Single-molecule biomechanical measurements, such as the force to unfold a protein domain or the lifetime of a receptor-ligand bond, are inherently stochastic, thereby requiring a large number of data for statistical analysis. Sequentially repeated tests are generally used to obtain a data ensemble, implicitly assuming that the test sequence consists of independent and identically distributed (i.i.d.) random variables, i.e., a Bernoulli sequence. We tested this assumption by using data from the micropipette adhesion frequency assay that generates sequences of two random outcomes: adhesion and no adhesion. Analysis of distributions of consecutive adhesion events revealed violation of the i.i.d. assumption, depending on the receptor-ligand systems studied. These include Markov sequences with positive (T cell receptor interacting with antigen peptide bound to a major histocompatibility complex) or negative (homotypic interaction between C-cadherins) feedbacks, where adhesion probability in the next test was increased or decreased, respectively, by adhesion in the immediate past test. These molecular interactions mediate cell adhesion and cell signaling. The ability to "remember" the previous adhesion event may represent a mechanism by which the cell regulates adhesion and signaling.