The binary cumulant is defined fur joint probability distributions on binar
y sequences of finite length. Tnt: binary cumulant is bounded, in magnitude
, by unity, and is shown to vanish if there exists any bipartition of the l
etter positions into statistically independent blocks. Probability distribu
tions on binary n-sequences are shown to map injectively into their binary
cumulants for all subsets of the set of letter positions. An inversion algo
rithm is established, recovering the joint distribution from its collection
of binary cumulants. (C) 2000 Academic Press.