Published online by Cambridge University Press: 30 October 2009
On-farm field demonstrations provide a useful method of technology transfer for researchers and extension specialists, and the proponents of alternative agriculture have adopted on-farm experimentation as an important means to establish the value of sustainable methods. For on-farm research long narrow strips harvested by machine are often employed in place of small hand-harvested plots commonly used at research facilities. The use of this approach, however, could bring the methodology into question if field variation should make determination of statistical differences difficult. Analysis of harvest data from ten irrigated and four dryland site-years for small hand-harvested corn plots and for machine-harvested long narrow strips was conducted to determine the impact of sampling method upon experimental variation. Analyses of variance indicated that six of the ten machine-harvested irrigated sites had significant yield response to applied nitrogen while only one of the ten hand-harvested sites exhibited significant nitrogen effects (P ≤ 0.10). The difference in statistical precision was not a result of increased range in yields, but rather of more consistent yields for each treatment. An economic analysis was performed, and the results indicated that optimum nitrogen rates depended on corn and nitrogen price ratios. When 1989 corn and nitrogen prices were used, the calculated marginal profit levels between the reduced nitrogen and recommended nitrogen rate ranged from $8.92 to -$8.14 per ha. For the irrigated sites, when experimental variation (F-test of total sum of squares) was compared by harvest method, hand harvest had significantly greater variation than machine harvest in five experiments. Machine harvest was significantly more variable in one comparison. Coefficients of variation for machine harvest were less than for hand harvest in nine of ten irrigated experiments. For the dryland sites, differences between harvest methods were significant in one of four experiments. These data suggest that under certain conditions machine-harvested long narrow strips provide more accurate estimates of treatment differences than do small hand-harvested plots despite potential differences in soil texture, soil fertility, or available soil water.