Beowulf survived to modern times in a single manuscript, written in ink on parchment, later damaged by fire. The manuscript measures 245 × 185 mm.
The poem is known only from a single manuscript, estimated to date from around 975–1025, in which it appears with other works. The manuscript therefore dates either to the reign of Æthelred the Unready, characterised by strife with the Danish king Sweyn Forkbeard, or to the beginning of the reign of Sweyn's son Cnut the Great from 1016. The Beowulf manuscript is known as the Nowell Codex, gaining its name from 16th-century scholar Laurence Nowell. The official designation is "British Library, Cotton Vitellius A.XV" because it was one of Sir Robert Bruce Cotton's holdings in the Cotton library in the middle of the 17th century. Many private antiquarians and book collectors, such as Sir Robert Cotton, used their own library classification systems. "Cotton Vitellius A.XV" translates as: the 15th book from the left on shelf A (the top shelf) of the bookcase with the bust of Roman Emperor Vitellius standing on top of it, in Cotton's collection. Kevin Kiernan argues that Nowell most likely acquired it through William Cecil, 1st Baron Burghley, in 1563, when Nowell entered Cecil's household as a tutor to his ward, Edward de Vere, 17th Earl of Oxford.
The earliest extant reference to the first foliation of the Nowell Codex was made sometime between 1628 and 1650 by Franciscus Junius (the younger). The ownership of the codex before Nowell remains a mystery.
The Reverend Thomas Smith (1638–1710) and Humfrey Wanley (1672–1726) both catalogued the Cotton library (in which the Nowell Codex was held). Smith's catalogue appeared in 1696, and Wanley's in 1705. The Beowulf manuscript itself is identified by name for the first time in an exchange of letters in 1700 between George Hickes, Wanley's assistant, and Wanley. In the letter to Wanley, Hickes responds to an apparent charge against Smith, made by Wanley, that Smith had failed to mention the Beowulf script when cataloguing Cotton MS. Vitellius A. XV. Hickes replies to Wanley "I can find nothing yet of Beowulph." Kiernan theorised that Smith failed to mention the Beowulf manuscript because of his reliance on previous catalogues or because either he had no idea how to describe it or because it was temporarily out of the codex.
The manuscript passed to Crown ownership in 1702, on the death of its then owner, Sir John Cotton, who had inherited it from his grandfather, Robert Cotton. It suffered damage in a fire at Ashburnham House in 1731, in which around a quarter of the manuscripts bequeathed by Cotton were destroyed. Since then, parts of the manuscript have crumbled along with many of the letters. Rebinding efforts, though saving the manuscript from much degeneration, have nonetheless covered up other letters of the poem, causing further loss. Kiernan, in preparing his electronic edition of the manuscript, used fibre-optic backlighting and ultraviolet lighting to reveal letters in the manuscript lost from binding, erasure, or ink blotting.
The Beowulf manuscript was transcribed from an original by two scribes, one of whom wrote the prose at the beginning of the manuscript and the first 1939 lines, before breaking off in mid-sentence. The first scribe made a point of carefully regularizing the spelling of the original document into the common West Saxon, removing any archaic or dialectical features. The second scribe, who wrote the remainder, with a difference in handwriting noticeable after line 1939, seems to have written more vigorously and with less interest. As a result, the second scribe's script retains more archaic dialectic features, which allow modern scholars to ascribe the poem a cultural context. While both scribes appear to have proofread their work, there are nevertheless many errors. The second scribe was ultimately the more conservative copyist as he did not modify the spelling of the text as he wrote, but copied what he saw in front of him. In the way that it is currently bound, the Beowulf manuscript is followed by the Old English poem Judith. Judith was written by the same scribe that completed Beowulf, as evidenced by similar writing style. Wormholes found in the last leaves of the Beowulf manuscript that are absent in the Judith manuscript suggest that at one point Beowulf ended the volume. The rubbed appearance of some leaves suggests that the manuscript stood on a shelf unbound, as was the case with other Old English manuscripts. Knowledge of books held in the library at Malmesbury Abbey and available as source works, as well as the identification of certain words particular to the local dialect found in the text, suggest that the transcription may have taken place there.
The scholar Roy Liuzza notes that the practice of oral poetry is by its nature invisible to history as evidence is in writing. Comparison with other bodies of verse such as Homer's, coupled with ethnographic observation of early 20th century performers, has provided a vision of how an Anglo-Saxon singer-poet or scop may have practised. The resulting model is that performance was based on traditional stories and a repertoire of word formulae that fitted the traditional metre. The scop moved through the scenes, such as putting on armour or crossing the sea, each one improvised at each telling with differing combinations of the stock phrases, while the basic story and style remained the same. Liuzza notes that Beowulf itself describes the technique of a court poet in assembling materials, in lines 867–874 in his translation, "full of grand stories, mindful of songs ... found other words truly bound together; ... to recite with skill the adventure of Beowulf, adeptly tell a tall tale, and (wordum wrixlan) weave his words." The poem further mentions (lines 1065–1068) that "the harp was touched, tales often told, when Hrothgar's scop was set to recite among the mead tables his hall-entertainment".
Debate over oral tradition
The question of whether Beowulf was passed down through oral tradition prior to its present manuscript form has been the subject of much debate, and involves more than simply the issue of its composition. Rather, given the implications of the theory of oral-formulaic composition and oral tradition, the question concerns how the poem is to be understood, and what sorts of interpretations are legitimate. In his landmark 1960 work, The Singer of Tales, Albert Lord, citing the work of Francis Peabody Magoun and others, considered it proven that Beowulf was composed orally. Later scholars have not all been convinced; they agree that "themes" like "arming the hero" or the "hero on the beach" do exist across Germanic works, some scholars conclude that Anglo-Saxon poetry is a mix of oral-formulaic and literate patterns. Larry Benson proposed that Germanic literature contains "kernels of tradition" which Beowulf expands upon. Ann Watts argued against the imperfect application of one theory to two different traditions: traditional, Homeric, oral-formulaic poetry and Anglo-Saxon poetry. Thomas Gardner agreed with Watts, arguing that the Beowulf text is too varied to be completely constructed from set formulae and themes. John Miles Foley wrote that comparative work must observe the particularities of a given tradition; in his view, there was a fluid continuum from traditionality to textuality.