Coupled with corpora, usage-based construction grammar aims to provide cognitive plausibility for linguistic phenomena. In this vein, this paper combines construction grammar and usage-based approaches to analyze evidentiality in Turkish. While Turkish has been analyzed from a usage-based perspective, evidentiality has not been taken up in a usage-based constructionist approach. By using corpora, association measures, and construction as a notion and a framework, this paper defines the Unevidentiality Construction in a taxonomic space. First, it outlines its semantic properties and then it uses association measures such as faith, delta (∆) p, and ITECX to determine its usage pattern and statistical biases based on corpora. The paper demonstrates a superordinate and lower-level, item-specific instantiations of the construction. The results from association measures and the family of unevidentiality constructions can serve for future linguistic endeavors.