INTERACTIVE COMPANIONFIELD NOTEANIMAL BEHAVIOR

The Partner-Choice Explorer

A live model accompanying “Killer Whales Are Manufacturing Grooming Tools, Mostly With Family”

Study: Field note
Review: Peer-edited by the club review board
Article: 13 sectionsabout 20 min
Files: LaTeX source16 kB

Weiss and colleagues fitted a Bayesian model relating two whales' odds of allokelping together to how related they are and how close in age they are, and reported which effects came back strong and which came back weak or absent [see the article, §8]. They did not publish the model's exact coefficients, intercept or link function — only the direction and posterior strength of each effect. So the calculator below is not a reproduction of their model. It is the club's own toy score, built to have the same ranking of effect strengths the paper reports — kinship strongest, age similarity close behind, skin condition weaker and tentative — so you can feel how those three real findings interact, without us pretending to know numbers the paper never gave us.

Everything below labelled "illustrative" is the club's own construction. The four numbers in the grey box under each model are the paper's real, reported statistics, unmodified.

Model 1. Would These Two Whales Groom Together?

Relatedness between the pair,r0.50r = 0.5 is a mother–calf pair or full siblings; 0.25 a half-sibling or grandparent; 0.125 a first cousin; 0 unrelated. These are standard relatedness coefficients, not values the paper reports per pair.Age gap between the pair, years0One of the pair has visibly moulting or dead skin

Illustrative partnership score: -Driven mostly by: -

Maternal kinship: β ± SE = 0.91 ± 0.21, posterior P(β>0) > 0.99 — strongest effect
Age similarity: posterior P(β>0) = 0.98 — nearly as strong
Skin moulting: posterior P(β>0) = 0.93 — weaker, called "tentative" by the authors
Age on its own: posterior P(decline) = 0.53, no support — this is why there is no plain "whale's age" slider above, only an age gap

Slide relatedness down to zero and age gap up past twenty years, with no moulting, and the score collapses toward the bottom of the scale — the toy model's way of saying that an unrelated pair, a generation apart, with clean skin, matches none of the three correlates the paper found. Push relatedness to 0.5 and the age gap to zero and it maxes out on kinship and age similarity together, which by the model's own ranking is the combination the real data most favoured.

Model 2. Where Would Your Guess Land?

The paper reports a median and a maximum bout length, not a full distribution. This model does not invent one. It just places a guess of yours on the same number line as those two real figures.

Guess a bout length1:19

Your guess vs. the real median: -Your guess as a share of the longest bout on record: -

What the two models are, and are not

Model 1 turns three real, directionally confirmed correlates into a single illustrative number so their relative strength is easier to feel than three separate posterior probabilities are. Model 2 does no modelling at all; it is a ruler laid against two numbers the authors already published. Neither model can tell you whether any specific pair of real whales groomed together, and neither should be quoted as the paper's own statistic. Read the article for what was actually measured.