Safety Culture & Behavior-Based Safety

Every site claims a strong safety culture, and almost none can say what would prove it. The word gets used to describe an attitude, which makes it unfalsifiable and therefore useless. What follows treats it as something with observable properties — and takes on the argument about behaviour-based safety directly rather than around it. This Safety Culture & Behavior-Based Safety Toolbox Talk (Safety Talk / Tailgate Talk) is about what actually distinguishes a real culture from a claimed one.

Here is the distinction that carries this whole talk, and OSHA has already drawn it. There are two kinds of safety incentive programme. One rewards people for reporting near misses and hazards. The other rewards people for low injury rates. OSHA's position is that the first type is always permissible, while rate-based programmes are permissible only as long as they are not implemented in a manner that discourages reporting. That difference is the entire safety culture question in miniature: one pays people to tell you things, and the other can end up paying them not to.

The regulatory line, precisely#

1904.35(b)(1)(iv) prohibits an employer from discharging or in any manner discriminating against an employee for reporting a work-related injury or illness.

OSHA's 2018 clarification is worth stating accurately, because it is often misreported in both directions:

Incentive programmes are not prohibited. Programmes rewarding workers for reporting near misses or hazards, and encouraging involvement in a safety and health management system, are always permissible.

Rate-based programmes are also permissible — those rewarding an injury-free month, or evaluating managers on their unit's lack of injuries — as long as they are not implemented in a manner that discourages reporting.

Where it goes wrong is a negative action taken because someone reported: withholding a prize or bonus because of a reported injury.

And the test that matters most on site comes from OSHA's earlier interpretation: discipline is considered an adverse action if it would discourage a reasonable employee from reporting a work-related injury or illness. Not whether it was intended to. Whether it would.

OSHA also concluded in the rulemaking record that blanket post-injury drug testing policies deter proper reporting — which is why post-incident testing should be aimed at the incident rather than at the person who reported it, testing everyone whose conduct could have contributed rather than only those who got hurt.

The honest case for behaviour-based safety#

This deserves a fair hearing, because the criticism of it has become fashionable and is often overstated.

It reaches work as it is actually performed. Inspecting conditions finds conditions. Watching work being done, and talking about it, reaches the method people are really using — which nothing else on a site does.

It produces leading information. Injury rates tell you about the past and about luck. Observations tell you about now.

It involves the crew rather than treating safety as something done to them by a department.

And done properly, it surfaces conditions. A good observation conversation ends with a broken tool replaced or an anchor installed, not with a worker corrected.

The honest case against#

It locates the cause in the worker. The name puts behaviour at the centre, and once behaviour is the finding, the condition that produced it survives untouched. A behaviour repeated by most of a crew is a design signal, not a discipline problem.

It can suppress reporting. Where the programme is tied to rates or to incentives, the cheapest way to improve the number is to report less — which is exactly the harm 1904.35(b)(1)(iv) exists to prevent.

It counts easily and measures badly. Cards are countable, so cards get counted, and the number stays healthy while the information goes to zero.

And it can become surveillance if observations carry names and feed discipline.

Reconciling the two#

The position worth holding, and it is consistent with everything else in this library: behaviour observation is a good instrument pointed at the wrong target.

The observing is genuinely valuable — it reaches work as performed, which almost nothing else does. The error is the inference: concluding that the behaviour is the cause, rather than treating it as the visible end of a chain that includes the tools, the access, the method, the time allowed and the state of the person.

So keep the instrument and change the conclusion. Watch the work, have the conversation, ask why the method made sense — and then act on the condition rather than on the person. A programme that does that is not really behaviour-based at all, whatever it is called, and it is the version that works.

What actually indicates culture#

Not the posters, and not the injury rate. Rates are lagging, heavily influenced by luck and exposure, and — importantly — a rate can fall because reporting fell. A site with an improving TRIR and a collapsing near-miss count is usually getting worse, not better.

Things that indicate something real:

Bad news travels upward fast. The single best indicator there is. In a weak culture, problems move sideways and disappear.

Near-miss reports are increasing. Counter-intuitive and correct — rising reports usually mean rising trust.

People can name a time work was stopped, and what happened to the person who stopped it.

Reported conditions get closed, visibly and reasonably quickly.

Supervisors raise problems about their own areas rather than defending them.

New starters are corrected kindly and consistently by peers, not only by the safety adviser.

Nobody is surprised by an inspection, because nothing is being hidden between them.

And the crew's account of how work is done matches the written method — or the difference is known and being addressed.

What can go wrong?#

Rate-based incentives that make an injury expensive to report.

A prize or bonus withheld because of a reported injury — the specific act OSHA identifies.

Blanket post-injury drug testing, which OSHA concluded deters proper reporting.

Discipline that a reasonable employee would find discouraging, whatever the intent behind it.

Observation cards with names on them feeding a disciplinary process.

Counting observations rather than counting what changed.

TRIR used as the culture measure, so improving the number becomes the objective.

Culture treated as an attitude to be exhorted rather than a set of conditions to be built.

How do we manage this properly?#

Reward reporting, not the absence of reports. OSHA says the first is always permissible; the second only works if it never discourages anyone.

Apply OSHA's own test to any policy: would this discourage a reasonable employee from reporting? If yes, change it, regardless of what it was meant to do.

Aim post-incident testing at the incident, testing everyone whose conduct could have contributed, not only the person who was hurt.

Measure changes, not counts. Tools replaced, anchors installed, methods rewritten, hazards closed.

Watch near-miss reporting as a health indicator, and treat a falling count as a warning rather than an achievement.

Keep observation nameless and non-disciplinary.

Act on conditions, not on people, and say so out loud so the crew can see the difference.

Have leaders raise their own problems first. Culture is set by what senior people do about bad news, not by what they say about safety.

Before you start#

• Confirm whether any incentive here rewards low injury rates rather than reporting.
• Confirm nothing would cost you if you reported an injury today.
• Confirm you know what happened to the last person who stopped work.
• Confirm near-miss reports here are going up rather than down.
• Confirm observations carry no names and feed no disciplinary process.
• Confirm reported conditions actually get closed, and how quickly.
• Confirm the written method matches how the job is really done.
• Confirm you could tell a supervisor bad news about their own area.

Talk it over#

• What happens here if you report an injury?
• Is our near-miss count going up or down, and what do we take that to mean?
• Would you tell the site manager something they did not want to hear?
• What has actually changed on this job because someone raised it?

The bottom line#

OSHA has already drawn the line that defines safety culture. Under 1904.35(b)(1)(iv) an employer may not discriminate against an employee for reporting a work-related injury, and OSHA's clarification distinguishes two kinds of incentive programme: those rewarding the reporting of near misses and hazards, which are always permissible, and rate-based programmes rewarding injury-free months, which are permissible only as long as they are not implemented in a manner that discourages reporting. One pays people to tell you things; the other can end up paying them not to. The operative test: discipline is an adverse action if it would discourage a reasonable employee from reporting — not whether it was intended to. OSHA also found that blanket post-injury drug testing policies deter proper reporting, so testing should target the incident and everyone whose conduct could have contributed. On behaviour-based safety, the honest position is that it reaches work as performed and produces leading information, but locates cause in the worker, can suppress reporting and counts easily while measuring badly — so: it is a good instrument pointed at the wrong target. Keep the observing; change the inference; act on the condition. And judge culture by whether bad news travels upward fast, whether near-miss reports are rising, and whether reported conditions get closed — never by a rate, because a rate can fall because reporting fell.

Frequently asked questions about safety culture and behaviour-based safety#

Does OSHA ban safety incentive programmes?

No. OSHA's clarification states that incentive programmes are not prohibited. Programmes rewarding workers for reporting near misses or hazards and encouraging involvement in a safety and health management system are always permissible. Rate-based programmes rewarding an injury-free month are also permissible as long as they are not implemented in a manner that discourages reporting.

When does an incentive programme become a violation?

When a negative action is taken because someone reported — for example, withholding a prize or bonus because of a reported injury. That is the specific circumstance OSHA identifies under 1904.35(b)(1)(iv), which prohibits discharging or in any manner discriminating against an employee for reporting a work-related injury or illness.

What is the test for whether a policy discourages reporting?

OSHA's own wording: discipline is considered an adverse action if it would discourage a reasonable employee from reporting a work-related injury or illness. The test is the likely effect on a reasonable employee, not the employer's intention — which means a well-meant policy can still fail it.

Is post-incident drug testing allowed?

Yes, in appropriate circumstances — but OSHA concluded that blanket post-injury drug testing policies deter proper reporting. Where testing is used to evaluate the root cause of an incident, it should be aimed at the incident: testing everyone whose conduct could have contributed, rather than only the employee who reported an injury.

Is behaviour-based safety a good idea or a bad one?

Both halves are real. For: it reaches work as it is actually performed, produces leading rather than lagging information, and involves the crew. Against: it locates cause in the worker, can suppress reporting where it is tied to rates, counts easily while measuring badly, and becomes surveillance if observations carry names. The resolution is that it is a good instrument pointed at the wrong target — keep the observing, change what you conclude from it, and act on the condition.

Why is the injury rate a poor measure of culture?

Because it is lagging, heavily influenced by luck and exposure, and — critically — a rate can fall because reporting fell. A site with an improving injury rate and a collapsing near-miss count is usually getting worse. Rising near-miss reports normally indicate rising trust, which is why they should be read as a health indicator rather than as bad news.

What is the single best indicator of a real safety culture?

Whether bad news travels upward fast. In a weak culture, problems move sideways and disappear. Close behind: whether people can name a time work was stopped and what happened to the person who stopped it, whether reported conditions get closed visibly, and whether the crew's account of how work is done matches the written method.

Download the safety culture toolbox talk PDF#

Get this safety culture and behaviour-based safety toolbox talk as a print-ready PDF — available in English, Spanish, Portuguese, and Turkish. Print it, hand it to the crew, and collect signatures on the included attendance sheet.

Download the PDF — free account required. New members get 5 free downloads.

Related toolbox talks#

• Near-Miss Reporting
• Peer Observation
• Common Unsafe Behaviors
• Stop Work Authority

Sources#

• OSHA, Clarification of OSHA's Position on Workplace Safety Incentive Programs and Post-Incident Drug Testing Under 29 C.F.R. §1904.35(b)(1)(iv), 11 October 2018: https://www.osha.gov/laws-regs/standardinterpretations/2018-10-11
• OSHA, Interpretation of 1904.35(b)(1)(i) and (iv), 19 October 2016: https://www.osha.gov/memos/2016-10-19/interpretation-190435b1i-and-iv
• OSHA, 29 CFR 1904.35 — Employee involvement: https://www.osha.gov/laws-regs/regulations/standardnumber/1904/1904.35

This talk summarises published regulation and interpretation for training purposes and is not legal advice. Whether a particular policy or programme complies depends on the facts and on the law in your jurisdiction.

Written by FieldSafetyTalk's safety professional — a CSP, ASP, CHST and OSHA Authorized Outreach Trainer with 14+ years of international construction safety experience across federal, heavy civil, and industrial projects.

Hazards covered