Perception of statistical relations often deviates from mathematically correct values: individuals infer a statistical relation when there is none, find a relationship that has the opposite sign of the genuine contingency, or ignore relevant third variables in the observed sample. Contingency assessment is based on sampling and integration of information, whether sampled from memory or from the environment. Information sampling in turn may depend on various factors, including the valence of events observed, direct consequences of sampling a piece of information, the amount of information given per sample, or prior knowledge and beliefs. Whenever individuals sample information, such factors come into play and may lead to asymmetries in the available information. For instance, more positive events than negative events might be observed, more instances of one alternative as compared with another, more information about the ingroup than the outgroup, and so on. Such asymmetries or skewed frequencies might provide the basis for pseudocontingency inferences. According to the pseudocontingency heuristic, contingencies are inferred on the basis of observed marginal frequencies: When observing skewed marginal frequencies of two variables (e.g., option A more frequently than option B and gains more often than losses), individuals relate the frequent observations per variable with each other (e.g., option A and gains) as well as the infrequent observations (e.g., option B and losses). In this chapter, we review empirical evidence on the influence of information sampling on the inference of genuine and pseudocontingencies. We provide some answers to the following questions: Which information is sampled? When do reasoners use marginal frequencies in order to form a judgment and thereby infer a pseudocontingency? May information sampling foster genuine contingency assessment? And what part do expectations or prior beliefs play?