Old English phonology

[19] The fricative /x/ (spelled ⟨h⟩) came to be lost when single between voiced sounds; since only long /xx/ remained in this position (in words such as hlihhan, hweohhol), its length was no longer contrastive.

Proto-Germanic *z (which existed only as the Verner's Law counterpart of *s) regularly developed to Old English /r/ (a sound change called rhotacism).

Such spellings occur regularly in Late West Saxon and Kentish texts from around 900 onwards, suggesting both sounds had come to be pronounced [x] in this position (compare the devoicing of final /f/).

[54] However, because palatal ċ and velar c alliterate in English poetry up through at least the late tenth century, Minkova 2014 assumes that they were still allophones of a single phoneme before 1000.

[14] Likewise, word-initial palatal ġ and velar g alliterate with each other in early Old English verse (before the latter changed to [ɡ], circa 950 AD[38]), which Minkova 2014 interprets as evidence that [j] and [ɣ] constituted allophones at this point in time,[55] despite the existence of /j/ from Proto-Germanic.

Velar [k ɣ sk] can be found before unstressed back vowels in words such as dīcas, plegode, æscas,[81] whereas palatal [tʃ j ʃ] can be found before unstressed back vowels in words that originally contained an etymological *j or *i after the consonant, such as sēċan, wierġan, wȳsċan from Proto-Germanic *sōkijaną, *wargijaną, *wunskijaną.

[82] Velar [k ɣ sk] can be found before an unstressed front vowel in class II weak verbs with an infinitive ending in -ian; e.g. wacian, dagian, āscian.

[83] The front vowel /i/ is here derived from umlaut, unrounding, shortening and raising of original -ō-: e.g. Proto-West-Germanic *makōn 'to make' was replaced with *makōjan, which is hypothesized to have developed through *makœ̅jan and *makejan to Old English maci(ġ)an.

In circumstances where the palatal affricates [tʃ] and [dʒ] came to be followed by another consonant due to syncope of an intervening vowel, they were eventually replaced with the corresponding velar plosives, [k] and [ɡ] respectively.

The spellings ⟨hw⟩, ⟨hl⟩, ⟨hn⟩, ⟨hr⟩ probably represented two-phoneme clusters, /xw, xl, xn, xr/, where /x/ was pronounced [h] (its usual allophone in syllable-initial position).

[93] There is an alternative hypothesis that holds that (at least in later periods) ⟨h⟩ in these sequences was not pronounced as an independent consonant sound, but was only a diacritic marking the voicelessness of the following sonorant.

[95] The merger of /xr/ and /r/ seems to have been completed earliest, by the middle of the eleventh century, based on frequent interchange of the spellings ⟨hr⟩ and ⟨r⟩ in glosses from that time period.

[97][98] The digraphs ⟨rh⟩, ⟨nh⟩ and ⟨lh⟩ are attested to some extent in Middle English texts; e.g. the 12th-century Ormulum contains ⟨rhof⟩ but also includes forms spelled with simple ⟨r⟩.

[94] A merge of the cluster /xw/ with /w/ is also attested in some historical and many current varieties of English, but has still not been completed, as some present-day speakers distinguish the former as [ʍ].

That is suggested by the vowel shifts of breaking and retraction before /l r/, which could be cases of assimilation to a following velar consonant: Based on phonotactic constraints on initial clusters, Fisiak 1967 proposed interpreting ⟨wr⟩ and ⟨wl⟩ as digraphs representing the velarized sounds in prevocalic position, in which case the distinction would be phonemic, as exhibited by minimal pairs such as wrīdan [ˈrˠiːdɑn] "to grow" vs. rīdan [ˈriːdɑn] "to ride" or wlītan [ˈɫiːtɑn] "to look" vs. lītan [ˈliːtɑn] "to bend".

[101] However, this hypothesis is inconsistent with orthoepic and orthographic evidence from the Early Modern English era,[102] as well as borrowings into and from Welsh, which has [wl] and [wr] as genuine initial clusters.

Furthermore, in Old English poetry, ⟨wr⟩ and ⟨wl⟩ can alliterate with each other as well as with ⟨w⟩ followed by a vowel, as in "Wēn' ic þæt gē for wlenco, nalles for wræcsīðum" (Beowulf 338).

[105][106] The long and short versions of each vowel were probably pronounced with the same quality, although some reconstructions assume accompanying qualitative distinctions.

[125] As with monophthongs, their length was not systematically marked in Old English manuscripts, but is inferred from other evidence, such as a word's etymological origins or the pronunciation of its descendants.

However, in response to these proposals, further arguments have been made in support of the proposition that short digraphs did in fact represent phonetic diphthongs.

[62] Ringe & Taylor 2014 assume that by the 9th century, the second component of ea had become lowered and unrounded (aside from in the minority of regions where the alternative spelling ⟨eo⟩ was used for this diphthong).

[149][150] However, Ringe & Taylor 2014 do not find height harmony convincing as a general rule, arguing that the later development of ie īe points instead to the value [iə̯ iːə̯].

[151] Hogg 2011 considers the lowering of the second element of diphthongs to be related to the development of unstressed vowel qualities; while acknowledging that the height of the first element affected the outcome of the second, Hogg rejects height harmony as an overarching principle, and supposes that io came to be pronounced [io] in Old English, with [iu] only being its early or archaic value.

However, rather than indicating the development of a diphthong, these spellings might have just been a convention for marking palatal consonants before a back vowel,[155] since the modern English descendants of such words do not display the typical evolution of the diphthong ⟨eo⟩ to a front vowel: Peter Schrijver has theorized that Old English breaking developed from language contact with Celtic languages.

[157] In dialects other than West Saxon, i-mutation instead turned ea, ēa into e, ē and left io, īo unchanged.

The cluster /sk/ could be found word-medially before a back vowel, e.g. in the words þerscan and discas, although the lack of palatalization in such forms might imply that the /s/ was shared between the first and second syllable.

This assumption is false: there are languages where a syllable can end in an obstruent followed by a resonant, as demonstrated by modern Icelandic, where vatn, býsn, segl, gísl are all monosyllables.

[197] There is evidence that this type of coda cluster eventually became disallowed in Old English, because many such words show a spelling with a vowel letter inserted before the consonant, such as hleahtor.

Some of these were inherited from Old English (drink and drench, day and dawn), and others have an unpalatalized form loaned from Old Norse (skirt and shirt).

Mercian was spoken in the middle part of England and was separated from the southern dialects by the Thames and from Northumbrian by the River Humber.