Causal and conditional inferences: A comment on Cummins (1995)