Interrupting visual feedback in writing. Perceptual and Motor Skills, 57

Perceptual and Motor Skills, 1983, 57,963-978.
@ Perceptual and Motor Skills 1983
INTERRUPTING VISUAL. FEEDBACK I N WRITING
GLYNDA A. HULL1 AND WILLIAM L. SMITH
University of Pittsburgh
Summary.-To test the effects of interrupting visual feedback on writing,
two groups of subjects-9 experienced and 9 inexperienced writers-were prevented from re-reading as they produced written texts. Their texts were then
analyzed for error, syntax, over-all quality, and sentence connectedness. While
both groups of writers were able to produce relatively well-formed sentences
when they could not re-read, both were nonetheless hindered at the level of
discourse production. Each group, however, reacted to that hinderance with
different linguistic strategies.
In recent years it has become commonplace to talk about the recursive
nature of the writing or composing process. Using case study evidence, Per1
( 1979) characterized composing as a kind of "retrospective structuring." On
the basis of their protocol research, Flower and Hayes (1980) have suggested
that composing consists of three major processes-planning,
translating, and
occur
recursively.
Sommers
(1978)
concluded, also
reviewing-which
can
from case studies, that revision is a recursive process which occurs throughout
composing. Researchers have noted that composing does n o t occur in a simple
linear progression. Rather, movement forward, whether to continue text production or to edit or plan, occurs in conjunction with movement backwards, to
read what one has written, and to reflect upon one's text.
If composing is a recursive activity, it seems reasonable to expect behavioral
evidence of that recursion, such as re-reading one's text. And in fact, researchers
have singled out particular behaviors of good writers which imply recursive
activity. Stallard (1974) noted that the "good" student writers in his study
stopped to re-read their texts more frequently than did the "average" student
writers. Similarly, Pianko ( 1979) found that her "traditional" college writers
re-scanned three times more often than her remedial group and concluded that
the reflective activity implied by such re-scanning is the parameter which
separates good and poor writers. Re-reading or re-scanning ( a term usually
synonymous with re-reading part of one's text) has been viewed both as evidence of the recursive nature of the writing process and as a characteristic
behavior of "good" writers.
Apart from this correlational evidence that good writers re-scan more,
'We thank Leonard Epstein for his comments on earlier drafts of this paper and for
suggesting our "invisible i n k methodology. W e also thank Deborah Arnowia and
Constance Wanner Ruzidr for their help with data analysis. W e particularly- appreciate
-Mrs. Ruzick's Tnsight and assistance in &vising a methodology for &dying sentence connectedness. Address all correspondence to Glynda A. Hull, Learning Research and
Development Center 606, University of Pittsburgh, Pittsburgh, PA 15260.
964
G. A. HULL
8r
W. L. SMITH
there have been several attempts to test the importance of re-scanning experimentally, by interrupting visual feedback; that is, researchers have taken away
writers' visual access to the texts that they write. Gould ( 1980) tested the importance of being able to re-scan when composing business letters by having
experienced adult writers compose with a wooden stylus on top of carbon
paper, which prevented their re-reading. H e found no differences in quality,
amount of time spent composing, or the number of proofreading changes between this "invisible" writing and either dictating or normal composing. His
findings, however, may be limited by the genre he chose. The possibility that
the business letter represents a common discourse schema which most writers
would have already internalized may explain why there were no significant differences berween invisible writing and normal composing.
Arwell (1981) asked college students-remedial
writers and average
writers-to
compose for 10 min. in each of two conditions, normal writing
with pen and paper and "blind" writing with worn-out ball point pens on top
of carbon paper. Analyzing discourse by means of propositional analysis, Atwell determined chat preventing visual feedback affected coherence, particularly
in the papers by remedial writers. However, because her subjects produced
only partial or very limited texts, her findings do not show the effects of such
a task on the production of a whole text.
There has been one attempt to investigate the role of recursion experimentally in children's writing. Teleman (1981) asked 12- to 13-yr.-old
Swedish schoolchildren to listen to a short fable and then re-tell it in writing, one
group using visible ink and the other, invisible. H e found it impossible to
distinguish one text from another, although texts and sentences were slightly
shorter in invisible ink. As Teleman points out, these results must be interpreted
in light of the fact that the children were provided both content and organization of text. This condition would, of course, decrease reliance on prior text
and thereby fail to provide an adequate test of the recursive activity which
might ,occur in a more complex writing task.
Based on the idea of recursiveness in writing, one would expect situations
which interrupt visual feedback to interfere greatly with composing, and writers
who are more dependent on negative feedback, what Sommers and others have
called dissonance, to experience greater disruption than writers who rely less
on recursion. It might further be expected, using the conclusions drawn by
Stallard and Pianko, that better (or more experienced) writers would depend
more upon visual feedback or make better use of it than poorer (or less experienced) writers. In fact, it is often claimed that poorer writers compose in a
linear fashion, which suggests that they do not take advantage of the feedback
available through re-scanning.
The purpose of the present research was to determine how intermpting
VISUAL FEEDBACK IN WRITING
965
visual feedback-by taking away visual access to what has been written-affected
the composition of whole texts by experienced, good writers and inexperienced,
poorer writers. W e wanted to determine the effects of interrupting feedback
on both the sentence and on discourse.
Sab jecu
Eighteen subjects participated in this study (two of the original 20 did not
complete all of the writing tasks). Nine of the subjects were inexperienced,
poorer writers who had been required to enroll in Basic Writing, a course designed for students who score below a specified level on a writing sample but
are not judged to be extremely deficient in writing skills. These students are
not, then, equivalent to the basic writers described by Shaughnessy (1977).
The nine experienced, good writers were teaching assistants who were expected
to write frequently and well as a part of their graduate studies.
This experiment was designed to compare writing in invisible ink and
regular ink for the two groups of writers. The design provided for comparison
both across types of writers (experienced, inexperienced) and within this
writer-classification (visible, invisible ink). The ideal design for the study
would be to counterbalance the order of presentation of essay, with half presented invisible ink first, and the other half, regular ink. However, because the
size of the sample was small, we felt it was impossible to control practice effects
by counterbalancing order. Consequently, we presented the invisible ink second
to minimize possible differences.
For each of the two writing tasks, the basic writers wrote at the same
time in a large conference room; the graduate students wrote on a different
day. W e gave subjects an hour to write their responses to the assigned topics,
and there was approximately a 1-wk. interval berween tasks. So that there
would be as little difference between the two writings as we could arrange, for
both tasks we supplied the students with fountain pens and with lined paper
which had been numbered on every other line. W e instructed them, under both
conditions, to write only on the numbered lines and only with the pens we had
given them. For the first essay, the fountain pens were filled with standard
blue ink. For the second, to prevent students from seeing what they had written, we filled the pens with invisible ink, a clear, colorless fluid which appears
only after the paper has been held close to a heat source, such as a candle or the
burner of a stove. Before the invisible ink task, we explained to the subjects
that they would not be able to see what they had written, and we instructed
them to use the numbered lines to help them keep their place on the page.
966
G. A. HULL
&
W. L. SMITH
The topics used in this experiment were designed to elicit persuasive
essays, essays in which a writer argues in support of a position. Research by
San Jose ( 1772), Perron (1976), and Crowhura (1780) has indicated that
this purpose in writing elicits the most syntactically complex prose and, we
hypothesized, should require more feedback from what has been written. Concomitantly, preventing access to one's text might alter that complexity. Furthermore, the persuasive purpose requires the conceptualization and sequencing
of ideas, but the ideas and their order are not pre-organized, as is often the
case with either narrative or descriptive writing. Finally, the persuasive purpose ought to require either extensive planning or extensive re-reading since
the ideas one presents must lead to a conclusion based on those ideas. Or, in
the case of an essay which moves from generalization to supporting evidence
(the pattern implied by the topics we chose), the persuasive purpose requires
a writer to refer to the generalization and the previously written text to maintain logic and coherence.
Since we could not use the same assignment for both tasks (students would
remember too much from the first essay when they wrote the second one only
a week later), we chose current events topics, assuming that the subjects would
have knowledge of both:
Topic 1 (used for "visible writing")
For some time now there has been a conuoversy over whether the United States
should compete in the 1980 Summer Olympics to be held in Moscow. Some people
feel that because Russia has invaded Afghanistan, American athletes should boycott the
games. Others, though, feel that politics should not enter the Olympics and that American athletes should be allowed to participate.
Write an essay in which you take a stand on the controversy and argue your position.
Topic 2 (used for "invisible writing")
Recently the issue of draft registration for women has received a great deal of
attention. Some people feel thac women should have to serve in the armed forces, while
others believe they should be exempt from service.
Write an essay in which you take a stand on this issue and argue your position.
W e pretested these topics before the experiment and found no differences in
the resulting essays that could be attributed to a topic variable.
Sentence-level Analyses
Syntax and fluency.-We analyzed each essay using the standard, global
measures of syntactic complexity, words per T-unit ( a main clause plus any
subordinate clauses) and words per clause. W e selected these metrics because
the existing research indicates that better persuasive writing typically has
greater T-unit and clause lengths, and also because both measures have been
shown to distinguish groups of writers who are several grades apart. W e
VISUAL FEEDBACK IN WRITING
967
expected, then, that constraining visual feedback might cause writers to reduce
words per T-unit, words per clause, and number of words.
that feedback interClause placement, type, and freqaency.-Assuming
ruption would have the greatest impact on the embedding of less-than-clausal
elements, we hypothesized that subjects might use more right-branching full
clauses when denied access to visual feedback. W e analyzed the types and
placement of clauses and the frequencies of each placement and type. These
clauses were classified as adverbials, relatives, or factives (including That-S,
Fact that, and That-S W H ) . Adverbials were further categorized according
to position within the sentence: initial, preceding the subject; final, following
the verb; or medial, following the subject but preceding the verb. W e classified
relative clauses according to two post-nominal positions, pre-verb and post-verb.
Factives were sub-divided according to "functional" position: use as the subject
or object, i.e., pre-verb or post-verb.
Error analysis.-We developed a taxonomy of all potential sentence level
errors that writers might make. This taxonomy was based on type of error
rather than cause of error, the kind of taxonomy proposed by Shaughnessy
(1977), because it seems impossible in many cases to determine reliably why
a writer makes an error-whether the mistake is a slip of the pen or whether
it indicates lack of knowledge about writing conventions. Also, the experimental condition of composing without being able to see what has been written
would further confound efforts to identify causation. Indeed, we intended to
utilize change in type of error or frequency as one broad index of the effect of
disrupting feedback, for one function of re-scanning might be to correct slips
of the pen.
Our error taxonomy included several general categories: spelling, punctuation, syntax (these categories were further sub-divided into particular types),
omissions/additions, and word-order mistakes. W e also had categories for
errors in verb and pronoun usage, and mistakes in what we called "concatenation" rules, failure to follow a lexical pattern established earlier in a sentence.
T o control the amount of writing each subject produced, we calculated
errors per various denominators. Since spelling errors occur on the lexical level,
we calculated them per number of words written. Because punctuation and pronoun mistakes, syntactic garbles, and incidences of omitced/added words occur on
the sentence level, we calculated these errors per T-unit. Because verb errors
and mistakes in concatenation occur on the clausal level, we calculated these per
clause.
Over-all Quality Analysis
The best available technique for determining the over-all quality of essays
is the holistic rating, for it does not isolate certain elements such as conformity
to editing conventions, but rather, views a whole as more than the sum of the
968
G. A. HULL
&
W.L. SMITH
parts (Cooper, 1977). In a typical holistic scoring session, raters read a set
of essays and assign each paper a score (ranging, for example, from a "1" for
poorest quality to a "4" for highest quality) based on their general impression
of the paper. In the present study, four raters were selected from a pool of
college teachers who had regularly rated freshmen essays using the holistic
scoring procedure, and had already been trained using the procedure. By
selecting raters who already were trained, we hoped to control for any inadvertent bias from their being trained by the researchers, a confounding variable documented by Freedman (1981).
Each of the four raters received a packet containing the essay topics and
all 36 essays, each of which had been typed and from which the author's name
had been removed. Raters did not know who had written the papers or that
some had been written in invisible ink. Raters were instructed to divide the
36 papers into four stacks of nine papers each, according to quality: the 9
papers which were best in over-all quality, the 9 which were second best, the
9 which were third best, and the 9 which were poorest. W e selected this
"forced choice" method of scoring because it would increase the possibility of
distinguishing within groups and between tasks. Had all essays been rated
without forced choice, we suspected that both essays written by the graduate
students might have received the highest rating; it would have been hypothetically possible for the 18 graduate student papers to have been rated "best" and
the 18 basic writer papers rated "worst." Our forced-choice method avoided
this problem.
To tabulate the ratings, we awarded a score of 4 to those papers a rater
considered best in over-all quality, a score of 3 to those rated second best, a score
of 2 to those rated third best, and a score of 1 to those rated poorest in over-all
quality. The ratings each essay received from the four raters were then summed,
providing a 4 to 16 scale.
Sentence Connectedness
Since it has been suggested chat writers re-read what they have written to
make connections between what has been written and what they are about to
write, we included among our analyses a measure of sentence connectedness.
But instead of measuring connectedness through conventional methods of
counting and categorizing what Halliday and Hasan (1976) have called "cohesive ties" (or the referents such as repeated words which link a sentence to
what has come before i t ) , we followed McCutchen and Perfetti (1982) in taking from the literature on speech comprehension a concept called the "givennew contract" and applying that concept to connectedness in writing. [See
Witte and Faigley (1981), for a review of the problems associated with the
use of Halliday and Hasan's taxonomy.]
Clark and Haviland (1977) have described how a speaker, in constructing
969
VISUAL FEEDBACK IN WRITING
a sentence, conveys two kinds of information: the given, or that which he believes his listener alceady knows, and the new, or that which he believes his
listener doesn't know. Clark and Haviland have also explained how a listener
uses the given-new strategy to comprehend what he hears. Vande Kopple
( 1982) demonstrated that paragraphs which facilitate the given-new strategy
are easier to understand and remember than are paragraphs which frustrate the
strategy, and so extended application to larger units of discourse. Odoroff
(1952) first suggested that the principle of given-new might prove a good
means of measuring cohesion in a written text, and McCutchen and Perfetti
(1982) devised a method for analyzing the c o ~ e c t e d n e s sof children's writing
via the given-new contract. Our own methodology is an adaptation of McCutchen and Perfetti's.
W e analyzed connectedness by means of the given-new contract first by
dividing each text into T-units ( a main clause plus any subordinate clauses) and
then each T-unit into given-new segments. W e then analyzed each given segment to determine whether its connection to a previous T-unit was local (that
is, it referred to the immediately preceding T-unit) or remote (that is, it referred to a previous T-unit other than the one immediately preceding). Although we did not count and label cohesive ties in the manner of Halliday and
Hasan--our intent was to find a more global and less tedious method for noting
sentence c o n n e c t i o s t h e given segments always contain at least one of Halliday and Hasan's "cohesive ties."
The following excerpt from an essay written by a graduate student ( i n
regular ink) illustrates our coding procedure. The double slash marks indicate
the end of a T-unit; the solid underlining, given information; and the broken
underlining, new information. Local and remote ties are shown by the superscripts L and R, respectively. It will be apparent from this excerpt that given
information often corresponds to the simple subject of a sentence and new
information to the complete predicate, though, we should note, this is not
always the case.
1 A recent Doonesbury comic strip depicted a conversation between an American
L
and a Russian.//
- - -
2
These two men, who were presumably lower echelon diplomats,
- - - - - - - - - L
were discussing the Russian invasion of Afghanistan.//
- - - - - - - - - -
3 The American questioned
the
- - L
Russian as to the USSR's motives for this action.// 4 The Russian, hesitant to speak,
fearing that he might be overheard, whispered to the American that he would tell him
- - - - - - - - - - - - - - - - - the reason if the American promised "not to tell anyone."// 5 When the American
- - - - - - - - - - - - - - - - - L
readily agreed to this condition, the Russian then leaned over and said in a low voice,
- - - - - - - - - - - - - -
970
G. A. HULL
&
W. L. SMITH
R
"We want to rule the world."// 6 Gary Trudeau's tongue in cheek treatment of the
- - - - - - Soviet's imperialistic tendencies reflects an anxiety many Americans, including myself,
- - - - - have about the Russians.//
- - - - - -
The first T-unit in the text is considered new, since nothing precedes it.
In T-unit 2, the phrase "these two men" represents given information, for it
refers to "an American and a Russian" in the immediately preceding T-unit.
Because "these two men" refers to a prior adjacent sentence, we consider this
tie "local." In T-unit 3, "The American" connects both to "These two men"
in T-unit 2 and "an American" in T-unit 1. However, we consider it a local
tie because it does connect to an adjacent sentence. T-unit 5 illustrates how we
segment T-units beginning with adverbial clauses. If an initial clause can be
moved to another part of the sentence, we consider that clause new, and look
for given information in the main clause which then begins the sentence. ( A n
exception is "if, then" clauses, where the "if" clause obviously contains given
information.) In T-unit 6, the given information refers, not to the preceding
adjacent T-unit, but to " Arecent Doonesbury comic strip depicted" from T-unit
1, or perhaps, to the entire example as developed by all the preceding T-units.
W e label it a remote tie.
Statistical Analysis
In all cases, we tested for statistical significance by means of a two-way
analysis of variance, repeated measures on one factor, with level of experience
as the between-groups factor and type of ink the within-group factor. W e used
P.05 as the criterion level for significant differences. Means and standard deviations or standard errors are also presented.
RESULTS
Fluency
The results of the analysis for fluency showed that both basic writers and
graduate students wrote fewer words in the invisible ink task. Graduate students
averaged 518.11 words in regular ink and 457.78 in invisible ink; basic writers
averaged 249.56 words in regular ink, 238.78 in invisible. However, these
decreases were nonsignificant. Disrupting visual feedback does not seem to
influence fluency. Graduate students did write significantly more words under
both conditions than did basic writers ( p < .001).
Syntax
The global syntactic analyses indicated a significant difference between
essays written under normal conditions and those written when feedback was
disrupted in number of words per clause: both basic writers and graduate
students significantly decreased words per clause when feedback was disrupted
( p < .001). Words per T-unit was not changed significantly in either group;
VISUAL FEEDBACK IN WRITING
97 1
however, graduate students wrote longer T-units under both conditions ( 9
.025 ) . Means and standard errors ate present in Fig. 1.
<
FIG. 1. Mean words per T-unit and
words per clause for each group on
each task with standard error
LXPERIENCED
WRITERS
WRITERS
0
REGULAR
INK
INVISIBLE
INK
REGULAR
INK
INVISIBLE
INK
Placement, Type and Freqzrency of Clauses
Our analyses of types, placement and frequency of clauses showed no significant differences between the two writing tasks for either group. (See
Table 1.) Seventy-five percent of all clauses used by graduate students were
post-verb for the tasks under normal conditions, as were 76% when feedback
was disrupted. For basic writers, 72% of all clauses were post-verb on the
normal task, and 73% when feedback was disrupted.
TABLE 1
Group
Adjective
Initial Final
Inexperienced writers
.78
Regular ink
Invisible ink
1.00
Experienced writers
1.67
Regular ink
Invisible ink
1.22
Adverb
Initial Final
1.44
Factive
Initial Final
All Clauses
Initial Final
1.87
2.00
2.11
2.33
.11
0.00
4.67
5.67
3.11
3.56
8.00
1.78
4.22
2.78
4.78
4.11
4.11
5.00
.44
.44
12.67
10.87
7.00
5.78
21.00
18.67
9.44
G. A. HULL
972
&
W. L. SMITH
Error
The analyses of error per various denominators showed that neither graduate students nor basic writers altered their error rate or error type when feedback was disrupted. W e also calculated the ratios of total errors per word
and total errors per T-unit, and there was no significant difference between the
tasks for either group on either measure. Graduate students made fewer errors
whether they wrote in visible or invisible ink. Means and standard deviations
for the error categories are presented in Table 2. Some categories have been
merged because either no errors occurred in a category or the number of subjects who made the error was too small (often, just one) to be insightful.
TABLE 2
MEANERROR
PROPORTTONS FOR EACH GROUPON EACH TASK
Group
Spelling
~ e r
dord
M SD
Inexperienced writers
,024 .024
Regular ink
Invisible ink
,020 .014
Experienced writers
Regular ink
.007 .006
Invisible ink
,004 .003
Note.-The "others" category
ornitted/added words.
Punctuation
Per
T-unit
Verb plus
concatenation
Per clause
Others
oer
T->nit
All error
oer
Gord
All error
Der
~Iunit
M SD
350 .315 ,077 .077 ,187 .I54 .065 .042 1.113 .SO6
.389 ,212
,004 .031
,170 ,113
,060 .030
.874 .417
.I42 ,059 .013 .011 .046 ,039 .020 .007 .339 .I13
.I47 ,132 ,011 .014 ,105 ,101 ,017 .007 ,331 ,163
is composed of errors in pronoun, syntax, word order, and
Quality
The results of the holistic scoring showed consistent patterns of decrease
in quality from normal conditions to disruptive feedback for both groups ( P <
.05). Under normal conditions, graduate students wrote papers judged better
than those by basic writers, as were their papers when feedback was disrupted
( p < .001). The means and standard errors for the ratings of quality are
presented in Fig. 2. This figure not only shows the difference between the
groups, but more importantly, shows the parallel decline between tasks for
both groups.
Sentence Connectedness
When writing in invisible ink, both inexperienced and experienced writers
produced significantly different percentages of local ties ( p < .01) and remote
ties ( p < .025) compared to those in their productions of visible ink. However, the percentages changed in opposite directions for the groups; that is,
while experienced writers used a greater percentage of local ties in regular ink
and fewer in invisible ink, inexperienced writers did the opposite, using a
VISUAL FEEDBACK IN WRITING
973
lesser percentage of local ties in visible ink and a greater percentage in invisible.
The pattern was reversed for remote ties. Here graduate students significantly
increased their percentage in invisible ink, while basic writers decreased their
percentage. These interactions are presented in Fig. 3, along with standard
errors.
fXPfRIfNCfD
WRITERS
FIG. 2. Mean quality rating for each
group o n each task, with standard
error
INK
INK--
Drscussro~
W e predicted that interrupting feedback would interfere with composing,
and we can infer that it did interfere: for both groups of subjects, essays written
in invisible ink had shorter clause lengths, were of lower over-all quality, and
exhibited different patterns of connectedness than did the "normally" composed
e s s a y s a n d this despite the fact that all subjects wrote their visible ink essays
first and could thereby be said to have had some practice in the experimental
situation. The groups responded similarly on two of the significant measures:
both graduate students and basic writers wrote shorter clauses of approximately
the same length in invisible ink, and both groups showed a parallel decline in
the quality of their essays from visible to invisible ink. It was only in the
analysis of sentence connectedness that we found evidence of one group responding differently from the other as an effect of writing in invisible ink.
The only significant difference on the sentence level, the reduction in
clause length, may have been a result of incapacitation of memory. When subjects could not see what they had written, they may have reacted by decreasing
memory load-putting fewer words or perhaps chunks of information into the
basic perceptual unit, the clause, as might be predicted by Daiute's (1981)
G. A. HULL
INK
INK
& W.
L. SMITH
INK
INK
FIG. 3. Mean percentage of remote ties and mean percentage of local ties, with
inexperienced and 0-0
experienced writers
standard errors by 9-0
work on memory constraints and composing. Words per clause can, however,
decrease without a concomitant decrease in words per T-unit for other reasons
besides fewer words being written in each clause. Additional subordinate
clauses and new main clauses could produce this effect, as could writing as full
clauses what normally would be written as less than clausal structures. Since
our second task was not a re-write of the first, we cannot say for sure which
technique was used.
Whatever the explanation for the reduced clause length, this reduction
cannot be used to explain the lower holistic ratings of the essays written when
feedback was interrupted, for as Faigley (1980) has shown, words per T-unit
and words per clause account for only three percent of the total variance in
holistic evaluations. Nor can sentence-level error or clause placement be
blamed, for frequencies and types remained similar across tasks. Since there
were negligible differences on the sentence level between the writing done in
invisible and visible ink, we would suggest that the lower holistic ratings of
the essays when feedback was disrupted indicate that students lost control of
their writing beyond the sentence and beyond the paragraph. Our analysis of
sentence connectedness suggests that they may have lost control by being forced
to alter their patterns of relating new information to previously given information.
Graduate students reacted to lack of visual feedback by relating new'hformation to previously given information through an increased percentage of
975
VISUAL FEEDBACK IN WRITING
remote ties, and a decreased percentage of local ties. Perhaps this strategy
resulted from the fact that they could not re-read previous sentences but presumably could recall the gist of their prior assertions. If a writer lost track of
the wording of a prior sentence, he could nonetheless begin a new sentence about
the general topic of his paper. Such a strategy would not, of course, necessarily
result in a well-formed text, for a sudden return to a general topic or a different
subpoint might give the impression of disjointedness.
In contrast, basic writers reacted by increasing their percentages of local
ties and decreasing their percentages of remote ties when writing in invisible
ink. Since better writers have been shown to use more local ties (Witte &
Faigley, 1981), this strategy might be judged a beneficial one, resulting in
greater connectedness between sentences. On further examination, however, we
noticed that the increased percentage of local ties in basic writers' invisible ink
papers had instead resulted in sentence connections that were strained and inappropriate. For instance, in the following invisible-ink paragraph by a basic
writer, the T-units are connected, in the main, by local ties to the word women,
yet these local ties are superficial joinings of repetitive information.
1 I feel that wome should serve time in the armed forces but not in combat or
----------------combat training but maybe in hospitals or desk jobs.//
- - - - - - - - - - - 2-Ever
-since
- the
-constitution
-L
L
was written up, women have argued for equal rights,// 3 and equally so I think they
- - -- - - - - - - - - - should have equal rights.// 4 Now that they are getting closer to being equal to men
----------------L
and have about the same rights as men, then they should have to be required to register
- - - - - - - - - - - - - - - - - L
L
also.// 5 If they want equal rights, then equal rights we'll give them.// G If women
- - - - - - - - - - - - - - - -are not required to register for the draft the I as a male will have to start campaigning
- - - - - - - - - - - - - - - - - L
for equal rights also,// 7 How can men and women
- - - - -war but women don't because their feminine.//
- - --- - -- -
be equal when men have to go to
- - - - - - -
Perhaps this use of local ties is a sign of these writers' increased attention
to sentence-by-sentence production rather than the planning of the whole text.
If this reasoning is accurate, then we might also speculate like Atwell (1981)
that inexperienced writers have fewer strategies for the production of a whole
text than more experienced writers. W e cannot, of course, be certain that the
altered pattern of given-new information was the sole factor in reduced ratings
of quality for both groups or, indeed, whether this change significantly affected
quality at all.
Both groups of subjects wrote essays under both conditions which did
not differ significantly in total amount of text produced, words per T-unit,
types and frequencies of errors, and types, placement, and frequencies of clauses.
976
G. A. HULL
&
W. L. SMITH
This sentence-level constancy may have been the result of how we designed the
study. That is, subjects may have received enough "practice" on the visible-ink
task so that their sentence-level production was not affected by the invisible-ink
condition. This interpretation, however, is unlikely for several reasons: our
subjects did not know what their second task would be and could not "practice"
for it consciously. Even if our subjects had known during the first trial that their
next task would be to write an essay in invisible ink, it seems unlikely that they
could have consciously practiced controlling sentence level error or manipulating
syntax. Any unconscious practice effect resulting from the first task would
seem more likely to affect the over-all planning of the text than sentence production, yet, as we pointed out earlier, subjects performed less well on the
invisible-ink task in over-all quality and sentence connectedness.
A more likely. explanation
for the sentence-level constancy is that both
the inexperienced writers and experienced writers have an internalized written
style-an
habitual syntax p e r h a p s w h i c h they produce automatically in a
timed-writing situation in response to a given topic. The possibility that these
writers could produce such a style when feedback was interrupted perhaps
suggesrs that part of learning to be a fluent writer is developing a memory for
one's own written syntax. Or as Frawley (1981) suggests, the syntax of unsophisticated writers may conform to "natural" syntactic tendencies which apply
across languages, particularly as evidenced by these writers' habits of using
relativization and sentential complementation.
For another explanation, one can turn to Ktashen (1981), who has
hypothesized that second language learners use what he calls "acquired" linguistic knowledge in ordinary unmonitored speech, but that they can apply
consciously learned rules only when there is sufficient processing time to monitor their speech. The subjects in our study, then, may have relied on an
acquired linguistic knowledge in both tasks. Differences between tasks on the
sentence level would appear only in revision, when subjects would have time
to apply consciously learned knowledge about written language.
We infer from our data that the significance of visual feedback lies not
in editing or manipulating syntax, but in managing larger units of discourse.
This inference receives support from research by Flower and Hayes ( 1981),
which suggests thac pauses before "episodes" in think-aloud protocols denote
god-related activities, activities which go beyond sentence-level planning. If
the significance of visual feedback lies in discourse planning, one might also
be tempted to infer thac sentence-level production is "non-recursive" or can
proceed independently of feedback-at least in a timed-writing simation. W e
think, however, that such an inference would be hasty, for in this research we
blocked only one feedback mechanism-visual
re-scanning.
The fact that having writers compose in invisible ink disrupted the feed-
VISUAL FEEDBACK I N WRITING
977
back loop which governs composing only partially should, instead, serve to-remind us of how complex the composing feedback system is and how interrelated
its components must be. Emig ( 1978) makes a similar point when she discusses the "organic" structures involved in composing-eye, brain, and hand. In
isolating a component like re-scanning, we ought not, then, to forget that others
complement its function. Indeed, there may be such redundancy among feedback components that inhibiting one does not inhibit the whole. While rescanning allows a writer to compose recursively, other activities, perhaps unobservable or observable but ambiguous ones, contribute to recursion as wellboth on the level of the sentence and the level of discourse.
REFERENCES
ATWELL,M. The evolution of text: the interrelationship of reading and writing in the
composing process. Unpublished doctoral dissertation, Indiana Univer., 1981.
CLARK,H. H., & HAVILAND,S. E. Comprehension and the given-new contract. In R.
0. Freedle (Ed.), Discourse fioductwn and compehenswn. Norwood, N J :
Ablex, 1977. Pp. 1-40.
COOPER,C. R. Holistic evaluation of writing. In C . R. Cooper & L. Odell (Eds.),
Evaluding writing: describing, measuring, judging. Urbana, IL: NCTE, 1977.
Pp. 3-31.
CROWHURST.M. Syntactic complexity and teachers' quality ratings of narrations and
arguments. Research i n the Teaching o f English, 1980, 14, 223-231.
D A I U ~C., A. Psycholinguistic foundations of the writing process. Research in the
Teaching of English, 1981, 15, 5-22.
In C. R. Cooper & L.
EMIG,J. Hand, eye, brain: some "basics" in the writing
Odell (Eds.), Research on composing: points o f %p',",:;ei.
Urbana, IL: NCTE,
1978. Pp. 59-71.
FAIGLEY,L. Names in search of a concept: maturity, fluency, complexity, and growth
in written syntax. College Conrposdwn and Communicrtion, 1980, 31, 291-299.
FLOWER,L. S., & HAYES,J. R. Identifying the organization of writing processes. In
L. W. Gregg & E. R. Steinberg (Eds.), Cognitive fiocesses in writing. Hillsdale,
N J : Erlbsum, 1980. Pp. 3-30.
FLOWER, L., & HAYES,J. R. The pregnant pause: an inquiry into the nature of planning.
Research in the Teaching of English, 1981, 15, 229-243.
FRAWLEY,W. Universal grammar and composition: relativization, complementation,
and quantification. In W. Frawley (Ed.), Linguistics and literacy. New York:
Plenum, 1982. Pp. 65-90.
FREEDMAN,S. Influences on evaluators of expository essays: beyond the text. Research
in the Teaching o f English, 1981, 15, 245-255.
G~ULD
J. , D. Experiments on composing letters: some facts, some myths, and some
obsefvations. In L. W. Gregg & E. R. Steinberg (Eds.), Cogtzicive processes in
wrztzng. Hillsdale, NJ : Erlbaum, 1980. Pp. 97-127.
HALLIDAY,
M. A. K..& HASAN,R. Cohesion in English. London: Longman, 1976.
KRASHEN,
S. Secolzd language acquisition and second language learning. Oxford:
Pergamon, 1981.
MCCUTCHEN,D., & PERFETTI,C. Coherence and connectedness in the development of
discourse production. Text, 1982, 2, 113-139.
ODOROFF,E. Sentence connections and thinking. Paper presented at the 33rd Annual
Meeting of the Conference on College Composition and Communication, San
Francisco, March 1982.
PERL, S. The composing processes of unskilled college writers. Research in the Teaching of English, 1979, 13, 317-336.
978
G. A. HULL
&
W. L SMITH
D. The impact of mode on written syntactic complexity: Parts 1-111. Third,
fourth, and fifth grades. fmdies in Language Education, Report Nos. 24, 25,
and 27, Dept. of Language Education, U n ~ v e r s ~ofq Georgia, Athens, GA, 1976.
PIANKO, S. A description of the composing processes of college freshmen writers.
Research in the Teaching of English, 1979, 13, 5-22.
SAN JOSE, C. Grammatical structures in four modes of writing at fourth grade level.
Unpublished doctoral disserration, Syracuse Univer., 1972.
SHAUGHNESSY,
M. Ewors and expectations. New York: Oxford, 1977.
SOMMERS,N. I. Revision in the composing process: a case study of experienced writers
and student writers. Unpublished doctoral dissertation, Boston Univer., 1978.
STALLARD,
C. K An analysis of the writing behavior of good student writers. Research
in the Teaching of English. 1974, 8, 206-218.
TELEMAN, U. On visual feedback in writing. Proceedings I, Sixth International Congress o f Applied Linguistics, Lund, 1981, 424-425. (Abstract)
VANDE KOPPLE, W. J. The given-new strategy of comprehension and some natural
expository paragraphs. J o u m l of Psycholinguistic Research, 1982, 11, 501-520.
WITTE,S. P.,& FAIGLBY, L. Coherence, cohesion, and writing quality. College Composition and Communication, 1981, 32, 189-204.
PERRON,J.
Accepted October 4, I983.