Spectral absorption coefficients of total particulate matter a(p) (lambda) were determined using the in vitro filter technique. The present analysis deals with a set of 1166 spectra, determined in various oceanic (case 1) waters, with field chi a concentrations ([chl]) spanning 3 orders of magnitude (0.02-25 mg m(-3)). As previously shown [Bricaud et al.; 1995] for the absorption coefficients of living phytoplankton a(phi)(lambda), the a(p)(lambda) coefficients also increase nonlinearly with [chl]. The relationships (power laws) that link a(p)(lambda) and a(phi)(lambda) to [chl] show striking similarities. Despite large fluctuations, the relative contribution of nonalgal particles to total absorption oscillates around an average value of 25-30% throughout the [chl] range. The spectral dependence of absorption by these nonalgal particles follows an exponential increase toward short wavelengths, with a weakly variable slope (0.011 +/- 0.0025 nm(-1)). The empirical relationships linking a(p)(lambda) to [chl] can be used in bio-optical models. This parameterization based on in vitro measurements leads to a good agreement with a former modeling of the diffuse attenuation coefficient based on in situ measurements. This agreement is worth noting as independent methods and data sets are compared. It is stressed that for a given [chl], the a(p)(lambda) coefficients show large residual variability around the regression lines (for instance, by a factor of 3 at 440 nm). The consequences of such a variability, when predicting or interpreting the diffuse reflectance of the ocean, are examined, according to whether or not these variations in a(p) are associated with concomitant variations in particle scattering. In most situations the deviations in a(p) actually are not compensated by those in particle scattering, so that the amplitude of reflectance is affected by these variations.