Animal Behaviour / Trial designer
Design a test
for what a dog learned.
Use the study’s coloured cards to plan a test. Choose which cards go left and right. See whether the choices could tell apart following a colour, choosing the darker card or sticking to one side.
Loading the designer. Controls and setup links unlock when ready. Your linked plan stays in the address if loading fails.
Your trial plan
The proposed cards
Layout only · no animal simulatedEach marker names a rule that permits that side. “Either” appears on both sides. Card colours are uncalibrated screen illustrations; the written cue labels define this exercise.
Edit the selected trial
Which cards go where?
A swap can change what the rules predict. Repeating the same layout adds no new logical contrast in this exercise. Real studies can still need repetitions to assess variability.
Read the predictions
Same choice.
Different explanations.
Read across each row: what would that rule allow in every trial? Two rules are forced apart only if at least one trial requires opposite choices. “Either” leaves both choices open; it does not mean a 50/50 chance.
Forced apart
Still overlap
Change the explanations
What are you trying
to distinguish?
Keep at least two rules. These are deliberately limited possibilities, not a complete account of learning.
Try an imagined result
If the choice were…
Supply a hypothetical choice for the selected trial. This checks which listed rules still allow your pattern. It never records or generates an animal observation.
Compatible with supplied choices
Why this matters
A successful choice can
hide an ambiguous test.
Someone checking what an animal learned needs alternatives to predict different things. Swapping positions helps separate a card preference from a side preference. In a different setting, the same design question arises when you change a product’s colour and price together: which change caused the response?
This designer answers a narrow question before any observations: can these exact rules be forced to disagree using these four cards? It supplies no error model, confidence, required sample size or conclusion about a real animal. The original study notebook separately explains the published eight-dog result.
Inspect the rules, possible patterns and exact search
A finite design problem.
The palette contains dark yellow, light blue, light yellow and dark blue. Each trial offers two different cards, in order. That gives twelve possible layouts. A plan holds one to six trials.
The possible response patterns
L means left, R means right. Where a rule allows either side, both possibilities are retained. No tie-break, fixed probability or independent-dog assumption is supplied.
How the next trial is chosen
The search checks every subset of the unused layouts and finds the smallest addition that forces every listed pair of rules apart. It searches the declared palette, not all possible experiments. Multiple equally small additions can exist. Repeating an existing layout cannot add a new forced contrast.
What a hypothetical mismatch means
With no errors allowed, one incompatible choice excludes a listed rule. If every rule is excluded, your pattern is outside this limited rule set. That is not a verdict that an observation is wrong. A real study would need its own design, welfare review, measurements and analysis.