A hidden semi‐Markov model (HSMM) has previously been used successfully to model breakpoint rainfall data which provide a continuous record of steady rainfall rates and the times of rain rate changes. These data contain more information than the commoner fixed period rain accumulation data. Although the HSMM is a homogeneous model which does not explicitly account for seasonality, it does yield firm state classifications in the case of New Zealand breakpoint data and these were used to explore the nature of seasonality in the breakpoint data. These data and the state classifications were analyzed using simple graphical and numerical procedures. It was found that seasonality resides in the frequency of rainfall events rather than their intensity and that seasons start at varying rather than fixed times of the year and have varying length. Such a seasonal pattern is not well modeled by conventional periodic functions and further developments to the HSMM to include these features are discussed.