Shakespeare Electronic Conference, Vol. 6, No. 0891. Thursday, 10 November 1995.
(1)     From:   Don Foster <This email address is being protected from spambots. You need JavaScript enabled to view it.>
        Date:   Thursday, 09 Nov 1995 21:13:54 +0100
        Subj:   Re: SHK 6.0874  Re: Julius Caesar
(2)     From:   Don Foster <This email address is being protected from spambots. You need JavaScript enabled to view it.>
        Date:   Friday, 10 Nov 1995 11:21:35 +0100
        Subj:   Re: SHAXICON
From:           Don Foster <This email address is being protected from spambots. You need JavaScript enabled to view it.>
Date:           Thursday, 09 Nov 1995 21:13:54 +0100
Subject: 6.0874  Re: Julius Caesar
Comment:        Re: SHK 6.0874  Re: Julius Caesar
<Steve Sohmer asks:  Could Don Foster be so kind as to advise why he believes
Shakespeare played other than the title role in Julius Caesar? I knew Decius,
and Will didn't look a bit like him.>
It cannot be proven that William Shakespeare-the-player performed *any* of the
parts that the SHAXICON database selects as "designated Shakespeare roles."  It
can only be shown that these roles, subsequent to production, are ones that
Shakespeare-the-writer *remembers* (and recycles in his subsequent writing).
For whatever reason, Shakespeare's disproportionate lexical recall of
particular roles is indelibly stamped on his writing.(In the late plays--*Tmp.*
through *TNK*--and a few earlier ones, no statistically significant pattern
emerges).  It is now a demonstrable fact that Shakespeare (by a hugely
disproportionate margin) remembers the King roles in *AWW*, *1H4*, *2H4*,
*LLL*; the Antonio roles in *MV* and *TNt*; the Adam and Corin roles in *AYL*;
the Ghost and 1.Player roles in *Ham* (etc.) while "forgetting" other roles in
the respective texts; and when those plays were revived, or revised, or both,
Sh. "remembers" the corresponding roles all over again.  That Shakespeare also
*performed* these designated "Shakespeare roles" is not a fact, but a
justifiable inference.  (See previous postings on SHAXICON.)  I'd be quite
happy, with Steve Sohmer, to believe that Shakespeare performed the role of
Julius Caesar in *JC* (and, for that matter, of Polonius in *Ham.*), which
seems eminently plausible.  But while Shakespeare remembers the lexicon of
Decius in *JC* (and of the Ghost and 1.Player in *Ham*), the character-specific
lexicon of Caesar (and of Polonius) registers no lexical influence whatever on
Shakespeare's subsequent writing.
It bears repeating, however, that inquiry after Shakespeare's most probable
stage-roles has only limited usefulness unless such investigation is keyed to
other queries. SHAXICON can tell us less about casting than about early stage
history, sequence of composition, dating, and textual authority.  The
identification of Shakespeare's stage roles will never be ascertained beyond
doubt.  What makes Shakespeare's excessive mnemonic recall of particular roles
most useful is that these indelible patterns in the poet's language, when
collated with external evidence, can often indicate with considerable precision
when Shakespeare's plays were written, staged, or revised.
    --Don Foster
From:           Don Foster <This email address is being protected from spambots. You need JavaScript enabled to view it.>
Date:           Friday, 10 Nov 1995 11:21:35 +0100
Subject:        Re: SHAXICON
Some weeks ago, John Owen posted some incisive queries about SHAXICON which I
have no recollection of having answered. The SHAXICON Notebook has a chapter on
"Procedures for Verification"; but here are a few brief remarks to address the
very thoughtful posting of Prof. Owen:
Let's start with a metaphor, by thinking of the Shakespeare canon as a fixed
tapestry of language (a combined text comprising the plays that most scholars
agree are indeed Shakespearean).  Particular threads have been worked into the
web in a particular order (at least insofar as its linguistic threads were
entered by the same individual--but let's not assume anything, or rule out the
possibility that more than one individual may have worked on the canonical
texts, either simultaneously or otherwise).  Put simply, SHAXICON maps which
threads are touching which other threads, and it does so without altering the
position of the threads as we move from the study of one text to another.
The user of SHAXICON may establish chronological priority by tracing a network
of relationships that can be checked against one another for obvious tangles or
loose ends.  The chronological sequence generated by SHAXICON closely matches
the sequence derived from traditional textual scholarship; and the theatrical
runs identified by SHAXICON are fully supported by what limited archival
evidence we have concerning performances at court, etc. These are happy
outcomes, but SHAXICON in no way depends upon these external records. For
example, if the canonical plays were written from 1891-1913 instead of from
1591-1613, SHAXICON would still identify *Err* as an earlier play than *1H4*,
*1H4* earlier than *AYL*, and so on, simply on the basis of lexical contacts
that these texts have with other texts in the same linguistic web
(Shakespeare's combined plays and poems).
Here, then, is a brief response to John Owens's questions concerning how a user
of SHAXICON might follow a line of inquiry for test-text "Q":
John Owens:
>"1. Clearly, Mr. Foster used external sources after SHAXICON's stats >were
>generated in order to supplement or correct the apparent >assignment of roles.
>My question is, how much of this was incorporated >during the program's
>design. In other words, is this data truly a raw >count of rare words, or are
>we somehow getting a synthesis of what we >already expected to find?"
First, we can't really *expect* to find ANYTHING.  We can begin, however, by
measuring general lexical correlation between texts (this has already been done
for you, but my work can be tested for accuracy, in multiple internal and
external ways, by any informed user).  And what we find is in fact in keeping
with what Eliot Slater, Alfred Hart, and others have already told us: that
Shakespeare's rare-words tend to be clustered chronologically (albeit with
anomalies, as when MV rare-words are heavily present in OTH, or MND words in
TNK). There is nothing to assume: SHAXICON simply tells us which (other)
Shakespearean texts have the highest general lexical overlap with test-text Q,
by an electronic count of lexical contacts (quite apart from theatrical
performance, speaking character, and the like). There is no a priori guarantee,
of course, that lexical contacts will be clustered chronologically in the work
of any author--but as it turns out, they most certainly *are* clustered
chronologically in Shakespeare's work, to a degree that overshadows all other
generic, or topical, or source-related influences (about which, more in a
moment). We have not had, until now, any closed and fully objective system
whereby to measure internal evidence against traditional scholarship with
respect to dating; now we do, and in most cases SHAXICON agrees with
traditional scholarship.
This chronological clustering is true even of non-Shakespearean texts with
which Shakespeare became acquainted.  Thus, in 1598, words from Jonson's EMI
come pouring into Shakespeare's writing, with EMI forming a very sharp peak of
"lexical influence" with Shakespearean texts written in 1598/9. The same thing
happens again in 1604, when EMI-words again begin pouring back into
Shakespeare's writing.  This makes it looks as if Shakespeare for some reason
may have read Jonson's EMI in 1598 and 1604. But "influence" here refers to a
statistical surplus of lexical contacts between texts, without presupposing
which is the source and which is the recipient--e.g., Shakespearean texts of
1598 and 1604 might also, hypothetically, have exerted lexical pressure of
Jonson's EMI-F1.  Happily, SHAXICON also provides ways to test direction of
borrowing (most persuasively when we've got multiple texts by the other
author), and Shakespeare is in this case the obvious, or at least principal,
debtor. External evidence cannot change SHAXICON's lexical evidence concerning
Shakespeare's acquaintance with EMI, but it *can* provide us with an
explanation: Jonson himself informs us that EMI was first acted in 1598, with
Shakespeare among the cast; and we have record of at least one other
performance of EMI in 1604 (with or without Shakespeare).  The external records
and SHAXICON are thus in full agreement, though neither kind of evidence
depends upon the other. But our knowledge that EMI was acted in 1598 and 1604
cannot in any way alter the high frequency of lexical contacts between EMI and
Shakespearean texts written ca. 1598/9 and 1603/4.
>John Owen:
>"2. Does Mr. Foster have any references to provide background for
>SHAXICON's assumptions? ... can the validity of this method be >confirmed (has
>it been?) by some type of textual experiment outside
>the Shakespeare corpus, preferably where we have more external >information
>with which to corroborate the results?
The applications for this program are up to the user, not up to me, but the
resourceful user will have no trouble constructing "textual experiments" of his
or her own design, using both Shakespearean and nonShakespearean texts.  For
instance, we can test EMI to see whether Shakespeare as a writer "rembembers"
one role or another (relative to the character-specific lexicons of other
characters in the same play); or we may test whether EMI-Q or EMI-F1 exerts a
stronger general "influence" on Shakespeare's writing, or first EMI-Q, then
EMI-F1; and so on.  Cross-checks are only a keyboard away: we may test whether
Shakespeare's selective recall of, say, the lexical influence of the Egeon role
upon Shakespeare's post-ERR writing endures briefly, or only when he is writing
plays, or only when he is writing poems; or whether the Egeon role seems also
to influence Ben Jonson or John Fletcher (and in each of these instances, a
"yes" would send up a red flag). Many users will spend their early acquaintance
with SHAXICON hunting for internal contradictions or other irregularities in
the data.
Let's suppose, for now, that SHAXICON has already been cross-examined by a host
of skeptical Shakespeareans (as indeed it will be), all of whom have conducted
rigorous cross-checking, only to conclude that SHAXICON can indeed be trusted
when it tells us, for example, that ERR and TIT are early plays, that WT and
TMP are late, and so on.  We may then check whether Shakespeare indeed
"remembers" (uses and re-uses and re-uses) the particular words of particular
characters. What we find--always--is that a few hundred rare-words in each play
are simply "forgotten," never to be used again by Shakespeare-the-writer.  But
the rare-word lexicon of one character (or of two or more double-able roles) is
in each case remembered. This doesn't prove that Shakespeare studied and played
that role: there might be other explanations with with respect to CAUSE.  But
SHAXICON proves this much, that in a Shakespearean text of, say, 1599, one can
expect to find that the text will draw, disproportionately, on the designated
"Shakespeare roles" of the pre-1599 plays (i.e., the play written in 1599 will
draw chiefly on the character-specific lexicons that "influence" all other
Shakespearean texts to 1599); and the text written in 1599 will not so draw on
the designated "Shakespeare" roles later than 1599.
John Owen:
>"3[A]. How does Mr. Foster rate the evidence provided by SHAXICON? >3[B]Should
>it be used to correct, confirm, or supplement already >existing scholarship?
>Let me put this another way: SHAXICON supports >John Aubrey's statements about
>Shakespeare's performances, (Ghost in >Hamlet, Adam in AYL). If they were to
>disagree, would Dan [i.e. Don] >be confident enough to submit his version in
>place of the
>traditional account?
Replies (in order):
3[A]. SHAXICON can be used in conjunction with virtually any statistical
software, so that you can perform statistical checks (CUMSUM averages and the
like) to your heart's content.
3[B]. I might be confident enough to sumit my version in place of the
traditional account, but that in no way guarantees that anyone has to believe
me.  In some instances, the designated "Shakespeare roles" are very small
parts.  Although Shakespeare registers a highly disproportionate recall of both
the Egeon role and Dr. Pinch role in *Err*, the Egeon role is considerably
larger and Shakespeare's selective recall of this role has a higher
mathematical significance than his selective recall of the much smaller Dr.
Pinch role.
The most important point to be made is that the "designated Shakespeare roles"
are those that Shakespeare certainly *remembers*, not ones that he certainly
acted.  The inference that the man Shakespeare performed the particular roles
that the man Shakespeare remembered--that he remembered just one or two roles
after having created *all* of the roles in a given text--is not an
unjustifiable inference; and there is at least a family likeness in the roles
that Shakespeare remembered (usually, characters who are introduced in the
first stage entrance; and these are usually old men and friars, feeble kings
and attendant lords).   But I wish to underscore, once again, that the
principal value of these mnemonic registers lies in their use as a touchstone
for testing dates of composition, production, and revival; for evaluating
problems of revision and textual authority; for identifying markedly different
lexical pools within the same text; for tracking Shakespeare's reading and
source material (see earlier posts); and so on.  Focusing on the designated
"Shakespeare-roles" (as almost everyone has done so far in their queries aboutr
SHAXICON) is a little finding a set of keys and saying, "Oh, what interesting
keys"--without noticing that the keys can actually be used to unlock various
dusty chests in the cellar.
Other queries are, of course, welcome.  But I've been having a hard time
keeping up with the mountain of correspondence generated by SHAXICON queries,
so please be patient.
Don Foster

Subscribe to Our Feeds


Make a Gift to SHAKSPER

Consider making a gift to support SHAKSPER.