In the absence of a theory of everything, modern physicists need to rely on other predictive tools and turned to Effective Field Theories (EFTs) in a number of fields, including but not limited to statistical mechanics, condensed matter, particle physics, cosmology and gravity. The coefficients of an EFT can be constrained with high precision by experiments, which can involve high-energy particle colliders for instance but are generally left free from the theoretical point of view. The focus of this thesis is to use various consistency criteria to get theoretical constraints on the low-energy coefficients of EFTs. In particular, we construct a new model of massive spin-1 field by requiring that the theory is free of any ghostly degree of freedom. We then study its cosmological perturbations and ask that all propagating modes are stable and subluminal, reducing the space of viable cosmological solutions. Finally, we implement a method to get ‘causality bounds’, which are obtained by requiring infrared causality. This is imposed by forbidding any resolvable time advance in the EFT. We derive such ‘causality bounds’ for shift-symmetric and Galileon scalar EFTs, before turning to gauge-symmetric vector fields. We prove that our causality bounds can be competitive with positivity bounds and can even be used in scenarios that are out of reach of the positivity approach. The result of this thesis, by exploring several consistency criteria, is to provide compact causality bounds for low-energy EFT coefficients, in addition to constraints coming from the absence of ghosts, stability and cosmological viability.