The EGC articles so far have covered where the EGC goes to earth (250.50–250.70), how it's sized (250.122), and how the GEC runs (250.64) and bonds the pipes (250.104). This article is the EGC's last two stops before it reaches a device: 250.130 — where the EGC connects (at service equipment: the (A) grounded-systems / (B) ungrounded-systems bonding split; in existing installations: the six permitted connection points for receptacle replacement and branch-circuit extensions — 2023 adds snap switches) — and 250.148 — how the EGC stays continuous in every box along the way (the 2023 "connected together" rule, device-removal continuity, the metal-box dedicated connection now sized from Table 250.122 on the largest OCPD in the box, the nonmetallic-box arrangement, and the deleted (E) solder rule that 250.8(B) now carries alone).
Both sections quoted verbatim from the 2017 NEC (scan artifacts corrected — the list is in the source note under the quote block):
250.130 is a routing section. It answers three questions, each with a one-sentence answer:
The third question is the one that makes 250.130 worth its own article: existing installations. Where an old branch circuit has no EGC at all — and you're replacing a nongrounding receptacle with a grounding one (2023: or a non-grounded snap switch with a grounded one), or extending the circuit — 250.130(C) gives you six places the new EGC may connect. Two are at the source (the grounding electrode system or its conductor, per 250.50), two are at the circuit's origin enclosure (its EGC terminal bar, or a sibling circuit's EGC from the same enclosure), and two are in the service equipment (the grounded service conductor for grounded systems, the grounding terminal bar for ungrounded systems). The EGC itself is still sized per Table 250.122 on the circuit's OCPD — the section picks the where, the table picks the size. Note the scope discipline: this permission covers receptacle/switch replacement and branch-circuit extensions only — it is not a general license to skip EGCs on new work.
250.148 is the section that makes "run a ground wire to the box" mean something. Its 2017 lead-in is the vaguest sentence in Article 250 — "all equipment grounding conductor(s) associated with any of those circuit conductors shall be connected within the box or to the box" — and the 2023 rework is essentially a fix for that vagueness (more below). Subdivision by subdivision:
One exception, unchanged across editions: the 250.146(D) isolated-ground EGC (the green-wire-to-the-receptacle-only arrangement for noise-sensitive equipment) shall not be required to be connected to the other EGCs or to the box. Everything else in the box connects; the isolated ground stays isolated by design.
The headline: 250.130 gained a device class (snap switches) and a note; 250.148 got the section the Mike Holt summary calls "in process for a while" — reworked so "which EGCs must be connected to each other or to the box" is explicit, with the deleted (E) solder rule absorbed by 250.8(B) and a brand-new box-jumper sizing sentence. A targeted machine check (verify_art55.py, 80 phrase-level assertions reading the on-disk 2017 NFPA scan and the on-disk 2023 NEC directly) confirms every claim below, item by item:
| Section | 2017 → 2023 (machine-verified) |
|---|---|
| 250.130 lead — separately derived → 250.30(A)(1); service → (A) or (B) | IDENTICAL — both routing sentences present unchanged in both editions (only the intra-reference space in "250.30(A)(1)" differs — a scan artifact, disclosed) |
| 250.130(A) — EGC bonded to grounded service conductor AND GEC | IDENTICAL — full sentence present unchanged in both |
| 250.130(B) — EGC bonded to GEC only | IDENTICAL — full sentence present unchanged in both |
| 250.130(C) — six permitted points, all six items | IDENTICAL except item (4) — 2023 names the snap switch as a circuit origin inside item (4) (delta below); items (1), (2), (3), (5), (6) word-identical in both |
| 250.130(C) Informational Note 406.4(D) (GFCI receptacles) | IDENTICAL text — 2017 unnumbered single note; 2023 numbers it "Informational Note No. 1" (2023 adds No. 2, below) |
| 250.148 Exception — 250.146(D) isolated-ground EGC not required to connect | IDENTICAL — full sentence present unchanged in both |
| 250.148(A) — connections per 110.14(B), insulation not required | CORE IDENTICAL — 2023 adds the 250.8 cite AND the new "connected together" first sentence (delta below) |
| 250.148(C) — dedicated connection "used for no other purpose" | CORE IDENTICAL — the dedicated-connection requirement survives; 2023 drops the 2017 hardware list (grounding screw / listed equipment / listed device) and adds the Table 250.122 sizing sentence (deltas below) |
| 250.148(D) — nonmetallic box EGCs brought in and arranged for connection | CORE IDENTICAL — "one or more EGCs brought into a nonmetallic outlet box" unchanged; tail reworded (delta below) |
2020 position — stated, not asserted: the on-disk 2020 source (a SlideShare text extraction of the NEC 2020 PDF) is an excerpts document — it carries only a handful of 250.x references (250.102, 250.118, 250.119, 250.122, 250.126) and no 250.130 or 250.148 text at all. So this page documents 2017 → 2023 and does not assert any 2020 position for these two sections. The identity and delta claims rest on the two on-disk full sources (2017 NFPA scan + 2023 CSV), the standard this series uses. No Mike Holt 2023 change-summary entry exists for 250.130 (the on-disk Mike Holt 2023 summary has zero 250.130 hits — verified); the 250.148 entry is quoted in the delta boxes. The 250.130 deltas are therefore on-disk-verified against the two full-text sources, stated as such.
Every table lookup below comes from the on-disk Table 250.122 (2023 CSV; rows 15–300 A cross-checked cell-for-cell against the on-disk 2017 study print — the table itself is unchanged across these editions) encoded once in compute_art55.js and asserted by the public test suite; the one computed figure (EX2, the 250.122(B) proportional increase) uses the shipped ch9Row core (Chapter 9 Table 8 circular mil areas) under node. EX5 is qualitative — the compliance matrix, no math. compute_art55.js → art55_numbers.json.
| Largest OCPD in the box | EGC jumper Cu | EGC jumper Al | Cu circular mils |
|---|---|---|---|
| 15 A | 14 AWG | 12 AWG | 4,110 |
| 20 A | 12 AWG | 10 AWG | 6,530 |
| 30 A | 10 AWG | 8 AWG | 10,380 |
| 40 A | 10 AWG | 8 AWG | 10,380 |
| 50 A | 10 AWG | 8 AWG | 10,380 |
| 60 A | 10 AWG | 8 AWG | 10,380 |
| 100 A | 8 AWG | 6 AWG | 16,510 |
| 200 A | 6 AWG | 4 AWG | 26,240 |
| 300 A | 4 AWG | 2 AWG | 41,740 |
| 400 A | 3 AWG | 1 AWG | 52,620 |
| 600 A | 1 AWG | 2/0 AWG | 83,690 |
| 800 A | 1/0 AWG | 3/0 AWG | 105,600 |
Read the seam at 30 A: Table 250.122 has no 30/40/50 A rows — an OCPD of 30 A or more (up to 60 A) lands in the 60 A row (10 AWG Cu), because the table's column is "rating or setting not exceeding (A)." Under 2017 the box jumper size was a judgment call built from 250.122 + 250.148(C)'s hardware list; under 2023 it is a direct table lookup on the largest OCPD in the box — this ladder is that lookup. The circular mil column is the shipped ch9Row core (Chapter 9 Table 8).
| Item | Value (core) |
|---|---|
| The circuit | 30 A branch circuit — minimum ungrounded 10 AWG Cu |
| The upsizing | Circuit upsized to 8 AWG Cu for a long run (10 AWG = 10,380 cmil → 8 AWG = 16,510 cmil, ch9Row) |
| EGC per Table 250.122 @ 30 A | 10 AWG Cu (10,380 cmil) — lands in the 60 A row (EX1's seam) |
| 250.122(B) proportional EGC | 10,380 cmil × (16,510 / 10,380) = 16,510 cmil |
| Standard size at or above 16,510 cmil | 8 AWG Cu (16,510 cmil, exact match) |
The 2023 (C) sizing sentence gives you the floor (10 AWG Cu for the 30 A OCPD); 250.122(B) says when the circuit upsizes, the EGC upsizes proportionately by circular mil area. A one-size circuit step is a one-size EGC step — 10 AWG Cu box jumper becomes 8 AWG Cu. (The 250.122 article covers the proportional rule in depth; here it lands on the very jumper 250.148(C) sizes.)
| Item | Value (core) |
|---|---|
| The circuit | 20 A, 12 AWG Cu — no EGC in the branch circuit (existing installation) |
| New EGC per Table 250.122 @ 20 A | 12 AWG Cu (6,530 cmil) or 10 AWG Al (10,380 cmil) |
| Permitted connection point 1 | Any accessible point on the grounding electrode system (250.50) |
| Permitted connection point 2 | Any accessible point on the grounding electrode conductor |
| Permitted connection point 3 | The EGC terminal bar in the enclosure where the circuit originates |
| Permitted connection point 4 | An EGC of another branch circuit from the same enclosure |
| Permitted connection point 5 | (Grounded systems) the grounded service conductor in the service equipment enclosure |
| Permitted connection point 6 | (Ungrounded systems) the grounding terminal bar in the service equipment enclosure |
The swap itself (nongrounding → grounding receptacle) is the 2017-scope trigger; 2023 adds the same permission for snap switches (404.9(B) via IN No. 2). Note the asymmetry: the EGC may connect at any of the six points, but it is sized only by Table 250.122 on the circuit OCPD — 20 A → 12 AWG Cu, regardless of which of the six points you land on.
| Item | 15 A circuit | 20 A circuit |
|---|---|---|
| Table 250.122 row (row for OCPD ≤ 60 A governs both) | 15 A row | 60 A row |
| EGC Cu | 14 AWG (4,110 cmil) | 12 AWG (6,530 cmil) |
| EGC Al | 12 AWG | 10 AWG |
A double-gang box feeds a 15 A and a 20 A circuit. The box jumper is sized for the largest OCPD protecting circuit conductors in the box — the 20 A device — so the jumper is 12 AWG Cu, not 14 AWG. Under 2017 the "largest OCPD" key for box jumpers was a 250.122 + 250.148(C) inference; the 2023 (C) sentence makes it the rule's own words ("based on the largest overcurrent device protecting circuit conductors in the box").
| Question | 2017 answer | 2023 answer |
|---|---|---|
| Solder-dependent connection in a box? | Prohibited — 250.148(E): "Connections depending solely on solder shall not be used" | Still prohibited — 250.148(E) deleted, but 250.8(B) independently forbids it (word-identical); 250.148(A) now joins 110.14(B) + 250.8 as the connection standard |
| Metal-box connection? | 250.148(C): grounding screw used for no other purpose, or equipment listed for grounding, or a listed grounding device | 250.148(C): "a connection used for no other purpose" + sized from Table 250.122 on the largest OCPD in the box (EX1/EX4) |
| Which EGCs connect together in a box? | 250.148 lead: "all EGCs associated with any of those circuit conductors" (vague — which ones?) | 250.148 lead: "the installation shall comply with 250.148(A) through (D)"; (A): "all EGCs spliced or terminated within the box shall be connected together" (Mike Holt: "making it clear which EGCs must be connected to each other or to the box") |
| Continuity when a device is removed? | 250.148(B): removal must not "interfere with or interrupt the grounding continuity" | 250.148(B): removal must not "interrupt the electrical continuity of the EGC(s) providing an effective ground-fault current path" (device order flips: luminaire, receptacle, other device) |
| Nonmetallic-box EGC arrangement? | 250.148(D): "arranged such that a connection can be made… requiring grounding" | 250.148(D): "arranged to provide a connection… requiring connection to an equipment grounding conductor" |
| 250.130(C) scope (existing installations)? | Nongrounding receptacle replacement + branch-circuit extensions only | Receptacle replacement + snap-switch replacement (2023) + branch-circuit extensions; IN No. 2 → 404.9(B) |
The one-line summary: 2023 made explicit what 2017 left as inference — which EGCs connect (all of them, in the box), what standard they connect to (110.14(B) + 250.8), and how big the box jumper is (Table 250.122, largest OCPD in the box) — while deleting the solder subdivision that 250.8(B) already covered.
Scope boundary, stated: this article is exactly 250.130 + 250.148 — the EGC-connection and box-continuity pair. 250.126 (enclosures not required to be grounded), 250.132 (short sections of raceway), 250.134 (alternative grounding paths), and 250.136 (grounding of separately derived systems) are adjacent sections outside this article's scope, as are 250.140 (frames of ranges/ovens/dryers — the 250.140 entry in the on-disk Mike Holt summary is context only) and 250.146 (receptacle-to-box). 2026 edition (flagged honestly): no on-disk source confirms the 250.130/250.148 section numbers moved for the 2026 cycle, so this page cites them as stable across 2017–2023 and presents the 2017↔2023 deltas above. Verify section numbers against the NEC edition adopted in your jurisdiction.
Section text. Verbatim NEC 2017 (official NFPA text on disk, nec2017_full.txt, region extracted to art55_2017_flat.txt): 250.130 + (A), (B), (C) + all six items + Informational Note, 250.148 + Exception + (A)–(E). OCR line-wrap reflow normalized; disclosed scan artifacts corrected (listed in the source note under the quote block): "250.30(A) (1)" (intra-reference space, as the scan printed it) re-cited as "250.30(A)(1)" in the 2023-comparison context only — the 2017 quote block keeps the scan's form; page furniture removed. No other wording altered.
2023 text. On-disk 2023 NEC CSV (art35_nec_csv.csv), rows 250.130, 250.130(A), 250.130(B), 250.130(C), 250.148, 250.148(A), 250.148(B), 250.148(C), 250.148(D) — complete for both sections (every subdivision has its row; no 250.148(E) row exists in the 2023 CSV — the deletion is on-disk). The 2023 CSV carries a trailing markdown link per body (stripped for the identity test only — never changes code text). Section headings (e.g. "(A) For Grounded Systems.") are shown as they appear in the 2017 scan; the 2023 CSV rows carry bodies only, so a 2023 heading change, if any, is not asserted.
Table 250.122 (worked examples). The 19-row table is encoded once in compute_art55.js: rows 15–300 A verbatim in both on-disk editions (2017 study print art20_table250122.txt, OCR-clean, and the 2023 CSV — cross-checked cell-for-cell this session, the table unchanged across these editions); the 400 A+ rows from the 2023 CSV (the 2017 scan's 400 A+ copper column has OCR misreads — the same disclosure posture the Article 20 250.122 article used). The one computed figure (EX2, the 250.122(B) proportional increase) uses the shipped ch9Row core (Chapter 9 Table 8 circular mil areas) under node — zero hand math.
Edition-delta claims (machine-checked). A targeted assertion script (verify_art55.py) reads both on-disk sources and asserts 80 phrase-level checks: 22 probes on the 2017 text (verbatim, OCR artifacts as-scanned), 20 probes on the 2023 text (verbatim from the CSV), 13 identity claims (present in both, alpha-normalized so OCR hyphen/spacing artifacts cannot cause false fails — including the 250.8(B) solder prohibition word-identical across the 2017 scan and the 2023 CSV), 15 2017-only probes (each delta's old form: present in 2017, absent in 2023 — including the 250.130 lead without the snap-switch clause, the 250.148 lead "associated with any of those circuit conductors", the device order "receptacle, luminaire", the grounding-screw hardware list, and the (E) solder prohibition), 9 2023-only probes (each delta's new form: present in 2023, absent in 2017 — the snap-switch clause, the "connected together" sentence, the Table 250.122 sizing sentence, the 404.9(B) note, the effective-ground-fault-current-path standard, the "(A) through (D)" range), 2 Mike Holt probes (the 250.148 entry present + its "making it clear which" quote; no 250.130 entry anywhere in the 2023 change summary — stated on the page), and 5 scope checks (CSV complete for both sections, no 250.148(E) row, 2017 flat text furniture-free and ending at 250.148(E), the 2017 250.130 region carrying exactly one Informational Note with no 404.9 reference). All 80 pass (re-runnable by the test suite). No Mike Holt 2023 change-summary entry exists for 250.130 (zero hits on disk) — every 250.130 delta is on-disk-verified against the two full-text sources, stated as such; the 250.148 deltas are corroborated by the on-disk Mike Holt entry. 2020 position is not asserted: the on-disk 2020 source carries no 250.130/250.148 text (excerpts document — verified).