The present paper addresses the detection of structural changes using statistical learning methods with the objective of exposing the major limitations of defining baselines in which structural systems must be assumed unchanged. For this purpose, the data acquired during approximately two years in a cable‑,stayed bridge is used and was chosen since it comprises unchanged structural responses as well as measurements comprising changes imposed by creep and shrinkage. In addition to the study of baseline‑,based detection, the present paper proposes a simple yet effective baseline‑,free approach relying auto‑,correlation functions. The effectiveness of the proposed strategy in clearly distinguishing between changed and unchanged structural responses was demonstrated using the same data set.