A question turns up on the electrician forums in some form every few weeks, and one version of it collected more than two thousand upvotes: if I drove a copper rod into the dirt and ran a wire from it to the thing that shocked me, would the breaker trip?
No. It would not. And the fact that so many people in the trade find that answer surprising is the clearest evidence we have that the words we use for this are badly chosen.
One word doing two unrelated jobs
Grounding, in ordinary trade speech, covers two entirely separate functions that happen to share a color of wire and a bus bar.
The first is the connection to the earth itself — the rod, the plate, the concrete-encased electrode, the water pipe. Its job has nothing to do with your breaker. It references the system to earth so that the voltage on your conductors stays in a predictable relationship to the ground you are standing on, and it gives lightning and utility-side surges somewhere to go that is not through the house. That is worth having. It is not a safety device for branch-circuit faults.
The second is bonding: metallically tying together everything conductive that could become energized, and tying that whole assembly back to the source through a deliberate, low-impedance path. That is what clears a fault. The equipment grounding conductor and the main bonding jumper are not earth connections. They are a return path to the transformer winding, and the earth plays no part in it.
Run the arithmetic on the ground rod
The reason the copper rod cannot do the first job is dull and unarguable. Dirt is a poor conductor. Article 250 treats twenty-five ohms as the threshold at which a single rod is considered adequate on its own, which tells you the order of magnitude the code expects. You can read the section for yourself through NFPA's free online access to NFPA 70, which needs a free account.
Take that twenty-five ohms and put a hundred and twenty volts across it. Ohm's law gives you a little under five amps. Your twenty-amp breaker will hold that all day. It will hold it while the fault current warms the earth and while the energized metal sits there at very nearly line voltage, waiting for somebody to complete a better path than the dirt. Which, as it happens, a person standing on damp concrete frequently is.
Now do the same arithmetic on the bonded path. A copper equipment grounding conductor back to the panel and through the main bonding jumper to the neutral is a fraction of an ohm. The same hundred and twenty volts produces hundreds of amps, the breaker's magnetic trip sees it instantly, and the circuit is dead before anybody has decided what the noise was. That is the entire mechanism. The code calls it an effective ground-fault current path, and the phrase is doing real work — effective means low enough impedance to operate the overcurrent device, and nothing else.
What happens when the path is a person
NIOSH investigated a case in North Carolina in 1994 in which two men died in the crawlspace of a private house. An HVAC crew was setting a new unit and running duct. A worker was driving screws through aluminum hanger straps into the floor joists, and the edge of one strap cut into the branch-circuit wiring above it. The strap and the drill became energized. The investigators recorded that the circuit was completed through the worker and through pipes in the area to ground. The contractor, called down by the homeowner who had heard something, reached the man, became part of the same circuit, and also died. The report is FACE 94-17, and it is three pages long.
Read that case as a bonding case and it is very plain. There was a fault. There was metal at line potential. There was a path to ground available through the plumbing. What there was not, at the moment it mattered, was a low-impedance metallic route back to the source good enough to open a breaker. So the current used what it had, and what it had included people.
This is also why the bonding of metal water piping and structural steel is not housekeeping. If everything conductive in a building is tied together, there is no voltage difference between the pipe in your left hand and the conduit in your right. Take the bond away and you have manufactured a potential difference across a person. The gutter that shocked the homeowner in that forum thread was doing exactly this: something upstream had failed, the gutter had gone live, and the gutter was not bonded to anything that could carry the fault away.
There is a mirror-image error worth naming while we are here, because it is at least as common as the ground rod one. If you bond the neutral to the equipment grounding conductor somewhere downstream of the service — in a sub panel, in a detached building, in a range or dryer receptacle wired the old way — you have not added redundancy. You have put normal load current onto the metal parts of the building. It will work. Nothing will trip. But the conduit, the boxes and the appliance frames are now carrying a share of the return current every time the circuit is used, and any break in the neutral upstream of that point puts the full load voltage on them. The code is specific about where that connection is made precisely because making it twice is worse than making it once.
Why this one is worth going back over
Electricians are killed at a rate of 2.89 per 100,000 workers, against 0.11 across all occupations, in ESFI's compilation of the 2011 to 2024 data. Most of that is contact with energized parts, not subtleties of Article 250. But the subtleties are where the quiet failures live — the isolated ground bar in a sub panel, the neutral and ground bonded on the wrong side of the disconnect, the equipment grounding conductor terminated to a rod because it was closer than the panel.
Those do not announce themselves. The installation energizes, the lights come on, the inspector may well pass it, and the defect sits there for twenty years until a fault arrives and finds out what you built. That is the argument for understanding Article 250 properly rather than pattern-matching your way through it: it is the one area of the code where a wrong answer produces no symptom at all until the day it matters.
If Article 250 is the part you have always half understood, our guide to NEC bonding and grounding requirements works through it with the current paths drawn, which is usually the thing that makes it click.