Comparison of time-based frequencies, fractal analysis and T-patterns for assessing behavioural changes in broiler breeders fed on two diets at two levels of feed restriction: A case study

Three methods of describing behavioural activity - conventional statistics (number of bouts, mean bout length and total duration), detrended fluctuation analysis (DFA) and T-patterns (THEME) - were compared. Behavioural observations were obtained by focal sampling of feed restricted broiler breeders. Hybrid commercial females were fed to the commercial body weight target (SR) or to gain 40% more live weight (SE) during rearing. Half the birds were fed on a conventional diet (C) and half on a diet with more crude fibre (F) in a 2 x 2 factorial experiment. Videotapes of the behaviour of the birds in each pen were recorded for I h before feeding (B) and in the afternoon (A). Focal behavioural sampling was performed for 600 s on a single bird in each videotape and conventional data (number of bouts, mean and total duration), T-patterns and fractals (from DFA) were analysed by ANOVA of a factorial model with a split plot term for time within pen. SR had longer bouts of preening than SE. SE had more T-patterns and spent more time on them compared with SR whereas no differences were detected by DFA. Diet F increased drinking and pecking at the environment and decreased time standing compared with C. Birds fed on C had longer bouts and more time preening compared with F. C was associated with longer and more frequent litter scratching of similar bout length than F before feeding but the opposite was observed in the afternoon. No differences between the diets were detected by THEME or DFA. The conventional analysis showed that more time was spent drinking in B compared with A. More T-pattern bouts of longer mean duration and longer total duration were detected in B compared with A whereas no differences were identified by DFA. These results may reflect the anticipation of feed. We conclude that a technique to assess behavioural organisation such as THEME or DFA in addition to conventional analysis of focal behavioural sampling should be used in behavioural and welfare research. (C) 2006 Elsevier B.V. All rights reserved.