1 The social structure of psychological experimentation Psych 304 I noted that the establishment of psychological laboratories always involved the institutionalization of certain social arrangements. Psychological experimentation became a collaborative effort dependent upon the division of labor among individuals who carried out different functions in the experimental situation. Broadly between those who acted as the source of experimental “data” (subjects) and those who manipulated the apparatus/experimental conditions (experimenters). While this social arrangement was largely due to the “brass instruments” (technology) that were a part of the 19th c. institutionalization of psychology as an academic discipline, the subject - experimenter distinction also quickly became part of tradition. What was not noticed at the time, and not noticed for another century, was that this social arrangement created a very special kind of “social system” one – a system of “psychological experimentation”. The interaction between subjects and experimenters was regulated by a series of constraints that set limits to what passed between them. Their communication in the experimental situation was governed by the roles they assumed including implicitly adopted proscription and prescriptions. However, the specific features of this social system were not necessarily fixed. The division of labor between subjects and experimenters still left much room for local variation. Thus, there was nothing in the practical requirements of psychological experimentation that dictated a permanent separation between experimenter and subject roles. The experimental practices in Leipzig (Wundt’s lab) were ones of convenience; it did not necessarily prescribe that the roles of experimenter and subject could not be reversed from session to session. However, other factors did serve to distinguish the roles between experimenter and subject more permanently. Thus we can ask what factors might be expected to result in a permanent separation between experimenter and subject? One such factor is obviously that psychological experiments were not conducted in a vacuum. Thus, it is impossible to keep participants’ identity outside the laboratory from having an impact on their roles or identity inside the laboratory. Who would serve as experimenter and as subject may well be bound up with their status outside the laboratory. Similarly, the possibility of exchanging roles may depend on the social identity of those who are occupying these roles. Apart from these social factors external to the experimental situation, there is the question of how the distribution of roles ties in with other activities that are a part of the social practice of investigation. Thus, there are other activities that pertain to conducting the experiment such as its theoretical conception, data analysis, and the writing of the research report (article for a journal). The fact that these activities are today routinely done by the experimenter does not mean that this was always the case. Let’s see what the case was in Wundt’s lab. 2 If you examine Philosophische Studien (Wundt psychology journal) we see that the person who authored the article was not necessarily the experimenter. Nor was the person who conceptualized/designed the experiment necessarily the data source in the experiment. Wundt often appeared as the subject or data source in experiments conducted by his students, although he clearly also contributed to the conceptualization of the experiments. Interestingly, Wundt hardly ever took the role of experimenter which suggests that the person who functioned as the data source (subject) had a more sophisticated task than the person in the experimenter role did and which Wundt left to his students. Obviously, this seemingly practical exchange of roles meant that there was no reason to make the distinction between experimenter and subject permanent. In fact, in Wundt’s lab experimenter and subject often exchanged roles. The participants in the early Leipzig lab were engaged in a common endeavor in which they were all regarded as collaborators, including the data source or “subject”. Characteristically, authors of the published articles would refer to their experimental subjects as “co-workers” especially given that their proper names were often included in the article as well. Importantly, playing the roles of subject or experimenter was part of the participants’ relationship outside the lab. They often knew each other well, were friends, fellow students, or professor and student. Thus, experiment collaborators were not strangers. Of course, Wundt held a special position as professor but once he participated in a study his responses (data) were treated like everyone else’s. Most of the subjects in the Leipzig lab were members of the lab (usually students) but also others who had theoretical interests and were well-acquainted with experimentation. The scientific community at Leipzig produced their own data on the human mind! This style of experimentation was the result of (1) a tradition in German universities of linking the training of intellectual elites with the production of new knowledge, and (2) the idea that psychological experimentation was intended to investigate the general states and processes of the normal, mature human mind. That’s is, psychological experimentation was designed to analyze this psychological “object” just as physiological experimentation was designed to investigate the biological processes in the normal mature organism. Just as the philosophical practice of introspection had helped to construct the object (private experience) it was meant to investigate, so Wundt’s systematic introspection (which was really a report on the direct experience of the stimulus) constructed it own object. Experimental subjects were not studied as individual persons but rather as examples (or instances) of certain common human characteristics (e.g., sensory discrimination). Hence, the role of subject could be taken on by anyone in the research community since they did not represent themselves but their common mental processes. The “elementary mental processes”, as Wundt called them, were assumed to be “natural objects” that could be studied quite independently of the rest of the subject’s personality (very much like we assume today). All that was necessary to achieve this was to restrict the conditions of the laboratory and give the subject some training in “introspection”. But this preparation (both in restricting the conditions of the laboratory set-up and the training of the subjects in introspection) together with the social arrangement wherein the 3 experimenter and subject could be exchanged resulted in the construction of the very research “object” which the research program was designed to investigate. This did not make the “object” any less “real” but it did mean that the “object” of Wundt’s lab had to compete with other psychological “objects” constructed in different ways. Alternative models of psychological investigation Clinical experiment While Wundt’s model of experimentation has a profound influence in the early years of psychology (which was aspiring to become a “science”), it was not the only model of psychological experimentation at the time. The clinical experiment had different goals (objects) and very different social arrangements. The first major research program using this type of experimentation was the study of hypnotic phenomena. This French program involved the systematic use of experimental hypnosis as a tool of psychological research. In this kind of research various psychological functions were investigated under experimentally induced hypnosis. Unlike the Leipzig program, in the clinical method there could be no exchange of experimental roles insofar as the clinical method depended on the permanent distinction between experimenter and hypnotic subject. Moreover, the experimenters were almost always male and the subjects almost always female. This arrangement is much like the modern conception of experimentation in psychology in that there was clear segregation between experimenter and the source (subject) of the data. The clinical method originated in the medical context. Those who functioned as subjects (and were labeled “hysterics” or “somnambulists”) were usually patients of the physician who was also the experimenter. When normal or healthy subjects were used they were always used to make comparison to diagnosed clinical subjects (patients). The experimenters were physicians and had a medical background and hence prior to the experiment the experimenter and subject were in a doctor-patient relationship and the experiment was essentially a continuation of this medical relationship. Thus, the whole situation of the experiment was defined in medical terms. A crucial feature of this definition was the understanding that the psychological states and processes that were being studied were something that the subject or patient suffered or underwent (“patient” means to be “suffering” to passively undergo a disorder). In this context of the clinical method we find the first use of the term “subject” in a psychological experiment. These medically oriented experimenters quite spontaneously referred to their patients as “subjects” (French.: sujet) because that term since the 18th c. been a practice of medical care or “naturalistic observation”. Before that time, “subject” referred to a corpse used for anatomically dissection and by the early 19th c. one spoke of patients as being good/bad “subjects” for surgery (i.e., would they suffer surgery well or not). When hypnosis came to be seen as a medical matter (certainly in Paris 1880s) there was nothing more natural than to extend the already linguistic practice to yet another object of medical scrutiny. However, within the medical context we immediately get the formulation of “healthy subject” (Fr.: sujet sains) when it is a matter of comparing the 4 performances of normal and abnormal individuals. From this distinction it is a short step to the generalized use of the term subject to refer to any individual under psychological investigation – a step Alfred Binet, for example, took immediately. Binet after some work on hypnosis turned to the study of infants which was possible without altering the clinical method (something that was clearly not possible using Wundt’s introspective method). The English term “subject” had also, like the French term, assumed medical connotations. As mentioned above, in the 18th c. it was used to refer to a corpse for anatomically dissection but by the middle of the 19th c. it could also mean “a person who presents himself for or undergoes medical or surgical treatment” – hence its use in the context of hypnosis. Thus, in the English language literature the first use of the term “subject” was in the context of the study of hypnosis (G. S. Hall one of the first American psychologists, used “subject” in this context of hypnotic research and referred to the subject as “percipient” or “observer” when he reported on experimental work with normal individuals). James McKeen Cattell in 1889 appears to be the first to use the term “subject” in psychological experiments using normal adults as data sources (see Ch. 8). However he was not at all sure about this use as he writes “subject or observer” and also uses “subject” always in inverted commas, a common custom at the time probably indicating that it was borrowed from the French. The third method: Francis Galton Thus, the earliest years of experimental psychology were marked by the emergence of two very different models of psychological experiment. The Leipzig model allowed for a great deal of fluidity in allocating social functions between experimenter and subject and if anything the subject was more important than the experimenter. In contrast these roles were rigidly segregated in the Paris clinical method of experimentation. Here the experimenter was clearly in charge and only the experimenter was fully informed about the nature of the experiment. The roles of experimenter and subject were definitely not interchangeable. The “objects” of study that these two experimental situations were meant to clarify (explain understand) were also very different. The clinical experiment was meant to display the effects (which could be any manifest bodily expression) of an abnormal condition for the benefit of a knowledgeable experimenter (and his research community of other physicians/ experimenters). In contrast the Leipzig experiment was meant to display the universal states/processes that characterized all normal adult minds and the significance was therefore open to all (not just the experimenter). Now while these two models of psychological investigation were very different and both very influential, there were several other models that began to emerge before the end of the 19th c. The earliest of these was in 1884 at the International Health Exhibition in London where Francis Galton set up a laboratory for “testing” the “mental faculties” of members of the general public. Galton’s experimental situation was clearly very different from both the Leipzig and the Paris models. Thus the people Galton investigated (“tested”) were not medically stigmatized but ordinary people/citizens. Nevertheless the 5 roles of investigator and “tested” individuals were definitely not interchangeable. Also there was a clear difference in the status of the two roles. Galton had expert knowledge about the individuals he tested, and he was willing to share this knowledge with them for a fee. That is, Galton charged every person who availed himself of his services in exchange for a card on which he recorded the individual’s “anthropomorphic” test results. While there was a fee for service this was definitely not a medical procedure; in fact, Galton referred to those who presented themselves for testing as “applicants” (cf. the modern term “client”). In fact, there were no lack of applicants (9000 applicants before the World Exhibition ended) which would suggests that the idea of being tested was probably familiar to the times so much so that they were willing to pay for the service. One reason for this familiarity was the existence of competitive school examinations. There was also the common practice of phrenology (by this time discredited) but which had been very persuasive and pervasive in the previous generation. Phrenologists offered people a “measurement” of their mental faculties for a fee (even Queen Victoria used phrenologists for her children and Galton himself had gone to a phrenologist and had taken it very seriously). If we compare the Leipzig or Paris model of experimentation to the social arrangement or structure of Galton’s investigative practice we see some revealing divergences. Galton’s social practice contained utilitarian and contractual elements that are quite original. Obviously what Galton offered was deemed valuable to his applicants, namely information (on a contractual basis) which they believed could be useful to them. It is likely that Galton’s applicants were interested in this information for the same reason their parents had been interested in phrenology. That is, knowledge about one’s mental abilities was very useful at a time when they depended on having marketable skills to obtain work in a rapidly technologically changing society (machines etc.). If one could obtain scientifically objective and reliable information on one’s mental skills this was not only of instrumental value in obtaining work but it was also relevant to one’s self-image and the desire for self-improvement (cf. self-help). From Galton’s own perspective, his large-scale “anthropological measurements” were not to make money (Galton was independently wealthy) but a way of implementing his eugenics program of selective breeding among those with outstanding anthropological scores. That is, Galton’s interest in a selective breeding program would be facilitated if it could be based on scientific data of human abilities. Thus, Galton’s interests were just as practical as those of his applicants; the difference being that the latter were interested in their individual advancement while Galton was interested in developing a rational foundation for his social-political program of eugenics. (Galton was not only related to Darwin; he bought into his later speculations about eugenics, hence, Galton was a proponent of the first evolutionary psychology.) 6 Comparing the three models of experimentation The “object” of investigation that Galton and his applicants created through their structured social interaction in the testing situation was very different from the “objects” in either the Leipzig or Paris laboratories, although it also had elements in common. Galton essentially produced sets of individual performances that could be compared to one another. Obviously they had to be individual performance – and not collaborative ones. While the performances were individual, the fact that they were a product of collaboration between the anthropometrist and his subjects (applicants) was not considered important. The performances were thus defined as characteristics of socially isolated, individuals which were designated as “abilities”. An ability belonging to an individual (something an individual could do) and the “object” of research converged to become (1) the individual defined as an assembly of characteristics (abilities), and (2) the statistical distribution of performance abilities in the population [Galton played around with the idea “distributions” and their “moments” (measurements e.g., mean) as well as with Pearson’s correlation coefficients]. It was obviously the latter (comparison of individual performances) that interested Galton whereas the former (individual performance) were of interest to his applicants. Locating the object of research within the isolated individual person was of course a basic common feature of all forms of psychological investigation; that is, this is what constitutes Psychology as an intellectual discipline and psychological investigation rather than some other kind of scientific investigation. But Galton’s experimental method was uncompromising in this respect. As we saw, Wundt was uncomfortable about the way in which the experimental method isolates the individual from his/her social cultural context and, hence, he limited the experimental method to “elementary” mental processes (and experimental psychology to only a small part of the discipline of Psychology with the largest portion going to his Volkerpsychologie). The French clinical experimentalists were interested in the inter-individual process of hypnotic suggestion (the experimenter induced hypnosis in his subject) which they regarded as a fundamental condition for the appearance of their subject’s abnormal states/processes which they wanted to investigate using hypnosis. But Galton’s anthropometry was quite radical in its conception of severing the links between the individual’s performances from his/her social-cultural conditions of those performances. He accomplished this by defining individual performances as an expression of innate biological factors thereby sealing them off from environmental influences (thus, Galton treated the individual person on the analogy to an organism, and he held that it was inheritance (of abilities) that was crucial in the performance of abilities). In fact, the appeal of Galton’s anthropometric method was based on three features. 1. A radical individualism which claims that abilities were stable and unalterable individual characteristics which owed nothing to the social conditions of the environment. This was the necessary theory behind the notion that performance of 7 individuals in test situations could be used as a guide to their performance outside the test situation in the real world. 2. There was built into this method the implicit element of inter-individual competition; that is, what was of interest was not the individual first of all but the individual compared to other individuals. This meant that any one episode of experimenter-subject interaction (measurement) was only of interest as part of a series of such interactions each one with a different subject. While in the other forms of experimental investigation this was also true, in these other investigations the series was constituted by variations of the conditions with the same subject. But in Galtonian experiments the series of experimental episodes was essentially defined by differences among subjects. Of course, the duplication of an experiment with different subjects was known both in the Leipzig and Paris modes of experimentation, but this duplication was merely to replicate for the purpose of reliability of the observations made. In case of Galton’s anthropometry the experimental episodes with different subjects formed a statistically linked series, and it was this series, rather than any of the individual episodes, that formed the essential unit or object in this form of investigation. Only in this way could Galton generate knowledge of the desired “object”: namely, a set of performance norms against which the individual could be compared. 3. Whereas the Leipzig model of experimentation only made use of the division of labor between experimenter and subject roles, Galton added to this the multiplication of subjects as a necessary component of the method. This was closely linked to a third feature of the method’s practical appeal namely the statistical nature of the information provided by Galton. To be of practical value, the comparison of individual performances had to be unambiguous, so that individual and social policy decisions could be based on it. The way to achieve this was to assign quantitative values to performances thereby allowing each individual performance to be related precisely to every other performance. In this context the individual’s performance and the individual subject became a mere “statistic”. This was of course very different from the “case” which was the object of the clinical method, or the exemplar of the generalized normal adult human mind which was the object of the Leipzig method. These differences in the status of the knowledge of the object were linked to certain features of the interaction between investigators and their subjects. In Galton’s method, the experimenter and his subjects were strangers, collaborating just for the duration of the experiment or “testing” (and contractually it turns out, although this was not a necessary feature of Galton’s method). In the clinical method the experimenter and subject were related outside the laboratory as doctor-patient. In the Leipzig lab the participants in an experiment were fellow investigators whose common interests and friendships often extended for many experiments over years. Thus in each method there was a certain fit between the kind of knowledge “object” sought and the nature of the experimental interaction (social arrangement) that was used to produce this “object”. 8 1. Galton: brief contractual interaction between strangers was the perfect vehicle for producing an object whose construction required a brief encounter with a large number of successive human subjects. 2. Wundt: The collaborative interaction of colleagues in exchangeable experimentersubject roles nicely suited the construction of an object defined in terms of common universal processes of the normal adult mind. 3. Clinical method: the object constructed in clinical experimentation depended in large measure on the tacit agreement among participants in order that the medical care context can be carried over into the experimental situation. Each of the three investigative situations constitutes a coherent pattern of theory and practice in which it is possible to distinguish three kinds of interdependent factors. 1. The factor of custom which provides a set of shared taken-for-granted meanings and expectations without which the interaction of the participants in the experimental situation could not proceed predictably. Much of this also involves general social-cultural meanings. For example, I distinguished several varieties of custom: those of the German 19th c. university research institute (Wundt), those of medical investigative practices (clinical method), and those of school, examination competition and phrenological assessment (Galton). These were sources of social interaction that were adapted to the purposes of psychological experimentation. Hence, psychological research practices were not invention out of nothing; yet once established, these psychological adaptations themselves produced new variants and established their own traditions. 2. The actual practices of psychological investigation, though derived from cultural patterns, were not identical with these patterns and therefore constituted a second distinguishable element in the emergence of local styles. 3. A third element was constituted by the diverging knowledge interests that I have already alluded to. Thus, the three kinds of psychological research were by no means interested in the same kind of knowledge. There is a world of difference between knowledge of elementary mental processes in the generalized normal human mind (Wundt), knowledge of psycho-pathological states/processes (clinical method), and the knowledge if individual performance comparisons (Galton). From the very beginning of the founding of psychology in the late 19th c. psychological investigators pursued distinct knowledge goals and worked out experimental situations appropriate to achieving these goals. None of this was a matter of rational choice. The knowledge goals were rooted in traditions as were the experimental practices employed to attain these goals. Different theoretical interests and practices were usually absorbed in a single cultural complex of theory and practice (we might call this a “disciplinary project”). Each kind of disciplinary project led in turn to the construction of the kind of knowledge “object” it has posited. In sum, Psychology emerged with a number of different models of psychological investigation and the differences were rather profound. Psychology as an academic 9 discipline then exists only as a number of different historical incarnations. While some models achieve dominance for a considerable period of time, this should not lead to believe that there is only one method. While questions about the scope and limits of psychological experimentation usually presuppose one method, it is historically more meaningful to compare the implications of very different forms of experimental practice. The triumph of the aggregate Psychology may have started out with three different kinds of investigative practice but very quickly (within less than 20 years; certainly by WW I, around 1914) one had completely disappeared namely, Wundt’s. In contrast, Galton’s method was to achieve dominance for the first fifty years of the 20th c. When one examines psychology journals and the kind of knowledge claims that are being made, one finds that they always have a basic propositional form. Some set of attributes is said to be true, of certain subjects, under certain conditions. What the “results” section of the article says is that one or more persons (subjects) did such and such when in a particular situation. The knowledge objects that were the legitimate focus of interest for scientific psychology involve the attributes of a specially constituted subject, namely the research subject. Usually it is the attributes rather than the subjects whose attributes these are, that are considered problematic for psychological research. But if we are going to get some insight into the changing investigative practices of the discipline, we will have to reverse this perspective and recognize the problematic nature of the research subjects themselves. For when the nature of the subject changes then the nature of the attributes do as well. For instance when a subject is defined as an individual consciousness (e.g., Wundt) its attributes will be very different from a subject defined as a biological population (e.g., Galton). Subjects are constructed in accord with various conventions about the proper objects of psychological knowledge – and we need to deconstruct these subjects to understand something about the changing nature of investigatory practices in psychology. Virtually all research reported up to WW I, in psychology journals show that experimental results were attributed to individual experimental subjects. Individuals were identified by proper name, letter, or initials but the report is one of attributing certain response to specific individuals – usually competent “observers” (and not naive subjects). Thus, it was typically individual response patterns and not averages that were reported of individual competent observers whose testimony was deemed to be more reliable than naïve observers. The question is how was this pattern of reporting related to psychological knowledge claims? Obviously, the individually identified research subject could only function in this way because of the way the psychological object of knowledge was defined (as contents of an individual private mind). So the basic propositions were constructed with the individual mind as the subject of the sentence and the mental contents as the predicate of the sentence. Attaching the name of a reputable observer was a way of establishing the scientific credibility of the data. 10 Now this reporting of patterns of individuals observers did not mean that the phenomena studied were idiosyncratic; on the contrary, claims for the general or even universal significance of the data were commonly made – usually because of a presumed (same) underlying physiology say in case of the study of sensations. The claims for generality were supported by repeating the experiment with a few individuals. Where replication was not successful, relevant personal information on the observer was used to rationalize the lack of replication. There were essentially two versions of this model: (a) a strong version inspired by Wundt for whom the individual mind was a synthetic unity constituted by real processes of psychic causation, and (b) a weaker version of Titchener who dropped all reference to real causal processes (Titchener was a philosophical “positivist” whereas Wundt was a philosophical “realist” with respect to mental content). There was in fact continuity between this tradition of psychological experimentation and physiological experimentation. Thus what distinguished physiological experimentation from say biochemical/biophysical studies on the one hand and biological population studies on the other hand, was that physiological experimentation restricted itself to the universe of the individual organism. In fact, the physiologist Pavlov whose work with conditioned reflexes provided a kind of objective analogue to the experimental work in mentalistic psychology was similar to Wundt’s in that both saw the purpose of experiments in terms of exploring a specific system of causal processes, although for Wundt this system was mental whereas for Pavlov it was physiological. Both took the individual or organism as a unit of study. Obviously there was an asymmetry between how Wundt allowed mentalistic knowledge claims to be a function of the individual and Pavlov who maintained that physiological knowledge claims were always strictly a function of the universal organism. Nor did knowledge claims as attributions of individuals have anything to do with whether the claims were qualitative or quantitative in nature; in fact, whether the claims were physiological or mentalist the data was usually quantitative (as witnessed by the quantitative nature of the psychophysical experiments which were still the mainstay of psychology). Error was calculated on a population of observations not on individual subjects. Using more than one subject as in conducting another experiment was merely a replication. If inter-individual variability was large this was deemed to be prima facie evidence that the attempted isolation of the critical determining factors had failed to be controlled for in the experiment. Alternative basis for knowledge claims: psychological consensus Where this early tradition of experimentation proved to be inadequate was when psychology changed it object. Psychological knowledge claims were now attributed to the collective rather than individual subjects. In other words, like Galton’s method, the aim was now to make attributions of psychological features to aggregates of individuals. This meant that knowledge claims necessarily took on statistical form. For example, one could compare different age groups on some psychological characteristic. Usually such 11 comparisons were restricted to averages or ratios of the population meeting some criterion. This change in methodology (which reflected a move away from the model based on experimental physiology) was the result of the emergence of quantitative social research which had emerged earlier in the 19th c. Indeed, statistical studies of human conduct in the aggregate were already better established than were laboratory studies of psychological processes. Interest in statistics of crime, suicide, poverty, and public health was widespread even prior to the middle of the 19th c. both in Europe and NA. The statistical depiction of social problems seemed to promise a more scientific approach to their solution than simple political ideology (this was obviously important to the establishment of Sociology/Economics). On a practical level the development of social statistics was closely linked to the spreading use of the questionnaire as a method of investigation. Thus, G. S. Hall used the Berlin municipal statistical office model to investigate the minds of American children which, along with the developing field of medical statistics, led to the compilation of psychological information on children by circulating questionnaires to their mothers. The popularity of descriptive statistical inquiry in the 1880s made the questionnaire method seem like an appropriate tool for investigating questions of a psychological nature. Darwin used it in studying emotional expression and his cousin Galton used it to study mental imagery and heredity (see Chs. 5, 6). The use of statistical data based on questionnaires was used by many who were deeply influenced by Darwin and hence less concerned about individual minds than about the distribution of psychological characteristics in populations. Hall who became the chief promoter of child development studies in the US used questionnaires extensively to study children. What made the questionnaire an acceptable tool of psychological research was not just practical convenience, but there was the belief that mass data constituted a valid basis for psychological knowledge. However, this belief was quite new in the late 19th c. and required some social and conceptual reformulations to make it plausible. [After all what could mass data tell us about the individual mind?] Let’s see what the background to this belief was. The scientific appeal of statistical tables depended on more than the reduction of complex issues to countable “facts”. It depended upon numerical regularities that began to appear when the actions of many individuals were aggregated. Rates of crime for example remained fairly constant in a particular area and variations among areas could be accounted for in terms of identifiable environmental influences (living in poor neighborhoods). It was not experimental psychology but the repeated demonstration of striking regularities in social statistics that first convinced the public that human conduct was subject to quantifiable scientific laws. The major methodological implication of these highly selective demonstrations was that the inherent lawfulness would become only apparent if the observation on a large number of individual cases were combined. This led to the infatuation of large samples. We see here the convergence of administrative concerns of public bureaucracies (which instigated large scale statistical 12 information for practical purposes of legislating social policy) and the theoretical claim that large numbers were also necessary to establish scientific laws of human conduct. It was this convergence that was perpetuated throughout the 20th c. in psychological statistics. The road from social statistics to psychological statistics depended on two conceptual steps: (1) Quetelet the pioneer of statistical science early in the 19th c. explained differential but stable rates of social indexes such as crime in terms of propensities attributed to the “average” individual (e.g., “propensity to crime” as a disposition of an average individual). Thus, if crime rates varied with gender, social class, education etc. this was because these social factors affected the “propensity to crime” as an attribute of the average individual. By analogy one could think of propensities to suicide, homicide, insanity, etc. What Quetelet did was to substitute a continuous magnitude (statistically gathered on large numbers of people who commit say crime in a particular population) for distinct acts of separate individuals (so we get 50 acts of murder in 100,000 people and hence the rate of murder was 50/100,000 and this ratio was then attributed to the propensity for murder of the “average individual” in that population. (2) In Germany there was widespread criticism of this inference from the sample/population ratio to the individual; in England however this methodology quickly became accepted. Buckle for example maintained that statistical regularities were clear evidence for the lawfulness of individual actions. Obviously, such a claim depended on a very specific view of the relationship between collectivities and their individual members. Not only were individual attributes freely composed into aggregates (the continuity assumption, above) but conversely group attributes were regarded as nothing more than summations of individual attributes. The latter, as we have seen, was consistent with Darwin’s conception of a biological species as an inbreeding population of individuals rather than as a “type” (which was deemed a hangover from essentialism). This assumption was one that Galton also relied on in his research on individual differences. Galton’s contribution and those of his follower Karl Pearson (of the Pearson “correlation coefficient” fame), finally made possible a psychology that could plausibly claim to be scientific while not being experimental (“correlational psychology” – see Ch. 8). A new method for justifying psychological knowledge claims had become feasible. Thus, to make useful and interesting claims about individual people it was not necessary to subject them to experimental or clinical exploration (in the Leipzig and Paris traditions). It was only necessary to compare their performance with that of others, and then to assign the individual a place in an aggregate of individual performances. Individuals were now characterized not by observing what goes on in their minds, or even inside the organism, but simply by their deviation from a statistical norm established from a population with which they had been aggregated (this was functionalism in psychology). As a result of this Galtonian turn, Psychology was left with two very different frameworks for justifying its knowledge claims: (1) based on individual attribution of psychological data (mix of Wundt and clinical models); (2) the other based on statistical 13 norms (Galton model). Both maintained that their knowledge claims rested on the demonstration of regularities that could be interpreted as having psychological significance. Of course the nature of the regularities was fundamentally different in the two cases. In Wundt’s traditional experimental paradigm, the regularities demonstrated were of an immediate causal significance. That is, the (introspective) observations by a subject in a psychophysical system changed as the (stimulus) conditions to which he was exposed changed. This was a classically experimental method (physiological, or of physics for that matter). In contrast, regularities established on the basis of testing groups and the attributed, by inference, to an average individual was a statistical system constituted by the attributes of a socially constructed collective subject (i.e., the average subjects). This method was clearly not experimental but had more in common with census taking and social and biological population studies. The discrepancy between these two alternative knowledge claims was enormous. Experimental knowledge claimed its special status on the basis of the kind of knowledge the older laboratory sciences had built their success. Its appeal was to use skilled observers, unambiguous reactions/observations, and systematically controlled conditions. None of this was true for the population studies. Different criteria for assessing the worth of knowledge claims operated there. The simply practice of using large numbers of subjects (averaged) seem to have compensated for their lack of skill and experience. An intuitive notion of individual deviations cancelling out in a large samples became in due course more precise by the use of calculus of probabilities (probability theory). Here we see that statistics in psychology is very much bound to a particular method. The earliest statistical approach to psychological knowledge in the spirit of Quetelet was still fascinated in deriving average values from group data and where these averages were then regarded as indexes of some human type (like an age type or sex type). Just as such official categories as “suicide”, “crime”, “violence”, “sickness” etc. had been uncritically accepted as social phenomena, just so conventional social categories were now transferred to psychology. The relationship between individual and aggregate was still defined or mediated by the “type” concept (“Joe belongs to type ‘male’, ‘black’, ‘over 16 years of age’ and ‘English speaking’”) which was usually deemed to be biologically/socially given, but obviously this could hardly compete with the scientific credentials of traditional experimental psychology. Most psychologists wanted to emulate the harder sciences. Proper scientific knowledge was knowledge of particular changes in the psychophysical system under systematically varied conditions. Merely statistical knowledge (e.g., Galton or “correlational psychology”) was regarded as superficial and incapable of leading to explanations of determinate processes taking place in actual psychophysical organisms. From this perspective statistical knowledge could not advance science. Thus, the subjects in these studies were often children, the most unreliable of observers and their reactions were 14 complex, variable, and unanalyzed, and the conditions under which the information was collected were uncontrolled and open to the influence of unknown factors. Indeed, this kind of study (“correlational” studies) never attained the respectability in the discipline as a whole; yet it had influential supporters like G. Stanley Hall, and J. McKeen Cattell who had both worked in Wundt’s lab and were well-acquainted with experimental research. The reason they supported statistical knowledge claims (instead of the experimental method) was of course because they were interested in alternative kinds of psychological knowledge. They did not want to limit psychology to the study of individual minds but wanted to extend it to include the distribution of psychological characteristics in populations. What the early statistical studies had going for them was that they were socially relevant and close to life. They operated both for the choice of populations (“whites”, “males”) and for the choice of the psychological characteristics (“impulsive” “homosexual”) attributed to these populations. The groups investigated were defined in social and cultural terms or in administratively relevant categories such as boys and girls of a certain age in certain schools, or Columbia University undergraduate students born in America. The kind of psychological categories were also simply drawn from daily life such as, for example, the activities of dreaming, playing, writing, recollecting, etc. It was their practical utility that compared favorably to the seemingly esoteric experimental studies (Wundt) which depended on a sophisticated intellectual community. There were several developments that prepared the way for the popularity of these statistical studies: (1) universal education which had to deal with the evaluation of large numbers of students and involved preparing standard curricula, age grading, written examinations which created the kinds of statistical populations that Galtonian psychology took as its basis. (2) There was also the increasing tendency to conceptualize social problems in terms of populations especially in the US where urbanization, industrial concentration, and immigration were problems of individuals were conceived as problems of individuals as members of aggregates. Crime, delinquency, and feeblemindedness were easily attributed to the statistical distribution of individual characteristics. This meant that structural social problems were now seen as individual problems which were to be dealt with not by social/economic change but by administrative means (selecting individuals for particular tasks). Even William James, America’s first and foremost psychologist, was not immune from this kind of research and conceptualization: “all natural sciences aim at the practical prediction and control, and in none of them is this more the case than in psychology”. Controlling states of mind is what every educator, asylum administrator, business CEO wanted advice on; they want to improve the conditions/productivity of their students, clients, inmates, etc. Psychology could contribute to the management of all kinds of social problems one that could be defined in terms of statistical variation of individual traits. In sum, on the one hand there was the Wundt legacy of experimental psychology, analogous to the hard - physiology - sciences) but which Wundt gave only a limited place in the discipline and without immediate practical value; on the other hand, especially in the context of American progressivism, industrialization, expanding frontiers, demanded 15 socially relevant knowledge, knowledge that had to be applicable to social categories and practical concerns. This kind of knowledge the Galtonian method provided even as this kind of knowledge could never meet the demands and standards of genuine scientific research (it remains correlational). By 1914 (WW I) journals of psychology were publishing group data more frequently, especially in “applied journals” while individual data studies decreased. Collective subjects In distinguishing among various kinds of collective subjects that were featured in psychological knowledge claims, the most obvious division to make is between collectivities that are the result of the investigator’s intervention (experiment, or a certain treatment, independent variable conditions) and those that exist irrespective of this intervention (just eliciting subject, like undergraduate students, for “testing”). Laboratory science generally constructs its own artificial objects which do not exist in nature while field science works with naturally occurring objects. Thus age and sex grouping exist naturally whereas experimental studies include experimental treatment groups which are constituted in the experimental situation (usually as a treatment intervention). Of course, what psychology meant by “natural” groupings were really social-political categories of importance to everyday life [for example, heterosexual and homosexual are social categories, as are high or low intelligence a social category, or even ethnicity such as Asian versus Caucasians social categories]. The natural groups of psychological research were important to social life outside the laboratory; psychologists did not create these categories, rather they simply took them over from the prevailing social order/language. A very different situation existed where psychologists constructed the collective subjects of their research in terms of their own criteria. In this case the population about whom the research process provided specific information was not related to a specific social category but simply the product of the investigator’s intervention. The earliest examples of this were when the experimental responses of a number of subjects were averaged to yield a group mean (as in a standard experiment in psychology today). This seems innocent enough even as it implied a profound conceptual change. Such a group mean is obviously not an attribute of any particular individual but an attribute of a collectivity. But what kind of collectivity is this? It is the group of individuals who happen to participate as sources of data in a particular experimental investigation. Their common activity in the experimental situation defined them as a group. Before the experiment, the group did not yet exist as a social category (e.g., undergraduate students who become subjects in an experiment where the experimenter tries to induce happiness or sadness so that they come to belong to “happy subject” or “sad subjects”). This shift is an important one. It points the way to a science that supplies its own categories for classifying people (by way of experimental intervention – some independent variable manipulation) rather than rely on existing social categories. There are three basic ways in which the psychologist’s intervention produced new groupings that do not correspond to traditional social categories. 16 1. By way of common experimental treatment in which subjects’ responses are pooled. A learning curve for example is based on the average responses of a number of subjects. The relationship between some response measure and number of trials is attributed to a group of subjects and hence is not expected to be demonstrable in one individual person (member of the group). Such a group is then an artifact of the experimental situation, and what defines the group is a particular pattern of treatment intervention. Presenting average results for the group (instead of for each individual) suggest as that the function (learning curve) is a general scientific law that holds across all individual members of the treatment group. 2. A second kind of artificial group is created when the individual subjects in the experiment are divided on the basis of their differential exposure to different experimental treatment conditions. What is of interest now is not the actual pattern of group responses as much as the difference between these patterns in the different intervention groups (group-control design studies using different treatment groups or treatment and comparison groups). Obviously different groups have a different history because they are the recipients of different interventions (and hence are socially constituted as “subjects” in the experiment). 3. There is also the psychometric group. In this case the defining characteristic of group membership is based on the performance of some psychological measuring or assessment instrument. In the first half the 20th c. this is usually a psychological test, say an intelligence test, or a personality scale. Groups are then defined by their performance on this test or scale designed by a psychologist and hence these are not natural social categories. These groups are different from intervention defined groups since the latter are “modifying” (e.g., exposure to systematically varied stimulus conditions) whereas the former are deemed to be performances based on some stable characteristic of group membership (e.g., extroversion). Given that in any investigative situation the responses of psychological subjects will depend both on what they “bring” to the situation and on what the experimental situation brings to them, all three kinds of artificial groups are artificial in two senses: (a) the groups are created in the process of investigation and do not exist outside the laboratory in the social-historical world, and (b) since these groups do not exists naturally in society, they are also in some fundamental sense artificial at the psychological level. The existence of experimental and treatment groups were defined solely in terms of certain modifying factors (tasks required of the subjects or imposed on the subjects) in the investigative situation, and hence the more or less stable characteristics of individuals were relegated to the category of “error” (error variance). That is, individual characteristics were “sacrificed” for the sake of the isolating the effects of treatment conditions by averaging over the responses to different treatments. However, psychometric groups which were defined solely in terms of supposedly stable individual characteristics (e.g., IQ or introversion, or sex/gender) ignored the situational effects (in which IQ, introversion, or sex makes a difference or is put into practice). 17 Thus, all such groups were based on an abstraction (idealization). In the experimental context, the idealization was that of a collective organism that exhibited only modifiability (as a result of exposure to stimulus treatment), whereas in the psychometric case the collective organism exhibited only stable traits (e.g., sex, introversion, selfesteem). [The term organism goes back to Fechner who used it to describe statistically constituted entities; it also has Darwinian influence and reflected the extensive animal research in the Darwinian tradition.] The distinction is important for the kinds of knowledge claims psychologists were interested in making. In order to make universal claims (i.e., laws-like claims or “generalizations”), psychologists presented their data as attributes of collectives (averages) rather than individual subjects. These collectives were constructed and assumed the existence of a collective organism that already exhibited the assumed general characteristics on which their knowledge claims depended. Thus, to demonstrate the effects of supposedly stable characteristics, they constructed experimental groups defined by such assumed characteristics (e.g., all people have some degree of introversion so one could divide people into groups of “high” and “low” introversion), and to demonstrate the effects of experimental intervention they constructed groups entirely defined by their exposure to such intervention (e.g., “treatment” versus “comparison” groups). Such a procedure was inevitable if knowledge claims were to be made through the medium of a myriad of separate small-scale empirical studies. Problems arose only when the hypothetical nature of the collective subjects was forgotten and attempts were made to transfer such knowledge to the world of real subjects (hence the whole question of “external validation”). At this time in the history of psychology such transfers to the real world were often made in terms of social ideological positions. Thus, intelligence testers maintained that IQ was radically hereditary (e.g., in service of eugenics program that many psychologists adhered to) or the behaviorists claimed that they found techniques on which the effectiveness of programs of social control depended (Skinner’s schedules of reinforcement). It was probably these ideological positions (innate versus environmental “causation”) that were more important than their experimental procedure of using either experimental or psychometric means of constituting the artificial collective subject. Thus, it has often been pointed out that they shared the same undemocratic goals (e.g., using IQ to admit immigrants, or using learning curves to allocate resources). They also shared a simplistic intellectual style that was inseparable from their real political role in mobilizing support for broad social programs (e.g., Watson’s “give me a dozen healthy infants…”, or say Skinner’s reinforcement strategies to raise infants). Identifying the subject in psychological research Making psychological knowledge claims involves an ordered set of attributions attached to a particular kind of artificially constructed subject. The significance of these attributions depends in a crucial way on the nature of the subject whose attributes these are claimed to be. If the subject is an individual consciousness we get a very different 18 kind of psychology than if the subject is a population of organisms. The predominance of the latter subject in psychology is that it results in knowledge that is applicable to populations outside the lab but then that knowledge was also as abstract as the generalizations in the natural sciences (why does it work there but not in Psychology?). Thus, it meant working with artificially collective subjects in research situations and then imposing these findings on people outside the lab. That is, psychological research on populations tended to replace the social categories that defined populations in real life with populations defined in terms of non-social “treatment” categories. It is interesting that American psychology aimed to be socially relevant but it did not aim to be a social science. Rather, its approach was that of natural science. Hence, research on socially defined populations tended to be regarded as “applied” research, whereas “basic” research worked with abstract populations (defined in terms of experimenter interventions). In the extreme case, these abstract populations need not even be human; in fact rats, pigeons, and mice fulfilled the role of a population of completely abstract organisms more adequately than human populations could hope to do (easy to breed, cheap to keep, and you don’t have to worry about “ethics”). But even where human populations were retained, the making of knowledge claims faced a problem. Obviously each empirical investigation (each experiment) is a historically unique event, involving the interaction of particular participants at a certain time and place, even as the product of that research had to be presented as valid independently of historical context. To convince that the experimental social situation was indeed ahistorically valid (unlike, obviously, other social situations) psychologists had to note some special features of the experimental situation. These features were sought (1) in the nature of the participant/subject, (2) the nature of the activity/treatment in the experiment, and the (3) nature of the circumstances in which the activity happened and the participant took part. Let’s start with the last. 1. One could control the circumstances by reducing the complexity, ambiguity, and irregularity (or “meaningfulness”) of the circumstances. Thus, one could control the stimulus conditions carefully (e.g., listening to pure tones, reducing the visual field, presenting regular stimulus series, meaningless syllable, etc.). Note that such control over the stimulus also permitted the use of recording devices (brass instruments). 2. One could also control the nature of the activity to which subjects were limited. Activity was segmented into artificial units that count be counter and aggregated, thus leading to quantitative data that permitted the precise comparison of various treatment effects. 3. Both these features aligned psychology with the natural sciences and hence gave the impression that psychology was a natural science. But of course in psychology experiments something else happens namely the data were attributed to a (human) data source. The nature of that data source had much to do with the knowledge claims made. The apparent similarity of psychology to the natural 19 sciences resulted in the total neglect of this third point, namely that the subject in the psychology experiments were historical human beings who were being investigated at a particular time and place. In view of the latter we need to distinguish between the identity of the research subjects and their identification in the research report. These two need not coincide. John’s identity as a person in his society is not his identity (even such blatantly social categories as age, profession, education, sex, etc. were often not reported) in the research report. The former (identity as research subjects) was deemed irrelevant to the experiment (John could be any person- he was merely a “subject”) and nothing was said in the research report about the identity of John or anyone else (they were simply “subjects” as distinguished from the experimenter). Thus, the “experimental identities” of the subjects were constructed precisely to construct the “experimental situation”. Social identity of research participants The social identity of “data source” participants (subjects) may be highly relevant because the goals of the investigation treat the participant (data source) as a representative of a particular socially defined group (e.g., as belonging to the group of male, grade-three children). It is this function that makes Psychology into a social science. The social identity of other participant (experimenters) making the observations, recording them, and reporting them also required an appropriate social identity (e.g., PhD, professor, researcher, investigator, etc.) one which vouchsafes the reliability of the observations. These two functions Psychology shared with the other natural sciences. There is no reason in principle why these two functions should not be carried out by the same person. There may be practical difficulties in doing so of course but, as we haves, the early psychophysicists (like Fechner and Weber) combined the two functions and in Wundt’s lab they were interchangeable. Also in Hermann Ebbinghaus’ early work on memory he functioned in both roles (see Ch. 4). In this way psychology was like the natural sciences (except for the interpretation of the data of course). In case of physics the data were interpreted in terms of the observed object, in case of psychology interpretation was in terms of the observing subject. We have seen that the departure from the Leipzig model involved the investigation of psychological processes in representatives of populations with social identities that were believed to be psychologically interesting. The clinical model of investigation was of this kind: the blind, hysterics, somnambulists, intellectually deficient, or the musically gifted functioned as subjects of investigation much like those with pathological/physiological conditions might do in medical research. The difference was that the medical profession in selecting its subjects had already imposed it own medical categories, whereas psychology had essentially adopted social-cultural 20 categories and treated them as if they were psychological or biological. In any case, in this kind of research, social identity was directly relevant to experimental identity. There were also other studies that amounted to a kind of psychological census taking and hence these were essentially descriptive in nature. Mental phenomena were not studied as processes but as isolated instances to be counted. They were not related to any kind of context and the individuals who conducted these studies acted merely as a medium through which these phenomena manifested themselves. These studies seemed to be modeled on quantitative natural history, requiring no special observational skills from the data source. Thus, virtually any unimpaired adult with a certain level of literacy could act as a subject these investigations; the only requirements were that they were available in sufficiently large numbers - and undergraduates seemed perfectly fitted for this role. Children constitute the last major category of subjects for psychological research as these were a major source of data at least for American investigators. Children shared with undergraduates the advantage that they were available in large numbers and in a well-defined institutional context which compelled their participation. School children also presented potential objects for practical psychological intervention. Children also presented special deficiencies and were therefore appropriate for clinical experimentation. In addition to members of the general public, undergraduates, and children there were animals of course (and the rat bred especially well, and was cheap to maintain). All these subjects were used as “abstract subjects” whose only identity was that they participated in the experiment. That is, American psychology was largely derived from biology not from sociology or history. So any comparison among subjects was interpreted in biological categories (like “age”) or innate endowment (like “IQ”). Gradually these biological categories were replaced by descriptive categories of human action (behaviorism- see Ch. 10, 11). But the more this became a trend the more subjects became substitutable (e.g., one could study action – “behavior” – in animals or human being, children or undergraduates, the pathological and the normal) one for the other. Individuals simply became the media through which abstract laws of behavior expressed themselves (S-R or S-O-R psychology). Because it was these abstract laws (S-R laws) and not their carriers (subjects) which were the primary object of investigation, social and biological categories became less and less relevant. The choice of subjects was simply determined by convenience – and animals and undergraduates were very convenient. The rhetoric of experimental identities Referring to subject identities in research reports depended on prevailing norms and conventions among the community of investigators. These norms in turn reflected the prevailing conception of psychological knowledge. One might think this to be a trivial matter but it is not. Thus, initially the data source was referred to by name or as co-worker etc., later as subject or observer; but it took until the 1930s until the term 21 “observer” totally disappeared (of course, since “observer” is reminiscent of selfconscious introspectionism or systematic experimental introspectionism). Thus, it was the emergence of behaviorism in American psychology that set the terms “observer” and “subject” in conflict - because clearly in the behaviorist focus, the subject was never engaged in observation (in behaviorism it was the experimenter who did the observing). This conflict was one that set the “cult of objectivism” over and against those studies in which the “organism” appears as “agent” (free will etc.). The fact that there was such conflict over several decades reflected a continuing controversy over the essential function of the human data source. There appears to have been a general feeling that the data source ought to be referred to by what it actually does in the experimental situation. What was the function of the data source? The objectivists felt that it consisted of the providing material for the experimenter to observe and use and the term “subject” had long ago acquired this meaning (although the medical use of “subject” was forgotten it remained a practice). A minority who still held to the idea that psychology was about consciousness resisted and felt that the function of the data source was to competently comprehend the experimental task and to act appropriately. The latter of course were closer to Wundt original project which conceived of observer as the original data source in experiments. But both “subject” and “observer” strictly refer to an intra-experimental identity. That is, participants only have these identities insofar as they play a role in the social situation of the experiment. Of course, as I pointed out, these participants do not lose their social identity outside the lab. When their extra-experimental identity was at stake, participants were usually referred to as children or pupils, or boys or girls, occasionally men, women, and students. Professional journals differed as to the use of the intra-experimental or extra-experimental terms. Journals that reported applied research, research relevant to the world used extra-experimental identities whereas pure research journals used the term subject. Moreover, undergraduates were more likely to be referred to as subjects whereas children and non-academic adults were referred to by their extra-experimental identity. One reason was that the conferral of particular identities in experimental situations is that such identity is relevant to the experimental situation. For example the identity of children was so identified in the experimental situation because being a child was relevant to the experiment; in contrast, undergraduates were not identified because their identity was not deemed to provide specific information about college students (but rather about the general public). When educational status or age were provided these variables were deemed essential to the purpose of the research. When children were not referred to as children but as subjects this was a way to emphasize the pretension that the psychological research was about abstract relationships rather than people. Experimental psychologists were concerned to establish a kind of knowledge about human beings that would be a-historical and universal. Yet in order to obtain this knowledge they had to work with specific, historically and socially defined human data sources, and they had to extract their data from these sources in investigative 22 situations that too were historically specific. To cope with this paradox, psychologists commonly used rhetorical devices when reporting their data. Thus, the experimental results were attributed to subjects whose identity was entirely established within the experimental situation (as e.g., belonging to a specific treatment group). In this way the experiment stands apart from the historical context; it is an abstraction from life leading directly to universal – most frequently biological ones. There are other effects due to this practice of not attributing the results to historical individuals but merely to individuals who have an experimental identity (i.e., “subjects”). Thus, sometime the results are simply attributed to the “treatment” whose anonymous members have no individual existence apart from the experimental identity. This served to further isolate the lab from the personal and social contexts. Of course, just because scientific reports employed rhetorical devices does not mean that the knowledge claims are necessarily false. It only means that they have not been empirically secured – and that the issue remains wide open. Of course, Psychology as a scientific discipline, or at least one with scientific aspirations to autonomy, was intent on producing universal knowledge claims – that is, claims beyond individuality, history, and local meanings – but these claims were based on the appropriate manipulation of the identity of the sources of data. With a few exceptions claims to universality were not empirically grounded but established by fiat. The role of the experimental report was to create the illusion of empiricism. Thus, by the way the results were presented, psychologists created the illusion that the results were not really the product of social interaction among certain human personalities in historical time, but that they were the manifestation of abstract transpersonal and transhistorical processes. Increasingly psychological reports dealt with “abstractions” and became less and less accounts of particular investigations. See Ch. 13