The pattern of the spatial-temporal distribution of the fishing yields (catch rates) of swordfish (Xiphias gladius) from the Chilean industrial longline fleet it was determined using generalized additives models, autocorrelation analysis and Fourier spectral analysis. The basic information came from fishing logbook recorded by scientific observers between January 2001 and December 2005. The generalized additives models included five physical variables (latitude, longitude, date, lunar index and sea surface temperature (SST)), which affect the availability and vulnerability of swordfish, and two operational variables (length of the vessel and type of longline), which are directly associated with the effectiveness of the fishing system. The non-linear effects were established significantly (p < 0.01) for each of the independent variables. There was a typical annual cycle in the catch rates, with high values from March to July/August, and displacement of the fishing operation northward from the 38° to 32°S, in a SST range of 17° to 19°C. Subsequently, catch rates decrease in a northerly direction from 32°S in SST higher than 20°C. This spatial-temporal pattern was determinates with a high square coherency (79%) in the time series of nominal catch rates and latitudinal component analyzed in the Fourier spectral analysis. It also established for the time series of swordfish nominal catch rates, low-frequency fluctuations with periods of 28, 38 and 59 days. This intraseasonal variability in swordfish catch rates coincides with changes in mesoscale oceanographic conditions in the area where the fishery is developed off the coast of Chile.