Slipping in Some Medieval Welsh Texts : A Preliminary Survey

Nous avons isolé et examiné des occurrences de slipping dans un corpus de textes médiévaux gallois de différents genres – narratif, religieux et savant – aussi bien dans le texte original qu'en traduction. Le slipping se révèle être un phénomène rare, que l'on trouve dans chaque genre examiné sans toutefois qu'il apparaisse dans chaque instance. Il s'agit d'une catégorie prototypique contenant un certain nombre de caractéristiques clés, qui, cependant, ne doivent pas être toutes présentes pour qu'il y ait passage du discours indirect au discours direct. La plupart des cas de slipping dans ce corpus peuvent être analysés en termes fonctionnels comme étant des cas de « slipping par effet de style » (artful slipping), c'est-à-dire qu'ils servent à mettre en évidence l'importance d'un énoncé ; quant aux autres cas, qualifiés de « slipping par inadvertance » (inadvertent slipping), ils sont généralement involontaires et proviennent d'une confusion du rédacteur ou du scribe. Toutefois, il existe un certain nombre de slippings qui ne sont ni tout à fait volontaires ni tout à fait involontaires, apparaissant dans des contextes qui semblent susciter des citations indirectes, par exemple des cas présentant un verbe introductif (verbum dicendi), et à partir desquels un rédacteur peut glisser vers une citation directe.

The corpus ofmedieval Welsh texts consists ofboth prose and poetry ; the prose may be assigned, by modern standards, to the genres of secular and religious narrative, as weIl as of historiographical, legal, and leamed prose. The earliest ex tant Welsh manuscripts date to the mid-thirteenth century.
The existence in medieval Welsh narrative of the phenomenon here termed « slipping » has already been noted before. Morgan Watkin (YBH, pp. clxxii-clxxiii) drew attention to one instance in Ys torya Bown de Ha mtwn, the Welsh version of the Anglo-Norman Geste de Boeve de Ha umtone, of what he called mixing of direct and indirect quotation -there are others (see below). However, his particular example « 19) below) presents textual and other problems, and its discussion will therefore be postponed until the general properties of slipping in medieval Welsh texts have been more clearly defined. Brynley F. Roberts (1983, p. 172) mentioned the narrator's « ability to switch effortlessly from oralio recta to obliqua and back again » in the tale conventionally called Owein, because of its complex relation ta Chrétien's Yvain (1) Ac y dy wawt Owein gwbyl 0' e gerdet idaw -« ac yn ymgeissaw a 'r marchawc yssyd yn gwarchadw y ffynnawnn y myn nwn ry mot ».  And Owein told him the whole of his quest -« and contending with the knight who is guarding the fo untain 1 would wish to be ». (2) . A datkanu eu kyfranc a orugant idaw mal y datkanassei y uorwyn y nos gynt -«ac Owein a pallwys idi, ac am hynny y llosgwn ninneu hi ».  And they told him their story as the girl had told it the night before « and Owein failed her, and therefore we will bum her ».
He placed the origin of this narrative technique in the oral tradition of Welsh story telling and also commented briefly on its stylistic and narrative function : Nor is this [i.e., the narrator' s ab ility to switch from oratio obliqua to recta] a mere fo rmaI or acquired skill, fo r his ear unerringly tells him at what point the change needs to be made, precisely on the significant words which bear most emotional stress. (Roberts 1983, p. 172) ln his view then, these instances would be examples of « artful slipping », to adopt a term from Richman (1986, p. 283), that is « a technique fo r controlling and adjusting emphasis and meaning ». Roberts' observations were fo llowed up by Sioned Davies (1995) in her study of the stylistic properties of the body of medieval Welsh native narrative, the so-called Mabinogion-corpus, with reference to Deborah Tannen's definition, «an indirect quotation fa des into a direct one » (Davies 1995, p. 240). Davies (ibid.) also noted one further fo rmaI characteristic of this device, namely the absence of a tag or speech marker to identify the speaker, although there are sorne significant exceptions in my corpus (see below), and she also gives useful statisticaI information about its frequency in the texts of the Mabinogion-corpusl. Slipping in this body of texts is very infrequent, given the high incidence of direct quotations here -unfortunately Davies does not provide any statistics about the incidence of indirect quotation2. The incidence of direct quotation can at least be indirectly, and somewhat approximately, derived from her statistics of the use of different tags in individual texts, because more than one tag may be used to mark one instance of 1 It must be noted that her statistical findings do not necessarily agree with mine, see below.

2
A survey of the distribution of indirect discourse in Middle Welsh prose and of its formaI properties is needed ; fo r a useful description of the syntax of the noun clause in Modem Welsh see Thome (1993, pp. 373-82).
direct quotation, but the fo llowing numbers of instances will at least give a general impression of the numerical relation between slipping and direct quotation3 : typical or preferred direction of the shift as weIl as the relevant linguistic unit in which it happens, the quotation of a single speaker's speech. In order to identify unambiguously such a shift from indirect to direct discourse, various fo rmaI properties are indicative, which would seem to include a switch of personal reference and a break in the construction resulting in relative syntactic independence of the direct quotation -the latter may be indicated by a switch of mood and/or tense categories, the lack of a subordinating conjunction, or main clause word-order. However, Roberts' two examples quoted ab ove already show one significant deviation from this tentative first definition of core properties of slipping : They contain no real indirect quotation, but the verhs of speaking (dywawt « [ Owein] told », datkanu [ ... ] a orugant « [they] told ») are fo llowed by a direct object (gwbyl o'e gerdet «the whole of his quest ») or its semantic equivalent (eu kyfr ane «t heir story ») which denotes the abstract content of the discourse, rather thati reproducing it5. 1 therefore prefer to take the fo llowing example, also from Owein, as my starting point fo r a description and classification of slipping in the Mabinogion-corpus : (3) A pheri dyfynnu y holl gyuoeth y un lle drannoeth a orue y iarlles, a menegi udunt uot y hiarllaeth yn wedu, ae na elUt y ehynnal onyt 0 uareh ae arueu a milwryaeth -« ac ys ef y rodaf inneu ar awch dewis chwi, ae un ohonawch chwi a'm kymero i, ae vyg k annyadu ynneu y gymrut gwr a'e kanhalyo 0 le arall ».  And the countess had her whole realm mustered in one place the next moming and told them that her earldom was without protection and that it could not be defended except by horse and arms and strength « and thus 1 give it in your choice, either one of you takes me or you allow me to take a man from another place who may defend it ».
The bipartite indirect quotation is dependent on a verb of saying (menegi) and introduced by markers of syntactic subordination, the lenited verbal noun of « to be », uot, and the suhordinating negation na respectively. Pronominal reference to the speaker is realized as third person singular in the possessive pronoun y «her ». The switch to direct quotation is marked by changed pronominal references, tirst to the speaker with the finite verb rodaf « 1 give » in the first person singular, then to the addressees with a possessive pronoun awch «your » in the second person plural, and concomitantly by a syntactic break, the beginning of a new clause, which is effected by the absence of a subordinator and overt main clause word order in the direct quotation with initial adverbial (y)sef« thus, now »6. Furthermore, there is no tag to identify the speaker, as Davies noted, nor an initial address or interjection to mark the beginning of direct quotation, another frequent tagging device. Semantic cohesion 5 Compare similarly Ger. 917-9, 428-9. However, under a somewhat wider concept of the reproduction of speech than applied here these ex amples may become relevant.
between the indirect and the direct quotation is provided by the main clause connective arc) «a nd ». Modem editors conventionally seem to consider this connective to be part of the direct quotation and position quotation marks accordingly, but it would be equally possible to take it as the end of the indirect quotation. It should also be noted that direct quotation in medieval Welsh narrative is otherwise only rarely introduced by a connective arc) «a nd », with the exception of strongly reactive, non-initial utterances in conversations between two, or more, speakers -fo r example : (4) « Ef a geif hynny yn llawen », heb y gwr 0 'r got. «A minheu a 'e kymmerafyn llawen », heb y Pwyll.  « He will get this gladly », said the man from the bag. «And 1 will accept it gladly », said Pwyll.
For the further analysis of the material 1 will therefore consider the presence of the fo llowing fe atures as an indication of slipping : a semantically general verb of saying is fo llowed by both an indirect quotation and a direct quotation, and the switch from the one to the other is marked by a fo rmaI break at clause level, typically effected by changing the syntactic status from subordinate to main clause, with concomitant changes of the verbal fo rms, as weIl as changing the pronominal reference ; this syntactic break is bridged by semantic cohesion between the indirect and the direct quotations which is indicated by the presence of a connective and the lack of a specific tagging device such as the identification of the speaker or an interjection or address. The narrative function of slipping may be tentatively described as highlighting a contextually central new information or, fo llowing Roberts (1983, p. 172), « the significant words which bear most emotional stress ». An instructive instance in which the semantic cohesion necessary fo r slipping between the indirect quotation and the fo llowing direct quotation is absent, is (5).
Here the indirect quotation reproduces information supplied by an unspecified collective, whereas the direct quotation reproduces the specific utterance of one member of this collective and is not linked by a connective to the indirect quotation, but marked as a separate entity by an interjection, an address, and an identificatory tag instead : E chwedyl a doeth at Ua tholwch. Sef ual y do eth, dy wedut anfuruaw y ueirych ac eu llygru, hyt nat oed un mwynyant a ellit 0 honunt. «J e, Arglwyd », heb un, «d y waradwydaw yr a wnaethpwyt, a hynny a uynhir y wneuthur a thi ». (PKM32.7-l 1) The story came to Matholwch. Thus it came, he was told of the maiming of his horses and of their disfigurement until there was no profit that could be obtained from them. « WeIl, lord », one said, « an insult has been inflicted on you and this was intended to be done to you ».
Six further examples from my Mabinogion-corpus confonn to the general pattern of slipping as defined here7• No obvious norm can be observed with regard to the relative length of the indirect and direct quotations respectively ; in BR 8.20-27 the direct quotation is much shorter than the indirect one, in Ow. 788-792 the opposite is the case. 1 do not consider as examples of slipping instances in which a phrase naming a specific speech act such as erchi trugaret «to ask quarter » or cyfarch well «to greet » is fo llowed by a direct quotation which reproduces either this speech act or considerations arising from it8, and 1 also exc1ude instances in which the contents of the direct quotation is characterised by a preceding qualifying expression, such as geireu die anhygar «wrathful ugly words » (Per. 57.8-9), which is the nominal object of a verb of saying.
In example (6) a connective between the indirect and the direct quotation is lacking, and there is no switch in pronominal reference which would mark the direct quotation uncontroversially as such ; the context and the importance of the information imparted, however, make its interpretation as an instance of slipping into direct quotation highly likely, rather than as the narrator's description.

(6)
Yn teu a dy wawt nas atwaenat -« gwr oed a chwnsallt coch ar uchafy arueu, a tharyan velen ar y yscwyd ». (Per. 65.5-8) He said that he did not know him -« he was a man with a red surcoat over his armour and a yellow shield on his shoulder ». Example (7) appears to fulfill the general conditions fo r slipping. Here the indirect quotation is introduced by the subordinating conjunction ual « as » and is furthermore preceded by a short direct quotation announcing the speaker' s intention to supply 7 These are PKM  Compare A bwrw y gledyf 0 ' e law a oruc, ac erchi trugaret y Ereint -« a rOlryr, » heb ef, « y gadawd uyg camryuic a 'm balchder ym erchi nawd, ac ony chaf yspeit y ymwneuthur a Dyw am uym pechawt ac ymdidan ac offeireit, ny hanlryfwell 0 nawd. » (Ger. sorne requested infonnation9• Slipping from indirect to direct quotation would seem to highlight the central part of this infonnation. (7) « Llyma ol! », heb ynteu, a menegi ual yr lygryssit ac y diuwynyssit y grofdeu idaw, ac ual y doethant y llygo t idaw y'r grofd diwethaf yn y wyd -« ac un ohonunt oed amdrom, ac a deleis, ac yssyd yn y uanec, ac a grogaf inheu auory. [ .. .] ». (P KM 60.24-61.1) «H ere is aIl », he said, and told how his crofts had been laid waste and destroyed fo r him and how the mice had come to him to the last croft in his presence -«and one of them was very heavy, which 1 caught and which is in the glove and which 1 will hang tomorrow.
[ ... ] 1 0 » Example (8) is only relevant if the direct quotation ' can be assigned to the rider, not to Idawc ; unfortunately, the context does not provide any certain cIues. Here a semantically more specifie verb of saying is used, namely gofyn « to ask [something of somebody] », and such verbs of asking appear ta have a preference for a fo llowing indirect, rather than a direct, question 11 • The direct quotation is not linked ta the preceding context by a connective, nor is it marked by a specifie tagging device.
(8) Ac yna nachaf y marchawc yn eu gordiwes, ac yn gofyn y Idawc a gaffe i ran 0 'r dynyon bychein hynny gantaw -« y ran a weda ymi y rodi mi a'e rodaf; bot yn gedymdeith udunt ual y bum ynneu ». A hynny a oruc y marchawc a myn et ymeith.  And then, 10, the rider overtook them and asked Idawc if he could obtain a share of the little fe llows from him -« the share it is proper fo r me to give, 1 will give, to be a cornrade to them as 1 myself have been ». And this the rider did and went away.
There are two similar examples from Gereint12, both with the verb erchi « to ask [sornebody ta do something] », bath with a connective, but also with a tag within the direct quotation to identify the speaker : 9 There is a slightly different pattern in which a direct quotation announcing sorne information is fo Howed by the auctorial statement that it is given, without quoting it, compare PKM 56.25-26 : « Llyma, » heb ynteu, « uyg kyfr anc » -a 'e datkanu ol!. «Here, » he said, « my story » -and he told it aH. Similarly PKM 7.25-27, 8.1 1-13, 23 .7-8, Ger. 250-253, 416-419, and Pero 11.10-

13.
1 0 For the use of ac a as a relative particle compare Evans (1964, p. 63, note 3).
I l Compare the relative infrequency of the use of such verbs to introduce direct quotation in the statistics provided by Davies (1995, pp. 199-227  Ac erchi a oruc ynteu y Enyt yscynnu ar y march a cherdet 0 'r blaen a chymryd ragor mawr -«ac yr a welych ac yr a glywych », heb el, « arnaf i, nac ymhoyl di dra c heuyno Ac ony dy wedaf i vrthyt li, na dy wet ti un geir heuyt »0 (Gero 744-747) And he asked Enyt to mount her horse and to go ahead and to keep a great distance -« and in spite of what you see and what you hear », he said, « conceming me, do not tum backo And unless 1 talk to you, do not say one word either »o The direct quotation reproduces the central part of the speaker's order, and the presence of the tag is the only untypical fe ature hereo Example (10) is more unusual in that slipping occurs within a direct quotation in which its speaker first reproduces, in a very brief indirect quotation dependent again on erch i, what another speaker will say, and then slips into direct quotation of this speaker's expected words ; these give the central information, this speaker's reasoningo The connective used is causal canys « since », which also occurs occasionally (see below, example (16» 0 The indirect quotation consists of an infinite verbal-noun substitute fo r a verbal clause, which is one fo rmat fo r positive indirect quotation in Middle Welsh, with just one adjunct denoting the patient of the verbal action, kymryt y llamysten « to take the sparrow hawk » : ( 10) « [0 0.] canys yna y dy t Ma rchawc y Llamysten gostec, nyt amgen, erchi y'r wreic uwyhaf a gar kymryt y llamysten -« canys goreu y gveda i li, a th i a y keueist », med el, «yr llyned ac yr d"Hry. Ac or byd a y gwarauunho it hediw 0 gedernit, mi ay hamdiffy nnaf it »0 Ac am hynny », heb y gwr gwynllwyt, «y may reit y titheu uot yno pan uo dyt [. . .] ». (Ger. 267-272) «["0] since the Knight of the Sparrowhawk will then make an announcement, namely, he will ask the lady he loves most to take the sparrowhawk -« since it suits you best, and you had it », he will say, «since a year ago and since twOo And if there is somebody who grudges it you today by fo rce, 1 will defend it fo r yoU ». And therefore », said the grey-haired man, « it is necessary fo r you to be there when it will be day [" 0] »0 Although slipping here occurs within a direct quotation, this instance agrees with aIl the fo rmaI and functional characteristics established above, with the exception of the presence of a tag, but within the Mabinogion-corpus only Gereint shows this unusual preference fo r a tag in the context of slippingo Another semantically specifie verb of saying, namely gorchymyn «t o order », is used in (11) ; the indirect quotation is again fai rly short, as it is in (10) and in (15) below. The direct quotation is marked by an identificatory tag. Semantic cohesion is established by the coordinating conjunction arc) «and ». The direct quotation repro duces the speaker' s central statement.
(1 1) Ac yskynnu ar y uarch e hun a Drue a gorchymyn y'r uornyn gyrru y meirch -« ac nyt gwell im », heb ef, « dyw edut vrthyt ti no thewi, cany bydy vrth uyg kyghor ». (Ger. 823-826) And he mounted his own horse and ordered the maiden to drive the horses -« and it is no better fo r me », he said, « to talk to you than to be silent, because you will not submit to my decision ».
As already mentioned above, semantically more specific verbs of saying, such as erchi and gorchymyn may have a tendency to prefer an immediately fo llowing dependent clause, that is, indirect quotation, rather than introduce a direct reproduction of the speech act itself ; it is therefore possible that the use of such verbs creates a context conducive to slipping, in that they fo rmally invite an indirect quotation, but also allow subsequent slipping into direct quotation fo r narrative effect. This may have methodological repercussion for the descriptive categories used, particularly for the polarity of the notions of « artful » and « inadvertent ». My final example from Gereint (12) shows some rather curious and complex vacillation between indirect and direct quotation, and also the insertion of a narrator' s explanation13 : (12) Sef a oruc Gwalchmei, gyrru hwnnw y uenegi y Arthur uot Gereint yno yn uriwwedic, ac na deuei y ymwelet ac ef, ac vd oed druan edrych ar yr ansawd J! § §J!f!. arnaw (a hynny heb vybot y Ereint ac yn hustinc y rythaw a 'r maccwyj). «Ac arch y Arthur », heb ef, « nessau y bebyll ar y fo rd, cany daw ef y ymwelet 0 'e uod ac ef, ac nat hawd y diryaw ynteu yn yr agwed y mae ». (Ger. 1 179-1 185) This Gwalchmei did, he sent him to tell Arthur that Gereint was there wounded and that he would not come to see him and that it were sad to see the condition in which he is (and this without Gereint's knowing and by whispering between him and the squire). «And ask Arthur », he said, «t o move his tent to the road, because he will not come to see him of his own will, and that it is not easy to fo rce him in the condition he is ».
As Thomson (Ger., pp. 128-9) notes, uot (equating to «t hat he was ») and na deuei «that he would not come » signal indirect quotation dependent on y uenegi «(in order) to tell », as does the subordinating negation nat «t hat not » (aIl set in bold), whereas the imperative arch «ask » and the present tense fo rms yssyd, daw, and y 13 Because of the complexities of the passage and the problem to define the position of slipping, 1 reproduce Thomson' s punctuation.
mae (aIl underlined) signal direct quotation. Thomson suggests that yd oed (bold and underlined) is also part of the indirect quotation, probably because of its tense, imperfect indicative, but one might have expected uot instead to introduce indirect quotation here. Were it not fo r the subordinating negation nat l 4, slipping could begin somewhere before yssyd, perhaps between yssyd and yd oed, even though there is no clause boundary here at which slipping typically occurs, or alternatively yd oed could mark the beginning of direct quotation, with reversion to main clause word order and some modal sense of oed. The grammatical mixture in this passage remains odd 15 , and it is interesting to note that the medieval manuscripts of Gereint show no variation in this matter. Example (13) combines a number of deviant fe atures, but at least the position of slipping is cIear : the introductory verb of saying menegi « to say » is fo llowed first by a direct obj ect denoting the abstract content of the discourse (as in examples (1) and (2) above) and then by an indirect quotation which is introduced by the subordinating conjunction megys « how » (compare example (7) above) ; the direct quotation is specifically marked by an address, Arglwydes « lady », and an identificatory tag, heb y Te irnon « Teirnon said » -but semantic cohesion between it and the indirect quotation is established by the connective ac « and ». The direct quotation reproduces contextually important information and is quite long in comparison with the preceding indirect one.
(13) Sefymdidan a uu gan Te irnon, menegi y hoU gyfranc am y gassec ac am y mab, a megys y buassei y mab ar y hardelw wy, Te irnon a'e wreic, ac y megyssynt -«ac wely dy yna dy uab, Arglwydes », heb y Te irnon. «A phwy byn nac a dy wot geu arnat, cam a wnaeth. [ ... } »  This was Teirnon's discourse, he told the full story about the mare and the boy, and how the boy had been in their charge, Teirnon's and his wife's, and how they had brought him up -« and see here your son, lady », Teirnon said, «whoever told lies about you, did wrong [ ... ] ».
On the basis of this admittedly small number of examples from the Mabinogion corpus of medieval Welsh prose it was possible to deduce a preliminary set of formai fe atures which characterise the majority of instances of slipping in it -a sequence of indirect and direct quotation, in this order, marked by a fo rmai break, but Iinked by semantic, and also fo rmai, cohesion between them, which makes the use of special tagging devices fo r the direct quotation unnecessary. There is also a small number of deviant examples which diverge from this pattern with regard to one or more 14 One would expect nyt. Watkin (YBH, pp. c1xxiii, 147) notes one example from Ys torya Bown where one manuscript has nat and the other nyt and where he would expect main clause negation nyt, but here scribal confusion is the likely explanation. fe atures, but which can still be considered to be more marginal instances of slipping. From a functional perspective aIl my examples here seem to carry a specifie narrative load in that the direct quotation reproduces verbatim the most important part of a speaker's utterance. It must also be stressed that in view of the importance and frequency of direct quotation and dialogue in this corpus of texts, slipping remains a relatively unusual stylistic fe ature.
The occurrence, or non-occurrence, in translations of particular narrative devices could be highly instructive since, in theory at least, a comparison between a translation and its source should reveal its redactor's innovations as weIl as retentions and thus allow insights into the availability of such devices in his literary tradition. However, only too rarely can his immediate source be uncontroversially defined, aithough an extant text may be arguably quite close to it. Furthermore, in the specific case of slipping it wouid be imperative to identify any specific instance as functionally meaningfui and (predominantly) « artful », rather than as « inadvertent ». With aIl these reservations in mind, 1 will fo cus in my discussion of Ys torya Bown de Ha mtwn first on a comparison between some instances of slipping in the Weish text with the corresponding passages in the ex tant texts of the Anglo-Norman Geste de Boeve de Ha umtonewhich, however, were most likely not the immediate source(s) fo r the Weish text, aithough they are probably quite close to it.
My first example (14) shows the typicai pattern, with a general verb of saying as introduction, a fo rmaI break between indirect and direct quotation with concomitant semantic cohesion, and a narrative emphasis on the information contained in the direct quotation which serves as a kind of punch-li ne. It corresponds to direct quotation in the extant text of the Geste1 6 , and could therefore be taken, with aIl due methodologicai caution, as an instance of the Weish redactor's artful slipping.
(14) Ar hynt y duhunawd ef y gedymdeithon ac erchi udun wisgaw eu harfe u ar ffr wst a menegi udun ry daruot y 'r palmer y rodyssynt idaw y swper y nos gynt mynet a Io sian yn lIathrut -«a bei gwypei Juor hynny, ys drwc a wyr eu dihenyd vydem ni ». (YBH 1760-1 768) At once he woke up his companions and ordered them to arm themselves in haste and toid them that it happened that the pilgrim to whom they had given his supper the night before, had gone away secretly with Iosian � « and if Iuor knew tbis, we would be men with a bad death ».
Example (15) is slightly more problematic : the bishop's speech is reproduced as direct quotation in the extant text of the Gestel?, and if the Welsh redactor had a direct quotation in his source as weIl, he then started ta convert it into indirect 1935-1950. quotation and then slipped into direct quotation in the course of this admittedly very long passage. This occurs after a connector a «a nd » by a change of word order, but since the bishop's narration is still -about a third party, no switch of pronominal reference is required until the end of the passage when the bishop directly addresses Bown. It could perhaps be argued that the length of the passage invited the redactor to slip into direct quotation, at an appropriate point of narrative importance, and if this is indeed the case, then this would be an instructive instance of artful slipping motivated by a specifie context, that is neither completely « artful » nor completely « inadvertent ».
(15) Ac yna ymdidan a wnaeth yr esgob a'e nei a dy wedut y vot ef yn varchawc dewr pan enillei ef y ryw was hwnnw, a menegi itJ.aw ry dy uot y varwgywedyl ef a' e dienydu yn waradwydus 0 'r sarassinieit at Sebaot, y datmayth a'e athro, y gan vab Sabot a ry fuassei ymplith sarassinieit yn y geissaw ynteu -«ac 0 achos hynny y dechreuis Sebaot ryuelu ar yr amherawdyr, ac eissoes ny allawd ymerbynneit ar amherawdyr, ac wrth hynny y gorfu arnaw adaw y wlat a'e gyfoeth a mynet y casteil cadarn a wnathoed y my wn ynys yn y mor, ac ny cheU y casteil hwnnw yn dragywydawl hyt Ira barhaei fwyt yndaw, can ny ellit ymlad ac ef 0 vn ffo rd. Ac o'r casteil hwnnw y dygei Sebaot a' e allu kyrcheu y gyfoeth yr amherawdyr gweith hyt dy d, gweitheu hyt nos, ac y welly yd oedynt yn diffeUhaw kyfoeth yr amherawdyr yn fe nedic-wychyr ; ac 0 gwney vyggyghor, fi a ey attaw 0 ' e gymorth, a minheu a rodafyt yn nerth pump cant marchawc yn gyweir ». (YBH 2172(YBH -2202 And then the bishop talked with bis nephew and told him that he was a brave knight since he won this seNant, and told him that the news of his death and his scandalous execution at the hands of the Saracens had come to Sabaot, his fo ster-father and his teacher, from Sabaot's son who had been in the lands of the Saracens to look fo r him -« and fo r this reason Sabaot began to make war on the emperor, but he could not stand up to the emperor, and fo r this reason he was fo rced to leave his country and bis realm and to go to a strong castle that he had built on an island in the sea, and this castle could never be conquered as long as there was fo od in it, since one could not fight with him on the same road. And from this castle Sabaot and his troops would make inroads into the emperor's realm, sometimes at day-time, sometimes at night, and thus they kept harassing the emperor's realm boldly and fe arlessly ; and if you fo llow my counsel, you go to him to help him, and 1 will give you as reinforcements five-hundred well-supplied knights ». And the boys ran to their fa thers to take off their arms and asked their fa ther for garments and arms and asked him to dub them knights -« since we are big and strong ». As mentioned above in the context of (8), verbs such as gouyn and erchi seem to prefer a fo llowing indirect quotation, although there are counter-examples both in the Mabinogion-corpus and in Ys torya Bown20• A redactor may therefore be tempted by his use of such verbs to begin an indirect quotation, but to slip artfully into direct discourse soon afterwards at the moment ofhighest narrative emphasis.
There are two potentially significant examples of differences between the two medieval manuscripts of Ys torya Bown, the White Book of Rhydderch (W) and the Red Book of Hergest (R) respectively, with regard to the employment of indirect or direct quotation and to the position of slipping. It should be noted that in general the texts of Ys torya Bown in these two manuscripts tend to agree quite c1osely. In (17.1) the Red Book text has a typical and straightforward example of slipping from an indirect quotation, with a verb in a third person singular (imperfect) (parei) referring to the speaker, to a direct quotation, with the verb in the first person singular (present) referring to the same speaker (paraj). The beginning of the direct quotation is furthermore marked by the connective a «and », and the importance of its infor mation is signalled by the interjection lly ma. In the White Book (17.2), however, indirect quotation continues slightly further with the verb digawn «he is able » in the third person singular present indicative, semantically corresponding to first person singular present para! « 1 will bring about » in the Red Book, and the use of the present tense is grammatically and contextually odd. The extant Anglo-Nonnan Geste has direct quotation, see BdH 1541-1545. (17.1) Sef a wnaeth Bonffei yna, dy uot attei a' e didanu a dy wedut wrthi y parei ef idi gael diane y nos honno -« a llyma y ffu ryf y paraf: Mi a atwen lyssewyn ac a'e kaffaff yn y weirglawd obry ». (R) Bonfei did this then, he came to her and comforted her and said to her that he would make it possible fo r her to escape that night -« and this the fo rm in which 1 will bring it about : 1 know a herb and will get it in the meadow down below ».
(17.2) Sef a wnaeth Bonfe i yna, dy uot attei a'e ddidanu a dy wedut wrthi y ·parj21 ef idi diagk y nos honno, a llyma y ffurufy digawn -«mi a atwen lys ewyn ae a'e eaffaf yn y weirglod obry ». (W = YBH 1688-1693) Bonfei did this then, he came to her and comforted her and said to her that he would make it possible fo r her to escape that night, and this the fo rm in which he can do it -« 1 know a herb and will get it in the meadow down below ».
ln the second example only the Red Book (l8.l) switches to direct quotation in a context which makes this instance different from slipping proper. This direct quotation reproduces Bown's words22 which a messenger is told to relate, and is dependent on the verbal noun dy wet « to say », which is itself part of a long indirect order syntactically dependent on the verbal noun erchi «to ask [somebody to do something] ». A coordinating connective therefore cannot be used ; note the tag heb ef «he said ». The White Book employs indirect quotation throughout and, accordingly, third person singular fo rms referring to Bown as the speaker (18.2)23, but switches ta grammatically and contextually odd present tense fo rms in the part which is direct quotation in the Red Book -again the complete passage is rendered as direct quatation in the extant Anglo-Norman Geste (18.3).
( 18.1) A dy dgweith y bore y gelwis Bown ar un o'r marchogyon, a Charuus oed y enw, ac un 0 'r gwyr glewaf a dewraf oed el, ac erchi idaw mynet hyt yn Hamtwn at yr amherawdyr, a dy wedut idaw y mae Bown oed enw y marchawc a ry uuassei yn ymdidan ac ef y nos arall ae a 'e twyllwys ac yn ynghwanec dy wet idaw : «y paraI y grogi neu y dihenydyaw [oJ dihenyd ny bo gwell idaw no hwnnw, kan ys oetran gwr yssyd arnafi », heb el, «a mi a allaI gwisgaw arueu a 21 Probably a mistake fo r 3rd sg. imperfect parei, which is the fo rm in R. And one moming Bown called on one of the knights, and Carfus was his name, and he was one of the bravest and most valiant men, and asked him to go to Hamtwn to the emperor and to tell him that Bown was the name of the knight who had been talking with him the other night and who deceived him, and to tell him furthermore : « 1 will cause him to be hanged or executed with an execution that may not be better fo r him than that, because 1 have the age of a man », he said, « and 1 am able to carry arms and to ride, and a large number ofbrave, valiant, ugly, and strong knights are together with me ».
(18.2) A dy dgweith y bore y gelwis Bown ar vn o'r marchogyon, a Charfu s oed y enw, ac vn 0 'r gMyr glewaf a dewraf oed ef, ac erchi idaw mynet hyt yn Ha mtwn at yr amherawdyr, a dy wedut idaw y mae Bown oed enw y marchawc a ry fu assei yn ymddidan ac ef y nos arall ac ea24 twyllwys, ac yn yghwanec dy wet idaw y peiranna y grogi neu y dienydyaw [oJ25 dihenyd ny bo gwell idaw no hwnnw, canys oetran gwr vssyd arnaw ac y digawn wisgaw arueu a m arch ogaeth, ac amylder 0 varchogyon gleu-dewr anhygar-gryf yssvd gyt ac ef (W = YB H 2535-2552) And one moming Bown called on one of the knights, and Carfus was his name, and he was one of the bravest and most valiant men, and asked him to go to Hamtwn to the emperor and to tell him that Bown was the name of the knight who had been talking with him the other night and who deceived him, and to tell him furthermore that he will cause him to be hanged or executed with an execution that may not be better fo r him than that, because he is the age of a man and he is able to carry arms and to ride, and a large number of brave, valiant, ugly, and strong knights are together with him. One moming Boves got up and called a messenger, and he came ; he was not a boy, but a strong knight. The story says his name was Karfu. « Brother », said Boves, « may God send you health, go to Hampton, do not delay. Tell the emperor, when you have seen him, that the knight who was there a short time ago, is called Boves, if he has been deceived about tbis ; tell him that 1 let him know that he will hang, because 1 am strengthened with a coat of mail and a shield, and with good knights who are strong and powerful, and I have a giant who has much strength ; but say this presumptuously, Jesus may help you ! » 1 am currently uncertain about possible implications of these two examples fo r the relationship of the -Welsh texts of Ys torya Bown ; could they be an indication that the scribe of the White Book clumsily transformed into indirect quotation a passage in his source with indirect quotation slipping into direct quotation, as it is preserved in the Red Book, or did Hywel Vychan, the scribe of the Red Book, intelligently improve on the text either of the White Book or of the common source which both he and the White Book used ? Watkin's example of a mixing of direct and indirect discourse in Ys torya Bown mentioned above is (19), and here again the Red Book appears to preserve a some what better reading than the White Book : (19) A phan daruu udunt y dala, Brice 0 Vr isteu, Glois 0 Gaerloyw a Clarice 0 Leicetre oed truan ganthunt ac a vynnynt y diane, ac y dy wedassant wrth y brenhin y mae gogan mawr a gwatwar [R oedJ idaw am y gwr a welsant -« yn gwasanaethu gyr dy vron ac 0 'th fuol dy uot a myn et, ac wrth hynny nyt iawn it peri y lad. A pha ny bei daet y march a'e glotuorusset, ni a vynem y diuetha ». (YBH 301 1 -3022) And when they had taken him, Brice of Bristeu, Glois of Caerloyw, and Clarice of Leicetre were sad and wanted to let bim escape, and they said to the king that it is a great disgrace and mockery [R which it was] fo r him about the man they had seen -« serving before you and coming and going with your cup, and therefore it is not right fo r you to have him killed. And if the horse were not as good as its renown, we would wish its destruction ».
According to Watkin (YBH, p. cIxxii), the use of the (third person singular) present y mae of bot « to be » and of the possessive pronouns second person singular dy and lth are characteristics of direct quotation, whereas the preposition idaw «f or him » with a suffixed pronoun third person singular and the third person plural subj ect of the verb welsant «they saw/had seen » with reference to the speakers point towards indirect quotation. However, y mae may simply have a special grammaticalised function here, namely to introduce an emphasised constituent in an object clause. At least this is the interpretation implied by the reading of the Red Book with additional oed « (which) it was »26, and the White Book's scribe may have overlooked an oed in his source. Watkin is undecided whether the mixing of direct and indirect quotation already was a fe ature of the Anglo-Norman source the Welsh redactor used27, or whether it arose, perhaps by mistake, within the Welsh transmission on the basis of a direct quotation in the original Welsh translation28. But even in the version of the Red Book, there remains a problem with the position where slipping sets in, namely within a subordinate relative clause rather than at a clause boundary, as has been the case in aIl examples discussed so fa r. There is, however, one other example in Ys torya Bown in which an indirect quotation appears to slip immediately after the dependent verbal noun mynet « to go » into direct quotation, indicated by the present tense of the verb elwir « is caIled » in a relative clause -altematively direct quotation could begin with this relative clause, but the first option seems to be semanticaIly better, in that the complete directions are reproduced as direct quotation29 : (20) Yn a y menegis y!ford ac yd erchis idaw mynet -« trM)' y dinas a elwir Nubie, ac odyno y Gartage ac odyno ti a wely Mwmwrawnt. (YBH 1475-1 479) Then he told him the way and told him to go -«through the city which is caIled NubIe, and from there to Cartage and from there you will see Mwmbrawnt ».
Sorne potentially interesting, but still only tentative observations emerge from this discussion of slipping in Ys torya Bown : most instances of slipping in the Welsh text fo llow the typical pattern identified in the Mabinogion-corpus. They have no counterpart in the extant text of the Anglo-Norman Geste, and the majority appear to have been brought in «artfully » by the redactor to enhance the vividness of his narrative -it must remain uncertain wh ether he was motivated by models in his own oral or/and literary narrative tradition, or by fo reign models, or by a mixture of aIl. « inadvertent » as discrete descriptive contrasts is too simplistic, because slipping may be invited by certain contexts and still function meaningfully within themprobable fa ctors are the length of a quotation or its introduction by a semantically specific verb of saying (see also above, examples (8) to (11» . Richman ( 1986, p. 290), in his discussion of slipping in Old English, notes another and more general di ffi cult y, namely the problem of interpretation : «Several instances of slipping in Old English seem to me problematic -that is, probably inadvertent but arguably artful ». Finally, differences between the two medieval manuscripts of Ys torya Bown with regard to the position of slipping point towards the possibility of scribal fluidity in the use of slipping in the manuscript transmission of a text. Two examples of slipping in Kedymdeilhyas Amlyn conform to the regular pattern, KAA 29-32 and 502-504 ; the Latin Vi la closest to the Welsh text has no obvious parallel fo r the first example and indirect quotation fo r the second. The third example (2 1) lacks a connective between the indirect and the direct quotation depending on erehi, and there is furthermore an address inserted in the direct quotation. 1 still consider this to be a fo rmally less typical instance of slipping in which this narrative device is employed fo r narrative emphasis on the central message. The Latin text used fo r comparison has a very short direct quotation here.
(2 1) Ac erehi a wnaeth udunt, yr karyat y Gwr a diodefassei angheu dros bopyl Adaf, gwneuthur vn 0 deu peth : ae ymlad yn wyehyr drwy vot yn barawt y odef angheu yr keissyaw budugolyaeth, ae na delynt yn nes no hynny y'r vrnydyr, ony bei gymeint eu karyat ar Duw ae eu bot yn barawt y ode! angheu drostaw, or bei reit udunt -«medylyaw, arglwydi vrodyr, a dy lyyeh, pwy bynnae a odefo angheu yn y vrnydyr honn, y byd yn Uewenyd terynas nef kynn oeri y waet ». (KAA 732-739) And he asked them, fo r the love of the Man who suffered death fo r Adam's seed, to do one of two things, either to fight boldly by being prepared to suffe r death by attempting victory, or not to go nearer to the battle than this, unless their love fo r God was so strong that they were prepared to suffer death fo r him, if it were necessary -«y ou should consider, brothers and lords, whoever will suffer death in this battle will be in the joy of heaven before his blood is cold »30.  (22) is fo nnally even more marginal because of the short narrative clause gymeint a cheissyaw y dwyllaw « and sought his hands» inserted between the indirect quotation and the direct quotation3 1, and the marking of the direct quotation by an address and an identificatory tag. 1 am therefore inclined not to consider (22) as an example of slipping.
(22) A ryuedu yn uawr a orue y palmer, a goryn idaw paham y gwattwarei ef was Duw yn gymeint a cheissyaw y dwyllaw. « Yn herwydd itt », heb el, « arglwyd, goryn y mi y bore hediw yr hynn yd wy t yn y oryn yr awr honn. (23.1) Sef a oruc [CharlymaenJ edrych yn llitdyawc ... arnaw, a dy wedut y vot yn ynryt dibwyll, a bot yn amlwc bot drycyspryt yn arglwyd arnaw -«pwy », heb el, « a vyd am d iffy nnwr ar y blaeneit os Rolond a dric yn geitwat ar yr rei 01 ? » (W = y dCM 134.12-18) Charlemagne looked at bim angrily and said that he was fo olish and senseless and that it was evident that an evil spirit was his lord -«w ho », he said, «will be protector of the van if Roland will stay as protector of the rearguard ? » (23 .2) «0 deuot dy n ynuyt y dy wedy di », eb el, «ac amlwc yw ar dy ymadrawd bot kythreulyaeth ynot. Pwy a vyd keitwat », eb ynteu, « ar y blaenieit ot adawn i Rolant yn geitwat ar yr 01 ? » (Peniarth 10 = CR L.5-8) « y ou speak like a mad man », he said, « and it is evident from your speech that there is devilry in you. Who will be protector », he said, «of the van if I leave Rolant as protector of the rearguard ?» (24. 1) Ac yna yd aeth Rolond y ymdidan a Ch arlymaen, ac y adolwyn idaw dan anhyed ystynnu idaw y breint rydaroed y We nwlyd y varnu idaw, tr»}' y bwa a oed yn y law, 0 gwelit idaw y vot yn wiw y'r gwassannaeth hwnnw -«a minheu a adawaf yti », heb el, «yn gadarn, na dygwyd y bwa 0 'm llaw i yn y gymryt yr llyfyrder, mal y dygwydwys y llythyr 0 law We nlwyd ». (W = y dCM 135.10-17) And then Roland came to talk with Charlemagne and to ask him with coaxing to invest bim with the privilege that Gwenlwyd had happened to adjudge to him, with the bow that was in his hand, if it seemed to him that he was fit fo r this service -« and I promise you finnly », he said, « that the bow will not fa H from my hand from taking it timidly, as the letter fe ll from Gwenlwyd's hand ».
(24.2) Ac yna y dy uot Rolant wrth Ch iarlymaen : « Y brenin da, kywaethoc, bonhedic », eb el, « anrydeda vi 0 'r deilygdawt a varnawd Gwenlwyd ym, 0 gwelir yt vy mot yn deilwg ohonaw. Ac estyn ym », eb el, «a'r bwa yssyd y 'th law ; a mineu a adawaf yti yn gadarn na digwyd y bwa o 'm llaw ual y digwydawd y llythyr 0 law We nlwyd 0 gymraw ». (Peniarth 10 = CR LII. 1-5) And then Rolant said to Chiarlymaen : «Good, mighty, noble king », he said, «present me with the honour that Gwenlyed adjudged to me, if it seems to you that I am worthy of it. And invest me », he said, «with the bow which is in your hand ; and I promise you finnly that the bow will not fa ll from my hand as the letter fe ll from Gwenlwyd's hand out of fe ar ».

139
1 will now turn to examples of slipping in a diffe rent textual genre, although still in a translation of a fo reign-language source, namely in the apocryphal Efengyl Nicodemus, the Welsh version of Evangelium Nicodemi, in its oldest manuscript text from the White Book of Rhydderch (mid-fourteenth century). This version parallels closely the Latin text of Evangelium Nicodemi in Einsiedeln, Stiftsbibliothek MS 326, and therefore provides another opp ortu nit y to compare, at least tentatively, source and translation. As it turns out, the Latin text has direct quotation in all instances, and the Welsh version transforms their respective first parts into indirect quotation and then slips into direct quotation, on the words which Roberts probably would consider to bear «most emotional stress » (see above, Roberts 1983, p. 2). There are five examples which conform to the typical pattern of slipping33. In one of these, the direct quotation is introq.uced by the exclamative llyman «behold », corresponding to Latin ecce ; in another, the direct quotation is marked by an address, also occurring in the Latin text, and an identificatory tag.
Passage (25.1) presents two problems, the delimination of the exact place of slipping and the use of a non-finite verbal fo rm in the direct quotation (but compare examples (30) and (3 1) below) : (25.1) Ac odyna galv attav y gynnulleitua a dy vedut vrthunt bot eu deuavt yn dy d gwylua -«gellvg yvch vn o'r carcharoryonn. Y mae gennyf», heb el, «ygkarchar llourud kelein, Barrabas y eno, ac ny welafi ry haedu dim 0 fessu. A pha vn onadunt MY a vynnyvch chui y ellvg ? » (EN261.17-20) And then he called to him the assembly and said to them that it was their custom on a fe ast-day -« to release fo r you one of the prisoners. 1 have », he said, «in prison a murderer, Barabas his name, and 1 do not see that Jesus has merited anything. And which of them do you want to have released ?» with reference to the addressees, yvch « fo r you », corresponding to Latin uobis. This yvch might therefore conceivably be a translator's slip motivated by uobis. The word immediately preceding yvch is a non-finite verbal fo rm, the verbal noun gellvg « release » ; and although verbal nouns are often used fo r finite fo rms in main clauses in narrative contexts, their use in direct quotation would appear to be unusual. Slipping typically occurs at sentence-or c1ause-boundaries, and there is no such boundary within the Welsh copular clause bot eu deuavt yn dyd gwylua gellvg yvch vn o'r carcharoryonn « [that] it was their custom on a fe ast-day to release fo r you one of the prisoners ». However, a clause-boundary occurs in the Latin text after per diem azimorum, corresponding to Welsh yn dy d gwylua. The second part of the direct quotation, which contains an identificatory tag, clearly carries the communicative and emotional stress. In (26.1) the Welsh version slips into direct quotation after quoting directly Jesus's words within the indirect quotation ; the Latin version (26.2) uses direct quotation dependent on dixerunt quia throughout. According to Richman (1986, p. 280), this alternative use of dicere quia to introduce direct quotation caused problems fo r Latin scribes and Anglo-Saxon translators and sometimes led the latter to tum direct quotation introduced by quia into indirect quotation. Richman (1986) quotes the same passage (26.3) from the Dld English Gospel of Nicodemus as an example of the translator's « trouble with the dicere quia construction » and notes that he « begins with indirect discourse but soon slips to direct » -slipping sets in earlier than in the Welsh text, already before the quotation of Jesus's word, although Richman also points out that the shift in the references of the personal pronouns, from third person before the inserted quotation to first person after it, is the clearest indication of slipping34• The Welsh redactor's decision to begin with indirect quotation may have been motivated by similar problems with the dicere quia construction (see also below), and it is likely that slipping into direct quotation after the inserted quotation of Jesus's words is motivated by the Latin model and the influence of uidimus, corresponding to guelsam, rather than by rhetorical or pragmatic intentions.
(26.1) E tri hynny a doethant 0 Alilea y Gaerusalem ac a dy vedassant y tywyssogyon yr offeireit ac y'r a oed yg kynulleittua y sinagoga 0 fdeon ry welet fessu onadunt yr hvnn a grogessynt hvy ygyt ac vn ar dec 0 ' e disgyblonn yn ymdidan ac wynt yn eu kymerued ymynyd Oliuet ac yn dy vedut vrthunt, « Euch ar hyt y byt oll a phregethuch y baup, a betydyvch paub o'r kenedloed yn enw y tat, a'r mab, a 'r yspryt glann, amen ,' a phwy bynnac a gretto ac a betydyer, hvnnv a vyd iach », -« a guedy dy wedut hynny vrth y disbyblonn y guelsam ef ynn ysgynnv y nef». (EN264.6-12) These three came from Galilee to Jerusalem and told the high priests and those of the Jews who were in the assembly of the synagogue, that they had seen Jesus, whom they had crucified, together with eleven of his disciples, talking to them in their midst on the Mount of Olives and saying to them, «Go into the world and preach to aIl, and baptize aIl the peoples in the name of the fa ther and the son and the holy ghost, amen ; and whoever will believe and will be baptised, he will be saved », -«and having said that to his disciples we saw him ascending to heaven ». 3) hig soedon to pam ealdrum. J to pam moessepreostum. J to ealre poere gesommunge. poer hig to poera iudea sy noDe comon. P! oet] hig pone onhangena hoelaend gesawon. J »yD hys endlufo n leorning cnyhtas sp oec ! ... }.
They said to the eIders and the high priests and to aIl the assembly where they had come to that synod of Jews that they had seen the hanged saviour «and he spoke with his eleven disciples [ ... ] ». (Richman 1986, p. 281) 141 The Welsh translator seems to have had sorne problems with dicere quia ; in another passage (27.1) he translated dixit quia as heb ef « he said » as an introduction of direct quotation, and although heb as a preposed tag is rather rare, it does occur (compare (27.3» . The tag interrupts the typical sequence of indirect-to-direct quotation, and 1 am currently uncertain about the place of this unusual example within a typology of slipping in Welsh (compare (22) above). In the final example dicere quia again introduces direct quotation . in the Latin version ; in the Welsh version the subordination panyv which is used to introduce the clefted element in an object clause35, signaIs that indirect quotation is intended. It could either begin before the relative clause defining the clefted element Jessu Grist vab Duv (28.1.1.) or after it, that is, before the relative clause of the cleft construction (28.1 .2). The first option somehow seems more naturaL Examples (26. 1) and (28.1) are quite similar to the instances of slipping after the dicere quia construction discussed by Richman (1986, p. 281), in that the Welsh translators were influenced by quia to begin an indirect quotation, but then motivated by a trigger in the Latin text, uidimus and me respectively, to slip into direct quotation, and they would therefore appear to be more inadvertent than artfu}36 . However, the majority of the examples of slipping in the Welsh Efe ngyl Nicodemus probably result from the translator's/redactor's conscious or semi-conscious input and are therefore more artful than inadvertent, and this seems to apply to the majority of the other examples of slipping in translated texts considered so far.
1 will now tum to still another textual genre of translation, historiographical texts, and specifically to one of the Welsh versions of Geoffrey of Monmouth's Historia Regum Britanniae, the so-called Cotton Cleopatra version of Brut y Brenhinedd. It is available in a reliable modem edition, but its special interest is that in contrast to the Dingestow version, which is also available in a modem edition and which has been described as a fluent and free translation of the Latin source (BB-LI, p. xxix), it shows a still freer handling, more in line with native narrative con ventions : The Cotton Cleopatra version is yet another translation but it is briefer than the Historia. It fo llows the Latin in the order of events and in language but the translator summ arizes long speeches and cuts out many unnecessary descriptive passages. This version, however, contains much additional material and shows more attempt at harmonization with other sources than any other group of texts. (BB -LI, p. xxx ) The narrative part here [in the fust section of the Cotton Cleopatra version] is mostly in simple straightforward Welsh with a good many stereotyped expressions into which the translator falls whenever conditions pennit. The latter part of the narrative seems to be written in a somewhat more omate style ; the stereotyped expressions are still used but the translator is consciously striving fo r rhetorical effect. One mark of this is the more frequent use of those long strings of descriptive adj ectives and adverbs [ ... ]. (BB-CC, p. xv) Example (29)37 conforms to the basic pattern, with the extra fe atures of an address in the direct quotation. A second manuscript of the same version of Brut y Brenhinedd, the Book of Basingwerk, has present tense y mae for imperfect yd oed, and thus begins the direct quotation one clause earlier than Cotton Cleopatra, an instance of variation within a text's transmission (see also example (32) below). gryfder, onyd ogeluydit. Ap hebythei y mein hynny ymha val y maen yno, wynt asseuynt yn dragywydawl ».  This Merlin advised then, to go to Ireland to the place which was called Cor y Kewri [The Giants' Circle] on Mynyd Kilara [Mount Kil ara] , since "there were stones of marvellous appearance -«a nd there is not anyone, lord, in this age who would know anything about these stones, and they cannot be got by might or by strength, but by art. And if these stones were here as they are there, they would stand fo rever ».
In passage (30) slipping from indirect to direct quotation is clearly signalled by a change from third plural reference (in bold) to tirst plural (underlined). The syntax of the indirect quotation is rather complex and the use of the phrase sef achos oed is more reminiscent of main clauses than of subordinate ones. Furthermore, non-finite verbal fo rms are used in the direct quotation (bod) were one would probably expect more naturally finite ones (but compare (25.1) above).
(30) Ac yna y dy wat gwr doeth onadunt na ellynt gyt oeisi yn hedychaul yn er vn kyuoeth yndragywydaul, sef achos oed, pan deley cof vdunt y lladuaeu awnaethessit vdunt ar galanasseu, y dodeynt en ev bryd yny geffeynt ymdiala, gwedy delei Ilawer 0 amseroed ac amylhau yr bobyl athyuu ryuel ryngthunt, ny bydei ryued heuyt goruot 0 deu parth yr ynys ar ytraean -« a bod yn waeth an kyfl e yna no gyd gynt, a bod yn weil yn yr awrhon canys go ruuam arnadunt kymryt y gan pandrassus vrenhin groec ignogen y verch yn wreic y brutus an tywyssawc � a digawn 0 eur ac ariant, a gwin a gwenith, a meirch ac aruew a llongheu m3 8 dwyn y ynys arall lle my nno duw BI bresswyliaw yn hedwch yndragywydawl ». (BB-CC 15.2-11 ) And then one of their wise men said that they could not live together peacefully . in one realm fo rever, this was the reason that, when they remembered the slaughter which had been inflicted on them, and the massacre, they would keep them in their mind until they obtained revenge, after much time had gone and the people had multiplied and war had arisen between them, it would not be strange either that two parts of the island would conquer the third -«and our opportunity was [lit. being] "Norse then than before, and it is [lit. being] better now, because we have conquered them, to take from Pandrassus, the king of Greece, Ignogen, his daughter, as wife fo r Brutus, our prince and plenty of gold and silver, and wine and wheat, and horses and ships to bring us to another island where God wants us to live peacefully fo rever ».
A similar use of non-finite verbal fo rms in a direct quotation occurs in (3 1) (treissiaw, alltudaw, anreithiaw, dehol), and there is some uncertainty about the place of slipping which may set in either before or after the clause canys 0 arallwlat y dathoed yr ystrawn genedloed hynny yn ormes ar ynys brydein.
( 3 1 And then Dunad sent again to Austin to tell hirn that it was not appropriate for them to preach to that cruel race [-« ]since it is from another country that these foreign races came as a plague to the Isle of Britain, [ -«]and thraugh their treachery and their treason they killed our ancestors and our race, and fo rcing us from our rightful inheritance, and exiling us and harrying us and banishing us, sorne from the island, others to the fringes of the island. And therefore it does not appertain to us to preach to them or to submit to anybody, except ta the Archbishop of Caerleon, since it is he who is primate over the Isle of Britain ». (32) is an example of differences with regard to slipping within the manuscript transmission (see also (29) above) ; the text in Cotton Cleopatra has indirect quotation (32.1), the corresponding text in the Book of Basingwerk shows slipping (32.2), but with a brief and probably inadvertent return to the personal reference of indirect quotation in i urawt idaw ef « his brother ta him » (in bold), as in the Cotton Cleopatra text, where one would expect in direct quotation *dy urawt itti «yaur brother to you ». Two non-finite verbal fo rms are used in the direct quotation, eruyn and eoffau, where the sense would require a first singular and an imperative respectively.
[ ... ] and asked him, fo r the sake of the Being who . created him as a man in her body from something without anything and fo r the sake of the breasts which he had suckled, and fo r the sake of the pain and the anguish she had suffered fo r him before he came to this world, that he should moderate his anger and that he should not cause the shedding of so much noble blood that has been assembled here from every country, and that he should remember that his brother had not done him any wrong [ ... ].
(32.2) [ ... ] ac yn erch i idaw yr gwr ay kreawd ef yn dyn yny chorfhi 0 beth heb dim, ac yr y bronneu ry dy nassei -[Bk of Basingwerk] «ac ir y boen ar dolur a diodeuais i erot ti kyn dy dy uot ir byt hwnnw, aC,eruyn ytty arafheu dy yrlloned. Ac na phar di lad y sawl waet bonhedic a ry gynullwyt 0 bob gwlat hyt yno, a choffa u na wnaethoed i urawt idaw el dim or cam, namyn ti a wnaethoed y cam ith uraut ac yt dy hun pan elut i geisiaw porth bren in lly chlyn i oresgin ynys brydein iar dy urawt, yr honn a oed mwy dy ledus ith urawt noc oed iti ac arwyd na wnaeth dy uraut yt dim cam, namyn yn ymdiffi n i gyfi ownder ac ef ehun dy yrru di 0 vrdas uechan i un uawr ». (BB-CC 50) [ ... ] and asked him, fo r the sake of the Being who created him as a man in her body from something without anything and fo r the sake of the breasts which he had suckled -«and by the pain and ache that 1 suffered before you came into this word, and asking you to moderate your anger. And do not cause the shedding of so much noble blood that has been assembled here from every country, and remember that his brother had not done him any wrong, but you did the wrong to your brother and to yourself when you went to seek the support of the king of Brittany to conquer the Isle of Britain from your brother, the one which more properly belonged to your brother than to you, and note [ ?] that your brother did not do you any wrong, but by defending his equity and himself he drove you from a small honour to a great one ».
The fo llowing passage (33) is probably not an example of slipping : (33) Ac yna y dy waut llyuelis y rodei el idaw y ryw bryuet, ac erchi idaw ev briwau mevn dwfyr gwedy y delei adref, a dy uynnu paub y gyt or a oed yny deyrnas, abwrw y dwuyr hwnnw yn gyffr edyn ar y bobyl. Ac el a gadarnhaei y bydei varw y coranyeit ac nad argywedei ar y bryttannyeit. «EU ormes a dy weist yw dreic oc awch kenedil chwi a dreic arall 0 ystrawn genedyl yssyt yn ymlad [. .. ] ». (BB-CC 67. 1-6) And then Llyuelis said that he would give him a kind of insect and asked bim to crush them in water after he had come home and to summon together what was in his realm and to throw this water indiscriminately over the people. And he affirmed that the Coranyeit would be dead and that it would not harm the Britons. «The second oppression that you mention is a dragon from your nation and another dragon ofa fo reign nation who are fighting [ ... ] ».

147
The required semantic cohesion between the indirect quotation and the direct one is missing and no connective is employed. The direct quotation begins a new narrative unit and therefore does not represent the climax in a unit of thought presented as a sequence of indirect-to-direct quotation, as would be the case in artful slipping40. Passage (34) is a rare example of ambiguity : John Jay Parry, in his translation of the Cotton Cleopatra text, translates as slipping, «And he told them that since [ ... ]. «And fo r that reason [ ... ] ». » Syntactic subordination of a noun clause is fo nnally neither obvious nor necessary, and 1 think that it is at least possible to interpret both sentences as direct quotation, as in my suggested translation. It should be pointed out, however, that slipping would yield good sense functionally, since the last sentence represents the most important part of the speaker's utterance.
(34) A menegi ydunt : « Yr pan doeth Max en wledic a chynan meiriadauc gyntaf y ly daw a dy ledogeon ynys brydein gyt ac »ynt ny bu yr hynny hythediw yno a allei gynydu breint na y gynal, ac am hynny drwc yw gennyfi auch bot mor wan ac na ellwch ymdiala ac wynt ». (BB-CC 206. 1 6-207.3 ) And he said to them : « Since Maxen Wleddig and Conan Meiriadog came first to Brittany and the nobles of the Isle of Britain with them, there was nobody from that day to tbis who was able to increase privilege or to maintain it, and therefore 1 consider it bad that you are so weak that you cannot wreak vengeance on them ».
My final set of examples is taken from the medieval Welsh laws. Here slipping would appear to have a slightly different primary motivation from slipping in narrative contexts considered so far : the direct quotation reproduces the exact wording which is, or can be, expected to be used at a specific point in a legal procedure, as in (35). This text then continues to reproduce the ensuing dialogue between cl aimant and judge. Since it is not the complete legal statement, but oruy a part of it, which is given in direct quotation, there is still a c1ear functional parallel between narrative and legal slipping -in both the mûst important part of the utterance appears as direct quotation.
(35) 0 deruyd y yngnat barnu cam vrawt yn y dadleu, a dy uot y dy n y barnwyt arnaw y gamvrawt yn y herbyn, a gofyn y'r yngnat a gadarnhaei y vrawt -« ac os kadarnhey mi a rodai ryng gwystyl y 'th erbyn ».  If it happens that a judge gives a fa lse judgement in the court and the man against whom the fa lse judgement was given, comes against him and asks the judge whether he affirms his judgement -« and if you affirm it 1 will give my pledge against you ».
Example (36) illustrates nicely that the presentation of such legal statements as direct quotation is, however, not mandatory in medieval Welsh law texts. The idea that the claimant has a sufficient number of witnesses fo r his claim is first expressed with slipping into direct quotation and then in indirect quotation.
(36) a dy wedet el y ryuot el neu e tat kyn noc el eysted ar e tyr hun ar dayar truy estyn argluyd ac ar ac eredyc ydau athy ac anl/oed bluydyn a bluydyned ae re erru en agkyureythyaul y arnau ac -«0 syt a amheuho henne mae eme dvga un ae guyr fer revot el en kehyt a henneJ », ac 0 syt a amheuho er ry erru en agkyureythyaul bot ydau dygaun ae guypo. (AL vii.i. 27) [ ... ] and let him say that he, or his fa ther before him, has been sitting on that land and earth by investiture of a Lord, and has had tillage and ploughing and house and home fo r a year and years, and that he was unlawfully ej ected from there -« and if there is any who doubts that, 1 have enough who know it, that he was there as long as that », and if there is any who doubts that he was unlawfully ej ected, [let him say] that he has enough who know it.
scap varn. oc berum ver pann fa nga quiô her i dom fram. (Gragéts 387) « These five men are to speak in these terms at the court and swear oaths beforehand and say that this man was killed in that kingdomand name the man killed -«and that this man N.N. who is now prosecuted here was there at that time -and we give our word of honor on that and we pronounce this scratch verdict here before the court ». » (LEI 233) (37.2) Sa maôR er secô vil! gera scal nefa ser vatta. oc moela sva. Nefn i ec ijJat vette. At ec nefni pesa menn xii. meô mer isattar do m. eôa secôar dom oc gera secô .N. 0 slica sem ec vil gert hafa . oc ver verôum a sattir. (Gragas 28 1) « The man who means to impose outlawry is to name witnesses and speak thus : «1 name them to witness that 1 nominate these twelve men inc1uding myself to join a settlement court or an outlawry court and to impose such outlawry on N.N. as 1 wish imposed and we agree upon ». » (LEI 207)

149
On the basis of my analysis of instances of slipping in Middle Welsh prose of different genres, comprising native and translated secular and religious narrative as weIl as historiographical and legal texts, 1 venture the fo llowing generalizations : slipping occurs in aIl these genres, albeit not in aIl texts that 1 have looked at, but has an over-aIl low frequency. It appears to be an optional stylistic device based within the medieval Welsh tradition43, since slipping may be introduced into Welsh versions of fo reign-language sources, in which slipping does not occur in corresponding passages (as far as we can judge on the basis of the extant texts), or during the course of a text's manuscript transmission. Instances of slipping are characterized by a set of fe atures which require the presence of a thought-unit of indirect and direct quotation in immediate sequence, which is marked by a fo rmaI syntactic break, but linked by semantic, and often also fo rmaI, cohesion between them, typically without special tagging devices in the direct quotation ; the direct quotation reproduces verbatim the most important part of a speaker's utterance. For legal prose this functional description may need to be slightly modified : the direct quotation reproduces verbatim the most important part of a legal statement to be uttered at a specific point of a legai procedure. In only fe w of the examples discussed above does the direct quotation have the character of a relatively short punch-line. The typology of the relative length of indirect and direct quotations respectively in slipping deserves further attention. My findings led me to suggest a graded concept of slipping, in view of examples which 1 would still regard as representing slipping, but which fit the above criteria less weIl than others, due to, fo r example, the absence of a connective 43 1 do not want to commit myself at this stage as to its oral, literary, or mixed origin.