FFMI Limits and Natty Boundaries: What the Data Actually Shows

Every gym has the same conversation: someone posts a physique, comments decide whether they’re natural, the shorthand verdict is FFMI 25. The number is real, the source is a real paper, and the cap is approximately right at the population level. It is also misused as a courtroom exhibit most of the time it appears online. What the data actually says, where the number bends, and which signals carry more weight than a single FFMI snapshot.

What FFMI actually measures

FFMI is lean body mass divided by height in metres squared:

FFMI = LBM_kg / (height_m)²

LBM = bodyweight × (1 − body_fat_fraction)

Height-normalised (corrects for the height-² scaling bias):
FFMI_norm = FFMI + 6.1 × (1.80 − height_m)

Raw lean mass isn’t comparable across heights. An 80 kg lifter at 168 cm carries lean tissue very differently than an 80 kg lifter at 188 cm. Dividing by height² adjusts for the dominant scaling effect.

Kouri’s normalisation term corrects to a 1.80 m reference height because lean mass scales slightly steeper than height². The exponent is closer to 2.4 in elite strength athletes^[6]; raw FFMI under-reads on tall lifters and over-reads on short ones. A 168 cm lifter at 78 kg LBM raw-FFMIs at 27.6; normalised, 26.4.

The FFMI Calculator outputs both. Use normalised for cross-lifter comparison. Raw is only valid against your own past data at the same height.

The Kouri 1995 paper

The anchor citation for the FFMI-25 framing is Kouri, Pope, Katz, and Oliva, Fat-free mass index in users and nonusers of anabolic-androgenic steroids, Clinical Journal of Sport Medicine, 1995^[1]. Key features of the methodology:

• Sample size. 157 male subjects, roughly half known steroid users. Competitive or recreational bodybuilders from northeastern US gyms.
• Body-fat method. Skinfold calipers (Jackson-Pollock 7-site), Siri equation. Siri’s hydration assumptions break at very low body fat, which matters for the high-end numbers.
• Steroid-use classification. Self-reported, no urinalysis, no longitudinal verification. The non-user group is the floor of the methodology, not the ceiling.
• Headline result. Non-users clustered at FFMI 21–23, upper tail to 25. Users clustered at 25–27 with a tail above 30.
• What the paper claimed. Kouri proposed FFMI 25 as the approximate upper bound observed among non-users in this sample. "Suggests an upper limit", not "diagnostic of steroid use." The hardening into a verdict happened in citing literature, not in the original paper.

The cohort was self-selected and competitive; a randomly drawn population would have a lower ceiling. Male-only sample; female natural FFMI distributions cluster 2–3 points lower with a softer ceiling around 22.

Confounder 1: lean-mass overestimation in glycogen-loaded lifters

Lean body mass on any method (DEXA, BIA, Bod Pod, hydrostatic, skinfolds) includes everything that isn’t fat. Water, glycogen, and the water glycogen binds. Every gram of muscle glycogen binds roughly 3 grams of water^[4].

A trained lifter with depleted glycogen carries 200–300 g of muscle glycogen plus bound water. After two days of high-carb eating, the same lifter sits at 500–600 g glycogen with another 1.5–1.8 kg of bound water. Up to 2 kg of measured "lean mass" that didn’t exist 48 hours earlier.

Lifter, 180 cm, 85 kg, 12% body fat
LBM (depleted)        =  85 × 0.88 = 74.8 kg → FFMI 23.1
LBM (carb-loaded)     =  87 × 0.86 = 74.8 kg actual muscle
                      + 2.0 kg glycogen+water
                      = 76.8 kg measured        → FFMI 23.7
Same person, same year, same training. Different scan day.

Add creatine loading (another 1–2 kg of intracellular water), a high-sodium pre-scan meal, and post-training engorgement, and an honest natty reads 0.8–1.2 FFMI points higher on a measurement day than on a baseline day. A real FFMI 23.5 lifter can land at 24.5–24.8 on a contest-prep snapshot without any change in actual contractile tissue. The math is correct; the interpretation is wrong.

Confounder 2: extreme height and the height-² breakdown

FFMI assumes lean mass scales with height². In real strength-sport populations the exponent is closer to 2.4–2.5^[6]. The discrepancy is small around the 1.80 m reference and compounds at the extremes.

Two natural lifters, both at the genetic 99th percentile:

Lifter A: 168 cm, 78 kg, 8% body fat
  LBM = 71.8 kg
  Raw FFMI       = 71.8 / 1.68² = 25.4
  Normalised     = 25.4 + 6.1 × (1.80 − 1.68) = 26.1

Lifter B: 198 cm, 110 kg, 10% body fat
  LBM = 99.0 kg
  Raw FFMI       = 99.0 / 1.98² = 25.2
  Normalised     = 25.2 + 6.1 × (1.80 − 1.98) = 24.1

Same biological percentile. Different normalised number.

Lifter A reads as a clear ceiling case; Lifter B reads as merely strong. Most of the difference is geometry, not biology. Above 195 cm the normalisation under-corrects, which is why elite tall strongmen and basketball-frame lifters can carry 100+ kg of lean mass and still index at FFMI 24–25 while looking far larger than a 175 cm bodybuilder at the same number.

The implication runs both ways. Tall lifters get unfair benefit-of-the-doubt; their FFMI under-reads. Short lifters get unfair scrutiny; theirs over-reads. A 168 cm lifter at FFMI 25 carries the same absolute lean mass as a 198 cm lifter at FFMI 23.5. Calling either suspicious on the FFMI number alone fails on geometry.

Confounder 3: contest-day vs casual measurement

The Kouri sample was almost entirely competitive bodybuilders measured close to contest condition. Body-fat estimates at contest leanness (5–7% men, 11–13% women) trip the Siri equation’s hydration assumption: body fat reads low, which inflates measured LBM.

Practical effect: a lifter actually carrying 7% body fat may DEXA at 5%. The 2-point shift translates into 1.5–2 kg additional measured LBM at typical bodyweights, another 0.5–0.7 FFMI points of pure bias. The Kouri non-user upper tail sits on top of this systematic over-read.

For an honest natty’s observed FFMI:

True body composition: 80 kg, 9% body fat, 1.78 m
  True LBM    = 72.8 kg
  True FFMI   = 23.0

At contest, glycogen-loaded, Siri bias:
  Measured BF = 6%
  Measured BW = 81.5 kg (water+glycogen retention)
  Measured LBM = 76.6 kg
  Measured FFMI = 24.2

Same person. +1.2 FFMI from measurement-day artefacts.

The on-stage FFMI of a lifter in May is not the same number as their November off-season FFMI. The former sits closer to the Kouri sample condition; the latter to year-round biology. Treat the on-stage number as inflated by 0.8–1.5 points relative to walking-around physiology.

Confounder 4: ex-natty and retained mass

Lifters who used PEDs in the past, dropped them, and now train drug-free. Myonuclear addition during anabolic exposure persists for years; the fibres retain elevated capacity to rebuild mass and can hold meaningfully more lean mass than a never-used baseline^[5]. A former user can sit 1–2 FFMI points above their never-used ceiling indefinitely, passing a drug test today.

No field-accessible way to detect this from a photo. The signature is biographical: a long-term lifter who plateaued at FFMI 23 for years, jumped to 26 in eighteen months, then settled at 25 indefinitely is harder to explain as "I finally got serious." Trajectory is the tell, not the snapshot.

The realistic natural-lifter trajectory

Published cohort data^[2][3] plus field experience produces a typical progression curve. Numbers below assume a male lifter starting near the population median, training consistently, eating adequate protein, sleeping seven-plus hours.

Stage           Years training   FFMI range    Annual ΔFFMI
─────────────────────────────────────────────────────────────
Novice          0–1              18 – 21       +2.0 to +2.5
Intermediate    1–3              21 – 23       +0.5 to +1.0
Advanced        3–6              23 – 25       +0.2 to +0.5
Elite-genetic   6+               25 – 26       under 0.2

Plateau distribution among long-term naturals:
  ~75% plateau at FFMI 22 – 24 (median lifter at advanced stage)
  ~20% plateau at FFMI 24 – 25 (favourable genetics, full effort)
  ~5% plateau above 25         (rare; documented but uncommon)

The elite-genetic tail above 25 exists. It’s rare enough that a stranger claiming to sit there year-round is statistically more likely wrong about their measurement, exaggerating, or reporting contest-day than a true outlier. The ceiling is soft, and most people above it are not naturals.

The Muscle Gain Potential Calculator models this decay curve with a configurable plateau, useful for setting year-3 and year-5 targets.

Signals that should outweigh FFMI

FFMI compresses everything to one scalar. Better discriminators are qualitative, rarely a single number, and much harder to fake.

• Arm-to-shoulder ratio and trap shelf. Anabolic exposure preferentially grows tissue with high androgen-receptor density: traps, deltoids, upper chest. Naturals show more even distribution. A 47 cm flexed arm on 130 cm shoulders is plausible naturally; on 110 cm shoulders, less so.
• Vascularity at non-contest leanness. Persistent forearm and shoulder vascularity at 12–14% body fat is unusual for naturals. Most naturals don’t show heavy vascularity until under 9–10%. Off-season veins down the deltoids is a stronger signal than the on-stage version.
• Recovery rate and frequency tolerance. Six hard sessions a week, each muscle three times, recovering fully, progressing on every lift is something most naturals cannot do. The training log catches this; the FFMI number doesn’t.
• Lift progression curves. A natural shows rapid year-one progress, slow year-two-and-three, near-stall after year five. A jump after year three from a long plateau is unusual without a methodology change. Sustained top-end growth past year five at the elite tail is particularly hard to explain.
• Skin-fascia tightness and the "3D" look. Muscle bellies unusually full and rounded relative to limb circumference. Partly water retention, partly genuine hyperplasia under cycle conditions. Hard to quantify, easy to recognise after seeing enough physiques.

None singly is conclusive. Together, they’re a much better discriminator than FFMI alone.

A worked FFMI computation with confidence intervals

Take a real measurement and propagate the error. Lifter at 180 cm, 84 kg on the morning scale, body fat estimated at 13% by hand-to-foot BIA.

Inputs
  Bodyweight     84.0 kg ± 0.3 kg (daily fluctuation)
  Height         1.80 m  ± 0.005 m
  Body fat %     13%     ± 5% (consumer BIA)

Point estimate
  LBM   = 84 × (1 − 0.13)        = 73.1 kg
  FFMI  = 73.1 / 1.80²           = 22.6

Lower bound (BF = 18%, BW = 83.7)
  LBM   = 83.7 × 0.82            = 68.6 kg
  FFMI  = 68.6 / 1.80²           = 21.2

Upper bound (BF = 8%, BW = 84.3)
  LBM   = 84.3 × 0.92            = 77.6 kg
  FFMI  = 77.6 / 1.80²           = 23.9

Reported: FFMI 22.6 [21.2 – 23.9]

A 2.7-point window from a single measurement. Skinfold calipers with a trained tester (body-fat error 3–4%) narrow the window to ±0.6 FFMI. DEXA at standardised hydration narrows it to ±0.3. Any honest comparison needs error bars.

The Body Fat Percentage Calculator uses the U.S. Navy tape-measure formula (±3–4% error in trained populations). The Lean Body Mass Calculator takes that estimate and returns LBM, which the FFMI Calculator converts. Each step compounds error.

Why "natty or not" is usually the wrong question

The diagnostic framing assumes the answer matters in isolation. Usually it doesn’t. Better questions for an individual lifter:

• Whether your progression rate matches the natural decay curve. If yes, you’re progressing well and absolute FFMI is secondary. If no, either training is off (programming), nutrition is off (feeding), or you’re below your individual potential (patience).
• Whether your proportions match how you actually train. If you train chest twice a week, back once, and back dominates the silhouette, hypertrophy responsiveness is doing something interesting. Muscle-group balance answers this; FFMI doesn’t.
• Which plateau your genetic draw is heading toward. The honest target is your own asymptote. A lifter heading to FFMI 22.5 should aim for clean technical execution, recovery quality, and protein adherence rather than chasing the upper tail of a distribution.

For a stranger’s physique, "natty or not" usually doesn’t change what you’ll do. FFMI does not survive the leap from population description to individual diagnosis. Treat it as a soft descriptor of where someone sits in a distribution, not as evidence in a case.

Summary

• FFMI 25 is a soft statistical bound on the upper tail of natural male bodybuilders, derived from Kouri 1995 (n=157)^[1]. The 2020 reappraisal documented drug-tested naturals above 25^[2].
• Glycogen, water, contest-day measurement bias, and post-PED retained mass each add 0.5–1.5 FFMI points of artefact above true year-round physiology.
• Height extremes break the height-² scaling assumption. Use normalised FFMI for any cross-lifter comparison; tall lifters under-read and short lifters over-read.
• Trajectory beats snapshot. A long, flattening progression curve is much harder to fake than a single high reading.
• Use FFMI to track your own lean mass over years and to set realistic targets. Don’t use it to convict strangers on a number alone.

Tools: FFMI Calculator, Body Fat Percentage Calculator, Lean Body Mass Calculator, Muscle Gain Potential Calculator.