In this blog post, we will examine the limitations of existing criteria for distinguishing science, focusing on the perspectives of Popper and Kuhn, and explore the need for a new approach to define science more accurately.
The question of what criteria distinguish science from non-science—known as the “demarcation problem”—has been a constant source of contemplation for philosophers for thousands of years and remains a subject of lively debate today. In fact, scholars generally agree on which fields are scientific and which are non-scientific. For example, physics, chemistry, and biology tend to be classified as science, while astrology, the humanities, religion, and alternative medicine tend to be classified as non-science. However, no widely accepted, clear answer has yet been provided as to why these classifications inherently possess different natures.
In this blog post, I will examine the arguments of philosophers Karl Popper (1902–1994) and Thomas Kuhn (1922–1996), who addressed the demarcation problem, and aim to propose a new criterion for demarcation by critically revising and integrating their theories. In particular, I argue that Popper’s concept of “falsifiability” needs to be combined with Kuhn’s concept of “paradigm,” while conversely, the element of “not testing the paradigm” among Kuhn’s criteria is unnecessary for defining the concept of ideal science. Based on this, I will propose a new definition—“the activity of constructively studying falsifiable theories under a paradigm”—as a solution to the demarcation problem. Furthermore, I will demonstrate that this criterion harmoniously combines compatible claims and can effectively distinguish between science and non-science in practice.
Popper presented a simple yet powerful criterion regarding the demarcation problem: namely, the claim that only falsifiable theories are science (Popper, 1963:36). At the time, science was widely understood as a process of “proving” true propositions through inductive reasoning that generalized observed results. For example, this is the method of observing a thousand sheep and concluding that “all sheep are white” because they were all white. However, this conclusion is not logically valid, as we cannot rule out the possibility that black sheep exist in an unknown region. On the other hand, it is clear that the proposition “all sheep are white” is falsified if even a single black sheep is found.
To avoid the problem of induction, Popper argued that the core of science lies not in inductive proof but in “falsification through counterexamples.” Science, he posited, is a process of proposing falsifiable theories, testing how well they hold up through repeated experiments, discarding them if falsified, and then testing new theories. He viewed the role of the scientist as testing theories under the harshest possible conditions.
However, while it may be valid to establish falsifiability as a necessary condition for science, extending it to a sufficient condition is problematic. Popper cited astrology and psychology as examples of non-science, arguing that they could be classified as pseudoscience according to his criteria. However, these fields were regarded as non-science not because of unfalsifiable claims, but because they showed no progress despite repeated failures of prediction (Hansson, 2008). According to Popper’s criteria, any discipline that presents falsifiable predictions should be recognized as science; thus, his demarcation criterion fails at this point.
That said, there is no need to completely dismiss Popper’s position. The reason Popper’s approach failed is that his criteria focused excessively on the “logical form of a theory” while failing to consider the active aspects of science—namely, how theories are actually investigated and developed. For example, consider the proposition in astrology that “if the Sun and Jupiter form a 90° angle, those born on May 7 will fall ill.” Viewed in isolation, this proposition is falsifiable, and someone unfamiliar with astronomy or biology might even feel it is worth scientific verification. Indeed, if there were a discipline called “ideal astrology” in which such propositions were systematically analyzed under the methodology of modern science, it might appear to be a genuine scientific activity. Nevertheless, actual astrology is not recognized as science because, despite presenting numerous falsifiable predictions, it fails to evolve. This case demonstrates that, contrary to Popper’s expectations, science cannot be defined by simple logical structures alone, and it highlights the need to consider how researchers advance the field after a theory has been falsified.
It was precisely in this context that Kuhn proposed a new criterion. In *The Structure of Scientific Revolutions* (1962), Kuhn distinguished between “normal science”—conducted under a dominant paradigm—and “scientific revolutions,” which occur when fatal flaws accumulate within the paradigm. According to Kuhn, what functions as the core of the demarcation problem is precisely “the activity of normal science that constructively develops existing theories based on the paradigm.” Kuhn viewed the fact that researchers in normal science do not test the fundamental assumptions of the paradigm as a key characteristic. This contrasts with Popper’s position, which holds that all theories must be rigorously tested and boldly discarded if falsified.
Kuhn’s argument is significant in that it includes “research activities following the occurrence of results”—that is, observations of how the scientific field actually develops—which Popper’s approach lacked. Furthermore, his focus on normal science rather than scientific revolutions is also valid. According to Kuhn, revolutions are incidental phenomena that occur when problems accumulated during the maintenance of a paradigm reach an extreme level, and because most scientific achievements are made during the phase of normal science. Conversely, it is difficult to call a situation “science” where new theories are announced daily but there is no interaction between them, and each theory makes no progress on its own. Science must necessarily have a dominant paradigm, and productive and constructive research must continue within it.
However, questions have been raised as to whether one of Kuhn’s criteria—“even when errors are discovered, the basic assumptions of the paradigm are not tested”—is truly a necessary condition for defining the demarcation problem. According to Kuhn’s explanation, the reason errors do not immediately trigger a revolution is that many errors can ultimately be resolved within the paradigm, and researchers do not test the paradigm itself unless the situation is extremely serious. However, this merely explains the situation in which real-world scientists are constrained by factors such as manpower, time, and funding; it does not explain how an ideal science should be structured.
Let us imagine an ideal civilization. If scientists could strictly follow the methodology of modern science without constraints on manpower, time, or funding, whenever an error was discovered, some researchers could immediately move to a team dedicated to testing the paradigm itself. For example, if the first error—the retrograde motion of planets—had been discovered during the era when the geocentric paradigm was dominant, some researchers would have attempted to resolve the error within the existing paradigm by adding new parameters (the normal science team), while others would have questioned the geocentric paradigm itself and conducted research to test it (the paradigm-testing team). It is impossible to judge which approach is correct at the outset, but if the discrepancy can be resolved within the paradigm, research in normal science will yield productive results; if the problem lies with the paradigm itself, the testing team will become the catalyst for revolution. This structure can drive revolution and paradigm shifts far more efficiently than Kuhn’s model, while simultaneously ensuring that research in normal science continues uninterrupted. In this case as well, the research activity is unquestionably science in the eyes of anyone.
Therefore, while the condition that the basic assumptions of the paradigm are not tested may be useful for describing real-world science, it is not necessary as a criterion for defining the essence of ideal science. If this condition is removed, Kuhn’s criterion can be summarized as “constructive research conducted within the paradigm.”
Combining Popper’s principle of falsifiability with Kuhn’s modified criterion leads to the definition of science as “the activity of constructively researching falsifiable theories within a paradigm.” This is the new criterion for demarcation proposed in this essay. Popper’s criterion is adopted as a logical tool for testing theories at the propositional level, while Kuhn’s criterion guides the productive development of those propositions within a single academic system.
For this criterion to hold, Popper’s and Kuhn’s modified criteria must first be compatible. While the central issue for both scholars was “to what extent should a theory be rigorously tested,” this definition removes that condition, allowing the two criteria to coexist without contradiction. If human and financial resources were infinite, the rigorous testing Popper speaks of could be carried out; however, as Kuhn points out, revolutions have occurred and science has progressed even without fundamentally questioning the paradigm. This fact demonstrates that there is no need to include “rigorous testing” as a condition in the definition of science itself. Rather, this is an issue to be addressed when discussing the practical methodology of science.
One might ask whether falsifiability is still necessary. The conclusion is “yes.” Even a theoretical system that is unfalsifiable can allow for constructive research within a paradigm. For example, let us imagine a scientific system in which every proposition is transformed into a probabilistic statement, such as “Newton’s laws are more likely to hold as mass increases.” While research would be possible using statistical techniques and probability theory, it would be difficult to call this science because the constituent propositions do not allow for falsification. In reality, since unfalsifiable systems—such as religious dogma—generate almost no constructive research, falsifiability appears to be an implicit prerequisite. However, there is no need to directly apply real-world observations when addressing the demarcation problem. Especially during the stage of paradigm shifts, the “harsh test” advocated by Popper must be carried out to ensure that the shift leads to actual progress in knowledge rather than merely a change in public opinion regarding pseudoscience. Therefore, if a scientific system has adopted a paradigm, falsifiability plays a central role in the process of maintaining or shifting that paradigm.
Finally, let us examine whether the criteria proposed in this essay correctly distinguish between science and pseudoscience. As discussed earlier, all science satisfies these criteria. Therefore, it suffices to show that no pseudoscience exists that satisfies these criteria. Taking astrology—a common example of pseudoscience—as an example, Popper argued that astrology is non-science because it is unfalsifiable, while Kuhn claimed it is non-science because, although falsifiable, it lacks constructive research. Both criticized the other for misinterpreting astrology. However, since this criterion includes both falsifiability and constructive research, astrology is clearly classified as non-science regardless of which interpretation is adopted. Therefore, this criterion functions as a finer and more precise sieve for distinguishing science than the criteria proposed by the two aforementioned philosophers.
In conclusion, this essay critically analyzes Popper and Kuhn’s existing arguments regarding the demarcation problem, adopts the core elements of both, and presents a new criterion. Furthermore, since the dispute between the two philosophers regarding the intensity of testing has been resolved, the combination of the two criteria has been shown to be logically valid. The answer to the question “What is science?” is an important criterion that everyone must possess to make sound decisions in today’s information society, which is overflowing with vast amounts of information and unverified knowledge—not just scientists seeking truth. Moreover, the key clues to resolving the demarcation problem are already inherent in the differing arguments of these two philosophers. Only by integrating the strengths of these two positions and resolving their points of contention can we truly come closer to the essence of science.
