Make a Donation

Consider making a donation to support SHAKSPER.

Subscribe to Our Feeds

Current Postings RSS

Announcements RSS

Home :: Archive :: 1998 :: February ::
Foster on Shaxicon (Part Five of Six)
The Shakespeare Conference: SHK 9.0127  Wednesday, 11 February 1998.

From:           Don Foster <
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 >
Date:           Monday, 09 Feb 1998 09:09:32 -0500
Subject:        Foster on Shaxicon (Part Five of Six)

=============================================================
SHAXICON, PART 5, examination of my own work

Note: this message is intended for SHAKSPEReans, and for the SHAKSPER
List archive; not for unauthorized circulation. Don Foster

Let's turn now to the Hieatts' cross-examination of my own work. I'll
take it from the top, with the letters by the Hieatt brothers in the
_PMLA_ Forum (May, 1997).  This was an odd episode that passed largely
unnoticed: representing themselves as longtime users of Shaxicon, the
Hieatts ventured to discredit the database without first learning how to
use it.  Kent's letter revealed that he lacked even a rudimentary
acquaintance with Shaxicon:  in his Forum letter he guessed, and guessed
wrong, about the number and kinds of words included in the database.  He
supposed Shaxicon to index virtually the entire Shakespeare canon, with
omission only of those words that appear in more than twelve different
plays; when in fact the lexical pool is strictly limited to words
appearing no more than twelve times in all of the canonical plays
combined (with cross-reference to non-canonical and STC texts).  That
information may be obtained simply by _opening_ Shaxicon!  For examples
of Shaxicon-entries, Hieatt cited words that have never been indexed in
Shaxicon and never will be, because they do not fit Shaxicon's
principles of inclusion.  (This is rather like publishing a sour review
of a Spanish-English diction and citing, as examples, Portuguese words
that do not occur in Spanish.)  Kent dismissed the extraordinarily high
influence of Egeon upon FE, H8, TNK, and other texts, saying that the
Egeon-influence was statistically insignificant because the Egeon-role's
word-pool is quite small, while forgetting that the Egeon-influence
occurs in texts from 1594 onward; and forgetting also that this is just
one of dozens of matching distributions for FE and the late plays.  (An
analogy: rolling "snake-eyes" one time is insignificant; rolling
"snake-eyes" 16 times out of 20 is unusual.) Kent further explained the
high influence of Egeon upon FE as a function of Egeon's "lugubrious"
vocabulary, without pausing to see whether the rare Egeon-words that
appear also in FE, H8, and TNK are indeed lugubrious (but they are
not).  Kent's fuming letter (greatly tempered by PMLA's managing
editors, bless their hearts!) made it clear that he was damning a
database without so much as turning it on.  Yet his letter was crafted
over a period of several weeks, during which time he submitted multiple
drafts.  His letter was not just the product of a moment's rashness.

Charles Hieatt took another tack.  Shaxicon can be used to test various
patterns of lexical distribution in Shakespearean and non-Shakespearean
texts.  Most pertinent to the question of remembered stage-roles are the
<tokens : types>, and <types : types> distributions (as illustrated in
Shaxicon's Excel tables and Deltagraph charts, which the Hieatts seem
not to have consulted).  Charles Hieatt, working from my own brief
description of a sample test on the model, <tokens (sourcetext) : types
("influenced" text)>--and not knowing what to do with it-opined that the
<types : types> tests should be used instead, while confusing a  <types
: types> with a simple reversal of terms (i.e., <tokens ("influenced"
text) : types (sourcetext)>.  The Hieatts are mistaken.  For a full
text-profile of ERR or any other Shakespeare play, one must construst
_both_ the <tokens : types> and a <types : types> searches;  both kinds
of Excel distributions are necessary for the entire text (in this case,
ERR), and for each particular character-without which, one cannot draw
any confident conclusions.  To measure the relative lexical influence of
a text (or theatrical role) across the canon, all distributions must
agree on its distinctive influence upon Shakespeare's subsequent
writing.

The test that Charles Hieatt assails is the single most necessary place
to begin if one is to avoid drawing a false inference from Shaxicon
data. But first, a brief explanation of "tokens" and "types".  A "type"
is simply a lemmatized word. (An example:   _abject_ [ad.], which
appears in ERR, MV, SHR-F1, 2H4, 1H6, 2H6-F1, TRO-Q1, -F1, and in H8-F1;
also in Edward III, Sejanus, and more than 50 texts in Shaxicon's STC
cross-sample).  A "token" is any instance of a lemmatized word.  (Thus,
2H6 F1 contains 3 tokens of the type, _abject_ [ad.];  TRO-F1 contains 2
tokens of the same word, and TRO-Q only one.)  Such diachronic
repetition of rare-words is inevitable, but it occurs less frequently or
insistently than one might suppose.  Unlike _abject_ [ad.], most of the
rare words in Shakespeare are used only once and then "forgotten."
Diachronic repetition of some words and not others is partly conditioned
by the poet's subsequent return to texts studied at an earlier point in
his career, either as a reader (e.g., a re reading of texts in which
"abject" appears) or as an actor (in a revival of Shakespeare's own
plays in the theater).

Both kinds of repetition can be significant indicators of intertextual
contact. Each has its limitations.  Let's suppose that Shakespeare's
1592 reading includes Holinshed's account of King Henry VI, and Arthur
Broke's poem,  _Romeus and Juliet_.  To keep things simple, let's
suppose that these two texts (Broke's poem and Holinshed's life of Henry
VI) are of equal length and of comparable vocabulary richness.  Any word
frequently repeated in either or both of these two texts is more likely
to appear in Shakespeare's new writing in 1592-3 than words that appear
only once in those texts. This is a simple matter to demonstrate, not
simply as a likelihood but as an established fact.  In addition, any
word in these texts that is borrowed by Shakespeare in his 1592-93
writing (e.g., in ROM) is most likely to be repeated during the
intervals when ROM is in repertory.  Moreover, in Shakespearean texts
later than ROM, any word appearing in the roles of the Chorus and Friar
Lawrence are more persistently remembered (for whatever reason) than are
the rare-word lexicons of other characters in the same play.

The tokens : types test that Charles Hieatt assails is a simple and
necessary first step in distinguishing between diachronic repetitions
influenced by _synchronic repetition within a text_ (on the one hand)
from _diachronic repetitions influenced by repeated acquaintance with a
text or theatrical role_ (on the other).  Let's stay with ERR, since
this is the text challenged by Charles Hieatt.  Syracuse Dromio repeats
_buff_ (n.) and _rest_ (v. = arrest_), words that never reappear in
Shakespearean texts after 1594.  Egeon also repeats _desk_ (n.), which
reappears thereafter only in HAM.  Syracuse Antipholus repeats _sconce_
(n.=head), a word used also (one time) by Syracuse Dromio.  But _sconce_
(n.) reappears only twice after 1594, at HAM 5.1.102 (Q2, F1) and COR
3.2.99 (F1).  Syracuse Antipholus repeats _sorcerer_ (n.), a word
already employed by Sh. in 1H6 but that reappears after ERR only once,
in TMP (3.2.43). And so on.

Conversely, most rare Egeon-words are used only once in ERR, yet
initiate career-long adoption in Shakespeare's later writing.  (A
typical example: _to impose_ [v.}, which reappears in LLL (Q, F), 3H6
(F1), ADO (Q, F1), JN (F1), MM (F1), WT (F1), R3 (F1), and LR (F1).
Shaxicon actually _understates_ the persistent influence of this one
role on Shakespeare's later writing:  words spoken by Egeon that are
rarely found outside Shakespeare appear in Shakespearean texts more than
12 times from 1594 onward, and so fall out of Shaxicon's purview.

Once it has been established that a high rate of lexical influence is
not simply a function of synchronic repetition, one must provide a
cross-check in constructing a full-text profile using the <types :
types> test requested (but misunderstood) by Charles Hieatt.  The
Hieatts are mistaken in supposing that this test has been suppressed.
(In fact, one wonders why they didn't simply perform the test
themselves!)  As it happens, the <types : types> relation shows less
variability than <the tokens : types> measurement:

table 1         types:
                types

Egeon/ERR       82/653
                12.6

Egeon/FE-Q      6/29
                20.7

Egeon/H8-F1     12/58
                20.7

Egeon/TNK-Q1    12/45
                26.7

It seems apparent, from Charles Hieatt's confused letter to PMLA, that
he has never produced a single <types : types> search, much less an
entire profile.  Charles has merely reversed the left and right terms,
which produces the following results (with repeated tokens for FE, H8,
and TNK in the lefthand column):

table 2         types:
                tokens

Egeon / ERR     82/653
                12.6

FE-Q            11/30
                36.7

H8-F1           18/67
                26.9

TNK-Q1          14/49
                28.6

This simple reversal of terms advocated by Charles Hieatt preserves the
very problem that he supposes himself to be avoiding, which is the
simultaneous measurement of influence by synchronic as well as by
diachonic token-repetitions: he has merely placed the "influenced" text
in the left hand column. And Hieatt thus introduces a different kind of
potential bias, which is to allow for repetitions in the "influenced"
text to be counted as if repeat-tokens in writing had a significance
indistinguishable from repeat-tokens encountered in the poet's reading
or stage-study.

A simple reversal of the <tokens (source) : types (influenced text)> to
a <tokens (influenced text) : types (source text)> is not a waste of
time, but neither does it show what Charles and Kent Hieatt think that
it shows.  The test that the Hieatts are hunting for is a <types :
types> test shown above (table 1); which I'd have been happy to teach
them how to do, if they had only asked.  But the <types : types> testing
requires in each instance the extra step of deleting multiple-tokens in
the left hand column (and thus, hours of extra labor for a full-text
profile that includes all major character-roles).  The systematic
pattern of error that the Hieatts allege to have taken place with
Shaxicon is a figment of their own imagination, deriving partly from
their own confusion and partly from my own wish to make their use of
Shaxicon as user-friendly as possible.  In no instance does the
left-right reversal of terms significantly alter the evidence of a
<types : types> profile.  If the Hieatts believe otherwise, I urge them
to prove it.

The Hieatts wrongly imply that I suppressed testing by  <tokens (source)
: types (influenced text)> to a <tokens (influenced text) : types
(source text)> (as in table 2).  Hardly.  The apparent influence of the
Egeon-role, by Charles's test, is even more strongly registered on FE
than on H8 or TNK (table 2).  Moreover, other character and text
distributions confirm this inimitable match between FE and H8:   those
texts that most directly influence FE-not just ERR-are principally the
same texts that most directly influence H8 and TNK. And so on.  In other
words:  Kent and Charles are, again, mistaken.

The errors made by the Hieatts could have been avoided.  Kent lives just
a couple of hours away from Poughkeepsie.   In Jan.-Feb. 1997, Kent sent
multiple drafts of his letter to the editors of PMLA.  When I learned of
this development, I wrote Kent, and politely offered to assist him in
seeing his letter through to the press.  I had no wish to censor his
letter, only to make sure that he was not still confused about how to
use the database.  After answering a few of his questions, I received
responses from Kent in this vein:

====================================================
Date: Thu, 16 Jan 1997 15:03:35 -0500 (EST)
From: A. Kent Hieatt <
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 >
X-From: "A. Kent Hieatt" <
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 >
Subject: Re: yr letter
X-Sender: 
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 
To: 
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
  (Don Foster)
MIME-version: 1.0

Dear Don - Fine.  I'm sure you're right about tokens, types, and lemmas.
- I think I've been to the Forum once in my life, and unhappily won't be
there this time.  Shakespeare is not really one of my things. -Best,
Kent
====================================================

Kent meanwhile revised his letter and sent the new draft to the PMLA
Forum.

I invited the Hieatts to spend a day at Vassar (including Charles-I
didn't know at the time that he lives in Cambridge), so that I could
show them how to construct both <tokens : types> and <types : types>
profiles (the latter of which, as noted above, requires an intermediate
step in each individual search to suppress repetitions).   Hoping to
spare the Hieatts some embarrassment and myself some inconvenience
during an incredibly hectic period in my life, I offered to teach them
the basics of Shaxicon, Excel, and Deltagraph. This is the kind of
response that I received from Kent:

===========================================================

Date: Fri, 28 Feb 1997 11:47:58 -0500 (EST)
From: A. Kent Hieatt <
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 >
X-From: "A. Kent Hieatt" <
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 >
Subject: tutorials etc.
X-Sender: 
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
  (Unverified)
To: 
 This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 
MIME-version: 1.0

Dear Don - Many thanks for the first tutorial you sent me.  I've just
started on it, but am bothered by knowing that I'm working with SHAXICON
2.0 OF 1994, and that you are thinking of a SHAXICON of 1996.  I'll get
back to it after I've finished another 4-day task.  But why did you say
I 'requested' this tutorial?  I'm sure I didn't.  What you did do was to
suggest, unprompted, that my brother and I come to Poughkeepsie for a
face-to-face tutorial and a meeting with one or two associates. ...

===========================================================

That same week, Kent and Charles represented themselves to Ron Rosenbaum
of the New York Observer as experts in the use of Shaxicon:

[from the _Observer_:]

"I have been able to learn (exclusively! You read it here first) that
...  there will appear two letters alleging serious flaws in Professor
Foster's statistical methodology.  The letters were written by Kent
Hieatt, a respected Spenser scholar, and his brother Charles Hieatt, a
scholar based in Cambridge, England.  Both men have worked with
Professor Foster's SHAXICON database for some years, know it inside out,
and have spotted what they argue are some fairly glaring mistakes in
Professor Foster's calculations..."

Ron Rosenbaum, "The Edgy Enthusiast," _The New York Observer_ (24 Feb.
1997): 35.

I recommend this article to anyone wishing to evaluate claims made by
the Hieatts, and will myself be quoting from it at a later date.

The Hieatts did not have the courtesy to let me know that they had
released their letters to the press.  After reading Rosenbaum's article
in the Observer, I resolved to sit back and let the Hieatts say whatever
they pleased in the pages of PMLA.  They have now done so.   And with
ample column-space:   the Hieatt brothers got around the strict-word
limit on Forum letters by sending in two letters, signed individually.
One reader quipped that this was the old "cowboy hats on sticks until
the cavalry comes" trick!   I have a hunch that the cavalry will not be
forthcoming.

Kent Hieatt next went to SNL (and perhaps elsewhere) attempting to
peddle the same line of goods published first in PMLA, and now again on
SHAKSPER. For example:

KH:  "It's he [Charles] who discovered that every one of Foster's
calculations of these percentages was incorrectly extracted from
Shaxicon, because in WordCruncher, the program on which Shaxicon was
originally erected, Foster had consistently started with the wrong
half-screen in matching two texts, a very easy mistake to make."

This is pure rubbish, and will remain so, no many how times Kent Hieatt
repeats it, or in how many different forums.

I thank Hardy Cook for the opportunity to defend my work and conduct.  I
wish to close this unpleasant episode by saying that I bear no ill will
toward either Charles or Kent Hieatt.  I believe that their interest in
Shaxicon arises from a sincere concern for literary scholarship.  I have
never met Charles Hieatt, but I know Kent Hieatt to be a conscientious
scholar who has made a valuable contribution to Spenser studies.  I
regret having had a falling out with him over matters that began as a
mutual and entirely friendly discussion about dating the Sonnets.
 

Other Messages In This Thread

©2011 Hardy Cook. All rights reserved.