Current perspectives propose that observer models accounting for both optimal and suboptimal behaviour may yield real progress in understanding perception. We propose that such models could, in addition, be very useful for precisely characterising the variation in perception across healthy participants and those affected by psychiatric disorders, as well as the effects of neuromodulators such as oxytocin.

A more general form of optimality approach applied to the entire behavioral paradigm should be used instead of abandoning the optimality approach. Adding the cost of information processing to the optimality criterion and taking into account some other recently proposed aspects of decision optimization could substantially increase the explanatory power of an optimality approach to modeling perceptual decision making.

Rahnev & Denison (R&D) argue against normative theories and in favor of a more descriptive “standard observer model” of perceptual decision making. We agree with the authors in many respects, but we argue that optimality (specifically, reward-rate maximization) has proved demonstrably useful as a hypothesis, contrary to the authors’ claims.

Optimal or suboptimal, Rahnev & Denison (R&D) rightly argue that this ill-defined distinction is not useful when comparing models of perceptual decision making. However, what they miss is how valuable the focus on optimality has been in deriving these models in the first place. Rather than prematurely abandon the optimality assumption, we should refine this successful normative hypothesis with additional constraints that capture specific limitations of (sensory) information processing in the brain.

Rahnev & Denison (R&D) argue that whether people are “optimal” or “suboptimal” is not a well-posed question. We agree. However, we argue that the critical question is why humans make suboptimal perceptual decisions in the first place. We suggest that perceptual distortions have a normative explanation – that they promote efficient coding and computation in biological information processing systems.

We agree with the authors that putting forward specific models and examining their agreement with experimental data are the best approach for understanding the nature of decision making. Although the authors only consider the likelihood function, prior, cost function, and decision rule (LPCD) framework, other choices are available. Bayesian statistics can be used to estimate essential parameters and assess the degree of optimality.

Rahnev & Denison (R&D) addressed the issue of (sub)optimalities in perception but only made a passing reference to evolutionary thinking. In our commentary, we concur with the authors’ claim that evolution does not work toward optimalities, but argue that an evolutionary perspective on perception questions the Bayesian approach that the authors adopted.

Although the suboptimality of perceptual decision making is indisputable in its strictest sense, characterizing the nature of suboptimalities constitutes a valuable drive for future research. I argue that decision consistency offers a rarely measured, yet important behavioral metric for decomposing suboptimality (or, more generally, deviations from any candidate model of decision making) into ultimately predictable and inherently unpredictable components.

According to Marr, explanations of perceptual behavior should address multiple levels of analysis. Rahnev & Denison (R&D) are perhaps overly dismissive of optimality considerations at the computational level. Also, an exclusive reliance on standard observer models may cause neglect of many other plausible hypotheses at the algorithmic level. Therefore, as far as explanation goes, standard observer modeling is no panacea.

We highlight that optimal cue combination does not represent a general principle of cue interaction during navigation, extending Rahnev & Denison's (R&D) summary of nonoptimal perceptual decisions to the navigation domain. However, we argue that the term “suboptimality” does not capture the way visual and nonvisual cues interact in navigational decisions.

The disagreements among commentators may appear substantial, but much of the debate seems to stem from inconsistent use of the term optimality. Optimality can be used to indicate sensible behavior (adapted to the environment), globally optimal behavior (fully predicted from optimality considerations alone), locally optimal behavior (conforming to a specific model), and optimality as an empirical strategy (a tool for studying behavior). Distinguishing among these different concepts uncovers considerable common ground in the optimality debate.