Field and laboratory observations of surface wave fields show that the propagation speed of harmonic wave components differs from the predictions of linear theory. Angular spread fails to account for the observed discrepancy. An explanation is that the wave field does not consist of independently propagating Fourier components only but also contains groups of Stokes waves, forming envelope solitons. A mode field is proposed, and the power spectral density and wave number spectra, as well as cross spectra, are calculated and compared to observations. A method is also suggested for identifying the properties of solitons contained in an observed wave field from spectra and cross spectra.