Scientific Foundations: Research Methods and Statistics
Master AP Psychology research methods: experiments vs correlation vs descriptive designs, variables, confounds, correlation vs causation, statistics, and ethics.
What you'll do in this lesson
A voice-first session with the Crimsora tutor on Scientific Foundations: Research Methods and Statistics, then targeted practice and FRQs — with the tutor adapting to where you get stuck.
What this lesson covers
Every claim in psychology rests on evidence, and the AP exam wants to know whether you can tell strong evidence from weak. Can a study prove that caffeine causes better memory, or only that the two are related? This lesson gives you the toolkit psychologists use to answer questions carefully: choosing a research method, defining variables precisely, controlling for confounds, and reading the numbers that summarize results.
By the end you will be able to distinguish experimental, correlational, and descriptive research; spot independent and dependent variables; explain why only experiments show causation; interpret means, medians, standard deviations, and correlation coefficients; understand what statistical significance means; and apply the ethical rules that protect participants. These skills appear in multiple-choice items and in the free-response Article Analysis and Evidence-Based questions.
By the end you will be able to distinguish experimental, correlational, and descriptive research; spot independent and dependent variables; explain why only experiments show causation; interpret means, medians, standard deviations, and correlation coefficients; understand what statistical significance means; and apply the ethical rules that protect participants. These skills appear in multiple-choice items and in the free-response Article Analysis and Evidence-Based questions.
Three Research Methods and What Each Can Prove
Psychologists pick a method based on the question they can ask and the conclusion they are allowed to draw.
Descriptive research observes and describes without manipulating anything. It includes case studies (one person or group in depth), naturalistic observation (watching behavior in its normal setting), and surveys. It answers "what is happening?" but cannot establish cause.
Correlational research measures two or more variables to see whether they are related, producing a correlation coefficient. It answers "are these connected, and can we predict one from the other?" but still cannot establish cause because of the third-variable problem and directionality problem.
Experimental research manipulates one variable and measures its effect on another while controlling other factors. Because of random assignment and control, it is the only method that can demonstrate cause and effect.
A common exam trap: describing a survey or observation and then asking whether it proves a cause. The answer is no. Only the experiment earns causal language.
Descriptive research observes and describes without manipulating anything. It includes case studies (one person or group in depth), naturalistic observation (watching behavior in its normal setting), and surveys. It answers "what is happening?" but cannot establish cause.
Correlational research measures two or more variables to see whether they are related, producing a correlation coefficient. It answers "are these connected, and can we predict one from the other?" but still cannot establish cause because of the third-variable problem and directionality problem.
Experimental research manipulates one variable and measures its effect on another while controlling other factors. Because of random assignment and control, it is the only method that can demonstrate cause and effect.
| Method | Manipulates a variable? | Shows causation? | Main use |
|---|---|---|---|
| Descriptive | No | No | Describe behavior |
| Correlational | No | No | Measure relationships, predict |
| Experimental | Yes | Yes | Test cause and effect |
Variables, Groups, and Control in Experiments
In an experiment, the independent variable (IV) is what the researcher manipulates, and the dependent variable (DV) is what is measured as the outcome. A memory device: the DV "depends" on the IV.
An operational definition states exactly how a variable is measured or manipulated, so the study can be replicated. "Stress" is vague; "score on the 20-item Perceived Stress Scale" is operational.
Participants are placed in an experimental group (receives the treatment) or a control group (receives none or a placebo) for comparison. Random assignment — using chance to decide who goes in each group — is the key feature that makes groups equivalent at the start and rules out preexisting differences.
A confounding variable is any factor other than the IV that could explain the difference in the DV. If the treatment group also happened to be older or tested at a different time of day, those confounds cloud the results.
Control tools also include the placebo (a fake treatment that controls expectations), the double-blind procedure (neither participants nor researchers know who got the real treatment), and single-blind designs. On the exam, expect to identify IV, DV, operational definitions, and one confound in a described study, and to recommend random assignment as the fix for group differences.
An operational definition states exactly how a variable is measured or manipulated, so the study can be replicated. "Stress" is vague; "score on the 20-item Perceived Stress Scale" is operational.
Participants are placed in an experimental group (receives the treatment) or a control group (receives none or a placebo) for comparison. Random assignment — using chance to decide who goes in each group — is the key feature that makes groups equivalent at the start and rules out preexisting differences.
A confounding variable is any factor other than the IV that could explain the difference in the DV. If the treatment group also happened to be older or tested at a different time of day, those confounds cloud the results.
Control tools also include the placebo (a fake treatment that controls expectations), the double-blind procedure (neither participants nor researchers know who got the real treatment), and single-blind designs. On the exam, expect to identify IV, DV, operational definitions, and one confound in a described study, and to recommend random assignment as the fix for group differences.
Correlation, Causation, and the Coefficient
A correlation coefficient, symbolized , ranges from to . The sign shows direction; the number shows strength.
A positive correlation means variables move the same way (more sleep, higher grades). A negative correlation means they move oppositely (more absences, lower grades). A value near means little or no linear relationship. Importantly, is stronger than — students often think negatives are weak.
The cardinal rule: correlation does not prove causation. Two reasons. The directionality problem means we may not know which variable causes which. The third-variable problem means an unmeasured factor may cause both. Ice cream sales and drowning correlate, but hot weather drives both.
Correlations are displayed on scatterplots, where each dot is one participant's pair of scores. Tightly clustered dots along a line mean a strong correlation; a shapeless cloud means a weak one. On free-response questions, never write that a correlational finding shows one variable "causes" the other — describe it as a relationship or a predictor.
A positive correlation means variables move the same way (more sleep, higher grades). A negative correlation means they move oppositely (more absences, lower grades). A value near means little or no linear relationship. Importantly, is stronger than — students often think negatives are weak.
The cardinal rule: correlation does not prove causation. Two reasons. The directionality problem means we may not know which variable causes which. The third-variable problem means an unmeasured factor may cause both. Ice cream sales and drowning correlate, but hot weather drives both.
| value | Interpretation |
|---|---|
| Perfect positive | |
| Strong positive | |
| No linear relationship | |
| Strong negative | |
| Perfect negative |
Reading Descriptive and Inferential Statistics
Descriptive statistics summarize data. Measures of central tendency include the mean (average), median (middle value), and mode (most frequent). In a skewed distribution, extreme scores pull the mean toward the tail, so the median better represents the typical score. In a positively (right) skewed set, mean > median.
Measures of variability describe spread: the range (highest minus lowest) and the standard deviation (SD), which shows the average distance of scores from the mean. A small SD means scores cluster tightly; a large SD means they are spread out. In a normal distribution, about 68% of scores fall within one SD of the mean and about 95% within two.
Inferential statistics help decide whether results generalize beyond the sample. A result is statistically significant when it is unlikely to have occurred by chance alone, conventionally when the probability . Significance depends on the size of the difference, the sample size, and low variability.
A key misconception: statistical significance does not mean the effect is large or important in real life — only that it is probably not due to random chance. The exam may give you means and SDs and ask which group performed better or which data are more consistent.
Measures of variability describe spread: the range (highest minus lowest) and the standard deviation (SD), which shows the average distance of scores from the mean. A small SD means scores cluster tightly; a large SD means they are spread out. In a normal distribution, about 68% of scores fall within one SD of the mean and about 95% within two.
Inferential statistics help decide whether results generalize beyond the sample. A result is statistically significant when it is unlikely to have occurred by chance alone, conventionally when the probability . Significance depends on the size of the difference, the sample size, and low variability.
A key misconception: statistical significance does not mean the effect is large or important in real life — only that it is probably not due to random chance. The exam may give you means and SDs and ask which group performed better or which data are more consistent.
Research Ethics and Sampling
Ethical research protects participants and produces trustworthy data. Key requirements enforced by an ethics review board include:
Informed consent — participants agree to take part knowing what is involved. Informed assent applies to minors. Deception is permitted only when necessary and harmless, and must be followed by debriefing, explaining the true purpose afterward. Participants have the right to withdraw at any time, must be protected from lasting harm, and their data kept confidential.
Sampling affects whether findings generalize. The population is the whole group of interest; the sample is who is actually studied. A random sample gives every member of the population an equal chance of selection, improving representativeness. Do not confuse random sampling (how you choose participants, affecting generalizability) with random assignment (how you sort them into groups, affecting causal validity). The exam loves this distinction.
Sampling bias occurs when the sample does not represent the population, limiting external validity. A convenience sample of only volunteers may not generalize.
Informed consent — participants agree to take part knowing what is involved. Informed assent applies to minors. Deception is permitted only when necessary and harmless, and must be followed by debriefing, explaining the true purpose afterward. Participants have the right to withdraw at any time, must be protected from lasting harm, and their data kept confidential.
Sampling affects whether findings generalize. The population is the whole group of interest; the sample is who is actually studied. A random sample gives every member of the population an equal chance of selection, improving representativeness. Do not confuse random sampling (how you choose participants, affecting generalizability) with random assignment (how you sort them into groups, affecting causal validity). The exam loves this distinction.
| Concept | Purpose | Affects |
|---|---|---|
| Random sampling | Choose participants fairly | Generalizability |
| Random assignment | Sort into conditions | Causal validity |
Key terms
- Independent variable.
- The factor a researcher deliberately manipulates in an experiment to test its effect.
- Dependent variable.
- The outcome that is measured and expected to change in response to the independent variable.
- Operational definition.
- A precise statement of how a variable is measured or manipulated, enabling replication.
- Random assignment.
- Using chance to place participants into experimental or control groups, equalizing them and reducing confounds.
- Confounding variable.
- An uncontrolled factor other than the IV that could explain changes in the DV.
- Correlation coefficient.
- A statistic (, from to ) showing the direction and strength of a relationship between two variables.
- Standard deviation.
- A measure of variability indicating the average distance of scores from the mean.
- Statistical significance.
- An indication () that a result is unlikely due to chance alone, not a measure of importance or size.
Worked example
A researcher wants to know whether background music improves reading comprehension. She randomly assigns 60 volunteers to read a passage either in silence or with instrumental music, then scores each on a 10-question quiz. The music group averaged 7.2 (SD = 1.1); the silent group averaged 6.8 (SD = 2.6). The difference was statistically significant at . Identify the method, IV, DV, and interpret the statistics.
First, the method. Because the researcher manipulates a variable (music vs silence) and uses random assignment, this is an experiment, so causal conclusions are allowed.
Next, the variables. The independent variable is the listening condition (music vs silence) — what she manipulates. The dependent variable is reading comprehension, operationally defined as the score on the 10-question quiz. The silent group serves as the control group.
Now the descriptive statistics. The means show the music group scored higher on average (7.2 vs 6.8). The standard deviations reveal consistency: the music group's SD of 1.1 is much smaller than the silent group's 2.6, meaning music-group scores clustered tightly while silent-group scores were spread widely.
Finally, the inferential statistic. Because , the difference is statistically significant — unlikely to be due to chance. But note this does not mean the effect is large or practically important; a 0.4-point gap on a 10-point quiz is modest. Random assignment lets her infer that the music condition, not preexisting group differences, produced the effect.
Next, the variables. The independent variable is the listening condition (music vs silence) — what she manipulates. The dependent variable is reading comprehension, operationally defined as the score on the 10-question quiz. The silent group serves as the control group.
Now the descriptive statistics. The means show the music group scored higher on average (7.2 vs 6.8). The standard deviations reveal consistency: the music group's SD of 1.1 is much smaller than the silent group's 2.6, meaning music-group scores clustered tightly while silent-group scores were spread widely.
Finally, the inferential statistic. Because , the difference is statistically significant — unlikely to be due to chance. But note this does not mean the effect is large or practically important; a 0.4-point gap on a 10-point quiz is modest. Random assignment lets her infer that the music condition, not preexisting group differences, produced the effect.
Practice questions
A psychologist finds a correlation of between hours spent on social media and self-reported happiness. Which conclusion is best supported?
- Social media use causes decreased happiness
- As social media use increases, happiness tends to decrease
- Happiness has no measurable relationship to social media use
- Increasing happiness causes people to use less social media
Answer: As social media use increases, happiness tends to decrease
An of is a strong negative correlation, meaning the variables move in opposite directions. Because this is correlational data, no causal claim is valid — that rules out the two 'causes' options due to the directionality and third-variable problems. The 'no relationship' option is wrong because is a strong relationship. Only the descriptive statement about the negative association is supported.
A study reports two groups with equal means but very different standard deviations. Explain what the standard deviations tell you and why this matters when interpreting the results.
Answer: The standard deviations describe how spread out the scores are around the mean; the group with the larger SD has more variable, less consistent scores.
Central tendency alone can hide important differences. Two groups can average the same score while one is tightly clustered and the other is highly variable. The larger SD signals greater dispersion, meaning individuals in that group differ more from one another and from the mean. This matters because higher variability makes it harder to detect a true effect and can reduce statistical significance, and it tells researchers that the typical experience within that group is less uniform.
Which feature of a study most directly allows researchers to conclude that the independent variable caused a change in the dependent variable?
- A large random sample from the population
- Random assignment of participants to conditions
- A statistically significant correlation coefficient
- Operational definitions of all variables
Answer: Random assignment of participants to conditions
Random assignment equalizes the groups on preexisting characteristics, ruling out confounds so that any difference in the DV can be attributed to the IV — this is what licenses causal claims. Random sampling improves generalizability, not causation. A correlation cannot establish cause at all. Operational definitions aid replication and clarity but do not by themselves justify a causal conclusion.
FAQ
- What is the difference between random sampling and random assignment?
- Random sampling is how you select participants from the population, giving everyone an equal chance to be chosen; it improves how well results generalize. Random assignment is how you sort chosen participants into experimental or control groups; it equalizes the groups and allows causal conclusions. A study can use one, both, or neither.
- Why can't correlational studies prove causation?
- Because of two problems. The directionality problem means we cannot tell which variable influences which. The third-variable problem means some unmeasured factor might cause both. Only an experiment, with manipulation and random assignment, controls these issues well enough to support a cause-and-effect claim.
- Does statistical significance mean a result is important?
- No. Statistical significance () only means the result is unlikely to be due to chance. A tiny, practically meaningless difference can still be statistically significant, especially with a large sample. Significance is about probability, not the size or real-world importance of an effect.
- When is deception allowed in a psychology experiment?
- Deception is permitted only when it is necessary for the study's validity, does not cause lasting harm, and is approved by an ethics review board. Researchers must debrief participants afterward, fully explaining the true purpose and any deception used.
Learn this with a teacher, not a page
The Crimsora tutor teaches Scientific Foundations: Research Methods and Statistics live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.