6.3. The hidden record in an exactly soluble case
Set in Eq. (6.1). The two variables execute a damped rotation:
Two preparations with the same but different generally have different later values of . Therefore alone is not a sufficient completed state for the family of future measurements of . The omitted variable is a physical record of the preparation, even when the observer cannot measure it directly.
At , the full motion is reversible and is conserved. Nevertheless, eliminating leaves a memory integral in the equation for . Memory is thus compatible with reversible dynamics. When , the squared radius decays as because the model already contains damping. The damping and the memory have distinct origins in the calculation.
This distinction is useful when interpreting temporal asymmetry. A reduced state can forget information because its definition omits a variable; the full state can dissipate because its law couples it to an unmodelled reservoir; and an observer can infer an arrow because a special preparation makes some histories more probable than their reverses. A concrete model should identify which of these mechanisms is operating.
The same exact solution also shows where a simple Markov approximation fails. A first-order decay law for alone cannot reproduce the oscillatory return of information for arbitrary . It may still approximate a restricted preparation over a suitable time window. The quality of the approximation is determined by preparation, time-scale separation, and the observables to be predicted.


