Silicon–polymer composites fabricated by micromachining technology offer attractive properties for use as matching layers in high frequency ultrasound transducers. Understanding of the acoustic behavior of such composites is essential for using them as one of the layers in a multilayered transducer structure. This paper presents analytical and finite element models of the acoustic properties of silicon–polymer composites in 2-2 connectivity. Analytical calculations based on partial wave solutions are applied to identify the resonance modes and estimate effective acoustic material properties. Finite Element Method (FEM) simulations were used to investigate the interaction between the composite and the surrounding load medium, either a fluid or a solid, with emphasis on the acoustic impedance of the composite. Composites with lateral periods of 20, 40 and 80 μm were fabricated and used as acoustic matching layers for air-backed transducers operating at 15 MHz. These composites were characterized acoustically, and the results were compared with analytical calculations. The analytical model shows that at low to medium silicon volume fraction, the first lateral resonance in the silicon–polymer 2-2 composite is defined by the composite period, and this lateral resonant frequency is at least 1.2 times higher than that of a piezo-composite with the same polymer filler. FEM simulations showed that the effective acoustic impedance of the silicon–polymer composite varies with frequency, and that it also depends on the load material, especially whether this is a fluid or a solid. The estimated longitudinal sound velocities of the 20 and 40 μm period composites match the results from analytical calculations within 2.7% and 2.6%, respectively. The effective acoustic impedances of the 20 and 40 μm period composites were found to be 10% and 26% lower than the values from the analytical calculations. This difference is explained by the shear stiffness in the solid, which tends to even out the surface displacements of the composites.