In typical small-plot cotton research, sub-samples of seed cotton are collected and ginned on small laboratory gins for determination of lint fraction (used in calculating lint yield) and fiber properties. Literature is scarce on how results from these sub-samples relate to large samples processed through a commercial ginning sequence. The objectives of this study were to: 1) determine if, and by how much, lint fraction and fiber quality results from hand- and machine-harvested sub-samples differ from results obtained from large samples processed through a commercial gin sequence, 2) determine if there is a significant genotype x sample method interaction for lint fraction and important fiber traits, and 3) compare the precision of different sample methods.

Three types of samples were collected from the 1994 Early Maturing Cotton Variety Test at three Mississippi locations. Sample methods were: 1) 100 random bolls hand-harvested prior to mechanical harvest (boll samples, BS), 2) 400 to 600g sample of machine-harvested seed cotton (grab samples, GS), and 3) remainder of seed cotton (25 to 60 lb per plot) (whole plot samples (WP). BS and GS were ginned on a 10-saw laboratory gin; WP were ginned through a commercial ginning sequence on the USDA-ARS 20-saw micro-gin at Stoneville, MS. Lint fraction was determined from each sample method. Fiber samples from each sample method were analyzed on a Motion Control HVI at the Louisiana State University Fiber Testing Laboratory. Data were subjected to analysis of variance, combined over locations with cultivars considered as main plot and sample methods as sub plots.

Averaged across locations and sample methods, cultivars differed for all traits measured. Lint fraction was 5 percentage points lower from WP than from BS or GS. Fiber length from WP was 0.03 in. shorter than GS, and 0.05 in. shorter than BS. Fiber strength was similar for all sample methods. Micronaire from WP was 0.54 units lower than from BS, but only 0.04 units lower than from GS. A significant sample method x cultivar interaction was observed only for lint fraction--relative differences among cultivars were affected by sample method. Both sub-sampling methods introduced bias into data for lint fraction (and thereby for lint yield), when compared to WP and there was no clear indication if GS or BS most closely resembled WP. Precision, measured by coefficient of variation and r-squared, was greatest for WP for most traits. For most traits, BS appeared to give slightly better precision than GS for most traits at one test site, while precision of GS was similar to, or slightly better than, BS at the other two test sites.

There appeared to be no clear advantage of GS over BS for measuring lint fraction. GS more closely resembled WP for micronaire and fiber length, but BS and GS were similar for fiber strength. The lack of sample method x cultivar interaction for fiber traits suggests that BS and GS are similar in distinguishing relative differences among cultivars in a test. Averaged across locations, BS appeared to be slightly more precise than GS, but the difference was small. As long as relative differences among test entries, rather than actual values, are of primary interest, choice of sample method should be governed more by logistical considerations (e.g. labor availability and distribution, cost, and time) rather than concerns about the quality of the data. If actual values are of primary interest, GS would be preferred.