The “scientific method” is traditionally presented in the first chapter of science textbooks as a simple, linear, five- or six-step procedure for performing scientific investigations. Although the Scientific Method captures the core logic of science (testing ideas with evidence), it misrepresents many other aspects of the true process of science — the dynamic, nonlinear, and creative ways in which science is actually done. In fact, the Scientific Method more accurately describes how science is summarized after the fact — in textbooks and journal articles — than how scientific research is actually performed. Teachers may ask that students use the format of the scientific method to write up the results of their investigations (e.g., by reporting their question, background information, hypothesis, study design, data analysis, and conclusion), even though the process that students went through in their investigations may have involved many iterations of questioning, background research, data collection, and data analysis and even though the students’ “conclusions” will always be tentative ones. To learn more about how science really works and to see a more accurate representation of this process, visit The real process of science.

Scientists often seem tentative about their explanations because they are aware that those explanations could change if new evidence or perspectives come to light. When scientists write about their ideas in journal articles, they are expected to carefully analyze the evidence for and against their ideas and to be explicit about alternative explanations for what they are observing. Because they are trained to do this for their scientific writing, scientist often do the same thing when talking to the press or a broader audience about their ideas. Unfortunately, this means that they are sometimes misinterpreted as being wishy-washy or unsure of their ideas. Even worse, ideas supported by masses of evidence are sometimes discounted by the public or the press because scientists talk about those ideas in tentative terms. It’s important for the public to recognize that, while provisionality is a fundamental characteristic of scientific knowledge, scientific ideas supported by evidence are trustworthy. To learn more about provisionality in science, visit our page describing how science builds knowledge. To learn more about how this provisionality can be misinterpreted, visit a section of the Science toolkit.

Peer review helps assure the quality of published scientific work: that the authors haven’t ignored key ideas or lines of evidence, that the study was fairly-designed, that the authors were objective in their assessment of their results, etc. This means that even if you are unfamiliar with the research presented in a particular peer-reviewed study, you can trust it to meet certain standards of scientific quality. This also saves scientists time in keeping up-to-date with advances in their fields by weeding out untrustworthy studies. Peer-reviewed work isn’t necessarily correct or conclusive, but it does meet the standards of science. To learn more, visit Scrutinizing science.

In an experiment, the independent variables are the factors that the experimenter manipulates. The dependent variable is the outcome of interest—the outcome that depends on the experimental set-up. Experiments are set-up to learn more about how the independent variable does or does not affect the dependent variable. So, for example, if you were testing a new drug to treat Alzheimer’s disease, the independent variable might be whether or not the patient received the new drug, and the dependent variable might be how well participants perform on memory tests. On the other hand, to study how the temperature, volume, and pressure of a gas are related, you might set up an experiment in which you change the volume of a gas, while keeping the temperature constant, and see how this affects the gas’s pressure. In this case, the independent variable is the gas’s volume, and the dependent variable is the pressure of the gas. The temperature of the gas is a controlled variable. To learn more about experimental design, visit Fair tests: A do-it-yourself guide.

In scientific testing, a control group is a group of individuals or cases that is treated in the same way as the experimental group, but that is not exposed to the experimental treatment or factor. Results from the experimental group and control group can be compared. If the control group is treated very similarly to the experimental group, it increases our confidence that any difference in outcome is caused by the presence of the experimental treatment in the experimental group. For an example, visit our side trip Fair tests in the field of medicine.

A negative control group is a control group that is not exposed to the experimental treatment or to any other treatment that is expected to have an effect. A positive control group is a control group that is not exposed to the experimental treatment but that is exposed to some other treatment that is known to produce the expected effect. These sorts of controls are particularly useful for validating the experimental procedure. For example, imagine that you wanted to know if some lettuce carried bacteria. You set up an experiment in which you wipe lettuce leaves with a swab, wipe the swab on a bacterial growth plate, incubate the plate, and see what grows on the plate. As a negative control, you might just wipe a sterile swab on the growth plate. You would not expect to see any bacterial growth on this plate, and if you do, it is an indication that your swabs, plates, or incubator are contaminated with bacteria that could interfere with the results of the experiment. As a positive control, you might swab an existing colony of bacteria and wipe it on the growth plate. In this case, you would expect to see bacterial growth on the plate, and if you do not, it is an indication that something in your experimental set-up is preventing the growth of bacteria. Perhaps the growth plates contain an antibiotic or the incubator is set to too high a temperature. If either the positive or negative control does not produce the expected result, it indicates that the investigator should reconsider his or her experimental procedure. To learn more about experimental design, visit Fair tests: A do-it-yourself guide.

In a correlational study, a scientist looks for associations between variables (e.g., are people who eat lots of vegetables less likely to suffer heart attacks than others?) without manipulating any variables (e.g., without asking a group of people to eat more or fewer vegetables than they usually would). In a correlational study, researchers may be interested in any sort of statistical association — a positive relationship among variables, a negative relationship among variables, or a more complex one. Correlational studies are used in many fields (e.g., ecology, epidemiology, astronomy, etc.), but the term is frequently associated with psychology. Correlational studies are often discussed in contrast to experimental studies. In experimental studies, researchers do manipulate a variable (e.g., by asking one group of people to eat more vegetables and asking a second group of people to eat as they usually do) and investigate the effect of that change. If an experimental study is well-designed, it can tell a researcher more about the cause of an association than a correlational study of the same system can. Despite this difference, correlational studies still generate important lines of evidence for testing ideas and often serve as the inspiration for new hypotheses. Both types of study are very important in science and rely on the same logic to relate evidence to ideas. To learn more about the basic logic of scientific arguments, visit The core of science.

Deductive reasoning involves logically extrapolating from a set of premises or hypotheses. You can think of this as logical “if-then” reasoning. For example, IF an asteroid strikes Earth, and IF iridium is more prevalent in asteroids than in Earth’s crust, and IF nothing else happens to the asteroid iridium afterwards, THEN there will be a spike in iridium levels at Earth’s surface. The THEN statement is the logical consequence of the IF statements. Another case of deductive reasoning involves reasoning from a general premise or hypothesis to a specific instance. For example, based on the idea that all living things are built from cells, we might deduce that a jellyfish (a specific example of a living thing) has cells. Inductive reasoning, on the other hand, involves making a generalization based on many individual observations. For example, a scientist who samples rock layers from the Cretaceous-Tertiary (KT) boundary in many different places all over the world and always observes a spike in iridium may induce that all KT boundary layers display an iridium spike. The logical leap from many individual observations to one all-inclusive statement isn’t always warranted. For example, it’s possible that, somewhere in the world, there is a KT boundary layer without the iridium spike. Nevertheless, many individual observations often make a strong case for a more general pattern. Deductive, inductive, and other modes of reasoning are all useful in science. It’s more important to understand the logic behind these different ways of reasoning than to worry about what they are called.

Scientific theories are broad explanations for a wide range of phenomena, whereas hypotheses are proposed explanations for a fairly narrow set of phenomena. The difference between the two is largely one of breadth. Theories have broader explanatory power than hypotheses do and often integrate and generalize many hypotheses. To be accepted by the scientific community, both theories and hypotheses must be supported by many different lines of evidence. However, both theories and hypotheses may be modified or overturned if warranted by new evidence and perspectives.

A null hypothesis is usually a statement asserting that there is no difference or no association between variables. The null hypothesis is a tool that makes it possible to use certain statistical tests to figure out if another hypothesis of interest is likely to be accurate or not. For example, if you were testing the idea that sugar makes kids hyperactive, your null hypothesis might be that there is no difference in the amount of time that kids previously given a sugary drink and kids previously given a sugar-substitute drink are able to sit still. After making your observations, you would then perform a statistical test to determine whether or not there is a significant difference between the two groups of kids in time spent sitting still.

Ockham’s razor is an idea with a long philosophical history. Today, the term is frequently used to refer to the principle of parsimony — that, when two explanations fit the observations equally well, a simpler explanation should be preferred over a more convoluted and complex explanation. Stated another way, Ockham’s razor suggests that, all else being equal, a straightforward explanation should be preferred over an explanation requiring more assumptions and sub-hypotheses. Visit Competing ideas: Other considerations to read more about parsimony.

Rigorous and well controlled scientific investigations¹ have examined these topics and have found no evidence supporting their usual interpretations as natural phenomena (i.e., ghosts as apparitions of the dead, ESP as the ability to read minds, and astrology as the influence of celestial bodies on human personalities and affairs) — although, of course, different people interpret these topics in different ways. Science can investigate such phenomena and explanations only if they are thought to be part of the natural world. To learn more about the differences between science and astrology, visit Astrology: Is it scientific? To learn more about the natural world and the sorts of questions and phenomena that science can investigate, visit What’s natural? To learn more about how science approaches the topic of ESP, visit ESP: What can science say?

Knowledge generated by science has had many effects that most would classify as positive (e.g., allowing humans to treat disease or communicate instantly with people half way around the world); it also has had some effects that are often considered negative (e.g., allowing humans to build nuclear weapons or pollute the environment with industrial processes). However, it’s important to remember that the process of science and scientific knowledge are distinct from the uses to which people put that knowledge. For example, through the process of science, we have learned a lot about deadly pathogens. That knowledge might be used to develop new medications for protecting people from those pathogens (which most would consider a positive outcome), or it might be used to build biological weapons (which many would consider a negative outcome). And sometimes, the same application of scientific knowledge can have effects that would be considered both positive and negative. For example, research in the first half of the 20th century allowed chemists to create pesticides and synthetic fertilizers. Supporters argue that the spread of these technologies prevented widespread famine. However, others argue that these technologies did more harm than good to global food security. Scientific knowledge itself is neither good nor bad; however, people can choose to use that knowledge in ways that have either positive or negative effects. Furthermore, different people may make different judgments about whether the overall impact of a particular piece of scientific knowledge is positive or negative. To learn more about the applications of scientific knowledge, visit What has science done for you lately?