Lexical co-occurrence and association strength

The 1-million-word Brown corpus was searched for co-occurrences of semantically related pairs of concrete nouns appearing within an arbitrary window of 250 characters. Related pairs of nouns (OCEAN-WATER) co-occur significantly more often than matched, unrelated pairs (OCEAN-HAND), and this difference remained significant within blocks of text up to 1000 characters in length. Frequency of co-occurrence, corrected for chance, is significantly correlated with association strength. Lexical distance between co-occurring members of a given pair is inversely correlated with association strength. Significantly more co-occurrences were found, per unit text, in the fictional sections of the corpus.