A paper reports that two characteristics tend to occur together. While preparing a discussion with colleagues, the statement becomes “one produces the other”. The change looks small, but it changes the question. Describing an association does not establish that changing one characteristic would cause the other to change.
For psychologists and psychotherapists reading research, this step deserves attention before using a finding in professional material. We can examine it through a wholly fictional survey about work organisation, with no clinical data, simulated numbers or findings attributed to real people. The example is for reading a conclusion, not demonstrating the usefulness of a digital tool.
Begin with what was observed
Imagine a fictional questionnaire for professionals. On the same occasion, participants report how often they use a digital diary and how organised they feel at work. In our invented account, more frequent reported use goes together with greater perceived organisation. No data have been collected or coefficients calculated here. This is simply a situation for practising the reasoning.
A consistent description would be “in the imaginary questionnaire, reported diary use and perceived organisation are associated”. “The diary makes professionals more organised” introduces an effect. It also replaces the perception recorded with a general feature of work that the questionnaire did not directly measure.
The Australian Bureau of Statistics guidance distinguishes association from causation and explains that a correlation coefficient alone does not identify a cause. We retain that distinction without reproducing data or empirical examples from the guidance.
Identify the missing causal question
The descriptive question asks which responses occur together among the people observed. A possible causal question asks what would change in organisation for comparable professionals if they used this diary rather than a defined alternative. The second requires specifying use, comparison, outcome and period of interest. Merely observing associated responses does not complete that comparison.
In our example, we have not assigned different organisational methods or established a comparison over time. We do not know how people worked before adopting the diary. This absence does not demonstrate that the tool has no effect. It means that the causal statement is unsupported by the described situation alone.
When reading a real study, examine the methods to see how the compared conditions were created or observed. “The authors examined existing users” and “the authors assigned participants to different conditions” describe different procedures. Understanding sustainable conclusions also requires examining how the study was conducted, rather than stopping at the design label.
Consider the direction of the relationship
One invented possibility is that the diary contributes to feeling organised. Another is that people already accustomed to organising their work choose digital tools more often. Presenting both does not assign them equal probabilities. It acknowledges that the described questionnaire does not establish which pathway produced the association.
Collecting both responses simultaneously does not establish which characteristic came first. Even knowing that one event preceded another would not, by itself, demonstrate an effect. Time order is part of the reasoning, rather than a shortcut that resolves every alternative explanation.
Your note can state the specific question: “Does the study document organisation before adoption of the tool?” If that information is missing from the section consulted, leave the check open. Do not fill in a plausible sequence simply because it makes the title you want to use sound more persuasive.
Look for possible common factors
In our scenario, previous organisational habits might matter: they could influence both choosing the diary and describing one’s work. This is an illustrative hypothesis, not an established mechanism. It shows why a common factor might make “the diary caused the outcome” an insufficient reading of the association.
Chapter 25 of the Cochrane Handbook discusses confounding in non-randomised intervention studies. It also explains that statistical adjustment does not necessarily remove all confounding. Our fictional questionnaire is not a ROBINS-I assessment or a reproduction of the chapter’s clinical examples.
Not every difference between people is automatically a confounder. Its relationship with the variables and its timing need consideration. For an initial reading, record which explanations the authors consider and how they address them. A full methodological appraisal may require specialist expertise. Adding many variables to a list does not replace that work.
Do not ask a number for a cause
A relationship described as strong or precise still needs an account of the design and assumptions supporting its interpretation. The strength of an association and the basis for a causal conclusion address different questions. A finding does not become causal simply because it looks convincing in a chart or is described confidently.
We do not need to invent a coefficient to see the problem in our example. The question remains: which information separates an effect of diary use from alternative explanations? If a paper uses words such as “effect” or “impact”, examine how the methods and discussion justify them. The verb itself does not settle the issue.
Also avoid the opposite conclusion that observational research is useless. Absence of randomisation does not make a report worthless, and randomisation does not remove the need to examine its conduct. The reading task is to recognise what that particular result supports, under which assumptions and with which limitations.
Keep the conclusion proportionate
For our fictional questionnaire, you can retain this statement: “The association between reported diary use and perceived organisation does not, by itself, establish that introducing the diary improves organisation. Time order, previous characteristics and the nature of the comparison remain to be examined.” It concludes neither that an effect exists nor that it is absent. It defines the answer currently available.
Before presenting a paper, mark where your wording moves from description to explanation. To work on the wording of a non-clinical note, you can try PsyLab for free while keeping the original text beside it. An AI rephrasing does not verify the study design. Reading the sources and taking responsibility for the conclusion remain the professional’s tasks.