Records on final BW (kg), backfat depth (cm), and LM area (cm(2)) of pigs from a University of Nebraska Large White/Landrace composite population were analyzed to estimate the effects of pen mates. Measurements were at approximately 180 d of age for 3,524 pigs in 351 pens (9 to 11 pigs per pen) farrowed from 1999 to 2005. The area of each pen was 8.13 m(2). The full model (M1) included the fixed effects of contemporary group, sex, line, and the covariates of age and inbreeding coefficient, and included random direct genetic, genetic pen-mate, permanent environmental, pen, litter, and residual effects. A derivative-free algorithm was used to obtain REML estimates of variance components for final BW adjusted to 180 d of age with M1 and 7 reduced models, and with 4 reduced models for the carcass traits. For final BW, likelihood ratio tests showed that M1 did not fit the data better than model 2 (permanent environmental effect omitted from M1) or model 3 (pen omitted from M1). Model 2 was not significantly (P > 0.05) better than model 3, which shows that variance attributable to pen effects and permanent environmental effects cannot be separated. Large sampling variances of estimates of the pen component of variance for models with pen-mate effects also indicate an inability to separate pen effects from the effects of pen mates. When pen-mate genetic effects were not in the model, estimates of components of variance and the fit of the data were the same for models 4 (included both permanent environmental and pen effects), 6 (included pen effects), and 7 (included permanent environmental effects), which shows that including both pen and permanent environmental effects was no better than including one or the other. Models 4, 6, and 7 were significantly better than model 8, which did not include pen-mate effects and pen effects, implying that pen effects are important. The estimate of pen variance with model 2 was approximately (number of pen mates - 1) times the estimate of variance of pen-mate permanent environmental effects with model 3. Patterns of estimates of variance components with models 2, 5, 6, and 8 for backfat depth and LM area were similar to those for final BW. Estimates of direct genetic variance and phenotypic variance were similar for all models. Estimates of heritability for direct genetic effects were approximately 0.40 for final BW, 0.45 for backfat depth, and 0.27 for LM area. Estimates of heritability for pen-mate genetic effects were 0.001 for the 3 traits for models including either pen or permanent environmental effects. Under the management conditions for this experiment, the conclusion is that the model for genetic evaluation should include litter effects and either pen effects or pen-mate permanent environmental effects and possibly genetic pen-mate effects, in general agreement with the results of studies of different populations at other locations.