Gravar-mail: Heterogenising study samples across testing time improves reproducibility of behavioural data