How does science work?

• Scientific knowledge is an aggregate of research-based evidence; it is not based on any single source of information.
• Although different scientific disciplines may have different ways of gathering knowledge, in general, the scientific method comprises observation, experimentation, and then analysis of experimental data. This may be followed by reformulating the original hypothesis or idea and sometimes synthesis to formulate natural laws.
• When research papers are published, they are first scrutinised by peers in the discipline. After they are published, they are scrutinised by the broader scientific community and other researchers will try to reproduce, analyse and challenge what is presented. This is an ongoing process and ensures rigour and integrity of the scientific process. It is an essential part of the peer-review process.
• As new knowledge emerges, newer publications supersede older ones and they become the new reference point.
• Advances in scientific knowledge or understanding are usually communicated through research publications in specialised outlets. These are often too detailed, out of the reach of the broader community, or highly specialised, so social media platforms and other forms of public communication are used to share scientific knowledge with non-specialists.

Science is a system of knowledge: knowledge about the physical and natural world, knowledge gained through observation and experimentation, knowledge organised systematically. It is knowledge gained using the scientific method, commonly involving a hypothesis that can be proved or disproved, or a question that can be answered. Science-based knowledge is usually subjected to discussion, debate and further examination and review over time, especially as new information becomes available.

This process of testing, contesting and reviewing is what gives scientists confidence in the state of knowledge at a particular time; it is what they use to explain the physical world. The knowledge that we retain and build on (‘systematised knowledge’ ) can explain phenomena robustly.

The scientific method is often thought of as a straightforward process: form a hypothesis, test or try to disprove the hypothesis through experimentation, and then revise the hypothesis. But this view can discount the role of purely observational research and pattern recognition—so-called ‘discovery science’ —and understate the role of analysis and synthesis of concepts.

Scientists do not often use the word ‘proven’ to describe a current level of understanding. This is reserved for the well-tested laws of nature. Science works on the basis that in many areas there will be always more to know. Even an overwhelming body of evidence may be expanded, or modified, as further work is completed and evidence compiled. That body of evidence usually becomes more complete with more work, but is rarely overturned. This is science at work.

The processes of science

Different scientific disciplines approach the task of gathering knowledge in different ways. For example, an astronomer does not have the same opportunity to experiment that a chemist or physicist might have. A neuroscientist has a different approach to medical knowledge from an epidemiologist. There are, however, three main and mutually compatible approaches to gathering scientific knowledge. These are often combined and most scientists will use all three approaches in their research. The knowledge gained is then tested against established understandings, reviewed and contested, all in order to ensure that the new knowledge is robust.

Observation

Scientific observation involves the close examination of phenomena. Historically, natural philosophers watched, learned and recorded their observations using only their senses, sometimes assisted by simple instruments. Over time, devices that assist observation have become increasingly sophisticated, ranging, for example, from simple magnifying lenses to scanning electron microscopes that can detect and examine objects at finer resolutions than the human eye, through to radio telescopes that can observe space objects well beyond the limit of visible light and well beyond the visible light spectrum.

Observations are no longer limited to human senses. Technology allows us to gather and record data on any number of physical properties. Such technology ranges from the simple and everyday—such as a thermometer or a rain gauge—to the highly advanced—such as the IceCube neutrino observatory in Antarctica, which detects subatomic particles (neutrinos) that barely interact with other matter. Scientific observation might include identifying gene sequences and comparing them across species, or measuring radio bursts from distant stars, lasers or crystallography to identify the structures of molecules, or large scale observatories to identify subatomic particles.

Observations need to be meticulously recorded and reported so that they can be compared across time periods, with or against the observations of others, or against benchmarks and standards. Knowledge is drawn from these comparisons.

Scientific disciplines that rely heavily on observations include astronomy, genetics, taxonomy, anatomy and medical science, and subatomic physics.

Experimentation

Experimentation is the deliberate, procedural testing of the physical world. It can be thought of as extending observations by changing aspects of a system to see what effects those changes have. Experiments are carefully designed to ensure that the conclusions drawn are derived directly from the changes made and the observed results. An experimental system is usually designed to retain as much control of the system as possible, so that deliberate changes are under the control of the experimenter and the resulting observations can be assumed to result from those changes.

Again, the methods and results of experiments must be meticulously recorded and reported. Experiments need to be reproducible by others so that their veracity can be tested and results examined.

In experimental disciplines, knowledge is gained by testing hypotheses and exploring different aspects of a system. As the understanding of the various interactions grows, predictions can be made with greater confidence. Systems can then be harnessed in reproducible, reliable ways. In this way, an experimental system becomes an applied technology, as in a medical or engineering device.

Examples of experimental disciplines include chemistry, biochemistry and molecular biology, agricultural science, physics and medical science.

Analysis

The data gathered and recorded from observational and experimental sciences provide insights beyond their immediate context. Scientists can gather and synthesise data from different sources and conduct analyses on the aggregated dataset. Using the greater statistical power of more massive datasets, we can be more confident of the patterns and relationships that we find within them.

The starting dataset does not necessarily need to be a scientific one. For example, medical records used in hospital administration might reveal patterns of disease prevalence, which could lead to knowledge about how those diseases are caused and transmitted.

Analytical disciplines include statistics, epidemiology, atmospheric science, data science, genomics and proteomics.

Conceptualising and testing

All knowledge gained through scientific processes must be contextualised within the current understanding. Science means testing: testing assumptions, testing knowledge, testing boundaries, testing evidence. Regardless of the approach taken or the methods used, a scientist must maintain an essential scepticism, constantly examining their work to ensure it is robust.

Publishing and communication

Scientists usually publish their work as research papers in specialised journals. These provide a record of a discrete piece of work: a set of observations, a series of experiments or a full analysis.

Academic journals have always had an essential role in the quality control of research. Journals do not publish material without analysis and comment by people skilled in the field of the research to be published, a process known as peer review. Reviewers usually remain publicly anonymous to allow comments to be made without fear of repercussion. Based on advice from the reviewers, a paper can be published or not published, or the author can make changes based on the reviewers’ concerns and resubmit to the journal.

Many scientists also archive their work in data and research repositories, to make it available to other researchers. This is mandatory for much of publicly funded research. Such repositories—including preprint servers, for papers that have been submitted but not peer reviewed—are beyond the scope of this article, but provide other opportunities to disseminate work beyond formal journal publication.

Once a paper is published, it is subject to scrutiny by anyone working in the field. Research results can be integrated into other scientists’ research and may be reproduced, analysed and challenged. It is common to publish research that contradicts (or appears to contradict) another paper. By recording and analysing these differences and discrepancies, scientists can identify flaws and errors that need to be corrected. A paper found to have major flaws will be withdrawn, either by the authors or by the journal, with the reasons being outlined. Withdrawal of a paper is seen as a correction of the scientific record—a necessary and appropriate action in certain circumstances, but it is a rare occurrence. Expert peer review usually exposes flaws and misinterpretations prior to publication.

As knowledge grows, newer papers with additional evidence generally take the place of older papers. Older papers remain in the scientific record so that the lines of reasoning that led to the current knowledge can still be traced and understood.

Any given paper is open to challenge. Peer review provides only a certain minimum quality standard and does not raise the research above criticism. Once published, the paper and the data on which it rests can be scrutinised by the wider international scientific community. There is, however, something of a hierarchy of trust: a peer-reviewed paper is considered more reliable than a description of research which has not been reviewed, and a heavily scrutinised and cited paper may be regarded as more reliable (unless the citations point to flawed work) than a less-cited paper.

Other methods can be used to communicate scientific results. Communication papers are short peer-reviewed summaries of research that generally provide results in advance of a full paper. Review articles collect the results of several papers on a particular subject, providing a detailed summary of the available knowledge while citing the original papers. Monographs (books) do the same on a larger scale, addressing a broader branch of knowledge. And models are shared, synthesised collections of data that describe a particular system in detail.

In some fields, it is normal to release papers on preprint servers ahead of peer review and formal publishing. Papers on preprint servers conform to the standards of their discipline and meet academic publishing and format standards, but are not peer reviewed. Publication on these servers allows early community scrutiny and rapid dissemination of results ahead of publication, and provides a record of active research. Papers on preprint servers are often cited, but are treated with the caution owed to research that has not been peer reviewed.

News articles, essays, videos, magazines, websites, social media and other forms of public communication help communicate scientific knowledge to non-specialists. The Australian Academy of Science produces videos and articles for this purpose. Importantly, these are not research publications and are not afforded the same status as peer reviewed research. However, they can provide valuable opportunities for scientists to share their knowledge widely and to encourage an interest in science.

Scientific knowledge is an aggregate: it is not based on any single publication or work, but rather on continual conversation that publishing represents. A scientist can only communicate what they know and understand; the scientific system ensures that knowledge will continue to expand and mature, and that new knowledge will be created.