Vortex-induced vibrations represent a potentially critical phenomenon for bridges in a multitude of cases. The increasing optimization of modern bridge structures is giving rise to long spans and slender decks particularly sensitive to vortex shedding, which may generate deck oscillations unacceptable for the comfort of the users even for relatively low wind velocities. At the same time, the variety of geometric details characterizing bridge deck cross sections and the marked influence of an even limited variation of the flow angle of incidence add further difficulties to the prediction of bridge deck VIV response without an accurate and extended experimental campaign. In this context, VIV mathematical modeling represents a challenging but also an attractive issue for bridge decks, not only from a scientific point of view, but hopefully with the practical purpose of developing a mathematical approach able to support or at least partially reduce the amount of wind tunnel tests needed during the design phase. The present dissertation deals with the critical effect of significant cross-section geometric details and flow angle of attack on bridge deck VIV response and these factors are included in an extended study about two mathematical approaches for VIV response prediction applied to different configurations of a realistic bridge section. An overview of representative and meaningful studies provided by scientific literature about the influence of geometric details and angle of attack is firstly proposed, along with a description of mathematical modeling attempts developed over the years for VIV modeling of elongated cylinders with a constant cross section. The variability given by geometric features combined with angle of attack variation over a realistic range of values can be remarkable. This should be preferably taken into account by VIV mathematical modeling, usually developed, on the other hand, for simplified section geometries at zero angle of attack. In this work, a wake-oscillator model, derived from Tamura and Matsui’s one, and a modified version of harmonic model are deeply discussed and employed for a realistic cross-section geometry. For the purpose of exploring the influence of geometric details at different wind angles of incidence, calibrating both models and assessing their performances, wind tunnel tests were performed on a bridge deck sectional model. The modification of bridge section lower corners, the addition of two lateral barrier typologies and the investigation of different angles of attack gave rise to a large amount of tested section layouts. Aerodynamic force measurements were firstly performed on the stationary body and results were employed for mathematical model calibration and to formulate qualitative suppositions about the expected dynamic behavior of different cross-section layouts. Then, aeroelastic tests were carried out, with the model elastically suspended and free to vibrate. Response curves for different values of wind velocity were determined for each geometric layout. They were employed to complete wake-oscillator model calibration and compared to the predictions of both mathematical approaches. A critical effect of lateral barriers installed on the deck was observed, along with marked effects, even unexpected in some cases, produced by limited angle of attack variations. In addition, an estimation of the aerodynamic damping was conducted through free-decay tests. In the last part of the dissertation, VIV response mathematical modeling is addressed specifically. According to what previously stated, a Tamura-type wake-oscillator model and a modified harmonic model were investigated. The first one is a physically based two-degree-of-freedom approach, while the second one is a one-degree-of-freedom simplified model aiming to a quick and relatively easy prediction of the peak response at lock-in. A mathematical connection between the models was explored and both approaches were studied and applied preliminarily for a 4:1 rectangular cylinder and then for the bridge deck geometric configurations experimentally tested at different angles of attack. The modified harmonic model was calibrated only through results of static force measurements, while for the wake oscillator model, in absence of further static tests like flow visualizations, the response curves obtained at low Scruton number were employed to calibrate a part of model parameters. Virtues and limitations of both models are discussed. In particular, the necessity of adapting the wake oscillator design of the Tamura-type model for elongated cross sections, like a 4:1 rectangle or a bridge deck, was observed. Speculations about wake lamina geometry modification were formulated, with the purpose of achieving, for elongated sections, the appropriate values of model parameters without losing their physical origin. On the other hand, wake-oscillator model equations of motion were found able to reproduce a large variety of experimental response curve shapes, while limitations in peak response amplitude prediction were frequently observed, as well as for the modified harmonic model, especially where a considerable oscillation amplitude was found. Such limitations were attributed to the spanwise correlation increase of the vortex-shedding force with the oscillation amplitude; full correlation at lock-in was proposed as upper bound condition for the fluctuating force acting on the body. Pressure measurements for correlation effect quantification and further study about the wake oscillator structure for elongated sections, including the addition of a degree of freedom, are finally proposed as main possible advancements to improve the present work.