Subjective intensity follows a power function of physical intensity
Aliases: psychophysics · magnitude estimation
What it is
Stevens' power law proposes that perceived intensity can be approximated as a power function of physical stimulus intensity. The exponent determines whether sensation compresses or expands: doubling a physical quantity need not feel like doubling. It is an empirical psychophysical model, not a universal conversion for every visual judgment.
Why it happens
Sensory systems map physical energy to reportable magnitude nonlinearly. A power function accommodates compression when the exponent is below one and expansion when it is above one. The exponent depends on modality, range, method, and context.
The power function was not the first candidate. Fechner's law proposed a logarithmic relationship instead, derived indirectly: treat every just-noticeable difference (JND) as an equal subjective unit, then integrate Weber's law — that a JND is proportional to intensity — and a logarithm falls out. Stevens rejected that indirect chain and used magnitude estimation instead, asking observers to report numbers proportional to sensation directly rather than counting JNDs; power functions fit the resulting data better than logarithms across most continua, which is why the power law displaced the log law as the standard model. One account of why a single power form produces both compression and expansion is that sensory neurons' firing rates themselves scale as power functions of stimulus intensity, with different receptors having different dynamic ranges and saturation points — which channel differs how is a separate matter. Stevens' later formulation added a threshold term, ψ = k(φ − φ0)^n, where φ0 is the absolute threshold, because the plain power function breaks down near threshold.
Studying it
Magnitude-estimation tasks ask participants to assign numbers proportional to sensation, or compare subjective multiples between stimuli. Control reference, random order, adaptation, and instruction wording; fit across several magnitudes and inspect individual scale-use strategies.
The standard check for whether a power law actually holds is to plot both stimulus intensity and reported magnitude on log-log axes: a genuine power relationship appears as a straight line, whose slope is the exponent, with goodness-of-fit from a linear regression quantifying the deviation. The exponent estimate also depends on the paradigm itself: magnitude estimation typically yields a smaller exponent than magnitude production (where observers adjust the stimulus to match a given number) for the same continuum. This paradigm-dependent gap is the regression effect, and it means a reported exponent reflects the measurement method as much as the underlying sensation.
Where it stops holding
Different tasks can yield different exponents, and numeric reports are affected by anchoring, practice, and culture. Reading a chart also includes area estimation, coordinate reasoning, and semantics. An experimental exponent is not a precise interface-encoding formula.
The plain power law breaks down near the absolute threshold, where only the threshold-corrected form holds even approximately; applying the uncorrected function there systematically overstates perceived intensity near threshold. The range of stimuli used in a single session also shifts the estimated exponent: a narrow slice of the intensity range gets subjectively stretched to fill the available number range, a range effect that makes the exponent partly an artifact of experimental design rather than a fixed property of the channel — a different stimulus set can produce a different number.