Recent research suggests that people’s explanations of complex patterns tend to oversample easily-accessible inherent properties of the phenomena they aim to explain (Cimpian & Salomon, 2014). However, research on this inherence bias has not directly examined the contents of inherent explanations people produce. Prasada and colleagues (2013) present a novel conceptual framework that distinguishes between two types of connections between kinds and properties: principled and statistical connections. For example: tanginess is principally connected to orange juice, whereas coldness is statistically connected to orange juice. Principled connections license normative expectations (e.g., orange juice should be tangy, or else it might not be orange juice), whereas statistical connections do not (e.g., orange juice needn't be cold to be considered orange juice). Here, we examine whether people reason that the behavior of novel scientific entities is best explained by principally or statistically connected inherent properties. We predict that even when, for instance, a novel chemical is reported to have behaved anomalously, people explain these patterns in terms principally connected inherent properties.