Intrasubject experiments, Features and principles...

Intrasubject experiments

Features and Planning Principles

By intrasubjects we mean experiments in which each subject participates in measurements at each level of an independent variable.

Often the same experiment can be conducted both as an intra-entity and as an inter-group experiment. The choice of one of these options is connected with the decision of the question of how such a threat of the internal validity of the experiment will be controlled, as the individual features of the subjects participating in different experimental conditions. The use of the intra-entity plan automatically removes this problem, which means it reduces the probability of obtaining unreliable results.

I Assume, the teacher of literature arranging a contest of readers wants to give students advice on how to learn poems for the contest: memorize each quatrain separately or divide the poem into two equal parts and learn by larger parts. To address this issue, he turns to a psychologist. Thus, you faced the task of assessing the effectiveness of two different ways of memorizing poems. Since the length of the poems can influence how best to memorize them, let's say that we are only interested in products of medium length, consisting of 16 quatrains.

You can take 20 students, the ten of them are asked to memorize a poem by a quatrain (ie, read the quatrain until it can be played back but memory, and only then move on to the next), and the remaining ten pupils offer divide the poem into two equal parts and learn in the same way larger parts. We call conventionally the first method of memorizing by the method of short fragments, and the second - by the method of long passages. Suppose it turned out that a group of students who memorized the poem by the first method spent on average five minutes less time than a group of students who memorized the poem in larger parts. How can you explain the result?

First, you can conclude that the influence of the independent variable (the method of memorization) and the method of short passages is a more effective strategy for memorizing poems. Thus, it turns out that the teacher of literature can offer to recommend the contestants to use this method.

Another explanation may be the error of the first kind. those. random detection of differences in the experiment, which in reality is not.

And the third possible explanation may be the effect of individual differences. Since different levels of the independent variable (different methods of memorizing) were presented to different subjects, you can not be sure that the selected groups of subjects are equivalent in everything except the memorization method used. It is possible that just in the group of students who applied the method of short passages, there were students with more developed memory. And that's why, and not because the method used by them is better, the participants of this group on average remembered the poem faster.

If you used the same subjects to memorize poems in both ways, the problem of group equivalence would be removed by itself, since both levels of the independent variable would be presented to the same subject. And if this particular group of subjects were characterized by less developed memory, then the influence of this factor would equally extend to both methods of memorization and, thus, would be controlled.

Consider another example. Let's say you want to study the influence of swimming before going to sleep on the quality of sleep at night. Suppose that you have the opportunity to measure the quality of sleep (the number of spills, the duration of the deep sleep phase, etc.) with a portable device that you can give to the subjects at home and which automatically gives a rating of the quality of sleep on a scale of 0 to 100. This indicator will be the operationalization of the dependent variable. You can compare the two conditions between the two: experimental - when the subject for 2-3 hours goes to the pool and swims there for 40 minutes, and control when the subject is not engaged in swimming before going to bed. This will be different levels of the independent variable.

You use two groups of subjects: representatives of one group swim before going to bed, and the second group does not. If it is found that the subjects of the first group sleep better at night, this can be explained not only by the positive effect of swimming on sleep, but also by the fact that, for example, the generally more peaceful subjects, who usually do not experience sleep problems, those. groups of subjects may not be equivalent.

If you ask the same subjects to swim, then do not swim before going to bed, this problem will be removed. It will not matter to you how well this person sleeps outside the experiment, since it is he who will participate in the control. n in the experimental condition. His individual characteristics will be equally apparent in both cases, and therefore will be monitored.

So, the main advantage of intrasubject experimental plans in comparison with the intergroup is the removal of the problem of individual differences and the equivalence of groups of subjects. All subjective variables (that is, all individual characteristics characteristic of the subjects) that affect one level of the independent variable will equally affect all other levels.

The absence of individual differences between subjects participating in different experimental conditions leads to the fact that intrasubject plans have purely statistical advantages when performing mathematical processing of the results obtained. Even small differences in the arithmetic mean of the measured values ​​are more likely to be reliable, whereas when using an intergroup experiment to recognize the differences, the difference in the mean should be greater.

Intrasubject plans also have advantages from the point of view of organizing and conducting the research directly. The main of them is a smaller number of subjects. So, for example, if you want to examine 40 subjects in each condition, then when conducting an intra-subject experiment, you need only 40 subjects, and when carrying out an intergranial experiment - at least 80 (since the independent variable in the experiment must take at least two values). And the more conditions you want to explore, the more groups of subjects you will need.

Attraction of a smaller number of subjects naturally leads to the fact that less time will be spent not only on their search, but also on their instruction and training. This can greatly simplify your work as an experimenter, especially when it comes to the fact that for the correct performance of experimental tasks the subject must undergo a long phase of training.

There are cases when intrasubject plans may be more preferable than inter-group plans. For example, when participation in one of the levels of an independent variable is very brief (as occurs when measuring the reaction time). Participation in the experiment within just a few seconds can be uncomfortable for the subject himself.

In addition, intrasubject plans can also give an advantage when an experiment is conducted to study the characteristics of small populations - say, narrow specialists in some area of ​​knowledge, cosmonauts with long experience in orbit or the like. Then the range of potential subjects may be too small for the researcher to be able to implement the intergroup experiment.

However, it should be understood that in some cases the intra-entity plan may not be suitable for conducting a study, regardless of the expected duration of the test subject's work and the size of the studied population. Decisive in the choice of intrasubject or intergroup plan are not these formal characteristics, but the substantive aspects of experimentation.

Intrasubject plans are inapplicable in studies where, after participating in one of the conditions, the subject receives knowledge that does not allow him to participate in other conditions as if from scratch.

For example, after completing the first task, the subject can guess about a possible hypothesis of the experiment or form an idea of ​​what behavior the experimenter expects from him. Such limitations are often characteristic of research in social psychology, psychology of learning, etc.

So, if you study the productivity of two different ways of teaching an object, you can not apply both methods to the same students, because after using the first one they will already master the material.

Or, say, you want to study what information - arriving at the beginning (the first impression) or at the end - most strongly influences the formation of the impression of an unfamiliar person. You can compose a text describing a person, one part of which will describe its positive characteristics, and the second - negative. Then it is necessary to ask the subjects to characterize the described person after reading the text. In one version of the text, positive characteristics will be described first, and then negative, in the other variant, on the contrary. It is absolutely obvious that you can not present both of these texts (both levels of an independent variable) to the same subject, since after reading the first text the impression of the described person will already be formed.

In addition, intrasubjects can not be quasi-experiments, where independent subjective variables are used, i.e. some individual characteristics of people (for example, gender, personal characteristics or the like).

As for the less obvious cases of the influence of participation in one condition on the behavior of the subject under other conditions that can make the collected data invalid, then we will look at them in this chapter a little further.

If we try to characterize the key shortcomings of intrasubject experimental plans, leaving behind minor discussions such insignificant shortcomings as increasing the time of participation in the experiment of each subject, the main problem of intrasubject plans is related to the fact that the subject takes part in the study at all levels of the independent variable and with the need to do it consistently. After he took part in one of the levels of the independent variable, the experience he gained or the change in external circumstances can somehow affect how he manifests himself, performing the subsequent tasks. Thus, the way a subject manifests himself at a given level of an independent variable depends not only on the specifics of this level, but also on what the sequence of presentation of the levels of the independent variable was. Such influences in general form are called sequence effects, or order effects.

By the general name of the effects of the sequence, we mean any influences that are the consequence of where in a particular sequence of the presentation of the levels of the independent variable is a particular level. This includes both general (time, training, fatigue) and specific factors.

Sequence effects pose a serious threat to the internal validity of the intra-entity experiment. In the absence of control or insufficient control over these effects, the differences obtained in the experiment can not be uniquely explained by the influence of the independent variable, but may be a simple consequence of the uneven distribution according to experimental conditions of the factors underlying the emergence of sequence effects.

There are two main types of sequence effects: progression effects and transport effects.

By the effects of progression, such nonspecific influences are meant that can grow or decrease over time. These influences are not directly relevant to the nature of the study and are not related to specific embodiments of the levels of the independent variable. Regardless of what is being studied and by what means independent and dependent variables are operationalized, such influences will still manifest themselves. So, in the course of the study, the subject may become more and more tired, his concentration may be falling or he may become more and more boring.

These influences will lead to the fact that the farther from the first in the sequence the given level of the independent variable is presented, the worse the tasks forming it will be performed. Either, on the contrary, the effect of progression effects may turn out to be positive (for example, as time goes on, learning takes place, or the subject becomes more and more involved in the task, and becomes more and more interesting). In such cases, later tasks will win.

The effects of progression are named because it is assumed that the force of their influence from task to task increases (or decreases) progressively, i.e. linearly. And for each subsequent test (no matter what the sample is), the effect of these effects increases by the same number of conventional units.

Consider, as an example, the influence of the fatigue factor. When the first task is presented to the subject, the fatigue factor from performing the tasks has no influence, the magnitude of his influence is 0. When the second task is presented, the subject is already slightly tired after the first one, and suppose his fatigue increased by 1 conventional unit, i.е. on the second sample, the effect of fatigue will be 1. To the third sample fatigue increased by another unit, and the effect of this factor is already equal to 2. On the fourth test, the fatigue again increased by one, and here the effect of this factor is already 3 and so on .

The general idea of ​​the linearity of the increase in the effect is that the step with which the influence of the side variables on the subject's work increases from sample to sample, always remains the same. In the example above, it is equal to one conditional unit, which is reflected in Fig. 12.1.

Regardless of what the task of the subject is and what indicators are measured, we assume that the influence of the factors underlying the progression effects will increase linearly, in proportion to the ordinal number of the sample. And we can

predict how this impact will be relative to the previous and subsequent samples.

Fig. 12.1. Graphic representation of the linear increase in fatigue effect

To control the effect of such effects, various methods of positional adjustment are used, i.e. Such sequences of presentation of tasks are developed so that the effect of progressively increasing side variables is the same under different experimental conditions.

So if in the study of the effectiveness of methods for memorizing poems we planned to give the subjects two poems for memorizing by the methods of short and long passages (each on a single day), then the following experiment scheme will be the most optimal for balancing the linearly increasing effects of the sequence: on the first day the poem is learned by the method of short passages, on the second and third days - by the method of long passages, and on the fourth day again by the method of short passages.

In this case, the effect of fatigue on participation in the experiment on the first day will be 0, on the second day it will increase by one conventional unit and will be equal to 1, on the third day it will again increase by one unit and will be equal to 2, etc. Knowing this, we can calculate the strength of the influence of the fatigue factor on each of the two levels of the independent variable.

On the first and fourth days, one level of the independent variable was used (learning by the method of short passages), for them the average fatigue factor force will be (0 + 3)/2, i.e. 1.5 conventional units. On the second and third day, another level of the independent variable (learning by the method of long passages) was used, and the average amount of influence of fatigue on this level of the independent variable will be (1 + 2)/2, i.e. the same 1.5 conventional units. Thus, the overall impact of this secondary variable will be the same for different levels of the independent variable, which means that if we get differences in the experiment, they can no longer be explained by the effects of the sequence effects, since they were equalized.

If we asked the subjects for the first two days to apply the method of short passages, and on the third and fourth day - the method of long passages, then the effect of fatigue would not be able to equalize, since on the average on the first method the effect would be equal to (0 + 1)/2 = 0.5, and the second method - (2 + 3)/2 = 2.5. In this case, having received differences in the time of memorizing the poems in two ways, we could not unequivocally say what this is related to - the real effectiveness of this or that method or with the influence of the subject's fatigue. If, according to the results of the research, it would have turned out that the time required for memorizing the poem by the method of long fragments is greater, this could simply be explained by the fact that by the time the method was used for memorization, the subjects were already tired from the experiment, which led to decrease their effectiveness.

When planning and analyzing intrasubject experiments it is necessary to remember that the linearity of the progression effects is only an assumption, i.e. Supposition. And in some cases this assumption is erroneous. Usually, the actual dependence of the magnitude of the effect of a sequence on time is unknown, and in some cases there are arguments that such an effect may not be linear.

When it comes to hack factor, as learning, it is known that it grows nonlinearly. First, progress is stronger, and then slows down, i.e. affects every successive sample to varying degrees. This change is shown in Fig. 12.2. Here, at the first trials of the experiment, the effect of learning will be significant, and in subsequent samples it will be less and less noticeable. The step with which the influence of this factor is increasing will change with time.

Fig. 12.2. Graphic image of non-linear build-up of training effect

Nonlinear effects can not be easily controlled by positional adjustment. With nonlinear effects like it is represented -

You can try to fight in the previous figure by increasing the number of samples for each level of the independent variable and presenting them in a random order. However, in general, when suspicion of the appearance of nonlinear effects is desirable, it is desirable to change the scheme of the experiment from intrasubject to intergroup. At the very least, one must always remember the possibility of such influences, conduct a thorough analysis of the possible effects of the sequence even at the stage of experiment planning, and proceed from this choice in favor of an intra-subject or intersubject scheme.

Another kind of sequence effects - transfer effects. Here we are talking about the fact that participation of a subject in an experiment in one of the levels of an independent variable can somehow (positively or negatively) affect his work within another level of the independent variable. And this happens regardless of external to the content of the experiment and nonspecific effects of progression (such as fatigue or exercise). Here the specificity of the tasks is important.

As an everyday example of the effect of transfer, you can bring the teacher's check of school essays. If at first the teacher read a very good composition, and then a bad one, then the bad one might seem worse to him than if he had not read a good composition before. Conversely, if a bad essay were first read, the next good can be judged as better than it really is.

With respect to the experiment, the execution of tasks at the same level of the independent variable can lead to the fact that the subject develops some strategy for working with the stimulus material, which may be unsuitable or ineffective for working with tasks of another experimental condition, the second condition is worse. At the same time, if this second condition were presented first in the sequence, the results on it would be different.

Thus, the effects of transfer are that performing tasks on one level of an independent variable alone can lead to a deterioration or improvement in the performance of tasks at another level. In this case, one sequence of presentation of tasks can lead to obtaining results that would not have been obtained with another sequence of their presentation. The specificity of each experimental condition is important.

If we return to a fictitious example with the study of the effectiveness of methods for memorizing poems, then, for example, it may turn out that after the subject has memorized poems by the method of short passages, it is difficult for him to restructure himself on perception and memorization of longer parts, which leads to an increase in time , necessary for learning by the second method. The subject, using the method of short passages, could acquire the skill for memorizing precisely short passages, and this skill hinders the learning of longer parts. At the same time, if the subject immediately began to memorize by the method of long passages, he would not have experienced such difficulties and coped with the task more quickly.

The transfer using the intra-entity plan may be symmetrical or asymmetric.

By symmetric transfer it is meant that the effect of work in one experimental condition (condition A), when it precedes the second condition (the condition B), to work in the condition In turns out to be similar to the effect that work can have in the In condition set before condition A, on job A in condition A.

For example, if in an experiment in conditions A and In the subject solves mathematical problems of different types, the way he worked out for solving the problems of condition A can negatively influence the development of the method B. But at the same time, if the subject first solves the problem of the condition B, the work done by him for this also has a negative effect on his work in the following condition A. Suppose that first the performance of the condition A conditions worsens the performance in the condition In by seven conventional units and first performing the conditions In conditions affects performance in Condition A by the same seven units. Here we can talk about the presence of a symmetric transfer.

The effect of symmetric transfer can be overcome by using positional adjustment schemes in which each level of an independent variable follows each of the remaining levels, and also precedes each of them the same number of times (more details about such sequences will be discussed below).

For asymmetric transfers, the impact of work in Condition A going first is to work in the B condition and work in the In condition , going first, to work in Condition A turns out to be unequal. For example, the previous condition A may negatively affect the condition B, while the condition B may not affect work in condition A. Alternatively, the influence can be simply different: in one case, stronger, in the other - weaker. Also, the effect can be multidirectional: if the subject first works in condition A, this can have a negative effect on the work in the condition B, worsening the result, while working first in condition In can affect the performance in Condition A positively, improving the result.

To illustrate the asymmetric transfer, imagine this situation. You do your homework or are engaged in some other intellectual activity in silence. And suddenly neighbors include loud music. Obviously, the effectiveness of your activities will decrease. Or another situation: you work under loud music, which is heard from behind the wall, and suddenly the music stops. Obviously, the effectiveness of your activities should increase. However, there is a high probability that in the first case the effectiveness of the activity will fall much more than it will increase in the second case.

Consider another example. The manufacturer of the smartphone wants to find out which virtual keyboard is best used for typing text messages. There are two keyboards involved in the testing, which differ in the need for additional actions to access the most used punctuation marks.

The intrasubject experimental plan is used. Subjects type the same text using each of the tested keyboards. The speed of typing is measured. Suppose that if the subject first typed the text on the keyboard, where additional switching is not necessary to access punctuation marks, this does not affect the speed with which the dial-up is subsequently made using the keyboard with the need to switch. At the same time, if the subject first typed the text using the keyboard, where switching to access punctuation marks is necessary, this negatively affects the speed of his work with another keyboard. For example, he may be so used to pressing the switch key to access punctuation marks, which will press it even if it is not required, which will lead to a loss of time when typing each punctuation.

Suppose half the subjects first used the keyboard without switching (condition A) and the other half with a switching keyboard (condition B). The subjects who first typed the text using a keyboard without switching, spent 20 seconds working with this keyboard, and 30 seconds using a keyboard with switching. The subjects who first typed the text on the switching keyboard also spent an average of 30 seconds in this condition, and when they typed on the keyboard without switching, they spent an average of 40 seconds (Table 12.1).

Table 12.1

Hypothetical results of the study of typing speed on different keyboards with different presentation sequences (time in seconds)

Terms

A (keyboard without switching)

In (keyboard with switching)

First A, then In

20

30

First B, then A

40

30

Average

30

30

Thus, as a result of the research, it turned out that there are no differences in the speed of printing with the help of different keyboards (the average time for typing a text message for each type of keyboard is the same). It should be understood that in fact the speed of printing on the keyboard without switching is higher, and this result is only a consequence of asymmetric transfer, which can be seen by dividing the groups of subjects according to the sequence of presentation of different keyboards.

Asymmetric carry can not be overcome by means of positional adjustment. If it is suspected that an asymmetric transfer may occur in the study, it is necessary to abandon the intra-subject plan and use an inter-group plan, in which each subject will participate only in one experimental condition, and therefore there will be no effect of the conditions on each other.

So, when using intrasubject experimental plans, all levels of an independent variable are presented by the same subject. This reduces the number of subjects required, but can lead to undesirable influences caused by the order of conditions following each other (sequence effects). To compensate for these effects, specially developed methods of presenting the levels of an independent variable are used, designed to evenly distribute their influence on all the investigated conditions. However, this can only work if the assumptions about the linearity of the progression and the symmetry of the transfer are satisfied. If these assumptions are not fulfilled, the researcher must refrain from using the intra-entity plan in favor of intersubject.

thematic pictures

Also We Can Offer!

Other services that we offer

If you don’t see the necessary subject, paper type, or topic in our list of available services and examples, don’t worry! We have a number of other academic disciplines to suit the needs of anyone who visits this website looking for help.

How to ...

We made your life easier with putting together a big number of articles and guidelines on how to plan and write different types of assignments (Essay, Research Paper, Dissertation etc)