S U M M A R YWe have carried out two seismic physical experiments to acquire wide-azimuth P-wave 3-D seismic data with a scaled down model (1:10 000) and scaled-up frequencies (10 000:1). Our aims are to verify the physical basis of using P-wave attributes for fracture detection, to understand the usage of these attributes and their merits, and to investigate the effects of acquisition geometry and structural variations on these attributes. The base model consists of a fractured layer sandwiched between two isotropic layers (Epoxylite). Inside the fractured layer there is a dome and a fault block for investigating the effects of structural variations. The two experiments were carried out using different acquisition geometries. The first experiment was conducted to maximize the data quality, with an offset-depth ratio of only 0.68 to the bottom of the fracture layer. For comparison, the second experiment was carried out to maximize the anisotropy effects, with the offset-depth ratio to the bottom of the fracture layer raised to 1.34.For each experiment, about 20 km 2 of wide-azimuth 3-D data were acquired with a P-wave source. The physical modelling confirms that the P-wave attributes (traveltime, amplitude and velocity) exhibit azimuthal variations diagnostic of fracture-induced anisotropy. For the first experiment with noise-free data, the amplitude from the top of the fracture layer yields the best results that agree with the physical model parameters and free of the acquisition footprint. The results from other attributes (traveltime, velocity, AVO gradient) are either contaminated by the structural imprint, or by the acquisition footprint due to the lack of offset coverage. For the second experiment, despite the interferences from multiples and other coherent noise, the traveltime attributes yield the best results; both the acquisition footprint and the structural imprint are reduced due to the increased offset coverage. However, the results from the amplitudes are affected by the noise and are less reliable. Analysis of the two experiments reveals that the offset-depth ratio to the target is a key parameter for the success of the P-wave techniques. Smaller offset-depth coverage may only be applicable to amplitude attributes with high quality data; whilst large offset coverage makes it possible to use traveltime attributes. A reliable estimation from traveltime attributes requires an offset-depth ratio of 1.0 or more.