Negative concord and the loss of the negative particle ne in Late Middle English (1).

ABSTRACT

The presence of Negative Concord (NC) and the sentential negative particle ne is investigated in northern, southern, and mixed later Middle English prose texts from around 1400. The typology of negation proposed in Rowlett (1998) is taken as the basis for an examination of whether the loss of an overt Neg [degrees] head element is associated with the loss of NC. It is found that NC, though almost categorical in southern varieties, was showing signs of weakening in northern/northern-influenced texts. In these texts, the decline of NC was usually associated with the absence of ne. However, the converse relationship was not supported. It appears that loss of ne did not exert a direct influence on the grammar of NC in English, but that NC co-existed with the absence of a Neg head for a substantial period of time. This finding of a temporal disjunction is discussed in relation to the notion of cluster effects in parametrised syntax.

1. Introduction

Since Jespersen (1917), it has become customary in historical syntax to speak of a "negation cycle", in which sentential negators with relatively little phonetic substance are eroded and adverbial elements are pressed into service as negative elements, e.g. the process by which nawiht developed into naht and then not, while the Old English and Early Middle English sentence negator ne weakened and was lost. Jespersen believed that negative concord (henceforth NC) was related to the negation cycle--a language had NC if its principal sentence negator had relatively little phonetic substance, but lost it when it gained a principal negator with more phonetic substance. Early Middle English appears to fit Jespersen's characterisation well: it exhibited strict Negative Concord, that is, in a sentence expressing a single negation all non-assertive indefinites had to be marked negatively, e.g.:

1) be feond ne     mei   neden na mon to na     sune

   the devil NEG   may   compel no man to no    sin

   'The devil may not    compel anyone to any   sin'

   (AR 82b: 15; a. 1225).

The sentential negator ne was obligatory, at least in 12th and 13th century prose texts (Frisch 1997). It sufficed to negate a clause, although from the early 14th century onwards it tended increasingly to be accompanied by a secondary negator in some spelling form of not (Jack 1978; Frisch 1997; Iyeiri 2001).

Rowlett (1998: 87-89) captured the association between NC and Jespersen's "short" negator in terms of current syntactic theory. Building on the work of Haegeman (1995) and others, he analysed sentential negation in terms of an X' theory phrasal constituent NegP with a head and a specifier constituent (see also Beukema--Tomic 1996). He argued that a language has NC if and only if the sentential negator is the head of NegP, as in Serbo-Croat, where non-assertive indefinites must be accompanied by the head negator ni, e.g.

2) Milan   ne      daje     nitkome   nista

   M.      ne      gives    no-one    nothing

   'M.     isn't   giving   anyone    anything'

   (Progovac 1994).

Where a language has a negator in Spec NegP, e.g. Norwegian ikke ('not'), the language does not have NC, e.g.:

3) Jeg sa   ikke   noen (2)   gutt

   I saw    not    any        boy

   I didn't see    any        boy'

   (Johanneson 2003).

It is interesting from this perspective to consider the changes in English negation between, say, 1300-1600. The co-occurrence of multiple negated elements in a logically negative clause eventually went out of use in texts written by educated speakers (see Nevalainen 1998); polarity items such as any or ever became the norm within the scope of a formally marked negative expression. Well before this took place, however, the main sentence negator ne declined sharply in frequency and in the 15th century became extinct as an unsupported sentence negator (Frisch 1997). In the later Middle English counterparts of (1), therefore, one or more negated quantifiers could negate a clause in the absence of ne. This meant that in later Middle English (henceforth LME) Negative Concord could take the form of (i) the co-occurrence of ne and not, or (ii) the co-occurrence of not and a negative quantifier, and (iii) the co-occurrence of multiple negated quantifiers without a sentence negator. These three alternatives are exemplified by sentences (2), (3), and (4) respectively:

4) ... that he ne mowe nought selle his fish (Brembre II: 18; 1384).

5) the aduersairs of John Northampton should noght have be in non offices (Usk: 121; 1384).

6) ... that no man make none congregaciouns (Brembre I: 4; 1384).

Rowlett's (1998) analysis suggests that a transition was taking place from a language with a Head negator to a language with a Spec negator. If transitions in grammar change can be modelled as grammar competition (see Kroch 1989; Pintzuk 1996, etc.) we might expect to see evidence of competition between these two types. In the present case this would take the form of markedly higher incidence of NC in sources with retention of he, as compared with sources lacking ne, assuming NC and ne retention "cluster".

As regards whether the loss of ne was related to the demise in educated English of types (ii) and (ii) during the Early Modern English period, researchers seem to have reached somewhat differing conclusions. Jack (1978: 58-9) and Iyeiri (2001: 144) considered that multiple negation was not affected by the loss of he. Fischer (1992: 281-283) apparently took the view that the decline in the use of ne had a role in the weakening of negative concord, saying: "The disappearance of ne precipitates the erosion of multiple negation". (3) Frisch (1997: 33) suggested that "... the decline in use of ne is concurrent with a decline in the use of negative concord in general". The uncertainty displayed by previous researchers may be at least partly attributable to an insufficient analytic base. The use of the negative polarity item (NPI) any series in negative contexts in later Middle English has not yet been given sufficiently detailed consideration. An exception is Iyeiri (2002), who conducted a survey of NPIs in late ME verse and found a small but apparently non-negligible level of incidence in negated clauses. However, her results included many cases where the indefinite was in a dependent clause, e.g.

7) He was so full of veine gloire,
   That he ne hadde no memoire
   That ther was eny good bot he
   For pride of his prosperite

          (Conf. Amant.: 2799-2802, after Iyeiri 2002).

As shown by Ingham (2003), using the early 13th century prose works in the Penn-Helsinki parsed corpus of Middle English, indefinites in such constructions already routinely took the any-series form in EME, e.g.:

8) Nule        ich naut  pat  ani seo ow bute       leaue
   NEG-wish    I   not   that    ani see you except permission
   habbe       of  ouwer special meister
   should-have of  your  special master

   'I don't wish anyone to see you without permission of you special
   master'

  (CMANCRIW II: 47, 1. 438)

The innovation with which we are concerned here is the spread of NPI forms such as ani within clausebound contexts. Iyeiri (2002) reported a further finding, to the effect that the Northern text sampled, The Yorkplays, had the lowest incidence of NPIs in negated clauses. This would not support an association between head negation and NC, since sentential ne was apparently lost earliest in Northern texts. Unfortunately, it was not stated whether the finding obtained when only clausebound contexts were considered, so to that extent the issue must remain open.

Nevalainen (1998) and Kallel (2005) found that any series items were beginning to appear in later 15th century private correspondence. However, by the period covered by Nevalainen's and Kallel's studies the ne negator had long been defunct. We do not know how far the loss of the ne negator had an impact on NC prior to that.

In this study we seek to identify to what extent the loss of ne was associated with the loss of NC in educated written English. Fischer et al. (2000: 87) ob serve that the negative head ne was in decline in the later part of the 14th century. They go on to state that "the disappearance of ne precipitates the erosion of multiple negation" but they recognise that NC persisted well beyond the 14th century. Published work on the loss of ne and erosion of multiple negation in Late Middle English is not extensive. Jack (1978) found that, except in Chaucer's prose and London administrative texts, ne was uncommon in later Middle English prose. This automatically made the first type of NC obsolescent, but did not automatically target types (ii) and (iii), involving negative quantifiers, so we do not know from Jack's research how far they might have been affected by the loss of ne.

2. Rationale

In this study we investigate the earliest appearance in prose of NPIs in negated clauses. We analysed a number of Late Middle English texts covering especially the later 14th century, seeking to identify the co-occurrence or otherwise of a negated quantifier with other negated elements, either the sentence negator not or another negated XP, or both. Secondly, we sought to determine to what extent the survival of NC with a negated quantifier in these texts was related to a tendency to conserve the sentential negator ne.

We initially investigated late 14th century works from the Midlands or South of England. In so doing, we noted that some of these contained material appearing to have originated in the north of England, or to have been composed by an author who himself originated in the North. As will be seen, this appeared to have an affect the linguistic variables used. We pursued this broad diatopic approach by considering "pure" Northern LME prose texts, that is, works extant in northern manuscripts and composed by a northern writer. Since such works are unfortunately almost non-existent in the second half of the 14th century, we were obliged for this purpose to use texts whose manuscripts were in some cases written well into the 15th century.

In the main part of this study we report our analyses of these groups of texts, which aimed to discover to what extent they displayed a bias in their use of NC and in their retention of the sentential negator he. In the final section of the paper we return to the theoretical issues raised by Rowlett's (1998) model of negation and negative concord.

A terminological point should be noted: for the purposes of this study, we shall use the term "northern" English to refer to data sources located in Yorkshire and the North-East of England, and the term "southern and midland" English, when we are referring to those originating in the half of England located roughly south of the Trent, thus by and large conflating the traditional Southeastern, South-western, West Midlands and East Midlands dialect areas. This broad usage of "southern and midland" is simply a matter of convenience, to allow reference to the totality of English regions outside the north. It is in no way to imply that Midlands varieties were the same as the varieties used south of the Thames, any more than our use of the term "northern" is meant to imply that all varieties North of the Trent were identical. It is a way of distinguishing from the rest a northern area commonly supposed to have led the way in various aspects of linguistic change in the late medieval and early modern periods, and in which we believe innovation may once again have arisen.

3. Sources

In the previous section we recognised the problem that data sources for Middle English syntax of the 14th century are far from ideal, especially as concerns regional distribution. A large preponderance of prose texts come from the Midlands or the South and date from the last two decades of the century. Furthermore, a pervasive problem in medieval linguistic studies is to distinguish between date and locality of the original composition, and date and origin of the earliest surviving manuscript(s). Few medieval texts occur in a manuscript reliably representing the dialect area in which they were originally composed.

To establish a data sample representing "southern and midland" English varieties, we followed a policy of using for our analysis only editions of prose texts in manuscripts originally written in a southern and midland variety of English in the second half of the 14th century, and not thought to postdate the early part of the 15th century. In many cases this was reasonably straightforward, as with localisable administrative documents and Chaucer's works. It was felt desirable to include the late 14th century Vernon manuscript, which is a major source for the period, but some of its texts, though written by a southern scribe, were of northern origin. Only Vernon MS material not of northern origin according to Horstman (1895) was included in the "southern and midland" group, which was as follows:

                                        Estimated      Estimated
                                         date of        MS date
                                       composition

Chaucer: (4) Boece, Treatise on the
Astrolabe                             late 14th c.    early 15 c.

Trevisa: Polychronicon                late 14th c.       c.1400

Brut prose chronicle, continuation
1333-77                               late 14th c.       c.1400

"Appeal of Thomas Usk" and other
legal documents, London 1480s                            c.1385

London guild returns of 1389             c.1389           1389

Norfolk guild returns of 1389            c.1389           1389

The old usages of Wynchestre          later 14th c.   later 14th c.

Vernon MS tracts: Charter of the
Abbey of the Holy Ghost, A talkyng
of love                               later 14th c.      c.1390

Horstman's (1895) Vernon MS religious tracts originating in the North, i.e. Hilton's Scale of perfection, The Abbey of the Holy Ghost, and The mirour of St. Edmund, were put in a separate category, which we shall refer to as "mixed" texts. Also included here was the religious treatise De institutione reclusarum, which shows traces of Northern origin in its lexis. (5) Rolle's Meditations, found in a southem MS from around 1400, according to Horstman (MS Cambs L1 I. 8), was also counted in this group. We also wished to consider the extensive material attributed to Wyclif. Here the mixed source issue took a different form: Wyclif himself, though a native of Yorkshire, spent much of his working life in the south midlands and is thought to have been influenced to some degree by southern/midland usage. There is considerable uncertainty as to which works attributed to Wyclif are genuinely his. This applies in particular to the tracts of Wyclif edited by Matthew (1880); we therefore included only those where the editor stated his strong belief that the tract had been composed by Wyclif, who died in 1384, rather than being the work of his later followers. Data were taken only from tracts where the editor was able to identify some contemporary circumstances indicating a 14th century date, i.e. nos. I, IV, V, X, and XXI. We used the text of Wyclif's sermons edited by Hudson (1984), with the readings from manuscripts dated to the late 14th century (MSS K, O, T, Z, and [??]). We also used tracts no. 3 "Of holy chirche and of hir membres" from the edition by Lindberg (1993). Adding Wyclif to the list of "mixed" data sources gave us the following:

                                        Estimated date   Estimated
                                        of composition    MS date

Wyclif. tracts and sermons                 c. 1380       late C 14

Rolle: Meditations                         c. 1349        c. 1400

Vernon MS (texts of known or presumed
  northern origin)--Hilton, Epistle
  on the mixed life                        c. 1380        c. 1390

Abbey of the Holy Ghost                        ?          c. 1390
Mirour of St. Edmund                           ?          c. 1390
De institutione reclusarum                     ?          c. 1390

The purpose of the "mixed" group was that it allowed us to put in the "pure southern and midland" or "pure northern" groups only those texts which could securely be taken as representative of "northern" or "southern and midland" English, as characterised above, with no known disturbance in the textual tradition. But the mixed group is not simply being used here as a ragbag category: its texts have in common some affinity with northern English, so we can perhaps observe the effect that this factor might have, when they are compared with "pure" southern/midland texts.

The third category consisted of "pure" northern texts, corresponding to the "pure" southern and midland texts mentioned above. The earliest northern prose compositions available to us in a northern MS are the Epistles of Richard Rolle, (MS Cambs Dd V. 64), and the Prose rule of St. Benet, which have both been dated to around 1400. These two works were insufficient to form a sample comparable to the "southern and midland" texts. To supplement them, the northern prose treatises edited by Horstman (1895) from three northern MSS attributed to the period 1400-1450 (Harley, Rawlinson, and Thornton MSS) were included. Finally, we used a group of northern sermons thought to have been preached in York around 1415, found in a manuscript of the first half of the 15th century.

                                     Estimated date   Estimated MS
                                     of composition       date

Rolle, Epistles, MS Cambs Dd V. 64        350           c. 1400
N. Prose rule of St. Benet                 ?            c. 1400
Harley & Rawlinson treatises               ?            c. 1420
Thornton MS treatises                  later 14th       c. 1440
                                        century
MES MS Harley                           c. 1415        1420-1450?

As can be seen, the Northern prose sources are mainly from slightly later MSS than most of the other texts analysed, but this discrepancy was unavoidable given the material that has survived. Another unavoidable discrepancy was that the northern and mixed groups consisted exclusively of religious writings, whereas the southern and midland material represented a wider range of genres. It is not known whether this had an effect on the expression of negation in the data collected.

Altogether, these works present the opportunity to analyse a substantial body of prose data bearing on the status of negation in English around 1400. Nearly 170 contexts for NC or NPIs were identified, and some 750 negated clauses with or without ne.

4. Analysis of Negative Concord data

In our analysis of contexts for NC, we counted only finite clauses where a formally negated expression (either not or a negated indefinite) gave the sentence the interpretation of a negated proposition, rather than of phrasal negation. We analysed sentential negative contexts where Present Day English (PDE) would require an NPI, that is, where an indefinite XP was preceded by a formally negated element, either not or another negated XP. We did not count ne for this purpose. Examples of sentences included are given below:

9) ... that I ne con not thenken therof no bigynnying (A talkying of the loue of God: 366).

10) And shal neuer this vessel of no crafty man be made (Inst. Rec.: 46).

11) ... that no mannes cursynge hath (Wycliftracts: 289, 1).

Here the PDE NPI contexts (underlined) are provided by the italicised NPI triggers not, no, neuer and no respectively. As mentioned, contexts provided by the negator ne alone were not counted, e.g.

12) Common tretys of the Astrelabie ne maken noon excepcioun (Astrolabe II: 34, 1. 13).

13) ... pt essoyne of ober men ne be a-lowed in none manere (Wynchestre: 361).

Also excluded were instances where the target context stood in a co-ordinate phrase introduced by ne, even if a negative expression other than ne was present earlier in the clause, e.g.

14) He shuld not reste ne take no brebe (Brut continuation: 319, 1. 22).

In Early Modern English, negative co-ordination introduced by ne continued to license an N-word. It was therefore considered that this kind of context should be kept separate from our analysis (cf. Nevalainen 1998, who uses the same procedure). Finally we excluded cases where the trigger and the target context were in different clauses (Ingham 2003).

5. Results

The results of the foregoing analysis are shown in Table 1 :

Table 1. Frequencies of NC and NPIs in PDE
NPI contexts, Late Middle English prose texts

Northern prose                       NC          NPI       TOTAL

Rolle Epistles                       1            6          7
St Benet                             5            1          6
Four MES                             6            1          7
Prose treatises Rawl. MS             4            0          4
Prose treatises Harley MS            1            2          3
Prose treatises Thornt. MS          10            3         13
TOTAL                            27 (67.5%)   13 (32.5%)    40

"Southern and midland"'prose         NC          NPI       TOTAL

Vernon--Chart. Ab. Hol. Ghost        7            0          7
  --Talkynge of Holi Ghost           3            0          3
Chaucer--Boece                      26            0         26
  --Astrolabe                        1            1          2
Trevisa Polychronicon               11            0         11
"London English"                    12            1         13
Norfolk Guilds                       9            0          9
Ancient usages of Wyncestre          3            0          3
Brut continuation 1333-77           10            1         11
TOTAL                            82 (96.4%)    3 (3.6%)     85

"Mixed-prose                         NC          NPI       TOTAL

Wyclif sermons and tracts           12           12         24
Rolle Meditations                    4            0          4
De institutione reclusarum           5            2          7
Vernon MS Mirour of St Edmund.       1            1          2
  Hilton, Epistle                    2            1          3
  Abbey of the Holy Ghost            1            1          2
TOTAL                            25 (59.5%)   17 (40.5%)    42

The contrast between northern and "southern and midland" prose takes the form of a very low rate of NPI use in the latter, as compared with NPIs in nearly one-third of contexts in the former. The mixed text data actually show an even higher level of NPI use than the "pure" northern texts, but they are heavily skewed by the Wyclif data. Following are examples of NPIs in negated clauses:

15) God may not juge folily ony man (Wyclif Sermons: 237, 1.33).

16) And when a man wil no[??]t for any erthly thing wreth god (Rolle Epistles: 52, 1. 42).

17) Non innere mowe haue any entre to maken pi silence (Abbey: 324).

18) ... pat pu myth not haue set a nedyll poynte upon any place (Four MES: 140, 1. 295).

Yet all northern texts show some use of NC, and the overall rate of NC in pure northern texts, at nearly 70%, is so great that it must still be seen as very much the predominant option at this time in northern varieties.

It is interesting to consider to what extent NC was provided by co-occurrence of the target item with the sentence negator not, and to what extent by co-occurrence with another negated XP. The following table shows that in the pure Northern texts contexts were almost always provided by co-occurrence with not, while in the southern/midland texts and mixed texts such cases were in a clear minority.

Jack (1978) and Iyeiri (2001) gained the impression that not very rarely co-occurred with a negated indefinite, but this view is not upheld by these texts, which show as many as 62 out of 131 cases of NC (47.3%) co-occurring with not. Their judgment may derive from inspecting southern and midland or mixed texts, where NC most often arises between two negated quantifiers.

Let us now see whether there were any particular tendencies in the kinds of constituent that were affected by the weakening of NC in Northern and mixed texts. We investigated the use of NC or NPIs in different grammatical functions, notably Direct Object and Adverbial, the most common cases. The results were as follows:

Table 3. Frequencies of NC or NPIs in Direct
Object and Adverbial grammatical functions

                                     Direct Object   Adverbial

                                      NC      NPI     NC      NPI

"Pure" Northern texts
Epistles of Rolle                      0       0       1       5
Prose rule of St. Benet                1       0       4       1
Harl. & Rawl. treatises                1       0       3       0
Thornton MS treatises                  4       0       7       3
MES, MS Harley                         0       0       5       1

Total Northern texts                   6       0      20      10

Mixed texts

Wyclif tracts & sermons                1       7       7       8
Rolle: Meditations                     1       0       2       0
Vernon MS (known or presumed           0       0       2       1
northern origin)--Hilton, Scale of
perfection
Abbey of the Holy Ghost                2       2       1       0
Mirour of St. Edmund                   1       0       0       1
De institut. reclusarum                0       1       2       0

Total "mixed" texts                    5      10      14      10

Overall total                         11      10      34      20

The numbers are small, so interpretations must remain very tentative, but the "mixed" prose results suggest that in Northern-influenced writers NPIs were making headway in both grammatical functions. In the pure Northern prose itself, unfortunately, there is an imbalance between the numbers of object and adverbial contexts, with very few of the former, so the picture remains unclear. At least we can say that the grammatical function of the constituent see MS not to have imposed any strong restriction on whether NC or an NPI was used.

The "pure" southern and midland texts, as noted above, showed very little use of NPIs. Of the very few that occurred, Direct Objects as well as Adverbials were found with NPIs.

It was noted that NC in adverbials in "southern and midland" texts often took the form of the expressions in no manner, in/by no way, and in no wise. Of 44 adverbials showing NC, 16 (36.4%) had this wording. Such expressions may have been favoured for rhetorical purposes: they do not add to the representational content of the clause (Halliday 1994) but rather seem to convey the strength of the speaker's commitment to the proposition expressed. We wondered if Northern varieties tended to retain NC in such expressions, for special rhetorical reasons. However, in pure Northern texts the corresponding figures were only 4 out of 20, and in mixed texts 3 out of 13, suggesting that the persisting NC uses in Northern or Northern-influenced texts were not purely a matter of rhetorical choice, but continued to be generally a matter of grammatical concord, although this was now clearly an optional phenomenon in Northern varieties.

6. Analysis of sentential negator ne

We now turn to the question of how far a negator in [Neg.sup.0] was still present in LME prose towards the end of the 14th century. We took samples of the first 50 possible contexts for sentence negator ne, counting finite clauses where a formally negated expression (either not or a negated indefinite) gave the sentence the interpretation of a negated proposition, rather than of phrasal negation. Not all sources provided sufficient contexts: in particular, some shorter texts from the Thornton MS and the Vernon MS could not be included by this procedure. The percentages for the sixteen texts with at least 50 contexts are shown below:

Table 4. Percentage use of sentential negator ne in first 50 negated
finite clauses per text

Northern                  %

St. Benet                20

Rolle Epistles            2

MES MS Harley             0
2268

Tracts MS                 0
Rawlinson 285

Tracts MS                 0
Thornton

Av. Northern:
4.4%

Southern and              %
midland

Chaucer Boece            52

Trevisa                   0

Norfolk Guilds            8

"London English"         22

Brut Chronicle            4
1333-77

Av. Southern and
Midland: 17.2%

Mixed                     %

Wyclif                    0

Vernon--Hilton            4

St Edmund                22

Rolle, Meditiations       4

De institutione          26
reclu sarum

Av. Mixed: 10.8%

It can be seen that in every text ne-retention fell well below 30%, except in the work of Chaucer. The southern and midland texts had the greatest average rate of ne-retention, four times that found in the pure northern texts. The data from Wyclif and Hilton, both born in the North but whose texts have southern/midland influences, showed little or no use of ne.

Overall, these findings uphold the conclusion of Jack (1978) and Frisch (1997) that ne was generally obsolete by the latter part of the 14th century. However, in addition to the London English texts and Chaucer, noted by Jack, ne-retention also occurs at a non-negligible rate in certain other texts: St. Benet, the Vernon Mirour of St Edmund and De institutione reclusarum. These texts were either pure northern or with northern influences. Nevertheless, it is undoubtedly in the southern and midland texts that ne-retention survived longest. The relationship between ne-retention and the form of negative indefinites in these 15 texts is summarised in Figure 1:

[FIGURE 1 OMITTED]

As noted above, there was some indication of a geographical trend as regards ne-retention: the rate of ne use is lowest in northern and highest in southern and midland texts.

7. Discussion

NC between not and a negated indefinite was found in both northern and southern/midland texts. However, in northern English co-occurrence of negated indefinites was rare, whereas in southern and midland varieties it was very common, e.g.:

19) ... pat neuere non such was sene in no mannes tyme alyue (Brut: 316, 1.20).

20) ... pat no wicked planet haue noon aspect (Astrolabe II: 4, 1.36).

We cannot say that in northern varieties not was becoming an obligatory element in negated clauses, since sentence negation with a single negative indefinite was also found commonly in northern texts, e.g.:

21) ... pat he kan fynd na ioy in pis life (Rolle, Form: 43, 1.34).

22) Scho sulde no more seke hym (Bonaventura, Privity of the Passion: 215, 1. 36).

Our main finding was that NC, though still virtually undisturbed in southern and midland texts, was yielding ground to NPIs in the northern texts.

The mixed texts, sharing a northern factor either in the origin of the texts or of their author, also showed substantial use of NPIs. The eighty or so datapoints obtained from these two groups of texts, though relatively small in terMS of the numbers often found in diachronic grammatical investigations, nevertheless form quite a representative sample, taking taken into account the infrequency of the construction: indefinites in a clause already containing at least one semantically interpretable negator.

It is true that the slightly later dating of the Northern MSS leaves open the logical possibility that if we had late 14th century prose MSS from the North they would show virtually uniform NC as the southern and midland MSS do. Against that, however, is the fact that the mixed group, which actually show proportionately even more NPI use, come from late 14th century MSS, and the factor they share is the presence of northern influence. We consider, therefore, that the earliest substantial use of NPIs in English prose is to be associated with northern varieties. The linkage of ne-loss and the rise of NPIs suggested by certain earlier researchers, such as Fischer, and expected on theoretical grounds following Rowlett (1998), thus receives some support. Although the demise of ne was not far from completion in both North and South, it see MS to have been most marked in the North. The five sources in which it was totally absent all had northern associations, except for Trevisa, suggesting that the Jespersen cycle completed a new stage in northern varieties before southern and midland. We may then speculate that in the process took place, NPIs entered use earlier in northern varieties. This can be no more than a tentative proposal, given the limited sample size we have been able to use for the present investigation.

An alternative possibility that might be envisaged is that NPI use in negated clauses arose under the influence of Latin, in which NPIs were used rather than NC in equivalent contexts. Both Wyclif and Rolle are known to have also composed works in Latin, and their works account for the bulk of NPI use in the data analysed in this study. However, Trevisa, who translated a large amount of material from Latin, seemingly underwent no such influence, so we believe that the argument for Latin influence in the avoidance of NC is less than compelling. (6)

It was noted in an earlier section that the southern and midland texts were stylistically much more diverse than the other two groups, featuring historical works (Brut, Polychronicon), philosophical literature (Bocce), administrative documents ("London English", "Norfolk guilds") as well as a modest amount of religious prose. It is certainly conceivable that the formality of the administrative documents from London and Norfolk, as well as the literary aspirations of Chaucer's Bocce, may have encouraged the retention of NC as a conservative form. Perhaps the influence of French, in which NC was the rule, also exerted itself. It could be that for stylistic reasons NPIs are under-represented in the southern and midland data we have. Vernacular usage at this time cannot, of course, be directly accessed via any of the religious, historical, literary, and administrative written works analysed in this study. The question is whether the religious prose genre, to which the northern prose available from this period largely belongs, was relatively closer to that usage than others. We do not at present know of any reason to suppose that it was, and wish to leave the issue open. On the basis of the data we have, and for all their limitations, it can be said that the loss of ne and the loss of NC do not seem to have been contemporary. If there was a connection between them, these grammatical changes appear to have taken place in succession, rather than simultaneously.

This outcome illustrates an important issue in the relationship between theoretical linguistic enquiry and grammatical change as attested in the material surviving from earlier periods of the language. It is well-known that the observable effects of the loss of V [right arrow] Infl in Early Modern English, for example, took some time to work through in different constructions (e.g. Ellegard 1953; Kroch 1989). To expect that theoretically related characteristics of the language should all undergo change in lockstep, as it were, when an underlying parameter changes, would surely be unrealistic. Other relevant factors, such as lexical choice, high frequency of certain recurrent patterns, stylistic preferences in certain registers, and so on, may well intervene to affect the picture. A routine but crucial caveat is that in examining the textual record for evidence of grammar change we are seeing only indirect effects of changes in individuals' linguistic competence. Clustering effects in language change, then, may well not be synchronically clearly observable.

The above considerations raise numerous issues to do with the relationship between historical linguistic data and the constructs posited in diachronic syntax research to which we cannot do justice in the space available here. Suffice it to say that this study has attempted to present one way in which the problem can be addressed: a theoretically grounded claim regarding an association of two grammatical traits can be seen as making a prediction of the form that variation would take for the claim to be false: if the data revealed a tendency for NC to weaken in a language variety conserving a head negator. The late 14th century Middle English prose data show us that, on the contrary, where NC weakens in it is in areas where ne tended to be lost first. This outcome leaves open the problem, to which we hope to return in a subsequent study, of the nature of grammars acquired in the transitional period between the loss of ne, and the disappearance of NC.

8. Conclusion

In this investigation we wished to examine how far use of NPIs may have made progress in the prose texts available from the later 14th century, and in particular to look for evidence of regional tendencies. We have also sought to place the development of NPIs in relation to the loss of the ne sentential negator. It is apparent from the results obtained here that in Southern/Midland texts NPIs were still almost entirely absent, whereas in some texts with Northern associations NPIs were by no means rare. These concerned most notably the Rolle epistles and Wyclif's prose, though NPIs in negated clauses examples crop up in most of the northern and mixed texts investigated. The Southern and Midland texts, as expected from the work of previous researchers, showed a higher level of retaining ne as a sentential negator. These results, then, go in the direction predicted by a theoretical link between head negation and NC, insofar as the texts associated with Northern Middle English, where head negation was lost earlier, are those where NPIs developed earlier.

It remains to be acknowledged that the surviving data we have is far from ideal from the point of view of attempting any precise kind of historical socio-linguistic analysis, being often difficult to localise precisely. However, as a broad generalisation it seems reasonable to believe that the loss of the ne sentential negator may indeed have had repercussions on the expression of negation in English, in terms of favouring NPIs rather than NC, at least in the educated register of English transmitted to us by the textual record.

REFERENCES

PRIMARY SOURCES

AR see Tolkien (ed.)

Astrolabe see Benson (ed.)

Ayto, John--Alexandra Barratt (eds.) 1984 Aelfred of Rievaulx's De institutione reclusarum. Oxford: Oxford University Press.

Basington, C. (ed.) 1865 Polychronicon Ranulphi Higden monachi cestrensis, together with English translations of John Trevisa. Vols. 1-3. London: Longman.

Benson, Larry D. 1988 The Riverside Chaucer. (3rd edition.) Oxford University Press.

Boece see Benson (ed.)

Brie, Friedrich (ed.) 1906 The Brut or the chronicles of England. London: K. Paul, Trench, Trubner & Co. Brut continuation see Brie (ed.)

De institutione reclusarum see Ayto--Barratt (eds.)

Four MES see O'Mara (ed.)

Harley & Rawlinson treatises see Horstmann (ed.) Hilton, Epistle on the mixed life see Horstmann (ed.)

Horstmann, Carl (ed.) 1895 Yorkshire writers. Vol. 1. London: Swan & Sonnenschein & Co.

1896 Yorkshire writers. Vol. 2. London: Swan & Sonnenschein & Co.

Hudson, Anne (ed.) 1983 English wycliffite sermons. Vol. 1. Oxford: Oxford University Press.

Kock, Ernst (ed.) 1902 Three Middle English versions of the Rule of St. Benet. London: K. Paul, Trench, Trubner & Co.

Lindberg, Conrad (ed.) 1991 English Wycliftracts. Vol. 1. Oslo: Novus.

2001 English Wyclif tracts. Vol. 2. Oslo: Novus.

O'Mara, Veronica (ed.) 2002 Middle English sermons. (Edited from Harley MS 2268.) Heidelberg: Universitatsverlag C. Winter.

Prose rule of St. Benet see Kock (ed.)

Rolle see Horstmann (ed.)

Smith, Joshua (ed.) 1870 English guilds. London: Truebner.

A talkynge of the luue of God see Horstman (ed.) Thornton MS treatises see Horstmann (ed.)

Tolkien, J. R. R. (ed.) 1962 The English text of the Ancrene Riwle: Ancrene wisse. London: Oxford University Press.

Trevisa see Basington (ed.)

Usk see Smith (ed.)

Vernon MS tracts see Horstmann (ed.)

Wyclifsermons see Hudson (ed.)

Wycliftracts see Matthew (ed.)

Wynchestre see Smith (ed.)

SECONDARY SOURCES

Blake, Norman (ed.) 1992 Cambridge history of the English language. Vol. 2: 1066-111476. Cambridge: Cambridge University Press.

Beukema, Frits--Olga Tomic 1996 "Negation in English: A diachronic view", Folia Linguistica Historica XVI/1-2: 123-135.

Ellegard, Alvar 1953 The auxiliary do: The establishment and regulation of its use in English. Stockholm: Almkvist & Wiksell.

Fischer, Olga 1992 "Syntax", in Norman Blake (ed.), 207-408.

Fischer, Olga--Ans van Kemenade--Willem Koopman--Wim van der Wurff 2000 The syntax of early English. Cambridge: Cambridge University Press.

Fisiak, Jacek--Marcin Krygier (eds.) 1998 Advances in English historical linguistics. (Trends in Linguistics: Studies & Monographs 112.) Berlin: Mouton.

Frisch, Stefan 1997 "The change in negation in Middle English: A NegP licensing account", Lingua 101: 21-64.

Haegeman, Liliane 1995 The syntax of negation. Cambridge: Cambridge University Press.

Halliday, Michael 1994 An introduction to Functional Grammar. (2nd edition.) London: Edward Arnold.

Ingham, Richard 2000 "Negation and OV order in Late Middle English", Journal of Linguistics 36/1:13-38.

2003 "Negative concord in Middle English: A canonical agreement relation". [Paper presented at Linguistics Association of Great Britain conference, Oxford University, September 2003].

Iyeiri, Yoko 2001 Negative constructions in Middle English. Fukuoka, Japan: Kyushu University Press.

2002 "The development of non-assertive any in Later Middle English and the decline of multiple negation", in: Yoko Iyeiri--Margaret Connolly (eds.), 127-43.

Iyeiri, Yoko--Margaret Connolly (eds.) 2002 And gladly wolde he lerne and gladly teche: Essays on Medieval English presented to Professor Matsuji Tajima on his sixtieth birthday. Tokyo: Kaibunsha.

Jack, George 1978 "Negation in later Middle English prose", Archivum Linguisticum 9: 58-72.

Jespersen, Otto 1917 Negation in English and other languages. Copenhagen: Hoest & Son.

Johanneson, Janne 2003 "Negative polarity verbs in Norwegian", Working Papers in Scandinavian Syntax 71: 33-73.

Kallel, A. 2005 The loss of negative concord in standard English. [Unpublished Ph.D. dissertation, University of Reading].

Kroch, Anthony 1989 "Reflexes of grammar in patterns of language change", Language variation and change 1: 199-244.

Nevalainen, Terttu 1998 "Social mobility and the decline of multiple negation in Early Modern English:, in: Jacek Fisiak--Marcin Krygier (eds.), 263-291.

Pintzuk, Susan 1996 "Old English Verb-complement word order and the change from OV to VO", York papers in linguistics 17: 241-264.

Pintzuk, Susan--George Tsoulas--Anthony Warner (eds.) 2000 Diachronic syntax. Oxford: Oxford University Press.

Progovac, Ljiljana 1994 Negative and positive polarity: A binding approach. Cambridge: Cambridge University Press.

Rowlett, Paul 1998 Sentential negation in French. Oxford: Oxford University Press.

van Kemenade, Ans 2000 "Jespersen's cycle revisited: Formal properties of grammaticalisation: in Susan Pintzuk--George Tsoulas--Anthony Warner (eds.), 51-74

RICHARD INGHAM

UCE Birmingham

(1) This paper is a revised version of a study bearing the same name originally appearing in Reading University Working Papers in Linguistics Vol. 7 (2003)

(2) Noen is synchronically non-negative in form, and functions as a polarity item equivalent to PDE any.

(3) The same point is expressed using the same wording in Fischer et al. (2000: 87).

(4) In order to avoid biasing the data too much towards Chaucer's usage, we used only the two non-fictional texts, Boece and a Treatise on the Astrolabe. The Riverside Chaucer 3rd edition, based on the early 15th century Cambs. Univ. MS I. i. 38 for the former text, was used.

(5) See Introduction to the EETS edition, p. li.

(6) Although Chaucer's Boece is nominally a translation of a Latin original, he appears to have made considerable use of the Old French translation of Boethius by Jean de Meun, entitled li livres de confort.

Table 2. Negated indefinites co-occuring
with not or with another negated XP

Co-occurrence with:              Not       Negated XP   TOTAL NC

"Pure" Northern prose             24            3          27
"Pure" Southern and Midland       27           53          80
prose
"Mixed" texts                     11           13          24
TOTAL                         62 (47.3%)   69 (52.7%)     131