Table 3.

Number of relationships stratified by expert agreement and definition usefulness

	Not Useful	One Useful	Two Useful	All
Pre-Delphi	81	28	15	124
Post-Delphi	19	10	3	32
Near-Agreement	14	7	4	25
All	114	45	22	181

We stratified the selected 200 relationships by two dimensions: (1) expert agreement (based on the Delphi rounds), which served as a proxy for task difficulty where less expert agreement implied higher task difficulty, and (2) expert-rated definition utility, which served as a proxy for definition quality for a relationship where the greater number of useful concept definitions for a relationship implied higher task definition quality. Note that some strata contain very few relationships. (For row label explanation see ‘Task Difficulty’ on page 12. For column label explanation see ‘Definition Quality’ on page 13.)