Dr Eric Bumstead is to be commended for his recent paper on gold analysis, published in the March 1984 Proceedings. It is not often that someone in industry, rather than academia, takes the time to collect and report on such a volume of data, data which is of interest to everyone involved in gold exploration and development.After reading the paper I felt, as a geochemist and as ayer, that there was another side to this whole question of analysis that few people are aware of, i.e. the problem facing a laboratory and how the laboratory can try to give the customer maximum value for money spent.Analyses for gold are carried out for different reasons, depending on the aims of the geologist or metallurgist who presents the samples. These can broadly be separated into three classes, to answer the following questions.1. "Is there any gold in this sample?"2. "Approximately how much gold is there in this sample?"3. "How much gold is there in this sample?"In every case there is a different balance between accuracy and precision, cost and avoiding chemical problems. By "chemical problems", I mean such possibilities as samples containing high levels of sulphides, gold present as tellurides, cyanide being present, contamination with mercury, and other mineralogical oddities, plus problems such as samples containing grease.There is no such thing as one assay to cover all possibilities.Samples in category I come mostly from early stages of exploration.There are often large numbers of samples, their mineralogy is not well known and hence chemical problems could exist, but the explorer wants assays at a minimum level of precision and expects minimum cost. These characteristics do not easily come together, for although the precision required is low, the accuracy required is high at low levels. A small positive result when the true answer is nil could lead to a fruitless search and a waste of money. Conversely, a nil result, when the true answer is a small but significant level of gold, could lead to the passing over of an area that may warrant further work.Because there are conflicting requirements for this type of assay, the methods chosen must be a compromise and different assayers will balance these requirements differently and come to different conclusions. The methods cover sample preparation, assay sample size and variations in assay techniques.