Middle English <bacche>, <bache> 'batch', <þacche> 'thatch', and <wacche> 'watch' are structurally parallel. The latter continues the Old English weak feminine abstract <wæcce> 'wakefulness' . This noun is directly related to a weak verb <wæccan> 'to be awake, watchful, vigilant' of which only the participle <wæccende> is attested in West Saxon, though finite forms are found in Northumbrian, including a preterit <ge-wæhte>. The usual WS weak verb in this sense is <wacian>, corresponding to Old Frisian <wakia> and OS <wakôn>, while OHG has <wahhên>. Rather than reconstructing both a Class II *wako:n and a Class III *wake:n to Common Gmc. (thus Köbler, Ae. Wb. s.v. <wacian>), I prefer to see the Class III form as original, preserved in OHG, but remodelled into *wako:jan in Ingvaeonic, where many other Class III weak verbs lost their characteristic morphology. And
rather than explaining <wæccan> as a by-form of <weccan> 'to arouse, waken' (Gmc. *wakjan) whose umlauted vowel was replaced in early Old English by the /a/ retained in <wacol> 'wakeful' and the like (thus effectively Bülbring, Ae. Elemb. §177), I prefer to see a denominative formed from <wæcce> in Old English, well after the period of /i/-/j/-umlaut, by simple analogy with inherited Class I denominatives showing vowel-agreement with their nouns (e.g. <bend> 'band, fetter, string', <bendan> 'to fetter, confine with string, bend (a bow)', from Gmc. *bandiz, *bandjan). In this view the preterit <wæhte> follows the model of <wehte> to <weccan>, <þehte> to <þeccan>, etc., in which the umlauted vowel of the present has replaced the regular stem-vowel of the preterit <weahte>, <þeahte>, etc. And the wk. f. <wæcce> owes its form not to WGmc /j/-gemination but to
Kluge's Law, continuing Gmc. *wakko:n- 'wakefulness' , the protoform being *wog^nó:n-.
ME <bacche> is usually referred to an OE wk. f. *bæcce connected with <bacan> (st. VI) 'to bake'. It is obvious that this noun cannot come directly from <bacan>, any more than <wæcce> can come directly from <wacan> (st. VI) 'to awaken, be born, originate'. The gemination along with lack of umlaut points to a Gmc. *bakko:n- 'action of baking', easily concretized as 'amount produced by one act of baking', i.e. 'batch'. If Greek <phó:go:> 'I roast' is a cognate normal-grade present, the root is *bHeh3g- and the protoform of *bakko:n- must be the zero-grade *bH&3gnó:n-. In any case related Gmc. forms with /kk/ and /k/ are found. OHG has both <backan> and <bachan>, <bahhan> (st. VI) 'to bake', OS has the agent <bakkeri> 'baker', MD has <bakken> (weak) 'to bake', and ON has <baka> (wk. II). I find it implausible that Gmc. had *bako:n, *bakan, and *bakkan in identical senses,
or that two strong verbs with distinct protostems (*bH&3g-é-, *bH&3g-né-) would have been inherited with the same function. Instead, I think a Class II weak denominative *bakko:n 'to make a batch' was formed from *bakko:n- 'batch', and this weak verb interacted with *bakan 'to bake' in different ways as the meanings fell together in the individual languages. In the Low Countries, *bakko:n (later *bakko:jan) prevailed over *bakan. The weak cross *bako:n is represented by ON <baka> and Upper German <bachen>. The strong cross *bakkan competed with the original *bakan in OHG and MHG, and has now prevailed in HG <backen>. In English, *bakan prevailed over *bakko:jan quite early, since the latter is unattested, but ironically the winning verb became weak, with the participle <baken> going obsolete early in the Modern phase. Nevertheless, if my analysis is correct, only English has preserved the noun which generated
*bakko:n in the first place.
In the OED, <þacche> is cited from the Bodleian MS. of Trevisa (1398), but the word is corrected to <thetche> in the printed edition (1495). After 1600 however other citations show the noun <thatch> prevailing over <thetch>. The explanation given is hardly satisfactory, making <þacche> a "late collateral form of <thack> sb." whose auslaut is "conformed to <thatch> vb.", while the latter has "apparently taken its vowel from <thack>". That is, we are supposed to believe that <thack> 'roof, roofing material' (OE <þæc> st. n., Gmc. *þakam) lent its vowel to <thetchen> 'to roof over, cover' (OE <þeccan> wk. I, Gmc. *þakjan), which then lent its inlaut back to <thack>. Such a reciprocal trade, while not impossible, is rendered implausible by the chronology, and seems unnecessary anyway. We can understand <þacche> as the regular development of an OE wk. f. *þæcce,
from a Gmc. abstract *þakko:n- 'roofing', easily concretized, from a protoform *tognó:n-. The noun <thetche> then has been altered from <þacche> under the influence of the regularly inherited verb <thetch(en)> , and since this altered noun dies out after 1600, it may have been nothing but a hypercorrection of printers which never took root in the spoken language. On the other hand the modern verb <thatch>, a simple denominative, has replaced this inherited verb <thetch>, but there are plenty of similar examples. As with <wacche> and <bacche>, the recognition that the geminate in the form ancestral to <þacche> is due to Kluge's Law, and not to WGmc /j/-gemination, avoids the difficulty of accounting for the missing /j/-umlaut. In this case it spares us the scenario of nouns and verbs in late ME picking each other's pockets for substitute phonemes.