Therefore, it is not credible that the standard survey results really imply merely one-part or two-part relations. It is virtually inconceivable to suppose it is a one-part relation if we pay even slight attention to what it could mean: I trust, period, everyone and with respect to everything. The standard survey question refers to “most” people; but then, it also says “most of the time,” which could be taken to mean that the respondents must restrict the range of matters on which they would trust “most people.” This is not genuinely generalized trust. The respondents are forced by the vagueness of the questions to give vague answers, and it is a misdescription to label their responses as generalized trust.

QUESTION: What would a properly formulated survey question for generalized trust look like, and would it change measured levels?

Russell Hardin, Trust and Trustworthiness, loc. 317