Reassessing Rhetorical Abstractions and Planning Mechanisms 1 
Daniel D. Suthers 
Department of Computer and Information Science 
University of Massachusetts 
Amherst, Massachusetts 01003 
suthers@cs.umass.edu 
Abstract 
The utihty of rhetorical abstractions and certain text 
planning mechanisms were assessed from the stand- 
point of accounting for how an explainer chooses and 
structures content under multiple perspectives to meet 
knowledge communication goals. This paper discusses 
ways in which they were found to be inadequate, ar- 
gues for greater emphasis on an epistemological level of 
analysis, and proposes a mixed architecture matching 
computational mechanisms to the explanation planning 
subtasks they are suited for. 
Introduction 
Our research is concerned with explanation in its 
broad sense, as "the act or process of making plain 
or comprehensible; elucidation; clarification" (Ameri- 
can Heritage Dicgionary). In many physical science do- 
mains, an explanation can be based on a variety of mod- 
els of the phenomenon being explained. These models 
differ on the type of properties and relationships em- 
phasized; what ontology is used and whether the phe- 
nomenon is described at a macroscopic, microscopic, or 
atomic granularity; what factors are ignored or assumed 
to be constant; the use of general statements vs. con- 
crete examples; and in general on what concepts func- 
tion as primitives, providing the basis for understand- 
ing the topic phenomenon. Selection of an appropriate 
model of the topic is an important aspect of content 
selection which impacts on the interlocutor's compre- 
hension of the explanation and on its appropriateness 
for his or her purposes. We use the term perspeetlve 
to refer to an abstract characterization of the kind of 
knowledge provided by a class of models. Our notion 
of perspective is a composite of distinctions made by 
Falkenhainer & Forbus (submitted), Stevens & Collins 
(1980) Stevens & Steinberg (1981) and White & Fred- 
eriksen (1989). 
1 Copyright @ 1990, Daniel D. Suthers Permission granted to 
copy for non-profit academic purposes. 
Most existing work in explanation and text planning 
has been directed at other problems, and hence has 
utilized single-perspective knowledge bases to simplify 
the research. (Notable exceptions include McCoy, 1989 
and McKeown et al. 1985.) A number of such re- 
search efforts emphasize rhetorical abstractions for the 
analysis of natural explanations and for expressing a 
theory of explanation (Hovy, 1988; Mann & Thomp- 
son, 1986; Maybury, 1988; McKeown, 1985; Moore, 
1989). A variety of mechanisms for selecting and or- 
ganizing the content of explanations have also been ex- 
plored. This includes schema filling (McKeown 1985) 
and structure matching (Hovy 1988), graph traversal 
algorithms (Paris & McKeown 1986), and top-down ex- 
pansion planning (Cawsey, 1989; Moore, 1989). 
Our own work on choosing explanatory content from 
multiple-perspective knowledge bases has uncovered 
some limitations of rhetorical abstractions, and led us 
to question previous applications of the computational 
mechanisms listed above. The purpose of this paper is 
to present our perception of the roles and limitations of 
these items, and suggest some alternatives. (We do not 
emphasize our work on perspective and content organi- 
zation here: see Suthers & Woolf, 1990.) We begin with 
an example, used to illustrate some of the problems. 
An Example 
The following example explanation will be used to 
illustrate our points. The domain is elementary elec- 
tricity. We emphasize the communication of an under- 
standing of concepts such as "charge", "current", and 
"capacitance" within the context of qualitative reason- 
ing about the behavior of simple circuits and their com- 
ponents. The explanation is an edited version of a hu- 
man protocol. 
QI: How does a capacilor s¢ore charge? 
El: A capacitor 
E2: can be thought of as two fiat metallic plates 
137 
I j- Background -~ l .'C°nstit ...i 
I ~Attributive? 
1 2 3 4 
{ ~. Conclusion ? -......~ 
 se-Effect 
I F, Illustration 
5 6 7 8 9 
Figure 1: One possible RST analysis (partial) 
E3: situated close to each other and parallel to each 
other, 
E4: with air in between. 
E5: If you connect a capacitor to a battery, 
E6: as follows: 
I I + 
I I - 
E7: then positive charge builds up on one plate and neg- 
ative charge on the other, 
E8: until the voltage across the capacitor is equal to the 
voltage of the battery. 
£9: The charge stays there when you disconnect the bat- 
tery. 
One possible rhetorical analysis of the first explana- 
tion, using relations from McKeown (1985) and Mann & 
Thompson (1986), is given in Figure 1. (Whether or not 
this is an optimal analysis is not the point; rather we are 
concerned with the role of rhetorical abstractions and 
various characterizations of the content planning task 
• in accounting for the explanation.) The analysis does 
point out some features of interest. For example, we 
note that the explanation starts with Background ma- 
terial (El-E4) before proceeding to the primary expla- 
nation (E5-E9). The Background relation in this case 
describes the high level organization of the explanation. 
The question is how this relation should be manifest in 
the mechanism for generating the explanation. Also of 
interest are the explainer's determination that a process 
account of how the capacitor carries out its function is 
appropriate; use of an abstract structural model of the 
capacitor, simplified to what is needed to support the 
process account; and use of a concrete situation for the 
enablement condition of the process. 
Problems with Rhetorical Abstractions 
Rhetorical relations were an important development 
in explanation research, since they provided a first pass 
at abstractions for a general description of explanatory 
structure, and in bringing various roles of the parts 
of explanation into the foreground, pointed out phe- 
nomena in need of further study. They are also useful 
via their "relational propositions" (Mann & Thomp- 
son 1983), for conveying propositional information im- 
plicitly in the structure of the text, hence reducing its 
length and redundancy. However, we claim that rhetor- 
ical abstractions fail to make the necessary distinctions 
for further advances in a theoretical understanding of 
explanation. 
Potpourri. Rhetorical abstractions are descriptive of 
explanatory tezt, i.e. the end product of some expla- 
138 
nation generation process, and so describe with one de- 
vice structure due to a variety of distinct knowledge 
sources bearing on such a process. These knowledge 
sources operate on different levels of information, and 
hence need to be separated in a theory of explanation 
generation. For example, (drawing on relations in McK- 
eown, 1985 and Mann & Thompson~ 1986) some corre- 
spond directly to the fundamental structure of domain 
objects and processes (e.g. Constituency and Causal- 
ity), while others are about derived relations between 
concepts which may vary according to the context (e.g. 
Comparison). Relations such as Amplification, Back- 
ground, Evidence, and Illustration are primarily about 
relationships between propositions which arise in part 
out of consideration of what the interlocutor knows and 
needs to know to better grasp a point. Illocutionary 
acts are involved in the relations as well (most blatantly, 
Concession; others are not themselves illocutionary acts 
but only make sense in the context of certain such acts). 
Finally, relations such as Topic and Conclusion appear 
to be due to conventions governing writing style which 
direct focus of attention. Grosz & Sidner (1986) made 
similar criticisms from the standpoint of characterizing 
discourse coherence. They suggested that each rhetori- 
cal relation combines domain information with certain 
general relations between propositions, between actions, 
and between intentions. 
Implicit Features Unaccounted For. Rhetorical 
abstractions are also inappropriate for a theory of con- 
tent selection because, in describing the final text, they 
fail to identify important relations between the chosen 
content and external material, such as what is known 
about the user, or information left o~$ of the explana- 
tion. For example, E5-E6 is more specific than is neces- 
sary and includes a concrete example. A more general 
and accurate way to state the condition for initiation of 
the charging process would be "If a voltage is applied 
to the plates ...'. However, the explainer has opted to 
replace this with one of the many particular configura- 
tions which meet the condition. A rhetorical analysis 
of the text canno$ even $ell us $ha~ ~his has happened, 
let alone why, because it does not describe relations be- 
tween the contents of the text and what is no~ included, 
viz., other models of the process being described. It can 
only report that an illustration is being used. Another 
example is provided by E9. Retention of charge when 
the voltage is removed is what is meant by "storage" in 
this case, so the fact expressed in E9 is essential to an- 
swering the question. Rhetorically, we can only identify 
relationships E9 has to the rest of the text, e.g. that it is 
a Conclusion. This does not illuminate the relationship 
between its content and the goal of the explanation. 
Epistemological Analysis. The success of an expla- 
nation is primarily a function of choice and organiza- 
tion of knowledge. Hencc, to account for how these 
choices further knowledge communication goals, one 
must examine explanation in part from an epistemo- 
logical standpoint. Such an analysis examines how ex- 
planations are guided by: 
• the types of knowledge in a given domain, and its 
logical and etiological structure (Rissland, 1978); 
• the types of knowledge which, in principle, could 
fulfill a given request for information; 
• the role of an individual's knowledge in under- 
standing new concepts and situations, and hence 
in understanding a given explanation (Paris, 1987); 
and 
• the ways in which individuals are willing or able 
to undertake conceptual change (Goldstein, 1979; 
Hewson, 1981; White & Frederiksen, 1989; van- 
Lchn, 1987). 
As discussed in Suthers (1989), most previous work has 
offered solutions to the subproblems of explanation in 
the form of mechanisms and data structures which are 
in part the result of, rather than the expression of, epis- 
tcmological considerations. Epistemological problems 
have been avoided through direct selection of content 
based on well-formulated queries; the simplicity of the 
knowledge bases used, which only permit one way of 
discussing each topic; and through implicit conflation 
of epistemological constraints on organization of the ex- 
planation with those of rhetorical and linguistic origin. 
A major goal of our research (Suthers & Woolf, 1990) 
is an explicit theory of the epistemological structure of 
the activity of explaining. 
Roles of Computational Mechanisms 
In this section we illustrate how inclusion of back- 
ground material and the use of multiple perspectives 
pose problems for various mechanisms in the literature, 
and suggest a mixed architecture solution. 
Structure Matching. By "structure matching" we 
mean methods where abstract descriptions of the struc- 
ture of explanations are matched to a collection of 
propositions (or other content to be expressed) in order 
to organize this material. This includes bottom-up com- 
position of structural units (Hovy, 1988) and schema 
filling (McKeown, 1985). We question their adequacy 
for accounting for the prerequisite structure of expla- 
nations, such as the Background relation of the exam- 
ple. In our view, the explanation is organized this way 
139 
because the explainer recognized in his process model 
(expressed in E5-Eg) concepts the interlocutor may not 
be familiar with (the parts of a capacitor), and then 
added material prerequisite to understanding the pro- 
cess explanation (the structural description expressed in 
El-E4). The background material is not automatically 
part of the relevant knowledge pool for this question, 
and structure matching methods leave choice of con- 
tent and perspective to other mechanisms. Suppose, 
then, that some other mechanism accounts for inclu- 
sion of the background in the pool. It is included as 
background by virtue of relationships between the inter- 
locutor's assumed knowledge state and the conceptual- 
izations contained in the first attempt at a knowledge 
pool. Pattern matching techniques which are ignorant 
of such relationships and see only the composite pool 
would be unable to identify the part of the knowledge 
pool which plays the role of "background", and account 
for placement of background before primary material. 
Graph Traversal. These are algorithms for selec- 
tively following links in a knowledge base, with the path 
so traced out providing the content and structure of 
the explanation (Paris & McKeown 1986). Such meth- 
ods model how an explanation exploits the structure of 
knowledge. They implicitly embody the heuristic that 
it will be easier for the interlocutor to reconstruct the 
knowledge if it is presented such that each unit is in- 
troduced in relation to the previous unit. For example, 
parts of an object are introduced in relation to con- 
taining or adjacent parts, and process descriptions or- 
ganized to follow temporal and causal relations in the 
forward direction. However, graph traversal is limited 
to modeling local organization, or global organization 
which is a composite of local choices. Traversal meth- 
ods don't naturally extend to global organization which 
occurs at a higher level of abstraction than the links 
followed, e.g. the presentation of a coherent structural 
model before a process account begins. 
Top-Down Goal Expansion. Top-down expansion 
of a discourse goal (Cawsey, 1989; Moore, 1989) is a 
stronger candidate for a uniform mechanism for expla- 
nation, integrating content selection and structuring. 
The background problem can be handled with precon- 
ditions on plan operators, as Cawsey does. It simplifies 
modeling explanations such as our example if one can 
specify the kind of background knowledge required in 
preconditions at the highest level of plan operators. To 
illustrate, consider a rhetorical plan operator which con- 
tains an optional satellite for Background but does not 
specify what constitutes background knowledge. Ex- 
pansion of the satellite would have to be predicated on 
a comparison of the content selected by expansion of 
the nucleus with the user model. This decision could 
not be made at the time the operator is selected by the 
planning system, since the knowledge pool for the nu- 
cleus would not have been selected yet. Instead, one 
would have to place the satellite decision on hold, ex- 
pand the nucleus, collect together the knowledge se- 
lected at the leaves of its expansion, and perform the 
comparison before deciding on the satellite. (One could 
make the decisions concerning the need for background 
locally to each leaf, avoiding the need for a high level 
decision depending on knowledge selected at many lo- 
calities. But then one could not model the structure of 
explanations such as this one, where the need for prereq- 
uisite material is anticipated and provided in advance 
as a coherent model, rather than as an interruption to 
the flow of the process description.) Then, if it was 
decided that some background was required, expansion 
of the satellite would have to occur under a binding of 
some variable in the satellite to the concepts for which 
background is required. 
With regards to perspective, some problems emerge. 
Content selection in top-down expansion can be influ- 
enced by the current perspective if some mechanism for 
sharing perspective decisions across the expansion tree 
is provided. However, neither the choice of perspec- 
tive nor the actual mechanism by which it influences 
content selection are modeled appropriately by top- 
down expansion. Choice of perspective tries to balance 
the (sometimes conflicting) constraints of adequacy and 
comprehensibility. Adequacy constraints come from ex- 
amination of the informative goals (McKeown et al. 
1985). Comprehensibility requires answering the ques- 
tion: given the concepts which have been used in the 
dialogue so far, and/or which the explainer has evidence 
the interlocutor is familiar with, what other concepts 
are also likely to be familiar? This suggests a strength- 
of-association mechanism (McCoy 1989), which we com- 
ment on further in the next section. As Hovy (1990) 
points out, top-down planning is prescriptive, and does 
not handle conflicting goals easily. The influence of per- 
spective on content selection involves what Hovy calls 
restrictive planning: some perspective goals, once gen- 
erated, remain active throughout a span of discourse, 
and operate as preferences applied to choice points in 
content selection. They cannot be erased once satisfied, 
and they may change dynamically. Finally, McDonald 
& Pustejovsky (1985) point out that a uniform mecha- 
nism may not be desirable, as it incorrectly implies that 
all information is equally available during each point of 
the text planning process. A principled match of dis- 
tinct computational mechanisms to each subtask of the 
140 
(Query) --> Identify 
} 
V 
Refine < .... 
t 
v 
Retrieve 
f 
V 
Modify ..... 
t 
V 
Order --> (Realize) 
Figure 2: Content planning tasks 
explanation planning t&sk is one way to express one's 
theory of what kind of process explanation is. For these 
reasons, we postulate a separate level of planning for co- 
ordinating perspective decisions. 
A Mixed Architecture. In our current approach, 
planning occurs at two granularities: selection and re- 
trieval of coherent packages of knowledge similar to 
Suthers' (1988a,b) "views" or Souther, Acker, Lester, 
& Porter's (1989) "viewpoints"; and editing and order- 
ing the propositions and examples which make up these 
views. Planning at the granularity of views is concerned 
with purely epistemological constraints on identification 
of appropriate content, including choice of perspective 
and prerequisite explanations. At the finer granular- 
ity, further epistemological constraints governing the 
comprehensibihty of the explanation are apphed, and 
rhetorical and linguistic constraints play a role as well. 
The question at hand is: wA~z~ sort of t~zsl~ is cor~ter~ 
plar~r~ir~g ? We postulate several, and comment on possi- 
ble computational mechanisms for each. Figure 2 gives 
the rough relationships between the different tasks. The 
process is reminiscent of case-based design, where one 
identifies and refines design specifications, and retrieves 
and reconfigures a previous solution to fit the circum- 
stances. 
An explainer must first identify at least some of its 
goals. These are of two types, as discussed above: pre- 
scriptive informative goals, and restrictive goals such as 
comprehensibihty. We mentioned that the latter sug- 
gests a concept association mechanism. This could be 
done by activating weighted links between concepts, or 
though intermediate frames of attribute weights, as in 
McCoy (1989). However, one would have to install a po- 
tentially combinatorial number of associations between 
concepts. Because we wish to make an epistemological 
theory explicit and implement it with an abstract inter- 
face to the knowledge base, we are investigating use of 
a mechanism akin to prototype induction, to generate 
an abstract description of the desired perspective from 
the user model and dialogue history. 
At the view granularity, content selection may be seen 
as refinement of explanatory goals into a specification 
of the appropriate addition to the relevant knowledge 
pool. This refinement includes consideration of what 
type of knowledge, in principle, could fulfill the infor- 
mative goal; of the concepts the interlocutor is likely to 
understand; and of the models which have already been 
shared in the dialogue. We are attempting to treat this 
refinement task as top-down planning, though our work 
has not progressed fax enough to comment further on 
this. 
Retrieval requires a knowledge-base specific mech- 
anism: its only theoretical importance to us is that it 
correctly operationalize the dimensions used to describe 
the desired view (Suthers 1988a,b). We are examining 
a variant of compositional modehng (Falkenhalner & 
Forbus, submitted) for this purpose. 
Once a knowledge pool is available, some data driven 
activities occur at a finer granularity, resulting in mod- 
ification of the relevant knowledge pool. This includes 
filtering activities, such as removing particular proposi- 
tions likely to be famihax to the interlocutor; and aug- 
menting activities, such as illustrating abstract state- 
ments with examples. These are oppor~g~istic planning 
tasks, that can be approached with critics which match 
to the knowledge pool and user model, and specify re- 
placements or deletions to be made. As shown in fig- 
ure 2, data driven operators may also reinvoke content 
planning at the refinement level to access material in 
a different type of model. For example, we have seen 
how process propositions may involve use of structural 
concepts the interlocutor is not likely to understand, 
causing the explainer to plan a prerequisite explana- 
tion. An explicit record of the prerequisite relations 
between views is created, important for ordering the 
explanation. 
Ordering is sensitive to prerequisite and illustration 
links installed by the modification processes. Hence fig- 
ure 2 should not be interpreted as a claim that order- 
ing considerations do not arise during content selection. 
The ordering task involves two kinds of processes. One 
embodies epistemological constraints by ezploi~ir~g ex- 
isting structure in the knowledge pool. As discussed 
previously, techniques for traversing links in the knowl- 
edge pool apply here. The result will likely be a par- 
tial ordering. Further ordering requires imposition of 
structure for rhetorical, linguistic, and pictorial reasons 
141 
during realization (generation of text and graphics). 
Matching of rhetorical patterns to the knowledge pool 
may be more appropriate at this later stage. 
In summary, we have argued for a mixed architec- 
ture matching prescriptive, restrictive, and opportunis- 
tic mechanisms to explanation subtasks. Coordination 
of these diverse processes may, at the implementation 
level, require an agenda control mechanism as in Niren- 
burg, Lesser, & Nyberg (1989). 
Conclusions 
Single-perspective knowledge bases, i.e. those which 
provide only a single conceptual basis for a given de- 
scription or explanation, have dominated existing work 
in planning expository text. This research has over- 
emphasized rhetorical abstractions as the basis for the- 
ories of content planning, and used computational for- 
malisms such as top-down goal expansion, traversal al- 
gorithms, and opportunistic structure matching which, 
taken alone, fail to fully account for the search for ap- 
propriate conceptualizations during content selection. 
We suggest a greater emphasis on epistemological ab- 
stractions, and viewing content selection in terms of 
identification, refinement, retrieval, modification, and 
ordering tasks. Rhetorical abstractions retain a place 
in initial analyses of explanations, to point out features 
in need of further study, and in the later stages of text 
planning, to further constrain a partial ordering of con- 
tent and implicitly convey content via textual structure. 
Finally, the devices of top-down goal expansion, traver- 
sal algorithms, and structure matching retain potential 
utility for high level content planning, exploiting the 
structure of knowledge when ordering an explanation, 
and further ordering an explanation on a rhetorical ba- 
sis, respectively. 
Acknowledgments 
The author has been supported by the National Sci- 
ence Foundation under grant number MDR 8751362, 
and by Apple Computer, Inc., Cupertino, CA. Par- 
tial support was also received from the Office of Naval 
Research under a University Research Initiative Grant, 
contract no. N00014-86-K-0764. Thanks are due to my 
advisors, Edwina Rissland and Beverly Woolf; to Klaus 
Schultz for his expert explanations; to Matthew Cornell 
for assistance with the research, and to the reviewers for 
comments. 

References 
Cawsey, A. (1989). Generating Ezplanatory Discourse: 
A Plan-Based, Interactive Approach. Ph.D. Disserta- 
tion, University of Edinburgh. 
Falkenhainer £~ Forhus (submitted). Compositional 
Modeling: Finding the right model for the job. Sub- 
mitted to Artificial Intelligence. 
Goldstein, I. (1979). The genetic graph: a representa- 
tion for the evolution of procedural knowledge. Inter- 
national Journal of Man.Machine Studies, vol. 11, pp. 
51-77. 
Grosz, B. J. & Sidner, C. L. (1986). Attention, in- 
tention, and the structure of discourse. Computational 
Linguistics, vol. 12, no. 3, pp. 175-204. 
Hewson, P. W. (1981). A conceptual change approach 
to learning science. European Journal of Science Edu- 
cation, vol. 3, no. 4, pp. 383-396. 
Hovy, E. H. (1988). Planning coherent multisentential 
text. Proc. ~6th Meeting of the A CL, Buffalo, New 
York; Reprinted as ISI/RS-88-208, Information Sci- 
ences Institute, Marina del Rey, California. 
Hovy, E. H. (1990). Pragmatics and natural language 
generation. Artificial Intelligence, vol. 43, no. 2, pp. 
153-197. 
Mann, W. C., &: Thompson, S. A. (1983). Relational 
Propositions in Discourse. ISI/RR-83-115, Information 
Sciences Institute, University of Southern California, 
Marina del Rey, California. 
Mann, W. C., & Thompson, S. A. (1986). Rhetor- 
ical structure theory: Description and construction of 
text structures; Proceeding8 of the NA TO Advanced Re- 
search Workshop on Natural Language Generation, Ni- 
jmegen, The Netherlands, August 19-23, 1986. 
Maybury, M. (1988). Explanation rhetoric: The rhetor- 
ical progression of justifications. Proc. AAAI-88 Work- 
shop on Ezplanation, St. Paul, August 22, 1988, pp. 
16-20. 
McCoy, K. F. (1989). Generating context-sensitive re- 
sponses to object-related misconceptions. Artificial In- 
telligence, vol. 41, no. 2, pp. 157-195. 
McDonald, D. D. & Pustejovsky, J. D. (1985). 
Description-directed natural language generation. 
Proc. 9th Int. Joint Conf. on Artificial Intelligence, 
Los Angeles, August 18-23, 1985, pp. 799-805. 
McKeown, K. It. (1985). Discourse strategies for gen- 
erating natural language text. Artificial Intelligence, 
vol. 27, no. 1, pp. 1-41. 
McKeown, K. It., Wish, M., ~ Matthews, K. 
(1985). Tailoring explanations for the user. Proc. 9th 
Int. Joint Conf. on Artificial Intelligence, Los Ange- 
les, August 18-23, 1985, pp. 794-798. 
Moore, J. D. (1989). A Reactive Approach to Ezpla- 
nation in Ezpert and Advice-giving Systems. Ph.D. 
Dissertation, University of California, Los Angeles. 
Nirenburg, S., Lesser, V., & Nyberg, E. (1989). 
Controlling a language generation planner. Proc. 11th 
Int. Joint Conf. on Artificial Intelligence, Detroit, 
pp. 1524-1530. 
Paris, C. L. (1987). Combining discourse strategies to 
generate descriptions to users along a Naive/Expert 
spectrum. Proc. lOth Int. Joint Conf. on Artificial 
Intelligence, August 1987, Milan, Italy, pp. 626-632. 
Paris~ C. L. £~ McKeown, K. It. (1986). Discourse 
strategies for describing complex physical objects; Pro- 
ceedings of the NATO Advanced Research Workshop on 
Natural Language Generation, Nijmegen, The Nether- 
lands, August 19-23, 1986. Also published by Martinus 
Nijhoff, Dordrecht, 1987. 
Rissland, E. L. (1978). (Formerly Michener.) Under- 
standing Understanding Mathematics. Cognitive Sci- 
ence, vol. 2, no. 4. 
Souther, A. Acker, L., Lester, J., g~ Porter, B. 
(1989). Using view types to generate explanations in 
intelligent tutoring systems. Proc. Cognitive Science 
Conf., Montreal, 1989. 
Stevens, A. L. /k Collins, A. (1980). Multiple con- 
ceptual models of a complex system. In R. E. Snow, P. 
Fcderico, and W. E. Montague (Eds.), Aptitude, Learn- 
ing, and Instruction (Vo!. $). Hillsdale, N J: Erlbaum, 
1980. pp. 177-197. 
Stevens, A. L. ~ Steinberg, C. (1981). A typology of 
explanations and its application to intelligent computer 
aided instruction. Report No. 4626, Bolt Beranek and 
Newman Inc., Cambridge, MA. 
Suthers, D. D. (1988a). Providing multiple views of 
reasoning for explanation. Proc. InL Conf. on In- 
telligent Tutoring Systems, Montreal, June 1988, pp. 
435-442. 
Suthers, D. D. (1988b). Providing multiple views for 
explanation. Proc. AAAI-88 Workshop on Ezplana- 
tion, St. Paul, August 22, 1988, pp. 12-15. 
Suthers, D. D. (1989). Perspectives in Explanation; 
COINS Technical Report 89-24, Computer and Infor- 
mation Science, University of Massachusetts, Amherst. 
Suthers, D. D, & Woolf, B. P. (1990). Account- 
ing for the Epistemological Structure of Explanation. 
Spring Symposium on Knowledge Based Environments 
for Learning and Teaching, March 27-29, Stanford. 
Available as COINS Technical Report 90-36, Computer 
and Information Science, University of Massachusetts, 
Amherst. 
vanLehn, K. (1987). Learning one subprocedure per 
lesson. Artificial Intelligence, vol. 31, no. 1, pp. 1-40. 
White~ B. Y. & Frederiksen~ J. It. (1990). Causal 
Model Progressions as a Foundation for Intelligent 
Learning Environments. Artificial Intelligence, vol. 
42, no. 1, pp. 99-157. 
