1. Controlled trials, of which there are few, do not substantiate claims that diuretics play a role in causing magnesium deficiency. Consequently, the vast majority of patients taking conventional doses of thiazide diuretics (i.e. bendrofluazide 2.5 mg day-1 or equivalent) do not need magnesium supplements. On balance, potassium-sparing diuretics tend to increase serum and intracellular magnesium content; this should not be taken as evidence of prior magnesium deficiency. It remains theoretically possible that large doses of loop diuretics given more than once daily for long periods could induce negative magnesium balance and magnesium deficiency. However, it has been difficult to run appropriately controlled trials in conditions where such therapy is needed (i.e. heart failure) and until more reliable information becomes available no absolute recommendation can be made. 2. Methods for the measurement of intracellular free magnesium levels are now available and are more relevant to the assessment of magnesium deficiency than total intracellular magnesium content; the complex relationship between intracellular free and total magnesium content remains to be defined. Future work involving the effect of diuretics on intracellular free magnesium measurements should make every attempt to avoid the errors of trial design and multiple publication that litter current and past literature.