Splitter - Riving Knife

identity

• name: Splitter (also riving knife — a distinct, later mechanical evolution of the same job, not a synonym; see lineage) — regulatory bodies in both the US and Canada use a third formal term, "spreader," interchangeably in code text: OSHA's own workplace standard (29 CFR 1910.213, covering hand-fed circular ripsaws) and CCOHS's current fact sheet both use "spreader" as the governing regulatory word even though "splitter"/"riving knife" are the terms actually used on the shop floor and in retail listings [SR11][SR17].
• family / category: A table-saw safety accessory — a thin metal (or, in one modern aftermarket line, engineered plastic) plate/knife positioned directly behind the saw blade, in the same plane as the cut. It performs no cutting or shaping operation of its own — it is a kerf-control and anti-kickback device, not a tool in the sense of doing work to the material. (Scoped to the table saw as of Run 567's absorption — see the Correction note above; the entry's earlier radial-arm-saw and circular/miter-saw framing has been removed as unsupported.)
• one-line: A splitter is a fixed plate aligned behind a table-saw blade; a true riving knife is mounted to the blade-support assembly so it rises, falls, and bevels with the blade. Both help keep the freshly cut kerf from closing on the blade's rising rear teeth — a passive anti-kickback device that reduces one important class of table-saw incidents, not a device that prevents direct blade contact or every form of kickback. (The reviewer's corrected framing, absorbed Run 567 — replaces the original's "the mechanical fix for the single most common serious table-saw injury mechanism" superlative, which the CPSC data did not support.)
• scope note — what this entry covers and what it deliberately does not: This is a standalone entry for the device itself: its two real mechanical forms (fixed splitter vs. arbor-mounted riving knife), its market products, and its own dado-removal misuse tension. It does not re-tell the Table Saw entry's own extensive Petroski arc (splitter → riving knife → flesh-detection), UL 987 regulatory timeline, or the SawStop/CPSC AIM-rule litigation and rulemaking history — that entry already covers that ground in depth (Run 228) and this entry cross-links to it rather than duplicating it. What follows here is new material, direct corroboration, or a correction of what the Table Saw entry already claims — not a repeat.

✦ how NOT to use it

Not-for: jobs it should never do

• ★ Never rely on it as the sole safety measure for the specific non-through cuts it genuinely cannot physically accommodate — most reliably, a dado stack. (Corrected Run 567 — the original named "a dado stack, rabbet, or tenoning-jig pass" as automatically requiring removal; a compatible low-profile riving knife is designed to remain installed for most rabbet/groove/tenoning-jig work. Check the saw and cutter manual rather than assuming.) The device is not a universal guard that should always remain in place regardless of operation, but the automatic-removal case is narrower than it may look [SR26][SR30].
• Never install a riving knife thicker than the blade's kerf, or thinner than the blade's body — the device will either bind against the blade or fail to keep the kerf open, defeating its entire purpose (see history for the sourced thickness rule) [SR14][SR15][SR22].

Dangerous: misuse that injures the user

• ★ Removing the guard/riving-knife assembly and not reinstalling it — this entry's most concretely, government-data-backed DANGEROUS finding, now stated within its actual evidentiary limits. Among CPSC's own studied blade-contact injury cases, a riving knife/spreader was attached in only 19.6% and a separate anti-kickback pawl assembly in only 23.9%, with "guard removed" cited as the reason no blade guard was present in 75% of no-guard cases [SR28]. These are injury-population figures, not general-ownership or removal-rate estimates — they show the device is often absent at the moment of injury among people who were hurt, which is the load-bearing finding, without claiming to measure how often saw owners in general remove it. OSHA's own Machine Guarding eTool documents a named real accident report of a finger amputation from a dado-cut binding incident directly on its guidance page [SR10].
• Kickback — one important mechanism this device reduces, not the only serious table-saw injury pathway. (Reframed Run 567 per the reviewer's fix 3 — the original called kickback "the single most common serious table-saw injury mechanism," which CPSC's own data does not support; blade contact is the dominant injury mechanism generally, and kickback is one contributing factor among several.) If the kerf closes behind the blade (from internal grain tension, an out-of-parallel fence, or operator twist — all already documented in the Table Saw entry's own dangerous-misuse section), the blade's rising rear teeth can grab the workpiece and throw it back toward the operator with real force; a correctly installed and correctly sized splitter/riving knife reduces this specific risk but does not prevent direct blade contact or every form of kickback [SR11][SR28].

Damaging: misuse that ruins the tool or the work

• Forcing a mismatched-thickness riving knife into position can bind against the blade body, generate excess friction/heat, or fail to track the blade cleanly through height and tilt adjustments — a real mechanical consequence of ignoring the thickness rule above, reasoned directly from the sourced spec relationship rather than a separately documented failure account this run [SR14][SR22].

Beginner-trap: mistakes that look fine until they aren't

• ★ Correctly removing the riving knife/splitter for a dado pass (or another genuinely incompatible operation), then forgetting to reinstall it before the next rip cut. (Narrowed Run 567 to match the corrected removal case — dado stacks, not a blanket non-through-cut list.) This is the mechanism this entry's own history and matching sections both point to directly: the removal itself is correct and necessary when the cutter genuinely doesn't fit (Rockler's own guidance confirms this for dado work) [SR26], but the tool "looks fine" with it off — the saw still cuts — right up until a kickback event occurs on the very next through-cut. This mirrors and reinforces, rather than duplicates, the Table Saw entry's own already-documented beginner-trap language on guard removal "for convenience."
• Assuming the removal case is broader than it is — assuming a rabbet, groove, or tenoning-jig operation automatically requires pulling the riving knife, when a compatible low-profile knife is designed to stay installed for most of them. (New beginner-trap, added Run 567 — the mirror image of the original entry's own error, now named explicitly as a mistake a reader could make even after the correction, if they don't check their own saw's manual.)
• Assuming an older or bare-bones saw's missing riving knife means "my saw is just old or cheap" rather than treating the gap as an active retrofit opportunity — real, current, reasonably priced aftermarket options exist (MicroJig ~$16–25, Shark Guard ~$75–120) [SR23][SR24], so a maker working on an older saw is not stuck without a fix.

Lineage-misuse: treating it like its ancestor or descendant

• Treating a fixed splitter and a true riving knife as functionally interchangeable, when they are not. A fixed splitter is set back from the arbor far enough to clear the blade at full raise — meaning on shallow cuts, a real, sourced, mechanically-explained stretch of kerf near the blade goes unprotected before the splitter's fixed position takes over; a true riving knife's arbor-mounted tracking eliminates that gap at any blade height [SR15]. A maker who assumes an older saw's fixed splitter behaves identically to a modern riving knife is misunderstanding a genuine mechanical difference between two distinct generations of the same basic idea, not two names for one thing. A second version of this trap, corrected Run 567: treating an older-era fixed splitter's stricter removal habits (see human side) as still true of a modern low-profile riving knife. The historical practice of pulling the whole guard assembly for any non-through cut belonged to the older, bulkier hardware; it is not the correct model for a compatible modern riving knife.
• Treating the device as a substitute for flesh-detection technology (SawStop-style), or vice versa — a live, real, named practitioner distinction, not a settled either/or. A 2008 Sawmill Creek poster (michael osadchuk) named SawStop directly as "the safety feature that really works to virtually remove the hazard," distinct in kind from a mechanical guard [SR9]; CPSC's own 2023 rulemaking record independently and institutionally states that "currently available safety devices, such as the modular blade guard and riving knife, do not adequately address the unreasonable risk of blade-contact injuries" [SR13] — both a practitioner and a federal regulator draw the same real distinction between mechanical kerf-control and active flesh-detection, a distinction the Table Saw entry's own Petroski arc already documents in depth and this entry does not re-litigate.

correct use

(brief — the workbench holds deep operation; this entry documents the device's own scope, not table-saw technique generally, which the Table Saw entry already covers in depth.)

• Leave it installed for every through-cut/ripping operation. This is the device's designed job and the operation it is actually engineered and rated for — MicroJig's own spec sheet states its splitter is explicitly "for 90 degree thru-cuts only" [SR23], and CCOHS's own current guidance instructs: "Use a guard with a spreader (riving knife) and anti-kickback fingers for all ripping or cross cutting operations" [SR11].
• Verify blade-thickness/kerf compatibility before installing a replacement or aftermarket riving knife — the thickness must sit strictly between the blade's body thickness and its kerf width (see history for the sourced rule and real manufacturer numbers); a mismatched riving knife either binds against the blade body or fails to keep the kerf open.
• ★ Keep a compatible low-profile riving knife installed for non-through cuts whenever the saw manufacturer permits it. (Corrected Run 567, the reviewer's fix 1 — the original told readers to remove it for "rabbets, some grooving, tenoning-jig work," which is backwards for a modern low-profile knife.) Remove the standard knife only when the cutter or operation is genuinely incompatible — commonly a dado stack — and reinstall the specified guard/knife before the next compatible cut. Follow the saw and cutter manufacturer's manuals rather than a blanket rule; forgetting to reinstall after a correct, necessary removal is this entry's own named beginner-trap (below).

✦ matching (the workbench's first read — the tools section: matching a user's owned tools to their project)

• does: keeps the saw kerf open directly behind the blade; reduces one important class of kickback by preventing the workpiece from pinching the blade's rising rear teeth. It performs no cutting, shaping, or material-removal operation itself — a safety/kerf-control device, not a working tool in the corpus's usual sense. (Corrected Run 567: does not prevent direct blade contact or every form of kickback — see *identity.)*
• suits: ripping and other through-cuts on a table saw — the primary, load-bearing, manufacturer-confirmed use case [SR23] — and most non-through cuts (rabbets, grooves, tenoning-jig work) when a compatible low-profile riving knife is installed and the saw manufacturer permits it. Any project stage involving repeated rip cuts (breaking down sheet goods to width, resawing dimensional lumber) is where this device does its most load-bearing work. (Scope note, Run 567: the original also listed "and, less commonly, a radial-arm saw" here — removed; see *lineage.)*
• kit-fit: pairs with — and is routinely sold bundled with — the blade guard and anti-kickback pawls as one combined guard assembly [SR20][SR24]. Nothing in this research genuinely substitutes for its specific kerf-control function; it does not replace a push stick or push block (hand-safety, a separate concern the Table Saw entry covers) and does not replace verified outfeed support.
• not-for (project-keyed) — ★ dado stacks, corrected Run 567 to be the primary, reliable removal case rather than a broad list. The device's fixed kerf-width sizing physically cannot fit inside a dado stack's wider cut — a real, manufacturer/retailer-sourced physical limitation (Rockler [SR26]), not a matter of user preference. Other non-through operations (rabbets, grooves, tenoning-jig work) are compatibility- and manufacturer-instruction-dependent, not an automatic removal case — check the saw and cutter manual before assuming removal is required. A user whose project calls for dado work should expect to remove the riving knife/splitter for that operation and treat reinstalling it before the next rip cut as a mandatory step, not an afterthought.
• access: a beginner should be told plainly to leave this installed by default and to treat its removal as an exceptional, deliberate, temporary step tied to a specific, genuinely incompatible operation (most often a dado stack) — never a standing convenience. The genuine beginner trap here is not misunderstanding the device, but forgetting to reverse a correct, necessary removal (see how NOT to use it).
• revival: not really applicable in the corpus's usual "museum → act of making" sense — this is a current, actively regulated, actively manufactured safety device, not a historical tool a maker might rediscover. The closest analog: an owner of an older saw that shipped with only a fixed splitter (or nothing) may reasonably "revive" the safety margin of a modern saw by retrofitting an aftermarket fixed splitter (MicroJig) or a custom-built riving knife (Shark Guard) rather than replacing the whole machine [SR23][SR24].

anatomy

• the knife/splitter blade itself — steel on nearly every product surveyed (SawStop, Delta, Grizzly OEM parts); one modern aftermarket line (MicroJig's MJ Splitter, thin-kerf version) uses a specially engineered flexible plastic instead, marketed explicitly for "flexibility to prevent binding, rigidity to prevent kickback" [SR23]. A stainless-steel custom line (Shark Guard) is also sold [SR24]. Thickness sits, by design, between the blade's body thickness and its kerf width — see history for the sourced rule and real working numbers.
• mounting bracket / clamp assembly — on a true riving knife, this attaches to the arbor or trunnion/blade-support assembly so the knife's position tracks blade height and tilt automatically; on a fixed splitter, it mounts at a fixed height independent of blade position, which is the direct mechanical source of the "unprotected kerf distance on shallow cuts" finding above [SR15][SR16].
• quick-release mechanism — a genuinely documented modern feature, engineered specifically to remove the friction that made older assemblies "too inconvenient to reinstall." SawStop's own mechanism (lift a clamping handle, slide the knife clear of two positioning pins, lift out) is marketed in the company's owner's-manual language as "so easy to use you'll actually want to have the guard and riving knife on the saw at all times" [SR19]. Bosch's REAXX "Smart Guard System" bundles a 1/8"-thick adjustable riving knife with pawls and a barrier guard, adjustable via a screw/locking nut and removable in seconds [SR20]. Honest gap, flagged rather than filled with inference: this run could not find a study isolating removal rates on quick-release-equipped saws versus older, bolt-out-required saws — the CPSC injury-population data above (history) supports that the device is often absent at the moment of injury, but not specifically that quick-release mechanisms have changed that rate one way or the other [SR28].
• combined guard assembly — the riving knife/splitter, blade guard, and anti-kickback pawls are routinely sold, mounted, and discussed as one modular unit rather than independent parts: Shark Guard bundles its riving knife with a separate "Clamp and Anti-Kickback Pawls Assembly" as a companion product [SR24], and Bosch's REAXX system is engineered and marketed as one integrated "Smart Guard System" [SR20].

variations (living)

(Row removed, Run 567: the original table carried a "Radial-arm-saw and some circular-saw variants" row, sourced to Wikipedia and US Patent 5,063,806. The reviewer found the patent describes an unrelated carriage-speed-control mechanism, not this device — see the *lineage correction note. The row is removed rather than corrected, since no source this run or the absorbing run could independently establish the claimed cross-use.)*

trades

• Woodworking / cabinetmaking — the primary trade context; ripping sheet goods and dimensional lumber on a table saw is the device's core, everyday use case.
• General construction / jobsite carpentry — benchtop and jobsite table saws carry lighter-built guard assemblies with their own tolerances; some older jobsite saws still in the field predate UL 987's 2014 riving-knife compliance date entirely (documented in the Table Saw entry's own lineage-misuse section, not re-sourced independently here).
• Metalworking-adjacent note, honestly flagged as out of scope: this run found no evidence the splitter/riving-knife device (or its underlying anti-kickback principle) is used on metal-cutting circular saws or cold saws — those tools' kickback risk profile and guarding approach were not researched this run and should not be assumed identical.

lineage

• ★ the fixed splitter predates the riving knife by a long margin, and no clean single "riving knife invented by X" patent could be found — an honest gap, stated directly and held more firmly after absorption. A modern splitter patent (US 7,293,488B2, "Splitter for circular table saw and method of installation," fetched via Google Patents) cites US 482,507 (1892) as prior art — but the 1892 patent's own full text could not be independently opened this run or the absorbing run (Google Patents blocked the fetch outside its search-result provenance), so its exact claims remain unverified. (Correction, Run 567: the original entry called this "the earliest well-documented splitter-adjacent US patent" — the reviewer correctly flagged that an unread patent cannot be characterized or credited that way. It is now carried only as an unread prior-art citation inside a later patent, not as this tool's own established origin point. No defensible single inventor or clean origin date exists for either the fixed splitter or the riving knife.) [SR1]. Early anti-kickback pawl patents — a related but mechanically distinct device, commonly bundled with the splitter into one guard assembly (see anatomy) — include US 1,074,198 (a toothed roller), US 2,095,330 (rubber-covered anti-kickback fingers mounted on a splitter), and US 3,232,326 (finger-like leaves depending from a splitter) [SR2]. The riving-knife concept itself (arbor-mounted, tracking blade height and tilt, sitting much closer to the blade than a fixed splitter can) appears to have emerged as standard European sawbench practice well before it was named, patented, or required in the US — a genuinely disputed origin, not reconciled by this run's research [SR3][SR4].
• ★ the regulatory divergence between Europe and the US is the clearest, best-sourced lineage story this entry carries. A named 2008 Sawmill Creek forum poster (Brad Naylor) states: "All table saws have had to be fitted with riving knives for the last forty years or so... 'Splitters,' as you have, are unknown [in Europe]" — this figure traces to the EU's harmonized machinery-safety standard EN 1870-1 ("Woodworking machines — Safety — Circular sawing machines"), but this run could only confirm later revision years (1999, 2007+A1:2009) in standards-catalog searches; the "forty years" figure is a forum anecdote, not an independently verified mandate date, and is flagged as an open gap rather than treated as fact [SR4]. In the US, by direct contrast, Woodshop News (David DeCristoforo, Oct. 24, 2008) states plainly: "Since the mid-1980's, virtually every table saw sold in this country has been equipped with some kind of 'splitter'" — dating the fixed-splitter era in American manufacturing to roughly the mid-1980s, decades after the fixed splitter's own patent lineage above, and editorializing that manufacturers "fought against this requirement for years" [SR7]. The US regulatory timeline for the true riving knife, corrected Run 567 to the controlling CPSC record rather than trade press: UL 987's sixth edition, published January 2005, added the riving-knife requirement; industry adoption accelerated so that the new systems were widely available by 2008; separate modular blade-guard requirements took effect January 31, 2010; the riving-knife requirement's own compliance effective date was January 31, 2014 [SR5][SR13a]. (Fine Woodworking's June 2011 "2008/2014" shorthand, and Wikipedia's citation of it, collapsed four distinct dates into two — corrected here per the reviewer's fix 4.)
• ★ a real, sourced mechanical reason the riving knife is a genuine improvement over the fixed splitter, not just a rebranding — from two independent named forum accounts. Sawmill Creek poster Nissim Avrahami (a long-time, named contributor) explains, with photos, that a riving knife is arbor-linked via a pivoting bracket, so its position relative to the blade's curve stays constant at any blade height — while a fixed splitter must be set back far enough from the arbor to clear the blade at full raise, meaning on shallow cuts a fixed splitter leaves a longer stretch of kerf unprotected before the workpiece reaches it. Poster Joe Spear independently makes the same point in different words in the same thread — real cross-corroboration from two named posters, not one voice repeated [SR15].
• contemporaries, cross-checked against the sibling Table Saw entry: the blade guard (a shroud shielding the operator from incidental contact from above) and anti-kickback pawls/fingers (spring-loaded or gravity teeth that let wood pass forward but dig in against being thrown back) are routinely sold, mounted, and discussed as one combined "modular" guard assembly with the splitter/riving knife, not as fully independent parts a user mixes freely — confirmed both by the Table Saw entry's own anatomy section and by this run's market research (see anatomy and matching) [SR20][SR24].
• Correction, Run 567 — a cross-tool use claim removed. The original entry claimed the same device is "also fitted to radial arm saws and to some handheld electric circular and powered miter saws," citing Wikipedia and US Patent 5,063,806 ("Anti-kick forward device for radial arm saws"). The reviewer read the patent directly and found it describes a carriage-speed control mechanism, not a splitter or riving knife — it does not corroborate cross-use of this device on a radial arm saw. The claim and the patent citation are removed from this entry (see proof of existence). Whether radial-arm-saw spreaders or pawls exist as a genuinely distinct application is an open question for future research, not established fact.

history

• ★ the Petroski "form follows failure" story here is not about the device's own invention, but about a slower, harder-to-fix failure: the guard assembly's habitual removal. The device's basic physics — a thin blade riding in the kerf keeps the kerf from closing on the rising rear teeth — was understood and patented in some form by the late 19th century, though the specific earliest patent is unconfirmed (see lineage). The much longer, better-documented history is the decades-long gap between "the fix exists" and "the fix is actually left installed," which is why the misuse layer below, not the invention story, carries this entry's real weight.
• the thickness rule — the device's one hard physical constraint, directly sourced. Wikipedia's Riving Knife article, citing UK HSE guidance ("Circular saw benches — Safe working practices"), states the governing rule directly: a splitter's or riving knife's thickness "should be greater than the body of the saw blade but thinner than its kerf" [SR14]. Sawmill Creek forum posters supply real working numbers that confirm this in practice: a Powermatic PM2000's riving knife is 0.125" thick, meaning any thin-kerf blade narrower than that is mechanically incompatible with that specific riving knife [SR15]. Manufacturer spec sheets independently confirm the same principle at different scale: Grizzly's T30329 riving knife is 0.08" thick against a blade body of 0.07" and a kerf of 0.09" [SR22].
• ★ the removal case is narrower than the original entry stated — corrected here, the reviewer's safety-critical fix 1. A properly positioned, compatible low-profile riving knife is designed to remain installed for most non-through cuts, including rabbets, grooves, and tenoning-jig work, whenever the saw manufacturer permits it — SawStop's own Jobsite Saw manual names non-through work as a primary reason to use the riving knife [SR30]. The reliable, physically-necessary removal case is a dado stack: because the riving knife/splitter is sized to a single blade's kerf, it does not fit inside the wider channel a dado stack cuts. Rockler's own published guidance calls dado work "the one instance where removing the guard is necessary" [SR26] — Rockler's own wording supports dado specifically, not the broader "rabbets, some grooving, tenoning-jig work" list the original entry had generalized it into. (This is the same design tension the Table Saw entry's own Petroski arc frames at a higher level — splitter → riving knife → flesh-detection, each fix leaving a residual gap the next fix addresses — but the mechanism is now stated correctly: dado incompatibility, not a blanket non-through-cut incompatibility.)
• a real, dated, government-sourced number on how often the fix is actually in place at the moment of injury — this entry's strongest single citation, now correctly scoped to its own denominator. The CPSC's own Survey of Injuries Involving Stationary Saws: Table and Bench Saws, 2007–2008 found a riving knife/spreader was attached to the saw in only 19.6% of studied blade-contact injury cases, and a separate anti-kickback pawl assembly in 23.9% of those cases; separately, "guard removed" was cited as the reason no blade guard was present in 75% of no-guard injury cases [SR28]. These figures describe injury cases, not the general saw-owning or saw-using population, and cannot on their own estimate device attachment rates, removal rates, or effectiveness among all users — stated explicitly per the reviewer's fix 5, not just implied. What the figures soundly support is narrower and still real: the device is frequently absent at the exact moment it would matter, among the population of people who were hurt.

✦ human side

• ★ two named, dated 2008 Sawmill Creek forum threads supply real, first-person trade texture on the historical practice of removing the guard assembly — the strongest human-side material in this entry. In "Difference between a riving knife and a splitter?" (started by keith micinski, Jan. 20, 2008), poster John Thompson ("Sarge") describes the pre-2008-mandate era bluntly: "In those days there were two predominant forms of guards on a table saw: the 'OSHA' guard, and The-OSHA-Guard-removed. The OSHA guard inevitably sucked." Poster Jeff Duncan, from the same era, states plainly: "Both my jobsite Makita and my previous BT3000 had the safety stuff tossed." Posters Steven Wilson and Joe Spear give the practical mechanical reason removal was so routine: an older, fixed splitter/pawl assembly required a bolt out for every dado or rabbet pass, and was widely described by tradespeople as "more nuisance than they're worth" — a real historical account of practice on the older, fixed assemblies this entry documents, distinct from the corrected guidance for modern low-profile riving knives above [SR8].
• a second thread ("Situations where a Riving Knife would cause problems?", started by Daniel Hillmer, Oct. 23, 2008) independently confirms dado work as the standard, near-universal trigger for removing the device — corroborated across five named posters (Rob Russell, Jason Beam, Jimmie Mayfield, glenn bradley, Robert Strebler). In the same thread, poster michael osadchuk volunteers, unprompted, that "SawStop is the safety feature that really works to virtually remove the hazard" — an early (2008), practitioner-sourced framing of flesh-detection technology as a genuinely different category of fix from a mechanical guard, independent of and preceding the corpus's own already-documented CPSC/SawStop regulatory history in the Table Saw entry [SR9].
• a named CPSC-record finding, not forum texture, but worth carrying here as institutional "human side": the agency's own 2023 rulemaking record states directly that "currently available safety devices, such as the modular blade guard and riving knife, do not adequately address the unreasonable risk of blade-contact injuries" — a federal regulator naming this exact device, by name, as an insufficient (not absent, insufficient) fix, the institutional mirror of the forum posters' own lived complaint that the older guard assembly is "more nuisance than they're worth" [SR13].

✦ proof of existence

• Patents: anti-kickback pawl patents US 1,074,198, US 2,095,330, US 3,232,326 [SR2]; MicroJig's own aftermarket splitter patents US 7,293,488 and US 7,784,385 [SR23]. (Corrected Run 567: US 482,507 (1892) is no longer listed as a proven splitter-adjacent patent in its own right — its full text is unread and it is cited only as unread prior art inside a later patent, see *lineage. US 5,063,806 is removed entirely — the reviewer read it directly and found it describes a radial-arm-saw carriage-speed-control mechanism unrelated to this device, not corroborating evidence of cross-use.)*
• Regulatory/standards record: UL 987 (sixth edition, January 2005, added the riving-knife requirement; industry adoption widespread by 2008; modular blade-guard requirements effective January 31, 2010; riving-knife requirement's own compliance effective date January 31, 2014 — corrected Run 567 to CPSC's own controlling record) [SR5][SR13a]; OSHA 29 CFR 1910.213 (current US workplace standard using "spreader" as the formal term, covering hand-fed circular ripsaws) [SR11][SR17]; CCOHS's own current fact sheet (last revised 2025-06-30) [SR11]; the EU's EN 1870-1 standard (later revision years 1999, 2007+A1:2009 confirmed; the original mandate date unverified this run) [SR4]; CPSC's own 2023 AIM-rulemaking record naming the device directly [SR13].
• Government injury data: CPSC's Survey of Injuries Involving Stationary Saws: Table and Bench Saws, 2007–2008, read within its own injury-case denominator (see history) [SR28]; a peer-reviewed Professional Safety article (Current, Main & Main, Nov. 2020, via CDC/NIOSH) putting 2016 US table-saw ER visits at 7,200 (±2,300), with 5,600 (±1,900) finger-contact — a separate, earlier-year data point from any CPSC figure already carried in the Table Saw entry, not to be merged with it [SR12].
• Current market/manufacturer record: SawStop, Delta, Grizzly, Bosch, MicroJig, and Shark Guard product/spec pages, all directly fetched this run with real current prices (see anatomy and variations) [SR18][SR20][SR21][SR22][SR23][SR24].

references

⚠ Correction note (absorbed Run 567, 2026-08-29 — read before the entry)

The reviewer reviewed this entry on 2026-08-10 (ticket CLAUDE-018) and returned REVISE with 8 required fixes, one graded safety-critical. The reviewer's own words: "The entry is not safe to merge unchanged. Its most consequential error says a riving knife cannot remain in place for most non-through cuts." This ticket sat acknowledged but unabsorbed for 19 days (Haft Runs 470–566 were committed to fresh corpus builds under the standing "every run builds a tool" discipline) before this run made the judgment call to spend itself on absorption instead of a new build — see the Run 567 Tool Log entry for the full reasoning. All 8 fixes are now applied in place below; nothing in the original research was discarded, only reframed, scoped, or corrected. A full before/after account lives in the companion report, Splitter - Riving Knife — applied.

The one fix worth stating plainly up front, because it is the one that matters if a reader only reads one paragraph: a modern low-profile riving knife is designed to stay installed for most non-through cuts — rabbets, grooves, tenoning-jig work — following the saw and cutter manufacturer's instructions. The reliable removal case is a dado stack, because a dado's cut width doesn't match the knife's kerf sizing. The original text told readers the opposite. That is now corrected throughout.

Note on how this tool was picked (read before the entry)

This run began by reading Haft Build Queue first, per standing discipline — the Priority Queue is empty (row 30 still blank) and the Parking Lot's only remaining item, Brad Nailer, is explicitly flagged as contested and Jonathan's call only, not to be built. Per the fallback rule, the next step is the lineage tree — but a check of all nine named lineage-tree files in (Abrading, Bench-Joinery Plane & Scraper, Saw, Striking, Clamping, Wrenches-Spanners, Lathe, Powered Mortising, Router-Shaper-Moulder) found every single one's own "Gaps" section states its named candidates are exhausted and that the next move is "a fresh corpus-wide reconnaissance pass or a genuinely new, un-mapped corpus family." The Pliers family (the newest sibling-gap chain, tracked in run-log text rather than a tree file) is also fully closed as of Run 397.

A dedicated reconnaissance pass (cross-referencing every wikilink in every built entry across all six corpus locations — root, Staging, and four mirror/review subfolders — against a master list of ~332 already-built tool names) found no tool clearing the corpus's usual 2+-substantive-citation bar. The strongest near-miss was the splitter/riving knife, cited once, substantively, by Table Saw. Rather than force a manufactured pick or spend the run on housekeeping alone, this entry proceeds with that single-citation candidate — flagged honestly here rather than overclaiming a stronger discovery process than actually happened — because it is a genuine, standalone, commercially sold, safety-critical tool (not a passing detail), used on the table saw, with its own real market, its own real misuse literature, and its own research turned out considerably richer than the thin corpus citation suggested.

The reviewer's corrections (8)

• HIGH — 1. Correct the non-through-cut guidance — Exact claim: “it physically cannot sit inside ... most non-through cuts (rabbets, some grooving, tenoning-jig work)” and “Remove it only for ... dado stacks, some rabbets/grooves, tenoning-jig work.”
• HIGH — 2. Remove the radial-arm-saw and miter-saw scope claims unless separately established — Exact claim: “A table-saw (and, less commonly, radial-arm-saw) safety accessory” and “radial-arm-saw and some circular-saw variants.”
• HIGH — 3. Reframe the injury mechanism — Exact claim: “the mechanical fix for the single most common serious table-saw injury mechanism, kickback.”
• MEDIUM-HIGH — 4. Repair the UL 987 timeline — Exact claim: “UL began requiring riving knives on all new table-saw designs in 2008, with all designs required to comply by 2014.”
• MEDIUM-HIGH — 5. Keep the CPSC percentages inside their actual denominator — Exact claim: The figures are repeatedly used to support “routine removal,” “how often the fix is actually in place,” and “the most concretely ... dangerous finding.”
• MEDIUM — 6. Correct the MicroJig classification and “universal riving knife” language — Exact claim: “retrofitting an aftermarket universal riving knife (MicroJig)” and “Aftermarket universal splitter (MicroJig MJ Splitter).”
• MEDIUM — 7. Remove or re-research the earliest-patent claim — Exact claim: “The earliest well-documented splitter-adjacent US patent found is US 482,507 (issued September 13, 1892).”
• MEDIUM — 8. Tighten terminology and certainty throughout — Exact claim: “fixed or arbor-mounted blade,” “US-mandated,” and repeated use of “prevents kickback.”