Direct numerical simulation is used to investigate the nature of pressure fluctuations induced by surface roughness in a turbulent channel flow at Re = 400 for three-dimensional periodic roughness elements, whose peaks overlap approximately 25% of the logarithmic layer. The three-dimensional roughness elements alter the pressure statistics significantly, compared to the corresponding smooth-wall flow, in both the inner and outer ͑core͒ regions of the channel. The direct consequence of roughness is an increased form drag, associated with more intense pressure fluctuations. However, it also alters the pressure fluctuations in the outer layer of the flow, and modifies the length scales defined by two-point correlations. We also find that the depth of the roughness sublayer defined by the pressure fluctuations is very different from that given by the large-and small-scale statistics from the velocity field. © 2007 American Institute of Physics. ͓DOI: 10.1063/1.2482883͔ This paper focuses on the influence of surface roughness on pressure fluctuations for a better understanding of inner/ outer layer interactions in wall-bounded turbulence. We consider incompressible turbulent flow in a plane channel, the lower wall of which is covered by a regular array of threedimensional "egg-carton"-shaped elements. 1 The elements extend well above the viscous sublayer of the equivalent smooth-wall flow, with their peak-to-valley height h given by h + = hu s / = 21.6 ͑based on the friction velocity from the smooth-wall side of the channel, u s ͒, corresponding to h = 0.054␦ ͑␦ is the half-height of the channel͒. The roughness is prescribed via a virtual no-slip surface whose mean height is at y = −0.96␦, using an immersed boundary technique. For the Reynolds number used here, Re = 400, the halfwidth-toroughness ratio is ␦ / h Ϸ 18.5. Assuming the top of the logarithmic layer reaches to about 0.2␦, this implies that the peaks of the roughness elements extend approximately onequarter of the way into the logarithmic region. As a consequence, the roughness can be expected to directly affect the bottom of the logarithmic layer but not completely destroy it. This case is relevant in a number of engineering and especially meteorological contexts.Much remains to be learned about rough-wall boundary layers. In light of existing experimental results, it now appears that rough-wall boundary layers can be categorized according to whether the surface roughness does not affect the outer layer 2,3 or if it does affect the outer layer.