In overall, the tCJB test showed the expected results in the bias response to the ambiguous stimulus. No individuals were excluded from the study since they all reached the learning criterion. However, in the nCJB, the latency distribution did not differ across probe valences meaning that this measure might not be a suitable one and the valence did not influence the latency (GLM model 2).

A spill-over effect might have occurred when testing in a row the dogs with both tests. The dogs have learned to go a bowl because of the tCJB prior training and testing, possibly affecting the nCJB latencies.

These are suggestions for next studies investigating this possible new method.

• Increasing the cost to reach each picture.
• Using bigger pictures or screen displaying the entire body of the aggressive dog and smiling human to match more a real-life situation.
• Increasing the distance between the starting point and the picture/screen.
• Not doing the tCJB and nCJB on the same day. The dogs might have developed a learning bias, to quicker ignore the pictures and focus only on the treat.
• Perhaps inverting the tests order, starting by the nCJB, and finishing with the tCJB. Or testing on a different day/week to help avoiding any possible spill-over effect.
• Not rewarding every time during the nCJB. Maybe putting a treat only when the human face is displayed, and no treats with the dog face and ambiguous signal. By combining the nCJB with a little bit of the tCJB method, it might be possible to reach eye clear results.
• Having a speaker playing sounds matching the valence of the displayed face, since dogs appear to be able to detect if a sound is matching the valence of a conspecific or human face (Albuquerque et al., 2016)

It appears important to expose that some dogs are more “careful/shy” with not-known humans or dogs. So, the base line that these two cues suppositively represent (human face = attraction/positive; dog face = aversion/negative) might have been individually biased and therefore becoming useless as a comparison for when exposed to the ambiguous cue. As an anecdotical example, Polly appeared to be frightened by both the human picture and the dog picture and did not seem to know how to react in front of the morph; interestingly she also had the highest score for the Fearfulness factor trait compared to the other dogs. This within-individual difference might cause the nCJB test to not work as planned.

The nCJB did not seem to provide any solid or tangible results to be declared a working alternative method to a location ambiguity CJB test. But the tCJB did work as expected, with clear-cut results allowing immediate categorization of the bias (optimistic or pessimistic).

Perhaps by applying the suggestions, clearer results could be seen. Nevertheless, if these upgrades were being applied, the reasons for which this study was conducted would not be met; time saving, easy to implement for dog-owners and kennels.