51. True and pre-experimental designs (according to Campbell) of research in experimental psychology

There are quite a few criteria by which experimental designs can be classified. Campbell's criterion (regarding formal planning). This is the criterion of the true experiment.

Pre-experimental designs:

1. Analysis of a single case. Understandings of the single-case analysis:

1. Campbell: the «X only» or XO design. There is no control group, and the dependent variable is measured after the organization of the experimental influence.

X may be absent altogether.

- The task of examination (neuro- and pathopsychological diagnostics).

Here there is no organization of the levels of a particular independent variable, and the use of psychological techniques includes both the actualization of the processes under study and the psychodiagnostic orientation of the subsequent conclusions.

The description and explanation of the psychological properties of a specific subject are built by testing a multitude of hypotheses, the choice of which is directed by the situation within the framework of the task of describing an individual case. The single-case analysis is based on a theory intended, with respect to the methodological instrument used, to set up a system of reference points for discussing the properties and processes being actualized. Problems:

• There must be normative indicators (a «norm»). Being experimental results for previously conducted experiments, the same experimental dependencies may be regarded as criteria for distinguishing the subject being examined. Psychological hypotheses will not necessarily act as cause-and-effect ones.

• There are no prior restrictions in the field of psychological hypotheses. In an examination situation, one diagnostic hypothesis may replace another, one identified property may suggest a hypothesis about another, and so on. If it is not possible in advance to indicate the general propositions important for understanding the given case, then the path of advancing hypotheses will be inductive.

Functional tests are used as techniques, behind which the orientation of the hypotheses associated with them is, as it were, discernible, but their choice is not determined.

• The main regulators of the study are the psychologist's experience and intuition. One must also know the participant's life context, the goals, tasks, and occasions of the examination, and so on.

2. Between-group pre-experimental schemes:

An analysis of a single case, where such a «case» is represented by a group of people.

For example, pedagogical experiments with a single X-influence and a single group.

Problems: - possibly, the chosen group initially contained stronger students;

- possibly, the pupils' awareness of the fact of the experiment made them work more diligently;

- possibly, the result is connected with the personality of the teacher.

2. Preliminary and final testing on a single group (O’XO”). Compared with the first, this design is better, since it takes into account the magnitude of the change in the dependent variable from the first to the second measurement (that is, there is control of the dependent variable at the «before» and «after» levels of the influence). The reasons for the low reliability of the conclusions:

1. The «background» factor. First a primary measurement was taken (O’), but the measurement O” took place in a changed situation!

2. The «natural development» factor. Here it is the same as above, only the situation – the participants' age changed, and so on.

3. The «testing effect». The very fact of testing produces a training effect.

4. The «reactivity» factor of the participants. The measurement procedure itself may act as a stimulus for changing the behavior, assessments, or thoughts of the participants (+ the «Rosenthal effect»).

Conclusions on pre-experimental designs: in these designs there is no control over the levels of the independent variable, as a result of which one cannot only fail to draw a conclusion about the action of the independent variable, but also cannot reject the multitude of other explanations of the changes in the dependent variable – by «third» variables (that is, the indicated extraneous factors). These competing hypotheses cannot be rejected because of the absence of data on the dependent variable under analogous conditions for control groups (where there is no active level of the independent variable).

True experimental designs.

1.. Intraindividual schemes.

Ensuring internal validity is the main goal of planning a psychological experiment when developing intraindividual schemes for comparing the dependent variable. The factors of time and sequence are the main threats of confounding. The task factor is the third main source of threats to internal validity in intraindividual schemes.

The main intraindividual schemes are aimed at solving the problems of controlling both systematic and nonsystematic confounds. Let us enumerate the main ones:

-random sequence (the use of a table of random numbers)

-A quasi-random sequence includes an additional condition: the total number of trials is divided into equal micro-sequences.

-The scheme of regular alternation looks like the repetition of the same conditions in an unchanging order: ABABAB.

-The scheme of positional balancing is the fourth of the most frequently used.

2. Between-group schemes.

Experiments in which not individual participants but groups of participants take part. Advantages:

1. The conclusions can be extended to a broader population.

2. It is more economical than the individual one.

3. The possibility of applying more effective schemes of between-group comparisons.

Strategies for composing groups (comparison schemes) The use of real groups

Already existing groups are used. Many disadvantages (the groups are different).

3. Cross-individual schemes. (!)

- these are experimental designs presupposing the assignment of different sequences, including all the levels of the experimental factor, to different participants or different groups of participants.

Designs:

1) Reversive balancing.

To two groups of participants the conditions are presented in forward and reverse order (ABC – to the first group, BCA – to the second). The averaging of the dependent variable indicators here occurs for the total number of trials separately for each level of the independent variable. Cross-individual schemes are often used for multilevel experiments.

2) Complete balancing (control of sequence effects):

- the Latin square

- the balanced Latin square

Designs for presenting all the levels of the experimental factor both to the same participant and – in a single presentation of each micro-sequence – to different participants. The balanced square differs in that in it every level of the independent variable is once immediately preceded by every other level. But there are effects of centration and of the edge.

For cross-individual schemes, the selection of participants into groups is the same as in the case of between-group schemes. The advantage of cross-individual schemes lies in the construction, on their basis, of functional dependencies.