Losses of Fe and Na from a typical basaltic liquid during experiments using the wire-loop method have been determined for both Pt and Ag40Pd60 alloy wire by electron-microprobe analyses of the quenched glasses. The amount of Fe lost is dependent on the duration of the experiment and is particularly sensitive to the ratio of sample to metal in direct contact with it. Fe losses can be minimized by using the thinnest wire possible in conjunction with the largest amount of sample that will adhere to the wire and, if Ag-Pd wire is used, can be kept below the 1% level even in long experiments. Much more serious in terms of changes in the bulk composition are losses of Na, which appear to be in direct linear proportion to the length of the experiment and in inverse proportion to the mass of the sample. Even when samples in excess of 200 mg are used the Na losses are unlikely to be less than 5% except in experiments of less than a few days' duration or on samples with very low initial Na contents.