Step 5: Controlling Extraneous (Confounding) Variables

Instructions: Click on the sections below to see the explanations.

Choosing Your Research Design (Click)

There are two main research designs that you can choose from for your experiment, depending on how you want to put your participants in two different conditions:

  • Independent measures design: Participants are assigned to EITHER the experimental OR the control condition.
  • Repeated measures design: Participants are assigned to the experimental AND to the control condition (one, THEN the other.)


Both designs have methodological advantages and disadvantages as they might lead to or help avoid extraneous (confounding) variables. To choose between them, you can look at the one used in the original study you are replicating, but you may also make a different choice.

Table 8: Advantages and Disadvantages of Independent and Repeated Measures Designs

Mitigating Techniques (Click)

There are different techniques you can use to mitigate the disadvantages of your chosen research design:

  • Random allocation: Participants are individually assigned to the experimental or control condition using an unbiased technique that gives all of them equal statistical chances of being assigned to either condition. This helps decrease the risk of participant variables in independent measures design and is easily done with one of the many random groups generators available online.
  • Matched-pairs design: Participants are matched based on relevant characteristics so that comparable pairs can be assigned to the experimental and control conditions. This is another way to help nullify participant variables in independent measures design and can be combined with random allocation.
  • Counterbalancing: Participants complete the tasks involved in the experiment in different order. This helps nullify order effects in repeated measures design (not by eliminating them, but rather by creating two different sets of order effects that should theoretically cancel out). Note that counterbalancing can be also useful in independent measures design if the task(s) that participants have to perform  involve a risk of order effects.


Again, to decide which one(s) to use in your experiment, you can start by looking at the original study, but you may also make a different choice.

Controlling Other Extraneous (Confounding) Variables (Click)

To control extraneous (confounding) variables beyond the ones related to your research design, you can once again look at the procedures of the original study you are replicating. However, you should also consider the techniques listed below, which can help you turn extraneous (confounding) variables into “controlled variables”. As always, keep in mind whether you plan on conducting your experiment in person or online.

Table 11: Techniques to Control Extraneous (Confounding) Variables

Controlling Participant Variables (Click)

Since you are conducting an experiment (and not a quasi-experiment, for instance), participant characteristics should not play any role in your study. Participants in each condition should simply belong to the target population and be otherwise comparable.

  • Ways to avoid and control participant variables through your choice of research design (repeated measures) and mitigating techniques (matched-pairs, random allocation) have been explained above.
  • In addition, you might use purposive sampling, which is a technique enabling researchers to select participants based on a particular set of characteristics that is relevant to the experiment.
    For instance, a study on the effect of listening to music with lyrics while completing school work might only select participants who self-report having this study habit. This would help avoid the extraneous (confounding) variable that would come from participants being familiar/unfamiliar with the task. Importantly, purposive sampling does not make an experiment a quasi-experiment as long as ALL participants share the same relevant characteristic and researchers still manipulate the independent variable. 
  • In some cases, you might need to use exclusion criteria if certain individual characteristics would make a type of person unfit to participate in your experiment, either for technical or for ethical reasons. You should list such criteria ahead of time and find a respectful way to obtain this information and exclude such participants, either by not selecting them or by not including their results in your data (in which case data collection cannot be anonymous).
    For instance, deafness would be an exclusion criteria for a study on music.
Controlling Tasks/Stimulus Effects and Variables (Click)

If tasks and/or stimulus have to be different in the experimental and control conditions because of the manipulation of the IV, they should still be comparable in all other relevant respects for all participants.

  • This is first controlled by properly operationalizing the independent variable. The way it is manipulated should create two conditions that are consistent (each condition is the same for all participants in that condition) and otherwise comparable (conditions only differ in terms of the IV).
  • A related way in which task/stimulus effects and variables should be controlled is through the use of appropriate procedures and materials. For instance, if participants are asked to memorize and recall a list of words, they should all be given the same amount of time to do so (procedures). Likewise, if two different lists are used in a repeated measures design, they should present the same level of difficulty (materials). 
  • To control this level of difficulty, we might use a random words generator and test different lists during a pilot study. A pilot study is a practice test of an experiment meant to check that procedures and materials are appropriate. It could also help us check that the number of words on the list as well as the time allotted to memorize and recall them is appropriate, thus avoiding ceiling effects.
  • Finally, ways to control order effects through independent measures and/or counterbalancing have already been explained. An additional technique is the use of “filler tasks”, which are tasks that have no other purpose in the experiment than to distract participants or to keep them busy during a period of time. 

Controlling Situational Variables (Click)

Just like tasks and stimulus, if the situations have to be different in the experimental and control conditions because of the manipulation of the IV, they should still be comparable in all other relevant respects for all participants.

  • Again, this is controlled by properly operationalizing the independent variable and using appropriate procedures and materials. The manipulation of the IV and its implementation through procedures and materials should create two conditions that are consistent (each condition is the same for all participants in that condition) and otherwise comparable (conditions only differ in terms of the IV).
    Regarding consistency within conditions, students have different tastes in terms of music, so that the same situation (listening to the same song) might affect them differently depending on whether they know and like it or not. One way to control this potential extraneous (confounding) variable is to use procedures giving participants the choice to listen to any song (with lyrics) they want.
    Regarding comparability between conditions, an example comes from the fact that, if participants in the experimental condition only listened to music with lyrics while memorizing the words and not while taking the test, they would be trying to recall them in a different environment (in silence), while participants in the control condition would be memorizing and recalling the words in the same environment (both times in silence). This could be an issue because Context Dependent Theory explains that it is easier to recall information in an environment that is similar to the one in which it was encoded. One way to control this is to have participants in the experimental condition listen to music with lyrics both while memorizing the words and trying to recall them.
  • Another way in which situational variables should be controlled is by testing all participants in comparable environments. An obvious technique is to use appropriate and similar times and locations.
    For instance, we shouldn’t test one group at a time when they are still full of energy and another one when they are tired after a long day. Likewise, we shouldn’t test one in a location where it is easy to focus and the other one in a different location where it is hard. However, even the same location can affect different conditions differently. Thus, a noisy environment would negatively impact the performance of a control group supposed to memorize and recall words in silence, while it would not affect an experimental group listening to music with lyrics while completing this task.
  • If different environments have to be used (e.g., because you are conducting your experiment online), they can still be homogenized (i.e., made comparable) by identifying the aspects in which they should be similar and taking steps to control them. This will be reflected in your procedures and materials.
    For instance, we might plan on controlling noise levels, either by providing participants with earplugs or, if they are completing the task online, by having them do so in a quiet environment.
Controlling Researcher Biases (Click)

Your behavior as researcher (presentation of your study, conduction of the experiment, collection and interpretation of data) should only influence participants through the manipulation of the IV. In particular, you should be mindful of the different ways in which you might unconsciously try to confirm your research hypothesis.

  • To avoid biases, you must plan ahead of time everything that your group will be doing (experimental script) and saying (standardized instructions) during the experiment. As you conduct the experiment, you should not do or say anything that hasn’t been scripted or standardized ahead of time.
    For instance, if participants have to be redirected, this should be done by using standardized redirections.
  • If the dependent variable is not itself a number, you should create scoring guidelines ahead of time. You might also want to use triangulation (more than one scorer), ideally with scorers that are independent from the experiment and blind to its research hypothesis.
    For instance, we should determine ahead of time what we will count as “correctly recalled” words: do synonyms count?
Controlling Participant Biases (Click)

To provide a reliable answer to the research question, your experiment should investigate participants’ authentic and truthful behavior. This means that their behavior should not be influenced by the very fact that they are taking part in an experiment.

  • To avoid demand characteristics, you must adhere to a strict experimental script, including standardized instructions, so as to not give participants unintentional cues.
  • As explained above, using an independent measures design can help decrease the risk of good/bad participant effects, as it makes the manipulation of the IV, and thus the research hypothesis, less obvious. However, this does not completely eliminate the risk of good/bad participant effects, and there are other ways to control them that are also applicable to a repeated measures design.
    For instance, using volunteer sampling makes it more likely that participants are properly motivated and don’t display a bad participant effect. However, it can also make it more likely that they display a good participant effect.
    To avoid this, researchers can use limited disclosure, i.e.,  present the experiment in a way that is truthful (not deceiving) but does not let participants guess or misconstrue the research hypothesis.
    Of course, exclusion criteria should ensure that participants do not have prior knowledge of the research hypothesis; but the use of filler tasks can also help dissimulate the manipulation of the IV.
  • Importantly, if you have reasons to suspect that your participants’ behavior was not authentic and truthful, you should make necessary changes and re-run your experiment. To avoid this, you must plan ahead how you will prevent participants from “cheating” during the experiment.
    For instance, if you are conducting your experiment online, you might consider the use of tools such as exam.net that allow you to monitor participants’ behavior (just like you would in a traditional setting).
  • Finally, to control audience effect, you should think of ways to make the collection of data anonymous.This does not only help decrease the risk of audience effects, but also protects your participants’ privacy, which is one of the ethical standards that your experiment must follow.
    For instance, if using a Google Form to collect participants’ answers, you could make it so that their email address is not collected.

QUESTIONS: Click on the questions below to see the worked examples.

Answer all the questions below on your individual Experiment Creation Form.

Question 11: Justify the research design that you will be using in your experiment. (Click)

Your answer should include:

  • A description of the research design that you will be using
  • A justification of your choice based on its specific advantages in the context of your experiment.

Table 9: Example of Justified Choice of Research Design

Question 12: Justify the mitigating technique(s) that you will be using to address the limitations of your chosen research design. (Click)

Your answer should include:

  • An explanation of the limitation(s) of your chosen research design in the context of your experiment
  • An explanation of how mitigating techniques can help alleviate these issues

Table 10: Example of Explanation of Use of Mitigating Techniques

Question 13: Explain how you will control the extraneous (confounding) variables that you have previously identified. (Click)

Your answer should include:

  • A detailed description of the techniques that you will be using to control all the different extraneous (confounding) variables that you have previously identified.
  • A justification of your choice explaining how each technique helps control each extraneous (confounding) variable.

Table 12: Example of Controlled Variable