Unit conversion: why temperature is not like distance and weight

Ratio scale vs interval scale: the distinction that organizes everything

Before any formula, ask one question: does the quantity you are converting have an absolute zero or a zero by convention? The answer sorts units into two families, and it is what decides whether you multiply or must also add. Length, mass, area, volume, time, energy, all have an absolute zero: zero metre is the absence of length, zero kilogram is the absence of matter. Quantities like these form a ratio scale. Doubling the number doubles the physical quantity, which is why the conversion is a clean multiplication by a fixed factor.

Temperature in Celsius and Fahrenheit is the exception that fools almost everyone. Zero on those scales is not the absence of heat: 0 °C is water’s freezing point, a chosen reference, and 0 °F came from an 18th-century ice-and-salt mixture. They are interval scales, the intervals between degrees mean something, but the ratio between two values means nothing. That is why 40 °C is not "twice the heat" of 20 °C, just as the year 2000 is not "twice" the year 1000. A quantity where doubling does make sense is audio level in LUFS, which is also non-proportional: different scale families, and treating them like ordinary rulers produces error.

Ratio scale

Has a meaningful, absolute zero. Sums, ratios and "double" all make sense. Metres, kilograms, seconds, kelvin. Converting is multiplying by a factor.

Interval scale

Intervals are comparable, but zero is a convention. Differences make sense; ratios do not. Celsius and Fahrenheit. Converting needs an offset.

Temperature: the family with an offset

Three scales dominate everyday life. Celsius and Fahrenheit relate through a factor of 9/5 and an offset of 32, the signature of an interval scale. Kelvin, the SI thermodynamic scale, is Celsius plus 273.15: the degree is the same size, but zero becomes absolute zero, the temperature at which thermal motion stops. That makes Kelvin a ratio scale: in kelvin, 400 K really is double 200 K. This is why physics formulas and laws use kelvin, not Celsius.

1. Quick everyday checks25 °C: 25 × 9/5 = 45; 45 + 32 = 77 °F (a pleasant day). Body temperature: (98.6 − 32) × 5/9 = 66.6 × 5/9 = 37 °C. Boiling water: 100 + 273.15 = 373.15 K.
2. Worked example, is 40 °C double 20 °C?On the ratio scale that actually measures heat (Kelvin): 20 °C = 293.15 K and 40 °C = 313.15 K. The ratio is 313.15 ÷ 293.15 = 1.068, only 6.8% "hotter", not 100%. Now look at the same pair in Fahrenheit: 68 °F and 104 °F, ratio 104 ÷ 68 = 1.53. The same physical pair gives "2×" in Celsius, "1.53×" in Fahrenheit and "1.068×" in Kelvin. Three answers to one question prove that "double" has no meaning on an interval scale.

A fourth scale shows up in US thermal engineering: Rankine (°R), which uses the Fahrenheit-sized degree but starts from absolute zero. So °R = °F + 459.67 and, equivalently, °R = K × 9/5. Rankine is to Fahrenheit what Kelvin is to Celsius: the ratio-scale version of an interval scale. Pinning zero at absolute zero is exactly what turns an interval scale into a ratio scale.

Distance and mass: ratio scales and exact factors

Life is simpler here, and for a deep reason: each unit has an exact factor fixed by definition, not measured in a lab. The inch is exactly 25.4 mm; the foot, 0.3048 m; the statute mile, 1609.344 m. In mass, the avoirdupois pound is exactly 0.45359237 kg, and the ounce is 1/16 of it. The nautical mile is a different animal: exactly 1852 m, defined to match one arc-minute of latitude. The table separates the exact-by-definition value from the approximation you use in your head, and it matches the factors the unit conversion tool applies.

1. Worked example, 5 km to miles, with significant figuresThe raw arithmetic: 5 ÷ 1.609344 = 3.106855961… miles. But the factor is exact; what limits precision is the input. If "5 km" has 1 significant figure, the honest result is 3 mi. If you measured 5.0 km (2 figures), it is 3.1 mi. Writing 3.106855 mi from a rough "5 km" is false precision: it implies a micrometre accuracy the measurement never had. Going back: 5 × 1.609344 = 8.04672 km ≈ 8.05 km.
2. Mass, no surprises10 lb × 0.45359237 = 4.5359237 kg ≈ 4.54 kg. 8 oz × 28.349523125 = 226.796… g ≈ 227 g (half a pound). Pure multiplications, it is literally a rule of three: "1 lb is to 0.45359237 kg as 10 lb is to X".

The factors are definitions, not measurements

One detail changes how you read the table above: those numbers carry no uncertainty. They are not the best value anyone has ever measured for the inch, they are the inch, by international decision. Since the International Yard and Pound Agreement of 1959, six countries (the United States, the United Kingdom, Canada, Australia, New Zealand and South Africa) fixed the yard at 0.9144 m and the pound at 0.45359237 kg exactly. From then on, 1 inch = 25.4 mm stopped being a measured approximation and became an equality by definition.

1. 1929Nautical mile = 1852 m

   The First International Extraordinary Hydrographic Conference, in Monaco, fixes the international nautical mile at exactly 1852 m. The US adopts it in 1954; the UK in 1970.

2. 1959Yard and Pound Agreement

   Six countries fix the yard at 0.9144 m and the avoirdupois pound at 0.45359237 kg. From these follow, exactly, the foot (0.3048 m) and the inch (25.4 mm).

3. 1983The metre becomes light

   The 17th CGPM defines the metre via the speed of light: the distance light travels in vacuum in 1/299792458 of a second. The speed of light, c = 299792458 m/s, becomes an exact constant.

4. 20 May 2019SI redefined by constants

   The kilogram is now defined by the Planck constant h = 6.62607015 × 10⁻³⁴ J·s (exact) and the kelvin by the Boltzmann constant k = 1.380649 × 10⁻²³ J/K (exact). The physical prototype of the kilogram is retired.

The 2019 redefinition closed a loop: today, all seven SI base units derive from constants of nature fixed by number, not from artefacts or properties of water. For you, in practice, this means uncertainty left the conversion arithmetic and moved into the physical realization of the units, building the standard that materializes the kilogram. The factor 0.45359237 never "rounds badly" because it is not measured; what can vary is the scale you use, not the definition.

Mass is not weight, and the pound has two versions

In physics, mass is the amount of matter (measured in kilograms) and weight is the force with which gravity pulls that mass (measured in newtons). They are different quantities, with different units: weight = mass × gravity. A 1 kg object has the same mass on Earth and on the Moon, but weighs about six times less on the Moon, because lunar gravity is ~1/6 of Earth’s. The bathroom scale, despite its name, measures force, it feels your weight and divides internally by g to display "kg". In daily life, "weight" almost always means mass, and that is fine in conversation; in a technical context, the distinction decides whether the numbers add up.

Mass (kg, g, lb)

Amount of matter; unchanged by gravity. It is what a kitchen scale actually compares.

Weight (N)

The force gravity exerts on the mass. Changes from planet to planet and even with altitude.

Force (N)

What changes a body’s motion. 1 newton = 1 kg·m/s². Weight is just the gravitational force.

The mistakes that flip the result

• Using a simple ratio for temperature: guaranteed error, because the scale has an offset (it is an interval scale).
• Claiming 40 °C is "double" 20 °C: it only makes sense in kelvin, where the ratio is 1.068.
• Confusing mass (kg) with weight (N), or pound-mass with pound-force (~4.45×).
• Confusing the US gallon (3.785 L) with the imperial one (4.546 L): nearly a 20% difference.
• Rounding mid-calculation and reporting more figures than the measurement allows.
• Mixing decimal KB (1000 B) with binary KiB (1024 B) in file sizes.

Unit conversion and percentages are the two everyday calculations that most often produce silent mistakes. If your problem involves discounts, markups or the variation between values, see the guide on percentages: increase and discount. And if you work with measured data, spreadsheets, fits, trends, the care that significant figures demand is a cousin of the care that separates correlation from causation: in both, the error is trusting more precision than the numbers actually carry.