Scientific results are said to converge when they all support the same hypothesis, having been derived from different types of empirical methods (Matiasz et al., 2018; Munafò & Smith, 2018). For instance, the hypothesis that meditation, practiced correctly, improves one’s attention might be supported by both (1) an observational study that finds an increased likelihood of meditation practice in people with extraordinary attention skills1 and (2) an interventional study that shows an improvement in attention skills in people who are assigned a meditation class with a trained instructor. These different types of studies — one observational, the other interventional — support the same hypothesis and can thus be said to provide convergent evidence. As another example, we might obtain convergent evidence by combining data from both EEG and fMRI, which provide different perspectives on brain activity.
Evidential convergence is thus distinct from evidential consistency. Scientific results are said to be consistent when they all support the same hypothesis, having been derived from the same empirical methods. Scientists derive consistent evidence by showing that a particular finding is replicable across multiple iterations of the same type of study.
Although consistency, or replicability, is critical in science, we achieve a more holistic view of evidence by considering both consistency and convergence. And we achieve a more realistic view of evidence by acknowledging that the two concepts are not equally powerful.
Epistemologically, convergence is stronger than consistency. This fact is known intuitively by scientists and the general public alike. Consider, for instance, the following two hypothetical scenarios that a juror could encounter when serving on a trial:
- Consistency: A single witness delivers the same testimony three times, on three different days.
- Convergence: Three independent witnesses deliver testimony that is consistent with that of the other witnesses, without any knowledge of what the other witnesses said.
The consistent testimony of the single witness may be compelling, as the ability to stick to one story will usually boost the jurors’ confidence. But in most cases, the jurors will be far more persuaded by the convergent testimony of three independent witnesses.
Lines of evidence within science are similar: Consistent scientific results are persuasive, even if each successful replication is less surprising and thus less informative than the last. But convergent results are even more persuasive, as each of the different methods can lend complementary strengths that compensate for weaknesses in the others. As each additional method provides convergent evidence in support of a hypothesis, it becomes less likely that the results are due to experimental artifacts common to every method employed.
1. Yes, correlation is not causation; however, we must also keep in mind that where there is causation, we are guaranteed to find specific correlations. It’s therefore instructive to identify correlations via observational studies; this approach can confirm correlations that are consistent with a causal hypothesis without having to perform an intervention, which may introduce spurious correlations through experimental artifacts, despite our best efforts to control for confounders. ↑