FFMI 25 Natural Limit: The Kouri 1995 Study, Reappraised

Fat-Free Mass Index — lean body mass divided by height in metres squared — is the most-cited metric in discussions of natural muscular potential. Compute your own raw and height-normalised value with the FFMI Calculator. The usual quote is “FFMI 25 is the natural ceiling”, based on a 1995 paper by Elena Kouri, Harrison Pope and colleagues at McLean Hospital. This article walks through what the study actually showed, what subsequent research added, and where FFMI fails as a discriminator. FFMI is the muscularity lens in the full framework; for how it reads against BMI and body fat on one lifter, see BMI vs FFMI vs body fat.

What Kouri actually measured

The 1995 study^[1] calculated FFMI in 157 male athletes: 83 users of anabolic-androgenic steroids and 74 non-users, classified by their own report. Key findings:

• Non-users' height-normalised FFMI extended up to a well-defined limit of 25.0.
• Many users easily exceeded 25.0, and some exceeded 30.
• Twenty Mr. America winners from 1939–1959, before steroids were in use, had an estimated mean FFMI of 25.4: the best physiques of the era sat at or slightly above the line.
• The authors offered FFMI as a useful initial screen for possible steroid use and called the findings preliminary.

The FFMI 25 figure has since been hardened into “if you're above 25 you must be enhanced”. That is a stronger claim than a preliminary screen from one self-reported sample supports, and later cohorts with higher upper percentiles do not support it either.

The FFMI formula

FFMI = LBM_kg / (height_m)²

Normalised FFMI (height-adjusted to 1.80 m):
FFMI_norm = FFMI + 6.3 × (1.80 − height_m)

Example:
  82 kg bodyweight, 180 cm, 15% body fat
  LBM = 82 × 0.85 = 69.7 kg
  FFMI = 69.7 / 1.80² = 21.5 kg/m²
  (Normalised = same, since height is 1.80 m)

The height-normalisation term was added by Kouri to account for lean mass scaling more steeply than height². Without it, very tall lifters look artificially high and very short lifters artificially low: at 1.95 m the correction subtracts 0.95, at 1.70 m it adds 0.63. Many calculators use 6.1 instead of the paper's 6.3, which changes the result by under 0.1. The FFMI Calculator provides both raw and normalised FFMI so you can see what difference height correction makes.

Updates to the 1995 picture

Collegiate-athlete normative data

A 2019 study^[3] measured 209 male collegiate athletes across 10 sports by DXA, height-adjusted their FFMI by linear regression, and reported the 97.5th percentile as an upper limit. The cohort mean was 22.8 ± 2.8; football athletes averaged 24.3 and water polo athletes 20.7. The 97.5th-percentile upper limit was 28.3 across all athletes, 29.1 in rugby and 25.5 in baseball. These athletes were not screened for drug use, so the figures bound the metric rather than certifying the individuals — but they show the FFMI distribution in trained men extends well past 25.

Height and genetics

Because lean mass rises faster than height squared, raw FFMI drifts upward with height: a tall lifter scores higher than a shorter lifter of the same build. That is why Kouri subtracts for heights above 1.80 m, and why the collegiate study fitted its own height adjustment by regression and found adjusted and unadjusted means significantly different.^[1][3] A linear correction is an approximation; compare yourself with lifters near your own height where you can.

Skeletal frame

Wrist and ankle circumference, clavicle width, and pelvic width all influence how much lean mass an individual can support at a given height. Two naturals at 180 cm with identical FFMI 23 can look strikingly different depending on frame. FFMI captures none of this.

What FFMI tells you — and what it doesn't

FFMI is useful for:

• Tracking your own lean-mass progress over years, with height held constant.
• Positioning yourself within the lifter population (approximate percentiles).
• Sanity-checking unrealistic bodyweight goals — targeting FFMI 27 as a natural is probably not a realistic training outcome.

FFMI is not useful for:

• Diagnosing whether another individual is natural or enhanced. A single data point isn't enough.
• Comparing across heights without normalisation.
• Measuring actual muscle gain — FFMI rises with both muscle and non-muscle lean tissue (glycogen, water, organ mass). A high-carb refeed that adds 1–2 kg of glycogen and water raises FFMI by about 0.3–0.6 at 1.80 m without any real muscle gain.

The "is he natural?" question is where the 25 ceiling gets misused most.

Population percentiles

What measured cohorts show, by study (FFMI in kg/m²):

Cohort                                         Mean          Upper value
─────────────────────────────────────────────────────────────────────────
Kouri 1995: male non-users (n=74)              —             up to 25.0
Kouri 1995: Mr. America winners 1939–59        25.4          —
Collegiate men, 10 sports (n=209, adjusted)    22.8 ± 2.8    97.5th pct 28.3
Collegiate men, 10 sports (n=596)              21.5 ± 1.9    —
Collegiate women, 8 sports (n=1,365)           17.9 ± 1.8    —
Collegiate women (n=372)                       18.8 ± 2.1    97.5th pct 23.9

Sources: Kouri 1995,^[1] Currier 2019,^[3] Magee 2024,^[2] Harty 2019.^[6] Means differ by sport: in Currier's men from 20.7 (water polo) to 24.3 (football), in Magee's from 19.9 (volleyball) to 25.7 (throwers). The collegiate studies did not report drug testing.

FFMI vs other natural-ceiling estimators

Several alternative frameworks for natural potential exist:

• Ankle-wrist-height models (Martin Berkhan, Casey Butt). Predict maximum lean body mass from ankle circumference, wrist circumference, and height. Useful for individual prediction when you have frame measurements; less replicable at the population level than FFMI.
• Bodyweight × height-adjusted targets. Rules of thumb like “one pound per inch of height for men at contest lean” persist in bodybuilding circles. These align roughly with FFMI 22–23 at 5–7% body fat for 180 cm lifters.
• Relative muscle mass indices that decompose lean mass into arm, leg, and trunk components. More precise but require advanced body-composition measurement.

For a quick, cross-comparable estimate, FFMI is usually the right tool. For a personally tuned target (say, optimising toward your own frame), Berkhan/Butt-style models with your measured ankle and wrist circumferences are closer to bespoke.

Common misconceptions about FFMI

“My FFMI can only go up by training harder”

FFMI rises with lean mass gained, which depends on training stimulus, nutrition, recovery, and genetic responsiveness. The training stimulus is one input among four. Lifters with identical programs often progress at very different FFMI rates because the other three inputs vary.

“I should aim for FFMI 24 in year 2”

Setting absolute FFMI targets is a trap. What matters is your own progression rate — +1 FFMI per year in years 2–3 is good, and arriving at 22 versus 24 by the end of year 2 is a function of genetic starting point more than training effort.

“FFMI 25+ means steroids”

As covered, the evidence does not support this as a diagnostic statement: Kouri's own pre-steroid Mr. America winners averaged an estimated 25.4, and collegiate cohorts run to a 97.5th percentile of 28.3. Treat FFMI as a soft descriptor, not a prosecutor's Exhibit A.

How body-fat estimation quality propagates into FFMI

FFMI depends on lean body mass, which depends on the body-fat estimate. For an 85 kg lifter at 1.80 m, each percentage point of body-fat error moves FFMI by about 0.26:

Body-fat error     FFMI error (85 kg, 1.80 m)
─────────────────────────────────────────────
±1 point           ±0.26
±2 points          ±0.52
±3 points          ±0.79
±5 points          ±1.31
±8 points          ±2.10

An FFMI of 24 from an estimate that is 5 points off could really be anywhere from about 22.7 to 25.3. The better the body-fat method, the narrower that band; DEXA sits at the tight end and consumer bioimpedance scales at the wide one.

Time horizons for natural progression

For a lifter starting at FFMI 19, a rough rule of thumb for progression under good training, nutrition and sleep (a coaching heuristic, not a study result):

• Year 1: +2 FFMI points (to 21) is typical for novices.
• Year 2–3: +0.5 to 1.0 FFMI per year.
• Year 4–6: +0.2 to 0.5 FFMI per year.
• Year 7+: Sub-0.2 FFMI per year. Most naturals plateau around 22–24.

The Muscle Gain Potential Calculator projects a timeline with a similar declining-gain model (Lyle McDonald's yearly gains), starting from your frame measurements and training years rather than from FFMI. These are averages — genetically gifted individuals progress faster and plateau higher; unlucky genetic draws plateau earlier.

Hedges

Practically, any claim that “X person must be enhanced because their FFMI is above Y” should be read as weak evidence, not diagnostic. The distribution overlaps meaningfully at the tails.

• FFMI 25 was the upper bound of self-reported non-users in one 1995 sample, not a crisp ceiling.
• Height-normalised FFMI is more comparable across lifters than raw FFMI.
• Measurement error in the body-fat estimate easily adds a point or more of FFMI noise.
• Use FFMI to track your own progress and for population percentiles — not to diagnose strangers.

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

Population boundaries of Kouri's 1995 sample

Honest interpretation of FFMI 25 as "natural ceiling" requires understanding who Kouri et al. actually measured^[1]:

• n = 157 male athletes. A sample of men already training seriously, 83 of them steroid users. The distribution says little about recreational lifters or about what is biologically possible in general.
• Cross-sectional, not longitudinal. Kouri measured each participant once. Nothing in the study tracks how naturals progressed over time or whether any individual's personal ceiling was lower or higher than the group's observed maximum. The "ceiling" framing is an inference from the upper tail of a snapshot distribution.
• Steroid-use classification was self-reported. The non-user group was assumed to be genuinely natural; there was no drug testing, and the collegiate cohorts cited above did not report any. The "25 in naturals" figure rests on self-report.
• Female data absent. Kouri's paper is male-only. In 372 female collegiate athletes the mean height-adjusted FFMI was 18.8 ± 2.1 with a 97.5th-percentile upper threshold of 23.9^[6], against 28.3 in the comparable male cohort^[3] — a female distribution several points lower, not a rescaled copy of the male one.

Alternative-view framing: frame-based vs ratio-based models

FFMI assumes height-squared is the right scaling variable for lean mass. Two alternatives that handle the same question differently:

• Casey Butt / Martin Berkhan frame-based models. Predict maximum lean body mass from wrist circumference, ankle circumference, and height. The underlying assumption: skeletal frame dimensions reflect growth-plate thickness and overall structural mass-bearing capacity. Produces individualised predictions (e.g. "given your 17.5 cm wrist and 22 cm ankle at 180 cm, your maximum contest lean is ~80 kg"). Trades replicability for specificity.
• Allometric scaling. Ford and colleagues analysed the world weightlifting champions of 1993–1997 and found weight lifted scaled with height to the power 2.16, implying muscle mass scales with roughly height cubed rather than height squared.^[4] FFMI divides lean mass by height squared, so if lean mass really scales closer to height cubed the index does not fully normalise: taller lifters drift upward on the FFMI scale and shorter lifters downward, for the same build. Compare yourself against people near your own height rather than against the raw FFMI scale.
• Muscle thickness / CSA at specific sites. Ultrasound or MRI cross-sectional area at standardised sites (mid-thigh, upper arm) is the most physiologically direct measurement of muscle mass. Clinical only, but the gold-standard comparator for any FFMI-based population claim.

Worked example: tracking personal progression vs comparing against ceiling

A lifter, 182 cm, starts training at 76 kg and 18% body fat. Track FFMI over 4 years:

Year  Bodyweight   BF%    LBM     Raw FFMI   Normalised FFMI
───────────────────────────────────────────────────────────────
 0     76.0 kg    18%    62.3    18.8       18.7
 1     82.0 kg    15%    69.7    21.0       20.9
 2     86.0 kg    14%    74.0    22.3       22.2
 3     88.0 kg    13%    76.6    23.1       23.0
 4     89.0 kg    13%    77.4    23.4       23.3

Year-over-year ΔFFMI (normalised)
  Y0→Y1  +2.2  (typical novice year)
  Y1→Y2  +1.3  (intermediate, still strong)
  Y2→Y3  +0.8  (approaching natural asymptote)
  Y3→Y4  +0.3  (plateau region)

This lifter is progressing normally within the trained-male distribution, plateauing around FFMI 23.3, inside one standard deviation of the 22.8 collegiate-athlete mean^[3]. The comparison "am I above or below the 25 ceiling" is less informative than the comparison "is my year-over-year delta consistent with the natural-progression decay model." A lifter at FFMI 22.5 in year 4 who is still gaining 0.3+ per year is doing better than one at FFMI 24 in year 4 who plateaued in year 2 — the trajectory beats the snapshot.

Common failure modes

• Reading FFMI off a bathroom BIA scale. Consumer BIA scales can miss body fat by several percentage points; at 85 kg and 1.80 m a 5-point miss is ±1.3 FFMI, so a scale-based "FFMI 25" could really be about 23.7 to 26.3. For any serious comparison, use DEXA or skinfolds with a trained tester.
• Treating the 25 cutoff as diagnostic. A single data point at FFMI 25.2 is not evidence of steroid use. The height-adjusted 97.5th percentile in male collegiate athletes is 28.3^[3], so 25 is nowhere near the top of the observed trained-male distribution. Diagnostic claims on FFMI alone are weak evidence and irresponsible in public discussion.
• Height normalisation ignored. A 160 cm lifter at raw FFMI 22.0 and a 195 cm lifter at raw FFMI 24.2 both normalise to about 23.3. Same normalised body, raw numbers 2.2 apart. Compare only normalised values across heights.
• Chasing FFMI targets on a bulk. Glycogen supersaturation from high-carb eating can inflate measured LBM by 1–2 kg transiently (each gram of glycogen binds ~3 g of water). Pre- and post-refeed FFMI measurements can differ by 0.5–1.0 without any real lean tissue change.
• Ignoring frame. Two 180 cm lifters at FFMI 23 can look markedly different if one has 19 cm wrists and the other 15 cm wrists. FFMI doesn't encode frame, so using it as a visual-aesthetic target fails; use it as a lean-mass-for-height progress metric instead.
• Using FFMI to self-assess contest readiness. Contest physique is about leanness and muscle distribution, not aggregate lean mass. A lifter can hit FFMI 23 at 18% body fat and still look dramatically different at 6% body fat (same lean mass, much less fat mass, different visual presentation). FFMI is a rough size metric, not a contest-readiness indicator.
• Assuming FFMI growth is linear. Progression decays sharply with training age: year-1 gains of +2 FFMI are typical for novices, year-5 gains of +0.2 FFMI for advanced lifters. That shape is consistent field observation rather than a published model, but extrapolating a year-1 trajectory to year 5 produces wildly optimistic targets that compound into frustration when they don't materialise.

FFMI across strength sports

Observed FFMI distributions vary by sport selection, reflecting the body-composition that the sport actually selects for:

Sport                    Typical FFMI range   Notes
────────────────────────────────────────────────────────────────
Natural bodybuilding     22–25                Contest-lean, tightest range
Powerlifting (raw)       22–26+               Wider range; higher bodyweights possible
Olympic weightlifting    22–25                Height-biased sport[4]
CrossFit (competitive)   21–24                Balanced with aerobic capacity
Strongman                24–28+               Bodyweight is an asset
Recreational lifting     19–23                Pareto distribution
Endurance running        16–20                Explicit opposite selection
Rugby forwards           23–26                Positional selection

Notice the range shift: strongman selects for higher FFMI because bodyweight itself helps performance; endurance running selects against it because carrying mass costs economy. A rugby forward at FFMI 25 and a natural bodybuilder at FFMI 24 may have similar lean mass but very different leanness and visual presentation.