Keith Baar's tendon research — and what the videos actually prescribe
Dr. Keith Baar, professor at UC Davis and head of the Functional Molecular Biology Laboratory, argues that tendons don't heal with prolonged rest — they heal with short, low-load, long holds: isometric loading with zero movement, in doses of ten minutes or less. Below is his peer-reviewed work, notes taken from transcripts of his video appearances, and the two-session forearm routine those sources support for tennis-elbow-side pain — with the evidence limits and the places his sources disagree kept visible.
Subject: Keith Baar, Ph.D. — Dept. of Physiology & Membrane Biology, UC DavisFocus: wrist extensors / lateral epicondyleCompiled: October 24, 2026, from the sources listed at the foot of this page
Listen: the podcast
“Rebuilding Tendons, Without the Hype” — the podcast version of this research, generated from the script. Press play below, or download it to listen offline.
Every number below traces to a Baar paper or a transcript in the video notes. Confidence is moderate-to-high for the dose and mechanism; there is no Baar randomised trial for tennis elbow specifically — the elbow use is extrapolated from his engineered-ligament work, a patellar-tendon case, and his podcast prescriptions.
The injured-tendon dose
30 seconds on · ~2 minutes rest · × 4 holds · ≤10 minutes total
Ramp into tension over 3–5 seconds, hold, release over 3–5 seconds. Maximum two sessions per tendon per day, 6–8 hours apart. Pain ≤2 out of 10 during — Baar's cue is “ooh, not ouch.” It must not be worse the next morning; if it is, drop the effort rather than pushing through.
Parameter
Healthy tendon
Injured / chronic tendon
Hold length
10 seconds
30 seconds — the minimum Baar cites for longstanding injury
Rest between holds
50 seconds (10 on / 50 off)
~2 minutes
Holds per session
~10–12 in 10 minutes
4 in ~10 minutes (120 seconds of total tension)
Sessions per day
1–2
1–2, separated by 6–8 hours
Effort
Feel tension through the tendon; ~50% of max gave maximal adaptation in one experiment Baar describes
Tension, not pain — up to ~50% for longstanding injury, never shaking or crushing, never 80–90%
Type
Either
Overcoming first — push against something that cannot move. Yielding holds (holding a weight up) come later; picking the weight up creates jerk
Progression
—
Raise load or hold quality only when pain-free at the current dose and not worse next morning. Don't add sets or time by default
Sources: Tim Ferriss Show #797 transcript; Doctor Yaad Podcast transcript; Mark Bell's Power Project transcript (verified live, Oct 24, 2026); Baar, Sports Medicine 2017.
Hold timer
Runs the injured-tendon dose above: a 5-second ramp, a 30-second hold, then 2 minutes of rest — four rounds. Keep it beside you during either session.
Ready
0:30
Hold 1 of 4 · then 2:00 rest
Each hold starts with its own 5-second ramp — ease in, don't snap into tension. Pain stays ≤2/10 throughout.
Your two sessions — outside elbow
Baar is specific on this point: for tennis elbow you need both wrist extension and forearm rotation — “the extensors of the forearm… they're going to be rotational and they're going to be extensors, so you have to do both.” Loading only extension leaves the rotational fibres at the common extensor origin unloaded and still stress-shielded.
Read this first — how this split was built Synthesis
The morning-extension / later-rotation split is not a verbatim Baar prescription. In the Ferriss episode, Baar rotated positions within one 10-minute session. The split is used here because it respects his 6–8 hour refractory gap and gives each movement plane a full 4 × 30-second dose without breaching the 10-minute ceiling. One conflict is flagged in full below: in the Mark Bell episode Baar says the first session should be at night, at very low load, judged by the next morning. Apply that safeguard to your first-ever session and to any step-up, whichever time of day you train. If a clinician has prescribed a different structure, theirs takes precedence. This page is general information, not medical advice for a specific person.
Ground rules for both sessions: overcoming isometrics only to start — push against a door jamb, counter edge, table underside, or a racket wedged so it can't move. No joint movement. Start lighter than you think you need; in one irritable case Baar describes, the starting load was ~3 N before pain, progressing to 20–40 over weeks. Total time includes rest — no bonus holds for the same tendon in the same session.
Session A · Morning
Extension
Loads the common extensor origin through wrist extension — the plane stressed by gripping and backhand-type tasks.
Diagram only — not a photo. The hand pushes up into the ledge and does not move.
4 × 30 sec2 min rest~10 min total
Set up: elbow by your side, bent ~90°, forearm supported on a table, wrist just off the edge, palm down. Back of the hand against the underside of a fixed ledge. Wrist starts slightly cocked down — the bottom of a reverse wrist curl, the position Baar describes.
Push: over 3–5 seconds, gently try to lift the back of the hand into the fixed resistance. Nothing moves.
Hold 30 seconds at tension, pain ≤2/10. Release over 3–5 seconds.
Rest 2 minutes. Repeat to 4 holds.
If even this hurts >2/10: use the lightest tension you can feel, shorten toward 10 seconds and build up, or start from a less cocked-down wrist position. Don't switch to holding a weight to “get through it.”
Loads the rotational component of the extensors. Vocabulary from the demonstration video below: supination = palm up, pronation = palm down, elbow by your side.
Diagram only — not a photo. Grip a racket, stick or dowel; the head is fixed so rotation is an isometric effort, not a movement.
4 × 30 sec total2 min rest~10 min total
Set up: elbow by your side, bent 90°. Grip a tennis racket, stick, or dowel with the head wedged under a counter or against a wall so it cannot rotate.
Push: over 3–5 seconds, try to rotate the forearm into the fixed resistance — the external-rotation effort Baar set up for Ferriss, racket under a ledge, rotating into it without moving.
Hold 30 seconds, release over 3–5 seconds, rest 2 minutes.
Dose is 4 holds total for the session — for example 2 in each rotation direction, or 4 in the painful direction only. Do not double it to 4 + 4.
Alternative setup — People's Pharmacy transcript: instead of wedging the racket, hold it with its head in front of your body and use the other hand to rotate it back out, then back off until you feel tension that is “ooh instead of ouch” and hold 30 seconds. Same isometric idea, different fixed resistance.
Short on time? Baar's own Ferriss structure — one extension hold and one rotation hold at each of 2–3 elbow angles, rotating through in a single ~10-minute circuit — is the closer-to-source alternative. Trade-off: less dose per movement plane.
Progression — by elbow angle, one step at a time
Progress effort within a stage before progressing angle, and never progress angle, effort, and frequency together. Each step is earned by a next morning that is not worse.
1
Bent 90°, elbow by side. Least load in the lengthened position; mimics the backhand position.
2
Mid-range — elbow around 45°, forearm forward on the table.
3
Straight arm, forearm forward. A longer muscle-tendon position with different upper-arm contribution; Baar notes a long position “helps a little bit” for stress relaxation.
4
Overhead — only if relevant. This stage came from Ferriss's own case (lateral elbow pain plus lower-triceps pain from climbing). It is Ferriss-specific, not a universal step — skip it if overhead work isn't part of your picture.
Avoid or defer
Dynamic wrist curls for reps. The held bottom position is Baar's; the reps are not. Dynamic means velocity and jerk, and load bypasses the weak part of the tendon. Defer until isometric holds are pain-free and strong — Baar's general order is isometrics → slow heavy strength → dynamic.
Heavy or jerky gripping, max-effort grip trainers. Gripping loads the extensor origin, and the backhand-style jerk is Baar's proposed injury mechanism. Light daily grip for normal function isn't banned by any source — but if it provokes >2/10 or next-morning worsening, it counts as overload.
Aggressive end-range stretching. The overpressure stretch in the demonstration video is a position reference only. Baar argues passive stretching doesn't produce the same tendon adaptation as isometrics.
Momentum in any lift — bouncing out of the bottom of a rep is the jerk principle applied.
Sudden volume jumps over ~30% in a provoking exercise (Built With Science / Baar load-management rule: going from 4 to 10 sets a week is too much; add 1–2).
Stopping when the pain eases. Pain relief in 1–2 weeks is commonly reported, but the tissue course runs 4–8 weeks minimum in the Built With Science plan. Continue at maintenance frequency rather than stopping at the first pain-free day.
Between sessions
Keep the 6–8 hour gap genuinely low-load for that elbow: no extra isometric “top-ups,” no provoking sport or lifting, no testing your max grip.
Use the arm normally for light daily tasks — complete avoidance all day is immobilisation by another name, which is the thing Baar's work argues against.
If a task provokes sharp pain, modify it — bent elbow, palm-up carry, two hands, lighter load — rather than bracing through it or shelving the arm entirely. Note: this modification advice is standard tennis-elbow load-management guidance, not a Baar quote.
Training the other arm and general activity can continue. Baar's teams kept athletes playing while loading isometrics, because the isometric is the only time the shielded area feels load.
On rest and RICE: Baar criticises prolonged immobilisation because unloading removes the signal that orients collagen — but he is not saying “never use a boot.” For a post-surgical Achilles he describes removing the boot morning and evening for zero-pain, low-jerk isometrics, then replacing it. Don't discard a clinician-prescribed brace on the strength of a podcast.
The research behind it
Peer-reviewed work first, with an honest read on what each piece actually proves. The pattern to notice: the mechanism is well supported in the lab; the human elbow evidence is extrapolation.
Baar K. “Minimizing Injury and Maximizing Return to Play: Lessons from Engineered Ligaments.” Sports Medicine 2017; 47 (Suppl 1): S5–S11. DOI 10.1007/s40279-017-0719-x
The foundation paper. Baar's lab grew human ACL/hamstring cells in fibrin gels between anchors — engineered ligaments — and mechanically tested them. The sinews “quickly become refractory”: signalling switches off after short bouts, which is where the <10 minutes of activity with ~6 hours of rest rule comes from. The review also reports that estrogen decreases sinew function and that a factor released after intense exercise increases collagen synthesis.
Evidence: review of in-vitro work. Strong for the model and the dose logic — not a human tendon trial. PubMed
Paxton JZ, Grover LM, Baar K. Tissue Engineering Part A 2010; 16(11): 3515–3525 · and Paxton et al. 2012; 18(3–4): 277–284
The engineered-ligament model itself (fibrin + fibroblasts, brushite anchors), and the follow-up finding that signalling becomes refractory to stretch for ~6 hours and that intermittent loading was twice as effective as continuous loading at increasing collagen synthesis.
Evidence: primary in-vitro work. The 2012 results are cited here second-hand via Baar's slide deck; full text was not re-read for these notes.
Shaw G, Lee-Barthel A, Ross ML, Wang B, Baar K. “Vitamin C–enriched gelatin supplementation before intermittent activity augments collagen synthesis.” American Journal of Clinical Nutrition 2017; 105(1): 136–143. DOI 10.3945/ajcn.116.138594
Eight healthy young men, crossover design: 5 g vs 15 g gelatin + 48 mg vitamin C, one hour before 6 minutes of rope-skipping, three times a day, 6 hours apart. The collagen building blocks in blood peaked at ~1 hour; engineered ligaments grown in serum from the 15 g group contained more collagen and had better mechanics, and the blood collagen marker PINP roughly doubled at 15 g vs 0/5 g.
Evidence: small, mechanistic, healthy men, surrogate markers. A published editorial critique notes PINP reflects bone turnover more than tendon synthesis, and clinical benefit was unproven — so “doubles tendon healing” is not a claim these notes make.
Baar K. “Stress Relaxation and Targeted Nutrition to Treat Patellar Tendinopathy.” International Journal of Sport Nutrition and Exercise Metabolism 2019; 29(4): 453–457. DOI 10.1123/ijsnem.2018-0231
A single case: a professional basketball player with MRI-confirmed proximal patellar tendinopathy, treated with gelatin, heavy isometric stress-relaxation work, and whole-body training while continuing to compete. A systematic-review summary gives the dose as 15 g gelatin + 225 mg vitamin C twice weekly for 18 months, one hour before sessions. At 1.5 years an independent orthopaedic surgeon declared the tendon normal on MRI, with pain down and performance up.
Evidence: n=1 case report — illustrative, not generalisable proof, and a knee tendon, not an elbow. PubMed
Tam KT, Baar K. “Using load to improve tendon/ligament tissue engineering and develop novel treatments for tendinopathy.” Matrix Biology 2025; 135: 39–54
The formal stress-shielding model: compliant scar tissue sits in parallel with stiffer healthy tendon, so at any given strain the scar takes less stress, its cells receive less signal, and it stays immature. Sustained loading that allows viscoelastic creep — isometrics — is proposed as the way to transmit tension into the scar.
Evidence: mechanistic framework/review. As the Physiotutors summary cautions, direct human proof that a specific protocol “overcomes” shielding in the body is not established in this paper.
Yamamoto N et al. Journal of Biomechanical Engineering 1993; 115(1): 23–28 (Hayashi group; cited by Baar)
The starkest demonstration of what unloading does: a rabbit patellar tendon completely stress-shielded by a patella–tibia wire fell to 9% of control tensile strength after 3 weeks, while its cross-sectional area grew 131% — bigger, but far weaker, with fewer aligned collagen bundles.
Evidence: animal work, pre-Baar. Supports “unloading weakens tendon”; not a human immobilisation trial. PubMed
The mechanisms, in Baar's terms
Stress shielding. A tendon isn't uniform: the stiff, strong parts take the load and shield the weak, scarred parts — and scar forms precisely because the injured area never feels load, so collagen is laid down in random directions. Baar's demonstration: tug dynamically on torn paper and it doesn't tear at the nick, because no load reaches it; pull slowly and sustained, and the weak spot finally loads.
Stress relaxation / creep. Hold a tendon at a fixed length and the force needed falls off quickly — most of the relaxation is done by ~30 seconds, with diminishing returns after. The exact percentage varies by source and preparation (about 85% of the way to its floor by 30 s in the Ferriss discussion; ~45% of tension released in one secondary summary; ~60% load drop in another) — don't quote a single fixed number. The 30-second dose is what's consistent.
Refractory period. After roughly 5–10 minutes of loading, tendon signalling goes quiet for ~6 hours. More volume in one session is wear, not signal — hence two sessions a day maximum, 6–8 hours apart.
Jerk. The rate of change of acceleration. Tennis elbow, in Baar's framing, is a backhand jerk — racket one way, ball the other. Isometrics have zero velocity, hence zero jerk: the core reason he prefers them over fast dynamic work.
Why eccentrics “worked.” Baar's claim is that slow eccentric programmes helped because they were slow, not because they were eccentric. Velocity is the variable; zero velocity is the extreme end of it.
Early loading. In his thigh-injury lecture, Baar cites a Kjær-group study in which loading started 2 days after injury versus 9 days: about 60 days versus about 90 days to 50% of people returning to training — early loading better. Flagged: Baar also says “25%” in the same passage, which does not match the 60-versus-90-day figures as spoken, so this page quotes the day figures, not a clean 25%. The primary paper itself was not re-read for these notes. His related figures — mouse tendon losing 15–20% of collagen within 3 days in a cast, a tendon ~30% weaker — come from the podcast discussions, at anecdote level for humans.
Video notes — what was actually read
Six of the seven videos were covered from full transcripts. One — Ask Doctor Jo — is covered from its verified description and page only — that distinction is marked on it, and no timestamped quote is claimed for it.
Tim Ferriss Show #797 — Dr. Keith Baar: Simple Exercises That Can Repair Tendons Transcript read
Key timestamps: minimal dose 00:12:06 · refractory period 00:13:50 · tennis-elbow protocols 00:18:49 · isometrics vs eccentrics 00:20:28 · stress shielding 00:22:14 · hold duration 00:29:58 · overcoming vs yielding 00:36:02 · Ferriss's own prescription 00:47:11 · multiple positions 01:04:02 · collagen & vitamin C 01:07:26
The source for most of this page: pain ≤2/10, 3-second ramp, 30 s × 4 with 2-minute rests for an injured tendon; 10-second holds with 50-second rests for a healthy one.
Extensors need extension and rotation; rotate through elbow positions — bent 90°, straight, overhead — within one 10-minute window.
Overcoming isometrics first, because force builds gradually and jerk stays near zero. A cast-iron pan held up (yielding) may simply be too much.
Caveat: the multi-position routine was tailored to Ferriss personally — lateral elbow pain plus lower-triceps pain from climbing. It is not a generic tennis-elbow prescription.
Why Isometrics Are the Missing Link — Mark Bell's Power Project Transcript verified live
Short (<1 second) maximal isometrics are for performance, not tendon health; long holds are the tendon tool.
Dose restated: 4 × 30 s or 12 × 10 s — 2 minutes of active loading across an 8-minute window, then 6 hours of rest; twice daily possible.
The rule found only here: the first session is always at night, at very low load — and you judge it by the next morning before progressing. If the morning is worse, you don't progress. This directly qualifies any morning-first plan, including the one on this page.
Collagen: any gelatin or hydrolysed collagen works; Baar avoids bone sources (heavy metals are stored in bone) and mentions fish skin as an alternative.
It's Weird, But It Re-Builds Your Tendons In 30 Days — Built With Science Transcript verified live
Chapters: what a tendon is 00:40 · why tendons adapt slower 01:45 · mistake #1 — volume too fast 03:00 · mistake #2 — momentum 04:20 · why rest doesn't solve it 05:40 · the isometric solution 06:50 · recap 08:45
Baar, quoted: isometrics “maximize creep,” and “by 2 weeks, almost everybody is like ‘I can do pretty much everything again’” — an anecdote, not a trial result.
The presenter's protocol (not Baar's words): hold the painful motion ≥30 seconds × 3 sets, twice daily for 1–2 weeks, then 3–4×/week for 4–8 weeks, then 5–10 minutes of maintenance after workouts.
The presenter's pain ceiling is 3–4/10 — looser than Baar's ≤2/10 (see Conflicts). This page uses Baar's.
“30 days” is the video's title claim, not a Baar endpoint — and the video contains no wrist-extension/rotation tennis-elbow demonstration, so it isn't cited for elbow positions.
Sprains, Strains and Tears: The New Science of Healing Faster — People's Pharmacy Transcript read
Frames RICE as alleviating pain short-term but possibly slowing healing, and names Baar's approach 3LD: low-load, long-duration isometrics — support the joint so it isn't taking full bodyweight, find the tension that is “ooh instead of ouch,” hold ~30 seconds.
Tennis-elbow demo, now from the transcript: hold the racket with its head in front of the body, use the other hand to rotate it back out, then back off to tension that is “ooh instead of ouch,” and hold ~30 seconds. A stress-ball squeeze with the wrist slightly flexed, held 30 seconds, is given for pumping fluid out.
Jerk framing is explicit here: tennis elbow from the racket going one way and the ball the other, giving jerk through the extensor tendons. The longer holds are twice a day, feeling that “ooh” through the injured tissue.
Tennis Elbow Stretches & Exercises — Ask Doctor Jo Positions only · Not Baar
Included purely as a position glossary: fist palm-down with wrist up (extension) and down (flexion); thumb-to-ceiling deviation; elbow by side, palm-up supination to palm-down pronation; a top-heavy hammer for the rotation positions.
Doctor Jo's actual routine — dynamic sets of ~10 reps and an overpressure stretch held 30 s × 3 — is the opposite of Baar's held, low-jerk dose. This page borrows her positions and converts them into Baar-style holds; it does not borrow her programme.
Training to Optimize Muscle and Tendon Structure After Thigh Muscle Injury Transcript read
A lecture on load management in thigh-muscle injury rehab. It contains no elbow content — but its dose and mechanism findings are now quoted from the transcript, not inferred: 4 × 30 seconds with 2 minutes of rest beat dynamic loading at the same time under tension for tendon markers, and intermittent loading (10 minutes, 6 hours of rest, repeated) produced about twice the collagen of continuous loading.
Refractory period stated precisely here: after a 10-minute stretch, a second bout 2 hours later gave 40% less signal; it took a full 6–8 hours to get the full signal back — hence two sessions a day maximum.
Early loading, as in Mechanisms: loading at 2 days versus 9 days after injury, about 60 versus about 90 days to 50% returning to training. A boot is described here as a mechanical stress shielder — the same shielding argument as for scar tissue.
These disagreements are in the material itself. They're kept separate rather than averaged away, because which version you follow changes what you do.
Point
The versions
What this page does
Pain ceiling
Baar: ≤2/10, “ooh not ouch.” Built With Science presenter: 3–4/10.
Uses Baar's ≤2/10 for an irritable elbow; the 3–4 figure is the presenter's, not Baar's.
First session timing
Ferriss / Built With Science: morning + evening. Mark Bell: first session always at night, very low load, judged next morning.
Keeps the morning/later split but applies the Mark Bell safeguard: first session and every step-up are low-load and judged by the next morning.
Sets
Baar's injured protocol: 4 × 30 s. Built With Science: 3 sets.
Uses Baar's 4.
Session structure
Baar (Ferriss) rotated positions within one 10-minute session. This page splits extension and rotation across the day.
Split is labelled a synthesis; the Ferriss circuit is offered as the closer-to-source alternative.
Stress-relaxation %
~85% by 30 s (Ferriss) vs ~45% released (secondary summary) vs ~60% drop (another transcript).
Quotes none as fixed; the 30-second dose is the stable claim.
Collagen dose
15 g + 48 mg vitamin C (2017 study) vs 15 g + 225 mg (2019 case) vs 15 g + 200–250 mg (Ferriss) vs 5–10 g + 50–250 mg (Doctor Yaad).
Lists all four rather than picking silently — timing (30–60 min before) is the consistent part.
“30 days”
Built With Science video title. Baar's own timeframes: pain relief often 1–2 weeks, 4–8 weeks of continuation, 18 months in the patellar case.
No 30-day elbow cure is claimed anywhere on this page.
The collagen + vitamin C option
An adjunct, not the treatment. Baar's own estimate in the Ferriss episode: the loading is ~95% or more of the effect; collagen and whey together might be worth 2–5%.
If you use it: hydrolysed collagen or gelatin from skin sources (bovine hide or fish skin — Baar avoids bone broth because heavy metals are stored in bone), taken together with vitamin C, 30–60 minutes before a session, so the amino acids are circulating while loading pumps fluid through the tendon. In the Doctor Yaad episode he adds whey at ~0.25 g/kg afterwards as a second stimulus.
Doses found — pick one, don't stack them: 15 g gelatin + 48 mg vitamin C (Shaw 2017, the only dose tied to a trial — small, with surrogate markers) · 15 g + 225 mg (Baar 2019 case) · 15 g + 200–250 mg (Ferriss discussion) · 5–10 g collagen + 50–250 mg (Doctor Yaad).
Consistency over precision: in the Doctor Yaad episode, daily consistency matters more than exact timing, and post-session intake still helps muscle connective tissue even if it's less targeted at tendon. If the pre-session timing is impractical, taking it with the later session beats skipping it — just don't expect it to substitute for the holds.
When to get checked in person
These come from general clinical sources (Swansea Bay NHS physiotherapy; Tidewater Orthopaedics), not from Baar — lateral elbow pain has look-alikes that a video protocol can't distinguish.
Seek an in-person evaluation promptly if you have:
Numbness, pins-and-needles, or weakness in the hand, fingers, or wrist; pain radiating to the shoulder or neck; or symptoms in both arms — possible nerve involvement rather than simple tennis elbow.
A specific trauma — fall, direct blow, sudden pop — or major swelling, deformity, inability to move the elbow, wrist, or fingers normally, or severe pain after an injury.
Pain that keeps worsening despite 6–8 weeks of consistent, correctly dosed isometrics; night pain that doesn't settle; or a hot, red, swollen joint or fever.
A history that doesn't fit overuse — pain at rest with no provoking activity, unexplained weight loss. Mention any relevant drug history (including fluoroquinolone antibiotics, which Baar discusses as a tendon risk) to the clinician.
Any real doubt about the diagnosis: osteoarthritis, loose bodies, inside-elbow (golfer's) pain, and neck referral can all mimic this.
This page summarises published research and podcast discussion for general information. It is not medical advice, and it doesn't diagnose your elbow. Baar's elbow protocol itself has not been tested in a randomised trial for lateral epicondylitis — that gap is stated plainly in the research section above, and it's a fair thing to raise with any clinician you see.