Transitional probability

Transitional probability is the conditional probability that a particular element follows another in a sequential stream — the statistical regularity that makes one syllable predict another, one note predict the next, or one event predict its successor. In language, transitional probabilities are high within words (the probability of "t" following "ca" in "cat" is near-certainty) and low across word boundaries (the probability of "the" following "cat" depends on context). This simple statistical signature is sufficient for infants to segment continuous speech into word-like units, a foundational result in the study of statistical learning.

The mechanism operates across modalities. Visual sequences, musical patterns, and tactile rhythms all contain transitional structure that learners extract without explicit instruction. The extraction is implicit — the learner shows no conscious awareness of the probabilities — but it shapes subsequent behavior, from recognition to prediction to production. Transitional probability is not the only statistical cue available to learners: co-occurrence frequency, mutual information, and distributional clustering all contribute. But transitional probability is the simplest, most studied, and most paradigmatic case of how raw statistical structure drives the emergence of structured knowledge.