What most safety teams call “human performance” today is a subset of a much larger field. Human Performance Improvement (HPI) began in 1962 as a science for studying workplace results, 17 years before it entered safety through the nuclear sector.
Along the way, the field narrowed. The nuclear sector took the error tools, companies built review programs around them, and Todd Conklin renamed the approach as Human and Organizational Performance (HOP).
Learning from Normal Work builds on this full line. It adds applied psychology and human factors into a system for proactive work at scale.
The path from ISPI to Learning from Normal Work runs through the nuclear sector and the DOE handbook before reaching HOP.
Key findings
- Human Performance Improvement began in 1962 when a group of experts formed what became the International Society for Performance Improvement (ISPI), focused on workplace results outside the safety field.
- The nuclear sector combined HPI with James Reason’s error science after Three Mile Island, leading to the DOE handbook (DOE-HDBK-1028-2009) and its core ideas.
- Companies adopted the DOE guide mainly for error tools and new ways of running reviews, as Domtar’s path in pulp and paper shows.
- Todd Conklin spread five key ideas and the HPI approach was widely renamed as HOP, with regional splits in which label leads.
- Learning from Normal Work expands HPI by adding Safety-II and human factors into a system designed for proactive use at scale.
Where did human performance improvement start?
HPI started outside the safety field. In 1962, a group of experts founded the National Society for Programmed Instruction (NSPI), which later became the International Society for Performance Improvement (ISPI). Their question was simple: why does training alone fail to change how people work?
Thomas Gilbert, who had spent a year with B.F. Skinner at Harvard, gave the field its key insight. He published Human Competence in 1978 and set out the Behavior Engineering Model (BEM). Its core rule is that workplace factors should be checked before looking at the person (Gilbert, 1978).
Gilbert found that for every dollar gained by changing people, about ten could be gained by changing the workplace. Geary Rummler, Joe Harless, and Robert Mager built ways to find the root of poor results at the task, process, and firm levels (Rummler and Brache, 1990).
ISPI gave the field a home, a credential (the Certified Performance Technologist), and two journals. Field experts drove the science with tracked results across sectors from healthcare to defense. Learning from Normal Work draws on this same body of work rather than starting from ideas alone.
HPI began as a workplace science in 1962, decades before it entered safety.
How did HPI enter the safety field?
Three Mile Island in 1979 was the turning point. After the partial meltdown, the nuclear sector needed a way to understand how human error leads to system failure. INPO (the Institute of Nuclear Power Operations) formed in response and took a simplified version of HPI into nuclear work.
The institute mixed parts of HPI’s workplace study with James Reason’s error science. Reason showed how choices made by leaders years earlier create the setting for error at the front line (Reason, 1990). Human factors and systems science shaped the nuclear version further.
In 2009, the US Department of Energy codified this into DOE-HDBK-1028-2009, a two-volume handbook. It set out core ideas about human performance: people make mistakes, error-prone settings are discoverable, and how a company runs shapes how its people act. DOE-HDBK-1028-2009 gave the nuclear and energy sectors a clear way to think about human error (DOE, 2009).
Volume 2 gave hands-on tools for cutting error at three levels: STAR self-checking for workers, pre-job briefs for teams, and watching methods for leaders. For a full look at what the DOE handbook actually contains, the two volumes cover distinct ground.
The handbook kept the premise that roughly 80% of events trace to how a company is run (DOE, 2009), but it left behind the full review system that Gilbert and Rummler had built. Learning from Normal Work builds on both the nuclear tools and the broader HPI system the nuclear version dropped.
Yorio and Wachter (2014) later confirmed the premise across sectors, finding that human-performance-focused safety practices significantly predicted injury and illness rates. The evidence backed what the DOE had codified.
What happened when firms adopted the DOE guide?
Beyond nuclear energy, the DOE handbook spread quickly. Companies in oil and gas, chemicals, mining, and pulp and paper adopted its ideas and tools. HPI training programs focused on error tools and how to reduce mistakes one by one.
Domtar, a North American pulp and paper company, followed this path. The company built its human performance program around the DOE model, with training centered on error tools and a new way of running reviews after events. Leaders learned to ask what made the error likely rather than who made it. Supervisors used pre-job briefs and STAR self-checking. Review teams began looking at workplace factors alongside individual actions.
The DOE guide’s biggest effect was changing how companies looked at events. Instead of stopping at “human error,” review teams asked what factors made the error likely. The tools gave form to that question, and the shift spread across the company’s safety culture.
Companies adopted the error tools and the review reframing, and those changes produced real gains in how events were understood. What most programs did not adopt was the front-end analysis, cost-value review, or process-level methods from the broader HPI discipline. Safety teams got a toolkit for cutting error but not a method for studying how work is planned and run before events happen.
| What firms adopted | What stayed behind |
|---|---|
| Error tools (STAR, pre-job briefs) | Gilbert’s BEM analysis |
| Review reframing (beyond “human error”) | Cost-value review |
| Core DOE principles | Process-level redesign (Rummler-Brache) |
| Error precursors and error-likely settings | Front-end analysis (Harless) |
The tools helped with reviews, and reviews became the main use case. Learning from Normal Work fills this gap by taking the HPI science further and putting it to work before something goes wrong.
Most firms adopted the error tools and review reframing but left behind the full study method.
How does HPI overlaps with HOP?
Todd Conklin was at Los Alamos National Laboratory for 25 years. He took the DOE’s human performance principles and simplified them to: error is normal, blame fixes nothing, context drives how people act, learning is vital, and how leaders respond to failure matters. Conklin called them “a modern update of the building blocks of Human Performance” (Conklin, 2019).
Conklin added thinking from outside the nuclear line. Erik Hollnagel’s Safety-II shifted the focus from failure to what goes right. Sidney Dekker’s just culture work changed how companies think about accountability. The split between work-as-planned and work-as-done gave safety teams a new lens, though Gilbert’s experts had mapped the same gap for decades.
In many companies, the HPI label was swapped for HOP without anyone noting the switch. The team had always called it “human performance,” and the content changed while the name stayed.
In some parts of the world, Human and Organizational Performance (HOP) is the main label, while in others HPI is used. The split often traces to which expert or training firm led the rollout, rather than to any real gap in the core ideas. For a closer look at where HPI and HOP agree and diverge, the line matters because the label shapes what people believe they have.
Two paths emerged from the DOE base: HOP took the principles, while Learning from Normal Work kept the full HPI science.
Learning from Normal Work treats this split as a question of depth rather than brand. Both HPI and HOP carry the premise that how people act is shaped by the systems around them. What matters is whether the program behind the label holds the study tools and tracked results the full science asks for.
What does Learning from Normal Work add to HPI?
Learning from Normal Work takes the HPI base and expands it where the original field stopped.
The scope is wider. HPI drew mainly from the study of conduct and workplace results. The approach adds Safety-II (studying what goes right) along with applied psychology and human factors. The result is a broader view of what shapes how people perform than any single school gives.
The focus shifts from reactive to proactive. In companies where Psychology Applied has assessed HPI programs, the primary use case remains review after an event. Learning from Normal Work studies work that is going well and brings out the factors that shape normal work before they lead to failure.
A deployable system ties it together. The system covers rollout, results tracking, leader skill-building, and long-term support. This is what makes it usable at a large scale across a whole company.
At TE Connectivity, leaders trained through the Learning from Normal Work program showed clear improvements. Learning-focused inquiry went from 58% to 91% (p < .001), and the method found 96% of error traps compared with 37% from a standard Gemba walk (Psychology Applied, 2026).
The gap came from the method rather than the people. ISPI, the body that has governed what human performance improvement actually is since 1962, backed the Learning from Normal Work Competency Pathway. The body that sets the bar for sound performance work confirmed that the approach yields tracked gains.
Learning from Normal Work builds on HPI by adding Safety-II and the science of human conduct into a system for proactive use at scale.
References
Conklin, T. (2019). The 5 Principles of Human Performance: A Contemporary Update of the Building Blocks of Human Performance for the New View of Safety. Pre-Accident Investigation Media.
Gilbert, T. F. (1978). Human competence: Engineering worthy performance. McGraw-Hill. (Tribute edition 2007, Pfeiffer.)
Hollnagel, E. (2014). Safety-I and Safety-II: The past and future of safety management. Ashgate.
Psychology Applied. (2026). TE Connectivity Competency Pathway data.
Reason, J. (1990). Human error. Cambridge University Press.
Rummler, G. A., & Brache, A. P. (1990). Improving performance: How to manage the white space on the organization chart. Jossey-Bass.
US Department of Energy. (2009). Human performance improvement handbook (DOE-HDBK-1028-2009), Volumes 1 and 2.
Yorio, P. L., & Wachter, J. K. (2014). The impact of human performance focused safety and health management practices on injury and illness rates. Safety Science, 62, 157-167.
