Science students don't get experiment variables? Try causal diagrams instead

Christian Moore Anderson
1 day ago
7 min read

Updated: 1 hour ago

Experimental design is an uncontroversial component of science curricula. It helps students learn methodology, how to spot flaws in arguments, and, partly, how scientists create knowledge. Yet, much of what we teachers see in experimental design is hidden from students.

In schools, it is common for students to be tasked with creating lists of variables with peculiar names: independent variable, dependent variable. Teachers often decry the failed efforts in helping students remember which is which.

What is the cause of this problem? In my view, the list is the problem. Behind the list is a set of causal relationships the teacher understands and can mentally map. For the student, however, it is just a list of terms that they don't fully understand.

This caused another problem; they found it boring.

Influenced by the science of cause and effect (Pearl and Mackenzie 2019), I created a causal diagram suitable for experiment design in schools; an alternative to list-making.

It makes experimental design more meaningful and enjoyable for all. My students now actively engage in discussion about causal relationships and, in doing so, construct a better conceptual understanding of what they are testing.

I’ll build a diagram slowly and finish by discussing how I’ve implemented it in my science teaching.

Distinguishing between the independent and dependent variables

I’ve always found that students struggle with the terms of independent and dependent variables. Yet, they find the question, Does X affect Y? quite intuitive.

However, I prefer the phrase, If I vary X, does Y vary? Let’s say we wanted to find out if the ratio of soil to sand as a growth medium affects the post-germination growth rate of some seeds. We could draw it like this (Figure 1):

Figure 1. A causal diagram showing an independent (X) and dependent variable (Y).

At the moment, this remains a question. To make this clear to students, I pretend to pinch X and wiggle it up and down. I rephrase the question: If I wiggle X, will Y wiggle too? If varying X does cause a variation in Y, then the arrow represents this relationship. If it doesn’t, we will remove the arrow.

As students get used to this idea, I can later explain how the variation in Y is said to depend on the variation in X, but not the converse.

In other words we’re suggesting that X→Y, but not X⇆Y or X←Y. Therefore, X is independent of Y. Using this diagrammatic form, my students have found this intuitive. Next, however, they must distinguish the dependent variable from what is measured.

Distinguishing between the dependent variable and what is measured

When I ask students what they intend to measure, regardless of age or stage, they tend to reply that they’ll measure their dependent variable. In our example, they’ll repeat that they’re measuring the growth rate. When asking how they’ll do that, they get confused.

Consider the next diagram (Figure 2), which now clearly distinguishes between the two:

Figure 2. A causal diagram showing an independent (X) and dependent variable (Y) and the measure of the dependent variable.

Once I introduced these diagrams to my courses, the lack of distinction was no longer a problem.

In this example, the hypothesis is that the soil-to-sand ratio will affect the growth rate. In turn, the growth rate will affect the stem length after some time.

There are other ways of measuring the growth rate, of course. For example, the Y_measure could also be weighing the wet mass, measuring the root length, or counting the number of leaves.

Making the Y and Y_measure distinct in the diagram has helped my students explore the available measures without getting confused.

Distinguishing between the independent, dependent, and other variables

When controlled variables are discussed at school, students are often asked to create a list of things to keep the same.

The list, however, obscures why some variables must be controlled; the causal relationships are hidden in the text, if mentioned at all.

Consider the next diagram (Figure 3) in which I add a set of other variables, which I simplify to Z. If any of these vary, they will also vary X, Y or Y_measure.

All these variables must be prevented from varying, or, in other words, they must be controlled; they are necessary controlled variables of the experiment.

For example, if I pinch the Z of watering rate and give it a wiggle (by watering some seeds more or less), it will affect the growth rate, and we must prevent this. If both the soil:sand ratio and the watering rate is different between the seeds, which is the cause of the changes we observe? It's impossible to tell.

Figure 3. A causal diagram showing independent and dependent variables (X and Y), plus other variables (Z) that must be controlled

At this point, the diagram is at a level suitable for young secondary school science students. It signals two interesting distinctions that typical variable lists fail to reveal.

Firstly, the duration of the experiment is a variable to be controlled. You wouldn’t let some seeds grow for five days and others for ten. This variable directly affects the stem length (our Y_measure) rather than the growth rate (our dependent variable).
The second interesting distinction is the relationship between the growth rate, light intensity, and position in the room. The diagram shows that we don’t necessarily have to control the light intensity if all the plants are placed in the same position in the same room. In this case, all the plants will experience the same fluctuations in light intensity.

Pondering the mechanism through the diagram

As students progress through secondary school, they can build more complex diagrams.

Consider the next version (Figure 4) in which a high school student wants to hypothesise a mechanistic relationship between X and Y.

Figure 4. An elaboration of Figure 3 in which a causal mechanism is suggested that connects the independent (X) and dependent (Y) variables.

The diagram now proposes two major causal pathways from X to Y.

Nutrient availability: sand and soil offer different minerals in differing concentrations.
Water availability: sand retains less water than soil. (Water retention could be added as another variable, but simplification prevents diagrams from becoming overwhelming.)

Unknown variables

Finally, you can add “unobservable” or “unknown” variables to the diagrams, depending on what the designer wants to communicate to the reader.

Consider the final example (Figure 5) in which a student wants to communicate that there is a variable they can’t control, but recognise its confounding contribution to the experimental results.

Figure 5. An elaboration of Figure 4 in which an “unknown” variable is included.

In this case, a student could buy seeds from a shop, but they can’t know the genetic variation between the seeds. Nor will they know if there is any variation in the stored nutrient type or quantity. To address the problem, students may include a large sample size that could average away the variability.

Implementing causal diagrams in teaching

Below, I’ll give examples of how I have implemented the diagrams in secondary science courses. While causal diagram construction is embedded in my courses (see Difference Maker), my experience has shown that students of all secondary school ages find these diagrams intuitive and are easily introduced at any level.

Full science investigations

Our Year 7 (11-year-old) cohort is asked to design, carry out, and analyse an experiment, before presenting their work in a science fair. We ask them to list their variables and create a causal diagram to construct and communicate their reasoning. Their diagrams may look like Figure 3.

I also teach the IB Biology course (16–18 years old), in which students must do the same for 20% of their final grade. The introduction of the causal diagram in their work has vastly improved student understanding and the quality of our teacher-student conversations. Their diagrams may look like Figure 5.

Regular investigation design

Designing experiments (without actually carrying them out) is a common classroom activity and is regularly assessed in GCSE standardised science exams (16-year-olds). Yet there is often no framework for students to follow, except mnemonics. An example I’ve come across is CORMS, which stands for: Change, Organism, Repeats, Measure, Same.

Causal diagrams offer a framework that goes beyond assisting memory, which can also easily be carried out in any lesson. Instead of hearing groans when asking students to design an experiment, they now engage in deep discussion about which variables affect which.

Having embedded these diagrams in my courses, I have created a mark scheme that has been used to guide and assess our students’ ability to design experiments. As everything can’t be clarified in a simplified diagram, I’ve also asked students to write a short written response. In Table 1, you can see an example of a mark scheme for a biological experiment.

Aspect	In diagram (correctly connected)	In writing (correctly described)	Available marks
X	✓	✓	2
Y	✓	✓	2
Y – measure	✓	✓	2
Y – measure tool used		✓	1
Y – number of repeats		✓	1
Z – about the organism	✓	✓	2
Z – about the organism	✓	✓	2
Z – about a condition	✓	✓	2
Z – about a condition	✓	✓	2
Z – about the measurement	✓	✓	2
Extra details	✓✓ (in either)	2
			Total: 20

Table 1. A mark scheme for marking experimental designs that include a causal diagram.

Conclusion

Written text can only express ideas linearly (one dimension), sentence by sentence. Yet science is a subject of many interconnections. Diagrams can express our ideas better by existing in two dimensions. Students can see connections immediately instead of having to go back and forth between sentences.

The diagram I've created gives another benefit: it makes key distinctions clear. With a shared form of expressing our experiments, students understand what I expect.

Finally, by focusing on a causal diagram, the very essence of the science comes to the fore: the causal relations. If you want to read more about co-constructing diagrams with students, see Difference Maker.