' inflectional endings which end in a consonant (such as the masculine a- stem genitive singular ending -es) or once carried a 'circumflex' intonation ( Schleifton in German) in an earlier stage of the language (such as the ó- stem nominative plural ending -a from Proto-Germanic *-ô(z) from Proto-Indo-European *-âs ). The 'short' inflectional endings are the remainder (such as the masculine i-stem nominative singular -e from Proto-Germanic *-iz ) [see further Appendix D. below].

This condition on resolution is known as Kaluza's Law (see Appendix C. for Kaluza's own formulation). Additionally, the law is restricted to only such cases in which the resolvable syllables and the immediately preceding lift (=stressed syllable) stand in the same foot. This additonal condition rules out, for instance, verses of type 2C2 (e.g. on bearm scipes (35b)). Despite Kaluza's claims, his law does not apply regularly in Types B and E (see Fulk §172). Without going into fine detail, which does not concern our primary purpose of dating here, we may observe that in Beowulf Kaluza's Law applies regularly under secondary stress in Bliss's verse types 2A3a(ii), 2A4, 1D3 and 3E3 (see Bliss, Metre of Beowulf,  §§34-7 & Fulk, History, §§170-183) and verse-finally under primary stress in Bliss's verse types 2A1b, 2A3b, a1 (Hutcheson, 3.D). Appendix A. provides an exhausting listing and details about all verses in Beowulf to which Kaluza's law applies.

This section may be somewhat technical for some readers.  The essential points are (a) in Old English verse, under certain conditions, two syllables are treated as a single syllable by the metre ('resolution'); and (b) of one the conditions for resolution, in the oldest OE verse, is the etymological status of a syllable as 'heavy' or 'light'. We shall take up the implications of point (b) for the dating of Beowulf in the next section.

III. VOWEL LENGTH, METRE AND THE DATING OF BEOWULF
The importance of Kaluza's law with respect to determining the date of composition of Beowulf lies in the fact that the poem observes the metrical distinction between long and short vocalic endings. The phonological distinction between these two types of endings was subsequently lost, as may be observed from a survey of Old English poems.  Fulk (History, §§176-83) catalogued of the verses of Bliss types 2A3a(ii), 2A4 & 1D3 & 3E3 (those verses with in which Kaluza's law applies under secondary stress), both those which conform to Kaluza's law and those which violate it for the OE poems Genesis A, Daniel, Beowulf, Exodus, Elene, Fates of the Apostles, Juliana, Christ II, Andreas, Metrical Preface to the Pastoral Care, Metres of Boethius, Judith and Battle of Maldon - (Appendix A. lists all verses from Beowulf and Appendix B. offers a summary of the statistics for most of the poems in Fulk's test group) . He found that Beowulf contains 108 relevant verses, of which 106 conform to Kaluza's law (i.e. only 1.9% of the verses are exceptions to Kaluza's law), and one of these exceptions involves a consonantal ending and is thus not relevant to the distinction between long and short vocalic endings. On the other hand, unsurprisingly, poems known to be late such as those of Cynewulf show a much higher rate of violation of the law. In addition to the high rate of conformity to Kaluza's law by Beowulf is the fact that verses of the relevant types (i.e. the type to which Kaluza's law applies) are exceptionally frequent in contrast with comparable poems: 'For example, while there are 108 relevant verses in Beowulf, in Genesis A there are just 9, even though the latter poem is nearly three-quarters the length of the former, and is rich in poetic compounds. The difference is especially extreme in type I [resolution pre-caesura], where Beowulf furnishes 65 examples and Genesis A only 2' (Fulk, History, §179).

Hutcheson demonstrates that, in addition to applying under secondary stress in certain A & D types (see above), it also applies under primary stress at the end of a number of 'light' verse types (including 2A1b, 2A3b, a1) -- in that resolvable endings do not usually occur verse-finally in these types (Hutcheson, 3.D). This allows us to add another 71 verses to our list of relevant verses of which only 1 is an exception to Kaluza's law, thus providing us with a total of 179 relevant verses in Beowulf with but 3 exceptions to Kaluza's law (a study of these 'light' types also adds appreciably to the list of relevant verses for Genesis A - see Appendix B. below).     

Fulk concludes: 'In sum, Beowulf is the most conservative poem of the test group with regard to Kaluza's law. Cynewulf and the Andreas poet violate it with considerable frequency. Poems known or presumed to date to Alfred's reign or later, to the extent that they provide any evidence at all, contain few instances to offer any statistical significance, but they clearly do not conform to the law. Violations in Genesis A, Daniel, and Exodus are few, but so are the relevant instances, at least in verses of type I. The difference between Beowulf and the rest of the test group, in respect both to regularity of application and to overall incidence of relevant verse types, is so profound that it can hardly be dissociated from the elimination of the phonological distinction between long and short vocalic endings. Beowulf, at least, must have been composed before the shortening of the long endings' ( History, §182).

The crucial conclusion to be drawn from these facts is: 'Observance of Kaluza's law...depends upon the maintenance of the original distinctions in vowel quantities. The fidelity of Beowulf to the law demonstrates that the poem can only have been composed before the quantitative distinction [between short and long vocalic endings] was lost' (Fulk, History,  §183 [emphasis added]).

Before we turn to the question of when the distinction between short and long vocalic endings must have collapsed, let us first address two possible objections to the conclusion reached above. The first we may find in Bliss's The Metre of Beowulf, Appendix B(pp.118-21):-- Bliss proposes that the phonological distinction between the endings may have become part of the morphology of the poetic language.  That is to say that it would be learned in the same way as a modern English speaker learns that the plural of man is men and foot is feet and knife is knives, despite the 'regular' plural form in -s, e.g. books, swords, tables, etc. Specifically with regard to Kaluza's law, Bliss's proposal is that the poet would learn 'a very obvious analogy between morphological and metrical equivalences: [e.g.] nominative plural scipu and hus and genitive plural scipa and husa are equivalent both in grammar and in metre' ( Metre of Beowulf, pg. 120). What Bliss means is that despite the fact that scipu has two syllables and hús one, they are treated the same by the grammar (since they are both nominative plural) and by the metre, since scipu may be resolved and thus treated as a single syllable. This claim is untenable because the distribution of short and long endings is not systematic, and such analogies do not hold true in many instances: '[f]or example, in the feminine ó- stems the acc. sg. ceare is not metrically equivalent to laré , which in turn is metrically equivalent to dat. sg. ceare , which is not metrically equivalent to gen. sg. ceare , even though the appearance of none of these phonologically identical endings varies according to the weight of the root syllable [as in the case between scip and hús ]' (Fulk, History, §406). If the reader finds the present discussion obscure and perplexing, let him reflect upon the idea that the Anglo-Saxon poet (who surely composed what 'sounded right' to him and not by making reference to charts or tables) could have kept a distinction between these endings if they all sounded identical. Such a notion is clearly unbelievable.

The second objection to the claim that the poem must have been composed by someone who still spoke an early enough form of English which distinguished phonologically short and long vocalic endings is that the verses in question might be purely formulaic. In other words, that these verses, or close variants of the same, would have been part of a common oral poetic 'repository' and the poet would have simply memorised them without having to be aware of any phonological distinction between short and long endings. 'The formulaic nature of [these] verses...is apparent; for example, all fourteen of the "neuter wa-stem" verses that Bliss lists contain either -bealu or -searo, and all ten of the "masculine u- stem" verses contain -wudu . This strongly suggests that what must have originally been a phonological distinction is preserved in Beowulf by the formulaic tradition as well, or else there would be no way to explain why the law operates in A- and D-types, but not in C-types' (Hutcheson, pp.90-1). Hutcheson suggests that A- and D-type verses are an older part of Germanic metre than C-types and thus the older formulaic verses which preserve Kaluza's law are verse types A and D, but Fulk rightly objects that this claim is unlikely since type C is common to all early Germanic languages ( pg. 386n74). It is simpler to assume that in type C there are other metrical constraints which override the considerations of Kaluza's law, though the nature of these constraints may not be entirely clear to us at present (Suzuki, 'Preference Conditions', argues that Kaluza's law does indeed apply in C-types, but assumes a different criterion for distinguishing between 'long' and 'short' vocalic endings). To address the general suggestion that the appearance of these verses conforming to Kaluza's law in Beowulf is due to inherited formulaic phrases rather than preservation of a phonological distinction betweeen short and long vocalic endings, let us consider the results of Fulk's survey of OE poetry (see Appendix B. below). If the presence in Beowulf of the verses in question is due to a formulaic tradition, then why do not the other poems show a similar conformity to the law?  More tellingly even, is the fact that so few verses which are excepts to Kaluza's law exist in Beowulf. Even if we thought that the verses conforming to the law were simply memorised formulae, this does not explain the absence of verses violating the law. Furthermore, OE formulas are not syntactically frozen (see Russom, 'Artful Avoidance'), and thus the same formula may appear in different forms, some of which would be violations of Kaluza's law. Additionally, even if a phrase were to be frozen phonologically, since many of the inflexions are homophonous in later OE but often not equivalent in phonological length in earlier OE, even this would not guarantee conformity to Kaluza's law. For instance, the phrase módceare micel (l.1778a), which appears in the poem with -ceare as accusative singular, would violate Kaluza's law if a poet used -ceare in the genitive or dative singular, despite the fact that the phrase would be phonologically identical: módceare micel.  To put it another way, if these verses were just memorised, this still does not explain how did the poet would have known how to avoid verses violating Kaluza's law.

Thus the only reasonable conclusion regarding these verses must be the other stated above: that the poem must have been composed before the real, audible, phonological distinction between short and long vocalic endings was lost. 

IV. THE DATE OF BEOWULF
Having argued that Beowulf must have been composed whilst the distinction between short and long vocalic endings was still an audible, phonological feature of English, it remains to determine what the latest date is at which these distinctions could have existed.  For this we must examine the loss of phonological distinction in unstressed vowels in Old English overall. In general, there is a tendency in language to collapse various phonetic differences between vowels in unstressed syllables. This tendency is particularly prominent in English - in modern English in fact most unstressed vowels are reduced to a central vowel sound (known as schwa). In Old English there is a relatively early coalescence of unstressed -æ and unstressed -i into -e (see Campbell, §369), e.g. the 1st-per. sg. pres. indic. form hafe shows up as hafæ and the adjective micel as micil in some early texts. The indications are that these changes occured sometime during the end of the seventh century and the middle of the eighth in the Midlands and the South, and in the middle of the ninth century in Northumbria (see Dahl, pp.194-6 & Fulk, History, §415-6). But before this change took place, long final vowels were shortened - relevantly, long final -aé shortened to -æ (see Campbell, §355). After the change shortening long final vowels, the distinction between long and short vocalic inflexional endings would have been lost. Thus Beowulf would have had to be composed before the shortening of long final vowels, which change itself took place before the coalescence of unstressed  -æ and -i. Based on studies of phonological change (see, for instance, Labov), the shortest period of time between the shortening of -aé to -æ and the change of unstressed  -æ to -e one could reasonably expect is a about generation of thirty years (see Fulk §417).

The upshot of these facts is that Beowulf would have been composed before ca. 825 if Northumbrian, or before ca. 725 if Mercian (or East Anglian, following Newton). On the balance, it seems most likely that the poem originated in either Mercia or East Anglia, considering the poet's use of nemne and other southern Anglian dialectal features (see Fulk, History, §361 & §420), as well as other linguistic, cultural and geographic factors (see Newton). Based on other linguistic considerations, Fulk concludes that the poem was most likely composed sometime after ca. 685 (see History, §400).

Thus on the strength of the evidence offered by Beowulf's conformity to Kaluza's law suggests it was most likely composed between ca. 685-825 AD, and probably before 725 AD. The reader should consider other arguments made for a later date as, for instance, in the volume The Dating of 'Beowulf' edited by Chase and in Kiernan's book Beowulf and the Beowulf Manuscript, though the evidence offered by Kaluza's Law (which is, admittedly, somewhat complicated) points to a composition of Beowulf in the late 7th- or early 8th-century.         


Postscript: (November/December 2003)

There remain difficulties for an early date of Beowulf, at least based solely on the evidence offered by Kaluza's Law.

In the Dating of Beowulf volume, one of the most persuasive arguments for a late date is offered by Roberta Frank ('Skaldic Verse'), who points out some apparent 'Scandinavianisms' of Beowulf and the fact that the epithet 'Scylding' does not elsewhere appear to have come into use until the later Anglo-Saxon Period (esp. in connexion with the 'North Sea Empire' of King Cnut). We may say in response that the 'Scandinavianisms' may as also be representative of very early Old English verse (see below on the persistent conservatism of Old Norse), and 'Scylding' as an epithet may have come into use in some early Anglo-Saxon kingdoms which had very early North Germanic influence:-- such as East Anglia which seemed to maintained some contact with Sweden, even if the epithet only appears elsewhere in A-S England at a later time (for more on the Scandinavian character of early East Anglia, see Newton (1993), esp. on the Sutton Hoo ship-burial).

Hutcheson ('Kaluza's Law') demonstrates that, due to the nature of the data, one cannot establish a statistically valid demonstration of significant difference between Beowulf and the other texts (there is a statistical difficulty called 'the short cell problem', which is actually caused by the scarcity of instances of verses violating Kaluza's law in Beowulf), with respect to conformity to Kaluza's law.

Kiernan (personal comm.) has suggested to me that North Germanic linguistic influence (i.e., the mixing of elements of the Old Norse dialects of the 'Vikings' (primarily Danes & Norwegians) into Old English) may also invalidate the evidence of Kaluza's law as attesting to such an early date. Indeed, Norse is in many ways more conservative than Old English; in particular, Norse preserved distinctions between vowels in unstressed positions longer than English did (see Townsend (2002)). The qualitative distinction between vowels in unstressed position is one the bases by which the evidence of Kaluza's law is used to establish an absolute dating (i.e., to a particular span of years), as opposed to a strictly relative chronology of OE verse-texts.

All of these factors require further study, I note them here primarily to point out the difficulty of firmly establishing a date for Beowulf, even with a test as rigorous as the one provided by Kaluza's law. But, whatever the date of composition of 'our' Beowulf may be, the poem contains a higher percentage of verses conformity to Kaluza's law than in the case for any other Old English poem. Verses which would have been metrically correct in the earliest (6th & 7th c.) Old English. Thus the style of Beowulf (whatever its period of origin) remains one of the most archaic of our Old English texts.

[Further notes to be added in future]