Long before the attention economy, a century of psychology studied attention as a limited, effortful faculty — and that science names costs the business model ignores.
Attention was a well-studied scientific object long before it became something to buy and sell. It entered modern psychology as a scientific problem in the late nineteenth century, and over the following decades experiments established that it is limited, selective, and costly to sustain. The economic framing came much later: in 1971 Herbert Simon described attention as the scarce counterpart to abundant information, and only in the decades after that did a business model form around capturing and reselling it. The older science still describes what attention is and what happens to a person when it is fragmented, and read against today’s business model, it names costs that the model leaves off its books.
William James gave the faculty its lasting description in 1890, in the attention chapter of The Principles of Psychology. “Every one knows what attention is,” he wrote, and then set out a definition that has held up ever since: attention “is the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought,” and its essence is “focalization, concentration, of consciousness.” For James, selection required withdrawal: attention “implies withdrawal from some things in order to deal effectively with others.” Selecting one object means giving up the rest. He also tied attention to what a person cares about, in a sentence that reads strangely well in an age of recommendation feeds: “My experience is what I agree to attend to.” Which objects a person attends to shapes what their world consists of.
After James, experimental psychologists set out to measure selective attention and its limits, and the major results all pointed toward strict capacity constraints. Colin Cherry, working on the practical problem of why listeners lose track of voices when several people speak at once, ran experiments in 1953 in which people heard two spoken messages at the same time. They could follow one closely by repeating it aloud, a method called shadowing, but afterward they could report almost nothing about the meaning of the other, even though they noticed crude physical facts like a change from a man’s voice to a woman’s. Donald Broadbent turned this into a model in 1958, describing the mind as a channel of limited capacity fronted by a selective filter, an early gate that lets only some of the incoming signal through for full processing and holds the rest back. Around the same time, George Miller was cataloguing narrow limits in immediate memory and judgment, giving psychology one of its most cited examples of finite processing capacity. The details of the filter were debated for decades, but the core claim survived every revision: the human system cannot fully process everything arriving at the senses, so it must choose, and the choosing is the whole point.
Two later lines of work attached numbers to the costs of dividing and stretching attention. Task-switching studies measured what it costs to move attention from one task to another: when people alternate between two simple tasks — sorting a character as a letter or a number, then as a vowel or a consonant — they respond more slowly and make more errors on the trials where they change tasks than on the trials where they repeat. Rogers and Monsell showed in 1995 that this cost survives even when the switch is entirely predictable and the person has been given time to get ready, a residual cost that preparation cannot fully erase. Vigilance studies measured endurance instead. Norman Mackworth, studying radar operators who missed real signals during long watches, built a laboratory version in 1948 and found that detection of rare events fell off substantially within roughly the first half-hour of continuous monitoring. This vigilance decrement is one of the most reliable results in the field, and later reviews confirmed that holding attention steady is hard mental work rather than a passive state. Sustained attention declines with time on task, especially in monitoring tasks with rare targets.
By the middle of the twentieth century, attention had a settled scientific profile. It was a faculty of limited capacity, it worked by selecting some inputs and holding others back, switching it between tasks carried a measurable penalty, and keeping it steady wore it down. These findings rest on repeatable laboratory results measured under controlled conditions, which turn distraction from a vague feeling into measured changes in detection, response time, and accuracy.
The concept crossed into economics through Herbert Simon, an economist and cognitive scientist who saw where an information glut would lead. In a 1971 essay on organizing institutions for a world awash in data, Simon pointed out that growth in the supply of information carries a hidden cost. “What information consumes is rather obvious: it consumes the attention of its recipients,” he wrote, so that “a wealth of information creates a poverty of attention.” Abundance on one side forces scarcity on the other. As the supply of things to read, watch, and answer grew without limit, the fixed human capacity to attend to them became the binding constraint. Simon drew a practical lesson: a system meant to help people should be designed to conserve their attention rather than add to the flood, filtering and condensing information so that less of the scarce resource is spent. His framing kept attention on the side of the person doing the attending, as a budget to be protected. The measurement tradition had already supplied the numbers behind his intuition. When Simon said attention was scarce, Broadbent’s bottleneck, Mackworth’s decrement, and the switch-cost studies had already shown in what specific ways selection, switching, and sustained monitoring fail under load.
The phrase “attention economy” arrived later, in Michael Goldhaber’s 1997 essay on the internet and in Davenport and Beck’s 2001 business book, and it carried Simon’s scarcity forward with a change of emphasis. Simon had asked how a person or institution should spend its own limited attention. The business writers asked a different question: how a firm could capture other people’s attention and convert it into revenue. In this framing a person’s attention becomes inventory that a firm holds and resells to advertisers. Platforms that sell advertising are paid in proportion to the attention they hold, so the engineering goal became maximizing time on the service and the number of sessions per day. That is a coherent commercial logic, and it now shapes some of the most used products in the world.
The two descriptions of attention diverge sharply. The science treats attention as a limited faculty belonging to a person, whose selection, switching, and endurance each carry costs the person pays. The business model treats captured attention as a resource to be maximized and resold, and its accounts record the revenue from holding attention while saying nothing about what holding it does to the holder. A minute on the platform is counted by the business as an additional unit of engagement, while for the user it is a minute drawn from a finite capacity for concentration, recovery, and self-directed thought. Because the user’s cost never appears as a price the advertiser pays, the model has no reason to count it. Economists have a name for a cost produced by a transaction but borne by someone outside it: an externality. The standard example is a factory that earns a profit while the surrounding community absorbs the health cost of its polluted air. Fragmented attention fits the pattern closely.
Three externalized costs follow directly from the science. The clearest is switching: a notification that pulls someone away from demanding work can impose a switch and recovery cost of the same broad type Rogers and Monsell studied — slower, more error-prone performance on return — and the reconfiguring is not free even when the interruption is brief. Designs that optimize for frequent returns tend to increase the number of transitions a user has to manage. A second cost is the strain of interruption. Gloria Mark and colleagues showed in 2008 that interrupted work does get finished, and often faster, but at the price of higher stress, frustration, and effort, so the speed is bought with strain the worker absorbs. Her earlier field study found that everyday knowledge work is already chopped into segments only a few minutes long, and that returning to an interrupted task took an average of about twenty-three minutes. A third cost is the crowding out of sustained thought. Difficult reading, writing, planning, and mathematical reasoning need uninterrupted time because the object of thought is larger than a single stimulus and has to be held active while its parts are examined in turn. Frequent unrelated prompts make it harder to hold the continuous train of thought — the focalization James described — that this kind of work requires. The loss is easy to underestimate because each interruption seems small, but the costs accumulate: a day of frequent switching is a day in which the deep, effortful mode of attention rarely gets assembled at all, and the work that depends on it is postponed or done worse.
The evidence supports these specific limits without requiring the stronger, unsupported claim that attention is a fuel tank that empties on a fixed schedule. The robust results are narrow and sufficient: selection has a cost, switching has a cost, and sustained attention fades. Some tools that direct attention still help, when they serve the user’s own goal, arrive at the right moment, and reduce irrelevant search. Each of these costs is something the attention economy’s core metrics are built to ignore, because each falls on the user rather than the buyer. The word “engagement” hides the distinction, because it can describe either useful involvement or successful capture. A student engaged with a lesson is doing something valuable, while in advertising-funded systems engagement often measures successful capture rather than benefit to the person captured; the metric can rise when the user is learning, and it can rise just as well when the user is anxious, provoked, or unable to disengage.
Taking the older science seriously changes what the attention economy is understood to be selling. The scarce object is the person’s capacity to select, sustain, and recover attention, and that capacity has limits that show up as slower performance, stress, and lost continuity. A business can sell captured attention without pricing any of it, while the costs still occur in the human system that supplies the commodity. The science that began with James names the missing side of the transaction: capturing attention draws down a limited faculty, and the bill is charged to the person’s account rather than the advertiser’s.