THE INTEGRATION OF VERBAL AND VISUAL INFORMATION
IN TECHNICAL CpMMUNICATION
B.A., University of Colorado, 1972
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Science in Technical Communication
Department of English
This thesis for the Master of Science
in Technical Communication degree by
has been approved for the
Department of English
Terry, Carol (M.S., Technical Communication)
The Integration of Verbal and Visual Information in
Thesis directed by Assistant Professor Colleen Donnelly
We are always coordinating information from a
variety of sources. Technical writers can organize
visual and verbal information in documents to enhance
the users ability to comprehend and store information.
The existing research about verbal and'visual
information processing focuses on each system
separately. Since the information we receive verbally
and visually should provide us with coherent knowledge,
the function of one system needs to be examined in
relationship to the other. Then we can determine
whether certain kinds of information are more conducive
to one system or another.
Coordinating verbal and visual information can
involve making models. Experts and writers, among
others, use models to make information manageable.
Writers use models to make the function of a sentence
or paragraph explicit by placing it in relationship
with other sentences or paragraphs. When writers use
outlines, they are making information hierarchical.
Experts use models when they "chunk" like types of
information together and use a label to "represent"
These strategies coordinate verbal and visual
information, making it easier for users to retrieve
select pieces from long-term memory.
To make interactions between the verbal and
visual systems more explicit, I asked two sets of
volunteers to read different versions of a document. I
asked them to write down what they remembered about the
text both immediately after reading it and after one
week had elapsed. Volunteers who read the document
containing visual model-building cues performed both
immediate and long-term recall tasks better than did
users of the document which lacked visual cues to its
Questions point toward further exploration.
What is the optimum level of providing models without
limiting the users participation in the text? Is
participation dependent on the users familiarity with
the text? When in the comprehension process is the
expert user engaged in the model-making process?
A new look at the research and conducting an
experiment show that while models can successfully
integrate verbal and visual information, their
usefulness might be dependent on the user's familiarity
with the material in the text.
The form and content of
recommend its publicati
abstract are approved,
because of you, Andrew
Although I have never felt more alone than when I
was doing this thesis, there were many who helped me. I
couldnt have done it otherwise and I need to thank
Colleen was a careful and patient editor; she
offered the many suggestions which allowed me to
understand what I really meant.
Karen, Marjorie, Pat, Paul and Eileen were always
willing to take Andrew so he could have a good time
while I suffered.
Everyone at RTD, without exception, gave me
continual support. Mare was always flexible when I
asked for time off. Samer knew exactly what I was going
through. Karen made me laugh and cheerfully lent me her
classes. Mary Davis, who first hired me, and Dick
Bauman, who hired me again, supported my project
through RTDs intern program; it was an excellent
opportunity for me to integrate theory with practice.
And Cindy provided me with comfort, support,
babysitting and her blinding analytical skills whenever
II. PROCESSING TEXT.........................
Chunking Verbal Information
in Short-term Memory.............. . .
Storing Verbal Information
in Long-term Memory..................
A Cognitive Model of Text Comprehension
to Link New Information..............
III. USING VISUAL INFORMATION
TO OVERCOME COGNITIVE CONSTRAINTS
Processing Visual Information
Contraints on Visual Information
Using Visual Strategies to Recode
IV. VISUAL ASPECTS OF THE WAY
EXPERTS SOLVE PROBLEMS.........
How Representation Reduces
the Load on Memory.........
Creating Situational Models
V. INTEGRATING VERBAL AND VISUAL INFORMATION
IN THE WRITING PROCESS.............................41
Instantiating the Model..........................41
Seeing and Verifying
Relationships in the Hierarchy.................46
VI. EXPERIMENT: HOW CHUNKED DOCUMENTS
CAN PROVIDE A CONTEXT FOR NOVICES.................51
Problem: Creating a Context......................51
Method: Exposure to Unfamiliar Material.........53
Observations: Forming Gists......................54
Conclusions: Is What You See What You Get?.......62
VII. INTEGRATING VERBAL AND VISUAL INFORMATION
IN THE WRITING PRODUCT........................ 65
Making Sections Modular..........................65
Clearly Specifiying Agency.......................66
How Using these Strategies
A. Overview of the Disadvantaged Business
Enterprises (DBE) Program at RTD .....................77
B. OLD version of Attachment A........................78
C. NEW version of Attachment A........................82
D. Test materials and representative responses........89
Table 6.1 Types of Immediate Recall Answers:
OLD version........................ 57
Table 6.2 Types of Long-term Recall Answers:
OLD Version......................... 58
Table 6.3 Types of Immediate Recall Answers:
Table 6.4 Types of Long-term Recall Answers:
An array gives information a spatial aspect........44
An array allows vertical or horizontal access......44
NEW version users found information
about three times faster than OLD
Although we are always using the visual system,
we arent generally aware of processing visual
information. When we see a stop sign, for instance, we
don't need to read the word 'stop to know what to do.
We see the shape and process the information visually
rather than verbally.
Visual information has a universal quality. We
all know what a stop sign is and, given a cue, would
produce very similar mental representations of a stop
sign. Even though these mental pictures are, at some
level, unverifiable, the fact is verified in action:
we all stop.
Since other people cant see the visual
information in our minds, we need to translate it into
words to describe it to them. For example, when we
give directions to a lost friend, our task is to
somehow translate spatial information about three
dimensions into a two-dimensional visual picture, a
map, or into sequential verbal instructions.
Researchers have found that users of a map form a
better "situational model" or mental map than users of
the verbal instructions (Perrig and Kintsch 1985). To
give verbal instructions, we have to first generate an
image of the roads and landmarks our friend may see on
the way to his destination, and then transform that
visual, spatial knowledge into "linear form" because we
can only mention one place at a time (Linde and Labov
1975, 927). Once we have generated the route in our
minds, we ask our lost friend if he knows where the
white church is or the fire station is. To determine
common landmarks, we check, by asking questions, that
our friend is generating the same spatial information
(visual landscape) as we are. The process is something
like this: we describe the scene; he comprehends the
words; he generates an image; he describes it in his
words to verify that he is indeed "seeing" the same
church we are describing; then we confirm, saying yes,
its the white church with the stone fence.
If we give him a map, the process is shortened
considerably. But once our friend finds his
destination, the map becomes unnecessary. Once the
instructions are integrated, they become disposable.
Once we know what a picture means or represents to us,
we dont have to redefine or reidentify the symbol
every time we see it.
We are constantly integrating visual and verbal
information in everyday life. And although we arent
always aware of it, we integrate verbal and visual
information when we read text. Certain kinds of text
tend to evoke more visual images than others.
Successful novels are those which stimulate our
memories of sights, sounds, smells, emotions, and
feelings. The purpose of evocative, literary discourse
is to give us something personal, yet universal, which
we can connect to and which will continue to resonate
in our minds.
Technical communication is non-literary text
which has a purpose; it may give instructions to elicit
a performance or evoke a behavior. Research in
advertising and in task performance adds support to the
evidence that writing which exploits our visual system
improves performance (Maclnnis and Price 1987;
Marschark 1985; Perrig and Kintsch 1985).
This thesis examines the visual system as a
significant element in the process of comprehending and
retrieving the information in technical documents.
After providing a model showing how the verbal system
functions, I will explore some of the research about
the visual system in order to relate it to the
comprehension process. This perspective allows the
possibility of consciously integrating the two types of
processing systems to improve writing products.
I also devised an experiment using a specific
text which I revised utilizing elements to engage the
visual system. This experiment shows that visual
information can underwrite verbal information and
improve the comprehension as well as the storage and
retrieval of complex information.
When we write useful text, we want the reader
to create and to be able to easily access his knowledge
structures stored in long-term memory. By using the
visual appearance of the text to provide cues to the
structure, to the concepts, and to the hierarchy of the
text, the writer can alleviate the cognitive burden on
the readers short-term memory, allowing more cognitive
resources to be devoted to comprehending the text
itself. Integrating verbal and visual information is a
way of making the exchanges between short-term and
long-term memory more efficient and more effective.
Comprehending text is an interactive process;
several levels of interaction occur between the reader
and the text. As we read, we apprehend a unit of
meaning, the clause or proposition, and then we read
on. However, to remember the information in the text,
we need to represent it to ourselves, usually in non-
Chunking Verbal Information in Short-term Memory
Although the comprehension of text seems to us
a singular activity, it actually involves several
subactivities. Generally, when we have apprehended a
unit or chunk of meaning (usually a clause), we go on
to the next unit. Closure is our cognitive "signal"
that we are certain about the meaning of a unit and can
go on to the next unit. According to Hirsch, closure
occurs at the end of the shortest "resolvable" unit of
meaning that a reader can hold in short-term memory.
It does not matter whether the items to be
remembered are physical objects, binary digits,
decimal digits, letters of the alphabet, letters
plus digits, or monosyllabic words; in all these
cases, people begin to make mistakes of memory
when the number of items exceeds six or seven.
If, in reading, we encounter a series of units
which climbs above seven before being resolved in
a definite pattern, we will not normally remember
all the units held in suspension. We will have to
go back and check up (Hirsch 1977, 67).
When we can progress through the text without having to
stop frequently and reread, we say that the text is
readable. Rapid, stable closure reinforces the
"singular" sense of comprehension; closure underwrites
Yet, comprehension involves subtasks such as
integrating information in our short-term and long-term
memory. The well-known limit on what we can' hold in
short-term memory is linked to the size of the chunk of
text and the speed of closure (the time it takes to be
certain of its meaning). When we comprehend a unit of
meaning, it not only enters into our short-term memory
but also has to be related to something else: the
previous sentence, or the topic of the text, or
information in our long-term memory or, perhaps, all of
Storing Verbal Information in Long-term Memory
Generally, what we read is first processed in
short-term memory, where closure is achieved. Some
time after that, new information is stored in long-term
memory. James Miller defines short-term or working
memory as "a store that allows rapid access to its
contents and long term memory as a static repository in
which information can be stored or retrieved but not
actively processed..." (Miller 1984, 332).
We store information in long-term memory in a
different form than the original text. Researchers
have speculated that we store verbal information in a
reduced form, called a "gist." We have other ways of
transforming or structuring new information for storage
in long-term memory. We use frames, schemas,
hierarchies and images. In general, we store a
representation of the information we are comprehending,
something that stands for that information (Flower
What is crucial in comprehension as well as the
retrieval of information for some purpose, is that
short-term and long-term memory interact. We need to
retrieve what we already know (given information) in
order to make sense of or use the text we are reading
A Cognitive Model of Text Comprehension
Walter Kintsch, a cognitive psychologist, has
developed a fairly plausible model of text
comprehension. Plausibility in this area is a relative
term because we dont have so-called objective access
to mental images. We make deductions and inferences
from test results, and depend on what subjects tell us
about their mental images and representations.
Although these inferences are supported by other areas
of research, we cant (yet) prove the existence of
schemas or representations in a persons long-term
memory. We can, however, judge how knowledge
structures function during problem-solving tasks.
The main component of Kintschs model is the
proposition; taken together, propositions represent the
meaning of the text being processed. Propositions are
the smallest unit from which meaning can be derived;
they are composed of concepts and "must include first a
predicate, or relational concept, and one or more
arguments. The latter may be concepts or other embedded
propositions" (Kintsch 1978, 367).
According to Kintschs model, when we first
approach the text ("Cycle 1"), our short-term memory
(the "buffer") is empty. As we read the text, we
maintain a proposition in the short-term buffer to make
sure it is related to the next proposition. When we
find propositions to be coherent and an "overlapping
relationship with the previous proposition is
established," we continue to the next proposition in a
"fully automatic" way (Kintsch 1978, 372).
Kintsch defines the text base as a coherent
structured list of propositions (Kintsch 1978, 365).
The relationship between propositions is "probably the
most important single criterion for the coherence of
text bases" (Kintsch 1978, 367). Since coherent text
bases are constructed in cycles, propositions are
"checked" for an overlapping relationship with the
Processing text is not the same as storing the
information. Another element of Kintschs model
describes the process by which readers condense text to
its gist in order to store it in long-term memory.
This process, guided by macrorules, deals with the same
facts contained in the propositions but "from a more
global point of view" (Kintsch 1978, 366). Macrorules
help transform the local meaning of [micro]propositions
to the more global meaning of macropropositions. It is
unclear exactly when this transformation occurs. It
could happen during comprehension, or during
verification (when propositions are "checked" for
coherence), or it could happen after text is processed
completely. In any case, it is the schema, or formal
representation of our goals (in long-term memory), that
determines how we reduce text and what information will
be reduced. Theories of schema support this role:
The goal of schema theory is to specify the
interface between the reader and the text to
specify how the readers knowledge interacts with
and shapes the information on the page and to
specify how that knowledge must be organized to
support the interaction (Adams and Collins 1979,
When we read for a purpose, we are engaged in a
process of transforming "the propositions of the text-
base into a set of macroproppsitions that represent the
gist of the text" (Kintsch 1978, 372). During this
process, the text in the buffer interacts with the
knowledge structures in long-term memory. Since "world
knowledge" found in long-term memory is required for
the "operation" of macrorules, the gists and
macropropositions readers form will depend on their
familiarity with the text, level of reading skill and
their goals (Kintsch 1978, 366). In other words, what
readers already know will dictate what they find
relevant; what is in the readers long-term memory will
interact with what is in his buffer.
Macropropositions, then, conceptually weave
propositions together to form the high level of the
text-base hierarchy. "It has been observed that
propositions belonging to high levels of a text-base
hierarchy are much better recalled (by a factor of two
or three) than propositions low in the hierarchy"
(Kintsch 1978, 370). Even more crucial is the ability
of global information (gist, theme, or
macroproposition) to establish context, a meaningful
whole for the text-base. Kintsch says:
Relating propositions in a local manner is not
sufficient. There must be a global constraint
that establishes a meaningful whole characterized
in terms of a discourse topic (Kintsch 1978, 366).
Creating a gist from a text is a way of
encoding information to overcome short-term memory
limits. Encoding reduces and stores information that
we find relevant into long-term memory. We need to
organize it so that we can use, or retrieve, the
information later. Linda Flower says:
The underlying fact is that people create
internal, mental representations of new
information in order to remember it and think with
it, and these inner representations can be both
verbal and non-verbal (Flower 1985, 51).
We receive information about the external world
through several systems: visual, verbal, kinesthetic,
aural, emotional. Information in one system can relate
to or connect to information in another system,
depending on the task or situation. An example of this
"cross-referencing" between systems occurs when we hear
a song that was popular when we were younger and it
arouses the same emotion we felt at that time. The
music we hear "anchors" a feeling, then can act as a
Paivio, a proponent of dual coding theory,
illustrates how information in the.visual system can
link to information in the verbal system:
a given word can evoke any number of images
corresponding to different exemplars of a referent
class (e.g., different tables) or different
versions of a particular class member (e.g., my
dining room table imaged from different
perspectives). Conversely, a given object (or
imaged object) can evoke different [verbal]
descriptions (Paivio 1986, 63).
Spoehr puts it another way: "In order to hold
information in long-term memory, we need to generate a
representation that retains abstract information on
meaning" (Spoehr and Lehmkuhle 1982, 7). She
recognizes that information in one system relates to
information in another in at least one way: "subjects
organize some spatial information in a hierarchical
fashion, using propositions to maintain some of the
relative spatial information about the parts" (Spoehr
and Lehmkuhle 1982, 233). Schemas are one way which
knowledge can be represented in a hierarchical
structure. These representations can be either verbal,
non-verbal, or a combination.
A schema is a description of a particular class of
concepts and is composed of a hierarchy of
schemata embedded with schemata. The
representation at the top of the hierarchy is
sufficiently general to capture the essential
aspects of all members of the class (Adams and
Collins 1979, 3).
Opportunities for retrieving information are
increased by utilizing both the visual and the verbal
system during encoding. Spoehr points out that we can
"code concepts such as table in terms of both visual,
sensory imagery and in terms of verbal, semantic
characteristics..." (Spoehr and Lehmkuhle 1982, 203).
There are more cues which can then elicit or retrieve
the concept of "table" than there are cues which can
elicit a concept, say, such as "truth."
Representations encoded or created in both the verbal
and visual systems are richer because they have more
opportunities to be accessed. In addition, these
representations, whether verbal or visual, can have a
Forming Elaborations to Link New Information
Elaborations are the inferences, or
assumptions, we make when we read. If we are to use
new information, or our verbal or visual
representations of that information, then we first need
to link it to what we already know (given information);
elaborations are the information we add from our prior
knowledge to make these links. Since writers cant
predict exactly what readers already know and how that
is organized or represented in their long-term memory,
there will be gaps in the communication process. As
van Dijk says, "Many propositions need not be
expressed, as the speaker may assume that they are
known to, or can be inferred by, the hearer" (van Dijk
and Kintsch 1983, 191). Readers must make inferences
about what is not explicitly stated. Writing for an
audience is a balance of not assuming too much or too
little of readers.
What readers assume depends on what they know.
In an article on the effects of elaboration on the
recall of text, Mandl and Ballstaedt found that
these optional inferences are elaborative
processes. Through elaborations relations are
created that go beyond those present in the text
and that relate the text to the various knowledge
structures the reader possesses (Mandl and
Ballstaedt 1982, 483).
Elaborations are a way to link given and new
information and to link different types of information.
Readers will make elaborations which writers
cant predict. Early on, when we were taught to write
topic sentences, we were being taught to reduce the
meaning of sentences in the paragraph to a gist or
macroproposition. The resulting topic sentence headed
the paragraph. Reduction, creating macropropositions
and representations, is a necessary part of creating
topic sentences. The result is a generalization which
summarizes the major points of the paragraph. This
technique gives the paragraph a hierarchical structure:
"The important thing about hierarchies, as you know,
are that they create a focus by distinguishing major
points from minor ones and they show how ideas are
related to one another" (Flower 1985, 145).
Hierarchies separate the general from the specific and
offer us conceptual information about the organization
of the information. This "meta-information" can be
Hirsch, van Dijk and others emphasize the
importance of the relationship of generalization and
reduction: "A crucial aspect of complex understanding
is higher level concept formation, organization and,
hence, reduction" (van Dijk and Kintsch 1983, 195).
Hirsch puts it this way:
The psychological device of representation is the
principal device of the mind for coping with
complex wholes of all sorts. Instead of trying to
think about a whole meaning in its complex
fullness and its relation to other complexities,
we isolate a part (or we use a short symbol) in
order to represent the whole (Hirsch 1977, 124).
Once reduced, information gains a hierarchical
structure: the "symbol" is the higher level and
subsumes the more specific information "under" it.
Readers will make connections by elaborating on "higher
level concepts" in order to link conceptual information
to what they already know, and to generate expectations
about the more specific information to follow.
When we create topic sentences, themes, or
gists, information becomes more hierarchical: higher
level concepts are extracted from, and subsume, their
supporting details. Propositions lower in the
hierarchy "inherit" the properties and characteristics
of their higher level, more general, macropropositions.
Inheritance can be defined as "the transfer of
knowledge among items of the same or similar types or
subtypes" (deBeaugrande and Dressier 1981, 91). Lower
level, specific details inherit the properties of their
higher level generalities or abstractions.
Hierarchical structure makes information more
spatial and increases the possibility that readers will
apprehend the information visually. When there is more
than one level of meaning, hierarchies give those
levels spatial organization. Text becomes hierarchical
because it is organized by macrorules, which, according
to van Dijk and Kintsch,
relate proposition sequences with proposition
sequences at a higher level, and thus derive the
global meaning of an episode or a whole discourse
from the local, sequential meanings of the
discouse....It follows that the resulting
macrostructure of a discourse is a hierarchical
structure, consisting of several
levels.... Sequences are no longer organized only
at the local level, by linear relations of
coherence ...but also by higher level conceptual
units (van Dijk and Kintsch 1983, 190).
Several researchers have shown how important
hierarchies are for organizing information. Linda
Flower advises that writers use hierarchies as a
strategy for sorting out ideas: "...the writer has to
group the facts and create a set of organizing ideas
with an underlying hierarchical organization" (Flower
In an article about text mapping, Dansereau and
Holley show that if students used certain coding
strategies designed to aid in creating hierarchical
structures, they quickly identified important concepts
The networking processes emphasize the
identification and representation of (a)
hierarchies (type/part), (b) chains (lines of
reasoning/temporal orderings/causal sequences),and
(c) clusters (characteristics/definitions/
analogies).... Application of this technique
results in the production of structured two-
dimensional maps. These cognitive networks
provide the student with a spatial organization of
the information contained in the original training
materials (Dansereau and Holley 1982, 538).
Seeking hierarchies and gists may be a form of
pattern recognition. Recognizing patterns implies a
visual and/or spatial dimension in the text
comprehension process. Bernhardt points out that "the
written mode necessitates the arrangement of script or
typeface, a process which gives visual cues to the
verbal organization of the text" (Bernhardt 1986, 66).
As we read, we seek patterns, visual as well as
textural: "because of the opportunities it offers for
visual inspection, writing heightens the awareness of
categories and divisions..." (Bernhardt 1986, 66).
If, as Norman says, "the process of pattern
recognition requires that information from sensory
sources and from long-term memory interact," then
hierarchies increase the ability and speed with which
we can match new information to what is in long-term
memory at a general level (Norman 1976, 113).
Hierarchies increase our ability to find and match
(verify) new information with given information by
organizing and conceptualizing new text and therefore
reducing the possibilities of given information to
which it might be linked. Experiments in visual
pattern recognition by Palmer support this theory of
When the test pattern appears, the subject first
attempts to compare it to the representation at
the highest level of the hierarchy.... If the test
pattern is a good subpart, a match is soon found,
and the verification response is made quickly
(Spoehr and Lehmkuhle 1982, 80).
Since we organize information in long-term
memory into patterns, one aspect of comprehension
involves finding and verifying the correspondence of
new text against existing patterns and hierarchies in
long-term memory. In this context, pattern recognition
enables us to process text visually and spatially as
well as verbally. The cognitive burden on short-term
memory is relieved because some aspects of the
information are processed through the visual system.
Because of this "dual" processing, written
especially when visually informative, encourages
the writer to be exact about grouping related
ideas, delineating beginnings and endings, and
using cues to signal to the reader a graphic
representation of cognitive organization
(Bernhardt 1986, 67).
But writers have to resolve several tensions
before communication can occur. The struggle to
constrain intended meaning during text production
(writing) so that it is relayed accurately and
efficiently during text comprehension (reading)
requires reducing the tensions between the following
processes and systems:
1. given and new information;
2. short-term and long-term memory;
3. elaboration (inference) and reduction (gist);
4. local and global meaning;
5. micropropositions and macropropositions; and
6. linear and hierarchical structures.
Of course, perfect resolution isnt possible.
Writers cant predict exactly what readers already
know. Since it is likely that readers will process
text using some combination of information processing
systems, writers can reduce this uncertainty by
designing and organizing text to consciously exploit
the users other information processing systems. This
increases the liklihood that the reader will apprehend
the intended meaning of the text.
USING VISUAL INFORMATION TO OVERCOME
Research in cognitive psychology and
neurolinguistics suggests that visual information is
processed through a different system than verbal
information. But rules govern this visual perception
process. If we can structure text to utilize both
systems for information processing, we can relieve the
burden on short-term memory by processing certain
aspects of the information visually.
Processing Visual Information Simultaneously
Many studies have shown that the memory for
visual information is better than for verbal
information (Bieger and Glock 1986; Glass et al. 1981;
Maclnnis and Price 1987; Marschark 1985). According to
Spoehr, "two of the most remarkable and noticeable
aspects of visual memory are its large capacity and
long duration" (Spoehr and Lehmkuhle 1982, 173). Some
of the characterisitics of visual memory are similar to
verbal memory. For example, pictures are remembered
better if they have some coherent organization, just as
coherent text itself is remembered better. Subjects
remember gists of pictures, similar to the way they
remember gists of text. As Spoehr notes,
The very good picture memory reported in many
studies may not result so much from the subjects
ability to record details as from their ability to
use organization both within and between related
pictures to construct meaningful representations
of integrated visual ideas (Spoehr and Lehmkuhle
Gestalt psychologists, among others,, offer
several principles which govern the processing of
visual information. In contrast to the linear,
sequential model of text comprehension, the visual
system tends to allow us to apprehend the whole all at
once, simultaneously. "We first process information
about features of the overall configuration and then
proceed to the analysis of the smaller parts" (Spoehr
and Lehmkuhle 1982, 79). But the initial experience is
a "gestalt," a visual impression which is so integrated
that the subparts are not immediately recognized as
Paivio uses the human face as an example: we
first perceive the whole face before we can break down
our perception into the constituent parts that form it.
The structural information is characteristically-
organized in a synchronous or simultaneous manner
into perceptual hierarchies or nested sets. An
example is the human face, which consists of eyes,
nose, lips, and other holistic components that are
themselves composed of still smaller parts....All
are part of a synchronously organized hierarchical
structure...(Paivio 1986, 60).
But we see the face as a whole not as the
hierarchically structured nested sets composing it. We
experience a visual "gestalt" of the whole. Only when
we are cued to look for a specific subpart do we then
generate an image and scan it to find information about
the parts. The hierarchy derived from the whole acts
as a map. Paivio says,
The meaning of simultaneous or synchronous
organization is apparent at the perceptual level:
The organized elements of a face are seen together
in time because that is the nature of the visual
system. At the cognitive level, this
characteristic is illustrated by mental images of
complex objects, the parts of which are
simultaneously available for processing (Paivio
Constraints on Visual Information Processing
To retrieve visual information, we need to
generate the whole picture, then successively scan its
subparts. Paivio says "the part that is initially
imaged and the ordered scanning will depend on the
contextual cues that activate the representation..."
(Paivio 1986, 60). Cosmetic surgeons, for example, are
"cued" to look at noses in isolation from the whole
face. Generating or constructing, then scanning
visual or spatial representations takes more time than
visualizing representations we have already processed
and stored in our minds. (Neurolinguistic programmers
call the first type "constructed" images and the latter
type "eidetic" images.)
Gestalt psychologists specify several other
principles governing visual perception. The principle
of "good figure" suggests that readers seek bounded
space, to "pull figures out of the background, to give
them definition against the undistinguished field in
which they are located" (Bernhardt 1986, 72). When we
look at a page of text and nothing stands out, we
cannot gain any information about the contents until we
read it. By contrast, if text is visually informative,
we can look at a page and visually process some
information about the contents. This "meta-
information" can help us remember where to look for
something later, or it can define what is general and
what is specific information. This meta-information
can be revealed "through headings, and through the
groupings of related information which are set off by-
blank space" (Bernhardt 1986, 72).
If readers seek visual boundaries, it may be
because there is a type of visual closure similar to
the verbal closure which occurs in short-term memory.
Gestalt theory says that if good figure is lacking and
the reader has to strain to delineate groupings, "the
perceiver has a tendency to fill in the missing gaps,
to provide the missing definition..." (Bernhardt 1986,
72). This is a type of closure which operates
visually and effects "...our organization of elements
[and] tends to form them into simple, closed figures,
independent of their other continuation similarity, or
proximity properties" (Spoehr and Lehmkuhle 1982, 66).
Other principles suggest that we will tend to
see as homogeneous grouping "units which resemble each
other in shape, size, color, or direction" (Bernhardt
1986, 72). Bernhardt emphasizes the spatial impression
text will make on the reader and points out that "to
attend to the layout of the text requires considering
the text as a visual gestalt, focusing attention on the
total visual impact of the text on a prospective
reader" (Bernhardt 1986, 72). Consequently, the
appearance of the text on a page should be considered
another aspect of the communication dynamic.
Using Visual Strategies to Recode Information
Employing visual strategies can support verbal
information processing and increase the amount of
information about the text we can process. This
integration reduces the burden on short-term memory.
Since text must first be visually apprehended, the
spatial appearance provides an opportunity to "give
visual cues to the verbal organization of the text"
(Bernhardt 1986, 66).
By carefully exploiting our natural visual
tendency to apprehend the whole "figure" and the
tendency to visually group homogeneous units together
(a visual "chunking" or grouping into like types), we
can provide distinct locations on the page for
categories of information. Grouping information can cue
readers, giving them meta-information about how the
text is structured, what the key points are and how
they fit together.
Spoehr says "since the average person easily
coordinates visual and verbal processing all the time,
almost all activites involve some interaction between
the two systems" (Spoehr and Lehmkuhle 1982, 241).
Writers of technical documents can provide visual cues
which reveal the hierarchy of the information, giving
the reader spatial information which identifies the
parts and subparts of that hierarchy, delineating
general information from specific information. These
cues can enhance the readers encoding process by
causing it to happen sooner. This is crucial for
conveying intended meaning because as Linda Flower
Readers organize ideas into gists and natural
hierarchies. A final way readers understand and
remember what you tell them is by organizing your
discussion in a general hierarchical way. If you
dont do it for them, they will do it themselves
and the result may be far from what you had
intended (Flower 1985, 144).
When the organization of the text is visually explicit,
it can provide readers with visual cues for making
further connections with what they already know, as
well as for generating new connections between systems
of ideas. Visually explicit textural organization can
enhance the readers process of encoding new
information, or recoding given information, for storage
(and/or reorganization) in long-term memory.
Norman points out that "recoding is an
extremely powerful weapon for increasing the amount of
information we can deal with. In one form or another we
use recoding constantly in our daily behavior" (Norman
1976, 94). Recoding techniques vary with individuals
and often involves the visual system.
Paivio points to the power of imagery in the
recoding process: "imagery is likely to be more
efficient ('swifter) than language in encoding
multiple bits of concrete information into an
integrated whole...." (Paivio 1986, 201). Even in
short-term recall experiments, "concrete words and
sentences are remembered better than similar abstract
materials because of this dual availability of semantic
and perceptual information in memory" (Marschark 1985,
744). Marschark notes that the role of imagery depends
on the context and the type of text. He says that the
effect of imagery on recall "should be less pronounced
when linguistic units are presented in context (e.g.,
related sentences in paragraphs) because relational
processing will result in abundant retrieval routes for
abstract as well as concrete knowledge" and make
retrieval (of prior knowledge) from memory more likely
(Marschark 1985, 736).
Imagery, hierarchies, and other products of the
visual system therefore will have powerful effects when
context is at a minimum and/or the reader is unfamiliar
with the text. While short-term recall tends to be
data-driven, imagery and hierarchies can link details
or act as a "conceptual peg" to give the reader an
overview (Woods 1986, 1326). Other research on short-
term recall supports this finding. Childers & Houston
found that "redundancy between pictorial and visual
information in advertisements produced a superior
effect on recall for the advertisement..." (Maclnnis
and Price 1987, 477). Stein, et al., found that the
"subjects given pictures illustrating information in
the explanatory and verbal elaborations also tended to
have the highest levels of recall" (Stein, Brock and
Ballard 1987, 287 ) .
We can deliberately and consciously exploit the
visual system when we recode information. Norman notes
that in information theory the goal is to reduce input
by using another code:
There are many ways to do this recoding, but
probably the simplest is to group the input
events, apply a new name to the group, and then
remember the new name rather than the original
input events (Norman 1976, 92).
This process relates closely to the ability to
conceptualize or abstract information. The visual
system may be involved in recoding in several ways. By
generating scenarios, images, and models as well as
structuring hierarchies, trees or taxonomies, we are
recoding information to make it easier to transfer to,
or store in, long-term memory. Some strategies are
more visual, some are more spatial, but they all
require visual and spatial knowledge to "see" the
relationship of "bits" of information and make (recode)
them into larger, more meaningfully related systems.
Linda Flower calls this process conceptualization; it
allows writers to "synthesize or combine ideas and
create a new concept that expresses or encompasses all
they have in mind" (Flower 1985, 94).
Recoding enables the writer (and the reader) to
"visualize the whole [argument] and see how all the
parts might fit together" (Flower 1985, 95). Recoding
is not just generating ideas, but fitting them
together, linking systems of ideas with subsystems and
using the resulting hierarchy as a cognitive tool to
"spot the missing links in your thinking and generate
new concepts that will organize your ideas" (Flower
The end product of recoding can be a model, or
a way to see relationships and fit new information into
a context with given information. The situation we are
most concerned about is the one in which readers have
no context for the new information they are trying to
process. This lack of context, or lack of ability to
fit new information into a system of existing
information, is what often distinguishes novices from
VISUAL ASPECTS OF THE WAY
EXPERTS SOLVE PROBLEMS
Experts use models to manipulate large
quantities of information so that it can be retrieved
faster. Chunking information reduces the cognitive
"area" experts have to search to retrieve what they
need. Developers of computer systems mimic this
strategy; because they write computer programs, they
need to make the process of retrieving information from
memory explicit. Understanding how computer systems
retrieve information can inform us about visual
How Representation Reduces the Load on Memory
The ability to create and exploit complex
visual or spatial cognitive strategies distinguishes
experts from novices. Conducting research about how
physicists solve physics problems, Chi, et al., found
that "expert power lies in the experts ability to
quickly establish correspondence between externally
presented events and internal models for these events
(Chi, Feltovich and Glaser 1981, 123).
Experts tended to categorize problems into types
that are defined by the major physics principles
that will be used in the solution.... We view the
categories of problems as representing internal
schemata, with the category names as accessing
labels for the appropriate schemata (Chi,
Feltovich and Glaser 1981, 150).
This technique (categorization into types) guides
programmers in developing expert systems. The programs
they create operate based on "pattern-action rules."
The system of programs:
operates by determining at every step what rules
are satisfied by the current state of the system,
then acting upon that state by executing one of
the rules. Conceptually, this operation entails
testing each of the systems rules against the
current state. However, as the number of rules
increases, techniques are sought to avoid testing
all of the'them (Woods 1986, 1326).
A technique used to avoid testing a new rule against
all of the systems existing rules involves chunking
like types of situations. The resulting taxonomies
reduce the verification process by generalizing or
creating a level of abstraction which links concepts
together "by a relation of generalization so that
concepts more general than a given concept are
accessible from it" (Woods 1986, 1326).
A computer program called RESEARCHER uses this
generalization strategy to verify new information about
objects; it can retrieve information about similar
objects from memory. This technique reduces the area
of memory which has to be searched by the system.
Information about the new object is checked (verified)
against stored information about only objects of the
These computer systems are designed to
implement the process of 'situation recognition, also
known as instantiation: systems are essentially testing
the specific instance to ensure it belongs "under" its
category. Its as though the question were being
asked, "Which general category does this instance
belong to?" The program is testing an instance against
a category, or type, of instances. The RESEARCHER
program searches the descriptions of objects in
permanent memory, which are stored "in terms of a
hierarchy" of general types ("similarities among
representations"), and finds a more specific instance
(Lebowitz 1988, 1489). "This process consists of
discovering that some input data...can be interpreted
as filling roles in some concept known to the system
and therefore constitute an instance of that concept"
(Woods 1986, 1329) .
The computer program RESEARCHER stores
information about objects in its permanent memory.
The program then can build up a "conceptual
representation" for further use (Lebowitz 1986, 1488).
In this way, the program "learns": it can add to its
knowledge base as it encounters new information about
objects. It can maintain "information from the
relational segments of the text in short term memory
and then... determine how the appropriate objects relate
to each other...."
The basic approach to memory in RESEARCHER is to
store objects in terms of a hierarchy of
automatically generalized prototypes created by
noticing similarities among representations. With
this sort of memory we can take partial
descriptions of objects and determine whether they
correspond to objects or prototypes in memory by
searching through the hierarchy (Lebowitz 1988,
By storing propositional information which is
then accessible for verification and object
identification, computer systems can compensate for the
lack of the visual or conceptual ability people need to
verify that the instance belongs to a category. But to
be manageable, object descriptions have to be organized
hierarchically, and this organization reduces the area
to be searched.
How does spatial location in the hierarchy
increase accessibility? Expert systems use a "save and
skip" strategy: "RESEARCHER saves noun phrase words in
a stack until the head noun is reached" (Lebowitz 1988,
1488). Systems with huge memory capacities can hold
modifying information infinitely until verification is
accomplished and a match is made between the head noun
which the program holds in temporary memory and the
appropriate (same) category encountered in the systems
hierarchically organized permanent memory.
This strategy echoes the process expert
physicists use for solving physics problems. They
"classify according to the major physics principle
governing the solution of each problem," and thus
reduce the number of possible solutions to be
considered or tested, once they have determined that a
problem is a certain type (Chi, Feltovich and Glaser
1981, 125). This process, called "perceptual
chunking," involves forming a representation or "an
internal cognitive structure... to stand for or model
a problem" (Chi, Feltovich and Glaser 1981, 144). A
model is a reduced representation of a chunk (body) of
information which may be stored through the verbal
system, the visual system or both (or other systems).
Whether they are primarily verbal or visual, models, by
virtue of this reduction, have a spatial, hierarchical
aspect. Readers and writers can use them to recode
information in order to retrieve it more efficiently.
Once created, users can then scan their models to
retrieve what they know in much the same way as
scanning an image. Users can also modify models based
on new information so that the models better represent
what the users know.
Creating Situational Models
Models, then, like other internal
representations, integrate information in both the
verbal and visual systems. Generating models is a
potent way for experts (and others) to represent and
reduce verbal information. Since neither experts nor
writers automatically generate complete and
hierarchically organized bodies of information, they
develop conscious strategies to first generate
representations of that information. Then it is easier
for them to "see" the patterns of relationships.
Teachers of writing suggest this strategy.
Linda Flower, for example, tells writers to use an
issue tree to "model" ideas. An issue tree is "an
upside-down tree that puts your ideas in hierarchical
order"(Flower 1985, 95). Gabriele Rico (Writing the
Natural Way) suggests "clustering" or drawing circles
around idea essences and connecting them to make ideas
and their relationships (or lack of relationship)
visually explicit. She says clustering "makes an
invisible Design-mind process visible through a non-
linear spilling out of lightening associations that
allows patterns to emerge" (Rico 1983, 28). What these
strategies have in common is the ability to both
represent and reduce information, making associations
explicit, enhancing the ability of the user to see gaps
in those associations as well as to generate new ideas
While experts may not always have to make their
models explicit, writers of technical documents usually
want to increase the ability of the reader to apprehend
their intended meaning and thus they do need to make
their models explicit. Linda Flower puts it this way:
But much recent research in psychology shows that
it is more accurate to think of readers as hard at
work using our information like a set of tinker
toys, to build an idea structure of their own. And
if our writing doesnt help them build it, that
final structure may or may not resemble our own
(Flower 1985, 141).
As writers of technical text, we want to improve the
readers retention and/or performance which are both
based on the retrievability of information. To improve
retention, we want the reader to encode information so
that it can be retrieved quickly. To improve
performance, we want the reader to make better models,
that is, models that resemble what we intended as well
as exploit the same coding system.
Perrig and Kintsch have researched students
creation of what they call "situational" models of the
layout of a fictional town based on propositional and
visual instructions. Some subjects were given
procedural instructions while other subjects were given
a map of the town. "Subjects who actually saw the map
ended up with a much better image than those who had to
construct the image from a verbal description" (Perrig
and Kintsch 1985, 515). Subjects performed "faster on
survey inferences which corresponded more closely to
the spatial imagery representation of the map they had
seen" (Perrig and Kintsch 1985, 515). Once the
information is stored in a system, then retrieval
improves when the retrieval cue directly utilizes the
same encoding system because information doesnt have
to be transferred from one system to another.
A variety of research supports the notion that
retrieval is sensitive to encoding strategy -- that
when the system used by the retrieval cue corresponds
with the system used at the time of encoding,
performance is improved. Stein, et al., found
"subjects given pictures illustrating information in
the explanatory verbal elaborations also tended to have
the highest level of recall" (Stein, Brock and Ballard
1987, 287). Weldon and Roedigers research supports
the idea that "memory will be 'good to the extent that
procedures engaged at encoding and retrieval are
similar" (Weldon and Roediger 1987, 278).
If models are going to help us perform tasks or
use instructional text, then the power of the model
lies in its ability to help us retrieve information as
well as store information. Research indicates that
retrieval depends on the retrieval cue. For example,
if information stored in long-term memory is
hierarchically organized, then the model can provide a
map, or short cut, which reminds the user of the
location of specific information within the
Although recall experiments usually deal with
relatively short-term recall, memory for images and
other visual phenomena is less subject to decay than
for verbal, propositional information. This may be
either because in the visual system retrieval is
faster, or because the verification process is
simplified, or because differences between short and
long-term memory are less significant than in the
verbal system (or some combination of all three).
Speed of retrieval in the visual system seems to depend
on whether we have to construct images (then scan them)
or whether we can access images which we already have
stored in long-term memory. All these possibilities
tend to improve the case for using a dual-coding
strategy to design texts: the research indicates that
in order to remember information over a longer period
of time, we are recoding it to make it more visual.
The ability to use models to enhance
comprehension assumes that writers can choose from a
continuum when organizing information in the text.
They may choose not to organize the information or not
to make that organization visually explicit. At the
other extreme, they can choose to chunk information in
the text and place it into systems of relationships
which are visually apparent to the reader. Writers can
choose to employ strategies ensuring that the texts
organization is visually accessible to the reader.
INTEGRATING VERBAL AND VISUAL INFORMATION
IN THE WRITING PROCESS
We use technical documents for a variety of
purposes. If we read sections of the document in random
order, we need to know their place in the hierarchical
structure of the document. This pattern should be made
visually explicit. In procedural or instructional text,
writers can use the agent of action or the actions
themselves as cues the reader can use to chunk and
organize the material.
Instantiating the Model
Although the model represents, or stands for, a
chunk of information, the reader must make this
representation specific and tangible for it to be
useful. Models alone arent adequate for understanding
text if the reader cannot generate an instance of the
model. This ability is called instantiation. "It is
the mental act of transforming an internal conceptual
representation... into a verbal, external analog"
(Flower 1984, 153).
One form of instantiation involves imagining
oneself doing what it is the text describes:
To understand the text we have to represent what
it is about. If we are unable to imagine a
situation in which certain individuals have the
properties or relations indicated by the text, we
fail to understand the text itself, (van Dijk and
Kintsch 1983, 337)
Maclnnis and Price note that "subjects who imagined
themselves performing the set of behaviors
significantly changed their behavioral intentions"
(Maclnnis and Price 1987, 481). By imagining oneself
doing something, readers can "imagine their reactions
to constructed scenarios better than they cain imagine
the reactions of others. Thus the imagined scenario
(including responses) is more concrete" (Maclnnis and
Price 1987, 481). When we imagine scenarios, whether
they involve us or not, they are more tangible because
we draw on the variety of experiences and information
in our memory. This information may be verbal, visual,
emotional, kinesthetic, or aural.
Perrig and Kintsch emphasize that what the
reader already knows is crucial to the formation of the
model or scenario: "the integration of information
derived from the text with the readers pre-existing
knowledge and its transformation thereby, produces the
situational model" (Perrig and Kintsch 1985, 515).
When writers can provide visual cues, such as bounded
sections with headings, the readers ability to scan
and then stop to read the text at any point (non-
linearly, or by random access) is enhanced. Readers
will find it easier to read for their particular needs,
linking the new information with their existing models
and scenarios or using textual cues to create new ones.
But to divide documents into sections which
have a clear place in the hierarchy, yet which can be
processed independently, writers first have to find the
organizing principles in the information itself. We
know when an organizing principle is lacking or the
existing structure of a document is inadequate: we
have to continuously back up and reread text and/or are
forced to read the whole document sequentially to find
a way to link the information to something we know. In
procedural or instructional text, this rereading means
the material hasnt been chunked; in other words, the
material isnt "modular."
In procedural text, a common element which can
be used for chunking may be the agent who is performing
an action, the action itself, the output of the
action, or a common destination of that output.
Chunking the procedural information this way offers
possibilities for the vertical or horizontal
configuration of information; since it now has a
spatial aspect, it can be expressed, or laid out, as an
! ! Procedure One I Procedure Two Procedure Three
I I *
! Agent 1 ! !
I ~ -a M-_n . ii st
II I I
II I >
!Agent 2 ! ! I
I ____- . | .___ ._____
II I I
I Agent 3 ! ! i
Figure 5.1 An array gives information a spatial aspect.
An array gives users the ability to access information
from the vertical or horizontal axis, depending on what
they need. In Figure 5.2 below, we might need to know
who is required to return Enclosure 2 (vertical access)
or we might need to know which enclosures a specific
contractor is required to return (horizontal access).
[Return [Return [Return
[Enclosure 1A [Enclosure IB [Enclosure 2
An array allows vertical or horizontal access.
Categorizing by agent or action not only
organizes the information but provides continual
answers to the questions "Who is doing this?" "What
are they doing?" Visualizing a situation involving
one subject/agent and one verb/action invites
modularity. E. Yourdan utilizes this principle in
If you can describe, accurately and honestly and
completely, what your module does (not how it does
its job, but simply what the job is) in a simple
English sentence containing a single verb and a
single object, then you probably have a
functionally cohesive module (Yourdan 1982, 22).
Unless the writer can describe who is doing what in a
reasonably simple way, the reader of the text will not
be able to form an adequate situational model for
We must verify information in the model to
ensure that it aligns with what we already know. To
chunk information when creating hierarchies and spatial
models, we need to verify relationships among the parts
as well as link those parts to the idea or category
which includes them. Linda Flower describes the
chunking process: "We look at two or more facts or
ideas and ask ourselves, 'What single new idea (or
chunk) would connect (or contain) those facts?"
(Flower 1985, 141). It takes less time for the reader
to accomplish this (establish these links) when the
text is hierarchically organized and when this spatial
information is visually accessible.
Seeing and Verifying Relationships in the Hierarchy
Verification is not the same thing as
comprehension: "That comprehension and verification
are logically distinct is obvious because we can
comprehend many sentences whose truth value we do not
know" (Glass, Miller, Beck, and Eddy 1985, 457). When
readers transform propositional information into
spatial information (using models, hierarchies, and
scenarios), they are involved in the verification
process at several levels. They need to verify
incoming (new) information with already comprehended
knowledge (given); they also need to verify the
relationship of new information within models they are
How we form models and hierarchies depends on
our existing schema, or prior knowledge. In research
exploring the differences between childrens and
adults representations of "Star Wars," Means and Voss
found that "the representation of Star Wars was
hierarchical for both younger and older experts" (Means
and Voss 1985, 754):
Analyses indicated that older experts* story
representations were qualitatively different from
those of younger experts, the former based upon a
conflict schema, the latter upon a military-based
good guy-bad guy schema (Means and Voss 1985,
The "older [individuals] have a more developed schema
involving the complexities of international
conflict..." and were able to identify more higher
level goals than younger 'experts (Means and Voss
1985, 755). Older experts perceived characters
motivations based on different, more sophisticated
schema. Differing responses show how hierarchies and
scenarios are based on individuals differing prior
knowledge bases and differing goals.
If hierarchical structures depend on the prior
knowledge embodied in schemas, then the verification
process involves relating new information to existing
schemas. This existing (given) information forms a
restrictive context for incoming (new) information.
Glass, et al., noticed that abstract sentences could be
verified as fast as concrete sentences only when "a
restrictive context was supplied for the abstract
sentences" (Glass, Miller, Beck and Eddy 1985, 463).
This context could give abstract information a
"specific representation" and thus provide the user
with a representation he wouldnt then have to search
for or create for himself. Such a supplied
representation frees the user from having to construct
his own, and also supports the conclusion that
representations allow us to verify information faster
by reducing (restricting) the amount of information we
will have to check it against.
In instructional or procedural texts, writers
can list particular processes or procedures as
coordinate items under the appropriate agent or action;
the agent or action then becomes the category.
Ultimately, this category becomes a title or a heading,
the most general level of the hierarchy. Hierarchies
integrate verbal and visual information, using
macropropositions to reduce the subsumed details.
Since we tend to remember the general levels most
readily in long-term recall, headings (formed from
macropropositions) should be visually prominent to
allow the user selective access (and re-access) to the
Hierarchies also organize visual images. We
generate the whole image and scan it for more
information about the subparts. With both visual and
verbal-type hierarchies, we continually seek coherence
in the relationships of details to their categories.
In fact, verification may be primarily a visuo-spatial
task, as suggested in the research of Glass, et al.,
who find that "visuospatial representation plays an
overt role in the verification of many sentences"
(Glass, Miller, Beck and Eddy 1985, 442).
In procedural and instructional text, the agent
of action must be clear so the reader can generate an
instantiation of the model. Instantiation is what
makes unfamiliar text more familiar and accessible. We
are making the general more specific and explicit, or
the unseen imageable by generating a tangible example
of a concept. Flower says that "studies of adults
reading difficult documents suggest that people may not
really comprehend a new concept unless they can
generate an instance of it" (Flower 1984, 154).
In complex legal writing, for example, the
passive voice and complex strings of modification bury
the agent of action, and inhibit the readers ability
to form and visualize an instance, or scenario, of
somebody doing that which the information describes.
Chunking information using agents or actions makes it
easier for users to create scenario. Users will not
have to search the text for the key building block of
the scene while still holding modifying information in
short-term memory. In recall tasks, the visual
prominence of agents and/or actions enhances the
retrieval of general, conceptual information containing
key elements of a scenario in correct relationship.
These results indicate that users form scenarios sooner
or are using supplied scenarios to enhance retrieval.
If the reader has to continually ask himself
"Who is doing this?" until he identifies correctly and
verifies the agent of the action, closure (certainty)
is postponed for too long. Postponing closure inhibits
the readers formation of a situational model and
inhibits instantiation; postponing closure prolongs the
comprehension process. However, if the agent and
action are readily discernible, readability is
How much context the writer provides depends on
what he assumes about the reader. The writer needs to
provide more context for novice users than for expert
users. The question becomes: How can the writer predict
whether readers will make their own models from the
cues (visual and otherwise) supplied in the text or
whether the reader will fully adopt the model provided?
EXPERIMENT: HOW CHUNKED DOCUMENTS
CAN PROVIDE A CONTEXT FOR NOVICES
Past research has shown that experts chunk
information and use models to organize it. This
experiment starts with two versions of a document: the
first version offers few visual cues to its internal
organization; the second provides visually prominent
sections and headings to give new users a "head-start"
in the model making process. "Novice" users of this
document retrieve more conceptual information in
immediate recall tasks.
Problem: Creating a Context
Attachment A is a document containing
instructions for potential contractors who may, if they
fulfill its conditions, provide goods or services to
the Regional Transportation District (RTD). (Please
see Appendix A, page 77, for a brief overview of the
DBE Program at RTD.)
The original document (OLD) has no visually
apparent structure; the pages look homogeneous and
undifferentiated. When Attachment A was revised (NEW),
I added an overview, gave the procedures headings
(e.g., "Meeting Contract Goals") and listed specific
details of procedures under their general headings.
(Please see Appendix B, page 78, for OLD version of
Attachment A; see Appendix C, page 82, for NEW
As a result, the NEW Attachment A has a
visually apparent hierarchical structure. While some
users of Attachment A are familiar with the information
it contains, other users are not. Overviews and
hierarchical organization provide an opportunity for
visual access, presumably making it easier for all
users to store a representation of the information in
The following experiment was designed to
determine if users of the NEW version would be able to
find information faster and retain it longer than users
of the OLD version.
Method: Exposure to Unfamiliar Material
Seventeen students volunteered to read either
the OLD or NEW version of Attachment A and take a short
test. Volunteers were from two undergraduate
communications classes, eleven from one class, six from
another. The first section of the test was a simple
matter of timing how long it took students to find
various specific items in the two versions of the
document. (Please see Appendix D, page 89, for the test
materials presented to the volunteers.)
Part of the second section of the test was a
free recall task. After asking them to "go over" the
document until they "felt comfortable with it," but
for a maximum of 15 minutes, students were asked to
write down whatever they remembered about the text.
They were alloted two minutes.
Five students were given a "visual
interference" task after they had gone over the
document. Before they wrote down what they remembered,
I asked them to describe to me the kitchen of the house
they lived in when they were 12 years old. The
intention was to see if this (visual) interference task
impacted their ability to form a representation of the
document and to store the information in long-term
An additional variable was introduced: a time-
lag component tested long-term memory of the text.
After a week had elapsed, volunteers were asked to
recall and write down what they remembered about the
Observations: Forming Gists
In the timing test, students found information
three times as fast in the NEW version of Attachment A
(see Figure 6.1 below). Subjects averaged 31 seconds
to accomplish simple search tasks in the OLD version,
10 seconds in the NEW version.
Figure 6.1 NEW version users found information about
three times faster than OLD version users.
Note: There Is no question #4 on OLD version
Table 6.1 (page 57) is an array showing the
types of answers which users of the OLD version gave on
the immediate recall test. Six out of nine users of
the OLD version produced fairly rote responses; these
students repeated or attempted to reproduce verbatim
various definitions they had read in the document.
Seven used a list format which usually had random
order: general information was mixed in with specific
information. Four of the nine OLD version users
responded with complete sentences on immediate recall
tests. A representative response by a user of the OLD
version is shown in Appendix D.
On the long-term recall test, five of the nine
users of the OLD version shifted to a less random, more
hierarchical organization in their responses (Table
6.2, page 58). These subjects* responses evolved from
the list-like formats given at the immediate recall
test to overview-type statements given at the long-term
recall test. Their responses included many key words
they had used in the first recall test. Often, in
making conceptual, overview-type statements, OLD
version users added an agent and an action which they
hadnt used in their earlier list-type formats.
Subjects using the NEW version of the document
gave responses to the immediate recall test which
contained more conceptual information (Table 6.3, page
59). Seven out of eight responded with gists and
general scenarios (procedural models representing a
situation). All volunteers used complete sentences,
and half of those sentences included an agent and an
action in correct relationship with each other. When
users of the NEW version wrote definitions or single
word lists, they were usually preceeded by a general
statement, or macroproposition, such as: "RTD is to
develop a program where DBE people are to be involved
as subcontractors" or "It concerned contracts issued
by RTD to build, etc. So much of this work has to be
fulfilled by DBE groups."
On the long-term recall test, users of the NEW
version wrote answers with strikingly similar syntax to
their immediate recall answers (Table 6.4, page 60).
Often, the second version was a "reduced" version of
their immediate recall answer, using the same key
words, but with less detail.
Most subjects using either version retained a
basic notion of who the players were in this fairly
complex contracting scenario. The long-term recall
test illustrated that most of them remembered who was
contracting with whom. Students who had the OLD
version tended to remember the RTD part of the liaison;
students with the NEW version remembered the DBE part.
Types of Immediate Recall Answers: OLD Version
I* I* I*
1 2 3 4 5 6 7 8 9
Definitions X X X X X X
Acronyms X X X X X X X
List format X X X X X X X
Complete Sentences X X X X
Definitions: subject attempted to list definitions verbatim
Acronyms: subject listed acronyms from document
List format: subject listed items rather than using paragraphs
Overview: subject made a general statement about the
Hierarchical: general information preceeded specific
Agent/Action: subject stated major agents in correct
Scenario: subject made procedural statements about situation
* interference task was given to this subject
before immediate recall test
Types of Long-term Recall Answers: OLD Version
I* I* I*
1 2 3 4 5 6 7 8 9
Definitions X X X
Overview X X . X X X
Agent/Action X X X X X X X
Scenario X X X
Complete Sentences X X X X
Reduced X X
see Table 6.1
Reduced: subjects used fewer words than in their
immediate recall response, but those words echoed
the syntax of the earlier response
interference test was given to this subject
before immediate recall test
Types of Immediate Recall Answers: NEW Version
1 2 3 4 5 6 7 8
Definitions X X X
Overview X X X X X X X
Hierarchical X X X X X
Agent/Action X X X X
Scenario X X X X X X
Complete Sentences X X X X X X X X
see Table 6.1
I* interference test was given to this subject
before immediate recall test
Types of Long-term Recall Answers: NEW Version
6 7 8
Overview X X X X X X
Agent/Action X X
Scenario X X X X
Complete X X X X X X
Reduced X X X X X
see Table 6.1
Reduced: subjects used fewer words than in their
immediate recall response, but those words echoed
the syntax of the earlier response
I* interference test was given to this subject
before the immediate recall test
In the NEW version, a DBE is defined and is visually
more prominent. Perhaps because of this prominence,
NEW users assumed that DBEs were the more important
component of the relationship.
Interference seemed to affect the immediate
recall answers of users of the NEW version. Users of
both versions processed data into gists and
macropropositions by the second recall test. The
immediate recall responses of subjects using the OLD
version exhibited the tendency to reproduce information
directly from the document. These responses were often
in a random order: there was no organization and no
hierarchical (general to specific) order evident.
Because NEW users immediate recall responses echoed
the structure presented to them visually in the
document, it is possible they were able to incorporate
the model supplied (visually) in the document or they
were making their own models sooner than OLD versions
users. NEW users may have been more vulnerable to
interference at immediate recall because they hadnt
yet integrated models into any existing schema/system
of their own. The fact that NEW users who were
interfered with listed some definitions on the
immediate recall test seems to reinforce this
conclusion. It is possible that interference forced
NEW users to either delay their model formation or
delay the integration of the given model into any of
their existing schemas.
Conclusion: Is What You See What You Get?
This experiment explores how and when the
visual system enters into the complex process of text
comprehension when the user is unfamiliar with the
document. Pinpointing when schema or model formation
enters into the information processing procedure would
require research beyond the scope of this thesis.
There seems to be clear support for the notion
that we form gists and schemas of propositional
information, whether it is chunked in the document we
use or not. However, if the information is chunked, it
seems that we can utilize supplied models in concert
with the visual information processing system to
The evidence produced by this experiment
indicates that models and schemas were formed by users
of the OLD version by the time of second recall task.
These users gave overview statements to describe what
they remembered even though they had originally used
more random, list-type formats. The evidence also
suggests that they created models based on what they
remembered from the immediate recall task itself rather
then what they remembered from the document. They may
have reduced their immediate recall answer even further
and re-organized it. Perhaps the immediate recall task
itself acted as a kind of rehearsal, influencing what
subjects remembered later.
Users of the NEW version either integrated the
information with existing scenarios at the first recall
task, or they were-able to use the models built into
the document. They often correctly ascribed actions to
agents and used complete sentences. Although it cant
be proven on the basis of this experiment, it seems
likely that users of the NEW version integrated the
overview that was offered within the text.
Although it is hard to determine when model
formation, or recoding takes place, the OLD version
users seemed to form models sometime after immediate
recall but before long term recall. NEW users may have
integrated models provided in the text before the
immediate recall task was given. This is desirable for
first time users of unfamiliar text because, when the
text facilitates the use of models through
hierarchical, visual array, users can:
1. find what they need in the document
sooner and then read it for their
2. know to which agents to ascribe
specific actions and procedures; and
3. develop a model/scenario sooner and
reduce the possibility of basing the
model on their own, possibly incorrect,
interpretation of the situation
described in the text.
When designing documents for various users,
writers can employ strategies which either guide novice
users to create models and hierarchies or offer models
which can be integrated into their existing scenarios.
When writers chunk documents and organize material to
make the hierarchical structure visually explicit, they
are making sure the reader will know which is the
general information and which is the specific
information supporting it, as well as who the agents
and what the actions are. This enhances the ability of
a variety of users to build models and scenarios.
This experiment supports the thesis that models
enhance the users ability to retrieve information
from within the document as well as from long-term
memory. This experiment also suggests that
hierarchically structured text gives users an
opportunity to integrate what is presented in the text
with familiar scenarios as well as a way to integrate
visual/spatial information processing with verbal
INTEGRATING VERBAL AND VISUAL INFORMATION
IN THE WRITING PRODUCT
When designing documents, writers can use
several strategies to integrate verbal and visual
information. They can:
1. use modular (bounded) sections for
visual reinforcement of structure and
2. clearly specify agency so readers can
visualize actions (present tense, first
3. give examples to enrich situational
models and scenarios.
Making Sections Modular
Chunking information and making sections of the
text visually finite (modular) allows the user to begin
to read the document at any point. This is crucial if
a document is intended for a varied audience. Bounded
sections increase the confidence (ethos) of readers and
visually hint that they wont have to read the entire
text to extract the information they need. Modular,
bounded sections in the document reduce retrieval time.
Clearly Specifying Agency
Whether readers create their own scenarios or
use those supplied in the text, writers need to provide
them with actors and actions. Writers should begin
statements with a specific, identifiable agent of an
Although Attachment A is written for DBEs and
contractors who subcontract with DBEs, it would take
technical assistance to understand what is supposed to
happen in this situation (OLD):
If the contractor has exhausted every reasonable
means to meet the RTD specified goal and has
failed to do so, it will furnish documentation of
its good faith efforts to justify a grant of
relief from the goal set forth in this attachment.
In the NEW version it is clearer agency that
makes the scenario possible to imagine:
In order to award a contract to a bidder/proposer
that has failed to meet the DBE contract goals as
stated in that specific contract, RTD will
determine whether the bidder/proposer made good
faith efforts and actively and aggressively sought
DBEs to meet those goals.
Agency is more definite; it is easier to visualize who
is doing what (and with whom). The NEW version of the
sentence also take the opportunity to more clearly
define what good faith efforts are. The OLD version
separates the action (exhausting "every reasonable
means to meet the RTD specified goal") from the
terminology "good faith efforts."
Instantiation is a way to anchor a
generalization. A situation is easier to visualize
once writers explicitly state who the agent is and what
the actions are. Writers can then instantiate the
scenario by providing a specific example.
The OLD version of Attachment A provided a
general scenario which missed an opportunity to place
the agents in a relationship to each other. When
revised, the overview established the agents of action
first in a general way and later with the specific
details which would echo or modify this basic scenario.
It was the key agent that was buried in the
overview type statements made in the OLD version:
OLD (placed halfway down on the page)
It is the policy of the RTD that equal opportunity
to participate in its procurements will be
provided to certified DBEs/WBEs.
NEW (first sentence)
It is the policy of RTD to provide all certified
DBEs maximum opportunity to participate in its
construction activities, professional service
contracts, and in the procurement of common goods
In the OLD version, you dont know who the
contracting partner is until you get to the end of the
sentence; the DBE is discounted as an agent of the
(contracting) action. There are two consequences of
burying the fact that DBEs are active agents: implying,
or emphasizing, a passive role in the contracting
process; and postponing closure for too long.
How Using These Strategies Improves Communication
Incorporating these methods in designing
documents helps users integrate verbal and visual
information, and can resolve the tensions between five
1. given and new information;
2. short-term and long-term memory;
3. elaboration and reduction;
4. local and global meaning; and
5. linear and hierarchical structures.
What follows is a brief summation of how each strategy
help to resolve tensions in these contrasting systems.
Linking Given and New Information.
1. Modular sections provide readers with more
opportunities to link information in the text with what
they already know. Sections with headings which reveal
the texts organization help users to create general
concepts and scenarios from which to build hierarchies
that are meaningful to them.
2. Clearly specifying agency in procedural and
instructional text gives users the tools to instantiate
models, allowing the user to put specific agents and
specific actions into a relationship with each other.
Users can employ their own scenarios to link agents and
actions in the document to more familiar situations.
3. Giving examples is a way of instantiating
the model and restricting its context (further
constraining its meaning) while at the same time making
a situation more imageable and allowing users to employ
knowledge structures in their visual and verbal
Relating Short-term and Long-term Memory
1. Modular sections can facilitate the movement
of information from the users short-term to long-term
memory. When information is chunked under a general
category heading, the user can reduce the amount of
given information in long-term memory he has to
manipulate in order relate it to the new information.
Reducing the time it takes the user to link given and
new information facilitates the transfer of new
information into long-term memory.
2. Users form models sooner and remember them
longer when they can easily identify agents and actions
in procedural text.
3. Examples provide another opportunity for
users to link specific details in the text to their own
scenarios stored in long-term memory, using them to
create new ones as well as transform existing ones.
Making Elaborations and Reductions
1. Modular sections and titles are visual and
verbal exercises in elaboration and reduction: titles
are the reduction of information which is to be
elaborated on in the ensuing paragraph.
2. Agents and actions provide users with
imageable elements which link verbal and visual
information. Relating agents and actions clearly and
quickly in the language of the document provides users
with a way to visually elaborate on how the procedure
3. Giving examples is a way which the writer
constrains the users elaboration process. If the
writer doesnt provide an example to illustrate the
intended meaning, the user might make incorrect
Linking Global and Local Meaning
1. Writers can visually identify information
which is global (more conceptual, general) and
information which is local (more specific). Headings
give visual, global cues to help users retrieve local,
specific details from within the text and from long-
2. Identifying agents and actions in procedural
text provides the user with more possibilities for
anchoring more general concepts with local (specific)
3. By providing examples, writers can give
users a way to make global information more specific.
Examples provide a way for the user to cognitively
rehearse the relationship of the global to the local.
Forming Linear and Hierarchical Structures
1. Writers should use titles and modular
sections to visually reveal the hierarchical structure
of the document. Users will remember general
information more readily. The hierarchy provides
conceptual and visual boundaries for the information.
Spatial structure provides a visual location on the
page for details which are processed linearly.
2. Clear agency and proceduralized actions can
make linear information more imageable. By presenting
information within a hierarchy, users can manipulate
larger amounts of details and "see" their
3. When writers provide examples they relate
higher level conceptual information to the details of
the text. This reinforces the hierarchy and makes it
functional by putting concepts in a visual (imageable)
relationship with specific agents or actions in
By employing these strategies, writers of
technical documents can facilitate the users ability
to comprehend and retrieve information. We can
tentatively conclude that providing visually accessible
models in the text improves the novice users ability
to perform immediate and long-term recall tasks.
However, we must take into account the
possibility that providing visual model-building cues
may limit some users participation in the text and
increase their vulnerability to visual interference.
Further research could clarify whether users differing
levels of familiarity with the text impact their
ability to use provided models for retrieving
information in immediate and long-term recall tasks.
Adams, M., and A. Collins. 1979. A schema-theoretic view of
reading. In New Directions in Discourse Processing, ed. Roy
Freedle, 1-22. Norwood, N.J.: Ablex Publishing.
Bernhardt, S. 1986. Seeing the text. College Composition and
Communication 37(1): 66-78.
Bieger, G., and M. Glock. 1986. Comprehending spatial and
contextual information in picture-text instructions. Journal of
Experimental Education 54(4): 181-188.
Chi, M. and P. Feltovich and R. Glaser. 1981. Categorization and
representation of physics problems by experts and novices.
Cognitive Science 5: 121-152.
Cooke, N., and J. McDonald. 1986. A formal methodology for
acquiring and representing expert knowledge. Proceedings of the
IEEE 74(10): 1422-1430.
Dansereau, D., and C. Holley. 1982. Development and evaluation of
a text mapping strategy. In Discourse Processing, eds. August
Flammer and Walter Kintsch, 536- 551. Amsterdam: North Holland
deBeaugrande, R., and W. Dressier. 1981. Introduction to Text
Linguistics. London: Longman.
Eddy, J., and A. Glass. 1981. Reading and listening to high and
low imagery sentences. Journal of Verbal Learning and Verbal
Behavior 20: 333-345.
Fletcher, C. 1981. Short-term memory processes in text
comprehension. Journal of Verbal Learning and Verbal Behavior 20:
Flower, L., and J. Hayes. 1984. Images, plans and prose. Written
Communication 1(1): 120-160.
Flower, L. 1985. Problem-Solving Strategies for Writing. San
Diego: Harcourt Brace Jovanovich.
Glass, A., D. Miller, L. Beck, and J. Eddy. 1985. Representation
of images in sentence verification. Journal of Memory and Language
Goetz, E.f R. Anderson, and D. Schallert. 1981. The representation
of sentences in memory. Journal of Verbal Learning and Verbal
Behavior 20: 369-385.
Hirsch, E. 1977. The Philosophy of Composition. Chicago:
University of Chicago Press.
Kintsch, W. 1978. Toward a model of text comprehension and
production. Psychological Review 85(5): 363-394.
Kosslyn, S. 1983. Ghosts in the Minds Machine: Creating and
Using Images in the Brain. New York: W.W. Norton.
Le Ny, J. and W. Kintsch, eds. 1982. Language and Comprehension.
Amsterdam: North Holland Publishing.
Lebowitz, M. 1988. The use of memory in text processing.
Communications of the ACM 31(12): 1483-1502.
Linde, C. and W. Labov. 1975. Spatial Networks as a Site for the
Study of Language and Thought. Language 51(4): 924-939.
Maclnnis, D., and L. Price. 1987. The role of imagery in
information processing. Journal of Consumer Research 13: 473-491.
Mandl, K., and W. Ballstaedt. 1982. Effects of elaboration on
recall of texts. In Discourse Processing, eds. August Flammer and
Walter Kintsch, 482-494. Amsterdam: North Holland Publishing
Marschark, M. 1985. Imagery and organization in the recall of
prose. Journal of Memory and Language 24: 734-745.
McGuinness, C. 1986. Problem representation: the effect of spatial
arrays. Memory & Cognition 14(3): 270-280.
Means, M., and J. Voss. 1985. Star wars: A developmental study of
expert and novice knowledge structure. Journal of Memory and
Language 24: 746-757.
Miller, J. 1984. Working memory. In New Methods in Reading
Comprehension, eds. David Kieras and Marcel Just, 327-348.
Hillsdale, N.J.: Erlbaum Associates.
Norman, D. 1976. Memory and Attention. New York: John Wiley &
Paivio, A. 1986. Mental Representations: A dual coding approach.
New York: Oxford University Press.
Perrig, W., and W. Kintsch. 1985. Propositional and situational
representations of text. Journal of Memory and Language 24:
Rico, G. 1983. Writing the Natural Way. Los Angeles: J.P. Tarcher,
Simon, H. 1969. The Sciences of the Artificial. Cambridge, Mass.:
Spoehr, K., and S. Lehmkuhle. 1982. Visual Information Processing.
San Fransisco: W.H. Freeman.
Stein,B., K. Brock, and D. Ballard. 1987. Constraints on effective
pictorial and verbal elaboration. Memory & Cognition 15(4):
van Dijk, T. and W. Kintsch. 1983. Strategies of Discourse
Comprehension. New York: Academic Press.
Weldon, M., and H. Roediger. 1987. Altering retrieval demands
reverses the picture superiority effect. Memory & Cognition
Woods, W. 1986. Important issues in knowledge representation.
Proceedings of the IEEE 74(10): 1322-1334.
Yourdan, E. 1982. Managing the System Life Cycle. New York:
Overview of RTD s Disadvantaged Business Enterprise
The Regional Transportation District (RTD) was
created by the Colorado General Assembly in 1969 to
plan, build, and operate a transit system. To carry
out this mandate, RTD depends on federal funds; the
U.S. Department of Transportation (DOT) requires that
contracts in any federally funded project be awarded to
contractors who utilize a predetermined percentage of
minority and women businesses (DBEs).
The Disadvantaged Business Enterprise Program (DBE)
program is composed of sets of procedures which guide
DBE participation in RTDs construction, purchasing and
professional service activities. These procedures are
set forth in the DBE program document, which was last
revised in 1984,
Attachment A is included in the bid document
supplied to all potential bidders on federally funded
projects. This legal contract governs the relationship
of RTD and the contractors and/or DBEs (they could be
the same entity) for the duration of the project.
Among other requirements, Attachment A enumerates "good
faith efforts". The contractor may be released from
DBE obligations if and only if he has made documentable
good faith efforts to locate DBEs. Attachment A lists
types of activities which the contractor must have
engaged in while trying to find DBEs to participate.
If he can prove that he engaged in these activities and
was still unable to locate appropriate DBEs, it is said
that he made good faith efforts and is still in
compliance with federal regulations. This phenomenon
is also referred to as "DBE unavailability."
AFFIRMATIVE A CTION/D IS ADVANTAGED BUSINESS ENTERPRISE (DBE) and
WOMEN BUSINESS ENTERPRISE (WBE) UTILIZATION
The RTD will annually set overall DBE/NVBE goals for work to be performed under RTD
contracts, including construction activities, procurement of common goods and
services, and professional service contracts. While the expected percentage of
certified DBE or WBE participation may vary from contract to contract due to the
availability of DBEs or WBEs in any given line of work, the RTD believes the overall
goals to be realistically obtainable over time with the assistance of the federal
government, the business community and DBE and WBE organizations.
In order to meet Federal requirements and to provide for maximum practicable
participation of certified DBEs/WBEs, the RTD will specify a minimum required
percentage of participation in contracts with subcontracting opportunities. (The
percentage may be zero (0)).
The RTD has specified a DBE goal of________% and a WBE goal of_______% for the work to
be performed under this contract(s). Since the RTD sets separate DBE goals and WBE
goals, this indicates they are separate and independent of each other, and are not to be
added together, nor can one be substituted for the other.
The amount of DBE/WBE participation will be determined by the dollar value of the
work performed by certified DBEs/WBEs as compared to the total value of all work
performed under this contract(s) and/or by the percentage of the net profit which the
parties agree will be shared by a certified DBE/WBE where a joint venture is entered
into for the completion of the project.
INSTRUCTIONS TO BIDDERS AND PROPOSERS
It is the policy of the RTD that equal opportunity to participate in its procurements will
be provided to certified DBEs/WBEs. To accomplish this objective the RTD requires, as
applicable, all bidders and proposers to complete and return this Attachment
(enclosures 1 thru 5) which obligates the Contractor to meet the goal(s) or assert a good
faith effort to attain the specified goal for certified DBE/WBE participation.
A bidder/proposer may satisfy the requirements of this Attachment by having certified
DBE/WBE status, by subcontracting portions of the work to certified DBEs/WBEs,
and/or by entering into a joint venture with certified DBEs/WBEs.
In order that all DBEs/WBEs meet the same eligibility criteria requirements for
certification (i.e., business size, socially and economically disadvantaged status, etc.),
the RTD in conjunction with other agencies in the State of Colorado have agreed that
the Colorado Minority Development Agency (CMBDA) will be recognized as the
certification agency for DBEs and WBEs. The RTD, however, reserves the right to
institute other certification processes as appropriate that are .different from those
adopted by CMBDA upon the approval by UMTA.
The Regional Transportation District reserves the right to accept or reject a firm's
certification from other than CMBDA as a certified DBE/WBE on RTD projects. In
addition, the RTD reserves the right to temporarily certify a DBE or a WBE while
certification from CMBDA is in process. In addition to CMBDA's certification forms,
the RTD will utilize *f9CFR, Part 23's Schedule A form as its basis for temporarily
certifying DBEs and WBEs.
REQUIREMENTS, TERMS AND CONDITIONS OF ATTACHMENT A
1. A DBE is defined as a small business concern that is owned and controlled by
socially and economically disadvantaged individuals. These socially and economically
disadvantaged individuals must own 51% of the business, and they must control the
management of the business. Socially and economically disadvantaged individuals
include Black Americans, Hispanic Americans, Asian-Pacific Americans, Native
Americans, Asian-Indian Americans, or any other minorities or individuals found to be
disadvantaged by the RTD or the Small Business Administration (SBA) pursuant to
Section 8(a) of the Small Business Act. _
Women Business Enterprise or WBE means small business concerns that is at least
51 percent owned, controlled and operated by one or more women.
2. A bidder or proposer who fails or refuses to complete and return the applicable
enclosures to this Attachment shall be deemed non-responsive and will not be awarded
the contract. Where a bidder or proposer intends to attain its goal for DBE/WBE
participation by subcontracting or use of a joint venture, it shall complete and submit
the following enclosures: (1A and/or IB) DBE/WBE Affidavit; (2) Schedule of DBE/WBE
participation; and (3) the Letter of Intent to Perform as a Subcontractor.
3. Enclosure (5) must be completed to indicate the equal employment status of the
U. All bidders or proposers; excluding construction contracts, are required to submit
written assurance of meeting RTDs DBE/WBE contract goals in their, bids or proposals.
In construction contracts, within five days after bids have been opened, RTD will
require all bidders to submit: (1) names and addresses of certified DBE/WBE
participating subcontractors, (2) a description of the work they are to perform and; (3)
the dollar value of each proposed certified DBE/WBE subcontract(s). In order to be a
responsive bidder/proposer, a contractor must meet the specified DBE/WBE contract
goal or demonstrate sufficient good faith, efforts to do so. Meeting the DBE/WBE
contract goals or making sufficient good faith efforts to do so, no less than meeting
technical specifications or complying with bid or proposal procedures, is a necessary
condition of .responsiveness. Among responsible and responsive bidders that is, those
bidders that meet the DBE/WBE requirements or demonstrate good faith efforts to do
so among other things the bidder offering the lowest price, if that price is reasonable
will be awarded the contract.
5. The contractor's commitment to the percentage of certified DBE and WBE
utilization during the the term of this contract will be stated in the DBE/WBE
6. The contractor's commitment to a specific goal is to meet DBE and WBE
objectives and is not intended and shall not be used to discriminate against any qualified
company or group of companies.
7. The contractor's attainment of the Specific Contract Goal for DBE and WBE
utilizatiSn is required by this Attachment. If the contractor has exhausted every
reasonable means to meet the RTD specified goal and has failed to do so, it will furnish
documentation of its good faith efforts to justify a grant of relief from the goal set
forth in this Attachment. The contractor shall complete Enclosure (4), DBE/NVBE
Unavailability Certification as evidence of its good faith efforts which documentation
will also include assurances that the contractor has done the following as a minimum:
Good Faith Efforts
a) Notified community organizations and minority focus media that the
contractor has subcontractor opportunities available and maintenance of
records (all dates should be recorded) of the organizations' responses.
b) Maintained a file of the names and addresses of each minority subcontractor
referred to it, and action taken with respect to each such referred
c) The contractor will avail himself or herself of grantee technical assistance
and resources, and will disseminate its DBE policy externally by notifying
and discussing it with ail subcontractors and suppliers (by advertising in
news media, specifically including minority news media).
d) Engaged in specific ongoing personal (both written and oral) recruitment
efforts directed at DBE/WBE contractor organizations, DBE/WBE recruit-
ment organizations, and DBE/WBE assistance organizations.
e) Whether the contractor selected portions of the work to be performed by .
certified DBEs and/or WBEs in order to increase the likelihood of meeting
the DBE/WBE goals (including, where appropriate, breaking down contracts
into economically feasible units to facilitate DBE/WBE participation).
f) Whether the contractor provided interested DBEs and/or WBE as required by
the goal set with adequate information about the plans, specifications and
requirements of the contract.
g) Whether the contractor provided written notice to a reasonable number of
specific certified DBEs and/or WBEs as required by the goal set in this
contract/proposal that their interest in the contract was being solicited, in
sufficient time to allow the DBEs/WBEs to participate effectively.
h) Whether the contractor acted in good faith with interested DBEs/WBEs not
rejecting DBEs/WBEs unqualified without sound reasons based on a thorough
investigation of their capabilities.
The DBE Coordinator will review the data submitted under this section, to determine
whether the DBE/WBE requirements have been satisfied through good faith efforts.
For further guidance and additional steps to take concerning the good faith efforts, see
49 CFR 23, Federal Register, 10/1/82, Vol. 46, No. 80, final rule, 4/27/81, as amended
by Vol. 48, No. 141, final rule, 8/22/83.
8. A prime contractor must make a good faith effort to replace a defaulting
DBE/WBE with another certified DBE/WBE. The prime contractor must notify the
RTD's, Disadvantaged Business Coordinator immediately of the DBE's/WBE's inability to
perform^and the contractor's intent to obtain a substitute certified DBE/WBE. If the
contractor obtains a substitute DBE or WBE, the contractor will notify the RTD's
Disadvantaged Business Coordinator and the DBE and/or WBE and provide copies or
descriptions of new or amended contracts and a completed certification form for each
new DBE/WBE and any applicable certificate of good faith effort as stated in item 7 of
9. The RTD, through its Disadvantaged Business Coordinator, will review the con-
tractor's DBE/WBE efforts during performance of the contract. If the contractor meets
its goal or if the contractor demonstrates that every reasonable effort has been made
to meet its goal, the contractor shall be presumed to be in compliance.
Where the Disadvantaged Business Coordinator finds that the contractor has failed to
comply with the requirements of this Attachment, the RTD's Contracting Officer will
be so informed and will in turn notify the contractor, who shall, after receipt of such
notice, immediately take corrective action. If the contractor fails or refuses to comply
promptly, the RTD's Contracting Office shall issue an order stopping all or part of the
payment and/or work until satisfactory corrective action ;has been taken. No part of
the time lost due to any such stop orders shall be made the subject of claim for
extension of time or for excess costs or damages by the contractor. Where the
contractor is found to have failed to exert a good faith effort to involve certified
DBEs/WBEs in the work provided, the RTD may declare that the contractor is ineligible
to receive further RTD contracts for a period of up to three (3) years.
10. The contractor will maintain those records and documents for three (3) years
following performance of the contract which indicate compliance with this Attachment.
These records and documents, or copies thereof, will be made available at reasonable
times and places for inspection by any authorized representative of the RTD and will be
submitted to the RTD upon request, together with any other compliance information
which such representative may require.
11. Nothing in this Attachment shall be interpreted to diminish the present contract
compliance review and complaint programs.
RTD AFFIRMATIVE ACTION/DISADVANTAGED
BUSINESS ENTERPRISES (DBE) PROGRAM
A. Overview of RTD's Policy
It is the policy of RTD to provide all certified DBE's maximum
opportunity to participate in its construction activities, professional
service contracts, and in the procurement of common goods and
services. A DBE is a small business concern that is owned and
controlled by socially and economically disadvantaged individuals as
defined on page 6 of this Attachment.
To accomplish this objective, RTD requires. that all bidders and
proposers complete and return applicable enclosures to Attachment A.
This obligates the contractor to meet the contract goals specified in
this Attachment or to assert a good faith effort to do so.
1. Overall Goals
RTD sets annual goals for work to be performed by DBE's. RTD
will specify a minimum required percentage of participation in
contracts with subcontracting opportunities to meet Federal
requirements and to provide for participation of certified DBE's.
The expected percentage of certified DBE participation may
vary from contract to contract depending on, availability of
DBE's in any given line of work. However, RTD believes the
overall goals to be obtainable over time with the assistance of
the federal government, the business community, and DBE
2. Contract Goals
RTD has specified a DBE goal of ________;______for the work to
be performed under this contract.
DBE participation is determined by the dollar value of work
performed by certified DBE's compared to the total value of all
work performed under this contract and/or, where a joint ven-
ture is entered into, by the percentage of the work which the
parties agree will be shared by a certified DBE for the comple-
tion of the project.
RTD's commitment to a specific goal is to meet DBE objectives
and is not intended and shall not be used to discriminate against
any qualified company or group of companies.
B. Requirements of Attachment A
The contract will be awarded to that responsible and responsive
bidder/proposer who offers the lowest reasonable price. A bidder/
proposer who fails or refuses to complete and return the applicable
enclosures to this Attachment will be deemed non-responsive and will
not be awarded the contract. The contractor's commitment to the
percentage of certified DBE utilization during the term of this
contract will be stated in the DBE Affidavit (Enclosure A).
1. Meeting Contract Goals
The bidder/proposer may meet RTD contract goals
a. by having certified DBE status;
b. by subcontracting portions of the work to certified DBE's;
c. by being an RTD-certified joint venture with other certi-
2. The Responsive Bidder/Proposer
To be considered a responsive bidder/proposer, the following
three conditions must be met:
a. when a contract goal is specified for a project, a respon-
sive bidder/proposer must meet the goals referred to in the
bid documents and Attachment A or assert a good faith
effort to attain these goals;
b. a responsive bidder/proposer must comply with bid or
proposal procedures/requirements; and
c. a responsive bidder/proposer must meet the technical
specifications of the project.
3. The Responsible Bidder/Proposer
A responsible bidder/proposer is one who has the necessary
equipment, facilities, resources, judgment and skill to perform
the task. In addition, the responsible bidder/proposer must
submit the following "written intent to comply" with RTD's DBE
RTD requires all bidders on construction contracts to submit
Enclosure 2 within five days after bids have been opened.
Enclosure 2 must contain the following information:
a. names and addresses of certified DBE participating
b. a description of the work they are to perform;
c. the dollar value of each proposed certified DBE contract;
d. documentation of good faith efforts, if applicable.
All Other Contracts
All other bidders and proposers are required to submit written
assurance of meeting RTD's DBE contract goals at the time of
submission of bids or proposals. This includes the following:
a. names and address of certified DBE participating
b. a description of the work they are to perform;
c. the dollar value of each proposed certified DBE contract;
d. documentation of good faith efforts, if applicable.
4. Enclosures 1-5
The contractor's commitment to the percentage of certified DBE
utilization during the term of this contract will be stated in the
DBE Affidavit, Enclosure 1A. Bidders/proposers must complete
and return the applicable enclosures as follows:
Enclosure 1A. The DBE Affidavit must be filed by any non-DBE
prime contractor to acknowledge the percentage of DBE
Enclosure IB. The DBE Affidavit must be filed by all DBE prime
contractors to verify DBE status.
Enclosure 2. Schedule of DBE Participation must be filed by all
prime contractors who intend to use DBE subcontractors.
Enclosure 3. Letter of Intent to Perform as a Subcontractor
must be filed by the prime contractor and signed by DBE
Enclosure 4 (for construction projects). Unavailability Certifica-
tion is to be filed by the prime contractor within five days after
bids have been opened, along with additional documentation of
good faith efforts.
Enclosure 5. Employment Data Form, defines the make-up of
the company's work force and must be filed by every prime
5. Joint Ventures
A joint venture is an association of two or more persons,
partnerships, corporations, or any combination thereof, formed
to carry on a single business activity which is limited in scope
If the joint venture is participating in a contract in which DBE
goals have been set, then the joint venture must be certified by
RTD's DBE Office prior to award. The DBE component of the
joint venture musts
a. meet the requirements for DBE as stated in the definitions
section of this Attachment;
b. have a share in the ownership, control and management
responsibilities, risks and profits of the joint venture and in
order for the joint venture to be considered a DBE joint
venture have a 51% share of the joint venture;
c. be responsible for a clearly defined portion of the work to
The distribution of money reflects the DBE's contribution to the
joint venture. Thus, the dollar amount credited to the DBE goals
set forth in the contract will be the same as the amount which
the DBE component contributes to and is paid for its participa-
tion in the joint venture.
1. Certification Agency
RTD will act as certification agency for DBE's. RTD reserves
the right to institute other certification processes as appropriate
upon approval by UMTA. In addition, RTD reserves the right to
accept or reject a firm's certification from other agencies.
2. Good Faith Efforts
In order to award a contract'to a bidder/proposer that has failed
to meet the DBE contract goals as stated in that specific
contract, RTD will determine whether the bidder/proposer made
"good faith efforts and actively and aggressively sought DBE's to
meet those goals.
RTD, through its DBE Office, will review the data submitted to
determine whether the DBE requirements have been satisfied
through good faith efforts. RTD will base its judgment on the
following list, not intended to be exclusive or exhaustive, to
determine whether good faith efforts have been made by
a. notified community organizations and disadvantaged busi-
ness media that the contractor has opportunities available
for subcontractors and maintained records, including all
dates, of the responses of those organizations;
b. maintained a file of the names and addresses of each DBE
subcontractor referred to it and the action taken with
respect to each such referred subcontractor;
c. utilized RTD's technical assistance and resources and dis-
seminated its DBE policy externally by notifying and
discussing it with all subcontractors and suppliers through
advertisements in disadvantaged business news media;
d. engaged in specific ongoing personal recruitment efforts,
both, written and oral, directed at DBE contractor organi-
zations, DBE recruitment organizations and DBE assistance
e. selected portions of the work to be performed by certified
DBE's in order to increase the likelihood of meeting DBE
goals, including where appropriate, breaking down con-r
tracts into economically feasible units to facilitate DBE
f. provided interested DBE's adequate information about
plans, specifications and requirements of the contract;
g. provided written notice to a reasonable number of certified
DBE's that their interest in the contract was being solici-
ted in sufficient time to allow DBE's to participate
h. contacted and effectively used the services of available
DBE organizations, contractors' groups, state and local
offices, etc., that have knowledge of available DBE firms
or the means to locate such firms. .
RTD will review the documentation submitted under this section
to determine whether the DBE requirements have been satisfied
through good faith efforts. For further guidance concerning
good faith efforts, refer to 49 CFR 23, Federal Register,
3. Defaulting DBE Subcontractors
A contractor must make a good faith effort to replace a
defaulting DBE with another certified DBE. The prime contrac-
tor must notify RTD's DBE Office immediately of the DBE's
inability to perform and of his intent to obtain a substitute
certified DBE. The substitute DBE must receive prior approval
by RTD. When the contractor obtains a substitute DBE, the
contractor will notify the RTD liaison and provide copies or
descriptions of new or amended contracts and a completed