Association strength: meaning and application

A report assigns ‘reliable’ an association-strength score of 82. It seems clear until someone asks: 82 of what? The share of participants, response speed, a rating or weight in a decision? A number becomes useful when its measurement definition is visible. Otherwise, a precise-looking score acquires a meaning the study never established.

Association strength describes the strength of a cue–response relationship within a specified measurement method. In free-association norms it may be expressed as relative response frequency. A scale of perceived importance or decision weight operationalises something different. Do not treat these measures as interchangeable; always explain how the score is produced.

Researchers can investigate a relationship through different questions. How often does response B follow cue A? How quickly is B recognised after A? How important does someone say the connection is? Each question requires an appropriate design and supports a different conclusion.

Nelson and colleagues report word-association norms including forward and backward strength. Direction matters: A may frequently evoke B while B less frequently evokes A. Strength is therefore not necessarily a symmetrical property of a word pair. Nelson et al., 2004

A practical target group profile may use a custom rating scale. Specify who rates the item, what scale points mean and which evidence supports the judgement. A researcher’s estimated priority is different from a directly recorded response proportion, even when both are displayed as a number between zero and one hundred.

This table supports research briefing; it does not provide conversion rules between measures. If two columns are both labelled ‘strength’, inspect their definitions before comparing scores, combining them or placing them in one ranking. Identical ranges do not imply identical meanings.

Imagine a fictional task in which 100 valid participants each give one response to the same cue. Twenty provide response B. The observed proportion for B is 20/100, or 0.20. This is an arithmetic illustration, not a research result or a rule for every association method.

You can display that proportion as 20%. You cannot conclude that B causes twenty percent of a purchase decision: no decision effect was measured. Specify what constitutes valid participation and how missing responses are handled. Changing the denominator changes the result.

If the task requests multiple responses, decide whether the analysis counts people, responses or response positions. Repetition by one participant must not silently become several independent people. A first response and a later response may also play different roles in the study. Preserve enough detail to inspect that distinction rather than reducing every answer to an undifferentiated total.

A scale from zero to one hundred is a display choice until its underlying measure is specified. For a Neurofactor-specific score, state whether it is a normalised measurement, a rating or a composite index. Explain zero too: does it mean absent, unobserved or the minimum of the selected scale?

A score of 80 does not necessarily indicate twice the decision influence of 40. Such a ratio requires a suitable measurement definition. Round values to a precision justified by the method. Two decimal places do not improve the accuracy of a small sample or subjective judgement. Where uncertainty materially affects a comparison, report it alongside the central estimate.

The Small World of Words publication addresses cue selection and the collection of multiple associations. It illustrates why association data require an explicit task description. De Deyne et al., 2019

Do not compare a one-word task with an extended interview as if only the audience differed. Check instructions, language, answer options, timing, coding and participant selection. When the method changes, a separate comparability study may be needed before interpreting a trend. A changed chart can reflect a changed instrument rather than a changed audience.

A fictional training provider investigates a course. ‘Time’ is mentioned often, whereas ‘recognition by my employer’ rarely appears spontaneously. Follow-up conversations suggest that time is mainly a scheduling question for some people. Employer recognition may be a firm requirement for others.

The team retains two findings separately: mention frequency and the role suggested by follow-up evidence. It then needs to test whether clearer scheduling and information about recognition make relevant differences. It does not rank both topics solely by spontaneous mention counts.

The example shows why strategic decisions may need different kinds of information. A large circle in an association map is not, by itself, a decision about budget, messaging or product development. The report should make the additional reasoning explicit so a later reader can assess it.

1. Name the cue, response and relationship direction.
2. Define the measure and scale.
3. Describe the sample, instructions and response options.
4. Explain the denominator, coding and missing-data treatment.
5. Report uncertainty and relevant within-group differences.
6. Separate measured strength from estimated strategic importance.
7. Test decision claims using data that concern the decision itself.

Ask another reader to explain the score in one sentence. If they say ‘this causes the purchase’ when the study only counted words, the reporting remains unclear. Revise the column heading and legend before the number is reused elsewhere. A precise definition should travel with the score into presentations and dashboards.

• Ruling out frequency as any possible operationalisation of association strength.
• Presenting a custom 0–100 score as a universal scientific standard.
• Equating positive valence with strong association.
• Omitting direction, denominator or valid sample size.
• Treating a group average as every participant’s characteristic.
• Describing differences between methods as changes in the audience.

A useful strength score states what was measured, among whom and under which conditions. Its strategic meaning depends on the appropriate follow-up question. Keep the associative relationship, its evaluation and its weight in a decision identifiable and separate.