Sunday, 13 August 2017

Pagan Anglo-Saxons, Phonology, and Christian lexis 3

In this series of posts I will put forward the following set of ideas. They are not entirely new or original, but I think that there is a strong philological case to support them.

In southern Britain in the fifth century there was a significant Latin-speaking Christian British population. This is evidenced by pre-550 Latin loanwords in Old English and in particular by Christian terminology. The loanwords are most likely to be due to this population adopting Old English as their first language. This may be why there is also an important group of religious terms coined in Old English rather than borrowed from Latin. And it also explains why there are virtually no British loanwords in Old English.

III. Christian lexis in the light of phonology.

In my first post, I suggested that eight Christian vocabulary items must have been acquired in Old English between the settlement period and the Augustinian mission, and neither before the settlement not after the mission. In my next post, I set out the phonological background against which the acquisition of loanwords from Latin took place.

I will now consider the words in more detail.

deofol ‘devil’
engel ‘angel’
abbod ‘abbot’
biscop ‘bishop’
preost ‘priest’
munuc ‘monk’
mynster ‘monastery’
cirice ‘church’

deofol and engel

First, some further points against their being part of the Anglo-Saxon settlers’ vocabulary.

Campbell’s theory (Grammar §492) that these words entered Germanic from Greek seems to me unnecessary and unlikely. It has to assume the words entering Germanic in the east via Gothic and being transmitted across Europe.

The form of the Gothic equivalent of ‘devil’, is in any case against the possibility that it was transmitted from Gothic into West Germanic. But the form of the word in the West Germanic languages is such that each constituent language could have independently borrowed it from the late Latin reflex of Latin diabolus, or one could have borrowed it and transmitted it to the others. The Old High German form tiufal shows borrowing earlier than the High German consonant shift (d > t) but not earlier than  the period when Latin intervocalic /v/ began to be borrowed as <f> rather than <b>, which makes it an Old High German borrowing (contrast sauber (Latin sobrius), an earlier borrowing with /b/).

Could deofol have been borrowed after the conversion in 597? Phonologically, this seems too late for the development of Latin /ia/ to OE io or eo, especially when we consider that many scholars have assigned this development to a Primitive Germanic borrowing of /ia/ as /iu/. The word diaconus, which is patterned very similarly to diabolus, and arguably must have been an early borrowing after the introduction of church services to Anglo-Saxon life, takes the OE form diacon (the Middle English and modern forms seem to imply that there may have also been an OE variant *dēacon with the /eːa/ diphthong). Had diabolus been borrowed around 600, arguably it would have taken the form diafol.

engel is even less likely a term for pagans to need. It seems pointless to posit ‘Gmc. *angil-’ as Campbell does, or even a West Germanic ancestor. But importantly what we see in this word is:

  • the application of the West Germanic/primitive Old English rule C5 above, by which late Latin */angelu-/ became */angila-/, with /i/ from /e/ in the post-tonic syllable;
  • and as a result of this the operation of A5 (i-umlaut), by which */angila-/ became */engil/.

This places the borrowing somewhere between 450 and 550.


OE abbod shows the operation of the following change:

  • B6: the late Latin development of intervocalic /t/ to /d/ (around the end of the fourth century)

It also shows that the following changes had not yet happened at the time of borrowing:

  • B10: Gallo-Romance fronting and raising of /aː/ to /æː/, which would have given OE e (early sixth century)
  • B11: the further change of intervocalic /d/ to /ð/ (6th–7th century)
  • B12: the simplification of late Latin long consonants (dated 6th–7th century)

It also indirectly shows that the following change had occurred before the time of borrowing:

  • A1: Common West Germanic /aː/ > /æː/ and /ai/ > /aː/, because the second syllable shows /o/ from an earlier /a:/ in a lower-stressed syllable, exactly as in native Old English earfoð (from */arvaiθ/.

This makes it likely to have been a late fifth or early sixth century borrowing.


The form of the word itself pinpoints fairly accurately the date at which it must have been adopted from late Latin into a Germanic language, whichever one that was.

It has to have been borrowed after:
  • B.7 Late Latin medial /p/ > /b/ (around the end of the fourth century) */episkopu/ > */ebeskobu/
and before
  • B. 11 (when the two resulting instances of /b/ would have become /v/, i.e. */ebeskobu/ > */eveskovu/ (sixth to seventh centuries)

It also shows:
  • C.2 substitution of /p/ for the second /b/
  • apocope of the initial e-, which must have occurred after change B.7.
  • A.3 palatalization of -(s)c-

In which language did this apocope happen? By rule C.1, one would expect that if the word still had the form /ebiskobu/ in late Latin, it would have been adopted into Old English with the stress shifted to the initial /e-/ **/ˈepiskop/  (with the first /b/ replaced by /p/ according to C.2, like the second). That a borrowed form so stressed is possible is shown by Welsh esgob, Old Irish epscop. There is not much evidence of avoidance of stress on an initial vowel by apocope in loanwords in Old English.

On the other hand, such apocope was possible in Romance, as is shown by Italian vescovo, although in this case the apocope comes at a later stage in the word’s development (*ebescobo > *evescovo > vescovo). On balance the likely option is that initial e- disappeared in late Latin soon after the intervocalic voicing of the –p-, most likely (as specialists have previously suggested) by metanalysis of *ille episcopus, or rather *ele ebescobo, to *ele bescobo.

However, this change did not happen in Gallo-Romance, the form of later Latin most likely to have influenced surrounding Germanic dialects. Here *ebescobo continued on its way, unmetanalysed, to *evescovo (eventually becoming évêque). Therefore the word was not borrowed from a known Gallo-Romance form, but from an unattested form in some unidentified dialect of late Latin. Moreover, since the Old English word has /i/ in the first syllable, this form must have avoided the late Latin sound change of stressed short /i/ to /e/, which was carried through in most Romance languages; hence the unidentified dialect showed a second peculiarity:

Absence of lowering of short /i/ to /e/ in a closed syllable

There is an alternative theory that the word was borrowed into Old English from Augustine’s Frankish interpreters. It might be imagined that the interpreters were heard to refer to Augustine himself by this word and hence it was naturally be taken into OE in place of the Latin equivalent. This would mean that the late Latin form of episcopus had already entered Frankish as *biskop. This raises the questions:

(i) what evidence is there that Frankish had the form *biskop in 597?
(ii) given that Frankish was spoken on Gallo-Romance soil, why would the Frankish form not reflect the Gallo-Romance vocalization, *ebescobo, with two instances of /e/ in two initial syllable(s)? And
(iii) why did the Old English word not have /sk/ as the medial consonant group?

The third point is particularly important. It is an unexpected feature of the word biscop that the sequence –sc- contains palatalized c (so that sc becomes English sh) rather than /k/. The palatalization of c after s occurred in initial position both before back vowels and (as expected) before front ones, but medially it was unusual in a back environment (though unpalatalized –sc- often became –cs-), e.g. ascian~ascode (giving ask), tusc (giving tusk), etc.

It is generally thought that palatalization of (s)c before back vowels was later than that of (s)c before front vowels; the latter must have occurred at the same time as palatalization before front vowels of /k/ (not after /s/), /ȝ/, and /g/.  However, it was still early enough to cause change A.4, the palatal diphthongization of following vowels, and that change is usually believed to have occurred before change A.5, i-umlaut. If this reasoning is correct, then biscop is a borrowing that predates the mid sixth century.

Campbell suggests that ‘bishop’ owes its medial /∫/ ‘to a strong medium stress, so that the initial of the second syllable developed as if before a stressed syllable’ (§441, footnote 3). This is tantamount to saying that biscop was treated phonologically as a compound. Perhaps it actually was understood as a compound, seen through the prism of folk etymology! To an Old English speaker ignorant of Latin and Romance the word could have appeared to be a compound of bi- (compare bismer, in which the original long vowel of bi- was certainly shortened) and scop: this would not make great sense, but scop does at least denote a learned, socially prominent, person, if a secular one.


This word also is explained as having been introduced to Old English by the Augustinian Frankish interpreters. The phonological transmission theory seems to be that

  1. OE preost reflects Frankish *priast
  2. Frankish *priast is hypothesized on the basis of OHG priast(ar)
  3. OHG priast(ar) is a variant of OHG pre:st(ar)
  4. OHG pre:st(ar) was derived from late Latin or early Romance *preːster(um) (supposedly attested by Old French prestre)
  5. late Latin *preːster(um) was derived from earlier *prevester
  6. *prevester is from Latin presbiter by metathesis

There are numerous weak links in this proposed chain of transmission. The strongest idea is no. 6, the metathesis of (a form resembling) presbiter to (another resembling) *prebester. Such a change is vital to explain the shape of the Old English word, and that it is a possible late Latin or early Romance change is confirmed by e.g. Latin vespertilio > *vipistrello > Italian pipistrello, in which the s has similarly migrated forward by the length of a syllable.

Step 1 assumes a fifth century Frankish form for which there is (I believe) no direct evidence. It also assumes that the Frankish etymon lacked the ending -er, like Old High German indeed, but unlike the forms in most West Germanic languages including Dutch (priester) which is the nearest thing to a modern form of Frankish. It also hypothesizes that a very small number of Frankish words (all of which happened to be concerned with ecclesiastical matters) were borrowed into Old English at the time of the Augustinian mission, despite the fact that no other Frankish loanwords are known.

Steps 2–3 rely on Old High German forms that contain unshifted initial /p/ and must therefore be later than the High German consonant shift, since the shift of /p/ to /pf/ is usually dated sixth to seventh century. These forms therefore cannot be taken as evidence for Frankish usage in an earler period. Moreover the assumption that sixth century Old English would have adopted /ia/ as /eo/ is at odds with the evidence of the word diacon (referred to above), which is likely to have been borrowed at the time of the Conversion.

Step 4 relies on a transmission (Gallo-)Romance *preːster, from *prevester, from *prebester, with lenition of medial /b/ followed by its complete loss. This is an impossibility at this date, since intervocalic /v/ by lenition was strongly maintained in Gallo-Romance and Old French except before /o/ and /u/ and after tonic /u/. But in any case Old French prestre is usually explained as from *presb(e)ter with syncope of the medial syllable and loss of /b/ as the midmost of three consonants.

The vocalization of OE preost is so anomalous that what is really required to account for it is a linguistic milieu which characteristically shows developments of late Latin that deviate from those attested by later recorded Romance, as was the case with ‘bishop’. (It is supportive in this regard that Old Irish cruimther must go back to *premiter, also a deviant form of the etymon presbyter.)

  1. Ignoring the -er ending for the moment, the primitive Old English form */preust/ requires a late Latin etymon */preust-/
  2. A late Latin form */preust-/ could have arisen from */prebust-/ by the early (third century) passage of intervocalic /b/ to /w/ followed by disappearance before a back vowel (B.2 above), seen in Old French taon, deu < late Latin *tabone, *debutu.
  3. A Latin form */prebust-er/ could have arisen from */presbuter/ by metathesis, as is already conceded for the alternative proposed form prevester.
  4. A Latin */presbuter/ could have existed as an alternative pronunciation of presbyter with /u/ representing Greek /ü/ rather than /i/. That Greek upsilon sometimes came into Latin (Gallo-Romance) as /u/ is shown in several words such as French bourse < Greek bursa and boîte ultimately < Greek puxis.
  5. At what point the -er ending was discarded is uncertain, but it was perhaps another aberrant characteristic of the late Latin dialect which is here hypothesized. It was also discarded in the anomalous word for ‘poppy’, OE popæg, etc., < Latin papaver.

This proposed etymology is admittedly based on thin evidence, and there are few intrinsic pointers to the date of the loan. However, if we discount the Frankish theory but still assume the word’s adoption into Old English at around 600 we have to explain why its Old English form isn’t a direct reflex of what would have been the then current late Latin or early (Gallo-)Romance form of the word, something like *presbeter. The etymology proposed here implies an earlier, though not a continental, borrowing.


The etymon is a late Latin */kirika/ (from Greek kyriaka). It seems to have been established that this term, standing in for ecclesia, was in current use in the north-western Empire for a limited period, round about the fourth century.

The Old English word cirice shows

  • A.3, palatalization of both the velar consonants

The common variant circe also shows:

  • A.7: loss of the unstressed medial vowel after a short syllable (like the inherited Germanic fyrsta, winstre, heolstor, seolfor (Campbell §389))
  • a range of root vowel variation between <i>, <ie>, <y>, and <e>

Old English glosses for ecclesia (beside cirican) include ciercan, cyrcean, cercean, and its Middle English forms show regional variation between <i> ~ <u> ~ <e>, which resemble variations in the reflex of Old English wyrcan, i.e. the kind that more usually occurred in words containing umlauted diphthongs or umlauted back vowels. This range of forms suggests that the root vowel underwent relatively early rounding to <y> and the subsequent southeastern change of <y> to <e>. So it is clearly a well-assimilated word.

This is the third word whose introduction into Old English has been attributed to the Frankish interpreters who accompanied Augustine. Again, there are serious problems with this idea:

(i) It is difficult to see why these interpreters would transmit to the English a Germanic word for ‘church’ which, by then, was used neither in church Latin nor in Gallo-Romance? Especially given that they would have been inculcating other ecclesiastical terms from Latin at the same time.
(ii) Why do we not find in OE any other Frankish-mediated Latin loanwords (apart from the proposed biscop and preost) or, significantly, any native Frankish words transmitted by these Frankish interpreters? If there were borrowings from Frankish at all, there is no reason why they should have been confined to Latin loanwords; the borrowers can’t be expected to have been able to distinguish one from the other.
(iii) Moreover it is not clear whether the existence of a Frankish *kirika, borrowed from late Latin *cirica at a period earlier than 597, is supported by evidence.

But in any case, the appearance of palatal consonants clearly points to adoption before the i-umlaut date of circa 550. Palatalization of velars before front vowels had stopped by the time the new front vowels, generated by i-umlaut, had become established.

If the reasoning is correct, it points to the persistence, in the relevant late Latin dialect, of a term that died out in other forms of continental Romance; another instance of idiosyncracy in this dialect. (The presence of the same loanword in continental Germanic dialects would be explained as independent borrowing.)


The late Latin etymon may either have been the regular monachus or an attested form monichus.

The late Latin etymon would have had the spoken form *monago or *monigo by 400, showing

  • Change B8, intervocalic /k/ > /g/ in the fourth century

Old English munuc shows:

  • C2 raising of /o/ > /u/ caused by both the nasal and the high front vowel following
  • C3 substitution of /k/ for /g/
  • suffix substitution of native –uc for late Latin -ig- (or -ag-): if the etymon was *monigo, this substitution of the suffix preceded the period of change C5, i-umlaut (an etymon *monago would not necessitate this dating)

Additionally, munuc has a feminine derivative mynecen (from primiitive Old English */munikiːn/). This shows

  • A3, palatalization of c
  • A5, i-umlaut

All the other Old English words with feminine –en which can show umlaut do so, and all look like old formations; mynecen is the only one with a Latin-derived first element.

This word is necessarily a pre-550 borrowing.


The late Latin etymon is *monisterium, attested from the fourth century (which is when monasticism first appeared in the Latin-speaking world).

The Old English word shows:

  • C2 raising of /o/ > /u/ caused by both the nasal and the high front vowel following
  • A5 i-umlaut

Additionally it shows some irregularity in the transmission, in that the fourth syllable of the late Latin etymon has left no trace (contrast the Latin suffix –arius, which, passing through West Germanic as approximately */aria-/, retains its second syllable in its Old English reflex –ere).

There are, however, other Latin loanwords in OE in which a trailing /i/ or jod which would have been expected to leave a trace either as final –e or as the i-umlaut of the root vowel has not done so, e.g. dinor, solor, ostre, orc, scrin; and also the place names Eotol, Reculf. This is not to say that the explanation is necessarily the same in all cases.

This has to be a pre-550 borrowing.

No comments:

Post a Comment