The historical phonology of Old English: a critical review

There is a widely accepted chronology of sound laws, covering the transition from Proto-West Germanic to Old English, found in every handbook of Old English. This chronology contains sound laws whose only function is to cancel the effects of previous ones, such as ‘retraction’ and ‘smoothing’, reversing ‘fronting’ and ‘breaking’. This chronology of sound laws is allocated to the relatively short period between the arrival of the Anglo-Saxons in the 5th century and the oldest Early Old English sources around 700. On close scrutiny, some aspects of the theory turn out to be problematic: the series of sound laws is fairly unique in the history of Germanic languages; some of the sound laws are phonetically unlikely (e.g. ‘Anglian smoothing’); the extensive, sometimes repetitive, sequences (up to 5 stages) of forms in only 250 years seem hardly realistic; none of the questionable developments is positively confirmed by runic evidence; the theory requires the interpretation of many attested forms as ‘merely’ spelling issues or signs of dialect mixture, instead of evidence of historical changes. This article offers a detailed discussion of these problematic issues, to conclude that the theory is in need of revision.


Introduction and central issue 1
There is a widely accepted chronology of sound laws, covering the transition from Proto-West Germanic to Old English, starting with Luick (1914-40) and later modified by scholars such as Campbell, Brunner, Fulk, Stiles, Hogg and Ringe. This comprises sound laws such as fronting, breaking, i-mutation and Anglian smoothing (for the details see section 1.3). I will refer to this theory as the Standard Theory (ST). Extensive overviews are given by Campbell (1977, 50-112), Hogg (2011Hogg ( [1992, 74-168) and Ringe & Taylor (2014, 167-336).
Adjustments to the standard chronology have been put forward on various occasions, e.g. by Ball & Stiles (1983), Toon (1987; referring back to criticism by Kuhn from 1939) and Toon (1992). Krupatkin (1970, 50) is bolder in his criticism and mentions an implausible "'zigzag' a > ae > a" when discussing the theory of restoration of fronted PGmc *a. Doubts have also been raised by Kortlandt (1999Kortlandt ( , 2008. Kortlandt (1999, 46), referring to the various stages as given in Fulk (1998), formulates his objections against this chronology in the following way: The main difficulty with [the Standard Theory]'s chronology is the unmotivated character of the sound changes: we find backing at stage l, fronting at stage 3, backing at stage 5, fronting at stage 6, backing at stage 7, fronting at stage 9, and backing at stage 11. What was the driving force behind these alternating developments?
An illustration of the current discourse is a paper by Howell & Somers (2008), who discuss Anglian smoothing. They state that this process is " […] difficult for scholars to adequately explain in either theoretical or phonetic terms." They observe that the forwardbackward movement as in *nēh > *nēoh > nēh 'near' is an "apparent paradox" (p. 187). The historical reality of smoothing is, however, not seriously questioned by the authors. An entirely different approach is offered by Schrijver (2014, 70), who suggests an influence from a Celtic substratum as the trigger of many of the Old English vocalic changes. The resulting imposition of the phonological system would cause a fairly abrupt change in the vowel system by imposing Celtic pronunciation habits on West Germanic vowels, with the result looking like the outcome of a complex set of consecutive sound laws. 2 A common stage or common chronology for Old Frisian and Old English is explicitly denied in the traditional approach e.g. by Campbell (1977, 52, §131), despite the many similarities between the two languages. Stiles (1995, 211) states that: "[…] it is not possible to construct the exclusive common relative chronology that is necessary in order to be able to establish a[n Anglo-Frisian] node on a family tree." Archaeological evidence, however, supports the idea that the Germanic settlers of Britain and the post-Migration Frisians have a shared origin and maintained close cultural contacts during the first three or four centuries after the landnam ('adventus Saxonum'; Colleran 2016, Nicolay 2005, 2017, Hines & Behr 2019. A comparison between various forms of Old English and Old Frisian shows that in terms of phonological shape it is especially the West Saxon form of Old English that deviates strongly from the continental sister languages, Old Frisian and Old Saxon, rather than the language of the early Northumbrian fragments, as will be illustrated further in section 1.4. The current descriptions of the phonological history of (pre-)Old English pose a number of methodological problems in the current theory, which will be discussed in the rest of this paper. Given the gravity of these objections, it can be concluded that the Standard Theory needs revision. An alternative reconstruction of the historical phonology of pre-Old English, based on the language of the Épinal and Erfurt Glossaries and the early Northumbrian material, will be presented in a second article (Versloot, to appear).
My paper is organised as follows: section 1.2 provides a more detailed formulation of the research question, namely: what should a theory of pre-Old English phonological changes cover and what should it exclude, and how can claims about the phonological changes be evaluated? The rest of section 1 addresses a series of general methodological issues relating to the Standard Theory. Section 2 discusses various contradictory aspects of individual sound laws that are part of the Standard Theory. The results of the analysis are summarised in a concluding section 3.
Hogg's A grammar of Old English Vol. 1, phonology ([1992] 2011) is a relatively recent presentation of the historical phonology of Old English with extensive argumentation and summaries of controversial issues relating to what I call the 'Standard Theory'. In order to limit the discussion of this theory, I confine myself largely to Hogg's descriptions and argumentation, occasionally extended with references to Campbell's (1977) presentation of the theory and the more recent overview by Ringe & Taylor (2014). My criticism should, however, not be understood as being specifically targeted against Hogg's interpretation of the facts.

Delimiting pre-Old English
The beginnings of the linguistic developments of Old English lie in the 5th century, when Germanic speaking peoples invaded present-day England on a larger scale (Higham & Ryan 2015, 70-125, Hines & Behr 2019 and resettled Frisia (Nieuwhof 2016, Nicolay 2017. The linguistic communalities between Old English and Old Frisian had not gone unnoticed, e.g. Siebs (1889). The 'old-school' way of reasoning about so-called Anglo-Frisian communalities found its climax in Schwarz's (1951) Goten, Nordgermanen, Angelsachsen. Schwarz formulates what seems to be the communis opinio about the origins of Anglofriesisch: "Es wird mit Recht geschlossen, daß die gemeinsamen Züge des Fries. und Ae. bereits auf dem Festlande ausgebildet waren." [It is rightly concluded that the common features of Frisian and Old English had already been established on the Continent]. The Anglo-Frisian hypothesis of a common idiom brought to England by the Germanic invaders was severely criticized in Hans Kuhn's (1955) seminal article. He argued for a fairly undifferentiated North-West Germanic language, prior to the movements of the Great Migrations. The relationships among and differentiation of the Germanic languages were carefully analysed in the work of Nielsen (1985Nielsen ( , 1998Nielsen ( , 2000, who shows a nuanced picture of a couple of early changes leading to a gradual differentiation between North Germanic, North Sea West Germanic and Continental West Germanic, before, during and after the Anglo-Saxon landnam. This delicate and complicated process, without sharp boundaries, but largely taking place in a language continuum, was also described in Stiles (2013). Most of the 'classical' North Sea Germanic features fall in the period during and after the landnam; Stiles (2013, 32) calls them 'Post-West-Germanic'. Little to nothing can be ascribed to a specific Anglo-Frisian idiom on the Continent, prior to the year 400. Interestingly, the core of these ideas can already be found in Chadwick (1889, 264-265).
The Standard Theory claims that essential steps such as fronting, retraction, breaking and i-mutation are prehistorical, i.e. were completed before the first attestations of Old English (Campbell 1977, vii, Hogg 2011. A theory of prehistoric (pre-attestation) developments in pre-Old English has the task to describe the developments in the period between reconstructed Proto-(West) Germanic and the earliest attestations of Old English. The early Anglian sources, encompassing the Mercian Glossaries and the early Northumbrian material (Toon 1992, 427) are the oldest substantial evidence of Old English. They date from the late 7th and early 8th centuries. Their exact attestation date is not relevant; what matters is that they are the oldest extensive written evidence of Old English. Some earlier runic finds may provide additional insights in the earlier period. The West Saxon sources represent later developments of Old English, most of them nearly 200 or more years younger than the oldest sources and they are therefore not the primary target of a reconstruction of prehistoric sound changes. There are two alternatives to interpret the differences between the early Anglian sources and later West Saxon: the phonological shape of West Saxon can be understood as the result of later sound laws on top of the prehistoric ones, applying to all pre-Old English dialects; this poses the question as to what extent the differences between Anglian and Saxon can be chronological rather than diatopical in nature -West Saxon can only be understood correctly with a (somewhat) different chronology of prehistoric sound laws In the first interpretation, an alternative theory can leave the reconstruction of these later developments open for further evaluation; in the latter case, a revision of the current chronology of prehistoric sound laws based on the mentioned sources, strictly speaking, only applies to the varieties they represent. An alternative reconstruction of the pre-Old English sound changes in Anglian, based on the development of PGmc *a in the oldest sources, is attempted in Versloot (to appear).
This article will evaluate the current Standard Theory with a view to the attested linguistic variation in (early) Old English and with reference to the early Anglo-Saxon runic inscriptions, which are expected to attest to the changes described as 'pre-Old English'. This evidence-based approach contrasts with Campbell's conviction that it is "[…] not possible to date any of these sound-changes by observing their gradual appearance in texts, and we can establish their approximate dates and arrange them in chronological order by theoretical means only." (Campbell 1977, 106, §246). Campbell's hypothesis is not falsifiable and allows for the atypical chronology of the Standard Theory, discussed in section 1.3.

The typological-chronological inversion in the arrangement of sound laws
The total number of sound laws for the traditionally distinguished varieties of Old English shows a remarkable inversion. The nulhypothesis is that a younger variety of a language is separated from the proto-language by more sound changes than an older variety. Under the interpretation of the Standard Theory, the situation for Old English is reversed. Table 1 presents the pre-Old English sound laws for the Old English dialects according to the Standard Theory. Note that the table contains only those items that are relevant for this discussion: general early sound laws such as the loss of nasals before fricatives are not included here. The relative order of sound laws is based on Hogg (2011), who differs only marginally from Campbell (1977, 109) or Luick (1914-40).  Table 1 shows that according to the Standard Theory, early 8th century Northumbrian and Mercian went through nine or ten 3 sound changes to end up with a shape of the language that hardly differs from Continental Old Saxon, as will be illustrated in section 1.4. To obtain this archaic linguistic shape, four sound laws, marked in the table with a grey shading, are only there to reverse the effect of earlier changes. In the approach of Campbell (1977, 55, § 143), the limited fronting of /a/ > /ae/ in Anglian, namely not before -lC (I leave out the context before -w here for reasons explained in section 2.2.2), is split into two sound laws: first a general fronting followed by another reversing sound law before -lC. For the more than 150 years younger Early West Saxon only seven changes are required, one of them still to reverse the effect of an earlier one. The setup of sound laws is summarized in Table 2.

Early Anglian Early West Saxon
Older Younger More Archaic shape More Innovative shape

More Sound Laws Fewer Sound Laws
Multiple 'Repair' Sound Laws One 'Repair' Sound Law This is a very unusual situation from a typological and methodological point of view. The expectation would be that the substantial differences between the early forms of Continental West Germanic and West Saxon, compared to Mercian and Early Northumbrian, are the result of later and additional sound laws that came on top of a fairly restricted set of changes that can be reconstructed for the oldest Anglian varieties. In particular, the 'repair' sound laws are typologically unusual, and I am not aware of any similar configurations in the history of other Germanic languages. It looks as if the theory of the early phonological history of Old English was built on the extensive West Saxon evidence and especially on the later sources, given the frequent problems with seemingly Mercian traces in Early West Saxon (see examples in the discussion in sections 1.6 and 2). To get it right for the (older) minor sources, a couple of 'repair' sound laws have been added to the theory.

Early Northumbrian and Continental Old Germanic
The unlikeliness of the standard chronology can be illustrated by a comparison between words from Caedmon's Hymn (CH) in the early Northumbrian versions with the West Saxon version (Tanner Ms. 10) and the corresponding forms in Old Saxon and Old Frisian. Many of the differences between Northumbrian and West Saxon concern the appearance of diphthongs in West Saxon versus monophthongs in all the other varieties, as can be seen in Table 3.  At first glance, there are very few differences between the Nhbr. forms and the Old Saxon forms in particular. The only substantial difference between the two languages comes from the more widespread implementation of i-mutation in Old English. Old Saxon has primary i-mutation (applying to short PGmc *a only), as in OE aelda OS eldea < PGmc *aldiō-against OE dryctin -OS drohtin < PGmc *druxtinawith secondary i-mutation. Both languages show the monophthongization of PGmc *ai, even with similar outcomes as evinced by OE ǣrist, hālig -OS ērist, hālag (next to more frequent hēlag). There is no a priori reason to assume that words that continue the PGmc vowels nearly unaltered into Old Saxon, such as in (u)uerc, would be the product of fronting, breaking, retraction and smoothing processes in Old English. This is, however, what the Standard Theory claims, as illustrated in Table 4 Table 4 illustrates the redundancy of the Old English sound laws from the perspective of the early Northumbrian evidence when compared with Old Saxon. All of the vowels that match the conservative Old Saxon vowel qualities (uard, uerc, fadur, aelda, barnum) supposedly all went through at least two intermediate steps. The shape of the vocalism of the words in Table 4 in Old Saxon and Early Northumbrian can be reached by only applying i-mutation; the effect of the other Old English sound laws is only there to cancel out each other's impact. The Standard Theory seems, however, to be tailor-made for the 200-year younger West Saxon, which encompasses the largest proportion of attested texts in Old English.

WS OE
In this paper, I will work with an alternative hypothesis, following more or less the chronology of sound changes for Old Frisian (Bremmer Jr 2009, 24-38), which is much simpler: 2. i-mutation 3. breaking of *a in specific contexts in the 7th and 8th centuries -only there where it is actually manifested in the written texts -a process that is the source of the synchronic alternation between gard and geard in the two Early Northumbrian versions of CH The details of this alternative, Anglo-Frisian (AF) chronology are presented in Versloot (to appear).

Ignored or lacking evidence
Available textual counterevidence for the Standard Theory has been ignored on various occasions. Attested spellings which are not in line with the idea that developments like fronting, retraction and breaking are considered prehistoric and pre-literary are consistently reasoned away by assuming a divergence between spelling and pronunciation, or they are dismissed as 'spelling errors', 'scribal confusion' or resulting from a mixture of different 'established spelling traditions'. In this section, I will present a couple of examples of ignored evidence or lacking evidence. More examples are presented in the detailed discussion in section 2. Section 1.7 addresses the way scribal practices are invoked as explanations for forms that do not fit the Standard Theory.
With reference to breaking, presumably an old and prehistoric development (Campbell 1977, vii), Campbell notes that "[i]n some early texts, forms occur in which the glide developed by breaking is not indicated in the spelling" (Campbell 1977, 54-55, § 140). He explicitly states in a footnote that unbroken forms in Bede appear in early manuscripts. This is confirmed by Hogg (2011, 83, fn. 5), and van der Schee (2015) reasons in the same way in a recent article. Campbell writes about the breaking of 'Prim. OE ae' thus: In Kt. and W-S […] while a appears freely for Prim. OE ae in early texts, ea rapidly asserts itself as the prevailing spelling. Thus, in Kt. charters of before 800, ea appears once only, Uuealhhunes g.s., Ct. 8 (dated 770), but in the following century ea becomes increasingly more common and appears practically always in the tenth-century KG. Similarly, in W-S, in earlier texts ea and a both appear, e.g. in the part of P[arker] C[hronicle] written by the first scribe ea occurs 29 times and a 104 times in accented syllables; but in late W-S ea becomes exclusively used." (Campbell 1977, 55, § 143). 5 The variation is apparently considered a matter of mere spelling, because the breaking to /ɛa/ is supposedly pre-Old English, and the observations are not taken as indication that breaking was possibly only spreading at the time of the attestations mentioned.
An example of lacking evidence for an assumed prehistoric sound change can be found in Hogg's Grammar of Old English 1. In section 5.23 (p. 86), Hogg writes that: the long vowel /e:/ is broken to /eu/, by the time of the earliest texts. This narrows down the body of evidence quite drastically: despite the postulated breaking of /e:/ before (earlier) x, the default realization in Anglian is the monophthong(!) ē. The evidence explicitly mentioned as proof of breaking (Hogg 2011, 87), with assumed eo < *ēx without smoothing due to the loss of *x (PsGlA 6 neolaecan 'to approach', neowest 'nearness' < PGmc *nēx-westi-and Kentish nior, neor 'nearer'), can be explained in a different way, as will be discussed in section 2.2.2. Scholars seem to ignore the fact that the oldest attestation in an Anglian variety of a word that is supposed to prove the early date of breaking is without breaking, namely neweste in Charter 98 from ca. 745 (as attested in the DOEEC). 7 It is not until 100 years later that we come across neoweste (PsCaA 1, PsGlA, from ca. 850). More logical would be to interpret the form neoweste as a later development from neweste. The same text, PsGlA, is said to contain another piece of evidence, namely nēolǣcan. I didn't find this exact form, but one comes across forms such as geneolaeceð or geniolaecað, but also genehlaecað. The last form contradicts the idea that the loss of x (and hence the lack of smoothing) are particularly old and prehistoric. The form with neolooks rather like a vocalization of /x/ as in PDE low < ME lāh. This would be in line with the interpretation of 'breaking' as a phonetic process of weakening of the consonant, developed by Howell (1991). 8 The form unneg 'un-near' on the 8th century Franks Casket (FC) is supposedly the product of breaking and smoothing developments, but it could just as well show that breaking (in Northumbrian) is rare and late, and does not apply to ē. Toon (1987:283) concludes on the basis of the material in the Épinal and Erfurt Glossaries (EpGl and ErfGl) (Pheifer 1974, Herren, Porter & Sauer 2020) that smoothing of /eō/ was not a prehistoric sound law, which contradicts the Standard Theory's statement about the prehistoric nature of both smoothing and then also breaking of */e:/ -if smoothing hadn't taken place, a form with <e> cannot be derived by the application of smoothing and hence the <e> attests to an originally unbroken vowel. Therefore, the form unneg on FC rather shows that such a breaking never took place in Anglian, at least not in Early Northumbrian.
The reasoning found in Hogg (2011, 87) in the discussion of breaking of short /i/ seems unnecessary complicated to me. For breaking before */x/, it is mentioned that "[…] most forms are found with i-umlaut of the breaking diphthong [...]". WS sihþ and Angl. sīþ 'sees (3.sg.)' < PGmc *sixiþ are presented as evidence for earlier breaking ("*sihiþ > *siuhiþ"), whereas the WS form shows in fact only syncope of the second vowel and the Angl. form attests to an early loss of the intervocalic *x and subsequent contraction. The only positive evidence for breaking comes from Kentish siohþ, 9 which resembles OFri. siucht, where breaking before /xC/ took place in the 7th or 8th century. A similar overinterpretation of the evidence is found in the examples WS filhþ, Angl. fīleþ < PGmc *filxiþ, which are supposed to support the earlier existence of a never attested form "*fiulhiþ". While none of the attested forms, except for the Kentish form siohþ, shows any real breaking, we are supposed to believe that broken forms existed at the intermediate stages in all varieties of Old English.
An interesting example of the different ways in which scholars deal with attested evidence comes from the discussion of the Caistor-by-Norwich knucklebone inscription raïhan, dated to the 5th century (Looijenga 2003, 284). Ringe & Taylor (2014, 171) state that "[…] it might […] exhibit inherited *ai with no change", which would provide an important datum post quem for most of the sound laws mentioned in Table 1. 10 Page (1994, 107) thinks that the inscription "might still be convincingly Scandinavian", a conclusion stipulated by the singlebarred h, in order to fit Luick's hypothesis -"in complete absence of direct evidence" (sic!) (Page 1994, 107) -that the transition from PGmc *ai > (pre-)OE ā should be dated in the 3rd or 4th century. 11 The 9 DOEEC does not contain any such Kentish form, only forsiohð in OccGl 49. 10 Compare Campbell's rejection of the possibility to date the changes, mentioned in section 1.2. 11 It is in fact highly unlikely that the form with the diphthong can be of Scandinavian origin. The *ai was monophthongized in Scandinavian exactly in the position before *x as evinced by various runic inscriptions. This process was already completed around 400, which makes a Scandinavian origin of a 5th centruy Anglian inscription with aï improbable (Versloot 2017, 294-296). in my view more favourable approach by Ringe & Taylor leads to the conclusion that all the developments in Table 1 must be squeezed into a period of no more than 250 years: the raïhan inscription in the early 5th century and the archetype of the Mercian Glossaries from the late 7th century (Pheifer 1974, lvii, xci).

The Parker Chronicle (Chron(A))
The first part of the Early West Saxon Parker Chronicle (PC) (Chron(A)) contains frequent spellings of <a> before -lC, which goes against the prediction made by the Standard Theory, namely that breaking of PGmc *a in West Saxon before -lC was pre-Old English. This contradictory evidence is, however, ignored (see the quotation from Campbell in section 1.5) or interpreted as a sign of 'Mercian influence' (Sprockel 1965, xix, Ringe & Taylor 2014. I analysed the text of the PC for all instances of <(e)al>+C. I distinguished hand 'A' (ca. 900 AD) from the other scribes, following Sprockel (1965, xviii-xxii). Hogg (2011, 129) describes the language of Hand A as one of the "[….] EWS texts which show signs of being (influenced by) an a-dialect", leaving it open whether this a-dialect was Mercian or a variety of West Saxon. Campbell (1977, 110, §258) states that "[s]pellings with a penetrate eW-S and eKt. extensively […]", due to "mutual influence of established orthographical systems". For each of the authors it is beyond doubt that breaking in this context must have been implemented in genuine West Saxon long before the Parker Chronicle was written. None of them considers the option that PGmc *a had not yet been broken in West Saxon at the time of the attestation of hand A.
The first section of the chronicle contains unbroken spellings in e.g., aldormann 'alderman' or salde 'sold'. My analysis focusses on instances of PGmc *a followed by -rC and -lC without i-mutation. A couple of lemmas can be found both in part 'A' and in the rest of the chronicle, rendering a comparison possible. Some morphemes appear in various personal names (PN), such as -wald (e.g. Oswald, Ecgwald) or Ealh-(e.g. Ealhmund, Ealhstan); -w(e)ard appeared both in names (e.g. Eaduuard, AEþelweard) and in adverbs (e.g. easteweard); see Table 5 for a summary of the distribution of attested forms.  The words with ample attestation in the entire chronicle and little or no breaking in hand 'A' have been marked with a grey-tone in Table  5. They all contain -ld or -ll clusters. Names with -wald as the second element lack breaking in all sections of the PC: the last, mostly unstressed, part of a word is less inclined to show breaking. The same tendency can be observed in the sequence -alh in the data from hand 'A': PN-walh appears only with a, whereas Ealh-PN is always spelled with a broken vowel <ea>. The contrast between first and second syllable is not relevant for the sequence -ard/þ, which shows breaking of *a from the very early parts of the PC, hand 'A' onward. The relevance of word structure can be illustrated by the contrast between the simplex healf (always with breaking in hand 'A'), and H(e)alfdene, which appears once with Half-. By the same token, the scribe of hand 'A' is more reluctant to apply breaking in the polysyllabic word aldormann, than in the simplex words h(e)aldan and s(e)alde/-on.
I tested all words with -r/l+d/ð/þ in the entire text of the PC for the following features: -r vs. -l, hand 'A' vs. the rest of the text, the position of the a in the word (first part of a compound or derivation vs. latter part) and the number of syllables (excluding inflectional endings for otherwise monomorphemic stems) in a logistic regression analysis. All evaluated features contributed significantly to the contrast between ea and a (see Appendix). For -r vs. -l and the text fragment, this comes as no surprise. The latter two factors comply with a hypothesis that breaking may have emerged from short *a that was lengthened before specific consonant clusters (see Versloot, to appear) -lack of stress in a second syllable prevents lengthening, while the word-isochrony principle constrains lengthening in polysyllabic words, even when the a appears in the stressed first syllable.
It is noteworthy that the alternation is always between <al>/<ar> and <eal>/<ear>, as in aldormon ~ ealdormon; next to 3 tokens with (-)bearn, there is one barn, and next to 36 tokens healf-, there is one half-, but no *baern or *haelf-12 ; compare also EpGl fosturbearn and ErfGl fo[s]tribarn, but no *baern, and Caedmon's Hymn 'Moore' -geard vs. 'Leningrad' -gard. This suggests that ae and ea exclude each other rather than that they are part of a chronological-phonological sequence.
Finally, one may consider the potential Mercian origin of the -alCspellings (Hogg 2011, 129). Hogg (2011, 157-158) explicitly ascribes non-WS spellings in the chronicle to Mercian influence. When comparing the progression of breaking in EpGl and ErfGl with the distribution of broken and non-broken forms in hand A, one can observe clear differences. The Glossaries show no breaking before *-lx or *-lf, as in salh 'sallow (tree)' or scalfor 'diver (bird), whereas Table 5 shows that these contexts are breaking contexts for the scribe associated with hand 'A' of the PC. There is no reason why a scribe, influenced by Mercian, would show this influence selectively. Moreover, it is hardly reconcilable with a scribal spelling influence interpretation that the observed variation between a and ea by hand 'A' follows statistical patterns, based on phonological principles. It seems more logical to consider the text of hand 'A' in the Parker Chronicle as a genuine form of early-West Saxon that shared some but not all of the distributional features of breaking with the language of the early Mercian Glossaries. The breaking appears significantly later in the Parker Chronicle in contexts that are resilient to phonetic lengthening, which lends support to the breaking-through-lengthening hypothesis. More important is the implication that breaking was an ongoing phonological process around 900 in West Saxon and did not take place in the prehistoric periods of English. The alternations between broken and unbroken forms in Caedmon's Hymn and the Mercian Glossaries (from around 700) and the unbroken forms in early manuscripts of Bede's Historia (see section 1.5), indicate that breaking in Old English had begun only briefly before, in the 7th century.

Scribal errors and practices
In order to explain discrepancies between the theory and the attested spellings, the Standard Theory often takes recourse to the idea of widespread spelling errors on the one hand and a mixture of established spelling traditions on the other. In all, Hogg invokes 'scribal error(s)' and 'scribal confusion' 15 times in his chapter 5, which deals with the vocalic changes.
The result is a myriad of early Mercian scribal influences on Early West Saxon and Kentish and an increasing impact of West Saxon on the other varieties of written Old English at a later date (for a nuanced discussion of this issue, see Toon 1992, e.g. 443-444). I do not deny the potential for an impact from scribal centres (see e.g. Rem [2003] for a thorough analysis of this effect for Middle Dutch in Holland), or the long path that some texts had from copy to copy, but the proponents of the Standard Theory apply these arguments in a sometimes very inconsistent and eclectic way (cf. the presumed Mercian patterns in the Parker Chronicle). The use of a phrase like "no doubt" often reflects circular reasoning rather than well underpinned arguments. 13 I will discuss three examples of this kind of problematic reasoning in more detail below.
Campbell ascribes the alternation between <ae> and <e(e)> in the early Mercian glossaries to "imperfect differentiation" of <ae> and <e(e)>, which were said to be "equivalent in the contemporary spellings of Latin" (Campbell 1977, 52, § 128 fn. 2; in a similar vein Hogg 2011, 61). The EpGl contains ca. 25% <ae>-spellings for PGmc *ē 1 . If these are random, equivalent spelling variants, one could expect a similar 'confusion' in the spelling of the <ae> from PGmc *ai with i-mutation. However, here the spelling <e(e)> appears in only 1 instance in EpGl (n = 23). This indicates that the alleged scribal confusion was constrained by knowledge about historical phonology, which is an unlikely scenario. The same pattern was observed in the previous section about the <a>spellings in the Parker Chronicle. Given the fact that the 'misspellings' for PGmc *ē 1 reflect the reconstructable pre-stage *ǣ, it seems much more logical to interpret the <ae> spellings for *ē 1 as a witness of the 13 There are 22 instances of the collocation "no doubt" or "undoubtedly" in Hogg's chapter 5, discussing the Old English vowels. last stage of the transition ǣ > ē. This transition is thus attested in the earliest Anglian writings and can therefore not be prehistoric. 14 Earlier scholars seem very convinced about such "[…] confusion of <ǣ> and <ē>." (Hogg 2011, 129, fn.1). 15 The spelling with <ae> for PGmc *au with i-mutation in Bede is supposedly "purely graphic", an opinion apparently also held by Luick, while the instances of <e> in early WS texts are "no doubt merely a sign of Merc. influence on EWS." (Hogg 2011, 129). The idea that WS <i(e)> might have developed from ē (as it e.g. did later in English in the Great Vowel Shift) is not given any consideration.
'Ad hoc' (Ringe & Taylor 2014, 183-4) argumentation by Hogg can be found in the discussion of uuiurthit 'becomes' in Bede's Death Song (BD). Hogg (2011, 91, fn.2) considers it to be a deliberate misspelling for *uuuirthit, in order to avoid the sequence <uuu>, where he argues that <ui> "of course" represents /y/. But Bede's Death Song and other Early Northumbrian sources contain many instances with <y> for /y/, such as CH dryctin 'Lord' < PGmc *druhtina-or BD yflaes 'evil' < PGmc *uƀila-, so there is no reason to assume an alternative spelling *<ui> here. The spelling with a sequence of <uuu> is found in thriuuuintri 'three winter' in EpGl. Bede's Death Song contains two instances of breaking of PGmc *e: uueorthae and uuiurthit, forms of the verb 'to become', where PGmc *e stands before /r/ + a voiced, non-velar consonant. The contrast between the diphthongs <eo> and <iu> can be understood in terms of height harmony with the following vowel: eo ~ ae and iu ~ i.
14 The confusion about the <ae>-spellings is emblematic. Pheifer (1974, lx) claims that Campbell (1977, 52, § 128 fn. 2) regards them as archaisms, which I can only understand as meaning that they are the reflections of an earlier *ǣ-stage in Mercian. But then Pheifer continues, stating that <ae>-spellings "reflect […] early uncertainty in the representation of non-WS. [e:]". The claim made is that the sound value was definitely "[e:]", which is reiterated by the statement that the development from *ǣ > [e:] "was certainly pre-literary", with a reference to Luick. Pheifer can be interpreted as saying that the scribes used only [e:] in their speech, but that they got confused by the nolonger existing, pre-literary pronunciation with *[ɛ:]. This presumes an unlikely knowledge of historical phonology among the scribes or the existence of an unattested corpus of earlier writing. Taking Campbell (1977, 51, §130 fn.2) at face value, one has to conclude that there is no evidence that Campbell considers these spellings to be archaisms. Campbell only says that they are "[…] probably due to imperfect differentiation of the symbols […]", so purely spelling without implications for the chronology. 15 Note the real confusion here, between the phonological interpretation ǣ (or rather: /ae:/) and spelling <ae>. Length was hardly ever marked and never spelled with macrons: <ǣ>, <āē>.
A theory that stays closer to the attested forms and needs less help from the discussed type of reasoning is to be preferred in my perception; see Toon (1992 esp. pp. 447-448) for a similar inclination towards a chronological interpretation of linguistic variation in manuscripts.

Detailed criticism of the Standard Theory
The Standard Theory, claiming a chronological order with developments taking place before the first attestations of English around 700, turns out to have various flaws, which, taken together, render this theory problematic. Here, I follow Hogg's (2011, 74-149) overview, referring to the section numbers in his treatment.

First Fronting and related issues (Hogg § 5.3-15)
The fronting of PGmc *a is clearly established in the history of English and Frisian. It is the structural-phonological mirror-event of the rounding of *a. Nasalization and rounding of PGmc *a before nasals is widespread in Old English, Old Frisian and Old Saxon, but at the same time phonologically unstable (for Frisian see Versloot 2014a), with o ~ a-alternations present in all languages (Ringe & Taylor 2014, 142-146). A couple of issues relating to the details of the First Fronting are discussed in this section.

Widespread fronting
A couple of words with PGmc *a are attested in early Anglo-Saxon runic inscriptions. Note the following inscriptions, containing the ansuz or aesc-rune (Waxenberger 2013(Waxenberger , 2019: positively futhorc characters and therefore k seems best interpreted as *aek(k)ō, with a fronted ae vowel, possibly before a geminated consonant, followed by a back vowel, cf. OE Acca (e.g. in Bede 3 & 5 and Chron D,E,F). The form reads as a masculine n-stem, pre-Old English *AEk(k)ɔ (Nedoma 2016, 10, fn.8) < *ak(k)ōn/-an-. Nedoma also mentions a Kentish Eacca. 17 Retraction or -in a different approach -absence of fronting before a back vowel is common but not uniform, with an intervening geminated consonant (Hogg 2011, 94 Therefore, if we assume a date of early 6th century for Spong Hill, Chessel Down II and Loveden Hill, applying a reading aelu */aelu/, aeko */aek(:)ɔ/ and baed */baed/ respectively, one may claim that there is runic evidence for general early fronting before 500, also before back vowels. That leaves the 6th century for the implementation of breaking, followed by retraction, as witnessed by Old English Acca, in the course of the 6th century, to be completed before i-mutation at around 600.
However, one could just as well claim that Spong Hill should be read alu */alu/, indicating that PGmc *a was simply never fronted before back vowels (the interpretation applied for Frisian), whereas Chessel Down II can receive the interpretation */aek:ɔ/, a context that shows fronting in the Mercian Glossaries, without any need to invoke a later retraction stage. The form Acca would represent the non-fronted version, which fits the ambiguous impact of the context *aC1C1Vback. Where Kentish Eacca should be derived from depends on the exact interpretation of breaking. Nedoma's interpretation of Loveden Hill, with b d, where represents an *[a] without any fronting and a specific onomastic loss of the final vowel *-u (Nedoma 2016, 18), fits in here as well. It implies the sequence *badu > -bad, with simple blocking of fronting before the following *-u in the 5th century and before the final vowel was lost, no later than 550. Otherwise, one would have to squeeze the whole development of fronting, breaking, retraction and loss of *-u into the period of 450-550.
Unfortunately, we have to conclude that the runic inscriptions do not provide decisive evidence for either of the two chronologies. Whitby I shows the well-known fact that PGmc *a was not fronted in Anglian before -lC, which is not a controversial interpretation in any approach.
The crucial issue is the ambiguous interpretation of the -rune and the potential allophonic realizations 'underneath' the use of this character, in particular advocated by Waxenberg. The proponents of the Standard Theory avoid mentioning absolute dates for the various steps: fronting, breaking and retraction (see Campbell's statement, quoted in section 1.2), but from other indications, discussed earlier (raïhan and the dating of i-mutation), we can conclude that these three processes must be fit between 450 and some time before 600. An additional complicating factor is the time span of the various dating methods. 19 Unfortunately, there are no words attested with a fairly certain reading in contexts where the Standard Theory predicts a broken vowel as an intermediate stage (cf. Table 4). 20 Early and widespread fronting is assumed to account for broken forms, because the breaking product ea is supposed to be derived from *ae (see Table 4). One of the puzzling aspects of this theory is that a 19 For examples of dating issues in general see, e.g., Imer (2011), Waxenberger (2011. 20 I have identified three potentially relevant inscriptions, but they are all not particularly old and they all have interpretive issues: -Mortain Cascet: gewarahtae 'made' (Looijenga 2003, 289), ca. 700-900 (RuneS: s.v. OE-GB-61), a past tense of OE wyrcan; Hogg & Fulk (2011, 275) mention similar forms, all 'Anglian', which is not the core region of breaking; -Carisbrooke Mount (Isle of Wight): ..gae(r/u)ae.. (Looijenga 2003, 293), ca. 775-825 (RuneS: s.v. OE-GB-15). This reading is doubtful, but if correct, one would expect a broken vowel before *rw in a southern dialect (Isle of Wight).
The pattern of attestations suggests that ea-developed directly from a-.
A similar observation was made in section 1.6. About the retraction of ae before l, Campbell writes: " § 143. ae retracted before l followed by a consonant in Angl. texts, including the early glossaries" (Campbell 1977, 55). Hogg (2011, 79-81) leaves the option open that PGmc *a had never been fronted before *l in Anglian, whereas Ringe & Taylor (2014, 185-186) adhere to Campbell's view that *a was first fronted and later retracted before *lC. Both Campbell (1977, 55) and Hogg (2011, 81) signal the fact that early WS texts show more <a> than <ea> before /l/. Campbell (1977, 110, §258) states that "[s]pellings with a penetrate eW-S and eKt. extensively […]", due to "mutual influence of established orthographical systems". Considering the early texts and their scarcity, it seems quite bold to me to talk about "established orthographical systems". Given the early date of breaking according to the Standard Theory, an allographic rendering of /ɛa/ by <a> would have to go back to a 5th century scribal tradition, where <a> was used to mark unbroken /a/. Hogg tries to solve the discrepancy between the theory and the actual attestations by assuming two different dialects in the south, an a-dialect and an ea-dialect, and the difference between early WS and late WS is interpreted as a spatial shift, "[…] although the geography of the change remains obscure" (Hogg 2011, 81). Despite the lack of evidence for Hogg's theory, and ignoring the obvious chronological cline in the data, the transition from <al-> to <eal-> is not interpreted as a chronological change, but it is assumed that a fully-fledged ea-dialect already existed from the very first moment of attestations (compare section 1.6).
That there is a conceptual tension between fronting of *a before *r,*l,* x on the one hand, and breaking in the same contexts -which presupposes back consonants -on the other, is also acknowledged by Ringe & Taylor (2014, 185-186 phonological parallel in Old High German, where the contexts *xC, *rC and *lC block fronting through i-mutation. Ringe & Taylor (2014, 185-186), however, argue that fronting is plausible by the following incremental reasoning. At first, it is claimed "[…] that fronting occurred even before the velar fricative /h/ […]" invoking some of the keywitnesses, discussed in section 2.2.2. Then it is stated that " [i]f a velar fricative did not inhibit fronting at that time, it is unlikely that /l/ would have done so, even if it was verlarized" (Ringe & Taylor 2014, 185-186). So, the first, phonetically unlikely, development should convince us of the likelihood of the second one. Consequently, we end up with series of unlikely events, all hinging on a few key-witnesses, which are the topic of the section 2.2.2. Although I voice this criticism, I do not claim that a solution is easy. Frisian shows no fronting before *x or *lC, but before *rC the situation is much more complex. When *rC is followed by a PWGmc back vowel *u or *ō or preceded by *b,f,p it comes out as PFri *[a̝ ], a vowel that develops into /a/ or /ɛ/ in different modern dialects (Hoekstra & Tigchelaar 2014). In the sequence *warC, the Proto-Frisian vowel may have even been [ɑ], considering the back-vowel realizations in some dialects (North Frisian uurt, East Frisian woort < OFri warte < PGmc *wartōn-'wart'). In most of the remaining contexts, however, the vowel comes out as Old Frisian <e>, as in therm 'intestines', merch 'marrow'. The Épinal Glossary show mostly <ea> in the context *arC, which is not very helpful for the identification of the exact vowel quality before breaking. The Erfurt Glossary is much more diverse in this respect, but does not show a clear one-to-one-correspondence with the Frisian situation, despite the fact that the contexts of Labial+*arC or *warC are known to be associated with Old English <a>, in particular in Northumbrian (Hogg 2011, 89-90;Ringe & Taylor 2014, 181-182). On the lexical level, however, the correspondences often fail: OFri bern, *bars, warte, ErfGl barn, baers, uaertae/uearte/uertae. An important difference is the differentiation between PGmc *zC (always <ae>) and *rC in Anglian, while no such contrast is found in Frisian. Another dissimilarity between Frisian and Anglian seems to be that Frisian consistently has /a/ before *x, as in nacht 'night' and fax 'long hair', whereas Anglian shows rather n(a)echt-and fex 'kelp'. 22 Therefore, when considering the differences between Old English and Old Frisian, but also within Old English, one should take into consideration that /r/, /l/ and /x/ may have had different realizations: [R]  differ strongly within a single linguistic continuum, with consequences for the inclination of preceding vowels to be fronted or broken.

Absolute and relative dating of the monophthongization of PGmc *ai
A related issue is the relative dating of fronting and monophthongization of PGmc *ai. The standard approach is that PGmc *a developed into *ae, also as part of the diphthong *au > *aeu. This places the monophthongization of PGmc *ai in the time before the fronting, because otherwise **ǣ would be the expected product of that monophthongization. This is considered to separate Frisian and English from the very beginning of their developments (Campbell 1977, 52, §131-2). The runic evidence tells a different story. The form [sk]amella 'footstool' < Latin scamellum (Fallward), from the early 5th century, originating from the so-called Weser-Elbe triangle, one of the main regions of origin of the Anglo-Saxons, as well as the raïhan 'roe-deer' < PGmc *raixan-(Caistor-by-Norwich), from the middle of the 5th century, show that both fronting ( = /a/) and monophthongization did not take place before the middle of the 5th century (for a description of the inscriptions, see Looijenga 2003, 239-240, 284). The combined evidence of Anglo-Frisian early futhorc-inscriptions shows a relative order of the developments: ca. 450 -rounding > fronting > monophthongization of *ai -ca. 550 (Waxenberger 2013, 41) 23 The chronology in the Standard Theory is based on the assumption that vowels in diphthongs behave in the same way as their monophthongal equivalents. This is demonstrably not true in many contemporaneous languages, e.g. Dutch /a/ = [ɑ], /au/ = [aw]. Another example comes from Wallis Swiss German, where OHG o was retained, but the diphthong ou developed into öi (Wipf 1910, 32, 38). So, there is no need to postulate the dating of monophthongization of *ai before the fronting and this order is not supported by the runic evidence (Ringe & Taylor 2014, 170-171).

PGmc *ē 1
Breaking of non-WS *ē > *ēo presupposes the raising of PGmc ē 1 to *ē at a very early stage in the chronology (Hogg p. 61; see also Table 1). It is generally held that PGmc *ē 1 developed to ē in Mercian at a fairly early date, except for the position before nasals, where we find <a(a)> or <ou> in EpGl and ErfGl. In other contexts, ErfGl has, apart from three instances with <ei>, only <e>; this demonstrates that the raising of *ǣ was already completed by the time of the attestation of ErfGl. The development from ǣ > ē took place in Kentish in the 9th century (Hogg 1988, 194-198). 24 Table 6 shows the distribution of the spellings <ae> and <e(e)> for PGmc *ē 1 and PGmc *ai with i-mutation in the Épinal Glossary. The figures in the table illustrate that these two spellings are not applied at random, but show a clear correlation with the etymology of the lemmas.

Épinal Glossary
PGmc *ē 1 PGmc *ai > *ā + i-mutation <ae> 9 23 <e(e)> 27 1 Table 6: The spelling of *ē 1 not before nasals, and *ai + i-mutation with <ae> or <e(e)>; the contrast is highly significant in a Fisher's Exact Test: p < 0.0001 The EpGl contains 28% tokens with <ae> in words with PGmc *ē 1 . I would argue that 'scribal confusion' -as discussed in section 1.7 -is not an attractive interpretation for so many deviating forms, given the fact that ǣ < PGmc *ai with i-mutation is nearly consistently spelled <ae> (Table 6). Therefore, I hypothesize that the transition from pre-OE *ǣ > ē was in its final stage in the EpGl-variety of Old English and hence cannot have been prehistoric. In Frisian, the raising of PGmc ē 1 to /e:/ post-dates the settlement of the North Frisian Islands around 700, which is in line with the developments in ErfGl. The crucial point of this reasoning is that it renders the whole reconstruction of -neg-'near' < PGmc *nēhwa-(Franks Casket), with early raising of pre-Old English *ǣ 24 This relatively late date of raising in Kentish was already identified by DeCamp (1958, 235), who also points out the reflex of proponents of the Standard Theory (here Luick and Sievers-Brunner) to discard counterevidence by assuming dialect mixture in the manuscript. DeCamp (1958, 240) explicitly mentions the raising of ǣ > ē as a 7th century innovation in Old English, except for West Saxon. I do not, however, adhere to his identification of i-mutation as a continental feature, originating in southern Germany. < PGmc ē 1 and subsequent breaking and smoothing *nǣh-> *nēh-> *nēoh-> *nēh-, obsolete.
The evidence from the Anglo-Saxon runic inscriptions partly supports this chronological interpretation, partly complicates the matter. There are two Anglo-Saxon runic words that seem to contain a PGmc ē 1 : maegae 'relative (dat.sg.)' on the Undley Bracteate (East Anglia) from the late 5th century (Waxenberger 2013, 46) and -mer?? as part of the name sigimer-on the Ash Guilton (Kent) sword pommel from the late 6th century (Looijenga 2003, 276; see also Hines 2006, 196). The former can be derived from PGmc *mǣgai, the latter from *mǣr(i)jaz. The two inscriptions confirm a chronological order of the change, but the dating of the change from *ǣ > *ē to the late-6th century is earlier than concluded above. Note, however, that *mǣr(i)jaz has an i-mutation factor which may have caused an earlier raising than in other contexts. 25 EpGl contains both words with <e(e)>: meeg and mere 'famous' (adj.).

Inadequate 'key-witnesses' for early breaking
It is typical for the Standard Theory that, on the one hand, so many forms are ignored or reasoned away (such as the al-spellings in early WS), while, on the other hand, the whole theory hinges upon only 2 or 3 key-witnesses. The forms ēa 'river' and slēan 'to hit' are generally invoked as the key-witnesses for early fronting and breaking of PGmc *a (Hogg 2011, 78-79;Ringe & Taylor 2014, 177, 189-190). The forms PsGlA neolaecan 'to approach', neowest 'nearness' and Kentish nior, neor 'nearer' serve as evidence for the breaking of *ē < PGmc *ē 1 .
All these key examples consist of words with a vowel followed by PGmc *x. An analysis of such words (Ringe & Taylor 2014, 177-179, 214-215, 314-317), and in particular the evidence from the earliest sources (the Épinal and Erfurt Glossaries, the somewhat younger Corpus Glossary and the early Northumbrian texts) makes it clear that the /x/ was still present in many words in the early 8th century, such as thuachl 'washing (s.)' < PGmc *þwaxla-, aehrian 'ears of grain' < *axizor suehor < *swex(u)ra-'father-in-law' (WS þwēal, ēar, swēor). In words with preserved /x/, there are no diphthongs. The Standard Theory interprets this as a consequence of Anglian Smoothing, but if such a process existed at all, it was only gradually implemented during the 8th century, as was shown by Toon (1987, 283; compare section 2.2.3).
Let us now return to the mentioned key-witnesses for early fronting and breaking: ēa, slēan and nēo-. The earliest attested form for 'river' is ēo-(Ep/ErfGl). The form ēa is a later/WS development. The form eocan easily be derived from *axwō-> *aexwu > *aexu > *aeu > eo, with contraction and without breaking of any sort; so also Kortlandt (1999, 49). 26 In the same way, slēan (e.g. in the somewhat younger Vespasian Psalter (PsGlA)), can be derived from *slaxan-> *slaehan > *slaean > slēan. Moreover, the verb slēan shows extensive variation in the paradigm due to strong analogical reshufflings, with root vowels involving not only ēa, but also ā, ǣ, ē, and their i-mutated variants īe, ē, y (Campbell 1977, 316-317, §744). This strong variation does not make it a solid candidate to be a 'key-witness' in the chronology, because a clear distinction between regularly developed vowels and analogical vowels is a hazardous enterprise. Compare also Kortlandt (1999, 49) for various other potential scenarios and Hogg (2011, 96, 179) for his vision on the variant slān.
The forms that the Standard Theory claim as evidence for the breaking of pre-OE *ē were discussed in section 1.5: PsGlA neolaecan 'to approach', neowest 'nearness' and Kentish nior, neor 'nearer'. A prerequisite for breaking to ēo is the early raising to *ē. For the lack of evidence for such an early raising in Anglian, see section 2.2.1. Hogg (2011, 86, 87) includes the Kentish examples here, suggesting that also Kentish had early raising of PGmc *ē 1 , but on page 61 he only mentions Anglian dialects. Also Campbell (1977, 50, §128, 99, §235 (2)) seems to operate with a Kentish ēo from an early *ē. Hogg (1988, 194-196) himself showed on the basis of 9th century charters that the raising from ǣ > ē in Kentish is of much later date than breaking in the Standard Theory (also mentioned in Hogg 2011, 204). A 9th century Kentish ē could never be the input to breaking in the Standard Theory, ranked much earlier in the chronology. We also do not need any breaking to interpret the Kentish forms. The PWGmc comparative form *nāxwōza-developed to pre-OE *nǣxora-; ǣ was later raised in Kentish and the intervocalic *x was lost, which resulted in neor-. Exactly the same developments took place in Frisian, which shows further accent shift to niār 'nearer' (compare Kentish nior; positive OFri. nēi). Old Frisian never had the type of breaking that is assumed for Old English in the Standard Theory. This implies that the development of diphthongs before a lost pre-OE *x was a different process than the breaking of short vowels before -r/lC and that the former can hardly be used as key-evidence to determine the chronology of the latter.
The development of *xw in PGmc *nēx-westi 'nearness' was different from that of the comparative form neor, due to the morpheme boundary between *nēx and *westi-: *nēx-westi-> nēwest (earliest attestation <newest>, 745). The later form neowest may be the result of a later glide insertion between the e and the w: *[eo̯ w]; compare EpGl beouuas < PGmc *bewwu/-a 'barley'. One may also consider the scenario sketched in section 1.5 with vocalisation of [x] > [u̯ ] from a not attested *nēxwest.
To sum up: the attested sequences such as from suehoras (EpGl) to sueoras (ErfGl) do not need to be explained by breaking and smoothing before the time of the first attestations; eo in such words can simply be interpreted as the result of contraction after loss of intervocalic *x words such as eo and slean can be understood in the same way the Mercian and Kentish examples with *nēx-fail as evidence for early breaking because of the unproven early raising of PWGmc. *ā all the way to *ē -the subsequent assumption that forms such as neolaecan are the result of early breaking, followed by loss of *x before the application of Anglian Smoothing, is strongly contested by attestations such as genehlaecað with <h> but no broken vowel, attested side by side with geneolaeceð -the existence, on the one hand, of diphthongs in Old Frisian and Old English, as in niār 'nearer' (OE nēor) and swiāring 'son-in-law' (OE swēor 'father-in-law'), and the lack of diphthongs, on the other, in Old Frisian in words with -r/lC, e.g. bern 'child' (OE bearn), jerne 'willing' (OE geornlice 'diligently'), indicate that these were two separate processes, where evidence from the context before *x (contested here) is not automatically applicable to the context before -r/lC; additional evidence for an independent interpretation of these two phonological contexts also comes from Old Norse, with breaking before -r/lC (ON gjarna 'willing'), but without breaking before *x (naer 'nearer') (Dyvik 1978) -I therefore conclude that the handful of 'key-witnesses' for an early date of the breaking do not stand individual scrutiny. There are plausible, different interpretations for each of these instances, and some interpretations from the Standard Theory are not supported by the data, so they cannot be taken as a reason to ignore so many other attestations that do not comply with the predictions of the Standard Theory.

The lack of smoothing of PGmc *eu and *au and OE *ĕo and *ĕa
One of the most intriguing sound laws in the history of English phonology is Anglian smoothing (see the quote in section 1.1). Apart from the phonological (un)likeliness, it turns out that such an effect was far from regular in the language of the Mercian Glossaries. There is only one example of a word with PGmc *eu in a potential smoothing context in the glossaries. The word does not show 'smoothing': -fliogae 'fly' < PGmc *fleugō-. The <io> is the regular continuation of PWGmc. *iu < PGmc *eu with height harmony.
Additionally, 'smoothing' of short eo (sometimes the text shows ea) (< PGmc *e) is not regularly applied in the early glossaries either (Hogg 2011, 143;Toon 1987). At the same time, one finds many unbroken forms in the glossaries which should qualify for breaking, but not for smoothing: eleven instances with <e> or <i> in contexts where the Standard Theory predicts breaking. The Standard Theory correctly predicts the vowel in tokens with PGmc *e in breaking and smoothing contexts in only 65% (26/40), as shown in Table 8.   Toon (1987, 283) in his eminent article on language change in the Old English period considers only instances represented in the second data row of Table 5, and concludes from this that smoothing was being implemented at the time of the writing of the Épinal Glossary. He does not consider the instances represented in the first data row in the table, which shows that the whole approach of 'breaking and conditional smoothing' provides in fact a poor explanation of the distribution of <e, i> vs. <eo, ea> in the Glossary.
Only for PGmc *au is there convincing evidence of a trend towards monophthongization before velar consonants during the 8th and 9th centuries, as can be seen in Figure 1. Taking all these facts into account, I conclude that pre-attestation 'Anglian smoothing' of broken vowels as it is operationalized in the Standard Theory did not exist and therefore cannot account for the unbroken, assumingly 'smoothed' forms in the early (Anglian) texts. In my view, the sound law is a reconstruction artefact.

i-mutation of diphthongs (Hogg § 5.82)
The i-mutation of non-WS *ea and *ēa is supposedly e/ē. Contrary to the WS development to <ie>, this non-WS development can hardly be called i-mutation, but rather a form of monophthongization. This is not the expected impact of i-mutation factors, for which I can quote Hogg (2011, 17) himself: "[…] it is improbable that i-umlaut had a general monophthongization effect." A more logical effect is a direct raising of *ae > e. Compare the following options in the Standard Theory and my alternative Anglo-Frisian hypothesis: ST: *a > *ae (fronting) > *ea (breaking) > *e ('i-mutation') AF: *a > *ae (fronting before palatal C) > *e (i-mutation) My alternative interpretation with breaking postdating i-mutation also accounts for the form feormat in EpGl (ErfGl <caeormad> = *faeormađ) < PGmc *farmjan-'to supply' (note the class transition from Weak I > Weak II).

Conclusion regarding the Standard Theory
In the previous sections of this article, I have discussed various flaws and contradictions of the Standard Theory. I believe that many of them pose serious challenges to the theory, and the accumulation of inconsistencies is, in my view, fatal. The criterion for falsification is the following: the theory claims that the mentioned developments are 'prehistoric' and took place before the attestation of the Mercian Glossaries. The fact that later stages of the language (e.g. 10th century WS, early Middle English, etc) positively attest to developments such as the monophthongization of (i-mutated) diphthongs, more extensive breaking phenomena or otherwise the raising, lowering or rounding of vowels does not contribute to the credibility of the Standard Theory, describing the phonological history of English between the landnam and the first attested texts in the early 8th century. To sum up: the axioma of general fronting of PGmc *a, both as a single vowel and as part of the diphthongs *ai and *au, is unproven and, given the raïhaninscription and the absence of fronting before *w, in fact, unlikely; but see Ringe & Taylor (2014, 175) for contemporaneous parallels for the development au > aeu the abundant number of intermediate steps with alternating frontedretracted or broken -smoothed forms, involving a high number of repetitive stages is methodologically unlikely; these reconstructed stages seem too many to reasonably fit in the timeframe between ca. 450 and 700 there is no positive evidence for the assumed intermediate steps from runic inscriptions there is a lack of attested dialects with the intermediate stage *ae before breaking the key-witnesses for early fronting and breaking (ēa, slēan, nēowest) can just as well and even more easily be explained in a different way; counterevidence from early 7th century attestations without breaking (e.g., in early attested names) and the 8th century form neweste is ignored these key-witnesses all concern the context before *x, which is not necessarily conclusive for the context before -r/lC the assumed early raising of PGmc *ē 1 in Anglian is unproven and ignores counterevidence from the early glossaries; without early raising, there can be no early breaking of *ē -'Anglian smoothing' is -if at all -only partially implemented; I rather consider it to be a reconstruction artefact to account for the lack of breaking in many non-WS forms -an attempt that, given the evidence from PGmc *a, *e and *eu, did not turn out to be successful for the interpretation of the data in the Glossaries the suggested pattern of monophthongization through i-mutation in non-WS dialects is typologically uncommon and the reconstructions are fraught with unattested intermediate steps

Additional points of criticism from earlier sections include:
the mismatch between the number of reconstructed sound changes for WS and Anglian, and their actual dates of attestation the high number of 'repair' sound laws for exactly the language of the oldest sources -the lack of parallelism with the chronologies for Old Frisian and Old Saxon, which -especially the former -start from the same common position of Proto-West Germanic and very often show the same or very similar outcomes as Old English the rejection of many instances of counterexamples from the attested data, which are ignored by invoking scribal confusion, 'Mercian influence ', etc All these points of criticism bring me to conclude that the Standard Theory is untenable and needs thorough revision. This revision has to come from a close scrutiny of the earliest attestations of Old English, the short Northumbrian fragments and the glossaries known as Épinal and Erfurt Glossaries (Versloot, to appear). Although not imperative, linguistic similarities between Old English and Old Frisian indicate -as a starting hypothesis for the interpretation of the Early Old English evidence -that the chronologies of Old English and Proto-Frisian may have been largely parallel or at least parallel to a larger extent than has been hitherto assumed.

Comments invited
PiHPh relies on post-publication review of the papers that it publishes. If you have any comments on this piece, please add them to its comments site. You are encouraged to consult this site after reading the paper, as there may be comments from other readers there, and replies from the author. This paper's site is here: https://10.2218/pihph.6.2021.6690

Acknowledgements
The content of this paper was first presented at the ICEHL2018 conference (at the University of Edinburgh). The text was critically read and commented on by the PiHPh-editor, by Stephen Laker and by two anonymous reviewers for English Language and Linguistics, who recommended to publish this part independently. I would like to thank all these people for their remarks and also Patrick Honeybone for his additional editorial work.

Logistic Regression Analysis
Definition of the independent variables: All four variables are binary and the value '1' refers to the feature that is expected to favour application of breaking.