Fear conditioning and extinction is a construct integral to understanding trauma-, stress- and anxiety-related disorders. In the laboratory, associative learning paradigms that pair aversive with neutral stimuli are used as analogues to real-life fear learning. These studies use physiological indices, such as skin conductance, to sensitively measure rates and intensity of learning and extinction. In this review, we discuss some of the potential limitations in interpreting and analysing physiological data during the acquisition or extinction of conditioned fear. We argue that the utmost attention should be paid to the development of modelling approaches of physiological data in associative learning paradigms, by illustrating the lack of replicability and interpretability of results in current methods. We also show that statistical significance may be easily achieved in this paradigm without more stringent data and data analysis reporting requirements, leaving this particular field vulnerable to misleading conclusions. This review is written so that issues and potential solutions are accessible to researchers without mathematical training. We conclude the review with some suggestions that all laboratories should be able to implement, including visualising the full data set in publications and adopting modelling, or at least regression-based, approaches.