On the long history of English adverbial subordinators

Matti Rissanen, University of Helsinki

1. Introductory

Adverbial clauses are traditionally defined as one of the three main types of subordinate clauses. They differ from noun clauses and adjective clauses in that the subordinators are much more numerous and varied, owing to the richness of the relations between the main clause and the adverbial subordinate clause. Perhaps for this reason, there is no detailed long-diachrony (from Old to Present-Day English) corpus-based survey of the appearance, development and distribution of the whole stock of English adverbial subordinators. [1] Kortmann (1997) is a most valuable general account of the development of the adverbial subordinators in English, although his study focuses on their typological aspects rather than on the details of grammaticalisation or borrowing. The most impressive account of Old English adverbial clauses is included in Mitchell’s Old English Syntax (1985). The main outlines of the developments from Old to Present-Day English can be found in the syntax chapters of The Cambridge History of the English Language (Vols I–IV, 1992–1999). A number of corpus-based article-length studies of the rise and development of individual adverbial connectives have appeared in the last decade or two. Lenker’s study (2010) on the history of adverbial connectors in English is an excellent long-diachrony survey of a group of connectives closely related to adverbial subordinators and also includes valuable observations on adverbial clause linking in more general terms (see also Lenker, this volume).

The purpose of my paper is to present a kind of introduction to comprehensive and detailed surveys of the history of English adverbial subordinators. The most important developments will be briefly outlined and illustrated with a few examples, with reference to grammaticalisation, borrowing and genre- or register-based variation. For details, earlier studies of individual subordinators are referred to, and topics for further study suggested. My survey also touches on relevant questions concerning the development of adverbial subordinators presented by Kortmann (1997: 291):

  1. To what extent has the inventory of adverbial subordinators been subject to fluctuation over the centuries (e.g., due to French influence), and in which semantic domains in particular?
  2. What are the major changes in the morphological make-up of adverbial subordinators and, more generally, what can we say with regard to the preferred sources of adverbial subordinators in older periods of English?
  3. Can the development of adverbial subordinators in English really be described as one towards an ever higher degree of semantic differentiation and a decrease in (semantic and syntactic) polyfunctionality?

Although the main focus of this article is on subordinators, i.e., on elements introducing a subordinate clause, it is impractical, or even impossible, to separate adverbial subordinators from the prepositions based on the same lexical items. In fact, some scholars define preposition-derived adverbial subordinators as prepositions governing a clause instead of a noun phrase (see Huddleston & Pullum 2002: 599ff). See my examples (1.a) and (1.b).


fela fortacna sculon geweorðan wide on worulde … ær ðam þe se dom cume þe us eallum wyrð gemæne. (Wulfstan’s Homilies 123 WULF3 HC) [2]

‘many signs will appear widely in the world … before the doom comes that will be common to all of us.’


nan man nys þe hyg wite ær þam miclan dome. (Prose Solomon and Saturn 32 SOLOM HC)

‘there is nobody who will know it before the great doom.’

In (1.a) the preposition ær forms part of the prepositional subordinator ær ðam þe introducing a temporal clause, while in (1.b) it is used as a preposition governing the noun phrase þam miclan dome (dative).

The main stages of development of adverbial subordinators can be fairly easily linked with the traditional period division of the English language, Old, Middle, Early Modern and Late Modern English (cf., e.g., Kortmann 1997: 291–335). In the Old English period, from the beginnings to about 1150, new ways of signalling adverbial subordination saw the light of day and important developments including grammaticalisation took place. Middle English (1150–1500) and Early Modern English (1500–1700) are interesting in view of the increasing variability of function words. This is not surprising: the Norman Conquest and the resulting radical changes in society and culture brought about a need for new ways of indicating relations between concepts and proposals, and the Renaissance, preceded by the Late Middle English reintroduction of English as a language of officialdom and literature, had a similar effect. Throughout the Middle English period borrowing from French and to a lesser degree from Latin enriched the stock of adverbial connectives; in Early Modern English the influence of Greek and Latin had a similar effect. The printed book gave a new dimension to people’s use of, and attitudes towards, language. In Late Modern English, from the beginning of the eighteenth century on, specialisation on the senses of subordinators and their style- and register-based distribution was further established.

This brief outline of the characteristics of the periods of English which should be taken into account in the discussion of adverbial subordinators suggests that my approach is strictly variationist and I will include observations on the influence of language-external factors in my discussion. The value of corpora is of course indispensable in this kind of research, and my examples and tables of occurrence are based on a large number of English historical and present-day corpora. My main corpus sources are: The Helsinki Corpus of English Texts (HC); The Dictionary of Old English Corpus (DOEC); The Corpus of Middle English Prose and Verse (CMEPV); The Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME); The Corpus of Early English Correspondence (CEECS/PCEEC); The Corpus of English Dialogues (CED); The Penn Parsed Corpus of Modern British English (PPCMBE); A Representative Corpus of Historical English Registers (ARCHER); The Corpus of Late Modern English Texts (CLMET); The Brown Corpus (Brown); The Lancaster-Oslo/Bergen Corpus (LOB); The Freiburg-Brown Corpus (Frown); and The Freiburg-Lancaster-Oslo/Bergen Corpus (1990s) (F-LOB).

2. Old English

As mentioned above, the texts dating from the Old English period show a number of interesting features related to the origin and development of adverbial subordinators. The most obvious of these is that adverbial connectors can be combined with the particles þæt or þe, and some of these combinations seem to be at least to some extent grammaticalised.

Gif ‘if’ can be regarded as the most established of the Old English adverbial connectives. It is used only as a subordinator in Old English texts and is not followed by the particle þæt or þe. [3]

The role played by the particle þæt or þe is relevant in the study of the development of Old English adverbial subordinators. Their function is to link the connector with the following subordinate clause and mark the subordinator in this way. Þæt and þe could be used as subordinating links even by themselves, þe mainly with adjective clauses and þæt with both noun and adjective clauses, and with some types of adverbial clauses, particularly those indicating purpose or result (see, e.g., Mitchell 1985: §§2825–2826).

It is most likely that subordinate adverbial clauses developed from sequences of coordinate clauses in which hierarchical ordering of the clauses was felt to be important (cf. Mitchell 1985: §1688). As a result, the connector, i.e., the adverb or adverbial phrase, indicating, for instance, a local, chronological, causal, or conditional relationship between the clauses, developed into a subordinator indicating the hierarchy between the propositions. It is not unlikely that the development of written Old English, supported by increasing knowledge of and translations from Latin, had an influence on the development and establishment of these subordinators. It would also be tempting to suggest – although impossible to prove – that allegro speech, which easily omitted clear breaks between clauses, supported the development of new grammatical devices of marking divisions and mutual relations between clauses.

Besides gif, there are a few other adverb-based simple subordinators in Old English, which were mainly used without þæt or þe, the most frequent ones being ær and siþþan. Both go back to temporal adverbs and thus probably represent a very early and common type of clause links. Siþþan was originally a compound adverb (siþ + þan instr; see Mitchell 1985: §2666; Molencki 2007: 98–107), but was probably grammaticalised at an early date, losing its compound appearance.

The most interesting group of Old English adverbial subordinators, from the point of view of the development of this word-class, is the one based on adverbial phrases consisting of a preposition (ær, æfter, be, for, mid, wiþ, etc.) and the oblique form of the demonstrative pronoun, þæm/þon/þy (examples 2–9). These phrases were probably first used as adverbial connectors but developed into subordinators in the course of the Old English period. In subordinator use, the particle þe could be appended to the phrase (examples 3, 5, 8). Þæt is also possible (examples 6 and 9). In examples (2), (4), and (7), the prepositional phrase is not followed by þe or þæt.


& þæt longe donde wæron ær þæt folc wiste hwonan þæt yfel come, buton þæt hie sædon þæt hit ufane of ðære lyfte come, ær þon hit þurh ænne þeowne mon geypped wearð (Orosius 108 OROS HC)

‘and they were doing that for a long time before people understood wherefrom the evil came, except that they said that it came from above, before it was revealed by a slave.’


He ys gehaten eac Crist; be þam cwæð sum witega for fela hund gearum, ær þan þe he acenned wurde (Ælfric’s Letter to Sigeweard 45 LSIGEW HC)

‘He is also called Christ, of whom some prophet spoke many hundreds of years ago, before he was born.’


he nyste hwæt þæs soþes wæs, for þæm he hit self ne geseah (Ohthere 17 OHTHR2 HC)

‘He did not know what the truth was because he did not see it himself.’


ic ondræde me for ðam ðe ic eom nacod (Hept. Gen. 3 AELFOLD HC)

‘I am afraid because I am naked.’


þis wæs geworden, forþon þæt se witedom wære gefylled þe ær gecweden wæs (Blickling Homilies 71 BLICK6 HC)

‘This happened so that the prophesy would be fulfilled that was mentioned earlier.’


þa ongan ic wundrian, for hwon wit to þam walle eodan, mid ðy ic on him nænige duru ne eahþyrl … on ænge halfe geseon meahte. (Bede 428 BEDEHE HC)

‘Then I began to wonder why we went to the wall, as I could not see any door or window in it … on either side.’


Hælend … hine het to him gelædon; & mid þy þe he him genealæhte, he him tocwæþ, Hwæt wilt þu þæt ic þe do? (Blickling Homilies 15 BLICK2 HC)

‘Christ … asked him [the man] to be led to him, and when he [the man] approached him, he asked him, “What would you like me to do for you?”’


Ac hie wæron micle swiþor gebrocede on þæm þrim gearum mid ceapes cwilde & monna, ealles swiþost mid þæm þæt manige þara selestena cynges þena þe þær on londe wæron forðferdon on þæm þrim gearum; (Anglo-Saxon Chronicle 89–90 CHRONA2 HC)

‘But they were much more seriously afflicted in those three years by the death of cattle and men, most of all in that many of the best thanes of the king who were in the country died in those three years.’

The use of þe with these subordinators increases in the course of the Old English period; this may be interpreted as an indication of the first steps of the grammaticalisation of the subordinator. Table 1 (from Rissanen 2007b: 190), based on roughly 2,000 occurrences, shows the increasing use of þe in the Old English period with three subordinators, ær, for, and mid + a demonstrative pronoun:

Table 1. Mean frequencies (/100,000) of ær/for/mid + a demonstrative pronoun as subordinators in the Old English section of the Helsinki Corpus. Percentages of occurrences with þe/þæt.


ær + dem. pron.

for + dem. pron.
‘for’, ‘because’

mid + dem. pron.
‘when’, ‘while’, ‘as’







OE2 850–950


9 (50%)


139 (24%)


9 (18%)

OE3 950–1050


14 (54%)


126 (31%)


19 (67%)

OE4 1050–1150


13 (93%)


164 (53%)


3 (75%)

The column for for is the most illustrative as the number of occurrences is not too low (more than 1,500 instances). The infrequency of zero with ær and mid in the latest Old English period is also worth noting. (cf. Mitchell 1985: §3155, fn. 167; see also Lenker 2010: 140–147).

It could be expected that the particle þæt, which was used in varying connective functions in Old English, would have been equally common with the prepositional subordinators as þe. That this is not the case [4] may be due to the analogical formation of new connectives with þe in the Old English period. In fact, there are only a few subordinators with which þæt, varying with zero, is more common than þe; the most notable of these are butan and (examples 10–11), and the prepositional phrase to þam/þon (12). It seems that the subordinating particle þæt introducing a clause of exception, destination or purpose is not necessarily as closely linked with the subordinator as þe and its use approaches that of the noun clause subordinator þæt (see Kilpiö & Timofeeva, this volume, referring to Mitchell 1985: §3155, fn. 167). This suggestion is supported by the use of þæt with butan only when the meaning is ‘except (that)’, ‘save only (that)’. When butan is used in the sense ‘unless’ (example 13), it is not followed by þæt (see Mitchell 1985: §3628).


Hwæt wille we eow swiðor secgan be ðisum symbeldæge. buton þæt maria cristes modor wearð on ðisum dæge of ðisum geswincfullum middanearde genumen up to heofenan rice. (Ælfric Hom. 259 AELFR2/29 HC)

‘What will we say more to you about this feast-day except that Mary the Mother of Christ was on this day taken from this troublesome earth to the Kingdom of Heaven.’


Nu ðu nelt gelyfan minum wordum, beo ðu dumb oð þæt þæt cild beo acenned. (Ælfric Hom. I 202.6 AELFR DOEC)

‘Now that you will not believe my words, be dumb until the child be born.’


þa feng Sameramis his cwen to þæm rice, & getimbrede þa burg Babylonie, to þon þæt heo wære heafod ealra Asiria. (Orosius 60 OROS HC)

‘Then Semiramis his queen became the ruler and she built the city of Babylon for the purpose that it would become the capital of all Assyria.’


Soð ic eow secge, þæt ge sylfe ne becumað into heofonan rice, buton eower rihtwisnys beo eallunga mare ætforan minum Fæder þonne þæra bocera is and ðæra Sunderhalgena. (Ælfric Hom. 533 AELFR15 HC)

‘I tell you truly that you will not come to the Kingdom of Heaven unless your righteousness be altogether greater before my Father than that of the Scribes or Pharisees.’

The items discussed above do not cover all types of Old English adverbial subordinators. There are some less common items, such as þenden ‘while’, sam … sam ‘whether … or’, swelc(e) ‘as if’. The comparative clause links formed with swa (swa swa ‘as’, swa sona swa ‘as soon as’, swa feor swa ‘as far as’, etc.) are also closely related to adverbial connectives.

3. Middle English

There are two main lines of adverbial subordinator development in the Middle English period: (1) the simplification of the prepositional subordinators in Early Middle English, with the popularisation of that as a subordination marker, and (2) the emergence of new subordinators, mainly borrowed from French or Latin, including a number of loan translations.

As early as the twelfth century the three-element prepositional phrase subordinators described above give way to the combination of the preposition and the subordinating particle that (examples 14–16). The adverbial connective use of þe disappears in Early Middle English (see MED s.v. the (conj.)); the change is rapid but not unexpected in view of the increasing functional load, caused by the establishment of the definite article þe > the. The subordinator preposition + demonstrative pronoun + þe (e.g. for þam þe) did not return to the connector form preposition + demonstrative pronoun (e.g. for þam), but developed the form preposition + that (e.g. for that). [5] The replacement of þe by that is not in itself surprising in view of the frequent subordinator use of this connective in other contexts. But it is of interest that that came to be used even with subordinators which were used independently in Old English, such as if, example (16).


sunnendei fond noe lond efter þet ure drihten hefde þet folc adreint. (Lambeth Homilies 141 LAMB14 HC)

‘On Sunday Noah found land after our Lord had drowned those people.’


Confitemini domino quoniam bonus; kneoweð ure louerd for þat he is wel god. and swo mild heorted. (Trinity Homilies 71 TRIN12 HC)

‘Thank Lord, for he is very good and so merciful.’


Þe seffnde god uss shall ȝet don
Þe Laferrd Crist onn ende,
Þurrh þatt he shall o Domess daȝȝ
Uss gifenn heffness blisse, ȝiff þatt we shulenn wurrþi ben
To findenn Godess are. (Ormulum 229 ORM DED HC)

‘Lord Christ shall yet do us the seventh good through that he shall on Doomsday give us the bliss of Heaven if we will be worthy to receive God’s mercy.’

The particle that is most common in late fourteenth and early fifteenth century texts (including Chaucer’s writings) and occurs fairly frequently even in later 15th-century texts (see Rissanen 1997). Its use diminishes radically in the Early Modern English period.

Why did that become so popular in Middle English that it came to be appended even to such unambiguous subordinators as if? Very tentatively, it could be suggested that, while the subordinating particle þe disappeared, the wish to mark subordinate adverbial clauses with another particle was enhanced by the fact that word-order, i.e., the clause-final position of the finite verb as a marker of the subordinate clause, disappeared at that time. Old Norse influence is also possible – most of the earliest examples of if that go back to Ormulum.

As to the new adverbial connectives emerging in the Middle English period, the most obvious sources are of course French, and to some extent, Latin. The most important native innovations are before, which competes with ere (< ær), and besides (that). Combinations of two prepositions can also be found, notably for to > forto > forte. Till and until go back to Old Norse (see, however, the account of the origin of till in Rissanen 2007a). The borrowings or loan translations from French and Latin include, among others, because, in case, provided, save, except, in despite of, according, outtaken/outnome (cf. Latin excepto), unless (cf. Old French à moins que), notwithstanding (cf. Latin/Old French non obstant[e]), and all be it (that) (cf. Old French tout soit il/ce que; see Sorva 2006 and Sorva 2007).

One feature distinguishing these new connectives from the Old English ones is that many of them are derived from verbs.

The lexical items from which the loan connectives were derived were also borrowed into Middle English. The question arises whether the grammaticalisation of the connective took place in Middle English or whether the connective had been grammaticalised in the source language, French or Latin. In most cases the dates of the first Middle English occurrences of the lexical item and the connective are too close to each other to allow sufficient time for grammaticalisation; thus the borrowing of the grammaticalised connective from the source language is more likely. To give an example: the earliest recorded Middle English occurrence of the verb except dates from 1393, example (17), and the earliest example of the connective use from about 1400, example (18).


To the povere and to the riche Hise lawes myhten stonde liche, He schal excepte no persone. (a1393 Gower CA (Frf 3) 7.2745 MED)

‘His laws should be alike to the poor and to the rich; he must not exclude anybody.’


Þer were suche eremites, Solitarie by hem-self and in here selles lyueden … Excepte that … an hynde oþer-while To hus selle selde cam and suffrede to be melked. (c1400 [?a1387] PPl.C Hnt HM 137. 18.9 MED)

‘There were such hermits who lived solitary in their cells … Except that … a hind sometimes came to his cell and let herself be milked.’

Yet the role of Middle English lexical uses is of importance in the establishment and further grammaticalisation of the connective. The introduction of a new grammatical item, particularly when it comes from a foreign source, involves the interaction between the speaker/writer and hearer/reader. When the speaker or writer uses a borrowed grammaticalised item, the reader most naturally associates it with other, non-grammaticalised uses of the same word or stem, and this process is important in the adoption and establishment of the grammaticalised form. Reference can be made to Traugott and Dasher’s (2002) theory of semantic change and their definition of the Invited Inferencing Theory of Semantic Change (IITSC). According to Traugott and Dasher (2002: 5) ‘the speaker/writer (SP/W) evokes implicatures and invites the addressee/reader (AD/R) to infer them’. All the lexical uses related to the borrowed connective in the receiver’s cognition affect his or her interpretation of the new grammaticalised item.

This process can be illustrated by the early history of the connective considering (that). The earliest Middle English examples of the subordinator are recorded from early fifteenth century texts (19), no doubt formed after the pattern of Anglo-Norman considerant. The verb consider appears in the late 14th century and the present participle form, used as in examples (20) and (21), certainly contributed to the establishment of the connective considering (that): the grammaticalised form in (19) differs from the use in these examples only in that a personal subject, such as Henry in (20) and Mayor and aldermen in (21), is no longer apparent.


Consederyng þe pacyent to be myche febled in þe day of accesse … if he ware þane flebomyed … he scholde be more feblid. (1400-25 Phlebotomy PHLEB 47 HC)

Considering the patient to be much weakened on the day of the attack [of the illness] … if he were then bled … he would be more weakened.’


In þis tyme, whil he was in Yrlond, þe duke of Lancastir, Herry, beryng heuyly his exile and eke priuacioun of his heritage, and considering who euel-beloued þe kyng was … cam into þe se with Thomas, bischop of Cauntyrbury. (Capgrave 211 CAPCHR HC)

‘At this time, while he was in Ireland, the Duke of Lancaster, Henry, enduring with difficulty his exile and also the deprivation of his heritage, and considering how much disliked the king was … went to sea with Thomas, Bishop of Canterbury.’


Therfor þe seyd Mair and Aldirmen, considering how such worthy persones as þe same knyght is … myght lightly and causeles renne in sclaundre … han awarded … that … (1418 Judgments 96 JUDG HC)

‘Therefore the said Mayor and aldermen, considering how such a worthy person as this same knight is … might lightly and without any reason be slandered … have decided … that …’

On a corpus-based diachronic study of considering, see also Meurman-Solin & Nurmi 2004 and Meurman-Solin & Pahta 2006.

The richness of new adverbial connectives in Middle English calls forth the question of the factors affecting the choice between roughly synonymous items. This question can be approached through the study of Late Middle English genre-based distribution. Some items go back to formal, official genres, while others become popular in less formal contexts. The uses of the connectives save/saving and except(ing/ed) can be taken as an example (see Rissanen 2002 and Rissanen 2009).

Of these two connectives, save appears in early fourteenth-century texts, earlier than except:


God ȝaf him [Adam] a gret maist[r]i Of al þat was in watir and londe … Saf o tre he him forbede. (?a1325 Þe grace of god Hrl 913. 46 MED)

‘God gave him [Adam] great mastery of all that was in water or on land … Save one tree he denied him.’

It is thus not surprising that in Late Middle English save is clearly more common than except. This is shown by Table 2, based on the Corpus of Middle English Prose and Verse.

Table 2. Save/saving and except(ing/ed) in Late Middle English texts in the Corpus of Middle English Prose and Verse. Absolute numbers.










The popularity of save compared with except can be seen if the occurrences in different genres are compared. The Helsinki Corpus with its chronological and genre-based arrangement offers the easiest basis for this kind of comparison. See Table 3.

Table 3. The occurrences of the grammaticalised (subordinator and preposition) forms of save/saving and except(ing/ed) in various texts and genres of Late Middle English texts in the Helsinki Corpus (ME3 and ME4, 1350–1500). Absolute numbers.

Statutes and documents11Statutes and documents8
Astrolabe1Reynes, Commonplace1
Canterbury Tales5Metham, Days of the Moon2
Early Planets4 Chyrurgie de Chauliac1
Cloud of Unknowing1Margery Kempe1
Siege of Jerusalem1Preface, Caxton1
Reynes, Commonplace3Private letters2
Serm. Innocents2
Capgrave Serm.1
Capgrave Chron.2
London Chron.2
Reynard the Fox3
York Plays2
Towneley Plays1
Private letters3

Table 3 shows that, in general, expressions of exception are typical of statutes and official documents: about 25 per cent of the occurrences of save and 50 per cent of the occurrences of except, while these genres only represent some 10 per cent of the texts included in these two sub-periods. The table also shows that save is more common than except in other types of text: 32 occurrences as against the eight occurrences of except. Particularly its use in the Canterbury Tales, Reynard the Fox, and the York and Towneley Plays is notable. Except is used mainly in handbooks, scientific treatises and other non-imaginative texts. Except undergoes even more markedly than save a change from above, no doubt introduced into the language through officialese.

4. Early Modern English

The Early Modern English period, the sixteenth and seventeenth centuries, witnesses the establishment and further grammaticalisation of Middle English adverbial subordinators. New connectives also emerge. Kortmann (1997: 300) lists, among others, the following: as that, being that, conditioned that, during that, for all (that), for fear that, granted that, how (that), how(so/some)ever, if case be (that), in respect (that), insomuch that, no sooner … than/but (that), say (that), seeing (that), supposing (that), what time (as). [6] As can be seen, many of these subordinators are phrasal and their level of grammaticalisation remains incomplete.

Differences in frequency and register distribution of the connectives are noticeable even in this period. This variability was certainly supported by the Renaissance explosion of printed texts and advances in science and literature. The changes can be illustrated by the frequencies of save, except and the newcomer unless (see Rissanen 2010) in the Penn-Helsinki Parsed Corpus of Early Modern English in Table 4.

Table 4. Occurrences of save/saving, except(ing/ed) and unless in the Penn-Helsinki Parsed Corpus of Early Modern English. Absolute numbers. (Mean frequencies per 100,000 words in brackets.)

1500–157049 (8.6)134 (23.6)33 (5.8)
1570–164036 (5.7)152 (24.2)65 (10.3)
1640–171022 (4.1)84 (15.5)100 (18.5)

As can be seen, except becomes clearly more popular than save in the sixteenth century. Its frequency decreases in the second half of the seventeenth century, mainly because of the increasing use of unless.

Table 5 shows that except becomes more common than save even in genres representing less formal discourse, such as private correspondence and dialogues.

Table 5. Occurrences of save/saving, except(ing/ed) and unless in the Corpus of Early English Correspondence Sampler and the Corpus of English Dialogues. Absolute numbers. (Mean frequencies per 100,000 words in brackets.)

CEECS 1 (1418–1638)35 (14.2)34 (13.8)
CEECS 2 (1580–1680)9 (4.4)23 (11.3)
CED1 (1560–1599)9 (4.5)35 (17.5)
CED2 (1600–1639)9 (4.4)11 (5.3)
CED3 (1640–1679)6 (2.3)8 (3.1)
CED4 (1680–1719)2 (0.7)14 (4.7)
CED5 (1720–1760)021 (9.4)

A sample extracted from the larger Parsed Corpus of Early English Correspondence confirms the growing popularity of except in comparison to save: in letters written in the sixteenth century the two connectives are equally common (absolute figures save 68 and except 63), while in seventeenth-century letters except is clearly more frequent (absolute figures save 28 and except 72).

5. Late Modern English

By the end of the seventeenth century the English language had reached a stage when its structure – and even vocabulary – were rather close to Present-Day English. The inventory of adverbial connectives was very similar to that of Present-Day English; some of the Late Modern English newcomers were assuming (that), given (that), in order that, insofar as/that, once, only (that) (cf. Kortmann 1997: 300). The genre- and register-based restrictions become, however, more and more obvious. To continue the survey of the frequencies of save and except, save remains a minority connective in Late Modern English. The figures based on ARCHER and the Corpus of Late Modern English Texts, which include texts from the beginning of the eighteenth to the twentieth century, are illustrative (Table 6):

Table 6. Occurrences of save/saving, except(ing) and unless in ARCHER and the Corpus of Late Modern English Texts. Absolute numbers. Mean frequencies per 100,000 words in brackets.

CLMET1 (1710–1780)13 (0.6)408 (20.0)306 (14.6)
CLMET2 (1780–1850)181 (4.9)803 (21.5)482 (12.9)
CLMET3 (1850–1920)314 (7.9)886 (22.2)518 (13.0)
ARCHER (1700–1799)3 (0.6)107 (19.9)62 (11.6)
ARCHER (1800–1899)26 (4.8)104 (19.3)64 (11.9)
ARCHER (1900–1990)9 (1.7)140 (26.2)44 (8.2)
LOB (1960s)19204154
Brown (1960s)7180101
F-LOB (1990s)6121111
Frown (1990s)713890

This distribution is somewhat surprising as monosyllabic save is shorter and easier to pronounce than except. One possible explanation for the lower frequency is that save was never grammaticalised to the same extent as except. Throughout the Late Modern English period, the lexical uses of save are more common than the connective uses, and that may have prevented its complete grammaticalisation. The grammaticalisation of except is more natural as the word and its derivatives are very uncommon in other than their connective uses. Table 7 shows the proportion of the connective save as against the other uses of the verb save in Late Modern and Present-Day English corpora.

Table 7. The connective and other uses of save/saving in CLMET, ARCHER, LOB, Brown, F-LOB, and Frown. Absolute numbers.


My final example will illustrate the possible influence of register and style on the frequency of an adverbial subordinator, the conditional connective provided (that). As can be seen in Table 8, it first appears in Late Middle English, in the Statutes and other official documents, but in the course of the Early Modern English period it spreads to other genres as well, and can even be found in highly informal contexts (examples 23 and 24, see Rissanen, forthcoming):

Table 8. The adverbial subordinator provided (that) in the Late Middle and Early Modern English parts of the Helsinki Corpus. Absolute numbers.

 ME4 (1420–1500)EModE1 (1500–1570)EModE2 (1570–1640)EModE3 (1640–1710)
Laws and documents10181324


The Rival cruelly vext; got a red hot iron, and comes again, tell her he had brought her a Ring, provided she would give him another kiss; (1684–1685 Penny Merriments 159 HC)


Hortensio   I promised we would be contributors,
     And bear his charge of wooing, whatsoe'er.
Gremio   And so we will, provided that he win her.
Grumio   I would I were as sure of a good dinner. (1623 Shakespeare, The Taming of the Shrew 1.1. 214–217)

Even in the late eighteenth century, this connective provided (that) is fairly common, as is shown by the figures based on the Corpus of Late Modern English Texts (Table 9):

Table 9. The adverbial subordinator provided (that) in CLMET. Absolute numbers. (Mean frequencies per 100,000 words in brackets.)

1710–178097 (4.6)
1780–185079 (2.1)
1850–192061 (1.5)
Total237 (2.4)

Some authors seem to use provided (that) as a stylistic marker as indicated by the high frequencies in Table 10, which includes the texts showing higher than five occurrences per 100,000 words.

Table 10. Texts and authors favouring the adverbial subordinator provided in CLMET. Absolute numbers. (Mean frequencies per 100,000 words in brackets.)

Sterne, Tristram Shandy, Sent. journey21 (13.3)
Chesterfield, Letters to his son17 (8.5)
Smith, Wealth of nations16 (8.0)
Hume, three philosophical texts11 (5.6)
Hogg, Memoires of a sinner8 (9.5)
De Quincey, Confessions of an English opium-eater3 (7.7)
Anne Brontë, Agnes Grey, Wildfell Hall13 (6.5)
Butler, three texts, varying genres12 (5.9)

Sterne’s and Hogg’s predilection of provided (that) is conspicuous. It seems that the increase in the popularity of this connective started in the eighteenth century, as a kind of ‘fashionable style’ marker. In the next century its use was fairly well established, and it gradually loses its special stylistic value in the course of the century.

In Present-Day English the use of provided (that) is restricted to written language (examples 25–27 and Table 11).


Today Mr James said: “Despite appalling difficulties during the last year the group has survived. Provided it can now achieve the essential level of new capital support to maintain and develop its market position, I believe it will again become a valuable investment.” (A38:67–68 F-LOB)


In certain cases students may be awarded support for pre-thesis studies on campus, provided that they intend to carry out their thesis research at Argonne. (H28:13–14 Frown)


I puzzled over and analysed the wording … just to make sure my initial interpretation was correct. By around four thirty a.m. I had come to the conclusion that it probably was, but then stewed over how the hell Sexton thought I could find where Stover had stashed his savings (provided there were any, of course, and that Stover had not blown them on some extravagance or other we hadn’t yet caught up with (L11:66 F-LOB)

Table 11. The adverbial subordinator provided (that) in late 20th-century corpora. Figures for miscellaneous documentary texts and scientific texts (H–J) and fiction (K–R) given separately. Absolute numbers. (Mean frequencies per 100,000 words in brackets.)

LOB (BrE 1960s)48 (4.8)26 (11.8)6 (2.7)
Brown (AmE 1960s)22 (2.2)11 (5.0)2 (0.8)
F-LOB (BrE 1990s)19 (1.9)9 (4.1)1 (0.5)
Frown (AmE 1990s)17 (1.7)8 (3.6)0

As can be seen from Table 11, the frequency of provided (that) is high in miscellaneous documentary texts and scientific texts (middle column, H–J), but extremely low in fiction (right-hand column, K–R). The higher frequency of provided (that) in LOB is worth noting. It seems that detailed genre- and author-based diachronic studies would reveal new interesting features in the development of borrowed, verb-derived adverbial connectives.

6. Concluding remarks

In this paper I have made an attempt to describe some of the typical features of the long-diachrony development of English adverbial connectives, mainly subordinators and prepositions. In Old English, the creative period of English grammatical vocabulary, most subordinators were derived from adverbs or adverbial phrases, often with the addition of the subordination marker þe or þæt. Grammaticalisation of these items was well on its way. It seems that the particle þe appended to the connective represented more advanced grammaticalisation than þæt, which was also common as a demonstrative pronoun. In Middle English, þe disappears and that becomes common as an element in adverbial subordinators, while borrowing appears to be the most important source of new items. The need for more accurate ways of indicating relationships between propositions increases, and grammaticalisation of the connectives continues. In Modern English, genre- and register-based factors influence the uses and frequencies of subordinators, and the grammaticalisation of items may remain incomplete.

I have illustrated my survey with references to the development of some adverbial connectives. In Middle English, the development of save shows the strong influence of French; the form lying behind the Middle English subordinator/preposition was most likely the French grammaticalised form sauf. The earlier use of the Middle English verb saven, supported, however, this development. The influence of the non-grammaticalised uses on the establishment of the borrowed connective is even more probable in the development of considering. The gradual spread of the adverbial subordinators to a greater variety of genres and the resulting changes in their relative frequencies can be seen when the early histories of save and except are compared. The discussion of the Modern and Present-Day occurrences of provided (that) further illustrates the roles of register and style in the choice of the subordinator.

The importance of corpora in this kind of variation-based analysis is unquestionable. Quantitative observations derived from a number of corpora of different size and structure, focusing both on chronological changes and genre and register variation, will provide new kind of information on the long diachrony of the grammatical items.


[1] In this paper, the terminology referring to clause-linking devices follows that of Lenker (2010: 1, 25–26). The most frequently used terms are “adverbial subordinator” (= a subordinating conjunction introducing an adverbial clause) and “adverbial connector” (= an adverb or adverbial phrase connecting two main clauses); the term “connective” refers here to any linguistic item linking clauses.

[2] The capitalised abbreviated title of the text in the reference line of the Helsinki Corpus examples refers to the list of texts in the Manual (Kytö 1996: 169–230). The abbreviated title of the corpus from which the examples are taken is given at the end of the reference line.

[3] According to Mitchell (1985: §3671) there are only two uncertain gif þæt examples in a total of c. 10,000 occurrences of gif.

[4] There are only seven occurrences with þæt out of the roughly 2,000 on which Table 1 is based (4 instances with for and 3 instances with mid).

[5] The most thorough discussion of this development is included in Kivimaa (1966).

[6] The earliest instances of some of these items (e.g. for all that, how that, seeing (that), and what time (as)) can be found even in Late Middle English texts.

Corpora used in the present survey

Names of compilers or corpus project leaders, and/or contact persons are given in brackets. More information on available corpora, historical and present-day, can be found in the Corpus Resource Database (http://www.helsinki.fi/varieng/CoRD/) and in the ICAME Corpus Collection http://clu.uni.no/icame/cd/. See also Rissanen (forthcoming).

Old English (750–1150)

  • The Dictionary of Old English Corpus (DOEC), about 3.5 million words (project leader Antonette diPaolo Healey).
  • The Helsinki Corpus of English Texts (HC), Old English part, about 500,000 words (project leader Matti Rissanen).

For further information on the Helsinki Corpus, see Kytö (1996) and Rissanen et al. (1993).

Middle English (1150–1500)

  • The Helsinki Corpus of English Texts (HC), Middle English part, about 500,000 words (project leader Matti Rissanen).
  • The Corpus of Middle English Prose and Verse (CMEPV), about 18 million words (project leader Frances McSparran)

For further information on the Helsinki Corpus, see Kytö (1996) and Rissanen et al. (1993).

Early Modern English (1500–1700)

  • The Helsinki Corpus of English Texts (HC), Early Modern English part, about 500,000 words (project leader Matti Rissanen).
  • The Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME), about 1.8 million words (project leader Anthony Kroch).
  • The Corpus of Early English Correspondence Sampler (1417–1681) (CEECS), Helsinki, about 500,000 words (project leader Terttu Nevalainen).
  • The Parsed Corpus of Early English Correspondence (1410? –1681) (PCEEC), 2.2 million words (see CEECS above).
  • A Corpus of English Dialogues 15601760 (CED), Lancaster and Uppsala, about 1.2 million words (project leaders Merja Kytö and Jonathan Culpeper).

For further information on the Helsinki Corpus, see Kytö (1996) and Rissanen et al. (1993); on the CEEC corpora, Nevalainen & Raumolin-Brunberg (2003) and Nurmi (1999); on the CED corpus, Culpeper & Kytö (1997) and Kytö & Walker (2006).

Late Modern English (1700–1990)

  • A Representative Corpus of Historical English Registers (1650–1990) (ARCHER), University of Northern Arizona and University of Southern California, about 1.7 million words (Douglas Biber and Edward Finegan; later versions: project leader David Denison).
  • A Corpus of Late Modern English Texts (1710–1920) (CLMET), Leuven, about 10 million words (Hendrik De Smet).
  • The Penn Parsed Corpus of Modern British English (1700–1914) (PPCMBE), about 1 million words (see PPCEME above).

For further information on the ARCHER corpus, see Biber et al. (1994). For further information on the CLMET corpus, see De Smet (2005).

Late 20th century

  • The Brown Corpus (1960s) (Brown), 1,000,000 words (Henry Kucera and W. Nelson Francis)
  • The Lancaster-Oslo/Bergen Corpus (1960s) (LOB), 1,000,000 words (project leaders Geoffrey Leech and Stig Johansson)
  • The Freiburg-Brown Corpus (1990s) (Frown), 1,000,000 words (project leader Christian Mair)
  • The Freiburg-Lancaster-Oslo/Bergen Corpus (1990s)(F-LOB), 1,000,000 words (see Frown above).


Biber, Douglas, Edward Finegan & Dwight Atkinson. 1994. “ARCHER and its challenges: Compiling and exploring a representative corpus of historical English registers”. Creating and Using English Language Corpora. Papers from the Fourteenth International Conference on English Language Research on Computerized Corpora, Zürich 1993, ed. by Udo Fries, Gunnel Tottie & Peter Schneider, 1–13. Amsterdam and Atlanta, GA: Rodopi.

The Cambridge History of the English Language, I–IV. 1992–1999. Cambridge: Cambridge University Press.

Culpeper, Jonathan & Merja Kytö. 1997. “Towards a corpus of dialogues, 1550–1750”. Language in Time and Space. Studies in Honour of Wolfgang Viereck on the Occasion of His 60th Birthday, ed. by Heinrich Ramisch & Kenneth Wynne, 60–73. Stuttgart: Franz Steiner Verlag.

Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.

Kivimaa, Kirsti. 1966. Þe and Þat as Clause Connectives in Early Middle English with Especial Consideration of the Emergence of the Pleonastic Þat. (=Commentationes Humanarum Litterarum, 39/1.) Helsinki: Societas Scientiarum Fennica.

Kortmann, Bernd. 1997. Adverbial Subordination: A Typology and History of Adverbial Subordinators Based on European Languages. Berlin and New York: Mouton de Gruyter.

Kytö, Merja (comp.). 1996 (1991). Manual to the Diachronic Part of the Helsinki Corpus of English Texts: Coding Conventions and Lists of Source Texts. (3rd ed.). Helsinki: Department of English, University of Helsinki.

Kytö, Merja & Terry Walker. 2006. Guide to A Corpus of English Dialogues 1560–1760 (=Studia Anglistica Upsaliensia, 130.) Uppsala: Acta Universitatis Upsaliensis.

Lenker, Ursula. 2010. Argument and Rhetoric. Adverbial Connectors in the History of English. Berlin/New York: De Gruyter Mouton.

Meurman-Solin, Anneli & Arja Nurmi. 2004. “Circumstantial adverbials and stylistic literacy in the evolution of epistolary discourse”. Language Variation in Europe (=Papers from ICLaVE, 2), ed. by Britt-Louise Gunnarsson, Lena Bergström, Gerd Eklund, Staffan Fridell, Lise H. Hansen, Angela Karstadt, Bengt Nordberg, Eva Sundgren & Mats Thelander, 302–314. Uppsala: Universitetstryckeriet.

Meurman-Solin, Anneli & Päivi Pahta. 2006. “Circumstantial adverbials in discourse: a synchronic and a diachronic perspective”. The Changing Face of Corpus Linguistics, ed. by Antoinette Renouf & Andrew Kehoe, 117–141. Amsterdam and New York: Rodopi.

MED = Middle English Dictionary. 1956–2001. Ed. by Hans Kurath, Sherman M. Kuhn, Robert E. Lewis et al. Ann Arbor, Mich.: The University of Michigan Press.

Mitchell, Bruce. 1985. Old English Syntax, I–II. Oxford: Clarendon Press.

Molencki, Rafal. 2007. “The evolution of since in medieval English”. Connectives in the History of English (=Current Issues in Linguistic Theory, 283), ed. by Ursula Lenker & Anneli Meurman-Solin, 97–113. Amsterdam/Philadelphia: Benjamins.

Nevalainen, Terttu & Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. London: Longman.

Nurmi, Arja. 1999. “The Corpus of Early English Correspondence Sampler (CEECS)”. ICAME Journal 23: 53–64.

Rissanen, Matti. 1997. “Optional THAT with subordinators in Middle English”. Language History and Linguistic Modelling: A Festschrift for Jacek Fisiak on his 60th Birthday (=Trends in Linguistics, Studies and Monographs, 101), ed. by Raymond Hickey & Stanislaw Puppel, 373–383. Berlin: Mouton de Gruyter.

Rissanen, Matti. 2002. “‘Without except(ing) unless…’: On the grammaticalisation of expressions indicating exception in English”. Of Dyuersitie & Chaunge of Langage: Essays Presented to Manfred Görlach on the Occasion of his 65th Birthday, ed. by Katja Lentz & Ruth Möhlig , 77–87. Heidelberg: C. Winter Universitätsverlag.

Rissanen, Matti. 2007a. “From to till: Early loss of an adverbial subordinator”. Connectives in the History of English (=Current Issues in Linguistic Theory, 283), ed. by Ursula Lenker & Anneli Meurman-Solin, 61–75. Amsterdam/Philadelphia: Benjamins.

Rissanen, Matti. 2007b. “The development of adverbial subordinators in early English”. Change in Meaning and the Meaning of Change: Studies in Semantics and Grammar from Old to Present-Day English, ed. by Matti Rissanen, Leena Kahlas-Tarkka, Marianna Hintikka & Rod McConchie, 173–210. Helsinki : Société Néophilologique.

Rissanen, Matti. 2009. “Grammaticalisation, contact and adverbial connectives: the rise and decline of save”. English Philology and Corpus Studies: A Festschrift in Honour of Mitsunori Imai to Celebrate his Seventieth Birthday, ed. by Schinichiro Watanabe & Yukiteru Hosoya, 135–152. Tokyo: Shohakusha.

Rissanen, Matti. 2010. ”On the history of UNLESS”. Language Change and Variation from Old English to Late Modern English : A Festschrift for Minoji Akimoto (=Linguistic Insights, 114), ed. by John Scahill, Merja Kytö & Harumi Tanabe, 327-347. Bern: Peter Lang.

Rissanen, Matti. Forthcoming. “Corpora and the study of English historical syntax”. English Corpus Linguistics: Crossing Paths, ed. by Merja Kytö. Amsterdam: Rodopi.

Rissanen, Matti, Merja Kytö & Minna Palander-Collin, eds. 1993. Early English in the Computer Age: Explorations through the Helsinki Corpus. Berlin and New York: Mouton de Gruyter.

De Smet, Hendrik. 2005. “A Corpus of Late Modern English Texts”. ICAME Journal 29: 69–82.

Sorva, Elina. 2006. “The concessive connective albeit: A diachronic corpus-based study”. Types of Variation: Diachronic, Dialectal and Typological Interfaces (=Studies in Language Companion Series, 76), ed. by Terttu Nevalainen, Juhani Klemola & Mikko Laitinen, 121–148. Amsterdam: Benjamins.

Sorva, Elina. 2007. “Grammaticalization and syntactic polyfunctionality: The case of albeit”. Connectives in the History of English (Current Issues in Linguistic Theory, 283), ed. by Ursula Lenker & Anneli Meurman-Solin, 115–143. Amsterdam/Philadelphia: Benjamins.

Traugott, Elizabeth Closs & Richard B. Dasher. 2002. Regularity in Semantic Change. Cambridge: Cambridge University Press.