‘which’(Modern English at)
The source of this word is ultimately PGmc *þat, but whether individual instances of ME at, including at Pe 536 (‘and wyrkeʒ and dotʒ þat at ʒe moun'), represent the rel. pron. at from ON at (with the loss of initial consonant, cp. the cognate OE ðæt) found in N dial of ME and MnE (as Goll takes it) or simply the native word is debated. Most subsequent commentators have followed Gordon and Onions (1932: 135, citing Sisam; see also EVG 536n and McGee), in interpreting the phrase þat at here as a development of þat tat < þat þat (and thus from OE) via assimilation and haplology (and the same explanation can presumably be offered for other ME instances of þat at).
Proposed ON Etymon (OIcel representative)
(1) at ‘who, that’
(ONP (1) at (conjunct.))
Other Scandinavian Reflexes
(1) Far at, Icel að, Norw at, Dan at, OSw at, Sw att
ðæt (conj. and adv.) ‘so that, in order that, after that, then, thence’
Phonological and morphological markers
(1) The rel. pron. at is attested in N dial from the early 13c onwards, surviving into MnE (OED, EDD). See also Discussion by text. (2) The phrase þat at (that at) appears in a handful of MED's citations (s.v. that rel. pron.), mainly N/EM (but notice also e.g. c1450(c1415) Roy.Serm.(Roy 18.B.23)).
Occurrences in the Gersum Corpus
Vant takes issue with Gordon’s argument against interpreting at as (1) (because at ‘does not occur as a relative elsewhere in this group of poems’, EVG 536n) by reading such an instance at Gaw 2205, where he maintains MS at (cp. emendations in TGD etc. and see Onions and Gordon 1933: 179).