PsychologyCounterintuitive resultExplained★ Famous
Murphy's Law
Also known as Sod's Law, Finagle's Law
If anything can go wrong, will it? The saying is a joke — but the engineering principle behind it is dead serious.
Counterintuitive result. Nothing is wrong with the reasoning. The result is simply correct and human intuition is simply unreliable here. Not a contradiction at all.
Discover
Edwards Air Force Base, California, 1949. Project MX981 is trying to find out how many Gs the human body can survive. A rocket sled is being used to slam a volunteer — Air Force doctor John Paul Stapp — into deceleration forces that make a car crash look gentle.
The sensors are critical. They measure exactly how hard Stapp's body is being hit so that the researchers can understand what a human being can actually endure. Sixteen sensors are attached to the harness. Sixteen sensors means sixteen connectors. Each connector can be installed in exactly two ways: right and wrong.
A technician wires all sixteen in the wrong configuration.
Captain Edward A. Murphy Jr., the engineer responsible for the sensors, stares at the result and makes an observation about the technician. Depending on which account you read, it is either colourful or professional, but the meaning is the same: if there are two ways to do something, and one of them will cause disaster, someone will do it that way.
A few months later, at a press conference celebrating the fact that Stapp had survived 46.2 Gs — at the time, the highest deceleration ever survived by a human — the project director credited Murphy's observation as part of why the tests had been safe. They had designed for the wrong installation, so it could not happen. The press wrote it down, and it spread from there.
Understand
The version that became famous is weaker than the original. Anything that can go wrong, will go wrong is a joke about bad luck. Murphy's actual observation was an engineering principle: if a failure mode exists, a failure will eventually occur through it. The difference matters.
The joke version claims the universe is malicious. The engineering version claims something more modest and more useful: systems will be used in every way they can be used, including the ways you did not intend. If a connector fits in two ways, a user will eventually try the wrong one — not because they are careless, but because they are human, the label is small, the light is bad, and they are in a hurry.
The practical conclusion is poka-yoke — the Japanese engineering term for mistake-proofing. Design the connector so it only fits one way. If you cannot do that, make the wrong way obviously, irreversibly wrong before it causes damage. If you cannot do that either, assume the mistake will happen and design for recovery.
Why does it feel true as a joke?
The pessimistic version — of course the toast lands butter-side down — persists because of a real cognitive mechanism. Bad events are more salient than good ones. We remember the traffic jam we hit after taking the slow road, not the hundred times we took the fast road and it was fine. We remember the USB cable that took three attempts, not the thousands of cables we plugged in first try. The joke is a story about how memory distorts the record of randomness.
There is also genuine mathematics here. In a complex system with many components, the probability that at least one will fail is higher than any single component's failure rate, sometimes much higher. A car with two thousand parts, each individually very reliable, will still spend time in the shop.
What Murphy's Law does not say
It does not say failure is inevitable in any particular case. It says failure modes should be closed, not left open on the grounds that a reasonable person would never trigger them. Reasonable people, pressed for time, in poor conditions, do trigger them.
Examples
| System | Failure mode left open | What happened |
|---|---|---|
| MX981 sensor harness | Connector fit in two orientations | Technician wired all sixteen wrong |
| USB-A plug (pre-2014) | Required specific orientation, not marked | Three-attempt insertion became cultural shorthand for frustration |
| Therac-25 radiation machine | Software race condition reachable in use | Six patients received massive overdoses, three died |
| Boeing 737 MAX MCAS | Single sensor failure could activate automated pitch-down | Two crashes, 346 deaths |
| Hospital drug dosing | Syringe port accepted wrong connection | Epidural drugs injected intravenously |
The last three are not jokes. They are what happens when Murphy's original engineering principle is treated as folklore rather than design constraint.
Explore
Finagle's Law is the darker cousin: anything that can go wrong, will — at the worst possible moment. It adds a claim about timing…
Finagle's Law is the darker cousin: anything that can go wrong, will — at the worst possible moment. It adds a claim about timing that Murphy's formulation does not make. The timing version is probably selection bias: the moments when failures are memorable are precisely the moments when failure was costly, so the failures we talk about are disproportionately the badly-timed ones.
Sod's Law is the British variant, with a slightly different emphasis: not that failure will happen, but that it will happen to you specifically, as if the universe were watching. Same cognitive mechanism.
The anthropic angle. A more serious observation: we only get to study systems that failed in ways we could study. Failures that killed everyone left no record. This is a form of survivorship bias that haunts accident investigation generally.
In software. Murphy's Law is the informal name for a set of practices that formal engineering calls fault tolerance, defensive coding, and chaos engineering. Netflix famously runs Chaos Monkey, software that randomly kills production servers, on the grounds that if a server can die, it will die, and better to find out while you are watching than at 3am when you are not.
Historical context
1949. The incident at Edwards Air Force Base. The exact wording Murphy used is disputed — several people were present and none of…
1949. The incident at Edwards Air Force Base. The exact wording Murphy used is disputed — several people were present and none of them agree. What is agreed is that Stapp used a form of the observation in public, crediting Murphy, and that journalists wrote it down.
1952. Anne Roe, a psychologist interviewing scientists, records a version of the saying — evidence it was already in technical culture within three years of the incident.
1955. Scientific American prints a version, cementing it in print.
1977. The Macquarie Dictionary (Australian) includes Murphy's Law — one of the first dictionary inclusions.
1980s onwards. The phrase enters popular culture globally, losing most of its engineering content along the way. T-shirts, calendars, and coffee mugs present the pessimistic folk version. The engineering community, quietly, keeps using the original.
Why it matters
Murphy's Law as engineering principle has probably saved more lives than any number of motivational posters. Design for misuse.…
Murphy's Law as engineering principle has probably saved more lives than any number of motivational posters.
Design for misuse. Every interface that accepts only the correct input, every physical connector that fits in only one orientation, every drug dose that cannot be wrong by a factor of ten — these are Murphy's Law applied. The question is not will anyone do this wrong? but what happens when they do?
Failure mode analysis. Modern safety engineering requires identifying every possible failure mode before a system is deployed. This is Murphy's original insight formalised: list the ways things can go wrong, then close them, before anything goes wrong.
The cost of optimism. The alternative to Murphy's Law is the belief that careful users and clear instructions are sufficient. Careful users are not always available. Instructions are not always read. Systems are used under conditions their designers did not anticipate. The history of engineering disasters is substantially a history of designers who believed their users would behave as intended.
Modern relevance
Software reliability. Every production system of any scale runs with the assumption that hardware will fail, network packets will…
Software reliability. Every production system of any scale runs with the assumption that hardware will fail, network packets will be lost, and code will encounter inputs it was not designed for. The discipline of building systems that continue to function under these conditions is Murphy's Law institutionalised.
Security. Security engineers assume that any attack surface that exists will eventually be exploited — not by theoretical adversaries, but by real ones who have time and motivation. This is exactly Murphy's original framing applied to an adversarial context.
Medical device design. The FDA requires, for any medical device, analysis of failure modes and their consequences. Connectors that could be confused are redesigned. Labels that could be misread are rewritten. Doses that could be entered incorrectly are constrained. The regulatory framework is Murphy's Law made mandatory.
AI systems. Language models and reinforcement learning agents are reliably used in ways their designers did not anticipate. They are given tasks outside their training distribution, jailbroken, used as components of larger systems that interact in unexpected ways. The field of AI safety is partly an attempt to systematically enumerate the ways things can go wrong before deploying systems that cannot easily be recalled.
Sources & further reading
Paper
(2003)Murphy's LawAnnals of Improbable Research, 9(5).
The most thorough historical account of the law's actual origin and the gap between the engineering principle and the popular version.
How this connects
- Related to
- Goodhart's LawBoth are named engineering or institutional observations that became cultural shorthand. Murphy's Law warns that systems will be used wrong; Goodhart's that they will be gamed.
- Streisand EffectBoth describe situations where an attempt to prevent an outcome reliably produces it. Murphy's is about design failures; Streisand is about information suppression.