"We show that face-to-face interaction decreased by approximately 70 percent following the redesign [to open plan]."

Physical environment: the open-office evidence is bad

In 2018, Ethan Bernstein (Harvard Business School) and Stephen Turban published "The impact of the 'open' workspace on human collaboration" in Philosophical Transactions of the Royal Society B (2018, Vol 373, 20170239). The study examined two Fortune 500 headquarters before and after open-plan office conversions.

The researchers used electronic sociometric badges and digital communication logs to measure actual interaction, not self-reported interaction. The findings were unambiguous: after the conversion to open plan, face-to-face interaction time decreased by approximately 70% (one case study) and 72% (the second). Email communication increased by 22-50%. Instant-message communication increased by 50-75%.

The mechanism Bernstein and Turban proposed: in open-plan environments, workers self-regulate by avoiding eye contact, wearing headphones, and withdrawing into digital channels. The physical openness that was supposed to encourage spontaneous collaboration actually triggered the same social avoidance behaviours that were present before the conversion, but now without the private spaces that made concentration possible.

Five mechanisms by which open-plan offices degrade focus-work productivity:

• Visual exposure and social self-regulation. The knowledge that you are visible to colleagues creates a performance layer that interrupts deep cognitive work. Self-consciousness (the loss of which is condition 3 of flow) is harder to achieve when you are physically visible.
• Ambient noise and involuntary attention capture. The brain processes speech involuntarily even when not directed at you. A nearby conversation triggers partial attention capture that disrupts sustained concentration. This is not a willpower failure; it is auditory physiology.
• Temperature variance. Open-plan offices are frequently over-cooled (to compensate for the higher heat load of many people in one space) and unevenly distributed. Thermal discomfort reduces sustained cognitive performance.
• Privacy removal inhibits difficult conversations. Sensitive conversations (performance feedback, conflict resolution, personal concerns) that previously happened in private offices now happen via Slack, where they produce worse outcomes.
• The desk drive-by. Physical accessibility makes Class 4 interruptions (tap on shoulder) structurally unavoidable in a way that they are not in office or remote environments.

Noise-cancelling headphones (Sony WH-1000XM series, Bose QuietComfort series, Apple AirPods Pro) are the default cultural adaptation to open-plan offices. They address the ambient noise problem effectively for steady-state background noise; they are less effective for variable speech nearby (the brain's speech-recognition circuits still partially process identifiable human voices even through active noise cancellation).

The social signal matters as much as the acoustic function: visible headphone usage is a "do not disturb" signal that has become widely understood in technology offices. The challenge is that some managers interpret visible headphone usage as anti-social or disengaged, which creates a cultural friction that undermines the focus-protection effect. Teams that normalise headphones as a focus marker, not an avoidance signal, get the acoustic and social benefits without the management friction.

Four physical environments with evidence for deep-work support, in order of research backing:

Enclosed private offices for senior engineers and managers. The traditional model, strongly supported by DeMarco and Lister's Peopleware (1987/2013) research and consistent with the Bernstein & Turban findings. Higher real-estate cost per seat; measurably higher cognitive output for roles requiring sustained deep work.

Small team rooms of 4-6 people. The middle path: some ambient collaboration without the acoustic and social exposure of full open plan. Enough enclosure to make difficult conversations possible; enough proximity for spontaneous pair interaction.

Library-quiet floors or zones. Hard-rule quiet (no voice calls, no conversations above a whisper), self-enforced by convention. University libraries have demonstrated the viability of this model for decades. Some technology companies (including certain teams at Google and Apple) operate designated quiet floors.

Home offices (for remote work). When set up appropriately (dedicated space, door that closes, reliable internet, camera positioning that signals professional context), home offices can provide better deep-work conditions than an open-plan corporate environment. The Bernstein & Turban research predates the 2020 remote-work expansion but is consistent with the observed productivity outcomes many engineers report from home.

Open-plan conversions are frequently sold to senior leadership as either cost-saving (higher seat density) or collaboration-boosting. The Bernstein & Turban evidence directly contradicts the collaboration claim. The cost-saving claim is usually true but modest: per-seat real estate cost falls, but the productivity cost per engineer often exceeds the real estate saving.

A rough calculation: if a 500-person engineering-heavy organisation loses an average of 30 minutes per engineer per day to open-plan distraction (a conservative estimate), at a fully-loaded $120/hr, the annual productivity cost is 500 x 0.5 x $120 x 240 days = $7.2M. A move to higher-cost private or semi-private offices might add $3,000-8,000 per seat annually in real estate cost (depending on market), totalling $1.5-4M. The real estate saving from open plan does not cover the productivity loss.

Not every role benefits from enclosed space. Some collaborative roles - design critique, creative agency work, early-stage startup founding teams - genuinely benefit from ambient overhearing and spontaneous interruption. The Bernstein & Turban research applies most strongly to roles requiring sustained cognitive work on complex individual tasks: software engineers, writers, financial analysts, and researchers. For highly collaborative roles where the work is primarily interpersonal coordination, the open-plan critique is weaker.

The argument is not "open plan is always bad." It is "open plan as a default configuration for engineering-team deep work has strong evidence against it, and the evidence has been available since at least 1987."