The probability density function for the CPUE data is important to estimate possible outcomes of fishing events and as auxiliary information for the application of generalized linear models (GLM) in fisheries science. The negative binomial model fitted well CPUE frequency distributions of the pole and line skipjack tuna (Katsuwonus pelamis) fishery off southwest Atlantic, and is an alternative model to enable applications of GLM for this fishery. Since the normal distribution has not shown to be suitable for the CPUE distribution, expected fishery outcomes estimated from average calculations are biased and unreal.