ABSTRACT

A series of six 88 graeco-latin square experiments was conducted to investigate reliability of visual estimates of grain abundance. Artificial mixtures prepared from sand-size fragments of 9 minerals plus basalt and siltstone were used. Variables studied include observer bias and precision, type of grain estimated, true percentage abundance, number of background materials, observer's level of geologic training, and effect of using or not using a percentage comparison chart. Visual estimates are unbiased but not very precise. Standard deviations of observation errors ranged from 16 percent when the material estimated comprises 55 percent of the mixture to 5 percent standard deviation for materials comprising 10 and 95 percent of the m xture. There were no effects in accuracy or precision which could be related significantly to number of background materials or level of geologic training. Mineral type was insignificant in 5 out of 6 squares. There is not significant improvement in accuracy or precision when the observer uses a percentage comparison chart.

Error curves for visual estimates vary with the true percentage in a manner similar to errors in a binomial population; a visual estimate has a precision approximately equal to an estimate based on counts of 16 grains. Experimental error in percentage determinations in mixtures can be cut in half by using a sample count of size 64 rather than a visual estimate with or without a comparison chart. The binomial approximation is valid only if point count spacing equals or exceeds mean grain diameter.