Verbal hypotheses about multiple proportions abound across behavioral science. We provide a general-purpose theory-building  framework, grounded in deductive  science. Using the accompanying online app, scholars can codify a joint system of hypotheses into a single mathematical formal model, thereby disambiguating their verbal claims. No mathematics knowledge beyond high-school level is required. Instead of conceptualizing their scientific theory as a, potentially ill-specified, alternative to a `no-effect' strawman Null hypothesis, scholars can compare multiple competing  substantive claims theoretically and empirically. The proposed  theoretical framework enables one to quantity the parsimony (hence refutability) of each hypothesis, characterize and quantify the overlap among competing claims, formally model heterogeneous populations, characterize the predictions of each hypothesis succinctly, and provide insights into the edge cases of each hypothesis. We also cast predictability and replicability as testable mathematical models. By helping the scholar steer clear of over-specifying, over-fitting, and/or over-predicting, the framework opens up new avenues of depth and nuance in the development and analysis of theories about multiple proportions.