Three common mainstream scale construction practices are based on both a faulty conception of concepts and on erroneous logical/statistical assumptions: (a) two states of affairs (or item responses designating same) correlating means that they are indicative of the same concept; (b) one and the same concept cannot designate both of two mutually incompatible states of affairs (or item responses designating same); and (c) using measures consisting of a cloud of correlated, but not necessarily conceptually coherent, variables entitles one to report results about a target concept (e.g., “conscientiousness”) even though their referent is not real-world instantiations of this concept, and even though the resulting construct routinely fails to capture anything close to established consensual usage of this target concept, thus resulting in potentially misleading conclusions. Finally, and more positively, existing research methods and findings that capture the actual concepts that people employ when they think and act in their worlds, several of which are described in the article, are argued as vastly preferable ways to establish concept meanings for incorporation into psychometric measures.