Acceptance of a Theory: justification or Rhetoric?


Noting that all observations are theory-dependent, Gergen (1991) argues that it is no longer possible to justify a theory objectively by appealing to empirical data because the data used to justify a theory are themselves theory-dependent. By the same token, psychological research is not really neutral with respect to moral or cultural values because psychologists necessarily see things in terms of existent cultural norms and moral values. What psychologists choose to investigate is influenced by these values. He further argues that, as empiricism is not viable, psychologists have to adopt an alternative means to choose theory. The tenor of his argument seems to suggest that psychologists may do well to adopt rhetorical analysis as an alternative.

It is Gergen’s (1991) contention that, in the absence of the veneers of objectivity and of being value-neutral, the practice of (i) juxtaposing non-creative theory and empirical data, and (ii) emphasizing methodological niceties, statistics and citation is just another form of rhetoric. Psychologists are made unaware of the rhetorical nature of their practice by (a) using the passive voice, (b) not using the first person pronoun, and (c) not using emotive language. Theoretical discussions are consequently intelligible only to the practitioners themselves.

It is bemoaned in the rhetoric-analytic account that novice psychologists are discouraged from "speaking beyond the warrants of their data" (Gergen, 1991, p. 25). Psychologists may do well if they follow their colleagues in economics and anthropology to adopt a more sophisticated form of rhetorical analysis. Psychologists are urged to free themselves from the current form of narrative, as well as the present practice of being detached from non-conceptual considerations. They are encouraged to develop creative value-sensitive and culturally responsive theories. Improved rhetorical skills will render these sophisticated and challenging theories more readily intelligible.

This synopsis of the rhetoric-analytic view suffices to make explicit the following issues which will be considered in subsequent sections:

[Q1] Do psychologists still subscribe to the type of empiricism represented in the rhetoric-analytic account?

[Q2] What are the difficulties of empiricism to psychologists as envisaged in the rhetoric-analytic view?

[Q3] Does the fact that there is no atheoretical observation render objectivity or non-circularity impossible?

[Q4] Is objectivity not possible when an individual’s action is being evaluated, or characterized, in terms of social conventions or cultural mores?

[Q5] Is scientific research, in the final analysis, necessarily value-dependent?

[Q6] What renders a psychological theory a good theory?



To use the term empiricism without making explicit as to what it refers to is neither informative nor conducive to well-informed discussion. It seems that empiricism is used as a catchall term (including characteristics of positivism) in the rhetoric-analytic account. Gergen (1991) attributes a host of methodological characteristics to contemporary psychological research. Some of the salient characteristics are (a) aspiring to be objective, (b) using experimental data to substantiate theoretical statements,(c) attempting to provide causal explanations for phenomena, (d) endeavoring to establish functional relationships among observable variables, (e) relying on operational definitions of theoretical terms, and (f) treating generality as a criterion of adequacy of theoretical statements. To the extent that one is dissatisfied with this list of methodological characteristics (particularly when the list is read casually), one may find the rhetoric-analytic account persuasive.

However, individual psychologists often subscribe to only a selective few of these methodological features. Nor do they necessarily commit themselves to those selected methodological characteristics with the same degree of thoroughness. Moreover, psychologists understand the exact nature of some of the aforementioned methodological features differently. The truth of the matter is that neither empiricism nor positivism is a reasonable characterization of contemporary psychological research, particularly research in cognitive psychology.

To complicate the matter even more, psychologists often use positivistic terminology to describe their methodology even when they have ignored some of the positivistic prescriptions. This positivistic veneer found in psychological research seems to be an historical accident. There was a particular way to look at Newtonian physics. Psychologists somehow have learned to treat that approach as the epitome of science (Greenwood, 1991). It is for these reasons that it remains an open question as to whether or not contemporary psychologists are closet positivists (see Tolman’s, 1992, collection of essays). Important to the present discussion is the realization that one can be dissatisfied with positivism, or the catchall empiricism, without finding the experimental method of cognitive psychology wanting.


A specific issue is implicated in the empiricism question raised in the rhetoric-analytic account. The issue is one about the relationship between theory and data. Attributed to psychological research are the following characteristics:

(i) Atheoretical observations precede theorization.

(ii) Research rationale is based on induction by simple enumeration.

(iii) The only means to evaluate a theory is its "correspondence to available fact" (Gergen, 1991, p. 19).

The first two characteristics are not typical of contemporary psychology (see Greenwood, 1991), particularly cognitive psychology (see, e.g., Collins & Quillian, 1969; Craik & Tulving, I975; Sternberg, 1969). A casual commentator may, nonetheless, find the first characteristic true of behaviorism as exemplified by Hull (1943) and Skinner (1938). However, even this may be too much of a concession to the rhetoric-analytic view. More specifically, Hull (1943) chose which variables to examine in an experiment with reference to a to-be-verified empirical generalization. Moreover, research in both the Hullian and Skinnerian traditions were guided by some methodological or meta-theoretical decisions. For example, only mathematical and logical rules were used in deriving experimental hypotheses from empirical generalizations in the Hullian approach. Consequently, theoretical disputes could be (and often were) settled by appealing to methodological or meta-theoretical criteria, such as coherence, logical validity, clear operational definitions, and the like in the heyday of behaviourism. That is, even among Hullian and Skinnerian researchers, disagreements were not settled by rhetoric.1

As for the second characteristic, only someone taking Skinner’s (1938) metatheoretical position literally would consider induction by simple enumeration as the rationale of’ scientific research. Part and parcel of experimental psychologists’ data collection procedures are four of Mill’s (1973) inductive methods.2 This may be readily seen from the structure of a basic experiment whose design consists of two conditions (viz., the experimental and control conditions) which are identical in all relevant aspects (i.e., the control variables) but one (viz., the two levels of the independent variable). This is generally an exemplification of Mill’s (1973) Method of Difference (Boring, 1954, 1969; Chow, 1987a). At the same time, Mill’s (1973) Method of Concomitant Variation underlies many experiments (including those in the Hullian and Skinnerian approaches) in which a functional relationship is established between an independent and a dependent variable in the context of a set of relevant control variables. These inductive methods are more sophisticated than the induction by enumeration method attributed to psychologists in the rhetoric-analytic account.

The third characteristic attributed to cognitive psychology is incorrect. The actual practice of a majority of cognitive psychologists may best be described in terms of Popper’s (1968/1962) conjectures and reputations (or more commonly known as the hypothetical-deductive) perspective (Chow, 1987a, 1992). As will be seen in a subsequent section, evidential support for a to-be-corroborated theory is not just any "available fact", but observations collected in a failed attempt to falsify the theory.

In short, it is not Justified to attribute Characteristics (i) through (iii) to contemporary psychology. Nor are the three features accurate characterizations of psychological research before the advent of cognitive psychology. Questions [Q3] through [Q6] will be dealt with primarily from the perspective of cognitive psychology. Nonetheless, an attempt will also be made to show that the rhetoric-analytic approach does not fare any better outside cognitive psychology.


A central thesis of the rhetoric-analytic approach is that it is impossible to warrant acceptance of a theory with empirical data because what counts as empirical evidence depends on a researcher’s theoretical interpretation. Moreover, a researcher’s choice of methodology is influenced by the cultural milieu in which the research is carried out. Implicit in these rhetoric-analytic assertions is the contention that any appeal to empirical support is necessarily circular. However, this assessment of the relationship between theory and empirical data in empirical research (including experimental psychology) is misleading because the following four distinctions have not been given due consideration in the rhetoric-analytic account. The four distinctions are (a) a to-be-corroborated theory versus the theory which underlies evidential data, (b) prior (to-be-explained) data versus evidential data, (c) non-experimental research versus theory-corroboration experimentation, and (d) objectivity versus generality.


It is accepted here that all observations are theory-dependent. If’ Theory TA is the sole all-embracing theory, it necessarily follows that all observations owe their meaning to Theory TA in the final analysis. Does it not mean that appealing to empirical support for theoretical discussions is necessarily circular? This charge of circularity is debatable because one can accept that all observations are theory-dependent without accepting that there is only one theory which is all-embracing. The truth of the matter is that there are numerous independent theories belonging to different domains. Moreover, these numerous theories also belong to various levels of abstraction.


It is easy to find the central thesis of the rhetoric-analytic approach attractive if one has in mind collecting data in a non-experimental setting (e.g., interview or naturalistic observation). Suppose that, on the basis of an interview, a clinician concludes that a patient is suffering from acute depression because the patient has made frequent references to (a) moral disapproval and guilt, (b) self inadequacy and humiliation, and (c) hostility directed inward (e.g., injury to self and punishment of self). That is to say, Symptom Categories (a) through (c) are prior data vis-à-vis the depression diagnosis. When challenged about the diagnosis, what cannot be offered as evidential data for the depression diagnosis by the clinician?

The clinician cannot justify the diagnosis by appealing to Behavioral Categories (a) through (c) themselves because these categories of verbal responses owe their meanings (viz., qua symptoms of depression) to Gottschalk and Hoigaard’s (1986) theory of depression. The clinician’s argument would be circular if Behavioral Categories (a) through (c) themselves are also used as evidential data for the diagnosis. It is indeed impossible to justify a theory with empirical observation if psychologists reason in this manner.3


Most contemporary psychological theories are theories within Manicas and Secord’s (1983) or Greenwood’s (1991) realist framework. That is, causally efficacious hypothetical mechanisms are postulated (e.g., Theory T) to account for observable phenomena (e.g., Phenomenon P). Suppose that Observation O is based on Theory TO and that Theory TO is independent of Theory T1. If Observation O is used to justify Theory T1, the reasoning implicated will not be circular. Two distinctions can be, as well as have to be, made in such an event. First, the to-be-explained phenomenon, P, is prior datum vis-à-vis Theory T1; and Observation O is evidential datum vis-à-vis Theory T1. Second, there is the further distinction between a to-be justified theory (i.e., Theory T1) and the theory which confers meaning to Observation O in a context unrelated to the research in question (viz., Theory TO).

It is instructive to consider why the distinction between prior data and evidential data is overlooked by critics of experimental psychology. Psychologists are often introduced to research methods in the course of learning statistics. Concepts implicated in experimentation and the concept of hypothesis testing are often introduced to psychologists in statistics textbooks with questions like, "Is Drug D effective in reducing hyperactivity symptoms?" or "Is Fertilizer F more potent than conventional fertilizers?" Mook (1983) called this type of experiments agricultural-model experiments. A characteristic of this type of examples is that experimentation is presented in the context of answering a specific, concrete practical question without any theoretical import (i.e., no reference is made to any explanatory theory in answering the "Is it so?" question).

Given the fact that the research objective of an agricultural-model experiment is not to test an explanatory theory, there is no to-be-explained prior phenomenon (i.e., there is no prior datum). As the experimental results are used to answer an Is it so? question, the data are not used as evidence for an explanatory theory. Hence, experimental data do not play the role of evidential data. Consequently, the logical relationship between a to-be-corroborated theory and the data which can be used to corroborate the theory is overlooked. Under such circumstances, the research hypothesis, experimental hypothesis and statistical alternative hypothesis are indistinguishable because the question why, or by virtue of what mechanism, does not arise (Chow, 1992, in press).

Three serious consequences of repeated exposures to agricultural-model examples are that (1) statistical hypothesis testing is confused with theory corroboration, (ii) the necessary distinction is not made between a research hypothesis and an experimental hypothesis, and (iii) the statistical alternative hypothesis is erroneously identified with an experimental hypothesis (Chow, 1988, 1989, in press).4 Furthermore, we may fail to distinguish between prior data and evidential data vis-à-vis a to-be-corroborated explanatory theory.


Cognitive psychologists conduct experiments to corroborate explanatory theories which implicate unobservable mechanisms. They rely on experimental data of a particular kind in their attempt to justify acceptance of a theory. More specifically, cognitive psychologists conduct theory-corroboration experiments, as opposed to the agricultural-model experiments. The rationale of theory-corroboration experimentation may best be illustrated with Sternberg’s (1966) study of memory search. This study is chosen because it is familiar to many psychology students. At the same time, this familiarity seems to have brought about indifference. It is hoped that we do not lose sight of the logic, or rationale, of theory corroboration exemplified by Sternberg’s (1966) argument as a result of over-exposure. The following description of Sternberg’s ( I 966) experiment will serve to (a) show why it is incorrect to suggest that using empirical data to corroborate a theory is necessarily circular, and (b) illustrate why cultural values and linguistic norms do not render objectivity impossible.



The to-be-explained phenomenon investigated by Sternberg (1966) was the everyday observation that we could reliably retrieve with ease a few things which have been committed to memory. The research question was about the internal mechanism responsible for such a feat of ours. Sternberg (1966) presented his subjects I, 2, or 4 digits only once at a fairly rapid rate (viz., 1.2 seconds per digit). Subjects were then shown a single digit (called the probe digit), and asked to decide whether or not the probe digit was among those items in the previously shown memory set. Subjects pressed the "Yes" key when they judged that the probe digit had been included in the memory set. They pressed the "No" key if the probe digit was not a member of the memory set. A real-time clock was activated simultaneously with the onset of the probe digit. The clock ran until a key was pressed. Trials resulting in incorrect response were re-presented to subjects. The dependent variable was correct reaction time (viz., the interval between the onset of the probe and when a subject pressed the correct response key).5


Sternberg (1966) considered three theoretical possibilities. Included in all three theories in an internal hypothetical mechanism scanning whose function is to compare the probe digit to the internal representation of the memory items. These theories were not tested directly. Rather, they were tested via their respective implications in the context of the following auxiliary assumptions.

First, it was assumed that, in the course of performing on the memory-search task, subjects had to encode the probe stimulus, compare the probe to items in memory, select response and execute the response, regardless of which of the three contending theories is being considered. Second, each of these stages took a constant amount of time to complete. Third, the time a subject took to make a response on any trial was the sum of the times required to complete each of these discrete stages.6 These three assumptions may collectively be called, information-processing assumptions.


The first theoretical alternative is the serial self-terminating model of scanning. The internal scanning mechanism interrogates members of a memory set one at a time. A response is made as soon as a match is found between the probe and a memory item (hence, the self-terminating label). In the long run, it is necessary to search only half of the memory set in positive trials (i.e., the probe is indeed among the memory seta). However, the full set has to be searched in negative trials (i.e., the probe is not among the memory set). The following two observations must be obtained in order to accept this theory as tenable. First, reaction time should increase linearly with increases in memory-set size. Second, the rate of increase in reaction times of positive trials should be half of that of negative trials. These two theoretical expectations have been depicted in the top panel of Figure 1.

Figure 1: Theoretical prescriptions of three models of scanning

The second model is the serial exhaustive model of scanning. Again, the search mechanism interrogates the memory items one at a time. However, a decision as to whether or not there is a match between the probe digit and a memory item is made only when all memory items have been interrogated (hence, the exhaustive characterization). If this theory is true, the reaction time functions for positive and negative trials should increase linearly with increases in memory set-size at the same rate. This theoretical expectation is shown in the middle panel of Figure 1.

Third, it is assumed in the parallel model that all memory items, regardless of’ memory set-size, are interrogated simultaneously. A response is made only when the interrogation of all items is complete. Consequently, the same reaction time should be observed at all memory set-sizes, as may be seen from the bottom panel of Figure 1.

As may have been seen, the three alternative scanning models differ in (a) the number of times the comparison stage takes place and (b) when, as well as how, the comparison process is terminated. It is further assumed in all three models that a "No" response is made by default. That is, it is prescribed in all three models that it takes longer to make a negative response than a positive response.


The rationale of Sternberg’s (1966) study schematically represented in Table 1 may be used to show that cognitive psychologists do not "entify" a hypothetical mechanism with their rhetoric, Gergen’s (1991, p. 28) assertion notwithstanding. Instead, they use a logical argument to show that a hypothetical mechanism is warranted by data collected for the specific purpose of corroborating a theoretical statement about the hypothetical mechanism.

As may be seen from Table 1, corresponding to each of the three theoretical alternatives is a syllogism. More specifically, Propositions [P1] through [P3] form the first syllogism for evaluating the serial self-terminating model. Propositions [P4] through [P6] make up the second syllogism for assessing the serial exhaustive model. The third syllogism is made up of Propositions [P7] through [P9] for testing the parallel model.

In every case, the major premise is a conditional proposition whose antecedent consists of (a) a set of auxiliary assumptions, and (b) the to-be-corroborated theory (i.e., the conjunction of propositions after If). Its consequent (viz., the proposition after then) is an implication of the to-be-corroborated theory in the context of’ the set of auxiliary assumptions. The minor premise is supplied by experimental observation.

To ascertain which of the scanning mechanism is tenable at the theoretical level, Sternberg (1966) set up his test conditions in accordance with an experimental design whose underlying logic (inductive) was Mill’s (1973) Method of Concomitant Variations Sternberg (1966) found that both the positive and negative functions increased linearly at the same rate with increases in memory set-size. These data were compared to the prescriptions of the three contending theories (i.e., the respective consequents of the three major premises shown in Table 1 or the three panels in Figure 1).

As may be seen from the top and bottom panels of Figure 1, experimental data do not match the theoretical expectations of either the serial self-terminating or the parallel model. That is, the consequents of Propositions [P1] and [P7] are denied by Propositions [P2 ] and [P8], respectively (see Table 1). Consequently the antecedents of [P1] and [P7] are rejected by virtue of modus tollens (Copi, 1965).

Sternberg’s (1966) data matched the theoretical expectation depicted in the middle panel of Figure 1. What this amounts to is that the consequent of Proposition [P4] in Table 1 is affirmed. On strict logical (deductive) grounds, no definite conclusion about the antecedent of Proposition IP4] is possible as a result of affirming the consequent of Proposition [P4] (Copi, 1965). However, the various experimental controls found in Sternberg’s (1966) experiment (viz., a constant presentation rate, learning of the memory set was ensured, etc.) render possible a tentative acceptance of the antecedent of Proposition [P4] (Chow, 1987a, 1992). That is to say, what is left is the tentative conclusion that the serial exhaustive model is tenable (i.e., warranted by the set of data in question) in the context of the auxiliary assumptions.

It may also be seen from the syllogisms in Table 1 that, when cognitive psychologists attempt to justify a hypothetical mechanism, they do not appeal to the to-be-explained phenomenon itself (viz., Gergen’s, 1991, "available fact", p. 19). Nor do they pretend to measure the hypothetical mechanism directly. What cognitive psychologists do is to use an implication of the to-be-corroborated theory as a prescription, against which experimental data are assessed. Underlying this reasoning is a syllogism.

Table 1. The Rationale of Theory Corroboration Illustrated With Sternberg’s (1966) Experimental Study of Three Theories of Memory Scanning.

To-be-explained phenomenon

(i.e., Prior data)

We can retrieve a few things from memory with ease.

Experimental Result

(i.e., Evidential data)

Correct reaction times increases linearly with increases in memory set-size at the same rate for both positive and negative probes.

Serial Self-terminating Model

Major Premise:

Minor Premise:


If (AA* + serial self-terminating scanning), then top panel of Figure 1. [P1]

Data are not like top panel of Figure 1. [P2]

Serial, self-terminating scanning is not tenable. [P3]

Serial Exhaustive Model

Major Premise:

Minor Premise:


If (AA* + serial exhaustive scanning), then middle panel of Figure 1. [P4]

Data are like middle panel of Figure 1. [P5]

Serial exhaustive scanning is probably true. [P6]

Parallel Model

Major Premise:

Minor Premise:


If (AA* + parallel scanning), then bottom panel of Figure 1. [P7]

Data are not like bottom panel of Figure 1. [P8]

Parallel scanning is not tenable. [P9]

AA* = Auxiliary Assumptions


Many cognitive psychologists accept Sternberg’s (1966) conclusion because of its logic, not because of his rhetoric. At the same time, Townsend (1971, 1972) argued that the rate of processing multiple items simultaneously might be inversely related to the size of a memory set because the processing capacity of the working memory was limited. An implication of this limited capacity parallel model is also what is depicted in the middle panel of Figure 1. Hence, Townsend (1971) argued that Sternberg (t966) had not unambiguously justified acceptance of the serial exhaustive scanning model. Important to the present discussion is that Townsend (1971) questioned Sternberg’s (1966) conclusion not because he found Sternberg’s rhetoric wanting. Nor is Townsend’s (1971) limited capacity parallel model an alternative rhetorical account. Sternberg’s conclusion was questioned for a conceptual reason. In sum, theoretical discussion in psychology has been incorrectly characterized as a mere exercise in rhetoric by Gergen (1991).


The hypothetical-deductive method schematically represented in Table 1 is not without critics. To its critics, the fatal feature of the hypothetical-deductive method is the fact that the antecedent of the conditional proposition, If . . . then, is always a conjunction of (a) a set of auxiliary assumptions (i.e., AA in Table 1), and (b) the to-be-tested theory. Suppose that experimental data do not match the consequent of the major premise. Modus tollens dictates that the conjunction of the theory and the auxiliary assumptions be rejected. In other words, either the auxiliary assumptions or the theory in question may be the culprit. Hence, an experimenter can easily retain a favored theory by adjusting the auxiliary assumptions (Folger, 1989).

What the critics are saying about the possibility of abuse (by blaming the auciliary assumptions) is true at the level of deductive logic. It may also be true of how some experimenters behave. However, that some players do not observe the rules of a game does not speak ill of the rules or the game (Chow, 1989, 1990). The real issue is whether or not the inevitable presence of auxiliary assumptions necessarily renders the hypothetical-deductive method unsatisfactory as a research paradigm. Nonetheless, critics’ reservations may be allayed when the following extra-logical restrictions imposed on experimenters are made explicit:

(i) The set of auxiliary assumptions often consists of [a] well-established theoretical assertions in cognate areas, [b] methodological assumptions underlying the experimental paradigm in use, and [c] assumptions about the experimental task (see Meehl, 1978, 1990).

  1. None of these assumptions can be denied without impunity in the event that experimental data do not match the prescription of the to-be-corroborated theory. First, the experimenter is obliged to explain why a particular auxiliary assumption is responsible for the discrepancy between experimental data and the theoretical prescription. Second, the experimenter is obliged to conduct new experiments to substantiate the explanation of the discrepancy and to test the original theory again.
  2. The three types of assumptions identified in (i) differ in terms of how binding they are on an experimenter. As an example, consider the case when an experimenter "saves" a to-be-tested theory by rejecting a well-established theory in a cognate area (e.g., Theory TX). To begin with, the experimenter is obliged to provide, as well as to justify, a new theoretical alternative to Theory TX in the said cognate area. In addition, it is necessary to design a new experiment in light of the changes made to Theory TX so as to test the theory thus "saved". For these reasons, an experimenter would be the most reluctant to temper with this type of assumptions.

(iv) It is not the stipulation of the hypothetical-deductive approach that a theory be substantiated, or that the choice between contending theories by achieved, with a single experiment. In fact, it has long been recognized that experimental investigation of theoretical issues is best characterized as a series of converging operations (Garner, Hake, & Eriksen, 1956). Regardless of which auxiliary assumption is changed, the modified conclusion so drawn becomes a post hoc hypothesis vis-à-vis the experimental data. For this reason, the experimenter is obliged to test the original theory anew in the context of the new auxiliary assumptions.

In short, in adopting the hypothetical-deductive paradigm, an experimenter is obliged to observe (i) through (iv) because of the nature of the logic. The difficulty of the paradigm at the logical level identified by its critics should be (and often is) rectified by carrying out converging operations guided by the to-be-tested theory.


We use language to describe our theories and data. It is assumed in the rhetoric-analytic and the social constructionist (Gergen, 1985) accounts that linguistic convention is responsible for a theory’s being what it is, as well as for data’s being what they are. For the sake of argument, suppose that this language-dependent assumption is taken seriously. It follows that the serial exhaustive model, as well as the experimental result, might be very different had Sternberg’s (1966) study been conducted by a member of another linguistic community. However, this critique is not one about the relative merits of the three alternative models considered by Sternberg (1966). That is, it has no bearing on the objectivity, or non-circularity, issue. This may be seen from the argument underlying the language-dependent assumption.

To a rhetoric-analyst, cultural values and linguistic norms may affect research activities in the following way. First, there is a culturally shared and transmitted (by language) system of values and norms (loosely call it, Tcm). In addition to Tcm, there are numerous theories in various domains (e.g., T1, T2 T0, etc.) which are entertained by virtue of, or are meaningful in the context of, Tcm. That is, even though Theories T1, T2 T0, and the like may be independent of one another, none of them is independent of Tcm in the sense that none of them would be what they are without Tcm. Suppose that Observation O is used to justify Theory T1, and that Observation O is dependent on T0. Observation O is, hence, dependent on Tcm in the final analysis by virtue of the fact that T0 owes its meaning to Tcm.

In more concrete terms, a rhetoric-analyst may argue as follows. The all-embracing cultural milieu (Tcm) underlying the conjectures and reputations approach consists of the requirement that experimenters are knowledgeable and capable of applying both deductive and inductive logic. Success of this methodology also depends on the provisions that (i) an experimenter and the subjects interpret the experimental task in the same way (a process influenced by language), (ii) the subjects find it proper, as well as are prepared, to carry out the experimental task consistently throughout the experiment (a process determined by cultural values), and (iii) experimenters subscribing to the conjectures and refutations approach have made a commitment to achieve conceptual rigor by following the prescriptions of deductive and inductive logic.7

That is to say, the adoption of a particular research approach is under the influence of a system of cultural values and linguistic norms. Consequently, what counts as evidential data depends on the language, conventions and socially agreed upon codes of behavior. The rhetoric-analyst can then suggest that objectivity and non-circularity of empirical evidence are illusions under such circumstances.

This argument for the impossibility of avoiding circularity, or achieving objectivity, is not correct for the following reasons. First, the steps implicated in corroborating a theory are undertaken after a commitment has been made to achieve conceptual rigor. The exact nature of the steps required to achieve conceptual rigor, as well as their respective consequences, is not beholden to the reason why this commitment is made in the first place. This may be seen from the fact that a theoretical dispute is one between two or more theories within the same cultural milieu (including the linguistic environment). That is, whatever the influences TCM has on T1 at the metatheoretical or methodological level, it can be ensured that they are also found on T2 to the same extent as well as in the same manner.

The crucial question underlying rhetoric-analyst’s language-dependent argument is whether or not this state of affairs renders it impossible to justify a hypothetical mechanism by appealing to empirical data, the description of which is constrained in a specific way by the linguistic properties of a natural language. The answer is "No", as may be illustrated with Sternberg’s (1966) example.

The serial self-terminating and parallel models are also subjected to the same linguistic constraints as does the serial exhaustive model. Acceptance of the serial exhaustive model is warranted by rejection of the serial self-terminating and parallel models in the same linguistic environment. This is not to deny that the language we use may have an effect on our theorization. This example demonstrates that cognitive psychologists collect evidential data in a way which makes it possible to exclude linguistic constraints as alternative explanations of experimental result.

The rhetoric-analytic argument for the impossibility of objectivity and non-circularity seems like an attempt to question whether or not psychologists should adopt conceptual rigor as a guideline for good research. It seems that the rhetoric analytic argument is an attempt to replace conceptual rigor with some other criteria. This is a legitimate endeavor. However, if this is indeed the intention of the rhetoric-analytic argument, the issue is no longer a methodological one. It becomes unnecessary (as well as misleading) for rhetoric-analysts to criticize the methodology of cognitive psychology. It suffices for them to make known their agenda.

In summary, it is helpful to emphasize that theoretical alternatives entertained by cognitive psychologists are proposed, as well as tested, in the same cultural milieu and linguistic environment. It is true that how cognitive psychologists handle their theories is subjected to the influences due to the meta-theoretical assumptions of the conjectures and reputations approach, as well as the linguistic conventions, psychologists subscribed to. However, the crucial point to note is that these influences are identical for all theoretical alternatives. Differences among alternative theories, hence, are not determined by the meta-theoretical assumptions. Nor are they affected by linguistic factors.


Our discussion so far has been concerned with theories which implicate causally efficacious structures or mechanisms which are unobservable. However, not all theories in psychology deal with hypothetical mechanisms. There are theories which are descriptive in tenor, and they appeal explicitly to cultural values, social norms or linguistic conventions. Examples of such a theoretical formulation are Peters’ (1958, 1969) analysis of emotions and motives and Greenwood’s (1991) realist account of human actions.

Suppose that Individual X is characterized as being jealous of his brother because X has failed to attend his brother’s wedding. Further suppose that this characterization is challenged, and that Individual X is said to be envious of his brother instead. To a rhetoric-analyst, this disagreement comes about because the two characterizations are based on different perspectives; and the dispute can only be settled by rhetoric by virtue of its theory-dependent nature. This rhetoric-analytic argument is also incorrect because there is an objective way of’ choosing between these two alternative characterizations.

The choice between the characterizations, jealous and envious, depends on a set of appraisal criteria (Peters, 1969). More specially, the jealous characterization is appropriate when (i) Individual X sees that his brother is in possession of something (viz., the opportunity to start a family), (ii) the said possession is something X finds desirable, (iii) X wants the possession, and (iv) X has not considered whether or not it is realistic for him to have a family. On the other hand, X may be properly characterized as being envious when, in addition to (1) through (iii), and instead of (iv), X considers himself as deserving of having a family as his brother (i.e., X considers it his rightful fate to be married). In other words, the distinguishing features between the two emotions, envy and jealousy, are (a) whether or not X has considered some realistic questions, and (b) what X’s idea of his rights is. Hence, this dispute may be settled by collecting new empirical evidence in regards to these two distinguishing features, not by rhetoric. In Greenwood’s (1991) terms, linguistic expressions of emotions have linguistic objectivity.

That is, the new empirical evidence consists of realistic considerations and ideas about his rights on the part of Individual X. This evidence is not the original action which prompted the two contending characterizations in the first place. Hence, it is not circular to appeal to this new empirical evidence to warrant the original characterization. At the same time, it should also be reiterated that both the being jealous and being envious assessments of X are made in the same cultural milieu. In other words, the influences of the cultural norms or linguistic conventions on these two characterizations are the same. Hence, the everything is relative implication of the rhetoric-analytic argument is also incorrect when applied to Peters’ (1958, 1969) or Greenwood’s (1991) descriptive accounts.


As has been noted before, linguistic objectivity is recognized by Greenwood (1991) because there are well-defined sets of criteria for proper utilization of various concepts of emotions and actions or their linguistic expressions. However, a rhetoric-analyst may point out that such a set of criteria is well-defined in a particular linguistic community at a particular time. A different set of criteria may be implicated in a different linguistic community. No such criteria may be found in a third linguistic community. Furthermore, the criteria themselves may change within a linguistic community over time. That is, the very basis of Greenwood’s (1991) linguistic objectivity is not invariant. It is very tempting to conclude that objectivity is not possible after all. However, to do that is to confuse generally with objectivity.

To begin with, to fully describe Individual X’s being jealous of his brother, it is necessary to make explicit the intentional object and intensional characteristics of the emotion, jealousy. More specifically, the intention object of X’s jealousy in our example is the individual to whom the emotion is directed (viz., X’s brother). The intensional characteristics of jealousy are those characteristics which differentiate one emotion from another emotion (viz., a desire for a family, being unrealistic, etc.) .

It is a central thesis of Greenwood’s (1991) realist position that the intensional characteristics of an emotion are independent of the whim of an individual or social collectives. Hence, it is not sufficient for a rhetoric-analyst to suggest that there may be inter-cultural or inter-lingual differences. It is an empirical question as to whether or not there is indeed inter-cultural or inter-lingual differences in the intensional characteristics of concepts of emotions or human actions. In Greenwood’s (1991) words, "It is always an objective question whether there are instances of aggression or feudal duties in other cultures, or instances of racism and dishonesty in our own, according to any social definition of these phenomena" (P- 33, Greenwood’s italic).

The rhetoric-analytic case is not strengthened even if there are indeed inter-cultural or inter-lingual differences in the putative intensional characteristics of concepts of emotions or human actions for the following reason. What the differences amount to is that what is said about one culture cannot be applied to another culture. That is, the set of criteria identified has limited generality. However, this observation does not affect the objectivity of the choice between two emotions (e.g., between jealousy and envy) by people within the same cultural or linguistic community. In other words, lack of generality is not a measure of lack of objectivity.


Some debatable features of the rhetoric-analytic argument may now be discussed in greater detail with reference to the rationale of theory-corroboration experimentation.


As may be recalled from Table 1, the to-be-explained observation is an everyday phenomenon of our being able to recall with ease a few things which have been committed to memory. This phenomenon can readily be explained by one of three explanatory theories. As the theories are proposed to give an account of the phenomenon, the phenomenon is a datum prior to the theories (hence, the prior data characterization). However, contrary to what is envisaged in the rhetoric-analytic account, cognitive psychologists do not use this available datum to choose among the three alternatives. Their argument would be circular if they do so. Instead, they collect special-purpose data to make the choice. These evidential data are different from the to-be-explained phenomenon, and they are evidential data collected after the three explanatory theories have been proposed.

Likewise, in everyday appraisal of human actions, the to-be-appraised action is prior datum vis-à-vis the characterization attributed to the action. Underlying every appraisal characterization is a set of criteria. A characterization is to be justified by examining whether or not its underlying criteria are met. It is for this reason that linguistic objectivity is possible (Greenwood, 1991). Behaviors relevant to the said underlying criteria constitute the evidential data vis-à-vis the characterization. That is to say, the prior data versus evidential data distinction should also be observed in situations governed by descriptive rules or norms.


The meaning of the variable time may be dependent on a theory in physics. At the same time, using reaction time as a dependent measure requires the set of three information-processing assumptions. However, what is said in each of the three scanning models affects neither (a) the meaning of time in the context of the said theory in physics, nor (b) the meaning of reaction time in the context of the three information-processing assumptions, nor (c) the meaning of the two response options (i.e., the meanings of the "Yes" and "No" keys). Hence, the meaning of the dependent variable correct reaction time is said to be atheoretical with respect to the three contending models of scanning. This is the case because the relationship between the information-processing assumptions and each of the three scanning models is identical. As a matter of fact, the theoretical perspective which underlies a dependent variable of an experiment must be independent of a to-be-investigated theory (see Feigl, 1970).

This demonstration of’ the atheoretical nature of Sternberg’s (1966) dependent variable provides an answer to the question "How is it possible to determine whether competing theories refer to the same entities, without reference to some other theory not contained in those under comparison?" (Gergen, 1985, p. 267). It also shows how Observation O may be used to corroborate Theory T1 when (a) Observation O is dependent on Theory To, and (b) Theory To is independent of Theory T1. This point may further be emphasized with Greenwood’s (1991) observation,

... the ‘theory-informity’ (i.e., theory-dependence) of observations only poses threat to the observational evaluation of scientific theories if the theories that inform the relevant observations (i.e., the theory which renders the dependent variable meaningful, e.g., Theory TO) are identical with (or entail) the theories that are the object of observational evaluation (i.e., the to-be-corroborated theories, e.g., Theory T1). (p. 105, my explications in parentheses)

In short, the choice among multiple theoretical alternatives is achieved in an objective manner by setting up a corresponding number of syllogisms. Experimental data provide the minor premise for these syllogisms. These data owe their meaning to a theory unrelated to the to-be-corroborated theories. Untenable alternatives are excluded by virtue of the modus tollens rule in formal logic. Tentative acceptance of a theory is made possible by adhering to one of Mill’s (1973) four useful inductive methods.


Gergen (1991) found contemporary psychological research unsatisfactory because it is assumed in the rhetoric-analytic account that methodological sophistication is pursued by psychologists at the expense of conceptual quality or sophistication. That this is not the case may also be seen from Sternberg’s (1966) study. It can be shown that concerns with methodological rigor do not mean replacing conceptual sophistication with technical complexity. Rather, the quality of a theoretical idea determines whether or not methodological rigor can be achieved.

Questions about methodology in theoretical discussion of experimental studies are invariably concerns about controls (Boring, 1954, 1969; Chow, 1987a, 1992, in press). For example, to be able to say that correct reaction time increases with increases in memory set-size, Sternberg (1966) required a valid comparison baseline (viz., the memory set-size of 1 condition). At the same time, he had to be sure that changes in reaction time could not be due to the fact that subjects learned memory sets of various sizes differentially well. Hence, Sternberg (1966) repeated all incorrect trials until all subjects had 100% correct responses.8 Likewise, it was also necessary to have the probe digit appearing equally often in the 2 and 4 temporal positions in Set-sizes 2 and 4, respectively. These methodological concerns can best be illustrated with reference to Table 2.

Table 2. The Relationship Between Inductive Logic and Experimental Design Illustrated with Sternberg’s (1966) Experiment.

Level of Independent


Independent Variable

Control Variables

Dependent Variable
















Shortest RT?








Medium RT?








Longest RT?


Variable Name

Level of Variable


Memory set-size

1, 2 or 4 memory items


Rate of digit presentation

1 digit per second


Order of testing of various memory set-sizes

Randomly determined


Ratio of positive to negative trials

Equal number of positive and negative trials (i.e., 1:1)


Error rate



Temporal position of probe

Distributed equally often among all temporal positions of a memory set-size


Correct reaction time

To be measured

In holding constant Control Variables 2 through 6 while varying Variable 1 from trial to trial in a random fashion, Sternberg’s (1966) experimental design exemplifies Mill’s (1973) Method of Concomitant Variation. As the same level of Variable 2 was used (viz., digits were presented at a rate of’ 1.2 seconds per digit), Sternberg (1966) could exclude rate of presentation as an explanation of his findings. The same argument is applicable to the other control variables. Consequently, Variables 2 through 6 can be excluded as explanations of the experimental result.9

It is not self-evident why considerations of this sort would diminish the quality of a theoretical idea when technical considerations are made after the to-be-validated theory has been proposed and examined at the conceptual level. The opposite is true for the following reason. The choice of control variables in an experiment obviously depends on (i) the to-be-corroborated theory and its contenders, and (ii) the auxiliary assumptions. For example, Sternberg (1966) did not worry about the nature of his visual display because this variable did not have any impact on the three theoretical positions. The variable rate of presentation, on the other hand, was important because it determined whether or not the memory items used in a trial were indeed learned. In general terms, an experimenter chooses the control variables with reference to the auxiliary assumptions, as well as the to-be-corroborated theory and its contenders.

In sum, an experimenter pays attention to technical details of an experiment not for their own sake. An experimenter’s concern with technical details is dictated by inductive principles. An experimenter has to care about methodology so as to ensure that extraneous explanations other than the to-be-corroborated explanation can be validly excluded. As the choice of the control variables depends on how well-formed the to-be-corroborated theory is, it is closer to the truth to say that well-formulated theoretical ideas lead to successful technical details of experimental studies.



Psychologists are encouraged in the rhetoric-analytic account to replace appealing to empirical data with conceptual critique when engaging in theoretical discourse. Moreover, it is argued that data are not necessary in conceptual critique because "the place of fact in a world of theoretical intelligibility becomes increasingly marginal" (Gergen, 1991, p. 21). To support this assertion, it is said that "no hard data changed hands" (Gergen, 1991, p. 21) in Chomsky’s (1959) critique of Skinner’s (1957) attempt to account for verbal behavior to terms of operant conditioning. Two observations may be made regarding Gergen’s (1991) appeal to Chomsky (1959). First, it is true that a theoretical position may be assessed in terms of the theory’s internal coherence, logic, and the like. However, if an empirical theory is being examined, the critique will not be complete if no reference is made to data. Second, Chomsky’s (1959) critique was devastating to the Skinnerian account of verbal behavior because he showed that, contrary to what Skinner (1957) said, people did not behave in the way envisaged in the Skinnerian framework. That is to say, the Skinnerian account is contrary to facts.

Consider the verbal-response category tact. It is said in the Skinnerian account that a child learns to utter "Milk" in the presence of a bottle of milk as a result of having been reinforced repeatedly in similar situations in the past. This is an example of the Skinnerian category, tact. Skinner (1957), moreover, suggested that a tact response would be emitted readily in the presence of its eliciting stimulus. It follows from this assertion that an individual should keep uttering "Book" when the individual enters a library or book store (viz., one utterance elicited by each book). Chomsky (1959) pointed out that people did not behave in this manner when in a library. That is to say, Chomsky (1959) showed that Skinner’s (1957) account of the response category, tact, is contrary to everyday observations.


Another reason why current psychological research is not congenial to the rhetoric-ana4ytic perspective is that contemporary psychological theories are deemed not "creative", "sophisticated", or "intelligible". However, it is doubtful that these criteria are appropriate ones to use when evaluating (a) the explanatory function of a theory, (b) the conceptual rigor of how a theory is being corroborated, and (c) the methodological rigor of how evidential data are being collected. This is the case for two reasons. First, if a theory provides an adequate explanation for a phenomenon, whether or not the theory is "creative" or "sophisticated" is immaterial. After all, a psychological theory is not a literary creation. Second, what counts as creative, sophisticated, or intelligible may be in the eye of a beholder.

Consider the fate of the dependent variable reaction time. Psychologists did not find it useful after its introduction by Donders (1969) because of a conceptual difficulty in Donders’ (1969) subtraction method. Sternberg (1966) resurrected the usefulness of reaction time by modifying Donders’ (1969) method. What Sternberg (1966) did might appear mundane, or merely technical, to one psychologist, but sophisticated and innovative to another psychologist. Likewise, Townsend’s (1972) treatment of the parallel versus serial issue may appear "merely technical" to one psychologist, but sophisticated to another psychologist. In the final analysis, how Sternberg’s (1966) contribution is characterized is really immaterial. The important points are that (a) Sternberg (1966) resurrected a useful research method which is congenial to the conceptual rigor objective, (b) Sternberg’s (1966) serial exhaustive model provides an adequate explanation for the to-be-explained phenomenon when its contenders fail to do so, and (c) there is non-circular evidential support for the serial exhaustive model.

A more serious objection to using Gergen’s (1991) criteria to evaluate a theory is that it is not clear what the envisaged purpose of theorizing is in the rhetoric-analytic framework. The characterizations creative, sophisticated, and intelligible may be appropriate if the purpose of theorizing is (a) to show off one’s literary skills, (b) to edify an audience, (c) to foster empathy, (d) to amuse the public, (e) to promote political and special-interest agenda or (f) to mystify everyone. For example, it is not difficult to see that a politically correct anthropomorphic theory of the universe couched in fashionable clichés is more readily intelligible to many more people than Einstein’s theory of relativity. Should psychologists aim at the former kind of theory? What is amiss in the rhetoric-analytic account is the neglect of the question as to whose intelligibililty a theorist should be concerned with.

In short, the criteria creative, sophisticated, and intelligible are inappropriate if one holds the view that a psychological theory serves to enhance our understanding of mental mechanisms and processes. Such an understanding is achieved by attributing to the mechanisms causally efficaciously properties at the theoretical level. That is, a theory makes it possible to explain a phenomenon of interest. For such a purpose, a theory has to satisfy the following criteria.

  1. It must be consistent with the to-be-explained phenomenon.
  2. It has to be internally consistent, as well as consistent with existent theories in cognate areas unless there are good reasons to the contrary.10
  3. It is formulated in terms explicit enough for the derivation of criteria of rejection (e.g., Proposition [P1], [P4] or [P7] in Table 1).
  4. Emotive terms should not be used, lest incorrect implications are derived from the to-be-corroborated theory.

Suppose that Theory T satisfies these four criteria. Its proponent may be pleased if it is deemed creative or sophisticated by colleagues. The proponent may find it even more gratifying if the theory is also deemed intelligible to the public at large. However, these non-conceptual characterizations are not the reasons why Theory T satisfies Criteria (1) through (4). On the contrary, had its proponent aimed at being creative, sophisticated or intelligible for its own sake, Theory T may not satisfy the four criteria.


It has been shown that it is possible to have observations which are atheoretical vis-à-vis a to-be-corroborated theory despite the fact that all observations are theory-dependent because an individual entertains multiple independent theories. Of interest here are the facts that these independent theories belong to different domains, and that the individual can (and do) keep them apart with ease. For example, two physicists subscribing to the same theory of light may subscribe to different moral theories. As another example, two psychologists entertaining different theories of self may behave in accordance with the same ethical theory. These four researchers may all subscribe to the same aesthetic theory when they talk about classical music.

In other words, an individual’s value system does not have an inevitable influence on the individual’s scientific endeavor in an insidious way, the rhetoric-analytic claim notwithstanding. It is true that this value system may determine a scientist’s choice of research topic. However, this is a decision made prior to, and independent of, the actual research process. An individual should not refer to this value system in the course of conducting the research. Nor should it determine how data are interpreted. If a scientist’s value system intrudes into theoretical discussions and research, it is a deliberate decision on the part of the scientist to sacrifice conceptual rigor so as to satisfy some non-conceptual agenda. This state of affairs makes it debatable to claim that scientific theory cannot be value-free.



This paper is a critical examination of the assertion that no non-circular appeal to empirical data can justifiably be made in theoretical discussion because all observations are theory-dependent. This assertion is incorrect because evidential data are not dependent on the to-be-tested theory in a circular way. Evidential data, in fact, owe their status to a logical argument. Realizing that one can (and does) entertain many theories in various domains makes it clear that scientific research can be (and should be) value-free. Some weaknesses of the rhetoric-analytic framework are readily seen if distinctions are made between (1) prior and evidential data, (2) non-experimental and experimental research, and (3) a to-be-corroborated theory and the theory responsible for the meaning of the dependent variable. The rhetoric-analytic argument would also lose some of its apparent attractiveness if it is realized that generality is not an index of objectivity. It is incorrectly asserted in the rhetoric-analytic view that psychologists "entify" their hypothetical mechanism by rhetoric and that cognitive psychologists succeed in maintaining methodological rigor at the expense of the quality of their theoretical ideas., While it is indeed gratifying to psychologists if their theories are found "creative", "sophisticated", or "Intelligible", psychologists should not aim at satisfying those literary criteria when developing theories because they are not germane to psychological knowledge. Introducing irrelevant criteria to theoretical and meta-theoretical discussion hinders rigorous progress at the conceptual level.

Siu L. Chow

Department of Psychology

Universiy of Regina

Regina, Saskatchewan

Canada S4S 0A2

Acknowledgment This research was supported by an operating grant to the author from the Natural Sciences and Engineering Research Council of Canada (OGPOO42159). Requests for reprints should be sent to Siu L. Chow, Department of Psychology, University of Regina, Regina, Saskatchewan, Canada S4S 0A2


1 I am grateful to an anonymous reviewer for this observation.

2 Mill’s (1973) five methods of scientific inquiry are (i) Method of Agreement, (ii) Method of Difference, (iii) Joint Method of Agreement and Difference, (iv) Method of Concomitant Variation and (v) Method of Residues. Method of Agreement is not useful as a means to exclude alternative interpretation of data (see Cohen & Nagel, 1934; Chow, 1987a, in press).

3 It may be argued that Behavioral Categories (a) through (c) are components of more or less generally accepted definitions of depression which are not dependent on any theory. What this amounts to is to say that a patient is characterized as suffering from depression by definition. The circularly difficulty still arises when the clinician appeals to the same set of symptoms when asked to justify the original diagnosis.

4 Distinctions should be made among theory, research hypothesis, experimental hypothesis and statistical alternative hypothesis. This may be illustrated with the study of iconic memory as follows:

[a] Theory: There is a sensory store in the visual system, whose capacity is larger than that of the short-term store.

[b] Research Hypothesis: Partial report by a sensory cue at very short cue-delays is superior to whole report.

[c] Experimental Hypothesis: Partial report by row at a cue-delay of 0 msec is superior to whole report.

[d] Statistical Hypothesis: The mean number of items correctly recalled in partial report by row at a cue-delay of 0 msec is larger than that of whole report.

The relative superiority of partial report and whole report may be measured in terms of (i) the number of items correctly recalled or (ii) the number of errors made. Hence, [d] is not synonymous with [c]. To instruct subjects to recall a randomly designated row of items in a partial report is to provide subjects with a sensory recall cue (Sperling, 1960). However, subjects may be cued to recall with a color (von Wright, 1968). For this reason, [c] is not synonymous with [b]. That [b] is not the same as [a] may be seen from the fact that Sperling’s (1960) version of partial report is not the only method used to study iconic memory; iconic store may also be studied with Estes and Taylor’s (1964, 1966) detection paradigm.

5 A recent development is to criticize Sternberg’s (1966) study on the grounds that it lacks ecological validity in Neisser’s (1976, 1988) sense of the term (viz., Sternberg’s experimental task is artificial; Sternberg’s argument is thus irrelevant to understanding everyday memory phenomena). This ecological validity argument is debatable (see Mook, 1983; Chow, 1987b; Greenwood, 1991, for rejoinders).

6 Some cognitive psychologists find the discreteness assumption wanting. For example, McClelland (1979) suggests that these stages may overlap in time. This would complicate the derivation of the respective theoretical expectations of the contending theories. However, this complication does not alter the fact that the meaning and identity of the dependent variable, reaction time, are independent of the three models being investigated. Nor does the complication affect the rationale of theory-corroboration experimentation.

7 A rhetoric-analyst may even dismiss the experimental approach by referring to (a) Orne’s (1962) idea of demand characteristics, (b) Rosenthal’s (1966) idea of experimenter effects (particularly, experimenter’s expectancy effect), and (c) Rosenthal and Rosnow’s (1975) idea of subject effects. The rhetoric-analyst may point out that these factors are sensitive to social or cultural values. Although these "social psychology of psychological experiments" factors are often accepted uncritically (particularly in introductory textbooks), their putative insidious effects on experimental results are debatable. Two observations may be made. First, it has to be noted that the minimal requirement of a properly designed experiment requires collecting data in two conditions which are identical in all aspects but one (Boring, 1954). The aspect in which the two sets of observation conditions differ is the independent variable. Second, the difference between the two (or more) levels of an independent variable found in an experiment is not confounded with cultural or social values. This is the case because all levels of the independent variable are used in the same social or cultural milieu (see Chow, I 987a, in press, for a detailed critique of the putative social psychology of the psychological experiment).

8 Only correct reaction times were considered by Sternberg (1966) because he limited his inquiry to error-free retrieval. This is admittedly an arbitrary decision. Consequently, the serial self-terminating model may be true only for error-frec performance. Sternberg’s (1966) conclusion may be faulted for its lack of generality. It is, of course, of interest to examine the relative superiority of the three models in a less-than-perfect recall situation. However, this issue is not pursued here because it is not germane to the present discussion. Moreover, as has been argued by Greenwood (1991), generality is not a criterion of theoretical adequacy. What is important is that the rationale of theory-corroboration experimentation described here is still implicated when the new empirical question is tested experimentally.

9 In following Mill’s (1973) Method of Concomitant Variation, Sternberg (1966) ensured that his experimental conclusion enjoyed internal validity. However, its generality may be questioned for the following reason. The correct reaction data might be very different if a different kind of stimuli was used. In using only digits, Sternberg (1966) effectively precluded the possibility of an interaction between stimulus-type and memory set-size. This is true. However, this must not be interpreted as an inherent weakness of the experimental approach. Nor does it render the rhetoric-analytic critique of cognitive psychology valid. This state of’ affairs simply means that theory-corroboration is not achieved with a single experiment. As soon as the question of the generality or hypothesis validity (Wampold, Davis, & Good, 1990) of an experiment arises, cognitive psychologists would have to (and they do) conduct further experimentation. For example, Burrows and his associates (Burrows, 1972; Burrows & Okada, 1975) used words (as opposed to digits) in a Sternberg (1966) type experiment. Research conducted by these investigators extended the generality of, as well as set the limiting conditions for, Sternberg’s (1966) serial exhaustive scanning model.

10 In such an event, the experimenter is obliged to propose, as well as to justify with new experimental data, an alternative theory.


BORING, E.G. (1954). The nature and history of experimental control. American Journal of Psychology, 67, 573-589.

BORING, E.G. (1966). Perspective: Artifact and control. In R. Rosenthal, & R.L. Rosnow (Eds.), Airtifact in behavioural research (pp. 1-11). New York: Academic Press.

BURROW, D. (1972). Modality effects in retrieval of’ information from short-term memory. Perception & Psychophysics, 11, 365- 371.

BURROW D., & OKADA, R. (1975). Memory retrieval from long and short lists. Science,188, 1031-1033.

CHOMSKY, N. (1959). A review of B. F’. Skinnerr’s Verbal Behavior. Language, 35, 26-58.

CHOW, S. L. (1987a). Experimental psychology: Rationale, procedures and issues. Calgary:


CHOW, S. L. (1987b). Science, ecological validity, and experimentation, Journal for the Theory of Social Behaviour, 17, 189-194.

CHOW, S. L. (1988). Significance test or effect size? Psychological Bulletin, 103, 105-110.

CHOW, S. L. (1989). Significance tests and deduction: Reply to Folger. Psychological Bulletin, 106, 161-165.

CHOW, S. L. (l990). In defense of Popperian falsification. Psychological Inquiry, 1, 147-149.

CHOW, S. L. (1992). Positivism and cognitive psychology. In C.W. Tolman (Ed.), Positivism in psychology: Historical and contemporary problems (pp. 119-144). New York: Springer-Verlag.

CHOW, S. L. (in press). Research methods in psychology: A primer. Calgary: Detselig.

COHEN, M. R., & NAGEL, E. (1934). An introduction to logic and scientific method. London: Routiedge & Kegan Paul.

COLLINS, A. M., & QUILLIAN, M. R., (1969). Retrieval time from semantic memory. Journal of Verbal Learning and Verbal Behavior, 8, 240-247.

COPI, I. M. (1965). Symbolic logic (2nd Ed.). New York: MacMillan.

CRAIK, F.I.M., & TULVING, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 1104, 268-294.

DONDERS, F.C. (1969). On the speed of mental processes. In W. G. Koster (Ed. And trans.), Attention and performance II (pp. 412-431). Amsterdam: North-Holland. (Reprinted from Acta Psychologica, 30).

ESTES, W. K., & TAYLOR, H.A. (1964). A detection method and probabilistic model for assessing information processing from brief visual displays. Proceedings of the National Academy of Sciences, 52, 446-454.

ESTES, W. K., & TAYLOR, H.A. (1966). Visual detection in relation to display size and redundancy of critical elements. Perception & Psychophysics, 1, 9-16.

FEIGL H. (1970). The "orthodox" view of theories: Remarks in defense as well as critique. In M. Radner, & S. Winokur (Eds.), Analyses of theories and methods of physics and psychology. Minnesota studies in the philosophy of science (pp. 3-16), Vol. IV. Minneapolis: University of Minnesota Press.

FOLGER, R. (1989). Significance tests and the duplicity of binary decisions. Psychological Bulletin, 106, 155-160.

GARNFR, W.R., HAKE, H.W., & ERIKSEN, C.W. (1956). Operationism and the concept of perception. Psychological Review, 63, 623-643.

GERGEN, K. J. (1985). The social constructionist movement in modern psychology. American Psychologist, 40, 266 275.

GERGEN, K.J. (1991). Emerging challenges for theory and psychology. Theory & Psychology, 11, 13-35.

GOTTSCHALK, L. A., & HOIGAARD, J. (1986). A depression scale applicable to verbal samples. In L. A. Gottschalk, F. Lolas, & L. L. Vinev (Eds.), Content analysis of verbal behavior (pp. 105-122). Berlin: Springer-Verlag.

GREENWOOI), J. D. (1991). Relations and representations. London: Routledge.

HULL, C. L. (1943). Principles of behavior: An introduction to behavior theory. New York: Appleton-Century.

MANICAS, P. T., & SECORD, P. E. (1983). Implications for psychology of the new philosophy of science. American Psychologist, 38, 399 413.

McCLELLAND, J. L. (1979). On the time relations of mental processes: An examination of systems of processes in cascade. Psychological Review, 86, 287-330.

MEEHL, P. E. (1978). Theoretical risks and tabular asterisks: Sir Karl, Sir Ronald, and the slow progress of soft psychology. Journal of Consulting and Clinical Psychology 46, 806-834.

MEEHL, P. E. (1990). Appraising and amending theories: The strategy of Lakatosian defense and two principles that warrant it. Psychological Inquiry, 1, 108-141.

MlLL, J. S. (1973). A system of logic: Rationcinative and inductive Toronto: University of Toronto Press.

MOOK, D.G. (1983).In defense of external invalidity. American Psychologist, 38, 379-387.

NEISSER, U. (1976).Cognition and reality: Principles and implications of cognitive psychology. San Francisco: Freeman.

NEISSER, U. (1988). New vistas in the study of memory. In U. Neisser & E. Winograd (Eds.), Remembering reconsidered: Ecological approaches to the study of memory (pp. 1-10). Cambridge: Cambridge University Press.

ORNE, M. T. (1962). On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 17, 776-783.

PETERS, R. S. (1958). The concept of motivation. London: Routledge & Kegan Paul.

PETERS, R .S. (1969). Motivation, emotion, and the conceptual scheme of common sense. In T, Michel (Ed.), Human action: Conceptual and empirical issues (pp. 135-165). New York: Academic Press.

POPPER, K.R. (1968). Conjectures and refutations: The growth of Scientific knowledge (originally published in 1962). New York: Harper & Row.

ROSENTHAL., R. (1966). Experimenter Effects in behavioral research. New York: Appleton-Century Crofts.

ROSENTHAL, R. & ROSNOW, R.L. (1975). The volunteer subject. New York: Wiley.

SKINNER, B.F. (1938). The behavior of organisms: An experimental analysis. New York: Appleton-Century.

SKINNER, B.F. (1957). Verbal behavior. New York: Appleton-Century Crofts.

SPERLING, G. (1960). The information available in brief visual presentations. Psychological Monographs, 74, (11, Whole No. 498).

STERNBERG, S. (1969). Memory-scanning: Mental processes revealed by reaction-time experiments. American Scientist, 57, 421-457.

TOLMAN, C.W. (Ed.). (1992). Positivism in psychology: Historical and contemporary problems. New York: Springer-Verlag.

TOWNSEND, J. T. (1971). A note on the identifiability of parallel and serial processes. Perception & Psychophysics, 10, 161-163.

TOWNSEND, J. T. (1972). Some results concerning the identifiability of parallel and serial processes. British Journal of Mathematical and Statistical Psychology, 25, 168-199.

von WRIGHT, J. M. (1968). Selection in visual immediate memory. Quarterly Journal of Experimental Psychology, 20, 62-68.

WAMPOLD, B. E., DAVIES, B., & GOOD R. H. III (1990). Methodological contributions to clinical research: Hypothesis validity of clinical research. Journal of Consulting and Clinical Psychology, 58, 360-367.