A threat to internal validity…communication between groups.
Posted on: March 10, 2012
In the Psychology world communication between groups in an independent measures design can be something of a disaster. It can cause diffusion, whereby the act of someone from one treatment condition talking to someone in the conflicting condition can cause treatment effects to spread across groups as those learning of the different treatment may start to filter in a bit of what they’ve found out. This can make any possible difference in treatments decrease.
Compensatory equalization can occur. If in a research study a treatment group was administered a new drug for depression and the control group was given no intervention at all then the control group (upon finding out) may demand that they receive the same treatment as the experimental condition. If researchers choose to meet this demand then they have eliminated their control group and are left without a comparison for the remainder of the study.
Communication between groups can also affect performance both positively and negatively. It can cause compensatory rivalry whereby the control group increases their efforts in retaliation so results are unnaturally high therefore showing little difference between groups. It can also go the opposite way and cause resentful demoralization where the control group will in effect give up and intentionally decrease their participant effort due to resentment of the experimental groups receiving special treatment causing control results to fall lower than expected and allowing a possible Type 1 error to occur. Schwartz et al. (2001) conducted research into effects of exercise on fatigue for patients undergoing chemotherapy. They specifically chose a one-group design with no control to avoid potential compensatory equalization effects and resentful demoralization.
Communication between groups can have very serious effects in real world scenarios. To prevent bias in a trial the jury is on rare occasions ordered to be sequestered to allow for an objective decision, this is the total isolation of a jury. In the case of the George Huguely V murder trial (George Huguely was found guilty of the 2nd degree murder of Yeardley Love on February 22nd 2012 and sentenced to 26 years) great precautions were taken to pick a jury. Huguely’s lawyers requested a sequestered jury but the judge chose to pick a jury that swore under oath they had done all they could to shield themselves from the vast amount of media coverage the case had created. The murder of Yeardley Love has received huge attention by the media and many thought it impossible to find a single juror that wasn’t aware of the case. In the end the 12 jurors picked had all heard of the case and they were carefully selected as individuals who did not possess any fixed prejudicial notions.
http://www.cltv.com/topic/bs-md-huguely-jury-20120220,0,13726.story
7 Responses to "A threat to internal validity…communication between groups."
March 14, 2012 at 1:55 pm
I was pleasantly surprised to see such an interesting topic – something I had never really considered before.
I believe that the biggest threat that communication would have to research is using placebo studies. Imagine that there is one group being given a drug to increase hyperactivity and the other is given a placebo. Even if neither group knows which they have received, just them talking about what effects they feel can affect results. By this, I mean that the drug group can say that they feel dizzy, alert etc, then the control group may feel they should feel that too and so say that they do – oh dear!
In their analysis of over 900 ghostly experiences, Lange et al. (1996) noted that approximately 60% of reports mentioned some form of prior suggestion that the location was haunted (e.g., rumours, advertising, or prior knowledge of previous experiences reported in the location). So if this can affect ghost stories, prior knowledge gained from different research groups can certainly affect data!
Thanks,
Abby
http://www.questia.com/googleScholar.qst?docId=5000634330
March 14, 2012 at 5:00 pm
Very good blog, and a very interesting topic. You do get some other issues when groups communicate, it can influence opinions and factors such as social desirability. It can also have an effect on peoples opinions which can be changed and can also become more extreme (Stoner 1961).
References:
Stoner, (1961) http://www.citeulike.org/group/13195/article/7592316
March 14, 2012 at 9:08 pm
I agree with you that between groups design has effects on an experiment but it is good because each individuals score is independant of the other scores because each participant is only measured once and this reducing the fact of any other treatment factors. It is also a good design because it can be measured in a wide variety of research questions and when there are two or more treatment groups. However it is at disadvantage because much more participants are needed and like you say treatment effects occur for example what one person says might influence someones behaviour thus would not produce natural scores. It is also hard to restrict communication between participants and this can result in things being filtered into the data depending on what they have been told about the experiment.
March 14, 2012 at 10:46 pm
You make some interesting points about how communication between groups can cause havok with the internal validity of a study. When two or more groups can comunicate with each other they can influence each other and so change the results of the study than would otherwise be found. It requires effort to keep experimental groups separate and so it can be a trial to spend time and money maintaining the separation, but it must be done to maintain the integrity of the study.
April 16, 2012 at 12:01 pm
It seems to be a challenge to avoid this inter-group communication to happen in certain situations. There are so many studies which indicate the threat of internal validity due to compensatory rivalry. In one piece of evidence known as the John Henry effect, this treatment condition has lead to a tragic incidence due to overexertion (Cook and Campbell, 1979, p. 55). Therefore, I agree with you that inter-group communication is a serious threat to internal validity in independent measure design studies. In order to avoid contamination, the passing on of information between groups, to take place researchers should encourage their participants to take the investigation for serious and keep information to themselves until testing is over. You bring good point to the attention of designing valid and reliable research and your points are strong. It is also great how you link scientific evidence with examples that are part of everyday life.
http://dissertation.laerd.com/articles/internal-validity-an-overview-3.php
emcg1 
March 11, 2012 at 2:17 pm
You make good comments of the issues related to communication between groups in experimental designs. This is no only an issue in independent-measures sampling, but can also effect repeated-measures. This is the case as counterbalancing means people partake in different levels of the experiment in different orders, and knowing what task is next as a result from communication could confound the data. You mention about the competition being caused between groups. Sometimes this is caused deliberately, and is used as a measure. Take the Stanford Prison Experiment as an example – there was competition deliberately created between the two groups of prisoners and guards, simply to see the effects of the roles. However, like you say, it is more often than not a negative thing, as it would confound the data.
You raise a good point about jurors. It is definitely difficult to find individuals that have not been prejudiced by media distribution of information about a case, especially in large-scale and well-know cases such as the one you’ve outlined above. Any prior decision made by the juror could confound the decision made in court, and so this lack of communication of sorts is vital. The same goes for the judges of a case – they are not to know the details of the trial before the trial begins, as to not bias any outcomes.