The biological units-of-selection debate has centred on questions of which units experience selection and adaptation. Here, I use a causal framework and the Price equation to develop the gene’s eye perspective. Genes are causally special in being both replicators and interactors. Gene effects are tied together in a complex Gouldian knot of interactions, but Fisher deployed three swords to try to cut the knot. The first, Fisher’s average excess, is non-causal, so not fully satisfactory in that respect. The Price equation highlights Fisher’s other two swords, choosing to model only selection, and only the part that is transmissible across generations. The models developed here show that many causes of organismal fitness do not cause Gouldian complications. Phenazine methosulfate Only two kinds of elements must be added to the focal gene for a causal explanation of its selective change co-replicators that are associated with the focal gene and co-interactors that interact non-additively with the focal gene. Identical equations for co-replication and co-interaction describe interactions between gene copies at a single locus or at separate loci, and also for genes situated within the same individual or in different individuals. These results resolve some of the objections to the gene’s eye view. This article is part of the theme issue ‘Fifty years of the Price equation’.Though the Price equation in itself is simply a statistical identity, biologists have often adopted a ‘causal interpretation’ of the equation, in the sense that its component terms have been supposed to correspond to distinct causal processes in evolution, such as natural selection and transmission bias. In this paper, we bring the issue of causal interpretation to the fore, by studying the conditions under which it is legitimate to read causal meaning into the Price equation. We argue that only if substantive assumptions about causal structure are made, which can be represented in the form of a causal model, can the component terms of the Price equation be interpreted as causally meaningful. We conclude with a reflection on the epistemic uses of the Price equation, emphasizing the difference between the description, explanation and prediction of evolutionary change. This article is part of the theme issue ‘Fifty years of the Price equation’.By the Robertson-Price identity, the change in a quantitative trait owing to selection, is equal to the trait’s covariance with relative fitness. In this study, we applied the identity to long-term data on superb fairy-wrens Malurus cyaneus, to estimate phenotypic and genetic change owing to juvenile viability selection. Mortality in the four-week period between fledging and independence was 40%, and heavier nestlings were more likely to survive, but why? There was additive genetic variance for both nestling mass and survival, and a positive phenotypic covariance between the traits, but no evidence of additive genetic covariance. Comparing standardized gradients, the phenotypic selection gradient was positive, βP = 0.108 (0.036, 0.187 95% CI), whereas the genetic gradient was not different from zero, βA = -0.025 (-0.19, 0.107 95% CI). This suggests that factors other than nestling mass were the cause of variation in survival. In particular, there were temporal correlations between mass and survival both within and between years. We suggest that use of the Price equation to describe cross-generational change in the wild may be challenging, but a more modest aim of estimating its first term, the Robertson-Price identity, to assess within-generation change can provide valuable insights into the processes shaping phenotypic diversity in natural populations. This article is part of the theme issue ‘Fifty years of the Price equation’.The Price equation has found widespread application in many areas of evolutionary biology, including the evolutionary epidemiology of infectious diseases. In this paper, we illustrate the utility of this approach to modelling disease evolution by first deriving a version of Price’s equation that can be applied in continuous time and to populations with overlapping generations. We then show how this version of Price’s equation provides an alternative perspective on pathogen evolution by considering the epidemiological meaning of each of its terms. Finally, we extend these results to the case where population size is small and generates demographic stochasticity. We show that the particular partitioning of evolutionary change given by Price’s equation is also a natural way to partition the evolutionary consequences of demographic stochasticity, and demonstrate how such stochasticity tends to weaken selection on birth rate (e.g. the transmission rate of an infectious disease) and enhance selection on mortality rate (e.g. factors, like virulence, that cause the end of an infection). In the long term, if there is a trade-off between virulence and transmission across parasite strains, the weaker selection on transmission and stronger selection on virulence that arises from demographic stochasticity will tend to drive the evolution of lower levels of virulence. This article is part of the theme issue ‘Fifty years of the Price equation’.For decades, parts of the literature on human culture have been gripped by an analogy culture changes in a way that is substantially isomorphic to genetic evolution. This leads to a number of sub-claims that design-like properties in cultural traditions should be explained in a parallel way to the design-like features of organisms, namely with reference to selection; that culture is a system of inheritance; and that cultural evolutionary processes can produce adaptation in the genetic sense. The Price equation provides a minimal description of any evolutionary system, and a method for identifying the action of selection. As such, it helps clarify some of these claims about culture conceptually. Looking closely through the lens of the Price equation, the differences between genes and culture come into sharp relief. Culture is only a system of inheritance metaphorically, or as an idealization, and the idealization may lead us to overlook causally important features of how cultural influence works. Design-like properties in cultural systems may owe more to transmission biases than to cultural selection.