This paper analyses the logical structure of personal construct systems in terms of relations of partial entailment between constructs and the relative frequencies of positive and negative judgments concerning both self and others. It is argued that conditional hypotheses (‘anticipations’) which incorporate the positive poles of constructs (e.g. happy) as antecedent terms will have wider ranges of relevance, on average, if self is assigned to the positive poles of those constructs. Conversely, hypotheses with antecedents based on the negative poles of constructs (e.g. sad) will tend to have wider ranges of relevance when self is assigned to their negative poles. Some theoretical implications of these relationships for the adaptability of personal construct systems are elaborated.