Psychological experiment

The psychological experiment is one of the main research methods in psychology.

Compared to the model of the scientific experiment , the essential difference is that there is no "object", but a voluntarily participating and self-confident person, an experiencing subject, in the role of a test subject based on psychological instruction, certain tasks under the artificial conditions of a laboratory or in another standardized situation.

The possibilities and limits of the experiment compared to other methods and directions in psychology have often been discussed and remain controversial (for history and controversies see Experimental Psychology ).

Design of experiments

Hypothesis formation

Each experiment should take place in a theoretical reference system , also taking into account previous experiments, with knowledge of their results and problems, and have a specific question.

Every psychological experiment is preceded by the precise elaboration of the hypotheses and the alternative hypotheses , whereby the essential terms and methods are defined and explicated (see operational definition , explication ).

The exact formulation of the hypotheses also includes the statistical formulation of the hypotheses and the specification of the intended statistical test procedures. The experiment should enable the empirical decision between the hypotheses. Therefore, subsequent changes to the test plan with regard to the question, hypothesis formation, test execution, data processing and statistical testing run counter to the purpose of the experiment.

Experimental and control group

As a rule, the test subjects (subjects) are divided into experimental and control groups. In a “real” experiment, this classification is random; in a “ quasi experiment ”, different groups already exist (e.g. men vs. women; rural vs. urban residents). In the experimental group (EG) the independent variable (UV) is manipulated (the UV is the variable whose influence is to be investigated; the expression of the UV is called "treatment"). This does not happen in the control group (KG), or another treatment is used whose difference to the EG treatment is to be investigated. After the treatment, the dependent variable (AV, the variable that is supposed to change through the manipulation of the UV) is measured in the experimental and control group. The severity of the AV in the control group is called baseline. The difference between baseline and the measurement in the experimental group is attributed to the effect of the UV.

It is important to be able to exclude alternative explanations by keeping all variables constant except for the UV (see internal validity ). It is therefore crucial that there are no differences between the control and experimental groups from the outset. The results of an experiment can be falsified by "non-equivalent" EG and KG (see control of confounding factors ). Therefore, the random assignment of subjects to the test conditions is of great importance.

In the methodology of psychological experiments, a large number of simple and complicated research designs have been developed, which are often closely linked to statistical evaluation strategies or modeling .

Further principles of psychological-experimental work

Two of the defining characteristics of an experiment ( experimental psychology ), which Wilhelm Wundts frequently cited, relate primarily to psychophysics and are only valid to a limited extent in other areas. The point of view that the investigated process can be triggered according to plan (not by the subject, but by the experimenter) generally applies, but there are cases in which the occurrence of the process of interest must be awaited.

The requirement of a state of "tense attention" of self-observation also comes from psychophysics and is also generally important if precise observations are expected from the test subject or the test director.

An investigation must meet further conditions in order to meet the scientific requirements of the experiment:

• Informed consent of the test subject: Existence of a (written) declaration with the consent of the test subject and reference to data protection.
• Standardization of the implementation and evaluation of the experiment.
• Logging: Precise recording of the entire test planning: Hypothesis formation, selection and grouping of test subjects, entire test setup, planned test execution, strategies for controlling (limiting) investigator effects and method-related reactivity of test subjects ( reactivity (social sciences) ), recording and evaluating the data including the intended statistical procedures as well as logging of the actual test sequence, any special features and malfunctions.
• Repeatability (replication, reproducibility ): An experiment must be described so precisely that it can be repeated in the same laboratory or in a second laboratory by another qualified investigator if the defined test conditions are observed. However, this standard can hardly be achieved in short magazine publications.
• Decision about the hypotheses and interpretation: The result is to be interpreted in connection with previous research on this question and discussed with regard to any method problems that may exist and with regard to a constructive continuation.

Types of experiments

• Laboratory and field experiments: Laboratory experiments enable extensive control of the examination situation and certain disruptive factors , but, like field experiments , are subject to the influences of the reactivity of the subjects in the laboratory situation. The results of field experiments are more strongly affected by disruptive factors, but because of their proximity to everyday life, they can have a higher ecological or external validity.

• Real experiments and quasi-experiments : Real experiments have all of the above properties. In particular, they are characterized by a random (randomized) distribution of the test subjects between the experimental and control groups and the manipulation of the independent variables. In quasi-experiments, already existing characteristics of the test subjects (e.g. the severity of an existing disorder) or their self-selection (e.g. independent decision about which of several treatment conditions a patient accepts) determine their group membership. Causal statements in quasi-experiments are therefore difficult. The test plan of real experiments is called experimental design, and the test plan of quasi-experiments is called quasi-experimental design (see research design ).

Internal validity exists if the change in the dependent variable can be clearly traced back to the variation in the independent variable (no alternative explanation). External validity exists if the result in the sample can be generalized (generalized) to other people, situations and points in time.

Problems of the Psychological Experiment

Subject motivation

It is still largely unclear how the subjects' motives affect the test results. As a rule, volunteers take part in psychological experiments. B. Because they are interested in science or want the money that is often given as a reward for participating. In addition, psychology students are often required to serve a specified number of hours as test subjects. According to a study by Sears, 74% of the subjects in the socio-psychological experiments published in 1985 were students. This selection of the subjects ( convenience sample ) thus differs systematically from the general population and is not even representative of the subpopulation of students due to lack of random selection. The generalizability or external validity of the results is very questionable.

Experimenter effects

"Experimental effects", the effectiveness of which has been proven in numerous experiments, are relatively well researched. A distinction must be made between experimenter effects, which arise due to certain errors in the conduct of the experiment and evaluation, which are not random but in favor of the hypothesis, and experimenter expectation effects in which the examiner's expectations (motivation, attitude) and his own hypotheses are more or less unnoticed influence the results (see Rosenthal effect , Greenspoon effect ).

Investigator errors can be reduced by a computer-aided implementation and automatic registration, expectation effects by a double-blind experiment . The double-blind test is a research design in which neither the subjects nor the investigator know at the time of data collection whether the subject belongs to the experimental group or the control group (e.g. drug and ineffective preparation, placebo ). So that the experimenter's expectations do not influence the behavior of the test subject, the roles of the researching experimenter and the experimenter are separated.

Techniques for controlling disruptive factors

The most important technique for controlling confounders is randomization . H. Test subjects are randomly divided into experimental and control groups. This randomization is an important hallmark of a good experiment. This is to ensure that possible differences that influence the results (and possibly still unknown) that the subjects bring into the experiment are represented equally in all test conditions.

The greater the number of subjects, the greater the chance that the randomization will serve its purpose. In small number of subjects and to existing assumptions about important differences, the alternative procedure is the parallelization of the study groups and the matching . Of corresponding subjects Subjects are divided from the outset so that the feature in question in all groups equally present (Example: Equal gender distribution in all test conditions). If there are a sufficient number of subjects, several personal characteristics can be taken into account in this “statistical twin formation”.

The methodology of experimental psychology contains numerous rules and instructions on how possible disruptive factors can be taken into account, for example through systematic permutation of conditions. In a test plan with daily repetition of the data collection, the experimental and control conditions should be planned in a mixed sequence, i.e. not uniformly on day 1 condition A and on day 2 condition B, but also in reverse order, since otherwise there would be a confusion between treatments and situational features as well as the effects of repetition per se.

Philosophical Aspects and Criticism

Special position of the psychological experiment

Investigations with and on people have - apart from the professional ethical side - a special position in terms of scientific theory. With a view to Wundt's Leipzig laboratory, Kurt Danziger (1990) analyzed the systematic introduction of the experiment and its socio-constructive character from a psychological-historical point of view. Here the roles are divided between the organizing “test leader” and the “test person” providing the data, while in a scientific experiment the role of the observer is mostly limited to reading measured values ​​and registering data.

Experimenter and subject form a new reference system due to these dependencies scientific theoretical and practical consequences (see the definition of the general prompt character and the special demand characteristics and aspects of reactivity (Social Sciences) ).

Context dependency and generalizability

The aspects of internal and external validity must be supplemented by a more detailed discussion of context dependency and generalizability .

The question of the context specificity of explanatory hypotheses already belongs in the initial explication of the theory and in the corresponding decisions on methodology, i.e. H. the operational definition of the theoretical constructs (Gadenne, 1976, Westermann, 2000). Which types of events, with regard to which application situations and which population of people should be represented?

Even Carl Friedrich Graumann argues that in principle there is a context-dependence of the experimental results. Context here means the entirety of the relevant conditions that have an effect in addition to the experimentally controlled independent variable: the constitution of the test participants, the experienced situation of the investigation and the objective test setup ( setting ) and important general and social framework conditions, the method-related reactivity of the test person, the interaction of the test leader and subject and other contextual variables.

The question of generalizability arises when research results from the laboratory are to be transferred to the practical fields of psychology. To what extent can experimental results be generalized: to similar test conditions in the same laboratory, to other laboratories (cross-laboratory), to everyday conditions ( ecological validity )?

The question of empirical generalizability has gained more importance since the terms causal explanation and law have been replaced by more cautious formulations of statistical expectations in the more recent epistemological discussion of psychology, since the ceteris paribus condition associated with the formulation of laws usually cannot be affirmed in psychology (Westermann 2000).

In the generalizability theory by Lee J. Cronbach et al. a. (1972) a universe of observations is assumed that is defined by many facets, including items of the task, variants of the task (test forms), observers, experiments, examination days, framework conditions.

So-called G studies (generalizability) provide estimates for the proportions of variance in the observation conditions considered important in a measurement procedure. The examiner makes a systematic selection of conditions according to his theoretical and practical interests and has to weigh the associated risks in D studies (decision). Such generalizability coefficients give more differentiated information than the usual simple coefficients of reliability . Werner W. Wittmann expanded the concepts of Lee J. Cronbach and Raymond B. Cattell into a multivariate theory of reliability.

Individual differences

The true (reliable) variance , i.e. H. the difference in experimental test results, which cannot be attributed to errors and certain confounding factors, often contains a proportion of inter-individual variance that is greater than the variance that can be explained by the effects actually examined. If the most important characteristics (parameters) of the individuals and the examination conditions (setting) are not only seen as a disruptive factor and "noise", but are considered in terms of research strategy, for example as a covariate, the possibilities of recognizing regularities can be improved, realistic models and useful models Develop predictions.

Subjective proportions for independent and dependent variables

The experimenter depends on the test subjects correctly understanding and following the introductory psychological instructions for the experiment.

If, for example, emotionally meaningful images or words are used in an experiment on the neuropsychology of emotions , the investigator generally assumes that this material is qualitatively experienced by all subjects in the way that he planned it.

In principle, the subjects should react to the stimuli (words, images, tasks) as the experimenter expects, they should neither develop disruptive test subject behavior nor their own distracting hypotheses, i.e. avoid much of what characterizes them as conscious people ; in principle they should behave like one interchangeable person among others.

The experimental effects in the dependent variable are recorded in many psychological experiments by collecting self-assessments using standardized scales ( scale , psychometry ) and questionnaires. Because of the semantically fuzzy, often ambiguous gradations and for psychometric reasons because of the lack of an interval scale (there is no intersubjectively valid, numerical reference system), such self- assessments must be fundamentally differentiated from the objectively observable behavioral reactions and physiological measured values.

Systematic replications?

The reproducibility of the research results by other researchers is considered fundamental for an empirical science, especially in the natural sciences .

The simpler meta-analyzes cannot replace empirical replication attempts. In the literature banks of psychology there are individual publications on successful or unsuccessful attempts at replication, but it is unclear which of the experiments in the list of classical experiments in psychology can be considered replicated.

A systematic replication of selected experimental work was only started in 2012 by Brian Nosek and colleagues in the international Reproducibility Project (see Reproducibility (Psychology) ).

literature

• Jürgen Bortz, Nicola Döring : Research methods and evaluation for human and social scientists 4th edition. Springer, Heidelberg 2006 ISBN 978-3540333050 .
• Lee J. Cronbach, GC Gleser, H. Nanda, N. Rajaratnam: The dependability of behavioral measurements: Theory of generalizability for scores and profiles. Wiley, New York, 1972.
• Kurt Danziger: Constructing the subject. Historical origins of psychological research . Cambridge University Press, Cambridge 1990, ISBN 0-521-36358-6 .
• Volker Gadenne: The Validity of Psychological Investigations. Kohlhammer, Stuttgart 1976. ISBN 3-17-002801-4 .
• Carl Friedrich Graumann: Context as a problem in psychology. In: Zeitschrift für Psychologie , Volume 208, 2000, 55-71. doi : 10.1026 // 0044-3409.208.12.55 .
• Oswald Huber: The Psychological Experiment. An introduction. With fifty-three cartoons by the author. 4th edition, Huber, Bern 2005, ISBN 3-456-84201-5 .
• Siegbert Reiss, Viktor Sarris: Experimental psychology: from theory to practice. Pearson, Munich 2012, ISBN 978-3-86894-147-0
• Manfred Sader : role play as a research method . Westdeutscher Verlag, Opladen 1986, ISBN 978-3-531-11786-7 .
• Viktor Sarris: Methodological foundations of experimental psychology. Volume 1: Knowledge acquisition and methodology. Reinhardt, Munich 1990, ISBN 3-497-01111-8 . Volume 2: Design of experiments and stages of the psychological experiment. Reinhardt, Munich 1992, ISBN 3-497-01112-6 .
• Rainer Westermann: Theory of Science and Experimental Methodology. A textbook on psychological methodology. Hogrefe, Göttingen 2000. ISBN 3-8017-1090-4 .
• Werner W. Wittmann: Multivariate reliability theory. Principles of symmetry and successful validation strategies. In: John R. Nesselroade & Raymond B. Cattell (Eds.), Handbook of multivariate experimental psychology. Plenum Press, New York, 1988, pp. ISBN 0-306-42526-2 , pp. 505-560.

Individual evidence

1. ↑ Robert Rosenthal , RL Rosnow (1975): The Volunteer Subject. Wiley, New York
2. ^ DO Sears (1986). College sophomores in the laboratory: Influences of a narrow data base on social psychology's view of human nature . Journal of Personality and Social Psychology, 51, pp. 515-530