Synthometricon

The data are primarily sourced from the APA PsycTests Database with permission from the American Psychological Association (APA). The database holds more than 28,000 instruments and 71,000 individual scales and subscales, built from more than 571,500 items that were extracted by parsing PDFs using (visual) language models. A handful of instruments come from other sources, which are labelled as such in the results.

Instruments are hierarchical. A scale may contain subscales, which may themselves contain subscales, up to four levels deep. Every level gets its own vector, pooled from all the items beneath it, so an instrument's overall representation and each of its parts can be searched and compared independently.

Synthometricon estimates how strongly your items would correlate with each scale in the database if both were given to the same people. We checked these estimates against real data in two studies. In the first, with 22 well-known personality and well-being questionnaires (493 participants), the estimates correlated .87 with the observed correlations between scales. In the second, with 57 scales drawn to cover the variety in the PsycTests database (387 participants), they correlated .83. A typical estimate was off by about .12 in the first study and .16 in the second. The first study is a best case; the second is closer to what you can expect for the database as a whole.

Of 10 scales that truly correlate .70 or more with yours, about 9 appeared with a displayed value of .30 or higher in the second study, and about 7 at .40 or higher. In the first study, all of them reached .40. Of scales that correlate .50 or more, about 7 in 10 appeared at .30 or higher in the second study. This means: If nothing in your results lies above .30, a strongly overlapping measure is unlikely, but not ruled out.

To see 9 in 10 of the scales that truly correlate .70 or more with yours, read down to a displayed value of about .30. For scales that correlate .50 or more, the same share requires reading down to about .17. For well-known personality measures, .55 and .30 were enough. The engine shows how many results lie above any value, and the filter box helps narrow long lists.

Yes, you will get better results with three or more items.

Warning codes indicate potential data quality issues detected during automated AI-extraction from source PDFs. Most cases flagged by these errors are not problematic, but highlighted records may require additional scrutiny. Regardless of warning codes, we recommend verifying any critical information against the original source when possible. Four checks can raise a warning: The extracted item count or scale count differs from the instrument's catalogue record, the extracted item text may not be a verbatim quote of the source PDF, or one or more items were translated into English for this corpus.

What "Could not be verified" means. Some checks cannot run at all. An instrument outside the APA PsycTests catalogue has no record to compare counts against, and a PDF with a thin text layer cannot be checked for verbatim item text. Those cases are reported separately from checks that ran and found nothing, because treating the two as the same would quietly present unverifiable rows as verified ones.

Separately, a result may note that some items point against the keyed average of the rest of their scale. This information is infered by the embedding model and may also flag records that require caution. However, the search results assume that that the information provided in the PDF document is correct.

In summary, warning flags are overly sensitive by design.

In item-based search, cosine similarity scores are closely aligned with the correlation coefficients one would expect when administering two scales to human respondents. In label-based search, similarity scores reflect semantic relatedness between the search term and the scale and/or instrument labels in the database.

While cosine similarity theoretically ranges from -1 to 1, sentence vectors rarely yield negative scores.

Expanding a result shows a score for the instrument as a whole and one for every scale and subscale inside it. Each is computed against that level's own pooled representation, so a single subscale can match your query more strongly than the instrument it belongs to. The instrument pools every item in the document, including the ones your query has nothing to do with. The headline score of a result is the strongest match anywhere in its tree.

Unfortunately, due to copyright restrictions, we cannot display individual survey items (statements/questions) directly on the platform. However, you can follow the DOI link to obtain the items from the original source. To access full instrument details, please:

• Click the DOI link in search results to access the original source
• Contact the instrument authors for permission to use the full measure
• Check institutional access through your library's APA PsycTests subscription