Is the Traditional “Scientific Method” the Best Way to Learn Science?

The research: Shemwell, J., Chase, C. & Schwartz, D. (2015) “Seeking the General Explanation: A Test of Inductive Activities for Learning and Transfer.”

Journal of Research in Science Teaching, Vol. 52, No. 1, pp. 58–83.

Recently I have come across a number of quotes claiming that “Complexity is the enemy of execution,” or simply “Complexity is your enemy.” I agree that the world in all its complexity may be difficult to deal with and that, particularly in the realm of education, the complexity of the learning process can make our task feel overwhelming. Finding ways to simplify is perhaps a necessary part of helping us plan our way forward.

However, simplicity can also be the enemy. When we work from simplified models of reality, we can often forget that they are just that—simplifications, not reality itself. While such models can be helpful, putting too much faith in them can obscure some of the important nuances of reality.

Perhaps a better way forward is to have at our disposal multiple models between which we continually move back and forth, engaging with the genuine complexity of the learning process so that we can be fairly certain we are applying the best model to a particular context. The results of a recent study by a team of researchers in the United States point to the potential usefulness of such an approach.

In their recent paper titled “Seeking the General Explanation,” a team of researchers open the discussion with a critique of the so-called “scientific method.” In their words, “hypothetico-deductive practices of science are diverse and complex. They cannot be reduced to formulaic expressions of ‘the steps of the scientific method,’ which oversimplify the process and reasoning involved. Nevertheless, the tendency toward such over-simplification has been a persistent problem in science education” (p. 59).

The authors thus see the scientific method as a sometimes unhelpful simplification of the reasoning processes involved in doing science. They further point out that scientists also make use of inductive methods of reasoning to complement these hypothetico-deductive practices, but that these are rarely seen in science classes.

One reason why inductive reasoning—or the search for generalizable patterns based on specific examples—may not be prominent in classrooms is that, unlike the steps of the scientific method, teachers tend not to be familiar with simple models of inductive reasoning that may be useful in the classroom.

The researchers propose an approach designed by Francis Bacon, who wrote that the preliminary work of science involves assembling data in three tables. The first table is used to compile information about various phenomena that share a common element of interest. The second should be used to compile data about phenomena that do not exemplify the element of interest but are otherwise similar to the phenomena in table one. The final table should contain instances of phenomena that exhibit the element of interest in different gradations.

Bacon’s method allows for the extraction of general principles through careful observation of instances, non-instances, and gradation across instances. The study discussed here relates to a series of experiments where researchers from three universities in the United States compared Bacon’s inductive method with a more traditional hypothetico-deductive approach to support students in coming to an understanding of concepts in electricity and magnetism.

Experiment 1

With 80 students in an undergraduate engineering program participating, in the first experiment students in both the hypothetico-deductive and inductive conditions used the same computer simulation of three cases: one a positive instance, one a negative instance, and one a particularly strong instance of a change in magnetic field inducing a current. In the hypothetico-deductive condition, students predicted the results for each case in turn, then tested their predictions and explained the results. In the inductive condition based on Bacon’s thinking, students gathered data from all three cases and then tried to formulate a single general explanation that accounted for the data they had collected. The researchers examined the number of students in each condition who were able to identify the deep structure of the concepts and also measured student performance on a posttest involving novel but similar problems—a near transfer task.

The results of this experiment were as follows:

Significantly more students in the inductive condition were able to identify the deep structure, or big idea, underlying the three cases.

Students in the inductive condition did significantly better on the near transfer posttest than students in the hypothetico-deductive condition.

Students who had identified the deep structure during the computer simulation (regardless of condition) performed twice as well on the posttest as those who had not.

Sixty-seven percent of students in the hypothetico-deductive condition relied on surface-level information that was irrelevant to the deep structure of the cases to construct their explanations as opposed to only 29 percent of students in the inductive condition. This may indicate that students in the hypothetico-deductive condition tended to handle each case separately and did not attempt to integrate information across the three cases.

Experiment 2

A second experiment involving 316 students in the same course the following year was designed to investigate whether certain factors in the first experiment may have confounded the results.

These possible confounding factors were as follows:

1. In the inductive condition in experiment one, the worksheet the students used contained the correct outcome of each case in the experiment. This was not provided in the hypothetico-deductive condition, as it would have taken away the possibility of predicting. The researchers were worried that some students in the hypothetic-deductive condition may therefore not have produced the cases accurately enough to get a correct result. Therefore in experiment two, after completing each experiment, the students in the hypothetico-deductive condition also received information about the correct outcome before they proceeded to construct their explanations for each case.

2. Students in the inductive condition in experiment one also received a model of what a good explanation might look like. While the model was for a completely different concept, the researchers were concerned that the lack of provision of a model in the hypothetico-deductive condition may have meant that some students in that condition may not have understood the components of a high-quality explanation. Therefore, in experiment two, students in all conditions were given the model explanation.

The second experiment also introduced two new conditions:

1. A condition where the negative instance was replaced by a weak instance to test Bacon’s thinking that negative instances are crucial for induction. This condition was called the low-contrast inductive condition.

2. A condition where the cases were the same as in the inductive condition, but where students were not specifically instructed to seek a general explanation accounting for the data from all three cases. They were instead asked to explain the similarities and differences between the three cases. This condition was labeled “the compare and contrast condition.” It was included to investigate the importance of specifically prompting students to seek a general explanation.

The results of the second experiment were as follows:

More students in the inductive condition identified the deep structure of the concepts underlying the cases than in the other conditions, though unlike in experiment one, the result was not statistically significant. Students in the inductive condition performed similarly on both experiments. The difference was that the students in the hypothetico-deductive condition performed better in this experiment than in experiment one. The researchers believe that the provision of the model explanation may explain this difference.

Students in the inductive condition performed significantly better on the posttest than students in the hypothetico-deductive condition.

Students who had identified the deep structure during the computer simulation (regardless of condition) performed twice as well on the posttest as those who had not.

The general pattern of results was that students in the inductive condition performed best, followed by those in the compare and contrast condition, then the low-contrast inductive condition. The performance of students in the hypothetico-deductive condition was worst. This pattern held for both identification of the deep structure and for results on the posttest involving near transfer problems.

Students in the hypothetico-deductive condition again tended to focus more on surface features unique to a given case than students in the other conditions.

The fact that performance in the low-contrast condition was worse than in the other inductive conditions may indicate the importance of including negative cases, though this result was not statistically significant.

The fact that performance in the contrasting cases condition was not as good as performance in the inductive condition suggests the importance of specifically prompting students to seek a general explanation.

What might this mean for our classrooms?

The researchers do not claim that their results indicate the superiority of inductive over deductive thinking. Rather they suggest that both kinds of thinking are complementary. They believe that one possible reason that students learned more from the inductive approach was that, due to a lack of deep knowledge grounded in evidence, they were not really in a position to base their predictions on theory when using the steps of the “scientific method.” They believe that while hypothetico-deductive thinking may be useful in certain situations, “inductive synthesis may be especially useful for initial learning where the goal is for students to learn the deep structure of a new phenomenon about which they have few prior beliefs” (p. 62).

Hypothetico-deductive analysis, on the other hand, may work better when students already have a strong theory built up from multiple encounters with a phenomenon and may work particularly well for confronting mistaken beliefs. Both the steps of the scientific method and Bacon’s inductive tables allow us to simplify and structure thinking processes, but the results of this study seem to confirm that having multiple simplified models available and matching them to purpose and context is necessary for effectively supporting learning.

Three simple lessons also seem to emerge from this study. Firstly, if students do not identify the big conceptual idea or deep structure underlying the cases we give them, they are unlikely to be able to transfer their learning. Secondly, learners seem not to seek general explanations unless specifically instructed to do so. Finally, models of “what good looks like” can be incredibly powerful, whatever pedagogical approaches they are combined with.

---

Please fill out the form below if you would like to post a comment on this article:

Nickname

(this will appear with your comments)

Email

Comments

---

Comments

There are currently no comments posted. Please post one via the form above.

MORE FROM

GORDON ELDRIDGE: LESSONS IN LEARNING