The Poverty of Attention: Focus and Conversation in an Economy That Sells Yours

In 1971, Herbert Simon wrote the sentence that names the modern condition. In an essay titled Designing Organizations for an Information-Rich World, he observed that what information consumes "is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information sources that might consume it."

Fifty-five years later, the overabundance is no longer a forecast. Infinite feeds, messaging threads, notifications, and an emerging flood of machine-generated content compete for the same finite resource, and the systems competing for it are engineered by some of the best-funded teams in the world. The capacity they compete for has not grown. Simon's sentence stopped being a warning about the future and became a description of the present.

This article synthesizes what the measured research says about that competition: what fragmented attention actually costs at work, what divided attention does to conversation, and — importantly — where the published evidence is weaker than the popular discourse claims. The conclusion is not that attention is "under attack" by a vague enemy. It is that attention is a scarce resource with measurable costs of misallocation, that those costs concentrate in places people do not feel, and that deliberate control of attention is therefore a skill with unusually high returns.

1. The Economy Where Attention Is the Currency

Simon's argument has a precise economic form. Information is not scarce; the human attention needed to receive it is. Any market with a scarce good and an unlimited number of sellers develops competition for the buyer — and in this market, the buyer does not pay with money.

Tim Wu's The Attention Merchants (2016) documents the business model built on that asymmetry: platforms funded by advertising treat human attention as the raw material to be harvested and resold. The design goal of such systems is capture, and capture competes with everything else a person might attend to — including the person sitting across the table.

The measurable drift in attention is documented in the logging studies of Gloria Mark's group at UC Irvine. In 2004, computer users averaged roughly 2.5 minutes on a single screen before switching. By the late 2010s, the same methodology found the average at roughly 47 seconds — a figure Mark reports has been replicated by other groups within a few seconds. These are logged behavior measurements, not self-reports, collected across more than a decade of study (Mark, 2023). Whatever the exact number in any given year, the direction is not disputed: the unit of sustained attention keeps shrinking.

2. What Interruption Actually Costs

The most detailed picture of fragmented attention comes from direct observation. Mark, González and Harris (CHI 2005) shadowed 24 information workers for three full days each — over 700 hours of formal observation — and timed their activities.

The findings describe fragmentation as the normal condition of knowledge work:

• People averaged only about 11 minutes in a working sphere before switching to something else.
• 57% of working spheres were interrupted rather than completed.
• 77.2% of interrupted work was resumed on the same day — but when it was, resumption took an average of 25 minutes 26 seconds (SD 54 min 48 s).
• Before returning to the interrupted task, people worked on an average of 2.26 intervening working spheres.

The last number is the important one. The cost of an interruption is not the interruption itself. It is the chain it spawns: two or more unrelated topics attended to in between, with the physical and mental context of the original task dismantled in the process. As the authors note, a blinking cursor at the end of a sentence is a strong re-orientation cue — but after two intervening activities, with other windows opened, people often cannot even remember which document they were in.

A follow-up experiment (Mark, Gudith & Klocke, CHI 2008) found something less intuitive. People who were interrupted completed their tasks in less time, with no difference in quality. Interruption did not destroy performance — it restructured it. People compensated by working faster, and the compensation had a price: significantly higher workload, stress, frustration, time pressure, and effort, measurable after only 20 minutes of interrupted work. The cost of fragmentation is paid, but invisibly — people feel faster while the stress rises.

3. Why Task-Switching Has a Price

The cognitive mechanism behind these observations was mapped by Rubinstein, Meyer and Evans (2001). Executive control processes — the ones that shift attention between task sets — operate serially, one at a time. When two tasks compete, the control system must activate one set, suppress it, activate the other. The cost of that switching grows with the complexity of the tasks involved.

David Meyer has summarized this work with an estimate that shifting between tasks can cost "as much as 40 percent of someone's productive time" (American Psychological Association). The figure needs its denominator stated: it is an estimate about time lost to switching blocks, conditional on task complexity and switching frequency — not a measured constant, and not a property of every multitask. The defensible version of the claim is simpler: there is no parallel processing for attention-demanding tasks, only rapid switching, and switching bills the account every time.

4. Heavy Media Multitasking: The Finding That Got Smaller

The most cited study of chronically divided attention is Ophir, Nass and Wagner (PNAS, 2009). Participants classified as heavy media multitaskers were more susceptible to interference from irrelevant environmental stimuli and from irrelevant representations in memory, and — the surprising part — performed worse on a test of task-switching ability than light multitaskers, apparently because they were worse at filtering interference from the irrelevant task set. Heavy multitaskers were, in effect, more distractible than everyone else.

This result is cross-sectional: it compared existing multitaskers rather than assigning anyone to multitask. The direction of causation is unresolved — heavy multitasking may degrade attentional control, or people with more reactive, bottom-up attention may gravitate toward constant multitasking.

A decade of follow-up research allows the stronger version of the finding to be measured. Wiradhany and Nieuwenstein (2017) ran two replication studies and a meta-analysis of 39 performance-based assays across 12 studies, and found only a small association between media multitasking and distractibility (d = 0.17). Uncapher and Wagner's 2018 review of 21 reports concluded that the balance of evidence supports poorer performance among heavier multitaskers in several cognitive domains — while noting the literature is sparse and marked by both convergent and divergent findings.

The honest summary: an association between chronic media multitasking and worse attentional control is reasonably supported; its magnitude is small; the causal arrow is unproven. The 2009 finding did not survive at the strength the discourse remembers. It survived at d = 0.17.

5. The Phone in the Room: What Survived Replication

Ward, Duke, Gneezy and Bos (2017) reported one of the decade's most quoted experiments. In two studies, participants performed working-memory and fluid-intelligence tasks under three conditions: phone on the desk, phone in the pocket, phone in another room. Performance was worst when the phone sat on the desk — even though participants reported no difference in how hard they tried to stay focused. The authors named this the "brain drain" effect: the mere presence of one's own smartphone occupies limited-capacity attentional resources.

The follow-up record is more complicated, and worth stating precisely. A systematic review and six meta-analyses by Parry (2024) covering 56 effects and 7,093 participants found exactly one significant pooled effect: working memory capacity was impaired by phone presence. For the remaining cognitive functions — fluid intelligence included — the pooled effects were null, the field showed substantial methodological heterogeneity, and statistical power was generally poor. Parry's own conclusion: "limited support for a negative albeit smaller than theorized brain drain effect." A direct replication of the original experiments (Ruiz Pardo & Minda, 2022) likewise produced incongruent evidence.

The pattern across Sections 4 and 5 is consistent: the headline findings of the attention-crisis literature shrink under replication, but they do not all vanish. What survives is modest and specific — and modest, specific effects that recur across millions of attention-moments can still aggregate into something large. The mistake is believing the loud version of the claim.

6. Divided Attention in Conversation

The human-interaction thread of this literature is where the evidence is weakest — and where a careful reading matters most, because the claims are the most socially loaded.

The laboratory origin is Przybylski and Weinstein (2013). In two experiments, pairs of strangers conversed for ten minutes with either a mobile phone or a pocket notebook placed nearby, unremarked. In both experiments, the pairs exposed to the phone reported significantly lower connectedness, less trust, and less perceived empathy from their partners. The effect was stronger during a meaningful conversation than a casual one — the phone mattered most precisely when the conversation mattered most.

In 2021, Linares and Sellier ran the closest possible replication with a larger sample — 52 dyads against the original 37 — and found nothing. Relationship quality, closeness, trust, empathy: all null. The authors concluded that the effect is "at minimum harder to find than what was previously assumed," pointing out that null effects do not invalidate an effect and that power may still have been insufficient, but also raising a genuine possibility: people may have simply become habituated to phone presence since the original data were collected. The honest reading is that the mere-presence-of-a-phone effect on conversation is fragile — it was found once, in small samples, and could not be reproduced.

The observational record is more consistent. Misra, Cheng, Genevie and Yuan (2016) observed 100 dyads in Washington-area coffee shops during ten-minute conversations. Device presence was not assigned — it was simply observed, so this is an association with covariate controls, not a causal estimate. Conversations in which neither participant placed a device on the table or held one were rated as significantly superior in connectedness (b = 0.85, p < .05), and participants in device-free conversations reported higher empathic concern (5.85 vs 5.51 on a 7-point scale). Notably, the empathy gap was largest for pairs who reported a close relationship — the people who most expect full attention are the ones most bothered by its absence.

The strongest controlled evidence in this thread is from the screen-removal side. Uhls and colleagues (2014) studied 105 sixth-graders: 51 attended a five-day outdoor education camp with no screens, 54 continued their normal school routines. The camp group improved significantly more on recognizing nonverbal emotion cues (DANVA2 and CASP measures). This is quasi-experimental — intact groups from the same school, not random assignment — and the camp bundled three changes at once: no screens, immersion in nature, and days of face-to-face practice. The improvement cannot be attributed to screen removal alone, but the package demonstrates that short-term changes in how attention is spent move a social skill that was assumed to be fixed.

One further finding ties the thread to the individual rather than the dyad. Killingsworth and Gilbert (2010), using experience sampling — real-time reports collected as people went about their days — found that minds wandered in 46.9% of samples, in at least 30% of samples during every activity except one, and that people reported being less happy while mind-wandering regardless of what they were doing. Attention that is not on the present activity, including on the present person, carries a measurable affective cost.

The synthesis of this section: the robust signal is not that a visible phone reliably poisons a conversation — the replication record says it does not — but that where attention is allocated during interaction predicts how the interaction goes. Divided attention during conversation shows up as lower connectedness and less empathic concern across observational and quasi-experimental designs, and the mechanism is ordinary: a mind that is 47 seconds from its next switch is a mind that misses the facial cue, the tone shift, the hesitation.

7. Why Control of Attention Is Worth Practicing

The sections above share a structure worth noticing. In every case, the cost of misallocated attention is real but unfelt: interrupted workers feel faster while stress rises; phone-present participants feel they are trying just as hard; the intervening working spheres pile up invisibly. Because the costs do not announce themselves, they accumulate without triggering correction. Control of attention is not a virtue to be admired — it is the compensating mechanism for a blind spot.

The research supports a few specific practices:

1. Treat attention as an allocated budget, not a default state. Simon's formulation makes the error mode clear: not spending attention, but failing to decide where it goes. Allocation decisions made in advance — what the next block of time is for — are the only decisions made before the market starts bidding.
2. Design the environment rather than relying on in-the-moment willpower. The 2005 data show recovery, not the interruption, is where the cost concentrates; the 2008 data show people cannot feel that cost while paying it. Notification batching, away states, and single-window work remove the triggers that re-entry costs depend on. Willpower fights the moment; architecture removes the moment.
3. Protect re-entry. Since resumption averages 25 minutes and 2.26 intervening spheres, interruptions should land at task boundaries — not mid-task. When an interruption is unavoidable, writing a single visible re-orientation line before leaving the task restores exactly the cue that intervening work destroys.
4. Match conversation to full attention. The conversation literature is weakest on device mere-presence and strongest on this: attention visibly divided during interaction correlates with lower connectedness and less empathic concern, especially between people who are close. Device-free conversation is the intervention best aligned with the evidence — and it costs nothing.
5. Demand effect sizes, not vibes. The replication record in Sections 4–6 is itself a skill: several famous findings shrank or vanished when tested properly, while quieter, measured effects (re-entry cost, stress under interruption, mind-wandering) held. A person who shares the newest alarming study before its replication exists is practicing the same attention-credulity the feed rewards.

Synthesis

The framing inverts the common one. The problem is not that "the attention economy" destroys focus by some ambient force. It is that attention is a scarce, serial resource; that a large industry competes for it without any mechanism to bill the buyer; and that the costs of misallocation are systematically unfelt — paid in stress and re-entry time rather than in anything a person notices in the moment.

Three moves follow from the evidence:

1. Trust the measured costs, discount the loud ones. The re-entry cost (25 minutes, 2.26 intervening spheres), the affective cost of interrupted work, and the association between divided attention and degraded conversation are measured and repeatable. The brain-drain catastrophe and the conversation-ruining phone are not, at the strengths popularized.
2. Spend architecture before willpower. Every robust effect in this literature operates through re-orientation and salience — mechanisms that environment design attacks at the root and that in-the-moment resistance cannot even perceive.
3. Treat presence as an allocation, not a feeling. During conversation, the observable variable is where attention is. The evidence says the person across the table can tell the difference — and that the divider usually cannot.

Attention is the one input every activity consumes. In a market built to capture it, the scarce skill is not having more of it — it is deciding, before the bidding starts, where it goes.

Sources and References

• Simon, H. A. (1971). "Designing Organizations for an Information-Rich World." In M. Greenberger (Ed.), Computers, Communications and the Public Interest. Johns Hopkins Press. Source of the "poverty of attention" formulation.
• Wu, T. (2016). The Attention Merchants: The Epic Scramble to Get Inside Our Heads. Knopf. Commentary on the ad-funded attention business model; historical argument, not a measurement source.
• Mark, G., González, V. M., & Harris, J. (2005). "No Task Left Behind? Examining the Nature of Fragmented Work." CHI '05, 113–120. 24 observed information workers; 57% interruption rate; 77.2% same-day resumption; 25 min 26 s average resumption; 2.26 intervening working spheres. Observational field study.
• Mark, G., Gudith, D., & Klocke, U. (2008). "The Cost of Interrupted Work: More Speed and Stress." CHI '08. Laboratory experiment; interrupted tasks completed faster with no quality difference, at higher workload, stress, frustration, time pressure, and effort.
• Mark, G. (2023). Attention Span, Hanover Square Press; and "Speaking of Psychology" interview (American Psychological Association, 2023). The 2.5-minute (2004) to 47-second (late 2010s) screen-attention trend, from behavioral logging studies as reported by Mark and colleagues.
• Rubinstein, J. S., Meyer, D. E., & Evans, J. E. (2001). "Executive Control of Cognitive Processes in Task Switching." Journal of Experimental Psychology: Human Perception and Performance, 27(4), 763–797. Switching costs grow with task complexity. The "as much as 40% of productive time" figure is Meyer's expert estimate reported by the APA, not a measured constant.
• Ophir, E., Nass, C., & Wagner, A. D. (2009). "Cognitive control in media multitaskers." PNAS, 106(41), 15583–15587. Cross-sectional comparison; heavy multitaskers worse at filtering interference and at task-switching.
• Wiradhany, W., & Nieuwenstein, M. R. (2017). "Cognitive control in media multitaskers: Two replication studies and a meta-analysis." Attention, Perception, & Psychophysics, 79(8), 2620–2641. 39 performance assays across 12 studies; d = 0.17 for the multitasking–distractibility association.
• Uncapher, M. R., & Wagner, A. D. (2018). "Minds and brains of media multitaskers: Current trends and future directions." PNAS, 115(40), 9889–9896. Review of 21 reports; balance of evidence supports modest performance differences, with heterogeneity.
• Ward, A. F., Duke, K., Gneezy, A., & Bos, M. W. (2017). "Brain Drain: The Mere Presence of One's Own Smartphone Reduces Available Cognitive Capacity." Journal of the Association for Consumer Research, 2(2), 140–154. Original two-experiment report of the desk-phone condition effect.
• Parry, D. A. (2024). "Does the Mere Presence of a Smartphone Impact Cognitive Performance? A Meta-Analysis of the 'Brain Drain Effect'." Media Psychology, 27(5), 737–762. Six meta-analyses, k = 56, n = 7,093; only working memory capacity shows a significant pooled effect; field marked by heterogeneity and poor power.
• Ruiz Pardo, A. C., & Minda, J. P. (2022). "Reexamining the 'brain drain' effect: A replication of Ward et al. (2017)." Acta Psychologica, 230, 103717. Direct replication; incongruent evidence.
• Przybylski, A. K., & Weinstein, N. (2013). "Can you connect with me now? How the presence of mobile communication technology influences face-to-face conversation quality." Journal of Social and Personal Relationships, 30(6), 747–756. Two experiments, 37 and 34 dyads; lower connectedness, trust, and perceived empathy with a phone present.
• Linares, C., & Sellier, A.-L. (2021). "How bad is the mere presence of a phone? A replication of Przybylski and Weinstein (2013) and an extension to creativity." PLOS ONE, 16(6), e0251451. Direct replication with 52 dyads plus extensions; null results; authors label the effect fragile and note power limits.
• Misra, S., Cheng, L., Genevie, J., & Yuan, M. (2016). "The iPhone Effect: The Quality of In-Person Social Interactions in the Presence of Mobile Devices." Environment and Behavior, 48(2), 275–298. Naturalistic observation of 100 dyads; device presence observed, not assigned — an association with covariate controls, not a causal estimate.
• Uhls, Y. T., Michikyan, M., Morris, J., Garcia, D., Small, G. W., Zgourou, E., & Greenfield, P. M. (2014). "Five days at outdoor education camp without screens improves preteen skills with nonverbal emotion cues." Computers in Human Behavior, 39. 105 sixth-graders; camp (n = 51) vs. school control (n = 54); quasi-experimental pre/post design with confounded camp conditions.
• Killingsworth, M. A., & Gilbert, D. T. (2010). "A Wandering Mind Is an Unhappy Mind." Science, 330(6006), 932. Experience-sampling study; mind-wandering in 46.9% of samples, associated with lower reported happiness.

Evidence scope: this article synthesizes the works listed above, read this session from the published papers, their full texts, or author-reported summaries. Mark's 2005 and 2008 figures were verified against the papers' own text. The "47 seconds" trend is reported from Mark's logging studies as she summarizes them in the APA interview; it was not re-verified against the raw logs. The Meyer "40%" figure is an expert estimate, not a measurement. The Ophir, Ward, and phone-presence findings are cross-sectional or quasi-experimental and are reported as associations; the Linares & Sellier null is reported as a null, not as evidence of harm. No vendor or marketing statistics are cited.

Written with Nyeker — AI assistant, for Faqih (dibotak). Bot disclaimer: this article was written with AI assistance.