Emotional and interactional cues are often treated as if they require a special kind of intelligence or a special kind of mystery. This paper argues for a narrower and more practical view. Systems already handle weak physical signals through bounded representation, local comparison, provisional interpretation, and governed escalation. Interactional cues can be treated the same way. Timing shifts, wording changes, tone mismatches, user self-report, and other local signals need not be treated as either meaningless noise or hidden truth. They can be represented as mixed partials, checked against short-horizon expectations, and escalated only when deviation becomes informative. The point is not machine feeling, mind reading, or a full theory of emotion. The point is architectural discipline: represent first, escalate attention second, and require stronger evidence before persistence or authority changes. This short bridge paper introduces a compact model for treating interactional cues as bounded mixed partials. It outlines operational primitives for escalation, non-persistence, evidence requirements, and scope discipline, then grounds the approach through brief physical-signal and interactional examples.