Previous discussions of observation selection effects (OSEs) have ignored the distinction between observation and evidence. Evidence for a hypothesis, I argue, is distinct from the observation of that evidence. This shows that the fact that evidence is unobservable does not entail that the evidence does not obtain. What is required for an OSE is that evidence is guaranteed, not that counter-evidence is unobservable. With the evidence-observation distinction in hand, apparent counterexamples fail. I then show that observer perspective can change whether or not an agent is subject to an OSE, even when knowledge is shared between perspectives.

This article proposes an epistemic account of complex phenomena. It first reviews and then develops some challenges for the recent and influential ontic framework of complex systems developed by James Ladyman and Karoline Wiesner. It next sketches an epistemic account of complexity, shows how the account complements rather than replaces Ladyman and Wiesner's framework, and shows how the account dissolves the challenges. It closes with a case study indicating how the account can be used to explain scientists multi-model strategies.

The possibility of testing contradictory statements about the factual world has been suggested but barely discussed in the relevant literature. In this paper, I argue that if we assume that there are contradictory observation sentences, it would be logically impossible to falsify them. Accordingly, the extension of the dialetheist programme into empirical science would be non-advisable for it would introduce logically unfalsifiable claims.