Daily feed intake (DFI) was described as a function of days on test using a spline random regression model. Order of fit for the spline random regression model was varied, models were compared using Schwarz's Bayesian Information Criterion. The objective was to investigate whether there are genetic differences in DFI patterns. Data were DFI records of 257 growing gilts, which had ad libitum access to food. Gilts were 94 days old at start of test. All gilts were from a Dutch Landrace line. Number of random regression coefficients was varied from zero to nine for both random effects (animal and permanent environment), the model with four random regression coefficients resulted in the most optimal fit. Six traits were derived to capture differences in DFI patterns, average DFI over test, variance of DFI over test, DFI at day 5 on test, DFI at day 50 on test, DFI at day 95 on test and the difference between DFI at day 95 and day 5. Heritability for DFI decreased from 0.53 at 5 days on test to 0.24 at 95 days on test. Genetic correlations between DFI at 5 days on test and 95 days on test, and between DFI at 50 days on test and 95 days on test were low (around 0.3). The results suggested that DFI at different days on test could not be regarded as repeated measurements of a single trait with constant variance and heritability. Based on the results presented, it can be concluded that changes in feed intake patterns through selection are possible. (C) 2003 Elsevier B.V. All rights reserved.