This dissertation investigates deformations of reinforced concrete (RC) elements subjected to short-term tension, accounting for the long-term processes, i.e. shrinkage and creep, realised in the concrete before the mechanical load; the yielding of steel reinforcement limits the considered deformation range. The effect of heterogeneity of the concrete structure on the deformation of the tensile elements is analysed, revealing limitations of the deterministic modelling approaches. The tensile test methods are explored discussing the corresponding beneficial aspects and constraints. The deformation models and, in particular, the modern numerical simulation techniques are also discussed. The results gathered by the author and reported in the literature proved the importance of the tension-stiffening model as an input parameter for adequate prediction of the deformations and cracking. However, a substantial scatter of test results, related to the structural defects of the concrete having stochastic nature, in combination with time-dependent effects acting in the concrete complicates the modelling. Thus, the reliability of the test results even gathered during the short-term tests, is insufficient, owning the long-term processes in the concrete. The finite element simulation results indicate that strain distribution in the concrete of the tensile elements varies not only along the element length but also within the cover depth. That proves the limited suitability of the effective area concept, accepted in the design codes, which adequacy is dependent on the loading and deformation conditions, and shape of the element. A stochastic approach to the modelling, suitable for a probabilistic design, has been proposed in this work. The heterogeneity of the concrete was related to the tension-stiffening model. The proposed methodology is accomplished using Monte Carlo simulations that allows generating an extensive collection of structural responses with a frequency related to the a priori probability distribution. Based on the array of the generated structural response, the a posteriori probability distribution of the deformation response of a particular RC member is assessed. The application of the stochastic principles enables predicting not only the average deformation response of RC elements but also the probabilistic bounds of these predictions. In contrast to the common practice, shrinkage of concrete is taken into account in the short-term numerical simulation. The dissertation contains the introduction, four chapters, general conclusions, and lists of literature references and publications by the author on the topic of the research. It also includes five annexes. Nine papers are published on the research topic: five articles in the journals referred in the Clarivate Analytics WoS Data¬base (DB), two works in the WoS Proceedings DB and two in other proceedings. Four presentations were made at national and international conferences.