Science: Experiments

This is Unit Twenty of ACT Preparation: experiments. Unit nineteen read data; this unit asks how the data were produced, and whether the design lets them mean what a study claims. One habit governs it: name what was changed, what was measured, and what was held the same, before reading any result. By the end of this part you will be able to tell a hypothesis from a prediction. To say what results can and cannot do. To identify a study's variables. To find its control group. And to recognize a fair test and a confounding variable.

Five entries on the scientific method, from question to hypothesis to what results can show.

The method as a cycle.

ACT 1901 The Scientific Method

The scientific method runs from observation and question to hypothesis, prediction, experiment, analysis and conclusion. Wilted seedlings by a window raise the question of whether heat or light is responsible. When results fail to support a hypothesis, the cycle begins again with a new one.

The hypothesis.

ACT 1902 Hypothesis

A hypothesis is a suggested explanation that can be tested, and it must be falsifiable: some possible result would show it wrong. "The seedlings wilt because of heat from the window" can be tested by cooling the window. An explanation no result could contradict is not scientific.

The prediction.

ACT 1903 Prediction

A prediction states what should happen if a hypothesis is true, often as if-then: if heat causes the wilting, seedlings shaded from heat but given full light should not wilt. A hypothesis explains; a prediction forecasts a measurable result of it.

What results can do.

ACT 1904 Science Tests, It Does Not Prove

Experiments can support or disprove a hypothesis, but science does not claim to prove anything final; its understandings stay open to new evidence. Ten fitting trials support a hypothesis; one well-run contradiction calls it into question. An answer saying an experiment proved a hypothesis overstates it.

Which hypothesis was tested, step by step.

ACT 1905 Which Hypothesis Was Tested?

Find what the researchers changed on purpose, find what they measured, and choose the hypothesis linking the two. In an invented study, beans in identical pots grow under red, blue and white light, and stem height is measured: the hypothesis tested is that light color affects growth. Water was the same in every pot, so a hypothesis about water was not tested. A factor held constant tells you nothing about its own effect.

Four entries on the variables of an experiment.

Independent and dependent.

ACT 1906 Independent and Dependent Variables

The independent variable is the factor changed on purpose; the dependent variable is the result measured in response. Statistics books call them explanatory and response variables. In the bean study, color is independent and height dependent. The independent is set before a trial; the dependent is read after.

Treatments and units.

ACT 1907 Treatments and Experimental Units

Treatments are the different values of the independent variable an experiment applies; an experimental unit is a single object or individual receiving one. Red, blue and white light are three treatments; each pot is a unit. One pot measured five times is still one unit.

What stays the same.

ACT 1908 Controlled Variables

Controlled variables are the factors kept the same across all trials, so that only the independent variable differs. In the bean study, pot size, soil, water and temperature are controlled. A factor left free to vary becomes a second possible cause.

Identifying the variables, step by step.

ACT 1909 Identifying the Variables in a Study

Find what was deliberately set, the independent variable and its treatments; what was recorded, the dependent variable; and what stayed the same, the controlled variables. In an invented study, the same ball is dropped from four heights onto the same floor, three times each, and the bounce measured: drop height is independent with four treatments, bounce height dependent, and the ball, floor and measuring method controlled. The repeated drops are not a variable; they make the measurement more reliable.

Eight entries on control groups and fair tests.

The control group.

ACT 1910 The Control Group

A control group matches the experimental group in every feature except the treatment being tested, showing what happens without it. Beans under ordinary white light serve as the control for beans under colored light. Without a control, there is nothing to compare against.

Two names that sound alike.

ACT 1911 Control Group Versus Controlled Variables

The control group is a comparison group receiving no treatment or a standard one; controlled variables are conditions held constant for every group. The white-light beans are the control group; equal water for all pots is a controlled variable. A question about what was held constant asks for controlled variables.

Finding the control, step by step.

ACT 1912 Finding the Control

List the groups and what each received, find the one with no treatment or the standard condition, and check it differs only in the treatment. In an invented fertilizer study, plot A gets fertilizer, B twice as much, and C none, with the same soil, sun and water: C is the control. Had C also been shaded, it would differ in two ways. Without C, plots A and B could show which dose did better, but not whether fertilizer helps at all.

One change at a time.

ACT 1913 One Change at a Time

A fair test changes one factor at a time; if two change together, a difference in the result cannot be assigned to either. Moving beans to blue light and a warmer room at once leaves the cause unknown. To test a second factor, run a separate trial.

Random assignment.

ACT 1914 Random Assignment

Random assignment places subjects into groups by chance, so other differences spread evenly and only the treatment differs systematically, as when names are drawn from a hat. Letting volunteers choose their own group lets their differences, such as motivation, travel with the treatment.

Confounding variables.

ACT 1915 Confounding Variables

A confounding variable is a factor, other than the independent variable, that affects the result and differs between groups, so it can be mistaken for the cause; statistics books also call it a lurking variable. Fertilized plots that also got more sun have sun as a confound. Random assignment and controlled variables are the defenses.

Spotting a confound, step by step.

ACT 1916 Spotting a Confound

State the comparison, look for any other difference between the groups, and ask whether it could affect the result. In an invented study, students who chose an online review course scored higher than those who did not, but those who chose it may have been more motivated, and motivation could raise scores on its own. Random assignment would have spread motivation across both groups. When subjects choose their own treatment, suspect self-selection first.

The placebo.

ACT 1917 The Placebo

A placebo is an inactive treatment, such as a sugar pill, given to the control group so both groups expect an effect: half take a vitamin, half an identical pill without one. People on a placebo can still improve slightly, the placebo effect, which is why the comparison group needs one.

Look at what you can now do. You can trace the scientific method and tell a testable hypothesis from the prediction it makes. You can say that results support or contradict a hypothesis, never prove it, and name the hypothesis a study tested. You can identify the independent, dependent and controlled variables. You can find the control group and keep it apart from the controlled variables. And you can recognize a fair test, random assignment, a confound and a placebo. Part two adds blinding, study types, and reading experiments together.