2.1 Old English frequencies as predictors of lexical survival

In my recent study of the diffusion of Latinate lexis in Old English (Timofeeva Reference Timofeeva2017), I argue that high-frequency loanwords are more likely to survive into the Middle English period. This applies, for example to god-spell with over 900 attestations in the DOE, but not to cȳþere ‘witness; martyr’ that occurs only c.100 times in the same corpus.Footnote ² Although the limitations and under-representativeness of the extant Old English texts with their focus on religious matters and narratives have to be borne in mind (Timofeeva Reference Timofeeva2017, Reference Timofeeva2018), the number of occurrences in Old English is typically a good predictor of the survival chances of individual lexemes into the later periods. For the present study, to verify this point further, I solicited the assistance of the DOE team to generate a list of religious words that are attested more than 500 times in the DOEC.Footnote ³ As the lemmatisation of the corpus is still an ongoing process, the list that I obtained contains headwords only from A to H (table 1). Nevertheless, generalisations about the Old English lexicon are possible even at this incomplete stage. I reproduce the list here, in descending order by number of occurrences. Each headword is provided with a definition in modern English and the date of last attestation (if obsolete) according to the OED and MED (if the datings differ).

Table 1. Old English headwords with 500+ occurrences (based on DOE, A to H)

Even though table 1 gives only a crude picture of high-frequency religious lexemes, with many of them being polysemous and the frequencies of the individual senses subsumed under general counts, it demonstrates unequivocally several important facts about the surviving record of Old English:

• • religious lexis constitutes a large portion of high-frequency content words (for comparison, only function words hē, hēo, hit, pron. (taken together, 200,000 occ.), and, conj. (172,000), and bēon, v. (100,000) have higher counts than god, n. (33,000); only cweþan, v. (17,500) and habban, v. (12,700) have frequencies higher than 10,000; between 10,000 and 5,000 we find dæg, n. (9,100), ān, num./pron. (9,000), for, prep./conj. (9,000), dōn, v. (8,900), cuman, v. (8,600), cyning, n. (8,000, with many of those probably being used in the religious sense too), be, prep./conj./adv. (7,600), æt, prep./adv. (6,000), būtan, adv./prep./conj. (5,600); between 5,000 and 1,000 we find hēr, adv. (4,100), gān, v. (3,700), hand, n. (3,700), hātan, v. (3,400), gear, n. (3,250), eald, adj. (3,000), faran, v. (2,600), gōd, adj. (2,500), hūs, n. (2,000), bōc, n. (1,750), drincan, v. (1,200), etan, v. (1,200), here, n. (1,150), full, adj. (1,100));

• • the high counts of religious terms reflect the thematic imbalance of the Old English corpus;

• • they have a high survival rate (all 52 lexemes are still attested in Middle English; 36 lexemes (or 69 per cent) survive into Present-day English);

• • all native-based lexemes (42 out of 52, or 81 per cent) represent either semantic loans, with an extension from secular to religious or from pagan to Christian domain, or loan translations.

From these we can conclude that high frequencies in OE and survival rates in ME correlate and, further, speculate that high frequencies in OE would suggest diachronic, diatopic, diastratal and diaphasic diffusion of these general religious terms already in OE, although the unavailability of statistics per individual sense makes it difficult to trace the exact trajectories of diffusion. Some of these trajectories will be addressed below, as we continue from generalisations to concrete examples.

While it may seem natural and consistent with the observed tendencies that such terms as god and hālig should survive, it is far less straightforward that such terms as dryhten and hǣlend should die out. As the analysis of the first five high-frequency terms on the list will demonstrate, the stories of individual lexemes are more nuanced than the general counts would suggest, whether they belong to the survivor or the obsolescent category. It is both the former and, especially, the latter categories that merit closer attention.

With 33,000 occurrences in the DOEC (or c.10,880 per million words), god continues as the dominant religious term into the ME period. French-derived deus, deu and de are attested only as parts of swear words and interjections: parde, depardeus; deu merci, deuleset (Käsmann Reference Käsmann1961: 41).

Although dryhten is the second most frequent religious word in OE (c.15,500 occurrences; c.5,110 per million words), its use drops already in early ME, but it only becomes obsolete towards 1500 (MED, s.v.). Apparently dominant in OE, dryhten encounters a powerful competitor in hlāford (c.1,400 occ.; c.462 per million words). In spite of its higher frequencies, the former is restricted to either poetry or, especially in late OE, to religious contexts, while the latter can refer to both ‘feudal lord’ and ‘the Lord’ without any genre or register restrictions. Käsmann relates the eventual demise of dryhten with its lack of semantic connection to the secular feudal world and its hierarchy (Reference Käsmann1961: 42). In LAEME the comparative distribution of the two lexemes becomes markedly different: lord scores 1,795 (c.2,767 per million words), while dryhten only 531 (c.818 per million words); moreover, sir(e) enters this semantic domain as early as c.1200 (Ancrene Riwle) and features 109 times in the corpus (of these only 23 are used as titles). Käsmann observes that the expansion of lord at the expense of dryhten is paralleled in the history of German: herro ousts truhtin as both ‘feudal lord’ and ‘the Lord’ (Reference Käsmann1961: 42–3). Although LAEME Maps function does not allow us to distinguish between the distributions of the sense ‘the Lord’ as opposed to ‘feudal lord’,Footnote ⁴ mapping the three lexemes seems nevertheless meaningful, especially considering that religious bias is present in LAEME Footnote ⁵ too (figure 1).

Figure 1. Geographical distribution of lord, dryhten and sir in LAEME

We can see that in many parts of England dryhten can still be used alongside lord and even alongside lord and sir. It serves very rarely as the only option, however, and alternation between dryhten and sir seems to be impossible. The innovation of sir is more pronounced in the West Midlands and the North, although the denser coverage for the West Midlands is a notorious problem in Middle English dialectology (Laing Reference Laing, Taavitsainen, Nevalainen, Pahta and Rissanen2000; Laing & Lass Reference Laing and Lass2007a: §1.3; Gardner Reference Gardner2014: 42–3; and discussion below).

Käsmann's observation that dryhten becomes obsolete in secular senses in all OE texts except poetry deserves closer consideration (Reference Käsmann1961: 42; the DOE s.v. dryhten points to the same conclusion). This would suggest that dryhten was a marked poetic, typically non-secular, term as early as late OE, which greatly reduced its survival chances into later periods. The German cognate truhtin seems to tell the same story. Perhaps it was precisely for their elevated connotations that these, probably archaic, terms were chosen or even reanimated in both languages. Diachronically, however, they were unable to resist the pressure of their unmarked synonyms.

The term that features third on the list, hālig, with 8,400 occurrences in OE (c.2,770 per million words), remains frequent in early ME: it occurs 1,476 times in LAEME (c.2,275 per million words) and is clearly dominant as an adjective (cf. only 3 attestations of the lexel saint as an adjective). hālig is never used with names of saints, however, this function being restricted to various reflexes of the Latin sanctus – sanct, sant, san, seint (tagged as a noun-title ‘n-t’ in LAEME, including abbreviations). The lexel saint is attested 1,164 times as a noun-title but only 22 times as a noun in its own right (cf. Käsmann Reference Käsmann1961: 62–4), in which function the reflex of the OE noun hālga (c.950 occ. in DOEC; c.313 per million words) is still more frequent – 88 attestations (c.136 per million words) in LAEME – becoming obsolete only after 1500 (MED, s.v. halwe). The functional distribution is thus apparent: hālig is used only as an adjective, hālga only as a noun, and the reflexes of sanctus are strongly preferred as titles.

Two nomina sacra come next on the DOEC list: crist (c.5,500 occ.; c.1,813 per million words) and hǣlend ‘Saviour, Jesus’ (c.5,000 occ.; c.1,648 per million words). crist is used 1,387 times in LAEME (c.2,137 per million words), typically in <crist> spellings, <ch> surfacing only in abbreviations. The history of the loan translation is much more complex, however.

Ælfric, who relied on the established vernacular tradition, grounded in the Gospels (Matthew 1.21), Jerome's Liber Interpretationis Hebraicorum Nominum and Isidore's Etymologies (vii.2), explained the meaning of the name Jesus as follows:

1. (1) Iesus is ebreisc nama. þæt is on leden Saluator. and on englisc Hælend. (ÆCHom II, 12.2 (122.420))

2. ‘Iesus is a Hebrew name, that is in Latin Saluator and in English Hælend’

Accordingly, the collocation Hælend Crist for ‘Jesus Christ’ is extremely common in OE (over 500 occurrences, of these more than 200 in Ælfric). It is not limited to West Saxon, however, featuring also in the North, for example, in the interlinear glosses to the Lindisfarne Gospels, nor to the Bible and biblical commentary, occurring, e.g., in liturgical texts and charters.Footnote ⁶ The use of the name Jesus <Iesus, Ihesus> is, in contrast, very rare (cf. OED s.v. Jesus) and, for the most part, restricted to the explanatory passages, such as the one quoted in example (1). Although, when used on its own, Hælend is typically preceded by a demonstrative, its status remains somewhere between a common noun and a proper name. This is why its demise in early ME should be examined next to the rise in the use of both saviour and Jesus.

Unfortunately, mapping proper names is not possible with LAEME, as they are not treated as lexels, so figure 2 represents only those text languages that use Helend (148 occ. in 24 texts) and saviour (1 occ. in # 295, MS Cotton Vespasian A.iii containing a version of Cursor Mundi and other religious poems, dated to C14a). Thanks to the kind assistance of the LAEME team I was able to obtain a list of all proper names from the corpus.Footnote ⁷ I then manually selected all instances of Jesus and checked the files that contained it for date and localisation.

Figure 2. Geographical distribution of hǣlend vs saviour in LAEME

Although still more frequent in the late twelfth century (figure 3), its relative portion being just over 50 per cent compared to Jesus, Helend is probably construed less and less as a proper name, as the collocations with Christ are not numerous (seven altogether in 5 texts: ## 301, 1200, 1300, C12b2; # 184, C13; # 246, C13b1). Explanations similar to that given in example (1) above are also found in early homily collections, the Trinity (# 1200, C12b2) and Lambeth Homilies (# 2001, c.1200), which are well known for their close kinship with the OE homiletic tradition (Swan & Treharne Reference Mary and Treharne2000), e.g. in this commentary on the Creed:

1. (2) & in ihesum christum. and ich ileue on þe helende crist. filium eius unicum. his enlepi sune. dominum nostrum. ure lauerd he is ihaten helende for he moncun helede of þan deþliche atter. (# 2001, Morris 1868: 75)

2. ‘& in ihesum christum. and I believe in Jesus Christ. filium eius unicum. his only son. dominum nostrum. our lord. He is called helende because he has healed mankind from the deadly poison.’

Figure 3. Relative frequencies of Helend, Jesus and Saviour between the late twelfth and early fourteenth centuries

It is illuminating that the scribe of # 2001 is still able to pick up the word play: healer/saviour who has healed/saved us from poison/corruption. The etymological relationship between the verb helen and helende crist was probably lost upon the scribe of # 1200, for he uses alesede ‘delivered, redeemed’ in his version of the commentary (cf. Käsmann Reference Käsmann1961: 154).

From the early thirteenth century, Jesus, typically in the form <Iesu> from French (Käsmann Reference Käsmann1961: 51; OED s.v. Jesus), if spelt out (28 per cent of occurrences), or in the form <IHU>, if abbreviated, gains increasingly more ground, ousting Helend completely by the end of the century. It is soon after this that Saviour is introduced from French. Käsmann connects the demise of Helend with its obsolescent morphology: indeed, all reflexes of OE agent nouns in -end die out in the course of the early ME period, e.g., alesend ‘redeemer’ or sheppend ‘creator’ (Reference Käsmann1961: 51). Although morphological unanalysability is an important concern, I would argue that its semantic and categorical arbitrariness was at least an equally valid factor. Again, we are dealing with a marked lexeme, with special connotations, belonging to a high register. Its juxtaposition with the verb heal (OE hǣlan), from which hǣlend was derived, is revealing. Its ME reflex hēlen is a polysemous and unmarked verb, which means ‘to cure, heal; become healed; reform, improve, save’ (MED, s.v.). As example (2) shows, these senses can still trigger an educated pun, which probably suggests that the connection between the verb and its derivative was no longer straightforward for more ordinary speakers, who could not easily associate a ‘healer, leech’ with ‘reformer, improver’ and ‘saviour’. The obsolescent morphological make-up of Helend made things even more ambiguous. Moreover, a special devotion to the holy name of Jesus that was developing in the thirteenth century under Franciscan influence (Renevey forthcoming), and the increasingly more expected practice of the recitation of the Creed and Articles of Faith by laymen (Reeves Reference Reeves2015), could have contributed to the marginalisation of Helend as a sacred name, making its usage in prayers archaic and unwanted.

All in all, word counts in OE are a good predictor of survival in ME, provided high-frequency terms are unmarked in ME. This finding is illustrated by several maps: of doom (vs judgement) and heathen (vs paien and saracen), taken from the 500+ list (figure 4), and of god-spellere (c.270 occurrences in OE; vs evangelist) and witega (c.450 occurrence in OE; vs prophet) (figure 5).

Figure 4. Geographical distribution of high-frequency terms doom vs judgement (left) and heathen vs paien and saracen (right) in LAEME

Figure 5. Geographical distribution of lower-frequency terms gospeller vs evangelist (left) and witega vs prophet (right) in LAEME

My last illustration attempts to relate lexical change to social innovation more generally. We have seen in the case study of Helend that the changes introduced into devotional practices in the thirteenth century may have contributed to the demise of this lexeme. Likewise, we would expect ecclesiastical innovation to be reflected in the growth of the lexicon (explored more fully in section 2.4).

Figure 6 shows the geographical distribution of friar (22 tokens, attested from C13b2 on) and of to preach (89 tokens from C13a2 on), preacher (14 tokens from C13a2 on) as opposed to bodian ‘to preach’ (30 tokens until C13b2). As ‘friar’ was a new concept whose introduction into medieval English society had an immediate impact on the lexicon (Wenzel Reference Wenzel1983; Durkin Reference Durkin2009: 3–7), the expansion of preach at the expense of bodian from the second quarter of the thirteenth century appears to be closely related to the innovation of friar. The two lexemes can be pinned down to the same survey points in the West Midlands and South West, and their spread may have gone hand in hand in other regions too. It is quite possible that there was also a semantic association between the new style of preaching and the mendicant orders, while bodian and bodung, although expected of the parish priests, were taken as something different and gradually fell into disuse. The relation between social innovation, geography and lexical changes is explored in the remaining sections.

Figure 6. Geographical distributions of friar (left) and preach, preacher vs bodian (right) in LAEME

2.2 Maps explained in more detail

The distribution maps in figures 4–6 are not colour-coded for period and need more detailed explanation in relation to ME dialectology and its written record. The tendencies observed are directly related to the regional coverage of LAEME across subperiods of early ME. The word counts per subperiod and dialect area are available in Gardner's (Reference Gardner2014) study of ME derivation, which uses LAEME for the early period. Gardner divides corpus files into five subperiods of roughly forty years (Reference Gardner2014: 40; for the rationale and advantages of forty-year time frames, see ibid.: 40–1) and then cross-tabulates diachronic and regional data, distributing LAEME files into the traditional ME dialect groupings.Footnote ⁸ Her tables are reproduced here as tables 2 and 3.

Table 2. The division of LAEME into five subperiods (from Gardner Reference Gardner2014: 41)

Table 3. Regional coverage of LAEME across subperiods (from Gardner Reference Gardner2014: 43)

Texts from the West Midlands account for the majority of the corpus (c.43 per cent) and dominate numerically in EMidE II, EMidE III and EMidE IV. East Midland texts (c.22 per cent) are distributed more evenly across subperiods but predominate in EMidE I. The South West (c.13 per cent) is sufficiently covered only from EMidE III onwards, while the North (c. 10 per cent) and the South East (c. 6 per cent) increase in coverage only in EMidE V (percentages are taken from Gardner Reference Gardner2014: 42; for word counts per region, see ibid.). Regarding chronology, as we would expect, earlier texts up to c.1200 show more lexical affinity with the OE tradition. Thirteenth-century texts still largely respect OE lexical practice but also display new French terms; texts from c.1300 are increasingly more innovative. Not surprisingly, therefore, many conservative lexical features are observed in the East Midlands (this region providing most of the data for EMidE I) and parts of the West Midlands, while variation between old and new tendencies is typically more pronounced in the West Midlands (dominating EMidE II, EMidE III and EMidE IV), and lexical innovation shows up a lot in the North and South (northern and southern texts, taken together, outnumbering the Midlands in EMidE V, 78 per cent and 22 per cent respectively).

2.3 The West Midlands in the thirteenth century

Even a brief look at LAEME maps and survey points reveals that the West Midlands stands out where the density of coverage is concerned, especially in the north of the dialect area, which includes Gloucestershire, Worcestershire, Herefordshire and southern parts of Shropshire. This is hardly surprising since the intellectual prominence of this part of Mercia in the OE period and the continuity of its pre-1066 literary tradition have long been acknowledged by scholars (Bately Reference Bately, Kennedy, Waldron and Wittig1988; Millett Reference Millett and Edwards2004). For centuries before the Norman Conquest, these counties enjoyed a more intense and more protected cultural life (especially after the beginning of the Viking raids in the late eighth century) than other parts of England. Mercian scholars were instrumental during the Alfredian revival in the 890s, and the bishops of Worcester played key roles during the Benedictine reforms of the tenth century. After the council of Westminster (1070), when a number of English bishops were deposed and many vacancies had already been filled by the Normans, only a few episcopal sees were still administered by the people appointed during the reign of Edward the Confessor. Among them were Walter of Lorraine, bishop of Hereford (in office 1060–79) and Wulfstan, bishop of Worcester (1062–95). Wulfstan's authority was particularly important in the preservation of religious foundations and their archives in the region (Mason Reference Mason1990). Therefore, when the seeds of new methods and styles of preaching began to spread from the Continent in the second half of the twelfth century and especially after 1215, in the West Midlands, they fell on fertile soil. Historical interest in vernacular written texts, a concentration of old cathedral centres and new mendicant houses, the proximity to Oxford, and, critically, the bishops who were active in the pastoral reform, all contributed to the proliferation of literary production in English and the circulation of texts across the region (Millett Reference Millett and Edwards2004: 9–10). It is in this context that the Lambeth Homilies, Ancrene Wisse, the Katherine and Wooing collections were written and copied, and it is in this context that the Tremulous Hand of Worcester annotated OE manuscripts (Laing Reference Laing, Taavitsainen, Nevalainen, Pahta and Rissanen2000: 108; Franzen Reference Franzen1991, Reference Franzen2003).

Although the decisions of the Fourth Lateran Council provided a necessary impetus for the social changes that we observe from the 1220s onwards, i.e. a wider support for grammar schools at cathedral level that could provide basic education for the laity, the increasing availability of English texts for religious instruction and private reading, and the availability of preaching in English that we can deduce from this, the first indications of these developments are already present decades earlier at both legislative and executive levels. For example, canon 18 of the Third Lateran Council (1179) required every cathedral church to assign a schoolmaster to instruct the clerics of that church and children from poor families for free (Orme Reference Orme2006: 201–2). Similarly, the programme of popular preaching had been advanced and elaborated in the last quarter of the twelfth century by the great Parisian theologians Peter Cantor (d. 1197) and Alan of Lille (d. 1203). The influence of their doctrines can be seen in all texts of the AB language (Millett Reference Millett and Edwards2004: 9).

In 1215, a major programme of pastoral reform was launched. The decisions of the Fourth Lateran Council provided for instructors in Latin grammar and ‘in matters which are recognized as pertaining to the cure of souls’ to be appointed by all cathedral churches and other churches with sufficient resources (canon 11). Bishops were expected to recruit preachers to help them ‘minister the Word of God to the people’ in large dioceses and assist with ‘hearing confessions and enjoining penances’ (canon 10). At the same time, Christians of both sexes were obliged to confess their sins and receive the Communion at least once a year (canon 21). In response to this, ‘[t]he teaching syllabus evolved by the English bishops . . . required the laity to know and be examined on the Creeds, the Pater Noster, the Commandments and the Deadly Sins, sometimes supplemented by the Sacraments and later joined by the Ave Maria’ (Gillespie Reference Gillespie and Edwards2004: 129). It goes without saying that clerical instruction on such a universal scale had to be undertaken in the vernacular. The demand for popular preaching was answered by the existing cathedral and monastic structures, but even more so by the new mendicant orders that were making astonishing progress both on the Continent and in Britain. The Dominicans arrived in 1221 and established their first base at Oxford. The Franciscans followed suit in 1224 and made their urban settlements at Canterbury, London, Oxford and Northampton. By 1250 both orders already had dozens of foundations across the country, including Franciscan houses in Worcester (1227) and Hereford (1228), and Dominican houses in Shrewsbury (before 1232) and Chester (before 1236) (Bolton Reference Bolton, Szarmach, Tavormina and Rosenthal1998: 305; Millett Reference Millett and Edwards2004: 10). A well-known reformer, Robert Grosseteste, later also a Franciscan, was active in the diocese of Hereford between 1195 and 1220, and most of the bishops, active in pastoral reform in the West Midlands, were linked to him. Bella Millett observes,

A similar repackaging of older native lexical resources for new instructional purposes must have taken place: established frequent terms were preserved, while those associated with changing practices of, e.g., confession and penance, were replaced with French loans (Timofeeva forthcoming).

2.4 Reforms, lexical innovation and social networks

Establishing the exact connections between linguistic change and social change has been the principal task for historical sociolinguistics, at least since the 1980s (Milroy Reference Milroy1987; Trudgill Reference Trudgill2000; Labov Reference Labov2001; etc.). The finding that rapid and/or large-scale lexical innovation and, especially, adoption of loanwords are indicative of social change is hardly a surprising outcome in this investigation. What needs to be addressed therefore is the question of the social networks that allowed thirteenth-century lexical innovation to happen, the patterns and conditions that defined its speed.

The reform movement of the late twelfth to early thirteenth century can be seen as the major driving force behind the introduction of new Romance-based religious terms in Middle English. Innovative lexical tendencies are more observable in the West Midlands because the initial stages of the new social developments and textual attestation were also more pronounced in this region. At the same time, both Midland areas show a stronger connection to the OE written tradition and its religious vocabulary.

With grammar-school education being typically available via the French medium, the higher schools and universities functioning in both Latin and French (Orme Reference Orme1973: 71–8; Ingham this volume), and Francophone schoolmasters and friars being more readily present in urban centres, we would expect lexical innovation to diffuse in episcopal and university cities and other large and rich towns that maintained grammar schools. Those immediately affected were the children of the nobility and social climbers who attended these schools (see Richter Reference Richter1979), but public preaching could also reach much broader audiences. Friars, Franciscans in particular, more mobile and resourceful, were trained to preach in a variety of situations – in ‘the cloisters, market squares, gardens, preaching crosses and other open places; . . . at fairs’ – and in more than one language – ‘in the vernacular, French and Anglo-Norman with Latin as the medium for clerical and scholastic congregations’ (Robson Reference Robson2017b: 20). Because friars created their own system of school and university education (O'Carroll Reference O'Carroll1980; Bolton Reference Bolton, Szarmach, Tavormina and Rosenthal1998), they were less dependent on the old written tradition in general and freer to generate new linguistic norms. In the latter process, common English terminology was naturally preserved, while rare and ambiguous terms were rejected. Considering that many friars travelled a lot both locally and across the country, it is easy to envisage how their French-influenced linguistic practices would spread together with them through preaching, charity work and spiritual ministrations. Moving easily among people of any social standing (Robson Reference Robson2017a: 456–7), they would also provide the necessary weak ties for the diffusion of new religious lexis both within and across social classes, including the very low and marginalised. Given the extent of their geographical and social mobility and the extent of multilingualism and language fluency expected from them, it is understandable why the introduction of French terms was often preferable to the reanimation of old or the creation of new lexical items in English. Not only could it facilitate less taxing code alternation between English, French and Latin on the part of the preachers, but it also allowed for a kind of lexical standardisation across the three (and possibly more) codes, if important terminology could be kept ‘stable’, that is to say, Romance.