An empirical test of signal detection theory as it applies to Batesian mimicry

Imperfect Mimicry and the Limits of Natural Selection

“…All of the predictions of this hypothesis have been confirmed in empirical systems. Many studies have documented selective surfaces that correspond to those outlined in Figure 2: animals or humans trained to respond to artificial prey (or flowers, in the case of plants) reduce attack rates nonlinearly as mimics approach models in phenotype (Schmidt 1958;Duncan and Sheppard 1965;Ford 1971;Caley and Schluter 2003;Lynn et al 2005;McGuire et al 2006). Furthermore, changing the relative abundance of models and mimics alters the amount of phenotypic space in which imperfect mimics receive protection: the precision of coral snake mimicry by scarlet kingsnakes (and selection for better mimicry) increases across the kingsnake's range as the abundance of coral snakes decreases (Harper and Pfennig 2007;Kikuchi and Pfennig 2010b), and the mimetic precision of andromorphs (malemimicking females) increases with the proportion of andromorphs:males in damselfly populations (females resemble males to escape sexual harassment; Iserbyt et al 2011).…”

“…An empirical study with human subjects selecting between two species of computergenerated prey was able to produce chaseaway, but the difference between models and mimics was quite small relative to the range of potential phenotypic difference (McGuire et al 2006). We are unaware of any study that adequately demonstrates a model evolving away from its mimic in nature, or resultant imperfect mimicry.…”

Which traits do observers use to distinguish Batesian mimics from their models?

“…Many theoretical predictions regarding predator behavior towards models and mimics are based on Signal Detection Theory (Green and Swets 1966) and assume that the predator has access to global information about the prey populations and can dedicate unlimited cognitive resources to the problem (e.g., Getty 1985;Sherratt 2002). There is experimental evidence to support the validity of models like these, but it tends to come from scenarios that include a limited range of simple stimuli (Lindström et al 1997;McGuire et al 2006). There is an increasing appreciation that, to understand mimicry, we must take into account the complexity of the prey community (Easley and Hassall 2014) and the limited knowledge of the predator .…”

“…Humans provide a useful model "predator", with fewer ethical and practical constraints, and experimental games in which humans "forage" for prey have previously generated large datasets to explore hypotheses about mimicry (Golding et al 2005b;McGuire et al 2006;Sherratt et al 2015;Morris and Reader, in review). Humans have been shown to make judgments about mimetic accuracy similar to those of birds, which are thought to provide the main selective pressure for mimicry in hoverflies (Dittrich et al 1993;Penney et al 2012;Sherratt et al 2015).…”

Two Models of Predator Decision Making Under Uncertainty

“…The classical model of optimal discrimination between classes of stimuli is signal detection theory (Green and Swets 1966), which has been used extensively in modelling problems in behavioural ecology (Getty 1985;Wiley 1994;Fawcett and Johnstone 2003;Holen and Johnstone 2004;McGuire et al 2006;Abbott and Sherratt 2013). Some work has also been done on models of speed-accuracy trade-offs alone (Edwards 1965;Froment et al 2014;Gendron and Staddon 1983;Palmer et al 2005), and on combining speed-accuracy trade-offs with signal detection theory, including simple models in which samples have constant costs (Getty 1996), models which incorporate optimal foraging to maximize energy intake (Holen 2013;Getty 1985), and incorporating opportunity costs of continued examination (Abbott and Sherratt 2013).…”