For all models (both stationary and approaching radar experiments), we used the Kenward-Rodgers degrees of freedom estimation method and restricted maximum likelihood estimation method. We checked all variables for normality, and log transformed those variables that were not normal. All results presented are the untransformed least squares means ± standard error. For the independent variables of time period (before, during, and after radar exposure) and radar treatment (radar off, magnetron radar, and solid state radar), we used pairwise comparisons (t-tests) to determine differences between treatments. We used a value of P = 0.05 as our significance threshold.