Witte, 1984, “Topical Structure and Revision”

Stephen. "Topical Structure and Revision: An Exploratory Study.” On Writing Research: The Braddock Essays, 1975-1998.
Ed. Lisa Ede. New York: Bedford St. Martin’s, 1999. 132-155.

Big Idea
Stephen Witte’s 1983 article reports on the quantifiable patterns of topical
structure in a sample of written revisions of a common text.  Through
comparative readings of revised texts and a common seminal text (from which the
revisers set out), Witte’s study surmises that the reviser’s treatment of
sentence topics correlates to the writer/reviser’s understanding of the text’s
discourse topic.  The relationship between a discourse topic and sentence
topic figures significantly into Witte’s work; he adopts a vocabulary of
specialized terms such as "topicalization," "theme/rheme,"
"macroproposition," and "hypertheme" in his elaboration of
methods.  According to Witte, sentence topics may or may not house
discourse topics, but the writer/reviser’s prior knowledge and readerly understanding
of the seminal text’s discourse topic guides the choices applied
throughout revision.  Witte acknowledges his methods are suited to
informational texts, collaboratively revised. Notably, he delivers some
eighteen-plus name references in the first three or four pages of the article (a
cluster of comp/rhet folks as well as several {unfamiliar-to-me} structural
linguists from the Prague

Wondering About
With all due respect, big chunks of this essay were a muddle; lesser chunks
were interesting in a structuralism-amuck, 1983-snapshot kind of way. 
Witte’s theoretical lead-in to the more empirical study sparked a few
interesting issues.  Without explicitly discussing sentence topics in terms
of links and relationships, Witte is centrally concerned with
syntactic cues, their systematic connections, and the role of the writer/reviser
in reshaping those cues toward a more coherent, unified discourse topic. We
could bend this analysis to rhythm or pulse–the discourse topic’s
dependence on the coordination of smaller units.   As provocative as
this is, the structuralist drawback impends: the study builds from a kind of
de-natured, de-cultured "writer’s hand(s)" (not unlike de Saussure’s
featureless talking heads).  The situation is absent: topoi sans kairos.  

The portions of the essay that scale the revisions from high-score to
low-score based on the sentence topic patterns (matches, deviations, etc.) were
hard to read.  It’s filled with statistical references, and it’s never easy
to connect the high/low assessments to specific texts (only a few of which are
sampled).  Witte notes that one of the setbacks in his study is the problem
of "no average text."  In fact, the whole piece is responsibly
self-conscious; he incorporates lots of reminders that this is "an
exploratory study," and it’s simply a frame for writing researchers to
consider.  But how should we use this? What other applications might
Witte’s work hold?  I don’t have a lot of ideas about this, but as I read,
I started to think that much of this analysis could be applied electronically
(especially the clause-length stuff).  In other words, when I want to see
revision (separated from the document), I simply use Word to compare
texts.  The changes are highlighted, easy to view.  I’ve never
considered the quality of a revision in terms of altered topic patterns;
instead, I simply have a glance at the depth of revision, the way the writer
responded to specific in-text suggestions or questions, and any oversights,
omissions, or clear decisions not to make changes.  And while I’m
not in favor of computers as stand-alone readers, I continue to wonder how
technologies can assist our reading by helping us see patterns in texts (not to
kick out sloppily composed standardized exams).  Witte’s approach, I think,
could be rendered into a software application–an application that might be
useful if we use it to see texts differently rather than measuring those texts
as successes or failures. 

Witte’s approach to measuring sophistication of revisions based on topic
patterns doesn’t acknowledge rhetorical strategies, deliberate re-arrangement,
topical abstraction or exemplification.  All of these forces ought to
figure into revision–even in classifiably informative texts, and studies of
topic structure alone might not reveal such developments.  It also sets up
knowledge of audience and revision in fairly narrow terms.  Revision isn’t
always (ever!) a sealed-off, exclusive, after-writing stage; knowledge of
audience, however carefully ascertained, is imperfect, incomplete. 
Similarly, while the sentence topics can be identified and tagged, discourse
topics spill, morph, shift–endlessly.  By this, I mean the sentence enjoys
punctuated boundaries; a discourse topic flows and is not frozen in time. 

"Although making inferences about composing processes from written
products is somewhat risky, the method I have outlined and applied to controlled
revisions of college writers appears to be a promising one for studying the
textual causes and effects of revision.  It is a method which may allow
researchers and teachers alike to study the decision-making processes writers
use during revision" (153).

"Whether the findings hold for other kinds of texts collected under
different circumstances and evaluated by different kinds of raters remains an
open question" (153). 

"In this regard, topical structure analysis–unlike the analytic methods
designed to examine the effects of the revision–enables the researcher to
explore the relationship between the textual causes of revision, the text
features to which the writer as reader responds, and the effects those changes
have on the revised text" (153).

"Thus in revising the original text, the high-score writers chose to
reduce the number of sentence topics and to develop more fully those retained,
whereas the writers of the low-score texts chose to increase the number of
sentence topics and to develop each of them less fully" (153). 

~muddle~ "Differences between the two sets of revisions can also be
attributed to differences in the mean number of t-units per sentence
topic.  The low-score revisions averaged 1.89 t-units per sentence topic, while
the high-score revisions averaged 2.59, about 27% more than the low-score
ones" (150). ~muddle~

"The two groups’ differing constructions of the gist of the
original governed their choices of sentence topics.  These different
sentence topics, in turn, led to different decisions about content which could
be deleted from the original" (149).

"But on what basis did the two groups decide which elements of the
original text to delete? I suspect that they based such decisions on their
constructions of a discourse topic or a gist for the original
text, because those constructions seem to differ in important ways" (147).

"When what is said (by the principal verbs in the text) about the
discourse topic is combined with the discourse topic, the product is the ‘macroproposition,’
‘gist,’ or ‘point’ of the text" (140).

"As I have explained it, topical structure analysis would seem to be a
useful tool for studying the textual cues which may prompt revision and for
studying the effects of revision on text structure, primarily because it
accounts for and illuminates the interaction of reading and writing during the
revision process. Topical structure analysis should enable researchers to chart
more efficiently the actual decisions writers make as they revise texts"

"Such a view of the relationship of subtopics (i.e., sentence
topics) to the discourse topic surmounts the problem of using the
orthographic boundaries of sentences and paragraphs as the principal semantic or
meaning markers in extended discourse. (Sentence boundaries can vary
independently of meaning when writers choose to produce compound or
compound-complex sentences, and I can find no evidence that writers segment
texts into paragraphs in consistent ways.)" (137).