Digitization, textual variation, and textual criticism

getting a better grip on the defiant multiplicity of textual traditions

C. M. Sperberg-McQueen

Oxford, 13 July 2010


Textual variation (general)

Some facts of life about text

Any work extant in multiple manuscripts or editions almost certainly has textual variants.
Uncomfortable, but true.

Does it matter?

Why does it matter? Because we may want:

Textual variation as dirty secret

Editors suffer from neglect. The audience does not always care.
Shocking, but true. They do not know, and do not want to know.
The facts of textuality make many uncomfortable.

Variation and the modeling of text

What is varying text, really?

Good question.
Er ...

What do we want to do with variants?

What is varying text? (take 2)

Take a simple case.

They flee from me that sometime did me seek
With naked foot, stalking in my chamber

Text is a stream.

Text as stream

But then we also have (Egerton MS)

They fle from me that sometyme did me seke
With naked fote stalking in my chamber

and (Tottel)

They flee from me, that somtime did me seke
With naked fote stalkyng within my chamber

Hmm ...

An echo

’49 Wyatt, 01549 Wyatt.
Coming sergeant.
Pick ’em up, pick ’em up — I’ll stalk within yer chamber.
David Jones, In Parenthesis, 1-6.

Text as stream (2)

How about ...

Text is a stream with multiple braided channels.
Cf. the Rhine Delta.

Methods of representing variation

Simple-minded diff / change markup

Some non-TEI vocabularies keep it very simple. For example, a standard xmlspec idiom:
Some applications 
<phrase diff="del">may</phrase
><phrase diff="add">will</phrase> 
require constraint capabilities not expressible in this
language, and so 
<phrase diff="del">may</phrase
><phrase diff="add">will</phrase> 
need to perform their own additional
diff = (del | add | chg), appears on any element.

Simple-minded change markup (2)

An additional attribute can say when the change was made:
Some applications 
<phrase diff="del" dg="may">may</phrase
><phrase diff="add" dg="may">will</phrase> 
require constraint capabilities not expressible in this
language, and so 
<phrase diff="del" dg="may">may</phrase
><phrase diff="add" dg="may">will</phrase> 
need to perform their own additional
Allows multiple change-marked renderings:
  • between 1.0 and now
  • between last public draft and now
  • between status-quo text and change proposal
But weak.*

TEI analogues of simple two-level change markup

Cf. TEI <reg>, <orig>, <add>, <del>. E.g. Tottel's changes vis-a-vis the Egerton MS:
<l>With naked 
<reg orig='fote'>foot</reg><add>,</add>
my <reg orig="chambre">chamber</reg>.
But what about stalking/stalkyng?

Location-references apparatus

A simple TEI method for transcribing existing apparatus.
Part 1: base text:
<l n="n1">They fle from me that sometyme did me seke</l>
<l n="n2">With naked fote stalking in my chambre.</l>
<l n="n3">I have sene them gentill tame and meke</l>
<l n="n4">That nowe are wyld and do not remember</l>
<l n="n5">That sometyme they put theimself in daunger</l>
<l n="n6">To take bred at my hand; and nowe they raunge,</l>
<l n="n7">Besely seking with a continuell chaunge.</l>

Location-references apparatus (2)

Part 2: apparatus:
<app loc="37.1">
<rdg wit="#D">boeth gentill</rdg>
<app loc="37.7">
<lem>seking with a</lem>
<rdg wit="#D">seking</rdg>
<app loc="37.9">
<rdg wit="#D">especiall</rdg>
<app loc="37.21">
<rdg wit="#D">What think you bye this that she hat deserued.</rdg>
A way out for retro-conversion.
Not suitable for new projects.

Double endpoint attachment

Marks the location more explicitly.
Part 1: base text:
<l n="w37.1">They fle from me that sometyme did me seke</l>
<l n="w37.2">With naked fote stalking in my chambre.</l>
<l n="w37.3">I have sene them 
<anchor xml:id="w37.3a"/> gentill <anchor xml:id="w37.3b"/>
tame and meke</l>
<l n="w37.4">That nowe are wyld and do not remember</l>
<l n="w37.5">That sometyme they put theimself in daunger</l>
<l n="w37.6">To take bred at my hand; and nowe they raunge,</l>
<l n="w37.7">Besely <anchor xml:id="w37.7a"/>seking with a 
<anchor xml:id="w37.7b"/> continuell chaunge.</l>

Double endpoint attachment (2)

Part 2: apparatus:
<app from="#w37.3a" to="w37.3b">
<rdg wit="#D">boeth gentill</rdg>
<app from="#w37.7a" to="w37.7b">
<lem>seking with a</lem>
<rdg wit="#D">seking</rdg>

Double endpoint attachment (3)

(1) More precise than location-referenced method.
(2) Allows overlapping variants: e.g.
Base: aaa bbb ccc ddd eee fff ggg ...
MS 2: aaa bxb cxc ddd eee fff ggg ...
MS 3: aaa bbb czc dzd eee fff ggg ...

(3) Easy* to do bilateral collations.

Double endpoint attachment (4)

(1) Not that easy to do collations. Hard to eyeball.
(2) Expensive to process.

Parallel segmentation

Probably the preferred method for new digitizations in TEI.
  • Easier to process.
  • No base-witness bias necessary.
  • Clutter in the document.
  • May require mechanical support.*
  • Difficulty with structural variation.*

Parallel segmentation (2)

Stolen* by TEI from Manfred Thaller.
Essential idea: no overlapping variants. Chop them into parallel segments.

The segmentation of Procrustes

The choice of segmentation points is not mechanical..

Base aaa bxb xxx cxc ddd eee
MS 1 bbb ccc
MS 2 czc yyy yyy dzd

Parallel segmentation (3)

  <rdg wit="Base MS2">bbb</rdg>
  <rdg wit="MS1"     >bxb xxx</rdg>
  <rdg wit="Base">ccc</rdg>
  <rdg wit="MS1" >cxc</rdg>
  <rdg wit="MS1" >czc</rdg>
  <rdg wit="Base MS1">ddd</rdg>
  <rdg wit="MS2"     >yyy yyy dzd</rdg>

Parallel segmentation of Wyatt

<l>Into a 
<app type="substantive">
  <rdg wit="#T">bitter</rdg>
  <rdg wit="#M #D #L"
   ><app type="orth">
      <rdg wit="#M #D">straunge</rdg>
      <rdg wit="#L" type="norm">strange</rdg>
<app type="orthographic">
  <rdg wit="#M #D">fasshion</rdg>
  <rdg wit="#T #L" type="norm">fashion</rdg>
<app type="orthographic">
  <rdg wit="#M #D #L">forsaking</rdg>
  <rdg wit="#T">forsakyng</rdg>
</app><app type="punc">
  <rdg wit="#M #D #L">;</rdg>
  <rdg wit="#T">:</rdg>

Open problems, challenges, possibilities

Some unsatisfactory situations: