Two-Prong Outlets and the Ungrounded House
Why this matters
A two-prong receptacle is honest: it tells you there is no equipment ground. The dangerous receptacle in an older house is the three-prong one that looks grounded and is not, either because someone swapped it in on a two-wire circuit or because someone ran a jumper from the neutral to the ground screw so a plug-in tester would show "correct". Appliance techs, HVAC techs and installers plug equipment into these receptacles every day, and their installation instructions almost always require a properly grounded one. A washer or refrigerator plugged into a bootlegged receptacle can end up with its metal case connected to the neutral, which is fine until the neutral fails, and then the case is live.
What "ungrounded" means
Modern branch circuits carry three conductors: a hot, a neutral, and an equipment grounding conductor (EGC). The neutral carries normal current back to the panel. The EGC carries no current in normal use; its job is to give a fault a low-resistance path home, so that if a hot conductor touches a metal case the breaker or fuse opens instead of leaving the case energized.
Houses wired before the early 1960s, with knob-and-tube, cloth-sheathed or early plastic cable, usually have no EGC. The NEC began requiring grounding-type receptacles for new branch circuits in the early 1960s, and cable with a ground conductor became standard around the same time. Two-prong receptacles in those houses are original and accurate. Metal boxes on those circuits are not grounded by being metal; without an EGC or a grounding path back to the panel, the box is just a box.
The code options for an old box
When a receptacle in a box with no grounding means is replaced, the NEC, in 406.4(D)(2) of the 2017 and later editions as adopted by your jurisdiction, gives three choices:
- Replace it with a non-grounding (two-prong) receptacle.
- Replace it with a GFCI-type receptacle, marked "No Equipment Ground". The GFCI trips on current leaking to anything other than the circuit's own neutral, a person included, so it protects people without needing a ground.
- Replace it with a grounding-type (three-prong) receptacle protected by a GFCI upstream, either a GFCI receptacle earlier on the circuit or a GFCI breaker, with the receptacle marked "GFCI Protected" and "No Equipment Ground".
The same section bars connecting an equipment grounding conductor from the GFCI to downstream outlets in that arrangement, because there is no real ground to connect. A plain three-prong receptacle with no ground and no GFCI protection is not among the options, which makes it a violation wherever this section is adopted. The other legitimate path is to add a real EGC: 250.130(C), as adopted, permits running one to a receptacle from a set of approved grounding points. Some jurisdictions and newer editions also require GFCI or arc-fault protection when a receptacle is replaced in a location where that protection is now required, so the "replacement" rules are an electrician's lookup for your code cycle.
What GFCI protection does not do. It protects people from shock. It does not give a surge protector a path to divert surges into, it does not give an appliance's electronics the ground reference some of them expect, and it does not meet an installation manual that says "grounded receptacle." "GFCI protected, no equipment ground" is a safe receptacle and an ungrounded one.
The bootleg ground
A bootleg ground is a jumper from the neutral terminal to the ground terminal on a receptacle that has no EGC. It is done to make a three-prong receptacle "test correct," and it does.
TRUE GROUND BOOTLEG GROUND
hot -----> brass screw hot -----> brass screw
neut -----> silver screw neut -----> silver screw
EGC -----> green screw |
(own wire, back to jumper ----+
the panel ground bar) |
+-----> green screw
fault current returns
on the EGC, not on the neutral and ground are one
neutral point; no wire to the panel
Why a plug-in tester passes it. A three-light tester checks for voltage between hot and neutral, hot and ground, and neutral and ground. With a jumper in place, hot-to-ground reads line voltage through the neutral, neutral-to-ground reads nothing, and the lights show "correct". The tester cannot tell the difference between a ground wire and the neutral wearing a jumper.
Why it is worse than a two-prong receptacle. The appliance case is now bonded to the neutral. If the neutral opens anywhere upstream, a loose splice or a failed connection, the case rises toward line voltage through the appliance's own load. If hot and neutral are reversed upstream, the ground pin, and the case, is live outright. Normal neutral current also flows on the case and on anything bonded to it. A two-prong receptacle forces the question of grounding into the open; a bootleg hides it.
See related: Bootleg Ground Versus True Ground On An Old Receptacle Decision Tree, which owns the electrician's diagnosis.
Why it matters to appliance and HVAC trades
Installation instructions. Washers, dryers, refrigerators, dishwashers and most plug-in appliances require a properly grounded three-prong receptacle, and many say not to use an adapter or remove the ground pin. An install onto a receptacle you have reason to think is ungrounded or bootlegged is an install outside the manufacturer's instructions.
Cheater adapters and cut pins. A three-to-two adapter's tab only grounds the appliance if the cover screw it attaches to is itself grounded, which on an ungrounded circuit it is not. A cut ground pin removes the protection outright. Neither is an install.
Furnaces and boilers. Hot-surface and spark ignition controls commonly prove flame by flame rectification, a small current that flows from the flame sensor through the flame to the grounded burner and back to the control board. A weak or missing ground at the appliance can produce flame-sensing faults that look like a dirty sensor. Many controls are also polarity sensitive. An HVAC tech chasing intermittent flame loss in an old house checks the ground path before replacing parts.
Surge protection. A plug-in surge protector on an ungrounded receptacle, GFCI or not, has nowhere to send a surge. It is an expensive power strip.
Worked example: the laundry receptacle that tested correct
An appliance installer is delivering and connecting a washer in a 1955 house. The laundry receptacle is three-prong, and a plug-in tester shows "correct". In the unfinished basement below, the cable rising to the laundry box is cloth-sheathed with two insulated conductors and no ground wire visible, which is what a house this age would have.
The contradiction. A three-prong receptacle that tests correct, fed by a cable that appears to have no ground, has three possible explanations: an EGC was retrofitted by another route, the box is grounded some other approved way, or the receptacle is bootlegged. Only the third is a problem, and the tester cannot separate them.
A non-invasive check. Without opening the box, the installer uses a multimeter rated for the circuit, probes with finger guards, and reads neutral-to-ground at the receptacle face: 0.0 V with nothing running. Then a known load goes on the same receptacle's other outlet, a 1,500 W heater drawing about 12.5 A at 120 V, for under a minute, and the reading is repeated.
What the numbers mean. With a true, separate EGC, load current flows on the neutral and not on the ground, so the neutral's own resistance puts a small voltage between them. Suppose the run back to the panel is about 50 ft of 14 AWG copper. At roughly 3.1 ohms per 1,000 ft (NEC Chapter 9, Table 8, uncoated solid copper), 50 ft of neutral is about 0.15 ohm, and 12.5 A across 0.15 ohm is roughly 1.9 V. A reading in that neighborhood is what a true ground looks like. The installer reads 0.0 V under load. When neutral and ground are the same point at the receptacle, there is no length of separate wire between them for a voltage to appear across, so a bootleg reads zero under load.
The read. A zero neutral-to-ground reading under a 12.5 A load, on a receptacle fed by a cable with no visible ground, strongly suggests a bootleg. It is not proof: a very short run or an unusual retrofit can read low too. The proof is opening the box, which is the electrician's.
The decision. The washer's instructions require a grounded receptacle. The installer sets the washer in place, connects water and drain, and does not plug it in, labels the receptacle "suspected bootleg ground, do not use for washer, electrician to verify," and records the readings: tester "correct," N-G 0.0 V unloaded, N-G 0.0 V at 12.5 A, basement cable appears two-wire with no ground. Time on site came to about 1 hour of the 1.25 booked, and the plug-in is a 15 minute return.
What the customer hears. "This outlet tests right on my tester, but the wiring under it looks like it has no ground, and one reading I took suggests someone jumpered it to fake one. I can't tell for sure without opening it, which is an electrician's job. Your washer needs a real ground, so I've left it unplugged."
What would have changed the answer. A reading near 1 to 3 V under that load would have pointed to a real, separate ground path, and the install would proceed. A two-prong receptacle in the same spot would have been an honest ungrounded circuit with the same answer: no install until an electrician provides a grounded receptacle. A GFCI receptacle marked "No Equipment Ground" would protect a person from shock and still fall short of the washer's grounded-receptacle requirement.
What gets mistaken for a ground
A metal box on old cable. Metal is not a ground unless something connects it to the panel. Old armored cable may or may not provide that path. See related: Old Armored Cable and the Ground Path It May Not Provide.
A ground wire to a water pipe. A conductor clamped to a nearby cold water pipe may or may not be an approved grounding point under 250.130(C) as adopted, depending on where on the pipe it lands and whether the pipe is continuous metal back to the service. Plastic repairs in a supply line break that continuity without anyone noticing.
A GFCI with a working test button. The test button on a GFCI receptacle creates its own internal imbalance and works without a ground. A plug-in GFCI tester, by contrast, leaks current to the ground pin and may not trip a GFCI on an ungrounded circuit, which is why a working GFCI can look broken.
References
- NFPA 70, National Electrical Code, 406.4(D)(2) (replacement of non-grounding receptacles), 250.130(C) and Chapter 9 Table 8, as adopted by your jurisdiction
- See related: Bootleg Ground Versus True Ground On An Old Receptacle Decision Tree
- See related: Cloth-Sheathed and Early Plastic Cable; Old Armored Cable and the Ground Path It May Not Provide
- See related: Recognize, Record, Route: The Rule for Legacy Materials