Adaptive cognitive-based selection is recognized as among the most significant open issues in adaptive web-based learning systems. In order to adaptively select learning resources, the definition of adaptation rules according to the cognitive style or learning preferences of the learners is required. Although some efforts have been reported in literature aiming to update the adaptation logic used for a specific learner by updating his/her profile through the use of complex questionnaires that estimate the cognitive characteristics of learners, still the cognitive profile used for a learner remains static for a significant period, leading to the same selection decisions independent from the previous interactions of the learner with the system. In this paper, we address the learning object selection problem based on learners' cognitive characteristics, proposing a cognitive-based selection methodology that is dynamically updated based on the navigation steps of learners in a set of hypermedia objects. The proposed approach utilizes the Cognitive Trait Model, that is, an approximation model for learner's cognitive capacity that provides a concrete method for identifying learner's cognitive characteristics based on learners' navigation steps. In our experiment we simulate different learner behaviors in navigating a hypermedia learning objects space, and measure the selection success of the proposed selection decision model as it is dynamically updated using the simulated learner's navigation steps. The simulation results provide evidence that the proposed selection methodology can dynamically update the internal adaptation logic leading to refined selection decisions. (Contains 4 figures and 6 tables.)