The concentration of CO₂ in the chloroplast is less than atmospheric owing to a series of gas-phase and liquid-phase resistances. For a long time it was assumed that the concentration of CO₂ in the chloroplasts is the same as in the intercellular spaces (e.g. as measured by gas exchange). There is mounting evidence that this assumption is invalid and that CO₂ concentrations in the chloroplasts are significantly less than intercellular CO₂. It is now generally accepted that internal conductance (g_i) is a significant limitation to photosynthesis, often as large as that due to stomata. Internal conductance describes this decrease in CO₂ concentration between the intercellular spaces and chloroplasts as a function of net photosynthesis [g_i = A / (C_i – C_c)]. Internal conductance is commonly estimated by simultaneous measurements of gas exchange and chlorophyll a fluorescence or instantaneous discrimination against ¹³CO₂. These common methods are complemented by three alternative methods based on (a) the difference between intercellular and chloroplastic CO₂ photocompensation points, (b) the curvature of an A / C_i curve, and (c) the initial slope of an A / C_i curve v. the estimated initial slope of an A / C_c curve. The theoretical basis and protocols for estimating internal conductance are described. The common methods have poor precision with relative standard deviations commonly > 10%; much less is known of the precision of the three alternative methods. Accuracy of the methods is largely unknown because all methods share some common assumptions and no truly independent and assumption-free method exists. Some assumptions can and should be tested (e.g. the relationship of fluorescence with electron transport). Methods generally require knowledge of either the kinetic parameters of Rubisco, or isotopic fractionation by Rubisco. These parameters are difficult to measure, and thus are generally assumed a priori. For parameters such as these a sensitivity analysis is recommended. One means of improving confidence in g_i estimates is by using two or more methods, but it is essential that the methods chosen share as few common assumptions as possible. All methods require accurate and precise measurements of A and C_i — these are best achieved by minimising leaks, maximising the signal-to-noise ratio by using a large leaf area and moderate flow rate, and by taking into account cuticular and boundary layer conductances.