Figure 1. What four studies measured

Each study measured a different outcome, in its own units and on its own schedule, so the results answer different questions.

Figure 1. What four studies measured. Each study measured a different outcome, in its own units and on its own schedule, so the results answer different questions.
Study	Listeners and music	How often heard	What was measured	What was found	What it cannot show
S2	2017 liking study: 15 listeners; 40 examples	Plays 1, 10, 19 and 28 over about four weeks	Liking rating from 0 to 10	Liking rose with more plays (omega-squared 0.197: plays accounted for about a fifth of the variation)	No universal exposure curve
S4	2019 brain-recording study: 40 participants in two groups; 8 or 12 excerpts	Three repeats in a row per excerpt	How closely listeners' brain signals moved together (EEG)	Brain-signal alignment fell for familiar-style music and persisted for unfamiliar-style music	No validated measure of boredom or value
S5	2023 earworm study: 44 participants; four sung choruses	One to four plays; test one day later	Recall of items in order, plus 0 to 100 ratings	More plays tended to increase interference with recall for some songs	No reliable earworm rule
S6	2024 melody-perception study: 96 recruited; two excluded; 96 excerpts	Two rating blocks; the number of plays was not varied	Five-point ratings, analyzed with statistical models	How listeners perceived a melody carried several links between its measured features and liking	No claim that more plays cause any effect

Does not prove: That more exposure has one meaning, that prediction determines pleasure or that a universal exposure curve exists
Scope: Four studies published from 2017 through 2024 using adult listener samples
Units: Study-specific ratings, EEG correlation, serial recall and model estimates
Denominator: S2: 15 listeners; S4: 40 participants; S5: 44 participants; S6: 96 recruited with two excluded
Date: Sources checked 1 September 2026
Transformation: Each study's published schedule, direction of effect and statistics are restated as reported. Nothing is normalized, meaning rescaled onto a common scale, and no unpublished means are estimated
Uncertainty: Each row keeps its own design, population, unit and limitation
Limitations: This is not a meta-analysis, which would pool the studies. What each study measured, and the size of each effect, are not comparable across rows.

Sources