This article seeks to identify the phonetic correspondence(s) of the digraph <cg> in Old English (OE) and Middle English (ME), assessing a range of sources: the etyma in early Germanic (Gmc) languages, the various spellings in OE and the spelling evidence in the Linguistic Atlas of Early Middle English. Almost all the textbooks on OE claim that <cg> was pronounced /dʒ/, i.e. as a phonemic affricate, in OE. Evidence is thin on the ground, and the argument rests on certain back spellings <cg> for words with etymological <d+g>, e.g. midgern <micgern>. Words with <cg> in OE go back to Gmc *g(g)j, which subsequently underwent palatalisation, and eventually assibilation and affrication. This article argues that the value [ɟj] is more likely for OE and early ME, and that such an interpretation agrees with the available spelling evidence for both OE and ME, in that there is not one <d>-type spelling in the entire historical corpus until late ME. It is also argued that the development of the voiced (pre-)affricate was later than that of its voiceless counterpart, as voiced fricative phonemes are a late, and infrequent, development in Gmc. Moreover, it is likely that the development of /dʒ/ was affected by the high number of French loans with /dʒ/ which entered the English lexicon after 1066. Thus, the English system of consonant phonemes may not have acquired /dʒ/ until the thirteenth century at the earliest.