The claimed mechanism turns on how much selection varies, rather than how strong it is. A target that sits still lets a trait arrive at it and then holds it there, and little further change accumulates along the branch. A target that wanders keeps dragging the trait after it, and each wander is recorded as change. Allen's account puts random drift in female preference in the role of the wanderer, and comparatively stable male-male competition in the role of the target that mostly stays put [6].
That asymmetry has a reason behind it in the model's structure. Ornaments sit inside a feedback loop with the preferences that judge them, so trait and preference can coevolve [3]. The Fisher process, which Allen describes as positive feedback between female preference and male ornaments, does not operate the same way on weapons [8], whose evolution the framework attributes mainly to competition between males [3]. The prediction traces back to McCullough, Miller and Emlen, who argued that weapons are not ornaments; according to Allen, nobody had tested it before [7].
Here is how the test worked. Rates were estimated from phylogenetic data across ten animal groups, with primate canine length and stalk-eyed fly eye span named as examples [4], which leaves eight groups unnamed in this account [11]. Ornaments came out faster, in the direction the model predicts [5]. The phys.org account does not give an effect size or say how traits measured in millimetres of canine were put on a common scale with plumage or eye stalks [10].
A rate comparison only shows that fluctuation happened, not which source produced it. Preference drift is one way to make selection wobble; anything else that changes what a female can see or compare would do the same, and phylogenetic rate estimates cannot separate them, because they measure how much a trait moved and not why. So the model supplies the mechanism and the comparative data supply consistency with it, which is a weaker thing than a demonstration.
My read: the mechanism is the durable contribution here, because it converts a verbal prediction into something with parameters someone else can break. The comparative leg is the softer one until the magnitudes are on the table. Ten groups is enough to establish a direction and not enough to make it a law, and a trait evolving fast is a statement about tempo, not about fitness.
