The Classification of Research Questions

Author(s): J. T. Dillon
Source: Review of Educational Research , Autumn, 1984, Vol. 54, No. 3 (Autumn, 1984),
pp. 327-361
Published by: American Educational Research Association

Stable URL: https://www.jstor.org/stable/1170452

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

American Educational Research Association is collaborating with JSTOR to digitize, preserve and
extend access to Review of Educational Research

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 Review of Educational Research
Fall, 1984, Vol. 54, No. 3, Pp. 327-361

The Classification of Research Questions

J. T. Dillon

University of California, Riverside

ABSTRACT. What are the kinds of questions that may be posed for research?
A dozen schemes proposing to classify research questions are surveyed,
analyzed, and applied to the understanding and practice of inquiry. The
extent to which the various schemes account for questions found in
educational journals is estimated. Some principles and issues are identified
to stimulate work on the classification of research questions in education
and other enterprises of inquiry. On the whole, little is known about the
kinds of questions that may be posed for research.

Since the time that Socrates first exemplified their use, questions have been
thought essential to the pursuit of inquiry. Aristotle was the first to distinguish the
kinds of questions that together form the domain of inquiry, proposing that
knowledge consists in answers to questions. Recent work in philosophy of science
and erotetic logic proposes to conceive of scientific and scholarly inquiry as a
question-answering system (Gale, 1978; Hintikka, 1981; Meyer, 1980). But little
systematic work has been done to distinguish the kinds of questions that are asked
and answered in research. As Gale (1978) observed, the available conceptions of
science are restricted to "a narrow class of questions" (p. 321), while existing models
of erotetic logic "only cover a small fragment of the kinds of questions raised and
types of answers desired in scientific contexts" (p. 329).
What are the kinds of questions that may be posed for research? Taken as a
theoretical issue, the answer to this question is itself constituted of a classification
scheme. Taken as an empirical issue, the answer is constructed by a classification
scheme. Here a dozen schemes for research questions will be surveyed, analyzed,
and then applied to the understanding and practice of inquiry. In that way the
review might serve to stimulate systematic work on the classification of questions
for research in education and other enterprises of inquiry.
Two distinctions will clarify the grounds for reviewing these 12 schemes (and not
others). They propose to classify questions for research rather than other purposes
and contexts; and the questions they classify are general to research rather than
specific to some subjects. In these terms, schemes that classify the questions asked
by teachers or interviewers, for example, are not included; nor are schemes that
classify the topical content of questions asked in curriculum studies or experimental
physics or any other field. In short, the schemes classify questions for research
generally.
Put otherwise, the schemes come under review for an importance of their own.
By contrast to the numerous schemes for questions in other pursuits (see Dillon,
1982)-dozens for the classroom alone-these 12 represent the few of their kind
proposed for the universe of research questions. More important, in conceiving of

327

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

this domain they distinguish the categories that serve as terms for a theory of
inquiry, and they define a framework for thought and action concerning research
questions. To that same purpose, this review exhibits the kinds of questions they
distinguish, examines their classificatory character, and assesses their uses for the
pursuit of inquiry.

Survey of Schemes
The schemes are surveyed in order of narrative convenience, save for the first.
The first to be presented is the first to have been constructed-that of Aristotle.
The second has been constructed by the reviewer and is newly proposed here. As a
survey device, the remaining 10 schemes are presented as they relate to or compare
with one another, including the present scheme as one term of comparison.

Aristotle (Posterior Analytics)

In Book II of Posterior Analytics, Aristotle analyzes the possible forms of inquiry.
His scheme consists of three sentences. He opens Book II by proposing, "The kinds
of question we ask are as many as the kinds of things which we know" (89b). In
the next sentence he proceeds to identify the number and kinds. "They are in fact
four: (1) whether the connexion of an attribute with a thing is a fact, (2) what is
the reason of the connexion, (3) whether a thing exists, (4) what is the nature of
the thing." After a paragraph he concludes: "These, then, are the four kinds of
question we ask, and it is in the answers to these questions that our knowledge
consists" (89b).
In the intervening paragraph Aristotle suggests a sequence of inquiry, a movement
from one question to another. Here is what he says, together with how it may be
construed. (His two sentences, which are separated in the original, are quoted
together in reverse order. The notation at left suggests categorial labels in existential
and cognitive terms, with movement between categories indicated by numbers and
arrows.)

1. Existence/ When we have ascertained the thing's existence,

1 Affirmation
2. Essence/ we inquire as to its nature.
Definition
3. Attribute/ When we know the fact
4 Description
4. Cause/ we ask the reason. (89b)
Explanation

As examples Aristotle gives the question, "Whether there is

next, "What, then, is God?" Further, "Does the moon suffer
"Why does the moon suffer eclipse?"
One is left in wonder about how the universe of question
knowledge, can be classified into four kinds. No other schem
broad a domain, or in so few terms. Yet the power of this sc
on the number of its categories.

328

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

Aristotle's magisterial scheme not only provides an analysis of knowledge. It also

in one stroke links questions and knowledge and correlates their categories; iden-
tifies the role and significance of questions; enunciates a principle for distinguishing
their kinds; formulates the generic questions, illustrates them with examples, and
suggests an ordering of their relations in a dynamic of inquiry.

Dillon (Table I, infra)

It was from Aristotle's few, short, and encompassing propositions that the present
scheme first took its inspiration. It classifies questions according to the kinds of
knowledge they entail about existential aspects of some phenomenon, and it
arranges the classes by priority and increment of knowledge. The principles yielding
the classification will be described first, then the categories of questions.
The domain or universe of discourse for classification comprehends questions
posed for scholarly inquiry generally, such as those found in journal studies and
those that might arise for study. The scheme distinguishes kinds of questions
according to the knowledge about some phenomenon P entailed in answer. The
classes are constituted by the generic question, What is it about P that the knowledge
in the question-answer proposition describes? The categories are grouped into
orders of information, and arranged therein, by a principle of containment (priority
and increment).
The use of knowledge and containment as classificatory principles rests on the
view that entailment of knowledge is a property of the question in three respects,
which may be put informally as follows: First, the question entails a proposition
which serves as its presupposition; second, the question entails a set of answers
which, true or not, affirm that presupposition; third, this question-answer propo-
sition serves in turn as presupposition to some further question. In that way each
successively ordered type of question is conceived to entail the preceding one and
to yield an increment of knowledge about the phenomenon in question.
It should be understood that the entailment described is of a logical kind,
irrespective of the "world-factual" truth-status of that which is entailed. The fact
that one question entails another does not of itself ensure the truth of the matter,
for the question may logically and correctly entail a question-answer proposition
that is false. The implications of this point will be discussed in a subsequent section.
Briefly, a higher ordered question contains "true" knowledge of a lower sort when
the lower question has been asked first and a true answer has been obtained. For
that reason, the principle of containment in this scheme orders the questions by
both priority and increment of information.
Table I presents the scheme in summary form. The table may serve as convenient
reference for subsequent sections of this narrative, as when its categories are used
as one term of comparison with other schemes surveyed here. The scheme comprises
17 numbered categories, with 12 lettered subcategories, grouped into 5 orders of
information; for each case the scheme supplies the generic question that constitutes
the category.
First-order categories of questions describe the properties of some phenomenon
P and entail knowledge of individual attributes. The six categories are identified by
a double label, the first indicating the existential aspect in question, the second
describing this aspect as known in logico-cognitive terms. For example, a question

329

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

about the nature or substance of P is also given as a question of definition (category

3). The arrangement of the six categories expresses a judgment about the order in
which the questions arise and their successive relations of entailment. (Lettered
subcategories are not ordered.) There arises first the fundamental question of P's
existence or, in logico-cognitive terms, of affirming P's existential reality. There-
upon arise questions of the instance of P, its substance, character, function, and
rationale, or, in logico-cognitive terms, questions of identifying, defining, describ-
ing, applying, and explicating P (noncausal accounts of P's connection with these
properties).

TABLE I
A Classification of Research Questions
Category of question Knowledge in question-answer

Zero order None

0. Rhetorical No knowledge or no answer.
First order: Properties Individual attributes of P, of Q
1. Existence/affirmation-negation whether P is.
2. Instance/identification whether this is a/the P.
3. Substance/definition what P is.
a. Nature -what makes P be P.
b. Label -whether "P" names P.
c. Meaning -what P or "P" means.
4. Character/description what P has.
5. Function/application what P does.
a. Modes -how P acts.
b. Uses -what P can do.
c. Means -how P does it or is done.
6. Rationale/explication why or how P has a certain attribute.
Second order: Comparisons Comparative attributes of P and Q
7. Concomitance whether P goes with Q.
a. Conjunction -whether P and Q are associates.
b. Disjunction -whether P and Q are alternatives.
8. Equivalence whether P is like Q, and wherein.
9. Difference wherein P and Q differ.
a. Disproportion -whether P is more/less than Q.
b. Subordination -whether P is part/whole of Q.
Third order: Contingencies Contingent attributes of P and Q
10. Relation whether P relates to Q.
11. Correlation whether P and Q covary.
12. Conditionality whether or how if P then Q, or if Q then P
a. Consequence -whether if P then Q, or what X if P.
b. Antecedence -whether if Q then P, or what X then P.
13. Biconditionality (causality) whether or how if P then Q and if Q
then P.
Extra order: Other Other attributes or ways of knowing P.
14. Deliberation whether to do and think P.
15. Unspecified to know P in other ways.
16. Unclear not known.

330

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

Second- and third-order questions describe comparative and contingent relations

between P and Q (and R, etc.). Knowledge of relational attributes is seen to
presuppose first-order questions of individual attributes.
Second-order questions describe comparative relations of three kinds. There
arises first the question of concomitance (conjunction and disjunction), then
equivalence and, on the negative, difference (disproportion and subordination).
Third-order questions presuppose knowledge of comparative attributes. They
yield knowledge of the mutually dependent attributes of P and Q in contingent
relations of four kinds. There arises first the general question of their contingent
relation and then the questions to know whether this relation is one of correlation,
conditionality, or biconditionality (causality). Knowledge of correlation is knowl-
edge of contingent relation whether this relation holds in fact between P and Q or
between these and some unknown X. To know further requires the question of
conditionality, describing antecedent-consequent relations between P and Q or
between P and some X. The next and last question is to know whether both
conditional relations hold or not. Biconditionality describes the conjunction of
conditionals with their converse.
As can be seen, the higher numbered categories and orders are classified as
containing the lower by priority and increment of knowledge. Questions in the first
first-order category and those in the last third-order category represent the funda-
mental/least and the ultimate/most knowledge of the phenomenon in question:
knowledge of P's existence and knowledge of P's cause. Causal questions figure
last for the further reason that they represent the kind of knowledge that scientific
inquiry is conceived ultimately to aspire to and that it is hoped eventually to attain.
Two additional orders group other kinds of questions that may arise and be
encountered in the journals. At the top of the scheme (Table I) is a zero order
constituted of a single class, rhetorical questions, which yield no knowledge in
answer. These include questions formulated in such a way as to foreclose inquiry
or to preclude an answer (e.g., loaded, tautological and contradictory questions,
many why-questions). At the bottom of the scheme is an extra order grouping
deliberative questions, "unspecified" questions of a possible kind not identified in
the scheme, and "unclear" questions. Deliberative questions involve knowing
whether to do an action P or to hold an attitude P. These are common in educational
research. For instance, the major questions identified for curriculum studies (Tay-
lor, 1979) are deliberative questions (e.g., "What should be taught?"). Indeed, one
of the major theories styles itself deliberative theory (Reid, 1981).
The principles of this classification waver in favor of its pragmatics, as the scheme
bends to accommodate the range of questions that may arise. One advantage of
the scheme as unbent is that it classifies theoretical and empirical research questions,
by reason of the logic of their propositions. For example, questions about the
conceptual implications of P or the observed effects of P are conditional questions
(consequence). For that same reason the scheme can also classify noninterrogative
problems, such as hypotheses to test and theses to sustain. Again for that reason
the scheme can classify questions expressed in various natural and artificial lan-
guages. The scheme's categorial approach also makes it useful in conjunction with
other schemes, such as those that specify topical content of a domain. For example,
this scheme can be used to classify the questions asked about each of the "com-

331

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

monplaces" of education-the kinds of questions about teaching, and about

learning, and so forth. The result is to map the kinds of knowledge about the kinds
of things in a given domain.

Lundstedt (1968)
The notion of using the logic of questions and grouping their types into orders
of information yields the scheme constructed by Lundstedt (1968) for research
questions in psychology, set forth under the modest title, "A Note on Asking
Questions." Lundstedt's major concern about research bears on the questions, and
his major concern about questions is their form. The formulation of questions
"transcends in importance the concern for method, and in itself constitutes a basic
method of inquiry" (p. 239). The form chosen affects the substance of the question
and "dictates as well the range of methods possible for answering it and suggests
too, mathematical models best suited" (pp. 230, 232).
Lundstedt borrows seven formal questions from traditional logic, arranges them
into three orders, and identifies the research methods and statistical treatments
appropriate to each. The seven questions are as follows: affirmative and negative,
conjunctive and disjunctive, conditional and biconditional, and contrapositive
(dismissed as a tautological form of the biconditional).
The affirmative type is called the most fundamental of the seven and hence,
together with the negative, a first-order question (p. 230). As an example Lundstedt
gives, Is group A task oriented?, a question whose form is said to dictate a method
such as survey and a mathematical treatment such as measures of central tendency.
The conjunctive and disjunctive are called second-order questions "because they
contain more information about relations" (p. 231), namely association and exclu-
sion. Is group A task oriented and (or, or) is group A socioemotionally oriented?
Correlational analysis is indicated for the first of these, and t tests for the second.
The conditional and biconditional types are third-order questions because they
assume association. These take the form: If X, then Y; and, If X then Y and if Y
then X. For example of the first, "If there is group pressure on individuals, then
will some individuals experience distortion of their perception?" (italics in original).
Experimental methods and analysis of variance are indicated for these two types.
The terms of these formal questions figure also in the scheme presented in Table
I, as does the grouping of types into orders of information. These are owed to
Lundstedt's work. The approaches are similar but the terms figure differently in
the two schemes.
Lundstedt begins with formal questions from logic and then appears to apply
their terms as syntactic descriptors. For example, he identifies conjunction and
disjunction with the syntactic operators and and or (p. 231). All of his examples
exhibit this view by classifying questions of varying contents and processes accord-
ing to one and the same linguistic form-labeled as and identified with logical
form (perhaps also identified with the logic of the formal questions).
Some instances taken from educational journals (Dillon, 1983) will show that
the six formal questions from logic can take on various linguistic forms. Here is a
question of correlation (conjunction) structured in affirmative/negative terms and
phrased in conjunctive/disjunctive terms: "Do schools get better or worse as state
and federal legislation increases?" An experimental study raises a question that has

332

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

the logic of conditionality but not its if/then linguistic form: "Do college students
evaluate their instructors differently depending on their knowledge of the uses of
the ratings?" From a journal of educational theory comes an instance of both
conjunction/disjunction: "Are equality and excellence mutually exclusive or inclu-
sive?" Another example takes the paradigmatic form: "The question might be
raised whether the two emphases can be separated in this manner."
When questions are classified into the six logical types that Lundstedt identified,
and when the types are grouped into the three orders that he constructed, the result
is a useful and intellectually satisfying scheme. But the logic of these formal
questions is not given by the syntax of the question-sentences that render them in
natural language. In the scheme in Table I, questions of yes/no, and/or, if/then
form may range variously across the categories and orders. On the other hand,
questions with the logic of affirmation/negation, conjunction/disjunction, condi-
tionality and biconditionality can be found only in the single category designating
that type.

Bunge (1967)
Another scheme that relies on logical form, but which derives from modern logic
and covers scientific research generally, is provided by Bunge (1967) in the chapter
on problems in his Scientific Research. Bunge identifies nine "elementary question
forms" divided into two "problem kinds," individual and functional.
Both kinds of problem ask for the value of one (or more) variables. The individual
kind asks about individual variables and the functional kind asks about predicate
variables. Four types of questions are distinguished as belonging to the individual
kind, and two for the functional kind. Each type is identified by interrogative
pronoun, formulated into a formal question-sentence, and given logical notation
of its form. For example, the why-type is formulated as, "Which is the p such that
q because p (i.e., p entails q)?," with logical form given as (?p) (p-q).
The four types of individual kind of questions are which, where, why, and
whether; functional questions are how and what. As classified by the scheme in
Table I, these six are questions of identification (category 2 in Table I), description,
antecedence, existence, function, and description. The whether type is classified as
a question of existence, requiring affirmation/negation, because it is described in
the text (p. 172) as a question about existence or universality (viz., Is P?), rendered
as a question about the truth value of the corresponding existential proposition
(viz., Is it true that /P is /?). Thus, all of Bunge's types but one constitute first-
order questions about individual properties; the why-type is a third-order question
about antecedent relations, a subclass of conditionality. Table I distinguishes
additional types and orders of questions.
To be sure, Bunge's scheme bears on "elementary" question forms and is meant
to be "illustrative, not exhaustive" (p. 173). Nonetheless, every elementary question
type is stated (p. 172) to be either of the two kinds identified (individual and
functional), and no indication is given as to the nature of nonelementary question
types. Ordinarily these would be understood to comprise combinations of elemen-
tary types or elaborations of the variables in one type. Such filigree would not alter
the type of question, only its complexity. Moreover, Bunge gives little indication

333

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

as to what thing it must be that the scheme is only illustrative of, and what extent
it must have such that the scheme describes it only partially rather than exhaustively.

Steiner (1978)
A pronominal principle was followed also by Steiner (1978) in constructing her
"schema of kinds of quantitative educational research questions," and she too
provides some logical notation for the kinds. Steiner starts with the view that
"problems which are relevant and fruitful for quantitative educational research
take an interrogative form" (p. 14). Accordingly, she proceeds to use the interro-
gative pronouns to distinguish kinds of questions. "Given the different interrogative
terms, the following kinds of quantitative research questions can be distinguished:
who, where, when, what, how, why, which, and whether" (p. 14).
The principle whereby these kinds of questions have been distinguished is
straightforward and the eight resulting classes are clearly drawn. But when presented
in Steiner's "Schema 6" (p. 15), the lines of the scheme, though clearly marked,
become conceptually tenuous or at least difficult to follow. Nonetheless it is
instructive to follow these lines.
The schema lists the eight kinds of questions by their pronominal labels. It also
draws lines from these labels to other categorial labels embracing the eight kinds
into successively superior categories. This schema raises a question about the
classification being presented. It depicts four steps of classification proceeding from
left to right of the diagram and yielding the eight kinds of questions listed at the
right; whereas the classification task must have begun beyond the right of the
diagram and proceeded in a single step to the left (namely, the step of distinguishing
the eight kinds of questions according to an interrogative-pronominal principle).
The first step as depicted in the schema is the bifurcation of the domain labeled
quantitative research questions into categories labeled hypothetical and verifica-
tional.
The second step depicted divides these two categories into three. Hypothetical is
somehow divided into a single class, questions of unknowns; verificational is
bifurcated into questions of alternatives and questions of acceptability. These latter
two classes can have been derived only by performing two other unspecified
divisions, neither one a bifurcation, rather than by the one bifurcation depicted to
have yielded them.
The third step depicted divides only one of these three categories, the unknowns.
The classes that result from this third division, theoretical and factual unknowns,
are aligned with the classes that will result from the fourth division of the other
two categories remaining undivided in the third step.
The fourth and last step depicted yields the list of eight kinds of questions with
which the classification must have begun. Theoretical and factual questions (of
unknowns) are each divided into three kinds: the theoretical what, how, why; and
the factual who, when, where. It is not known in what respect a given one of these
kinds (e.g., what) represents a theoretical rather than a factual question. Six of the
eight having been identified, there remain two kinds of questions that are to result
from the division of other categories. As there are no third-step categories from
whose division they may result, for there exist no other third-step categories to
divide, the last two kinds of questions result from a division of second-step

334

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

categories. Two second-step categories have remained undivided in the third step,
questions of alternatives and questions of acceptability; each of these is now divided
into a single kind, which questions and whether questions. The eight kinds of
questions have been identified.
The result is a scheme that conceives quantitative research questions to be either
hypothetical or verificational; and, following through with the same principle that
has made them to be of either a hypothetical or verificational kind, the scheme
would conceive verificational questions to be of either a which kind or a whether
kind, and hypothetical questions to be of what, how, why, who, when, and where
kinds. By yet the same principle, the scheme would conceive these hypothetical
wh- kinds to be either theoretical or factual wh- questions, contrasting with
verificational wh- questions of other kinds.
For purpose of comparison, the scheme will be taken to consist of the eight kinds
of questions labeled pronominally. The brief logical notation given for each sug-
gests a way to correlate these kinds with categories in the classification proposed in
Table I.
Steiner's three first-listed kinds (what, how, why)-otherwise called hypothetical
questions of theoretical unknowns-may be classified as questions of identification
(category 2 in Table I), application, explication, and/or conditionality, specifically
antecedence. Of the next three-otherwise hypothetical questions of factual un-
knowns-the who type is also identification, and the when and where types are
description. The remaining two types are which and whether-verificational ques-
tions of alternatives and of acceptability. The first, noted as (p? or q?) is a question
of disjunction. The second, (p?), appears to be a question of existence/affirmation,
verifying the existential status of some phenomenon or proposition P.
In short, Steiner's eight types of question may be classified into five categories in
Table I: three categories of first-order properties, one each of second- and third-
order comparative and contingent relations. Both schemes propose to classify
educational research questions, with the difference that Steiner's applies to quan-
titative questions only. The scheme in Table I does not distinguish between
quantitative and other questions. It does distinguish more types than Steiner's
scheme but the types apply equally to quantitative as to qualitative questions,
conceptual or empirical questions within the disciplines in educational research.

Shulman (1981)
The types of questions raised in the disciplines form Shulman's (1981) scheme
as part of his interesting survey of "Disciplines of Inquiry in Education." In contrast
to the preceding schemes, such as Lundstedt's (1968), formal logic plays no role
here. Like Lundstedt, however, Shulman relates types of research questions and
types of research methods. But whereas Lundstedt identified methods "dictated"
by the various question types, Shulman identifies questions that are "raised" by
various types of method. He proposes that what distinguishes the various research
methods is "the very types of questions they tend to raise," and that disciplines are
distinguished by "the manner in which they formulate their questions" (p. 6).
Shulman illustrates this case by examples of how various disciplinary methods
can be applied to the study of reading instruction. For six different cases he first
identifies a question, together with one or two others in the same vein, and then

335

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

relates these to different research methods and disciplines. For example, the first
question, "What makes some people successful readers and others unsuccessful?,"
is related to correlational methods from fields of psychology and agriculture.
Although the narrative describing the scheme proceeds from question to method,
the principle yielding the scheme proceeds from method to question. The categories,
then, are types of methods that raise types of questions. The six different methods
are correlation, experiment, survey, case study, conceptual analysis, and historical
analysis. For example, a question of case study is "How is reading instruction
carried out?"; a question of conceptual analysis is "What does it mean to be able
to read?"
Although these are six different things, they cannot be regarded as six different
kinds, or as six kinds of the same thing. Correlation, survey, and historical analysis
do not stand to one another in the same way that they stand to experiment, case
study, and conceptual analysis; they do not stand to these three in any one way;
and they do not stand to one another in one way. Hence these six might represent
various kinds of different kinds of things. They are distinguished according to
different principles and multiple and varying characteristics; these might include
their character as a technique, as an approach that uses various techniques, and as
a label describing a discipline that may follow various approaches and use various
techniques.
All together Shulman gives 16 questions as examples of the six varieties of
method. All 16 questions may be classified by the scheme in Table I as three kinds
of first-order questions about the properties of the phenomenon in question, reading
instruction. Six are questions of definition (category 3), 6 are description, and 4 are
function. None of the questions appears to have the character of second- or third-
order comparative and contingent relations. In other words, the identification of
questions according to research method may yield a restricted range of questions
for study.

Smith (1981)
It was for this sort of reason that Smith (1981) proposed his scheme, in order to
increase the range of questions (and methods) in educational research. In a
stimulating note entitled "Noncausal Inquiry in Education," Smith asserts that
because of researchers' preference for experimental methods, "only causal questions
tend to be asked," and he encourages them to ask a variety of noncausal ones (p.
23). Like Shulman, Smith chooses a reading program to illustrate his case; and like
Shulman and Lundstedt, but in still another way, he correlates questions and
methods.
The scheme first divides questions into causal and noncausal, then identifies
eight noncausal types. For each type the scheme provides a label, a sample question,
characteristic research methods, and fields as sources of methods. "Notice the
diversity of the significant questions that could be asked about this program and
the range of relevant inquiry methods" (p. 23). Indeed, the scheme specifies 21
methods from 15 fields for the 9 types of questions. For example, a question of the
type called constituents of quality, "What are the characteristics of the best CAI
materials being used?," is related to content and structural analysis from poetry
and film criticism.

336

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

The types of question are also diverse, to the point that they do not seem to
derive from either a formal principle or from a single one. Four of the types are
approximately comparable and may correspond to traditional philosophic cate-
gories. These types are labeled meaning, geographic distribution, constituents of
quality, and "what transpires"-corresponding to, or perhaps instantiating, cate-
gories of substance, location, quality (or composition), and function. The remaining
four constitute a mix of types formed by at least two additional principles. The
types are labeled cost, optimization, power and influence, public and group opinion.
These do not represent classes or kinds or types of any one thing.
The motive, purpose, and organization of Smith's scheme are clearly stated. But
the principle that distinguishes the types of questions goes unstated and has not
been divined here. Despite their diversity, some of Smith's types represent individ-
ual variants of one and another type of question, and types of various kinds, rather
than each of the types constituting one kind of question. The eight noncausal types
are eight different questions, and they are questions of different types, but they are
not eight different types or kinds of question.
In terms of the classification in Table I, the causal type in Smith's scheme is a
third-order question of contingent relations; one of the noncausal types is a second-
order question of comparisons; and the remaining seven describe first-order ques-
tions of properties. The causal type corresponds to a subclass of conditionality,
consequence (category 12a), less probably to biconditionality or causality. The type
labeled cost describes a quality of the reading program in comparison with others
and represents a second-order question of equivalence: "What is the cost effective-
ness of the program compared to other programs?" (viz., Is cost of P equal to cost
of Q and to cost of R?). The remaining seven are first-order questions about the
properties of the reading program. The labels and examples together suggest that
two of the types are questions of description, three of function/application, one of
explication, and one of definition. To illustrate, the type labeled meaning is a
subclass of definition.
All of Smith's noncausal types may be classified into five categories in Table I,
which distinguishes 20 noncausal types in all. In addition, it identifies a deliberative
type, a kind of question that readily arises in education, that is frequently posed
about reading programs, and that seems especially pertinent to ask within Smith's
subfield of evaluation research.

Johnston and Pennypacker (1980)

Whereas Smith's scheme was designed only for nonexperimental questions in
educational research, Johnston and Pennypacker (1980) designed theirs only for
experimental questions in behavioral research. Moreover, they enunciate at the
outset a clear principle to follow in constructing their taxonomy: "A useful
taxonomy of research questions can be immediately derived from an analysis of
the functions that the possible answers serve for different investigators" (p. 35).
Six kinds of questions are distinguished and labeled as curiosity, demonstration
of a new phenomenon, exploration of a new method, analysis of a "new" phenom-
enon, extension of the generality of a known phenomenon, and hypothesis testing.
This list makes it plain that the scheme has not followed a principle and does not
constitute a taxonomy.

337

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

Despite the identification of "function" as a generative principle, the nature of

function and its manner of generating the items remains unclear. The six varieties
seem unrelated to the function of an answer, unrelated to possible answers which
are to serve some kind of function, and unrelated to the investigators for whom
the answers are to serve different functions. For example, "exploration of a new
method" is illustrated by the case of using various schedules of reinforcement, and
"extension of the generality of a known phenomenon" by attempting to reproduce
with other species relations established for one species. Neither of these types
describes a possible answer, a function of an answer, a service to an investigator,
or a point around which different functions or answers may turn for different
investigators.
The scheme does not constitute a taxonomy for the surface reason that it does
not order anything, and for the prior reason that it does not constitute a classifica-
tion, hierarchical or not. As with some other schemes surveyed here, it specifies
and lists items. Lists and inventories can make useful schemes. But much depends
on the list. The items listed together in this scheme raise questions about what it is
that is being listed, and how the list is to be constituted of those items that are
specified on it.
The six items make an uncertain list. They do not represent six types, and they
are not six individuals. They make a list of no single thing. Curiosity, for instance,
is a thing of one kind and hypothesis testing and demonstration of a new phenom-
enon are two things of another kind; they are not three kinds of thing but three
things of at least two kinds; and the two kinds are not variants of one thing but of
two. The thing that is common to the items on the list cannot be function.
Hypothesis testing and curiosity, for example, can serve the same function, the one
can serve a function of the other, and both may receive the same answer, rather
than the two serving, as announced in principle, different functions and yielding
different answers.
The six varieties of experimental question specified on the list do not describe
the realm of experimental questions. All of them may be used equally for nonex-
perimentation, and several of them may be used for nonempirical research. There
are other kinds of experimental questions and functions not identified on the list.
It is not known why some are listed and others are not.
The principle of this scheme is unique among those surveyed here, and the
categories as well. Only one of the categories, hypothesis testing, can be classified
in terms of the scheme in Table I. As defined, hypothesis testing is a question of
conditionality, in the subclass of consequence (12a). But even as defined, this type
need not be an experimental question; "to affirm the consequent of a particular
proposition" (p. 38) may be achieved by logical analysis or theoretical research as
well as by experiment.

Laudan (1977)
Another taxonomy, also unique, has been proposed by Laudan (1977) in his
influential essay on the philosophy and history of science, Progress and Its Prob-
lems. Laudan conceives of science as a problem-solving activity, wherein problems
constitute the questions of science and theories constitute the answers (p. 13). He

338

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

observes that the literature of the methodology of science does not offer a "taxon-
omy of the types of scientific problems" (p. 13). In two chapters of narrative he
proposes a taxonomy, distinguishing broadly between empirical and conceptual
problems.
Empirical problems are called first-order questions about objects in the domain
of a science; conceptual problems are higher order questions about the theories
devised in answer to first-order questions (p. 49). Three types of empirical problems
and two of conceptual problems are distinguished "relative to the function they
have in theory evaluation" (p. 17).
Empirical problems are classified as unsolved (by any theory), solved (by some
theory), and anomalous (not solved by a given theory but by a competitor).
Conceptual problems are of two kinds: internal to the theory and external in its
relation to another theory. Each kind incorporates several classes. The two types of
internal conceptual problem are logical inconsistency and conceptual ambiguity.
External conceptual problems are of three types, describing relations between one
theory (T 1) and another (T2). Logical incompatibility is a type of problem consti-
tuted by T2 entailing the negation of T1. When the two are compatible but T2
entails the unlikelihood of TI (acceptance of the one reduces the acceptability of
the other), the problem is one of joint implausibility. The third type is mere
compatibility, where T2 entails nothing about T1, although it "ought to reinforce"
Tl (p. 53).
Immediately upon having distinguished these types of conceptual problems,
Laudan outlines another taxonomy-"a taxonomy of the various cognitive rela-
tionships which can exist between two (or more) theories" (p. 54). Five relations
are specified in sequence: entailment, reinforcement, compatibility, implausibility,
and inconsistency. Except for entailment, these relations can pose or generate a
conceptual problem, and they are listed in an increasing order according to
"different degrees of cognitive threat" (p. 54).
The second taxonomy is difficult to evaluate. Its kinds appear to be cognitive
relationships between theories, and its principle of arrangement is by increasing
degrees of cognitive threat. It may be noted first that theories do not exhibit
cognitive relations, that some of the relations specified are not cognitive ones, and
that they are arranged not according to any characteristic of the theories or of their
relations, such as may be said to increase, but according to the relations between
either the theories and their knowers, or between knowers of two theories.
Compatibility, implausibility, and inconsistency are instances of the first type of
relation, entailment, rather than being three other types. The second type, reinforce-
ment, appears unrelated to entailment. That is, the five specified relations are not
five kinds of relations but two. Moreover, of the five relations, only implausibility
and inconsistency bespeak of themselves a problematic relation; the others can
pose "cognitive threat" only in relation to some unspecified additional term of
correlation, for example, relating compatibility to expectation of reinforcement.
The scheme describes an increase in problematicity but does not describe the
principle by which it is to increase, or how it is that one cognitive threat is a degree
of another. The first item on the list is exempted from the scheme's sequence, on
the grounds that it does not represent problematicity (p. 54); that is, the first item
in the scheme does not fit within the scheme. The two subsequent items are to

339

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 J. T. DILLON

represent two degrees of problematicity although both are described in nonproble-

matic terms (reinforcement and compatibility), and although the second is defined
as T2 entailing nothing about T1 (p. 54).
As for the first taxonomy-empirical and conceptual problems-the classes
within the two major orders are distinguished according to different and multiple
principles. Kinds of empirical or first-order problems are distinguished first accord-
ing to the solution state of the problem (unsolved/solved) and then according to
the solver of the problem (anomalous-solved by one theory rather than another).
Kinds of conceptual or second-order problems are distinguished not by solution
state of the problem but by the interrelations of achieved solutions. Internal
conceptual problems are identified first by the logical quality of a theory (incon-
sistency) and then according to its cognitive quality (ambiguity). External concep-
tual problems are distinguished first by the logical relation between two theories
(incompatibility-entailment of negation), next by the cognitive consequence of
accepting one of the theories (implausibility-reduces likelihood of accepting the
other theory), and last by a logical relation as evaluated by an antecedent cognitive
state (mere compatibility-T2 ought to reinforce T1 but is only compatible with
it).
Moreover, the two orders of the taxonomy stand in uncertain relation-questions
about objects, and questions about the answers to those questions. It is clear that
these are two distinct kinds of questions, but their taxonomical ordering is not
clear: What is the thing of which conceptual questions are a higher order and
empirical ones a first order? They seem each to be a level from two different
orderings rather than two levels from one ordering. Similarly, the taxonomic
arrangement of categories within orders and of classes within categories remains
unclear. What, for example, is the relation between an internal inconsistency and
an external one? The second does not describe an increase in the force or extent of
the first, nor does it describe any inconsistency within either of the theories, since
these are inconsistent only in relation to each other.
Apart from issues of its quality as a classification, Laudan's scheme is unique
among the others surveyed here-in the categories it identifies, in the principles
whereby it distinguishes types of problems, and in the uses for which it is proposed.
(None of its categories correlates with those in Table I.) Its application may be
broadly put as historical and philosophical, to evaluate the answers (theories) to
questions posed over the course of science, and thereby to gain an understanding
of science as a problem-solving activity. Yet its compass is not the range of all
questions that have been pursued but only those few that served for a time as
paradigms of scientific understanding, as promoters and competitors in the progress
of science. As such, the scheme will prove its usefulness in the account that it yields
of the intellectual history of research. (Laudan proposes that his view can also be
applied "to all intellectual disciplines" [p. 13].)

Fischer (1970)
Another distinct scheme, incomparable to the others in yet a different way, is
Fischer's (1970) "Fallacies of Question-Framing," the first chapter in his Historians'
Fallacies: Toward a Logic of Historical Thought. Fischer argues that historical

340

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

research is "not story-telling but problem solving" (p. xii). His scheme appears as
the lead chapter perhaps because of his view that inquiry is initiated and controlled
by questions.

Without questions of some sort, a historian is condemned to wander aimlessly

through dark corridors of learning. Without questions of the right sort, his
empirical projects are consigned to failure before they are fairly begun. (pp.
3-4)

Fischer does not go on to specify the right sort of questions but the wrong sort, the
fallacious kinds.
The scheme identifies 11 fallacies of question-framing. Each is richly and hu-
morously illustrated by analyses of published historical research. For example, he
gives the form of false dichotomous questions as "Basil of Byzantium: Rat or
Fink?"-no more fanciful an example than the two dozen titles he cites from
historical publications.
The others on the list include the fallacy of many questions ("When did racial
segregation harden into its elaborate mold?"); metaphysical questions ("Was the
Civil War inevitable?"); fictional questions ("Did railroads alter the process of
American economic growth in a way that only railroads were able to do?"); and
semantical questions ("Was the political structure of 17th-century America demo-
cratic or aristocratic?"). Declarative questions forego inquiry by giving their own
answers (or are formulated after the answer is obtained by mimicking other
research), and counterquestions reiterate faulty premises while repudiating mistaken
conclusions. The remaining types include tautological questions ("Why do some
spectacular agitators forward their cause, and others do not?"); contradictory
questions ("What really happened in the summer of 1422, when, as every schoolboy
knows, an irresistible force met an immovable object?"); and last, "potentially
verifiable" questions, those which propose historical questions for sociologists to
answer.

Fischer's evaluation of these generally may be borrowed from his c

one of the types: "A working historian receives no clear signal fr
interrogatories as to which way to proceed, how to begin, what k
will answer the problem, and indeed what kind of problem is rai
basic principle seems to be that a fallacious question will infirm
answer, as is known from the adage, "Ask a foolish question, get a f
This scheme does not claim to be a classification, but it has qual
inventory. It borrows from logic the traditional types of fallacy
questions to set forth propositions, it classifies questions accordin
fallacy exhibited by the question. The scheme specifies varieties o
types, for example, three varieties of tautological questions; but it a
of what are varieties, for example, counterquestions and declara
operate in the same basic way. These specifications seem to have
convenience of narrative, as if to fit the examples of published
analyzed. Thus the scheme seems sound in principle and in the gre
execution, while following another logic for the remaining part.

341

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

None of the categories of this scheme compares to those of others. They derive
from logic but do not describe logical form, as do the schemes of Bunge (1967) and
Lundstedt (1968). The usefulness of the scheme seems demonstrated in its appli-
cation to published historical research. It may also find wider application to other
domains and to the analysis of questions in the process of being formulated as well
as to those already answered.

Aristotle (Topics)
The distinction of Aristotle's scheme is that it is the first to take a categorial
approach and, in accord with his view of knowledge as answers to generic questions
(Posterior Analytics), it is the first to constitute categories by questions.
In Book I of the Topics, Aristotle inquires into the number and kinds of things
about which arguments take place. To describe all the propositions that might be
formed about a given subject, he identifies 10 classes of predicates: essence or
substance, quantity, quality, relation, place, time, position, state, activity, and
passivity. "Such then, and so many, are the subjects on which arguments take
place" (103b).
Aristotle appears to have developed his scheme of categories by reference to the
kinds of questions they answer, for in the Greek he denominates the (first 6)
categories by interrogatives. (The list is also given in Categories at lb). Thus, the
number and kinds of predicates derive from the number of distinct questions that
can be raised about a given subject. Each question entails a set of answers which
together form the class and which individually, question-cum-answer, form prop-
ositions for argument. All together, the 10 question categories exhaust the domain
of discourse ("Such, then, and so many").
For present purposes this scheme may be construed as a classification of questions
for research, that is, for inquiry into a given subject. On the evident assumption
that the subject or phenomenon exists, there arises first the question of its essence
or substance; Aristotle words this category "What is it?" The other categories
identify questions about the attributes of the subject. Five of these are "How much?,
Of what sort?, Towards what? or Relative to what?, Where?, When?" For the four
remaining categories on Aristotle's list-position, state, activity, passivity-the
questions can be formulated as suggested by Kahn (1978): What is his position? or
How is he situated?; What is he wearing? or How is he disposed?; What did he do?;
and What did he suffer?
In terms of the classification in Table I, Aristotle's question categories may be
distributed as follows. The question, What is it? corresponds to the category of
substance/definition (3). Six of the nine other kinds all group within the category
of character/description-How much?, Of what sort?, Where?, When?, What is his
position?, and What is he wearing? (state or habitus). Questions of activity and
passivity are questions of function. Finally, Aristotle's question of relation, Relative
to what?, may well incorporate all second-order questions in Table I-concomi-
tance, equivalence, and difference. These are questions of the comparative attributes
of P and Q, just as Aristotle's question predicates something of a subject by
reference to something else. In that respect his single category might also be
interpreted to embrace as well all third-order questions of contingent relations-
relation, correlation, conditionality, and perhaps causality.

342

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

Otherwise put, the scheme in Table I may be taken to specify Aristotle's single
category, Relative to what?, by distinguishing seven kinds of questions in two
orders. On the other hand, Aristotle distinguishes six kinds of questions that specify
the single category in Table I, character or description of the qualities of P. In
addition, Table I distinguishes questions of identification, means, explication, and
deliberation, none of which appears to figure in Aristotle's scheme despite their
seeming incidence in argumentation and other forms of discourse. (For a sophisti-
cated criticism of Aristotle's categories in light of moder research, see Kahn,
1978.)
One is struck by the simplicity of Aristotle's scheme, yet one is perplexed by the
ordinariness of his little questions. How can it be that these tiny interrogatives,
after all daily heard ex ore infantium, describe the universe of discourse, detail the
possibilities of argumentation, or comprehend the kinds of questions which moti-
vate inquiry? Yet authorities in categorial theory (e.g., Kahn, 1978) stress that
categorial schemes must depend on such modest and elementary distinctions, and
"should be firmly rooted in humble, everyday questions like What is it? How big?
Of what sort or quality? In relation to what? Where? and When?" (p. 266). The
next scheme takes just such an approach and identifies just these kinds of questions
to classify the domain of scientific inquiry.

Rescher (1982)
Much as Aristotle had done, Rescher (1982), in his book Empirical Inquiry,
takes a categorial approach to distinguishing kinds of factual questions that we may
ask about the world around us. Indeed, his scheme figures as part of his chapter on
categories rather than in the subsequent chapters devoted to questions and inquiry.
Aristotle begins Book II of Posterior Analytics, as he usually does, with a
magisterial statement: "The kinds of question we ask are as many as the kinds of
things which we know" (89b). Rescher begins his chapter on categories with the
wonderfully succinct rendition: "Categories are correlative with questions" (p. 61).
He subsequently elaborates:

They delineate and canalize our efforts to secure information about the
world. They provide the conceptual frame of reference in terms of which we
pose our questions about the nature of things-the cognitive scaffolding we
employ in erecting our view of the world. (p. 62)

Rescher's questions too are simple. He describes his scheme as an "inventory" of

"protoquestions," "the most elementary, basic and fundamental sorts of issues to
be raised about the furniture of the world" (p. 63). Scientific questions are of the
same kind. They are sophisticated in that they may make finer distinctions and
may be answered in different terms, but they proceed from the protoquestions and
they proceed within the same categories of thought. The issues posed by the
protoquestions are "recurrent Leitmotivs throughout the unfolding of inquiry"
(p. 65).
Rescher's scheme is no mere inventory. It distinguishes 11 classes of questions
according to five traditional philosophical categories: substance, kind, quality,

343

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

relation, and rationale. For each kind a categorial process and a paradigm question
are provided.
The first three categories comprise one question each. Which one? is a question
of identification in the category of substance. For the category of kind the question
is, Of what kind? (a question of classification), and for quality, Of what properties?
what manner? (description). Related to the scheme in Table I, the first category
corresponds not to substance but to instance/identification (2), and both of the
other two correspond to character/description. Rescher's category of classification
seems oddly placed and nearly tautological.
The remaining two categories are relation, with six kinds of questions, and
rationale, with two kinds. Their correspondences with the scheme in Table I are
various.
Three of the six types within Rescher's category of relation correspond to a single
category in Table I, questions describing Ps character (4). These three are Rescher's
location, for which the questions are When? where? whence? whither?; composition,
Composed of what? how constituted?; and quantification, Of what size or magni-
tude (in this or that respect)? Rescher's other three types of relation are affiliation,
Like what or which?, corresponding to equivalence; subordination, To/of what or
which?, a subclass of difference, also labeled subordination; and process-character-
ization, In what way? by what means?, specified in Table I by two subclasses of
function, modes and means.
Rescher's last category, rationale, comprises two kinds of questions. Questions
of the first kind, function-specification, are To what end? for what purpose? with
what point? These correspond to the category of function in Table I, particularly
to the subclass of uses. Questions of rationalization or explanation are Why so?
how to be explained? by what agency? These correspond in part to the category of
explication, and in other part to conditionality and causality. In light of the scheme
in Table I, the category of rationalization seems insufficiently discriminated.
Hence, 9 of 11 classes in Rescher's scheme may be termed first-order questions
about the properties of some object. Two classes are second-order questions of
comparisons, and part of another class describes third-order questions of contingent
relations. On the whole, the two schemes describe much the same dimensions of
phenomena but specify these in different ways and with differing degrees of
distinction and ordering. One further difference is that Rescher's scheme is limited
to factual questions raised for empirical inquiry, whereas the Table I scheme
comprehends theoretical as well as empirical inquiry, and conceptual and deliber-
ative questions as well as factual ones. The two schemes are alike in the significant
respect that, like Aristotle's, they are constructed following a categorial approach,
and their categories are constituted by questions.

Analysis of Schemes
Some analysis of each scheme was provided in the course of the survey, and a
few points of comparison drawn among them. In this section all of the schemes
will be considered together for their general characteristics and next examined
particularly for their comprehensiveness. Other criteria and issues will then be
considered.

344

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

General Characteristics

All of the schemes surveyed are different and each of them may be considered
deficient. Their differences, and some of their deficiencies, are to be expected, for
classifications must differ according to a number of principles and they cannot
achieve a number of criteria set for them. In the view taken here, the differencies
are desirable and certain of the deficiencies unavoidable.
Classifications will differ according to a number of principles (see Koerner, 1981).
For example, they may describe different domains or they may be designed for
different uses. What is more, two schemes may differ in classifying the same domain
for the same purpose. In principle, no one classification can comprehend a domain
in such manner as to leave no place and no angle for another scheme. W. S. Jevons
(1874), a pioneer of modern classification theory, starts his chapter on classification
with the reminder that "there must generally be an unlimited number of modes of
classifying a group of objects" (p. 677); and he ends with the emphatic caution,
"we must not attribute exclusive excellence to any 6ne method of classification" (p.
722). Several schemes, then, will always be needed. Each will delineate selected
characteristics which are of interest to construing any given phenomenon or
domain.
Furthermore, no a priori rules govern either the choice of characteristics that
may be of interest, together with the manner of conceiving them and the strategy
of classifying them, or the relations that are drawn among the classes and the
judgments that these classes and relations express. As Jevons (1874) stressed, "there
will be no royal road to the discovery of the best system, and it will even be
impossible to lay down rules of procedure to assist those who are in search of a
good arrangement" (p. 690). Similarly, Gregg (1954) thought it not possible to set
down "antecedent directives" to follow in constructing a taxonomy (p. 30). That is
not to say that there are no rules. From another perspective, Broadfield (1946)
states that, to scientists who wish to have formulated into rules the principles on
which they deal in genera and species, classification is prepared to issue rules "with
some reluctance and many saving clauses" (p. 51).
In that spirit the dozen schemes may be compared on the most general charac-
teristics and their differences summarized. Table II notes the approach taken by
the schemes, the number of categories they distinguish, the domain they are
designed to describe, and the principle used for distinguishing types of question.
The most prominent difference among the schemes turns on whether or not they
adopt a categorial approach to identifying the various kinds of questions. Four of
them do. The others either do not or else they take a categorial approach only in
part (effectively, these are for that reason noncategorial approaches). It would seem
that, as a result, some of the schemes are more comprehensive than others.
A categorial approach leads the classifier to ponder all the possible aspects or
ways of being of a phenomenon. The result is a set of categories that may more
readily assume a comparable level of generality, an individually unique distinction,
and a collectively encompassing character such that the scheme systematically
comprehends the entirety of the domain being thought about. By contrast, a
noncategorial approach entails greater risk of omitting aspects of the phenomenon,
of identifying aspects according to divergent principles or characteristics, and of

345

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

distinguishing several specific and several general categories, to the effect that the
classes may overlap, may describe noncomparable dimensions, and may form
uncertain mutual relations within the scheme.
The general approach taken to classification is therefore the most significant
point of difference among the schemes and the most instructive. Next is the
characteristic selected whereby to distinguish the kinds or varieties of questions.
In general, the four categorial schemes in Table II distinguish questions according
to the kinds of knowledge in answer, that is, knowledge describing kinds of aspects
of the object in question. The other schemes use a range of characteristics. These
include the form of the question (Bunge, Lundstedt, Steiner), the fallacy it proposes
(Fischer), the method of answering it (Shulman, Smith in part), the function of the
answer (Johnston & Pennypacker), and the status of the answer (Laudan). More-
over, some of the noncategorial schemes apply the selected characteristic in change-
able ways and introduce others into the mix. For example, one scheme (Lundstedt)
distinguishes types of questions according to logical form but classifies some of

TABLE II
Selected Characteristics of the Schemes
Nof
categories/ Principle of distinguish-
General approach Domain
classes + ing question types
subclasses

Categorial
Aristotle (Posterior 4 all knowledge knowledge in answer
Analytics)
Aristotle (Topics) 10 all propositions ways of being (predi-
(arguments) cates)
Dillon (Table I, 5, /17 + scholarly inquiry knowledge in answer
infra) 12
Rescher (1982) 5/11 empirical inquiry categorial process
Semicategorial
Bunge(1967) 2/6 scientific research logical form and inter-
rogative pronoun
Laudan (1977) 2/5 + 5 scientific research (and 1. status of answerhood
other disciplines) 2. interrelations of
answers

Lundstedt (1968) 3/6 quantitative research logical form

(psychology)
Smith (1981) 2/9 noncausal inquiry (unspecified)
(education)
Noncategorial
Fischer (1970) 11 historical research logical fallacy
Johnston & Penny- 6 experimental research function of answer
packer (1980) (behavioral science)
Shulman (1981) 6 disciplinary inquiry method of answering
(education)
Steiner (1978) 8 quantitative research interrogative pronoun
(education)

346

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

them by linguistic form. Another scheme (Smith) distinguishes a few types of

questions according to one (unidentified) principle, then distinguishes the remain-
der according to at least more than one additional (unidentified) principle. Still
another scheme (Johnston & Pennypacker) distinguishes all of its classes by several
unidentified characteristics, and none of its classes by the characteristic it identifies.
Finally, one of the schemes (Laudan) distinguishes each of its classes according to
different characteristics or principles.
The domains described by the 12 schemes also differ. These can naturally yield
different schemes. Of themselves they make no difference to the classificatory
quality of the scheme; it is merely a matter of what the scheme sets out to describe.
Of course, that in turn affects the general utility of the scheme. Here the range is
great but not enormous, for all of the schemes have been selected for review by
reason of their addressing one and the same broad domain, research questions. Yet
each scheme circumscribes the domain in particular ways that need to be noted.
The most grandiose schemes in this respect are Aristotle's. In the Posterior
Analytics he classifies the entire realm of knowledge into four generic questions. In
the Topics he classifies into 10 categories all the ways that an object may exist, or
all the propositions that can be made about it. The remaining schemes are more
modest yet ambitious enough. One of them covers all empirical inquiry (Rescher),
another all scholarly inquiry (Dillon). Two cover all scientific research (Bunge,
Laudan). Others cover quantitative research only, in a given field (Lundstedt,
Steiner); experimental research only, in several fields (Johnston & Pennypacker);
one discipline only (Fischer); all disciplines in the field of education (Shulman);
and noncausal inquiry in education (Smith).
The domain that a scheme circumscribes for classification has a general bearing
on the utility of the scheme, granted a sufficiency of other qualities. Utility is
governed in part by the comprehensiveness of a scheme, the extent to which it
classifies the universe of research questions in its chosen domain. Therefore this
one particular feature seems of interest to select for further analysis.

Comprehensiveness
The domain of immediate interest to this review is education, a broad field which
exhibits a variety of modes of inquiry. Most of the schemes may be examined and
compared for the extent to which they classify the kinds of questions found in
educational research journals-whether these be quantitative questions, noncausal
questions, and so forth. (Three schemes will be excluded from this analysis for the
noncomparability of their categories-Fischer, 1970; Laudan, 1977; Johnston &
Pennypacker, 1980.)
The method of analysis adopts the device used in the survey of schemes, whereby
categories were correlated with the scheme in Table I according to the correspond-
ence of their definitions and descriptions. That device may be said to have yielded
a theoretical assessment of comprehensiveness by noting the varying range of
categories and fineness of distinctions. Now analysis seeks to discover the empirical
comprehensiveness of the schemes insofar as their various categories account for
questions found in the journals.
Rather than applying each of the schemes serially, the procedure was to apply
the scheme in Table I to the journal questions and then to estimate the application

347

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

of other schemes by attributing to them the proportion of questions accounted for

by the corresponding categories of the Table I scheme. Two conditions must be
anticipated for this method. First, all of the categories in the other schemes correlate
with categories in the scheme in Table I, as described in the survey. Second, all of
the questions in education journals are accommodated in the Table I scheme, as
will be noted. Failing these conditions, a scheme-by-scheme application would have
to be made.

The first step was to classify by the scheme in Table I all 924 questions found in
a sample of nine education journals (one volume each), three each of theoretical,
empirical, and general research character. Two graduate assistants independently
coded the questions, as reported in Dillon (1983). (This step was taken prior to the
review in order to test the present scheme.)
The second step was to attribute to the other schemes the proportion of questions
classified by corresponding categories in Table I. As noted, all of their categories
correspond to (some of) the categories in Table I. For instance, if a given scheme
included a category corresponding to Table I's "character/description," and if 20%
of the questions in journals were classified as description, then the given scheme
was estimated also to classify 20% of the questions as description. For, if the given
scheme were actually to be applied to the questions, its category of description
would be found to classify 20% of the questions. The estimates were made to allow
for the greatest proportion of questions to be classified by other schemes. For
instance, a scheme with a class of "quantity" would be estimated to classify all
questions describing character, although quantity is only one element in that
category.
The third step was to sum the categories, yielding the total proportion of questions
classified by a scheme. For example, if a scheme had only four categories and each
accounted for 20% of the questions, then the total for that scheme would be 80%
of the questions in the journals; if five and 20% each (or four and 25%), then 100%
total.

The results are set forth in Table III. Because the domains covered by the various
schemes differ, Table III reports results only for that domain addressed by a given
scheme, for example, empirical research only.
As described earlier, the scheme in Table I designs to cover all the kinds of
questions in all modes of inquiry in education. In Table III this scheme is seen to
classify 91% to 100% of the questions found in this sample of the journal literature
(9% "unclear"). The only other scheme that designs to be as comprehensive is
Aristotle's (Topics), in that it covers all propositions that can be formed about a
given subject; this scheme classifies about two-thirds of the questions in these
journals.
The domains covered by the other schemes are less ranging and are variously
circumscribed. In addition, most of them have previously been analyzed as iden-
tifying a comparatively limited conceptual range of questions within their domain.
What is their empirical range?
Four schemes are designed for empirical research only. Accordingly, Table III
reports only the extent to which they classify questions from the empirical journals

348

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

in the sample (e.g., American Educational Research Journal). Rescher classifies

83% of the empirical questions, Bunge and Steiner two-thirds, and Lundstedt one-
third.
The remaining two schemes are designed to cover more than empirical research.
They include theoretical research but not also the kinds of inquiry found in general
journals (e.g., Education). Accordingly, Table III reports their results for empirical
plus theoretical journals (e.g., Journal of Educational Research plus Educational
Theory). Shulman's scheme accounts for about half of these questions and Smith's
for two-thirds. (As an indication of how these schemes might cover the broader
range of inquiries, Table III reports in parentheses that they cover fewer of the
questions in all journals than they do for only the empirical and theoretical ones.)
On the whole, then, those schemes revealed by theoretical analysis as more
comprehensive are also found empirically to comprehend a greater proportion of
the questions found in education journals. But what is more, these same schemes
are also found to be more comprehensive in the various particular domains
addressed by those other schemes that are theoretically less comprehensive in the
classes they distinguish for that domain.
Table III shows that the scheme that has been proposed as theoretically most
comprehensive (Table I) accounts for 99% of empirical questions, whereas the four
schemes designed particularly for empirical research (Bunge, Lundstedt, Rescher,
Steiner) account for as little as one- and two-thirds of empirical questions. That
scheme also accounts for more questions in both theoretical and empirical journals
than do the two schemes designed for them (Shulman and Smith).
In addition, the results suggest that schemes that take a categorial approach to
classification are empirically more comprehensive than those that take a noncate-
gorial approach. That finding may serve to confirm the suggestion derived from
theoretical analysis.

TABLE III
Educational Research Questions Classified by the Schemes
Proportion of questions (n = 924), by journal type

Scheme Em pirical and

Empirical ical theoretical and
theoretical
general

Categorial
Aristotle (Topics) 89.1% 70.9% 61.9%
Dillon (Table I, infra) 99.2 90.4 91.2
Rescher (1982) 83.5 n.a. n.a.
Semicategorial
Bunge (1967) 64.4 n.a. n.a.
Lundstedt (1968) 34.2 n.a. n.a.
Smith (1981) 79.5 69.5 (62.8)
Noncategorial
Shulman (1981) 37.4 48.3 (42.9)
Steiner (1978) 68.4 n.a. n.a.

349

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

Table III shows that the schemes that are found to account for the greatest
proportion of questions are those that take a categorial approach (Aristotle, Dillon,
Rescher). They account not only for more questions overall, but also for more
questions in various specific domains. For example, Rescher's and Steiner's schemes
are designed for empirical research; the first takes a categorial approach, the second
does not; the categorial scheme accounts for 83% of the empirical questions, the
noncategorial for 68%. Rescher's categories also account for more empirical ques-
tions than do the two noncategorial schemes that cover both empirical and
theoretical research (Shulman and Smith). The same case results from the other
two categorial approaches. Moreover, somewhat the same case results from com-
paring the semicategorial and the noncategorial schemes. As a group, the semica-
tegorial schemes account for more questions in both empirical and theoretical
research.
This finding does not confirm a hypothesis but it does strengthen the suggestion
that a categorial approach yields a more comprehensive classification scheme. The
point is strengthened in that the theoretical and empirical analyses were undertaken
independently for complementary but separate purposes-the first to see what
kinds of questions the various schemes distinguished, the second to see how many
questions their various classes accounted for. Neither analysis was undertaken in
view of grouping the schemes into categorial and noncategorial.

Other Criteria and Issues

Other criteria are available for the analysis of classification schemes. They will
not be applied here to the several schemes under review. Rather, analysis turns to
the nature of these criteria insofar as they may affect the task of classifying research
questions and assessing the schemes that result. The discussion identifies the criteria
and evaluates their application, raising issues and illustrations where appropriate
to schemes that classify research questions.
Any number of theoretical, statistical, logical, and pragmatic criteria have been
proposed for classification, and diverse sets of criteria have been applied, or the
same criteria applied in different ways, to assess classificatory schemes in education.
(For illustrations of how a taxonomy of educational objectives has been analyzed
by a variety of criteria and procedures, see Furst, 1981, with Seddon, 1978; and,
for sophisticated general treatments in education, see Cullinan, 1969, and Travers,
1980.) For instance, two contemporaneous meta-schemes for codifying curricula
and classifications of curriculum (Ariav & Atkins, 1983; Harris, 1983) specify
different, even contradictory, lists of criteria, having in common only the single
criterion that the schemes be "comprehensive."
As an example and a useful summary of the range of criteria, Derr (1973, chap.
5) assessed his taxonomy of social purposes for schools according to four kinds of
tests. These may be described as follows, with suggestive questions added for the
classification of research questions.
1. Theoretical. To what extent is the principle yielding the scheme consistent
with its underlying rationale? What are the rationale and principle of the scheme?
What is the domain or universe of discourse, and how has it been circumscribed?
2. Reliability. How reliably can the scheme be used to classify the kinds of

350

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

questions it distinguishes? How precisely has each of the classes been defined? What
is the definition of the thing whose kinds are being classified "research question"?
3. Logical. Derr borrows four "rules" from various sources.
a. There must be only one defining characteristic at each step in the division.
Have all of the classes been distinguished according to the same principle? What is
the defining characteristic, and how may a research question be seen to exhibit it?
b. The classes must be mutually exclusive. To what extent may all research
questions that exhibit one certain characteristic be grouped into the one correspond-
ing class, and all those that do not possess that characteristic may not be found in
that class? (Derr's reliability test seems to apply here.)
c. The division must be exhaustive, such that (i) there is a place in the division
for all members of the universe of discourse, and (ii) the sum of the subclasses
equals the whole class that has been divided. What proportion of research questions
find no place in the scheme? What characteristics do these exhibit, and what new
arrangement might be conceived to incorporate them?-or, on what principle are
they to be excluded? This kind of test may be performed by pondering the logical
possibilities (e.g., via dichotomy) or by inspecting the empirical actualities (e.g.,
questions studied in research journals).
d. The successive steps in the division should be in accordance with one principle
of division. For a nontaxonomical classification, this principle is similar to (3a).
What is the principle whereby kinds of questions have been identified and ordered?
What are the relations among kinds thus exhibited?
4. Practical utility. For what uses has the scheme been designed? How well does
it prove to satisfy these uses? What kinds of thought and action about research
questions does the scheme permit?
Other classifiers have made much of the criterion of "neutrality," whereby a
scheme is not biased against incorporating phenomena of various possible kinds.
For example, Bloom's (1956) taxonomy was proposed to incorporate educational
objectives from all schools of thought. Two recent classifications of curriculum
diverge on this point, one making of neutrality a primary criterion (Harris, 1983),
the other specifying that it is not a criterion (Ariav & Atkins, 1983). The philosopher
of classification, Broadfield (1946), rejects the very possibility of neutrality, for an
interesting reason.

A common dictum is that classification should not be critical. Whatever

precautions a classification may take, it will be critical. For it is a system of
expressed judgments.... The endeavor to avoid criticism can have only one
result-the monumentalising of beliefs which are looked upon, wittingly or
otherwise, as being beyond criticism. Thus an attempt to achieve impartiality
can become an insidious form of dogmatism. (p. 78)

As for logical criteria, the list of 14 principles of classification specified by

Inhelder and Piaget (1964) may be taken as a useful summary of the possibilities.
Ten apply to additive classification (p. 48), four in addition to multiplicative
classification (pp. 152-153). For example, the first group formulates the familiar
principles of mutual exclusivity and joint exhaustiveness of classes, and one of the
second group runs as follows: All the elements of a class A, must also belong either

351

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

to A2 or to its complement A'2 (not characterized by the property a), but not to
both. It may be that too much is made of logical criteria, or too stringent an
application. Howsoever necessary they might be to the study of classification, they
probably cannot apply to schemes classifying research questions with the same
force that they apply in theory or in an exceptionally few domains.
Apart fromn the partitions of set theory and the domain of geometrical entities,
criteria such as mutual exclusivity and joint exhaustiveness of classes can only be
approximated in practice-so suggest classification theorists from different tradi-
tions (Broadfield, 1946; Koerner, 1981), and for good reasons. "The rule that
classification should be exhaustive is one that can seldom be followed, since
exhaustiveness is only guaranteed by insight into the necessary connections of
characters," says Broadfield (1946); "logic must yield to the establishment of types
with partial knowledge" (p. 28). Indeed, it is for the purpose of coming to know
the phenomenon that the task of classifying its kinds is undertaken in the first
place.
Classification may usefully proceed by methods of logical division, but it is not
itself logical division, nor are its principles the principles of logic. Broadfield (1946)
remarks on "how sharply classification declines from logical division" (p. 30). One
essential difference turns on the fact that we do not know all the characteristics of
the phenomena of interest (save those that we have constructed, such as geometrical
figures), otherwise we would not have to classify them. Other factors include the
necessity of choosing among characteristics whereby to classify the phenomena and
the difficulty of achieving exhaustiveness without miscellaneous divisions. Simi-
larly, Koerner (1981) suggests that the achievement of partition according to set-
theoretical principles is, when applied to nonmathematical entities, based on an
"unrealistic assumption" of exact or extensionally definite classes: "It disregards
the frequent occurrence of borderline cases; i.e., of objects that can with equal
correctness be accepted or rejected as members of a class" (p. 691). Moreover, a
classification scheme must always take its form also from the nature of the
phenomena or the domain it embraces. That is particularly the case in education
and social sciences generally, but also in biology (pp. 692, 693). In that connection,
Hempel (1965) observes in his chapter on taxonomy that the objects of scientific
research, or their characteristics fruitful for classification, often cannot be found to
have or not to have a given property so as to fall or not to fall in a given class;
rather than either/or, classification is a more-or-less matter (pp. 151-152). Finally,
in his book on moder symbolic logic and taxonomy, Gregg (1954) notes that "in
most taxonomic systems the categories overlap in various ways" (p. 67).
By no means do these theorists minimize logical criteria. Nor does this discussion
propose that they be dismissed. Rather it introduces a perspective that recommends
a measure of sophistication in applying logical criteria to schemes of research
questions. Classification theorists from various schools were quoted as expressing
caution in this regard. And, in recommending to educators Piaget's inventory of
classificatory operations, Travers (1980) too suggests that they be applied with
caution to schemes in education: "A classification system may meet all the criteria
of a logical classification system and still be completely trivial" (p. 15).
It might also be stressed that this discussion does not apply to the principles of
classification but to the principles of logic which are imposed by some on the task

352

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

of classification. The principles of classification can be observed with more rigor.

Accordingly, a scheme should, for example, distinguish and relate the species of
questions by one and the same principle for each genus, such that each of the kinds
represents, in Broadfield's (1946) neat phrase, one of "alternative realisations of a
common notion" (p. 36). For several fundamental reasons, some already noted,
this task cannot be made equivalent to logical division, set-theoretical partition, or
dichotomy.
Classification is, in principle, an incompleteable and uncertain activity. It is
foregone that all schemes will suffer logical deficiencies. Some will still be useful,
but that is another matter. Two good examples in this regard are alphabetical
indices and library classifications, which are illogical but useful schemes. Perhaps
the essential criterion (granted the sufficiency of others) is in the end a pragmatic
one: How does the scheme facilitate knowledge and action in the domain that it
classifies? What does a particular classification of questions permit researchers to
understand and to do about research questions?
In turn, this criterion depends on an ultimate one-the choice of characteristic
used to classify the phenomena. The characteristic chosen is the most fundamental
of all possible features of a scheme, because it distinguishes the kinds that determine
the classes in a scheme. It is the first and most important issue faced by any
classificatory effort: Which characteristic of the phenomena-research questions-
should be selected for use in classifying them?

Unfortunately, selecting the characteristic is a matter of prejudgment and near

luck. There are no rules and few guides to making the choice. There seems no way
to know beforehand which (several) of all possible characteristics (some presently
unforeseen or unknown in any event) will prove to be the most useful for classifying
and thus knowing the domain. Several characteristics must be chosen for the several
schemes that are needed. Which ones are of most interest? Further, which of these
can be used reliably for classification? At one extreme, for example, research
questions can be classified according to their length or linguistic form. These
characteristics will yield a reliable classification but one of little interest. At the
other extreme is the significance of the question, a characteristic most important
and fruitful to know but unreliable to use for classification.

The purpose for raising this issue here is not to propose a criterion for evaluating
the several schemes reviewed. It is to identify a matter underlying the very task of
classification, and to stimulate thought about how to approach it. What can we
most usefully know about research questions and to what extent can that be reliably
used to classify them? As noted, we need several schemes for knowing questions in
their several characteristics. Yet the choice of these characteristics remains a chancy
matter, while some of the most interesting ones (e.g., significance) remain a difficult
matter to classify. Then the pragmatic issue will arise: To what uses can this
interesting and reliable classification be applied?

Application of Schemes
In general, classification gives expression to the manner in which we conceive of
the world, and it provides a framework for our thinking and acting within it (see

353

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

Koerner, 1970). What are the particular uses of schemes that classify research
questions? Several uses will be outlined for the understanding, the practice, and the
pedagogics of inquiry.

Understanding of Inquiry
A classification scheme serves as one analytic tool for understanding inquiry at
three levels: the individual study, a corpus of studies, and the entire enterprise of
research in a given field.
Understanding of an individual study is enhanced by construing in light of a
scheme the particular question posed for study. The scheme (better, several
schemes) reveals selected characteristics of the question that in turn influence
characteristics of the proposed answer. These characteristics delineate the terms in
which the question-answer proposition assumes a particular meaning, especially in
contrast to other meanings that may permissibly be drawn from it in the absence
of the specification that a classification scheme provides. The understanding that
results may then be compared with the interpretations and applications that the
researcher has made of the findings.
A still richer understanding is yielded by a scheme that has been constructed
following a categorial approach. Such a scheme will reveal not only particular
characteristics of the research question but also the category of the phenomenon in
question; the scheme permits knowing something further about the question by
force of having known that one thing about it. What comes to be known about the
question for study is first, the generic question of which it is a particular variant;
second, the category of things constituted by the generic question; and third, the
relations between the question and (a) other particular questions of that kind, (b)
other generic questions, and (c) other categories of the phenomenon in question.
In that way a classification can reveal both the propositional and the contextual
meaning of the question-answer pair represented by the study's problem and
conclusion. The study is apprehended not only for its particulars but for its kind,
as well as for its relations with other questions and categories in other studies in
the literature-studies of the same kind and related kinds, studies that have been
reported and those that are now discovered, in light of the scheme, remaining to
be undertaken.
In that respect classification serves at a second level, to enhance understanding
of a body of studies on some topic. One can devise one's conception of the research
specialty by arranging the extant studies according to the questions they address,
since a classification scheme provides a conceptual framework to begin with. That
procedure stands in contrast to starting with an assemblage of individual studies,
trying to make sense of them one by one, and then grouping them into "clusters"
which they are made to form. A classification systematically distinguishes the
clusters to begin with, then allocates individual studies within them. In Broadfield's
(1946) revealing turn of phrase, we can use classification to arrange things, "but
the classification was an arrangement of our knowledge of their genera and species,
and was completed before the things were touched" (p. 43).
Classification is a conceptual act. What is classified are not the individual
questions but the kinds of question. The result is a set of concepts, or classes. When
familiar usage has it that we "classify" (individual) questions, our act consists in

354

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

attributing to the individual questions (one of) the kinds that we have conceived
questions to be, and then collecting the individuals into classes according to their
kind. So too in "classifying" studies by their questions: We begin with a classifica-
tion, then allocate individuals within it. The classification thereby permits under-
standing of the individual studies as kinds, their relations with other kinds, and the
kinds of studies that are emphasized or neglected in the body of literature. That
particularly yields a conceptual picture of the knowledge available in the domain,
for by proposing answers to questions of various kinds the studies are contributing
different kinds of knowledge.
Accordingly, a classification of questions is a useful device for organizing and
reporting a review of research, thereby proposing a public conception of the field.
Studies are reviewed by the kinds of question they raise and thus the kinds of
knowledge they propose in answer. In addition to helping distribute the individual
studies, the classification also exhibits the categorial distribution of knowledge in
the domain. It provides a way to assess at once the individual studies, the available
knowledge, and the further research needed. To that end, a categorial scheme might
further be used in conjunction with a topical one specifying the particular content
of the field. In effect, topical schemes of questions classify not the question but the
subject matter in question. They reveal that something is known about a thing,
rather than the kind of knowledge it is about that kind of thing. A categorial scheme
of questions reveals what kinds of question are being asked, and thus what kinds
of knowledge are yielded, about each of the various kinds of thing constituting the
subject matter of the domain under review.
At yet a third level, classification can serve to understand the entire enterprise of
inquiry in a given field. As shown in the preceding section, the various schemes
can be used to analyze inquiry in the field of education as a whole or subfields and
disciplines within. From such studies may come warrant for statements of a form,
"Inquiry in discipline A is characterized by X and Y kinds of questions but not Z
kinds; discipline B by Z but not X," and so forth. (For an instance of such a study
in the field of curriculum inquiry see Dillon, 1984.)
As a result, researchers can more clearly conceive of the enterprise in which they
work. This understanding serves in turn as a useful complement to a priori
prescriptions and assertions about what inquiry consists in.
To illustrate the contrast with assertions, Smith (1981) introduces his scheme of
questions by proposing that "causal research is the dominant mode of inquiry in
education" (p. 23). "Only causal questions tend to be asked," Smith asserts, "to the
exclusion of all other types of questions" (p. 23). He recommends that researchers
begin to ask a variety of noncausal questions, of which his scheme specifies eight
types. As noted, the eight noncausal types represent five categories in Table I. All
together these five were found (Dillon, 1983) to account for half of the questions
in the education journals examined, including half of the empirical studies. On the
other hand, causal questions of all types were found to account for only one-tenth
of the questions (30% in empirical journals). That is, the type of question asserted
to be used predominantly if not exclusively by educational researchers was found
to be used in less than a third of the cases examined, while the specified types
recommended for use were already being used in half of the cases.

355

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

As for prescriptions about inquiry, a prominent methodology text (Kerlinger,

1979) stipulates that a good research question "should express a relationship
between two or more variables" (p. 32). This same criterion is reiterated in several
other texts on educational research (e.g., Hopkins, 1980; Tuckman, 1978). What-
ever the merit of this prescription, it does not describe the majority of questions
examined in educational research (Dillon, 1983). All together only half of the
questions, including half in empirical journals, were found to entail a relation of
any kind between two or more variables-whether comparative or contingent
relations in any of a dozen categories and subclasses (Table I).
In that way a classification of questions can yield clearer theoretical and empirical
understanding of the enterprise of inquiry in a given field. It can help to comprehend
the variety of questions that might be raised and those that are actually being
pursued by researchers. In both respects, it can serve as well for the practice of
inquiry as for the understanding of it.

Practice of Inquiry
A classification of questions can serve for the practice of inquiry in ways that
complement its uses for understanding inquiry. Generally it can serve for identifying
and formulating the question for research, for conceiving of the study initiated by
the question, and for designing a program of research.
A classification helps the researcher to perceive the range of questions that might
be asked about the phenomenon for study. One and another scheme can identify
characteristics of the phenomenon that may be of interest and the kinds of
knowledge that might be sought about any given aspect. In addition to the
theoretical range of questions, the scheme helps to perceive specifically those kinds
of questions that have been raised in the research literature. It identifies selected
kinds of questions that a researcher might now fruitfully ask, either because that
(kind of) question has not yet been raised or that (kind of) knowledge thought of
most interest has not yet been provided. At the same time the scheme specifies
(e.g., through its generic questions) the formal ways in which the research question
can be phrased according to the kinds of characteristics and knowledge in question.
Rescher's (1982) scheme reviewed here is an example.
Beyond helping to identify and formulate the question, a classification can be of
use in conceiving of the study that the question initiates. Schemes that give the
formal terms for their various kinds of questions can prove especially useful in
anticipating the set of possible answers and the form of the eventual answer, as also
the choice of appropriate method. An analysis of one's question provides a priori
understanding of the limits of the answer and thus provides a way of conceiving
the inquiry that is being undertaken and of construing its significance once
completed. Of the schemes reviewed here, Lundstedt's (1968) is an example.
Certain schemes can be useful for designing a program of research, not simply a
single study. Most schemes describe a range of questions, but only certain schemes
(e.g., Table I) distinguish a series of questions. Using the first kind of scheme the
researcher identifies any one of several types of question to raise, now for this
study, now for that. Using the second kind of scheme, the researcher also identifies
the question of interest but thereby knows further at least two additional questions
that are prior and subsequent to raising it-whence a program for research.

356

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

Once the researcher identifies the question of interest, the scheme suggests a way
of determining whether that question can be raised. If it can, the question is
followed by one of the subsequent kind in the series distinguished by the scheme.
If not, a prior question must be raised and answered. These determinations will
take their form from the particular principle whereby a scheme orders its kinds.
For example, the scheme in Table I operates by a form of logical entailment:
Question B is conceived to presuppose A, and C to presuppose B. On that principle,
B cannot be raised in the absence of the resolution of A; more precisely, if A
remains indeterminate, then B does not arise. (This view relies on the theory of
erotetic presupposition formulated by Belnap [1969; Belnap & Steel, 1976].) It is
not quite the case that B is invalid. Rather, B is not known to be valid, and
therefore cannot be validly posed.
Using such a scheme, the researcher asks, "To what extent is it meaningful to
ask my question? Can it be safely asked?" Safety refers to the risk of having the
question corrected upon the answering, versus protected in the asking (see Harrah,
in press). The risk of basing a study or a program of research on questions whose
presupposition remains indeterminate is that a single later study, from another
tradition or line of research, may correct the question-and invalidate the findings
from the entire series of studies-by demonstrating the falsity of the presupposition
to some prior question in the series.
The logic of question sequences, such as inferential series, has not received much
systematic study. A recent reviewer (Harrah, in press) identifies these problems
remaining: "to evaluate sequences of questions with respect to their answer-yield,
their safety, or other properties, and to compare sequences with respect to various
concepts of containment and equivalence" (p. 73, ms.). These surely apply to all
schemes reviewed here. Nonetheless a particular risk attaches to schemes that
appear to promise more, or more complex, knowledge in answer to a higher order
question. For, whatever the sum and complexity of that knowledge, it may be false;
and it will be false if the lower questions have either not been answered or answered
falsely, or answered truly but otherwise than as presupposed by the higher question.
Only in that one case-where the lower question has been answered, and truly,
and as presupposed-does the higher order question yield reliable as well as more
knowledge.
In all hierarchical schemes it is by definition the case that higher questions
contain lower questions. Therefore it is the case that to ask the higher entails asking
the lower, and that the answer to the higher entails an answer to the lower. But it
cannot be the case that the answer to a higher question will supply the answer to a
lower, by reason of "containing" it. The higher may indeed contain the lower, but
the truth of the higher depends on, rather than demonstrates or supplies, the truth
of the lower. In the absence of answers to lower questions, all the higher question
can do is to presuppose their answers, not establish them (save logically). To the
contrary, the answer to lower level questions provides what is necessary to ask the
higher question. Hence a research question may be at risk if it is posed on the
strategy that it is better to ask a higher, more significant question because it contains
the lower, trivial questions.
A specific example is Braithwaite's (1953) theory of the structure of a scientific
system. His view is of a deductive system that orders scientific hypotheses by levels

357

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

of diminishing generality, such that "from some of the hypotheses as premises all
the other hypotheses follow" (p. 12), with the result that "the logical strength of the
hypotheses increases the higher their level" (p. 17). The system is safe in the
principle whereby the establishment of the system depends on the establishment of
its lowest level hypothesis (p. 13). The risk appears in the strategy of verification
whereby the evidence for any higher level hypothesis is said to become indirect
evidence for the lower hypotheses that logically follow therefrom (pp. 17-18). If a
research strategy were to be designed on the view that to test a higher order
hypothesis would represent a productive and economic way of settling the issue in
one fell swoop (rather than inching one's way up through the lower levels), then
the strategy would be at risk, because the higher order hypothesis is riskily posed.
Its resolution may be true; but if the lower hypotheses are not only indeterminate
because untested but also infirmed if tested, then the superordinate resolution must
prove false. The problem would lie in not knowing that the superordinate resolution
proves false, because the very posing of that hypothesis had presupposed the truth
of subordinate hypotheses. Indeed, the risky strategy presumes to demonstrate the
truth of that which it has presupposed to be true and hence the truth of which it is
not testing at all, let alone demonstrating.
In that respect, a classification that orders kinds of questions by priority as well
as increment of information (e.g., Table I) frames the question of whether the
increment of knowledge to be yielded by a higher question will also be a reliable
increment of true knowledge. The scheme identifies which questions that are of
interest to ask are also safe to ask. (Perhaps the significant questions are not safe
questions.) The scheme reveals in what order questions are safely asked, specifying
which question must be asked first, which one may come next, and which one may
follow in the program of research. Such a scheme incidentally also offers a fruitful
way of ordering and evaluating studies in a review of literature. (Perhaps it is time
to ask some of the trivial, low level questions.)

Pedagogics of Inquiry
The uses of classification for the understanding and practice of inquiry make
plain its pedagogical application in methodology courses, doctoral seminars, and
dissertation tutorials. Because no one scheme can comprehend the domain of
research questions, several schemes are well used together.
Schemes of different sorts can serve to instruct the student about the nature,
kinds, and functions of research questions. They can help the student to perceive
of the range of questions that might be asked, to identify the question for research,
to give it form, to understand its character, and to situate its place in relation to
other questions studied in the domain. The schemes can help the student to
conceive of the domain for study and to organize the review of research. They can
help in determining the appropriate method and in anticipating the eventual
answer. They provide a framework for conceiving of the dissertation study, for
construing its significance, and for planning a possible program of research to
follow the dissertation.
As part of their earlier training, graduate students might use the classification
schemes to analyze individual studies and modes of inquiry in the journals in a
given domain or discipline. As part of their later training, doctoral students might

358

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

construct their own classifications of questions suitable to their special domain or

topic, each category constituted by a generic question of which the question for
dissertation and the questions in related studies constitute particular and interre-
lated variants. The student's production of a scheme constitutes in itself an
advanced level of conceptualizing, because a classification embodies as well as
portends a theory of the field.
In that way both professors and students can join in using classification of
questions to understand and to practice inquiry at the level of the individual study,
the body of research literature in their domain, and the enterprise of inquiry in the
field of education.

Conclusion

What are the kinds of questions that may be posed for research? On the whole,
little is known of the answer. To know the answer, theoretically or empirically,
requires a classification of research questions; few classifications are available, none
satisfying the question.
The enterprise of inquiry generally, not alone in the field of education, lacks an
encompassing conception of the nature and kinds of questions for research.
Classification supplies this lack. It provides the terms for a theory of inquiry and
the framework for thought and action in the domain of research questions.
This review reveals the need for more and better classifications: a set of different
schemes to know the several characteristics of research questions; and schemes
more systematic in conception, more searching in compass, and more sophisticated
in classificatory character. Together these classifications will enhance the under-
standing and the pursuit of inquiry. For, as Aristotle first taught and as Socrates
first exemplified, it is in questions and their answers that our knowledge consists.

References
Ariav, T., & Atkins, E. (1983, April). Curriculum classification systems: Do they do us any
good? Paper presented at the annual meeting of the American Educational Research
Association, Montreal.
Belnap, N. D. (1969). Questions: Their presuppositions, and how they can fail to arise. In K.
Lambert (Ed.), The logical way of doing things (pp. 23-37). New Haven, CT: Yale
University Press.
Belnap, N. D., & Steel, T. B. (1976). The logic of questions and answers. New Haven, CT:
Yale University Press.
Bloom, B. S. (Ed.). (1956). Taxonomy of educational objectives. The classification of educa-
tional goals. Handbook I: Cognitive domain. New York: McKay.
Braithwaite, R. B. (1953). Scientific explanation: A study of the function of theory, probability
and law in science. Cambridge: Cambridge University Press.
Broadfield, A. (1946). The philosophy of classification. London: Grafton.
Bunge, M. (1967). Scientific research: Vol. 1. The search for system. New York: Springer.
Cullinan, P. (1969). Processes and problems in taxonomic studies. In D. E. Griffiths (Ed.),
Developing taxonomies of organizational behavior in education administration (pp. 3-25).
Chicago: Rand McNally.
Derr, R. L. (1973). A taxonomy of social purposes of public schools: A handbook. New York:
McKay.

359

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
J. T. DILLON

Dillon, J. T. (1982). The multidisciplinary study of questioning. Journal of Educational

Psychology, 74, 147-165.
Dillon, J. T. (1983). The use of questions in educational research. Educational Researcher,
12(9), 19-24.
Dillon, J. T. (1984, April). The character of inquiry in curriculum. Paper presented at the
annual meeting of the American Educational Research Association, New Orleans.
Fischer, D. H. (1970). Historians'fallacies: Toward a logic of historical thought. New York:
Harper & Row.
Furst, E. J. (1981). Bloom's taxonomy of educational objectives for the cognitive domain:
Philosophical and educational issues. Review of Educational Research, 51, 441-453.
Gale, S. (1978). A prolegomenon to an interrogative theory of scientific inquiry. In H. Hiz
(Ed.), Questions (pp. 319-345). Dordrecht, Holland: Reidel.
Gregg, J. R. (1954). The language of taxonomy: An application of symbolic logic to the study
ofclassificatory systems. New York: Columbia University Press.
Harrah, D. (in press). The logic of questions. In F. Guenthner & D. Gabbay (Eds.), Handbook
of philosophical logic. Dordrecht, Holland: Reidel.
Harris, I. B. (1983, April). Codification of educational practices: Characterization of a set of
essential topics. Paper presented at the annual meeting of the American Educational
Research Association, Montreal.
Hempel, C. G. (1965). Aspects of scientific explanation. New York: Free Press.
Hintikka, J. (Ed.). (1981). Scientific method as a problem-solving and question-answering
technique. Dordrecht, Holland: Reidel. (Special issue of Synthese, 47, Whole no. 1)
Hopkins, C. R. (1980). Understanding educational research. Columbus, OH: Merrill.
Inhelder, B., & Piaget, J. (1964). The early growth of logic in the child: Classification and
seriation. New York: Harper & Row.
Jevons, W. S. (1874). The principles of science. London: n.p.
Johnston, J. M., & Pennypacker, H. S. (1980). Strategies and tactics of human behavioral
research. Hillsdale, NJ: Erlbaum.
Kahn, C. H. (1978). Questions and categories: Aristotle's doctrine of categories in the light of
modern research. In H. Hiz (Ed.), Questions (pp. 227-278). Dordrecht, Holland: Reidel.
Kerlinger, F. N. (1979). Behavioral research: A conceptual approach. New York: Holt,
Rinehart and Winston.
Koerner, S. (1970). Categorialframeworks. Oxford: Blackwell.
Koerner, S. (1981). Classification theory. Encyclopedia Britannica (15th ed., Vol. 4, pp. 691-
694). Chicago: Encyclopedia Britannica, Inc.
Laudan, L. (1977). Progress and its problems. Berkeley: University of California Press.
Lundstedt, S. (1968). A note on asking questions. Journal of General Psychology, 79, 229-
239.

Meyer, M. (1980). Science as a questioning-process. Revue Internationale de Philosophie, 34,

49-89.

Reid, W. A. (1981). The deliberative approach to the study of the curriculum and its relation
to critical pluralism. In M. Lawn & L. Barton (Eds.), Rethinking curriculum studies: A
radical approach (pp. 160-187). New York: Wiley.
Rescher, N. (1982). Empirical inquiry. London: Athlone.
Taylor, P. H. (1979). Introduction: Curriculum studies in retrospect and prospect. In P. H.
Taylor (Ed.), New directions in curriculum studies (pp. ix-xii). Brighton, England: Falmer.
Seddon, G. M. (1978). The properties of Bloom's taxonomy of educational objectives for the
cognitive domain. Review of Educational Research, 48, 303-323.
Shulman, L. S. (1981). Disciplines of inquiry in education: An overview. Educational
Researcher, 10(6), 5-12.
Smith, N. L. (1981). Noncausal inquiry in education. Educational Researcher, 10(3), 23.

360

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms
 RESEARCH QUESTIONS

Steiner, E. (1978). Logical and conceptual analytic techniques for educational researchers.
Washington, DC: University Press of America.
Travers, R. M. W. (1980). Taxonomies of educational objectives and theories of classification.
Educational Evaluation and Policy Analysis, 2, 5-23.
Tuckman, B. W. (1978). Conducting educational research (2nd ed.). New York: Harcourt.

AUTHOR
J. T. DILLON, Assistant Professor, School of Education, University of
Riverside, CA 92521. Specialization: Question-answer processes.

361

This content downloaded from

39.40.25.84 on Mon, 25 Jan 2021 17:02:13 UTC
All use subject to https://about.jstor.org/terms