Abstract Calmodulin (CaM) is a small protein involved in calcium signaling; among the targets of CaM are a number of kinases, including myosin light chain kinases (MLCK), various CaM-dependent kinases and phosphorylase kinase. We present results of molecular dynamics (MD) simulations of 4-ns length for calmodulin in its three functional forms: calcium-free, calcium-loaded, and in complex with both calcium and a target peptide, a fragment of the smooth muscle MLCK. The simulations included explicit water under realistic conditions of constant temperature and pressure, the presence of counterions and Ewald summation of electrostatic forces. Our simulation results present a more complete description of calmodulin structure, dynamics and interactions in solution than previously available. The results agree with a wide range of experimental data, including X-ray, nuclear magnetic resonance (NMR), fluorescence, cross-linking, mutagenesis and thermodynamics. Additionally, we are able to draw interesting conclusions about microscopic properties related to the protein's biological activity. First, in accord with fluorescence data, we find that calcium-free and calcium-loaded calmodulin exhibit significant structural flexibility. Our simulations indicate that these motions may be described as rigid-body translations and rotations of the N- and C-terminal domains occurring on a nanosecond time scale. Our second conclusion deals with the standard model of calmodulin action, which is that calcium binding leads to solvent exposure of hydrophobic patches in the two globular domains, which thus become ready to interact with the target. Surprisingly, the simulation results are inconsistent with the activation model when the standard definitions of the hydrophobic patches are used, based on hydrophobic clefts found in the X-ray structure of calcium-loaded calmodulin. We find that both experimental and simulation results are consistent with the activation model after a redefinition of the hydrophobic patches as those residues which are actually involved in peptide binding in the experimental structure of the calmodulin–peptide complex. The third conclusion is that the calmodulin–peptide interactions in the complex are very strong and are dominated by hydrophobic effects. Using quasi-harmonic entropy calculations, we find that these strong interactions induce a significant conformational strain in the protein and peptide. This destabilizing entropic contribution leads to a moderate overall binding free energy in the complex. Our results provide interesting insights into calmodulin binding to its kinase targets. The flexibility of the protein may explain the fact that CaM is able to bind many different targets. The large loss of conformational entropy upon CaM:peptide binding cancels the entropy gain due to hydrophobic interactions. This explains why the observed entropic contribution to the binding free energy is small and positive, and not large and negative as expected for a complex with such extensive hydrophobic contacts.