The methods and techniques currently used to measure oscillator strengths (or, equivalently, f-values and transition probabilities) are reviewed. Both linear and non-linear optical methods are discussed.Following references to critical compilations and bibliographies, the definitions concerning the interaction between an electromagnetic radiation field and a medium (consisting of free atoms, for example) are given: linear and non-linear susceptibilities and radiative constants are introduced. Next, the basic principles of linear methods (dispersion, absorption and emission) and non-linear methods (interaction effects near and off resonance) are explained.Measurement techniques are described in detail. The subjects covered are as follows: introductory remarks on equilibrium, number densities, and relative and absolute oscillator-strength scales; methods based on anomalous dispersion (hook and fringe-shift methods, further interferometric techniques and magneto-rotation); methods based on emission and absorption (the curve of growth, absorption methods yielding absolute and relative $values, and emission methods). We introduce the designation 'Ladenburg method' for a combined method that permits oscillator strengths to be determined without assumptions on plasma state. Subsequently, we discuss the methods based on non-linear interactions (phase-matching in non-linear wave mixing, stimulated Raman scattering, dynamic Stark effect) as well as further methods involving lasers.We emphasise the progress toward more accurate measurements in quantitative spectroscopy and the concomitant new applications. We also point out areas where technological advances concerning light sources, spectrometers and standards, open new opportunities for further, more refined studies in experimental and theoretical atomic and molecular physics.