In 2014, Pickett and Baker cast doubt on the scholarly consensus that Americans are pragmatic about criminal justice. Previous research suggested this pragmaticism was evidenced by either null or positive relationships between seemingly opposite items (i.e., between dispositional and situational crime attributions and between punitiveness and rehabilitative policy support). Pickett and Baker argued that because these studies worded survey items in the same positive direction, respondents’ susceptibility to acquiescence bias led to artificially inflated positive correlations. Using a simple split-ballot experiment, they manipulated the direction of survey items and demonstrated bidirectional survey items resulted in negative relationships between attributions and between support for punitive and rehabilitative policies. We replicated Pickett and Baker’s methodology with a nationally representative sample of American respondents supplemented by a diverse student sample. Our results were generally consistent, and, in many cases, effect sizes were stronger than those observed in the original study. Americans appear much less pragmatic when survey items are bidirectional. Yet, we suggest the use of bidirectional over unidirectional survey items trades one set of problems for another. Instead, to reduce acquiescence bias and improve overall data quality, we encourage researchers to adopt item-specific questioning.