Attentional control deficits are often invoked to explain individual or aging-related differences in how people allocate attention. Yet what constitutes optimal attentional allocation, and how closely human behavior approximates it, remains poorly understood. We introduce a novel paradigm designed to test how people select attentional strategies in a prototypical allocation context. In each trial, participants could maximize piece-rate earnings by focusing on a primary task or, at a cost to earnings, check cues indicating which of two primary tasks was currently rewarded. Normative modeling of allocation/value functions revealed wide variability in optimal strategies across decision contexts, but also relatively flat utility surfaces, implying weak normative constraints. The relative time cost of checking the cues emerged as a key normative determinant of allocation differences across individuals and age groups. Empirically, participants were less sensitive to most decision variables that the model predicted, yet they showed pronounced individual and aging-related differences. Consistent with normative predictions, these differences were largely explained by variations in relative cue-checking costs: Because older adults were slower on the primary task, but not in cue checking, their effective cue-checking costs were smaller. This lead to more frequent cue use. We argue that a normative framework clarifies when attentional differences reflect adaptive, cost-sensitive optimization rather than control deficits, and outline testable implications for further research on how people allocate attention.