How to “mix” discriminants/control things into your one sample without containing it?

Any good ideas for avoiding over discrimination/specification with limited samples?

Should we swab the storage rooms?

Any ideas for promoting generalisation across individuals, stages, phases with very few samples?

Testing for generalisation vs. specification before imprinting on target /deploying?

Is discrimination training something they always do as a training step before fieldwork? Why? Or is it something they do if there is a challenge in fieldwork? And especially: what is their preferred method to do the training? Line-ups or go - no-go or search type scenarios? Why?