Observe animal reactions and stress signals

Write an observation form to record dog stress signals during training trials. Start with a one-line purpose, then list five observable behaviours to note (lip licking, yawning, tail tucked, whale eye, panting), a severity scale from 0 to 3, space for context (time, handler, trigger), and a short instruction to the observer to stop the trial if severity hits 3. Make it usable by Jenna in the clinic tomorrow morning.

Before I distribute this form to the animal behaviour team, make it easy for reviewers to act on: put the top three high-risk signals first, add a short 'what to do at level 2' line, bold the stop condition, and flag anything that could make a reviewer hesitate (ambiguous signal definitions or missing inter-observer calibration).

In yesterday’s trial Jenna wrote that a lip lick looked like a tongue flick and marked different severities between observers. The handler, my colleague Amir, and I must decide whether to retrain observers or change the form. I’m worried inconsistent scoring invalidates the session data. What’s the likely diagnosis and the best next move?

Across projects we lose confidence in behavioural data because observers interpret signals differently, forcing repeats. What single change to our observation routine — a two-page calibration guide, a short video training, or paired scoring for the first three sessions — will most quickly and reliably align observers?