Are Within-Subjects Designs Any Good?

Within-subjects designs are an essential part of research, especially in the context of sensitive topics. The basic definition for this design is when the same participants are used across all conditions of the study. This could be either across all control and testing conditions in an experiment, or a longitudinal study in which the same individuals are assessed over time. This obviously shouts out brilliant benefits for its use, however, it does result in some questionable effects.

When using the same individuals for every condition, the first advantage that comes to mind is that less people are required which therefore requires less of the experimenters’ time recruiting participants. This is especially beneficial when researching sensitive topics, for example domestic abuse. In this context people may be reluctant to come forward for research into abuse, and so the within-subjects design allows research to be condutcted with only a few participants. Another advantage of the within-subjects design is the removal of individual differences. As participants are the same across conditions, there are no differences between conditions. This reduces variance and confounding variables, which enables clearer analysis of the difference between conditions. 

However, the replication of participants isn’t always a good thing. Due to the same individuals being in every treatment, the conditions have to be carried out at different times which can result in extraneous variables.

  • History – outside events influence scores in one treatment differently to those in another.
  • Maturation – physiological or psychological changes that happen whilst the study is being conducted that effect scores. (E.g. illness)
  • Instrumentation – changes in the measurement apparatus that effect scores.
  • Testing Effects – participants’ scores are effected by an earlier condition. (E.g. order effects, fatigue)

These factors can threaten the internal validity of participants’ scores which can in turn threaten the external validity. This can be detrimental to research as it can cause the results to become inaccurate and falsely generalisable. Although within-subjects designs provide great opportunities for research into delicate areas and drastically reduce indiviual differences, it is impossible to have complete control over effects that may ruin validity. Therefore, all variables must be taken into consideration before deciding which design is best to use.

 

http://en.wikipedia.org/wiki/Repeated_measures_design#Uses_of_a_repeated_measures_design

http://www.psychexchange.co.uk/glossary/repeated-measures-design-263/

http://psychology.about.com/od/windex/a/within-subjects-design.htm

Gravetter. F. J. & Forzano. L. B. (2009). Research Methods For The Behavioural Sciences. Belmont: Wadsworth, Cengage Learning.

Advertisements

7 Comments (+add yours?)

  1. psuc28
    Mar 26, 2012 @ 15:40:41

    Within-subject designs seem to be easier because we need less participants however, as you pointed out there is the chance for individual differences and carry over effects. Say there are 3 conditions, A,B and C and we use no counter-balancing all participants take part in tasks A then B then C. However, before we start our research we do not know that treatment group B is actually a strenuous task therefore it affects performance on task C for all participants. If we were to use Counterbalancing techniques such as the Latin-Square Balanced Counter Balancing, each condition at one point will precede each condition at one point in time therefore excluding any carry over effect. So, providing that researchers try and eliminate these effects through counterbalancing then within-subjects designs could be considered a good research design.

    (For more info on Counterbalancing follow this link- http://www.experiment-resources.com/counterbalanced-measures-design.html)

    Reply

  2. PSUD00
    Apr 18, 2012 @ 16:51:15

    Deciding on a design is one of those issues where both situations have their advantages and disadvantages. I agree that the main advantage with using a within-subjects design is that there are less participants required. This can be helpful to reduce cost as well as you are more likely to find the amount for participants you need especially when the participants you need are rare cases (like your example of domestic abuse sufferers.) Also when using the same participants you are able to reduce the error variance associated with individual differences (http://web.mst.edu/~psyworld/experimental/within_subjects.html). By using the same participants in each condition then there are no longer any issues of ‘this group has more experience in the area’ or ‘this group have a higher average age’. Though difference may be small it can increase the variance.
    One of the big problems of within-subjects design is carryover effects. This is where the first condition has an effect on the participants during the second condition. (http://www.experiment-resources.com/within-subject-design.html). These effects are practice effects, where the participant has had time to grasp the task and therefore their ability to complete it has improved, and fatigue effect, where the participant is tired and bored of participating and consequently their performance decreases. These issues can be solved through counterbalancing though this is not always possible.

    Reply

  3. Trackback: Homework for my TA | PSUD00
  4. Trackback: Final Blog Comments! :) « All About Psychology
  5. psuc0e
    Apr 18, 2012 @ 21:25:21

    Personally, I much prefer between groups designs as I think, despite it being very heavy on numbers of participants used, it is better to only allow one participant to participate in one condition. There are times when it is not possible or what is being tested requires the within subjects design, but if its possible, try to use a between groups design.

    There are a number of problems that could prove problematic if you carry out this type of experiment, such as fatigue and participants getting used to the test and knowing what to expect. If participants get bored, can’t complete the study, or chose to drop out given how long it is taking, it could have a negative effect on your research.

    It’s not to say that the other types of research aren’t prone to their own problems, but I think that this type of research is more susceptible to issues and that these issues are much more easily remedied. Methods such as counterbalacing can help in a within-subjects design, but it is still easier to use one participant per condition. That’s what I think anyway.

    Reply

  6. Trackback: Final comments for Thandi « A Research Skills Blog
  7. psuc3d
    Apr 18, 2012 @ 22:45:55

    Hello, really enjoyed reading your blog! I would just like to add that within-subjects experiments are extremely useful in experimenting the same participant but in two different points of time (http://web.mst.edu/~psyworld/within_subjects.htm) – for example, once measure before an event and one measure after. Although this can be done similarly with between subjects if they are matched, I personally feel that within subjects is more appropriate.

    Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: