The premise of this paper is that interactions among strangers are largely based on the symbolic systems, reputation, and credibility generated through institutions. We propose competence, agency problem, and hazard remedy as the fundamental dimensions of institutional trust. And we measure an individual's trust in an institution from a behavioral perspective, i.e., behavioral responses to financial risks, civil offenses, and misconduct of public officials. A stratified random sample of 1,374 respondents from the northern part of Taiwan have completed face-toface interviews. The major findings are: (1) The behavioral approach to measuring institutional trust is valid. The assessments of the institutions on competence, agency and hazard remedy are correlated with the behavioral responses at the right direction. (2) Behavioral responses are markedly different for those who have actual experience and those who don't but project their behavioral responses to hypothetical situations. The experienced tend to rely on own means to deal with the crisis, while the no experienced on legal system. (3) For those who have actual experience, they would have chosen different strategies if the same offenses ever happen again to them. Behavioral responses reveal a trial-and-error learning process.