Twin Studies: Experimental or Correlational?

Travis DixonUncategorized

Twin studies are common in the study of genetics. But what "research method" are they?
In the new course students need to be able to discuss how and why research methods are used for every topic. They might, therefore, get exam questions that look like this:
  • Outline the use of one research method used to study the link between genes and behaviour.
The most common method used in studies on genetics and behaviour in my course, I would call “twin and adoption studies.” (The IB uses the far less common term “kinship studies” but I don’t know why). But this is not considered a “research method” by the IB. So a couple of days ago I posed a question on a forum about how to classify twin studies that compare DZ (fraternal) and MZ (identical) twins. From the discussion, it there emerged a central question:
  • Are these experimental or correlational?
The reason this might be important to know is because mark schemes often have very particular rules about what is and isn’t considered a research method. If students identify a method that isn’t considered by the mark scheme to be a “research method” (e.g. animal study), they run the risk of writing amazing essays that score terrible marks.
This is why I’d rather be safe than sorry and get some clarification. Here’s where we landed…

The case for quasi-experimental

There were some really good points made by experienced and highly qualified teachers for how twin studies can be considered a type of quasi-experiment. If we think about it from a basic definitions view point, it’d be easy to say that the IV is genetic similarities (DZ or MZ) and this is a naturally occurring variable, so this could be classified as a natural experiment (a sub-type of quasi-experiments).
Another point is that twin studies are considered an example of matched pairs design in the new guide. If they’re matched pairs, this means they’ve been matched across conditions of the IV, and if there’s an IV then they must be experimental.
Of course they’re not a true experiment, because the researchers aren’t manipulating the IV and the participants cannot be randomly assigned to either condition, so they could be best considered quasi-experimental.

The case for correlational

On the other hand, there were some equally good points made by equally qualified and experienced teachers saying how twin studies can be considered correlational. And this makes sense, as after all, they’re measuring correlations between differences of behaviour between DZ twins with differences in behaviour between MZ twins in order to determine heritability. So in this sense, there are two variables that co-vary:
  • Variable one: differences in behaviour between DZ twins
  • Variable two: differences in behaviour between MZ twins

Twin studies compare these differences to draw conclusions about heritability (the extent to which variations in behaviour between individuals can be attributed to genetic factors). They are not used to draw causal relationships, unlike experimental studies. So in this sense, it makes more sense to consider them correlational.

Furthermore, it’s my understanding that the strict definition of “quasi-experimental” is there needs to be an identifiable treatment (IV) acting on a DV, and sufficient controls should be in-place so that some type of effect is being observed. As Hugh Coolican proposes in his book, (Research Methods and Statistics in Psychology), a study that simply compares groups based on existing differences (e.g. gender, disposition, an presumably genetics?) should not be called quasi-experimental, but rather called a “group-differences study,” seeing as there is no other widely accepted term to describe these kinds of studies comparing groups without a “treatment.” (That being said, the definition of a quasi-experiment in the IB guide is when “participants are grouped based on a characteristic of interest, such as gender, ethnicity, or scores on a depression scale.”)

If one is to take this view, then we can’t classify twin studies as quasi-experimental because the existing differences between these groups enable so many potential confounds that we can’t confidently say that the genetics is an identifiable “treatment.” For example, fraternal twins are often different genders. How could we say that variations in behaviour between a set of twins that are boy-girl, and another MZ set are girl-girl are due to genetics, when gender and socialization are massive potential confounds?

As you see, there are good arguments on both sides.

So where to from here?

This leaves me with the obvious conclusion that when asked about “one research method used to study genes and behaviour,” students would be able to use DZ-MZ twin studies and identify this method as either being quasi-experimental or correlational, because either is a defensible position.

Which leaves me with another question that I’ve been asking for years:
  • if the distinctions between methods are not black-and-white, why are we asking kids to classify studies into black-and-white methodological categories in the first place?

Examination advice

This question won’t be answered any time soon, if ever, so I need to think about how to deal with this come exam preparation time. If students want to write about DZ-MZ twin studies in response to research methods, my advise is that:

  • For SAQs you (should) only need one method, so identify one method or the other (i.e. quasi-experimental or correlational).
  • Make some attempt to outline at least one reason why the twin-study could be considered a particular method (e.g. correlational or quasi-experimental). This will set the argument in context, and will also help show knowledge of the research method.
  • When explaining the use of the method, be sure that the reason is connected to the method identified: e.g. the value of studying naturally occurring variables (quasi-experimental); or, the value of drawing correlations between variables that exist between groups (correlational).
  • For essays asking about one or more methods, that’s simple – you can identify both.
  • For essays asking about only one method (which shouldn’t be an essay question, but you never know), students could identify one method and use as their core argument with one study to demonstrate, and then use a second study and argue why this method could also be considered the other type of method. Critiques of using twins could be part of the discussion/evaluation as well.

HL v SL

This may not be such an issue for HL students, depending on how you’re going to teach the HL extension on the use of animal studies to research genetics. In my course I’m planning on looking at how knockout mice are used in true experiments, so my HL students should have more than enough to write about the use of true experiments to study the links between genes and behaviour. In which case, all this becomes a bit irrelevant. But my SL kids won’t have that luxury, so it’s still very relevant for them.

I hope this was helpful. It’s certainly helped me think through the issue and how to deal with it.