Errors, Corrections and other Textual Problems in Three Copies of a Middle English Antidotary

Error analysis has been traditionally conceived as the step prior to any critical editing, providing the editor with grounded arguments to dev is a stemma that would accurately reflect the relationship between the ext ant copies. Yet, the scenario for texts other than literary changes, as with scientific tex ts, in which accuracy in terms of content stands out over faithfulness to the origina l i terms of form. Anyway, errors and other textual problems may provide clues as to how manuscripts circulated and scientific knowledge was disseminated. This article analyses scribal practice in three copies of the same Antidotary, focusing on scribal errors, corrections and other extual problems, which will serve to account for the diver gences and similarities they show. For the purpose, each copy is described and their i nd vidual textual problems are categorised and discussed. This will help to illust rate the dissemination of scientific knowledge, as well as varying scribal practice, whi ch will in turn point at the possible relation between the copies.


Introduction
Textual criticism has traditionally relied on error detection, which has been used to group manuscripts into different branches of stemmata (Reynolds and Wilson 1978: 190;Crossgrove 1982: 45) that reflect the links between the former. The notion of error has normally been based on the premise that scribes tended to systematically go wrong: the more removes a copy was at from the exemplar, the more errors it was supposed to display and the more defective it was (see also Donaldson 1970: 113). 2 As Vinaver puts it, "'textual criticism' implies a mistrust of texts " (1939: 352). The natural tendency was, in short, to simplify the text found in the exemplar (Reynolds and Wilson 1978: 199;Jacobs 1992: 61). Within this framework, any scribal intervention on the text being copied has also been assumed to be an error, and this includes all attempts on the part of scribes to improve the text from a lexical, syntactic or textual viewpoint, as Crossgrove reports (1982: 56). 3 Another classic supposition is that errors are the consequence of the copying process itself, which has been taken to be 'auditory' (that is, the result of dictation). However, the fatigue of the scribe, lapses of memory and even the imperfect use of sight have also been listed among the possible factors leading to copying errors (Petti 1977: 30). Vinaver also links errors to the very mechanisms driving the copying process (for instance, going back and forth from the exemplar to the copy), as opposed to the writing one (1939: 353).
More recently, several scholars have warned against a carefree approach to errors, since "the identification of textual error is linked to one of the most thorny issues in editorial theory, that of editorial interference, or emendation" (Rauer 2013: 148). In fact, evaluating errors implies editorial judgement being imposed on the text to be edited or analysed, something against which Vinaver also reacted (1939: 352). Moorman is reluctant to emendation, too, and advises that "[b]efore making any change, the editor should (1) make every reasonable effort to justify the MS reading and (2) make no change without having a clear, articulate, and positive reason -linguistic, textual, palaeographical, whatever -for doing so " (1975: 57). 4 Laing and Lass follow the same line of reasoning and highlight that "much of what tends to be dismissed as 'scribal error' rather represents writing praxis no longer familiar to us" (2009: 1). For that reason, this article builds on the individual examination of each copy, on the copying practices followed and on the resulting errors and textual difficulties, so that neither the reconstruction of the archetype nor the establishment of a stemma, which belong mostly to the domain of critical editing, become the main goals. Similarly, textual problems are pointed out, and these are discussed and assessed as likely errors or as instances of variation.
Many studies on scribal errors are based on literary texts (such as those on The Canterbury Tales -e.g. Blake's 1997 study on the language and style of the additions made to this work-, as well as Rauer's 2013 study on the Old English Martyrology), since these belong to the type of text that is normally edited, especially critically, which gives scope for the systematic analysis of the variant readings (along with the errors) in the extant copies. Yet, as Crossgrove points out, there is also a clear interest in other types of texts, such as scientific ones, even though the primary goal in these cases may not necessarily be the reconstruction of the lost archetype (1982: 58). 5 Actually, as Hudson explains, auditory copying might result in several original texts being produced at a time (1977: 45-46), a situation that would also rule out the possibility of tracing back a single original or archetype text, irrespective of the type of text produced. A further complication stems from the typical lack of concern for the concept of authority, which increases in scientific texts as opposed to literary ones: medical treatises could be more easily blended, excised, expanded, etc. upon the practitioner's needs, whereas literary works were more frequently perpetuated in a relatively fixed form. 6 Hence, traditions are far more flexible and open to change.
In the light of this, the present article focuses on the analysis of the scribal errors, corrections and textual problems found in three versions of the same Middle English (ME) scientific treatise, an Antidotary. This medical treatise is described in section 2, along with the witnesses that hold a copy of it. Section 3 describes the methodology followed to gather together the errors in each copy, along with the typology used for classification. This classification and the analysis of the errors and textual problems in each witness are provided in section 4. The results are discussed in the conclusions section, which also contains final remarks on the possible stronger connections between particular copies using the data presented as evidence.

The text and its witnesses
The treatise under scrutiny has been overlooked in the relevant literature, which has led to its wrong cataloguing (or no identification at all) in several library catalogues, a commonplace when dealing with scientific texts (Voigts 1995: 185-186). Yet, a recent textual examination of one of the witnesses (Marqués-Aguado 2008: 58-64) has evinced that this is a composite text that blends excerpts from two extremely popular medieval treatises: first comes part of Mondeville's antidotary, included in his Surgery (Nicaise 1893;Rosenman 2003), and then follows part of Chauliac's second doctrine of the seventh (and last) book in his Magna Chirurgia (Ogden 1971). Compilations drawn from auctoritates such as Mondeville or Chauliac were indeed common in medieval England, especially at the end of the fourteenth century (Wallner 1995: viii). This was particularly true of reputed medieval scholars whose writings spread all over Europe, like Lanfranc or the two aforementioned surgeons, besides classical and Arabic authors.
The present research stems from work on the witness preserved in Glasgow, University Library, Hunter 513 (ff. 37v-96v) -hereafter H513-(see Marqués-Aguado 2008). 7 The identification of its sources and of its ultimate author proved to be a demanding task, inasmuch as this antidotary had been catalogued as an anonymous text in both Young and Aitken's (1908: 421) and Cross's (2004: 35) catalogues on the Hunterian Collection.
Additional work led to the finding of other witnesses. The one in Glasgow, University Library, Hunter 95 (ff. 156r-184r) -hereafter H95-was identified by sheer coincidence, as it had been wrongly catalogued. Labelled as an antidotary, it had been tentatively placed under the ME tradition of Mesue the Younger's Antidotary (Young and Aitken 1908: 102;Cross 2004: 15), a completely different work whose author's identity has even been questioned (see the discussion in Marqués-Aguado 2008: 74-75). The finding of this copy led to the identification of the incipit in Voigts and Kurtz's electronic database (2000), which allowed adding the following list of witnesses to the two already mentioned: London, British Library, Sloane 2463 (ff. 153v-193v); London, British Library, Sloane 3486 (ff. 3-18); New York, Academy of Medicine, 13 (ff. 152r-188v); and Oxford, Bodleian Library, Ashmole 1468 (ff. 139-171). Out of all these, the two Glaswegian copies have been selected for the study, along with one of the Sloane manuscripts (hereafter, S2463).

Methodology: error identification and typology 3.1. Identification of errors
The task to be undertaken does not lend itself well to methodologies such as automatic retrieval, for instance, or to any other kind of automatic processing of the text. Convenient and time-saving as these are, errors of various kinds (ranging from spelling errors to omitted fragments; see section 3.2) and other textual problems have to be identified through attentive reading and by taking into consideration the context.
In this situation, transcribing all the copies under scrutiny has proved to be essential to spot errors of various types. Others were identified through the lemmatisation and tagging of the texts for the compilation of the corpora described (see footnote 7), since this process required delving into the texts and their intricacies to provide suitable lemmas and tags. As a matter of fact, by having to supply lexical and morphological information about each word in each text, duplications, alterations of word-order, omissions of necessary words and other difficulties become more evident.

Typology of errors, corrections and textual problems
Several classifications have been set up to account for scribal errors, although similarities among them are also noticeable. Nonetheless, despite the establishment of such typologies, some scholars have remarked that certain errors may be difficult to classify (Reynolds and Wilson 1978: 200;Jacobs 1992: 61), and that these may even co-occur (Vinaver 1939: 361-362). Petti, for instance, remarks that errors are more difficult to establish in vernacular languages during the late medieval period, when syntactic norms were far more flexible (1977: 29), a problem that will become evident in our analysis (see also section 1 and footnote 4 in particular).
For practical purposes, Petti's classification (1977: 30-31) will be followed for the most part, although reference to other taxonomies will be made whenever necessary to account for errors otherwise unclassified. 8 Accordingly, four types of errors are identified: omission, addition, transposition and alteration. 9 Omissions are said to be the most numerous group, and comprise instances of haplography (writing once what was twice in the exemplar) and of homoeoteleuton (eyeskipping part of the text due to the scribe's going back to another instance of the same word which is 8 Yet, those classifications departing from the scope of the article will be left aside, such as Jacobs's 1992 typology, which builds on the psychological dimension (that is, on causes and contexts leading to errors, rather than on errors themselves). The same applies to Vinaver's 1939 classification, structured into six types of errors which emerge from the movement in which they occur (e.g. from the exemplar to the copy, from the latter to the former, etc.). 9 A similar study on a scientific text has been recently conducted by Esteban-Segura (2012).
further down in the text) (see also Moorman 1975: 58). 10 Reynolds and Wilson expand this group with two more types: omission of a line of text (which is specifically linked to verse), and omission for no apparent reason, which will be most helpful in the subsequent analysis. 11 Three types of additions may be found: dittography (writing a syllable, word or phrase unnecessarily twice), contamination (inserting extraneous material from elsewhere on the page, as clearly exemplified by Vinaver (1939: 359-360)) and insertion (which reflects scribes' attempts to improve what they believed to be a defective text -although this does not mean that they were necessarily right). As with omissions, Reynolds and Wilson present two more types of additions, i.e. additions of glosses and "additions to a text of a parallel passage originally written in the margin of a book by a learned reader" (1978: 206); besides, they exclude contamination. In Moorman's classification, these errors are distributed into spontaneous and determined variation, since dittographies are spontaneous, but insertions (to correct metre, grammar or sense, or, more generally, to clarify) are determined and wilful acts on the part of the scribe (1975: 57-59). 12 The definition of transposition in the three classifications surveyed is fairly consistent and implies reversing the order that particular letters, words or phrases had in the original. If it is only letters that are involved, this may be termed metathesis (Petti 1977: 30). Moorman, in turn, remarks that transposition is particularly noticeable when dealing with word-order (1975: 58).
The last group is that of alterations, which can be unwitting (when the scribe does not understand the text or the handwriting and provides what might be a likely reading) or wilful (when the scribe modifies something purposefully). The most common form is mistranscription, which may be caused by the scribe's difficulties to understand the handwriting of the exemplar, its dialect or language; by the confusion of letterforms (for instance, the ever-present problem of minims); by the misunderstanding of abbreviations or even numerals; by an awkward word division in the exemplar; etc. 13 Scribal corrections are also worth exploring, since they reflect subsequent supervision or correction of the main text. 14 As with errors, Petti's classification will be followed (1977: 28-29). According to it, three types of corrections are identified: alteration, insertion and deletion, the latter of which comprises different mechanisms, such as cancellation (crossing out), erasure (scraping the ink from the writing surface) or expunction (placing a dot under the letter(s) to be deleted), among others. Marginalia can also be used to emend an inaccurate or incorrect reading in the main text, although this means a later user's involvement with the finished text. Occasional reference will be made to marginal notes as instruments for correction.

Analysis
The analysis begins with H513, and then the other two copies (S2463 and H95) are described and analysed, in such a way that crossreferences are established among copies whenever needed. As 13 Besides the main causes for errors listed in section 1, the relevant literature reports that many errors arise from palaeographical causes (Reynolds and Wilson 1978: 211), such as those regarding mistranscription. Yet, palaeography may at the same time be useful for textual criticism; see Petti (1977: 29-30) and Marqués-Aguado (2013: 35-36), among others. 14 Whether the same scribe was responsible for correcting his work in medieval times or not is still a matter of contention, although it is commonly assumed that a different person undertook this task (Petti 1977: 28). explained above, some figures are provided for each type of error, but no statistical study as such is conducted.

Glasgow, University Library, Hunter 513
This is a medical miscellany which has been dated to the first half of the fifteenth century on account of features such as the scripts used or the binding (Marqués-Aguado 2008: 50-52). The text is executed by two hands, but the change of hand occurs unexpectedly in f. 95r, where no textual boundary is found, that is, neither the break between the two sections (see section 2) -which is found in f. 88v-nor a new chapter.
There are two words which result each from the omission of parts of other two words, as if these were blendings. These are 'whicchith', from "whiche worcheþ" (f. 66v), and 'madder', from "made pouder" (f. 94r). These examples may reveal either a manifest lack of attention on the part of the scribe during the copying task, or else a lack of 15 In the examples provided, the readings obtained from the transcriptions are consistently used and offered in single inverted commas (' '). This implies that word-division, punctuation and line and folio division (marked by means of <|> and <||>, respectively) are retained. On the contrary, double inverted commas (" ") enclose correct readings and PDE spellings offered for clarification. Dictionary entries appear in italics. Finally, square brackets ([ ]) contain omitted material (taken from H95, as explained further down), and braces ({ }) are used for additions. acquaintance with the specific language in the text. Indeed, "ma(d)der" is also a possible ME spelling (s.v. madder(e in the Middle English Dictionary [MED]) 16 for the present-day English (PDE) noun "madder" -hence a different word-, which is also used in the text on two occasions (ff. 60v and 93v), something that may explain the scribe's confusion. The difficulties deriving from scribal unfamiliarity with specialised language have also been dealt with by Keiser, who remarks that Robert Thornton, the scribe of the Liber de Diversis Medicinis, "had difficulty in reading it, perhaps because of his unfamiliarity with the technical language -a common problem in the copying of vernacular medical books in 15 th -century England" (2005: 33).
Nonetheless, most omissions (up to 97) may be probably put down to scribal carelessness or be motivated "for no apparent reason" in Reynolds ' (f. 96r). In all these cases, ungrammaticality is likely to arise. In 17 other cases, the articles that should precede the nouns, or else the numerals (especially PDE "one"), are missing, as in 'take [ane] ounce' (f. 44r), 'a vessell of glasse vnder [þe] dunge' (f. 73v) or 'of [a] collerik man' (f. 82v). These omissions may constitute a scribal preference, though.
Finally, important medical information, like apothecaries' measures or the amounts of certain ingredients, is some other times withheld. This is particularly frequent with 'ana', which serves to introduce a similar quantity or amount of several ingredients in a medicinal preparation, as in 'bdellium serapinum [ana] dim' (f. 45r). Other important medical information is also sometimes left out, as in 'mede ne þat is regeneratiue [moste be drie in þe firste degre and | a medicine] incarnatiue' (f. 62v): 18 by omitting this, important specifications about certain medicines are not conveyed. The same occurs with the long passage omitted in f. 45r -'[and þen boile it eftsones in atinned pan|ne wiþ aliZte fire · and þen lete it wexe col-|de and coile it]'-, which results in an incomplete recipe, since a stage in its preparation is skipped.

Additions
12 dittographies have been counted, one of which refers to a whole sentence (ff. 44v-45r) and two to particular syllables that are copied twice but whose second occurrence is rendered separately from the previous one by a blank space or by a line break, as in 'of rosyne | {syne} and' (f. 59r). The remaining 9 instances attain to certain words (as in 'dissolue {dissolue} thoo', f. 53v) or strings of words ('emplaster for hit {for it} is experte', f. 95v). Two of these take place between the recto and the verso sides of the same folio, a change which may explain the unnecessary repetition, as in 'Stronger remollitiues be || {be} competent' (ff. 80r-80v).
As for insertions made to the base text -maybe to 'improve' it (from the scribe's perspective)-, 14 cases have been found. 19 Different explanations may be put forward to account for the curious description about the bark of frankincense, 'the whiche is moy=|ste and drie and most stiptyke' (f. 64r). The scribe could have misread the exemplar (possibly "most") as 'moyst' (the manuscript spelling), maybe due to the difficult handwriting of the exemplar or to carelessness. Yet, another reasonable explanation is that there was no such difficult handwriting, but rather that the scribe failed to recognise 'moyst' in the exemplar as an alternative spelling variant for "most" (s.v. most in the MED). Whichever the origin of the error is, and in the light of the following word (i.e. the adjective 'drie'), 'and' seems to have been added to turn this sequence into a coordinated adjective phrase providing an enumeration of qualities. Be it as it may, the PDE 18 A blotting prevents the reading of the first word in this excerpt as "mede<ci>ne". 19 Insertions to the base text have also been made if compared to the French tradition reflected in Nicaise's edition (see Marqués-Aguado 2008: 61-63); however, these fall outside the scope of this article, as explained in footnote 2.
reading "moist and dry" is not feasible either in the context of medieval science, where frankincense was classified as a dry substance, and not a moist one (Rosenman 2003(Rosenman , vol. 1: 1024. Other insertions are made within words. Instead of improving the readings, these distort them, hence leading to misspellings. An example of this is 'consol{o}d{it}atiue' instead of "consolidatiue" (f. 64v).

Transpositions
Instances of transposition are scarce: there are no examples of transposition of phrases, but two of transposition of words (as in 'be maye not' in f. 41r, instead of "may not be"), and 11 of metathesis. Some remarkable examples are 'antitodarie', in which <t> and <d> are systematically transposed (ff. 37v, 46r, 59r and 90r), or 'deprosye' instead of "dropesie" (f. 94r). It is worth highlighting that all the occurrences of metathesis -but for the case of 'fro' instead of "for" in f. 51v-belong to the specialised fields of medicine or botany, hence their likely connection with scribal lack of familiarity with such language.

Alterations
Many alterations in this treatise may be explained by the scribe's lack of familiarity with the handwriting of the exemplar, since most of them concern one single letterform which may have been confused with the one featured in the exemplar, or else with sequences of minims (which may incidentally pose difficulties even to the most skilled modern editor). Some examples are 'contorie' instead of "centorie" (f. 38r) or 'hen matuiatiues' instead of "ben maturatiues" (f. 67v). A particularly difficult pair of letterforms seems to have been (presumably) long <s> and <f> in the exemplar, since these are confused in 'slekked' (used instead of "flekked", f. 40r) and 'semigrek'/'semygrek' (used instead of "femigrek", ff. 58r and 58v), among others.
Word-division was not fully standardised in the ME period, as evinced in 10 cases, such as 'stronglie' ("strong lie" in f. 53r) or 'apo|lipum' ("a polipum" in f. 71v). Yet, none of these hinder the reading of the text.
There are some alterations, as also shown in S2463 (see section 4.2.4), which lead to changes in the morphological information or the word-class of particular words. This happens, for instance, with 'The ·3· made lie' (f. 72v), where a past form is used instead of the expected imperative "make". On other occasions, alterations may be simply put down to (apparently stylistic) variation, as in 'þat shall be reduced be | hoote medecine' (f. 82v), since the plural "medicines" is found in the other two witnesses.
The last set of alterations can only be classified as 'wilful', since these present words which cannot be confused on the grounds of the handwriting or the minims employed in the exemplar. A case in point is the use of 'oyle' instead of (presumably) "floures" (first occurrence, in bold) in the recipe in (2), where such oil appears twice in the list of ingredients: (2) The fyrste take oyle of Camo|mylle dille seed ana ·2· ounce · floure of femygrek and | lyneseed and of barly ana ·3· ounce · oyle of dylle and | of Camomylle ana ·i· ounce · (f. 46v)

Scribal corrections
As for deletion, erasure is the most common method, being used in 'encence' (f. 38v), 'spueme' (f. 46r) and 'poudred' (f. 76r), among others. Similarly, the <a> in 'and' in f. 46r was erased, but no letter was added afterwards. Expunction and cancellation are used only once (Fig. 1), in combination, to signal out the adjective 'hard', which had been mistakenly placed before the noun 'eyren', instead of after it:

London, British Library, Sloane 2463
S2463 is a quarto manuscript. Its size (larger than that of H513) suggests that it may have been intended as a copy for display, as opposed to the likely use of H513 as a vade-mecum for a practitioner to carry with him. The neat writing space and the consistent margins support this hypothesis. As opposed to H513, one single hand deployed the text in Secretary script. It has also been dated to the fifteenth century, according to the Catalogue of the Sloane Collection (Ayscough 1782, vol. 8: 108).

Omissions
The same instances of haplography are found in both H513 and S2463, but there are fewer cases of homoeoteleuton in the latter, because 3 of those in the former are deployed correctly (including the example given in 4.  ' (f. 193v), which is also attested in H513 (f. 96r).
Certain omissions lead to indeterminacies in terms of medical information. Such is the situation with the omission of 'ana' (which is skipped more than 15 times), as in 'whyZt popie blakke popie [ana] halff' (f. 156r). The same applies to cases such as 'The ffirste is [pouder] of arnement' (f. 175v), where the way in which vitriol is to be used is not given.
Finally, there are some recurrent omissions that may constitute a scribal fingerprint in terms of language use, so that these would count as conscious omissions of linguistic features, as with the regular deletion of the preposition 'of' after al-maner (s.v., MED), as in 'al maner [of] brennyng ' (f. 156v) or 'al maner [of] hardnesse' (f. 188r). The same pattern is found with partitions; e.g. 'a libra [of] whete ' [f. 160r,], 'a litell [of] vínegre' [f. 161r]). Another linguistic peculiarity is the tendency to occasionally omit the conjunction 'and', a feature that leads to a rather paratactic style, as shown in 'take amidum [and] caunfer ana | [and] grynde' (f. 156r).

Additions
The number of dittographies is substantially reduced if compared to H513, since there are only three, all of which are proper to this witness, as in 'The .4 th . is | {is} made' (f. 165v).
Some of the insertions represent, compared to H513, an improvement of the text, since information is specified or expanded. This happens, for instance, in 'wexe .2.ounce {the fatnesse of an henne and | of a malard an 1 dragme.} and make thyn oynement' (f. 159v). 21 Some insertions improve certain grammatical constructions which were somehow faulty in H513, as in 'The seconde {is} for offen=|ciouns ' (f. 192r). On other occasions, insertions bring about a distortion of the sense of the text, whether in grammatical or in conceptual terms, as in 'and {of} one partie of' (f. 186v). Finally, and as also found in H513, there are two cases in which words are added, but they result in the ungrammaticality of the passage, as in (3): (3) A duche | man þat was cledde all in skynnes with outen clothe | {þat} broute þis medecyne fyrst to parys (f. 177r) As in H513, insertions that lead to misspellings are also present, as in 'pel{e}ter' (f. 158v) and 'spu{e}me' (f. 160v).

Transpositions
As with omissions and additions, there are fewer transpositions than in H513. Only five cases of metathesis are found, and these replicate those in H513. Likewise, there are no examples of transposition of phrases. In turn, three cases of transposition of words are found, such as 'for .2. causes' (f. 184r), which is rendered 'for causes·2' in H513 (f. 81v).

Alterations
Alterations which may have been caused by misreading on account of the handwriting of the exemplar are plentiful. Many of them are shared with H513, such as the confusion of <f> and (presumably) long <s> in words such as 'slekked' for "flekked" (f. 156r), along with 'rate' instead of "rote" (f. 160r), 'renne' instead of "reume" (f. 184r) or 'dialetica' instead of "diabetica" (f. 192v).
The use of certain abbreviations has also led to alterations if compared to H513. An example that was catalogued as a case of metathesis in H513 is now an instance of alteration ('cucurme', f. 170r). Some numbers are also altered, as in 'mirre ana .9.ounce the ieuse' (f. 166r), which reads '3' in H513 (f. 54v). The abbreviation for the idea that errors are difficult to classify. See also section 4.2.4 on the different measures used in H513 and S2463.
'ounce' is occasionally substituted for the one for 'dragme' (at least, 6 times), which becomes one of the most common apothecaries' measures in S2463. In doing so, the composition of recipes is altered.
Word-division in S2463 is definitely close to PDE, with one exception that might raise confusion: the noun phrase 'attemperaunce' (including a determiner followed by a noun) runs together (f. 166r), and this may be taken to be the noun attempra(u)nce (s.v., MED). 'amillne' (f. 186r) is also found ("a millne"), but this one does not pose difficulties.
There is yet another group of alterations that result in changes in the morphology of words which may in turn lead to a different categorisation in terms of word-classes. Some examples are 'the medecíne þat regender flessch must be' (f. 167r), which is clearly singular, as opposed to the plural in H513 ('the medecines þat regender | flessh muste be', f. 56r); or 'pouders myZt | sufficent' (f. 168v), which should read "suffisen" for this to be a verb in the plural present indicative adequately conjugated. 22 Finally, some peculiarities which are probably wilful alterations on the part of the scribe can be encountered. As with other categories and types, many reproduce those in H513, as in 'floures' for "oyles" (f. 160v). Yet a substantial number (up to 14) present a different wordchoice, as in 'comune malewis' (f. 162r) instead of 'somme Malewys' in H513 (f. 48v), or 'techyn' (f. 171r) instead of 'seien' in H513 (f. 62r).

Scribal errors
Insertions, which amount to 6, are normally marked by means of two slanting strokes. The inserted material is placed above the baseline, as in 'muste be \ more dessicatíue þanne' (f. 167r). One case merits further comment: during the revision process, the scribe or corrector of S2463 realised that 'may' had been skipped during the copying process and inserted it. However, this was done in the wrong place, since the final reading is 'be // may not' (f. 156v; the same as in H513, f. 41r), instead of "may be not".

Glasgow, University Library, Hunter 95
H95, another fifteenth-century copy (Cross 2004: 15), also gathers a complete copy of the Antidotary whose palaeographical and codicological features go in the line of those displayed by S2463 rather than H513. The text is presented in two tidy columns (labelled a and b), the script is extremely neat and ornamentation is also in use, with decorated initials and some colour; likewise, the codex is larger in size. These features suggest that this was intended as a copy for display.

Omissions
In contrast to H513 and S2463, no cases of haplography or homoeoteleuton have been found, which represents an outstanding improvement in terms of the copying technique. Other types of omissions are found, though, but they are also far less frequent (totalling 9). In three cases, individual letters have been omitted in three words (i.e. misspellings), as in 'lite[l] ' (f. 171v, a). In the other six cases, certain words are missing, which hinders the understanding of particular sentences, as in 'summe [ben] ablucíouns and | summe ben pultes' (f. 163v, b).

Additions
Only one possible case of dittography has been found, and it occurs within a word, ('preísed{d}e', f. 183r, a). Insertions concerning letters, individual words, or even clauses or sentences, however, are more common. An example of the first type is 're-|solí{c}tíue ' (f. 171v, b), whereas the second one may be exemplified with 'for to make compounde medicines {of} fo-|ment þe place ' (f. 162v, b), where 'of' renders the clause almost meaningless. Yet, some additions of the third set (i.e. clauses or sentences) may not actually be insertions proper, but be rather part and parcel of the 'original' text, hence bringing to light further instances of homoeoteleuton in the other witnesses, as in (4): (4) and Zit {Zif þeí myZte be founded at al tym-|es and ín euerye place. Zit} summe medicines | be so dere þat pore men maye not haue | hem (f. 158r, a)

Transpositions
Transpositions occur sparingly, but they correspond to those in the other two copies. These are the 4 examples of 'antitodarie ' (ff. 156r, a; 160v, a; 166v, a; and 181r, b) and the spelling for PDE "fretting" -which refers to the action of corroding or scraping-('fírtinge ', f. 180v, a).
A possible instance of transposition at word-level may be 'Take·4·ounce·of olde oíle and 8·ounce | of þe spume of siluer' (f. 160v, a), a recipe whose ingredients and measurements are reversed in H513 (i.e. eight ounces of such oil and four of silver; f. 46r). With the linguistic data available, however, it is not possible to determine which of the two readings is more accurate, and medical information on the advantages and counter indications of these ingredients should be sought for clarification.

Alterations
Alterations due to problems with handwriting are, once again, quite numerous and some of them repeat those in H513 and/or S2463, such as 'fac cessíuelye ' (f. 157r, a). Here the confusion between long <s> and <f> persists -but not in other problematic words in the other two witnesses, like PDE "fenugreek"-and an additional omission (<n>) is noted. Alterations peculiar to this copy are 'mastíl' for "mastic" (f. 159v, a) and 'lesseny' for "lessenþ" (f. 170v, a). 23 Abbreviations and numerals may be responsible for deviant readings such as 'haþ þre maner' (f. 182v, a, which should read "two", since only two possibilities are described), 'cucurme' (f. 167r, b) or 'excerscences ' (f. 173r, b).
Finally, wilful changes on the part of the scribe, or else their lack of understanding of the text, may account for readings such as 'woundes and sínewes ' (f. 167v, b), which should rather read "of", or 'and Zif it be but of blood allonelye' (f. 156v, b), which reads 'not' in H513 instead of 'but', and was also counted as an error (in view of the context, the most adequate reading would imply getting rid of 'but' or 'not'). Possibly the scribe's lack of medical knowledge may help explain the reading 'aggeracíouns' (f. 176v, a), which is an altered reading of Galen's famous work Aggregations.

Scribal corrections
It is important to note that, along with the few errors found in the text, corrections are not abundant either, which seems to build on the idea that utmost care was taken to copy the text as neatly and correctly as possible.
Besides the occasional erasure, there is one case in which the correction is made via the use of a marginal note. 24 When discussing the types of poultices (in the chapter on cleansing medicines), ten types are announced in the main text. However, the ninth type is not explicitly marked (as it so happens with the other types), to the extent that it may be difficult to establish whether this is an alternative for the eighth type or a completely new poultice. Yet, a marginal annotation overtly marks it as the ninth type (Fig. 4), a correction that is not found in either H513 or S2463:  24 Marginal notes in H513 and S2463 are mainly aimed to sketch the structure of each chapter in the treatise, thus facilitating the localisation of particular information that may be important from the specialised standpoint. These, however, do not normally add new information to the text.

Conclusions
This article has shown that, as Jacobs stated, "[e]very textual tradition is to some extent the product of individual scribes, none of whom was an automaton reacting predictably to a definable combination of factors" (1992: 68), a remark that has been applied here, as he suggested, outside the limits of romance and verse texts.
The first conclusion that may be drawn is that the concept of error is indeed slippery and that careful examination is required to label a linguistic phenomenon as such. Within this framework, some omissions have been catalogued as possible scribal fingerprints rather than errors in the texts surveyed, insofar as they do not hinder readability or understanding of the text, let alone produce ungrammaticality. Likewise, scribal variation needs to be considered, as with certain wilful changes concerning word-choice, especially in a period such as ME, when no single standard was in use. In fact, texts for medical practice need not be that faithful to the original/exemplar in this respect, but rather in terms of content. Besides, selecting a particular witness as the reference for comparison may lead to considering an excerpt an addition or an omission, as shown with the renderings under analysis. In the light of this, this research has stressed the differences between the witnesses, specifying which of these lead to ungrammaticalities and which may rather relate to scribal variation, besides signalling those which alter medical content.
Concerning types of textual problems, omissions and, to a lesser extent, alterations, rank as the most frequent types of errors, while transpositions and additions are comparatively infrequent. The difficulties of allotting particular errors to a single category have become evident, which prevents us from running a precise quantitative analysis. There are clear differences between the copies in this respect: H95 is a much more polished and complete rendering than S2463 and H513, with fewer errors. S2463 is, in turn, a more refined version than H513, since some of the errors in the latter are deployed correctly in the former, although it also features errors of its own. As a consequence of the higher number of errors in the main text, more corrections are added to S2463 and H513 than to H95, even though neither of the former underwent much revision in the light of the manifold uncorrected errors.
In spite of the similar dating for the three manuscripts (i.e. the fifteenth century) on the grounds of palaeographical and codicological features, simultaneous generation of the copies is unlikely, given, for instance, the omissions and additions of material in each copy. It rather seems that these copies simply perpetuate the incorrect readings of the exemplar(s) they were copied from, in which case literatim scribes -copying faithfully the text in front of them-would be at work, adding nonetheless their own errors during the copying process, many of which can be explained by scribes' lack of specialised knowledge. Although the witnesses this far not analysed need to be checked before reaching any definite conclusion, and notwithstanding the shared errors in the three copies (as with 'antitodarie' or 'semygrek'), the similarities between S2463 and H513 regarding omissions and misspellings are noticeable, which might reveal a closer link between them. An illustrative example is the scribal correction of 'be // may not' in S2463, whose altered word-order is not emended in H513 but reproduced. The excision of material in H513 (compared to S2463) because of the possible confusion of abbreviations for apothecaries' measures (see section 4.2.2) reinforces this hypothesis. This supposition is put forward not so much because of the traditional assumptions that copies feature more errors than their exemplars, but rather because of what the errors found reveal. Research at other language levels, such as dialectal ascription, is expected to supply data as to these similarities, which may provide further suggestions in terms of dissemination of this text as well as book production.