This study asked how a question set in an invented scene can make its premise open to being shown wrong, and what its account of its making lets it claim. Its answer, from thirteen synthetic inquiries of Dionyas, is that the scene is reduced to a role and a question, that each record writes one test of its own, that seven of those tests say which result would weaken the question's premise, and that no test says who would run it. The records' one clause about their making, a model's draft checked and registered by RISCON, is enough to show that the shape of the test follows the series of source cards; it is not enough to explain why.
Discussion
The debate on vignettes has turned for decades on how much of a situation to describe and whether answers to invented scenes can be trusted. Alexander and Becker argued for concrete situations over abstract questions; Aguinis and Bradley asked for realism under experimental control; Hainmueller, Hangartner and Yamamoto checked hypothetical answers against real choices and found that differences in design changed how well they matched. The Dionyas records take an unusual position in that debate. They describe almost nothing. A role of a few words and a short question are the whole scene, and in three records the question is not even a question but a single sentence of situation. What they build instead is the comparison: the chorus put back or left out, the cover seen or not, the choreography met first or the live scene. The variation that a vignette study writes into its scenes is written here into a test, and the test points outward, at rooms, listeners and rates that exist outside the invented scene.
The debate on personas turned on whether an invented figure can carry evidence at all. Pruitt and Grudin held that personas engage teams and carry data; Chapman and Milham answered that a persona cannot be verified or falsified; Matthews, Judge and Whittaker found practitioners using personas to communicate rather than to design. The roles in these records are personas stripped to a job title, and the objection holds for them in full: no observation could show that a regular in the audience was the wrong person to ask. But the objection was aimed at the figure, and the records put their weight on the question. In seven of the thirteen, the question can be found to rest on too large an assumption by an observation the record itself names. That is a modest claim. It does not make the role valid; it shows that a question built around an unverifiable role can still be built to be checked.
What the records leave open is as plain. A test with no one to run it is a test in waiting. The mark on every record, registered and awaiting judgment, confirms that none of these tests has been run, and the records do not say who would run them, who would read the result, or who would decide that a premise had failed. Two of them send the result to a meeting; none sends it to a person. Read through Gebru, Morgenstern and others, the clause on provenance documents what the records are and roughly how they were made, but not the instructions behind each series of cards or what the check by RISCON looked for. The regularity by series that this study measured is visible because of that clause and unexplained because of its brevity.
Contributions
The study makes three contributions. It describes, with counts from the records alone, a form of question neither literature has examined: an invented scene compressed to a role, carrying a test line that is the record's own and often names the result that would weaken the question. It relocates the falsifiability objection to personas from the figure to the question, and shows on published records that the relocation can succeed where the figure fails. And it reads a one-clause account of provenance as a partial datasheet, showing both what such a clause makes visible, a difference in test form by series of source cards, and what it cannot explain.
Limits
The evidence is thirteen records, fetched once and translated by this study; another translation could shift the reading of a sentence, though not the counts, which run on the Korean. The three counts that matter most, a condition, a result that would weaken the premise and a prescribed procedure, are word tests, and a different list of words would move a record or two. The series of source cards are known only by the prefix of a card number; this study does not know what the series are, and attributes the regularity to none. The records are synthetic by their own statement, so nothing here is a finding about fans, labels, venues or the people who work there. The selection rule left out the 6 synthetic inquiries of the evidence lane, which may be built differently. And the study has not run any of the tests the records describe; it cannot say whether any of them would work.
Further research
Three lines of work would carry this further. The first is to read the 6 synthetic inquiries the rule left out and the 1,520 others with the same measures, to see whether a test line that names its own failure appears anywhere outside the synthetic records. The second is to follow the records over time: each gives a question version and a judgment version, and a later version could show whether a test was run, by whom, and whether a premise was given up. The third is to ask readers, in the manner of the vignette studies, whether a record with a full scene and a record with only a role and a question draw different answers, and whether a test line that names its own failure changes what readers decide. Each would show more of what this study has only described: how a question about music, invented for the purpose, can be built so that it might turn out wrong.