If you ask a random scientist, they might come up with something like this: science is a systematic approach to arriving at the truth.
What if this definition is inaccurate – or simply wrong?
What if even core features like reproducibility or falsifiability do not help us move forward?
Is science simply what the majority of scientists intuitively agree it to be?
Here are unexpected challenges in defining what science actually is.
PS from Patrick, the author: I hope it will be a fun and intriguing read for you. It’s a slightly longer article, as it’s meant to take you on a little journey, offering a brief escape from day-to-day stress.
At first, most scientific initiatives do not really aim to find a final truth.
That becomes very obvious if you think about what most of our research is about. It is, for example, about curing cancer or helping patients with Alzheimer’s.
It’s not really about looking for the perfect characterization of a cancer cell. Rather, we are looking for a treatment.
Did you ever read about the average distance a mitochondrion moves in a cell? Probably not. Although research has been conducted on mitochondrial transport (e.g., in neurons) braoder floating didn’t attract much attention since it does not seem to be involved in any research topic worth investigating.
Curiosity-driven basic research can address questions that are not directly applicable. However, the common goal is to understand principles behind evolution, metabolism, or protein folding that can later be applied to other questions through the investigation of a more accessible model.
What we normally don’t notice is that while we search for basic principles, we do not really do it for the sake of truth. Rather, we treat truth as a surrogate for reliability, certainty, or robustness.
For example, we rarely care about the “true response” of proteins, bacteria, or cells. We study the effects of our interventions in laboratory settings to derive an understanding of principles that can be used to better predict and shape the world around us.
A prime example is our research involving cell lines. It has been well established that cancerous cell lines carry a staggering number of mutations, telling us little about the behavior of a “normal” cell1,2. We use fairly simple 2D cultures instead of complex tissue matrices. We apply bovine serum despite it dramatically affecting the response of the cells cultured in it3,4.
In other words, we do not search for truth; we use models that help us assess our hypotheses or reach our goals.
Here is a weird yet instructive thought: if you heard someone say they studied the folding of proteins in Coca-Cola, you would not readily think of that as scientific.
However, Coca-Cola seems more standardized than FBS or Matrigel. Moreover, we could observe and reproduce findings – core pillars of scientific investigation.
Yet we immediately have the feeling that studying proteins in a sugary beverage does not really make sense; it seems useless, deranged, and contingent in its approach. Therefore, it does not feel scientific. However, the true behavior of proteins exists everywhere.
We seek novel insights that have meaning, rather than debating what truth is or whether abstraction from a large body of independent laboratory findings can ever help us truly understand complex real-world systems.
As a result, we measure the impact of our work in citations and recognition. Doing science is inherently connected to funding, but also to our human psyche, which assigns meaning only when there are implications.
But now we might wonder whether we are still talking about science or whether we drift into engineering.
Going back to a long-standing debate5, science is often understood by scientists as a more abstract approach to arriving at the fundamental principles and working mechanisms of the world around us. We do not simply try things out and stop there. We want to understand how they work.
And yet, we often behave differently. Think about one of the greatest scientific discoveries: the discovery of penicillin. It happened by accident. Furthermore, the scientific community did not really care about the finding(s) for almost a decade – it was perceived as having no meaningful implications.
It took many years of hard work from a few scientists and support from major pharmaceutical companies to turn the discovery of penicillin into a major scientific milestone by making it an effective treatment.
However, we could still argue that science is not a thing in itself; it is a methodology. Thus, it is not about penicillin, it is about the kind of work that characterized it.
I.e., it is a certain kind of approach that makes something scientific. We commonly hear features like observability, testability, or reproducibility cited as criteria for defining science.
Still, these also prove insufficient.
Let’s just ask ourselves how many of these features there are. What about quantifiability, generalizability, or predictive power? How many of them have to be met before we call something science? There is no final answer to this question.
Furthermore, when we try to define each of these features individually, we quickly realize how volatile they are.
Is reproducibility meant technically or pragmatically, and to what extent do we allow for variation in results? Overall, we have to admit that the results of a large fraction of scientific work are not properly reproducible.
But even on a theoretical level, these factors seem insufficient.
Testability (or, in Popperian terms, falsifiability) falls short at some point. For example, if someone argued that they were investigating whether thoughts can kill bacteria, we would struggle to call this a scientific endeavor.
By the same token, someone claiming that there is a piece of cheese at the core of the Moon feels ridiculous, but is it because we cannot properly falsify this? We could falsify it through thorough investigation. However, can you ever falsify that protein folding is temperature-dependent?
Maybe we should stop debating about one factor or another, because something needs sufficiently many of these features we listed to be considered scientific.
But then, what counts as sufficiently many?
There was an interesting debate in the 20th century when some “scientists” claimed that the shape of the skull determines intelligence or character traits. Intuitively, this does not sound scientific today.
And for many scientists, it was not then. But for several years there was an ongoing debate about whether this was actually science, not least because parts of the public perceived it as such. Universities were even about to open capacities for these topics5.
Obviously this approach was testable; it was, to some extent, quantifiable, and so forth. In the end, it fell short partially because one couldn’t really establish a proper system and therefore couldn’t reproducibly teach it to other people.
The entire framework could at best work retrospectively when one knew about a person’s character, as the bumps on one’s skull obviously had little to do with character traits. What’s important for us to note is that one couldn’t abstract the practice into a coherent biological system. However, this is not uncommon even for real scientific breakthroughs.
We know that paradigm shifts, before they are accepted, are often approached with skepticism (to say the least).
For us, epigenetics is an established concept6. But to postulate a number of decades ago that overfeeding mice would make the second generation of offspring more vulnerable to metabolic diseases and that this has to do with changes in DNA that are not related to the genetic sequence itself would have seemed more than questionable7. Indeed, very few people would have thought you were a proper geneticist or biochemist.
We judge science based on our current model of the world. Yes, innovative work needs to provide evidence, but more than that, it must be congruent with our established models.
That often happens once so much new information is available that we can arrive at a new coherent system – paradigm shifts don’t occur because we know what is scientific a priori.
We face similar challenges with fields such as neuroimmunology today. Let’s return to the idea that thoughts might kill bacteria. There are peer-reviewed case studies in which people, simply through meditative practices, become more resistant to streptococcal infections.
Scientific evidence? The key issue is how we connect these findings to our current understanding, not what one investigation or another finds.
Let’s therefore take this perspective: through our thoughts we can affect the type and level of neurotransmitters as well as hormones that are released. In turn, this affects immune cells.
There is a vast amount of literature showing that neurotransmitters regulate immune cells8,9, and that cytokines produced by immune cells can affect neuronal activity10 and even cause different mental states11. Even more obvious is that we salivate when we see food, and the lysozyme in our saliva can kill bacteria (even though these are rarely problematic pathogens).
Doesn’t it feel much more scientific to say: this investigation focuses on how thought patterns might affect the viability of pathogens through neuroimmune interactions?
Investigations can be confounded, and as famously noted by quantum physicists, all too often we can only see what we know we have to look for.
Of course, by conducting several studies and investigating problems from different angles, we are able to arrive at intersubjective concepts and mechanisms of the world. But what gets us there is not always considered scientific from the start.
Thoughts or character traits often feel unscientific because we generally associate science with the investigation of the world around us.
Introspective approaches are normally not considered scientific because they are too subjective and therefore difficult to quantify intersubjectively. It remains an ongoing debate whether psychology (or which branches of it) is science, or whether they would even benefit from what we call a scientific approach.
In that sense, any scientifically derived evidence must be obtained in a way that allows other people to arrive at the same conclusions.
However, could we then consider mathematics to be science? Or do we count it as logic? Without going too deeply into other philosophical debates about linguistics, we can extend this question to theoretical physics.
Here again, we see the struggle with “truth.” What we are doing as scientists is often neither final nor systematic. We do not test across a large range of concentrations, crossing with different time points and models just to find the “true effect.”
For us, it is much more about relevant novelty, even though that leaves many gaps in our understanding.
Asking whether mathematics is science, what truth is, what truly constitutes proof, or what constitutes true causation is perceived as a philosophical debate.
So science is not fully applied, but also not merely theoretical. It’s about principles – but somehow what we might call relevant principles, because we would think of somebody who does something that obviously has no merit as “deranged.”
And still, we see with penicillin and other breakthroughs that doing what seems useless can be an important, maybe even necessary, activity.
In other words, innovative thinking that is not accepted by the larger scientific community can still prove to be groundbreaking. (No doubt, we shouldn’t forget that it is also frequently the case that such ideas are eventually proven to be mistaken.)
It then seems that what most scientists are really doing on a daily basis is research?
We observe the world around us and sometimes discover new things by accident. Then we investigate them using a set of standardized approaches. If we find that they fit into our current system of knowledge, we publish them, and when enough other scientists agree, we consider them scientifically proven.
However, it is very important to note that there is no clear definition of what a researcher is either.
There is a set of instruments and methods that a typical biochemist, chemist, or experimental physicist uses.
There are also common practices that we consider scientific – for example, scientific publishing. Somebody presenting their data in a LinkedIn post would hardly be considered proper scientific practice, although this has nothing to do with how true the claim might eventually turn out to be.
Our understanding is context-dependent: the idea that thoughts can effectively kill bacteria doesn’t feel scientific, whereas the idea that the activity of neurons can lead to the release of molecules that have antibacterial effects very much does.
We don’t have a way to prove or determine what science is, thus it often becomes what we culturally determine it to be.
Nevertheless, let’s give it another try to define science:
Science is a standardized approach to the generation of knowledge.
Knowledge here means something that we consider reliable and intelligible within our current understanding of the world, and that therefore yields predictive power.
However, this approach has to be standardized. It must fit what we commonly assume to be technically reproducible and therefore intersubjectively accessible.
Additionally, it must be coherent, creating an intelligible system of insights with some predictive power under determined circumstances.
Importantly, we identify science intuitively. And what we consider “best practices” can therefore be misleading.
The assumption of independence among scientists is very much idealistic, because it is hard to claim that when investigating whether a drug works for cancer treatment, having no preference for the outcome is natural.
Of course, we use methods such as statistics in order to deal with our human biases. But in the end, science is not the objective estimation of truth. We do it to create and assess insights about the world around us.
This is also why this article is hardly considered scientific. It is difficult to fit it into a mental model that aligns with what other scientists have traditionally done.
Yet, there is an important takeaway here.
There will always be a certain kind of tension between what we call science and what we consider engineering, psychology, medicine, or spirituality.
However, it is up to us to refine and protect what we call science – even if we cannot fully put it into words.
Yes, science itself is dependent upon beliefs and assumptions because we are humans. But it is up to us to maintain values such as reproducibility, independence, and rigor as an approach to investigating the world around us.
But at the same time, this should also remind us to keep an open mind toward innovation and new ideas.
Because even though they might not seem scientific at first, they might eventually prove to revolutionize our understanding of the world.
Science – Scientific Method – Research; nobody truly knows exactly what these are, yet it is they that unite us around so many marvelous insights.
I hope you enjoyed this read. No doubt there is so much more to mention and discuss! Therefore, leave your thoughts in the comments below!
By Patrick Penndorf – courtesy of the IUBMB
References
- Kranjec, C., Eilertsen, I.A., Nunes, L., et al., 2026. Common gene mutations in 103 authenticated colorectal cancer cell lines. Oncogenesis, 15, 8. doi:10.1038/s41389-026-00599-0.
- Leroy, B., Girard, L., Hollestelle, A., et al., 2014. Analysis of TP53 mutation status in human cancer cell lines: a reassessment. Human Mutation, 35, 756–765. doi:10.1002/humu.22556.
- Warren, H.S., Fitting, C., Hoff, E., et al., 2010. Resilience to bacterial infection: difference between species could be due to proteins in serum. Journal of Infectious Diseases, 201, 223–232. doi:10.1086/649557.
- Liu, S., Yang, W., Li, Y., et al., 2023. Fetal bovine serum, an important factor affecting the reproducibility of cell experiments. Scientific Reports, 13, 1942. doi:10.1038/s41598-023-29060-7.
- Gieryn, T.F., 1983. Boundary-work and the demarcation of science from non-science: strains and interests in professional ideologies of scientists. American Sociological Review, 48, 781–795. doi:10.2307/2095325.
- Felsenfeld, G., 2014. A brief history of epigenetics. Cold Spring Harbor Perspectives in Biology, 6, a018200. doi:10.1101/cshperspect.a018200.
- Noble, D., 2015. Conrad Waddington and the origin of epigenetics. Journal of Experimental Biology, 218, 816–818. doi:10.1242/jeb.120071.
- Hodo, T.W., de Aquino, M.T.P., Shimamoto, A., et al., 2020. Critical neurotransmitters in the neuroimmune network. Frontiers in Immunology, 11, 1869. doi:10.3389/fimmu.2020.01869.
- Wood, L.B., Singer, A.C., 2025. Neurons as immunomodulators: from rapid neural activity to prolonged regulation of cytokines and microglia. Annual Review of Biomedical Engineering, 27, 55–72. doi:10.1146/annurev-bioeng-110122-120158.
- Zipp, F., Bittner, S., Schafer, D.P., 2023. Cytokines as emerging regulators of central nervous system synapses. Immunity, 56, 914–925. doi:10.1016/j.immuni.2023.04.011.
- Bennett, F.C., Molofsky, A.V., 2019. The immune system and psychiatric disease: a basic science perspective. Clinical and Experimental Immunology, 197, 294–307. doi:10.1111/cei.13334.