When someone notices something in the presence of others they invoke a connection between their current cognitive/perceptual state and some noticeable state of affairs in the world. But what they notice, and what---if any---response this action makes relevant next depends on the detail of how this sequence of events is organized interactionally. Someone may notice something as just-now mutually visible, or that something is missing, or may notice the actions of others in very different ways. Conversation analysts have tended to eschew definitions of noticings as prospective, paired sequences of action-in-conversation because there is little evidence that people design noticings as recognizably initial actions that routinely occasion noticing-specific responses in a subsequent turn. Noticings are difficult to analyze as such because their interactional usefulness often hinges on participants' situated inferences about tacit cognitive/perceptual or environmental states. Without the endogenous evidence produced by participants' routine patterns of initiation and response, sequential analysis may lack empirical access to people's ascriptions about what may be noticeable to others. This paper analyzes how people work interactionally to manage ambiguous ontological/epistemic states of affairs by comparing their practices of noticing between two perspicuous settings. Firstly, we see how geologists engaged in fieldwork in a wilderness landscape notice things in structurally similar ways to how gallery visitors first notice an unconventional artwork in a public exhibition. Their practices suggest that noticings as actions-in-conversation are structured retro-sequentially and are therefore only reflexively accountable and analyzable in retrospect. Both groups use noticings as contingent epistemic formulations of relevant states of affairs in the world, which they design to remain equivocal and therefore open to varying degrees of joint ontological reformulation. This supports skepticism about defining noticings as paired sequences of prospective actions. Instead, we argue that noticings are only ever reflexively accountable as such.