Choose Words That Match the Study
A sentence such as “increasing study time raises quiz scores” sounds clear, but it makes a cause-and-effect claim: changing study time is said to produce a change in scores. Whether that wording is appropriate depends on how the data were collected. If researchers only recorded students’ usual study time and scores, the data can show an association, but they do not establish that studying longer caused higher scores.
In “Association Versus Causation in Regression,” you learned that a fitted regression line summarizes an association in the data and can help predict a response from an explanatory variable. In “Experiments and Establishing Causation,” you learned that random assignment to treatments in a well-designed experiment can support a cause-and-effect conclusion about those treatments. This tutorial focuses on putting that distinction into words.
The aim is not to make every conclusion vague. It is to be precise about what the evidence shows. For an observational study, a strong sentence can name both variables, describe the direction of their association, and identify the cases or setting. It should not claim that deliberately changing the explanatory variable would change the response. For a suitable randomized experiment, causal wording may be justified, but the conclusion should still name the treatment and the experimental units rather than making a broader claim than the study supports.
Translate Causal Claims Into Associational Statements
When you revise a sentence, first identify its variables. Which variable is being presented as the cause, and which is the response? Next, check the study design. Did researchers impose and randomly assign a treatment, or did they observe values that already existed? Then choose wording that matches the design.
For observational data, useful phrases include “is associated with,” “tends to be higher among,” “is related to,” and “the data show a positive association between.” These phrases describe a pattern without saying that changing one variable produces a change in the other. Avoid causal verbs such as “causes,” “increases,” “raises,” “reduces,” “improves,” and “leads to” when the study design does not support that claim.
A regression slope needs the same care. As you learned in “Variables, Units, and Meaning in a Regression Model,” the slope describes the predicted change in the response for a one-unit increase in the explanatory variable. That interpretation describes the fitted model. It does not, on its own, say that making the explanatory variable increase will cause the response to change. In an observational study, include language such as “the model predicts” or “among these students, scores tend to be higher for students reporting more study time.”
Also keep the conclusion tied to the cases in the data. “Students who studied more tended to earn higher scores in this survey” describes an observed association. “Making students study more will raise their scores” proposes an effect of an intervention. Those are different claims, even if the same two variables appear in both sentences.
Worked Examples: Matching Claims to Evidence
Worked Example: Study Time and Quiz Scores
A fictional teacher asks 90 students how many hours they studied for a recent quiz and records each student’s score. The teacher does not assign study hours. A fitted regression line has a slope of 2.4 score points per additional reported study hour.
Original claim: “Increasing study time raises quiz scores by 2.4 points per hour.”
Check the design. The teacher recorded students’ existing study habits and scores. This is an observational study, not an experiment: the teacher did not impose study-time treatments or randomly assign students to them. As discussed in “Observational Studies and Confounding,” other variables could be related to both study time and quiz performance.
Interpret what the slope supports. The slope describes the fitted regression model’s predicted score difference for students whose reported study times differ by one hour. It does not show what would happen if a particular student were made to study one additional hour. The words “increasing” and “raises” in the original claim suggest a causal effect that this study cannot establish.
Rewrite. “Among the 90 students in this survey, each additional reported hour of study was associated with a predicted quiz score that was 2.4 points higher, on average.” This sentence names the cases, gives the direction of the association, and treats the slope as a model-based prediction rather than proof of cause and effect.
Worked Example: A Randomized Study Routine
In a fictional experiment, 48 volunteer students are randomly assigned to one of two study routines for a week. One group uses a spaced-practice routine; the other uses the students’ usual routine. Everyone takes the same quiz under the same conditions. The spaced-practice group has a higher average score.
Original claim: “Spaced practice caused the students’ quiz scores to be higher.”
Check the design. Researchers imposed the study routines and used random assignment. Both groups took the same quiz under the same conditions, so the design supports comparing the assigned treatments. This is the kind of randomized experiment described in “Experiments and Establishing Causation.”
Check what the claim covers. Causal wording is appropriate for the treatment comparison in this experiment. However, “the students” should refer to the 48 participants, and the conclusion should describe the routine as assigned and the quiz outcome measured. The participants were volunteers, not a stated random sample of all students.
Rewrite. “For the 48 volunteer students in this experiment, assignment to the spaced-practice routine caused a higher average score on the quiz than assignment to the usual routine.” This statement connects the cause-and-effect conclusion to the randomized treatment comparison and the measured response. It does not claim that the result must apply to every student or every quiz.
Worked Example: Screen Use and Sleep Duration
A fictional survey asks 160 teenagers to report their usual evening screen time and their usual sleep duration. The data show that teenagers reporting more evening screen time tend to report fewer hours of sleep.
Original claim: “Using a phone longer at night reduces teenagers’ sleep.”
Check the design. The survey records screen use and sleep as they occur; researchers do not assign participants different amounts of phone use. The study shows a negative association in the surveyed teenagers, but the design does not establish that longer phone use caused shorter sleep. A third variable, such as evening responsibilities, could be related to both variables, as you considered in “Identifying a Confounding Variable in a Scenario.”
Choose a description of the pattern. “Reduces” says that changing screen use produces a change in sleep. To report the observed pattern, state which values tend to occur together and name the cases.
Rewrite. “Among the 160 teenagers surveyed, those reporting more evening screen time tended to report fewer hours of sleep.” This is associational wording: it describes the direction of the observed pattern without claiming that reducing phone use would necessarily increase sleep.
Common Mistakes and Full-Credit Wording
A common mistake is to use a causal verb simply because one variable is labeled explanatory and the other response. Those labels identify the roles of variables in an analysis; they do not prove that the explanatory variable caused the response. Another common mistake is to treat a strong association or a high \(r^2\) as proof of causation. As covered in “What r-squared Does Not Tell You,” \(r^2\) summarizes variation accounted for by a linear relationship; it does not establish a cause-and-effect relationship.
A third mistake is to write an associational sentence so generally that its meaning becomes unclear. “There is a relationship” may be technically cautious, but it does not say which variables are related, whether the association is positive or negative, or which cases were observed. A complete associational statement identifies those details without overstating what the study proves.
- Do not turn a slope into a treatment effect. In observational data, say the model predicts a change in the response for a one-unit difference in the explanatory variable; do not say that deliberately increasing the explanatory variable will cause that change.
- Do not mistake “associated with” for a weak conclusion. It is often the precise conclusion supported by an observational study.
- Do not use causal wording just because the pattern is strong. The design, not the strength of the association, determines whether a causal conclusion is justified.
- Do not overgeneralize an experimental result. State which treatment comparison and experimental units the result concerns. Generalizing beyond the participants requires an appropriate basis.
- Do not omit context. Name the variables, the observed cases or participants, and the response when interpreting a result.
For an AP-style response, a full-credit associational sentence might say: “Among the students surveyed, reported study time was positively associated with quiz score; students reporting more study time tended to have higher scores. Because study time was observed rather than randomly assigned, this study does not establish that increasing study time causes scores to rise.” A full-credit causal sentence for a randomized experiment should name the assigned treatment, the comparison condition, the measured response, and the experimental units or participants.
Check Your Understanding
For each statement, decide whether the wording is causal or associational, then revise it if needed to match the described study.
- A survey records students’ usual exercise and resting heart rates. A student writes, “Exercising more lowers resting heart rate.” Rewrite the claim so it describes an association.
- Researchers randomly assign volunteers to two reasonable stretching routines and compare flexibility afterward. What kind of wording may the study support, and what should the conclusion name?
- A regression slope describes the predicted change in monthly water use for each additional household member. Explain why that slope alone does not show that adding a person causes water use to rise.
- A report says, “The strong association proves that the new tutoring program improves scores.” Name the wording error and identify what information about the study design is needed.
- Write one sentence interpreting a negative association between daily outdoor temperature and home heating use without implying that the data-collection method establishes causation.