For rationalizing molecular analysis of field-collected roots in diversity studies on arbuscular mycorrhiza, we compared three different approaches. After DNA extraction from 50 root samples of Plantago lanceolata grown on monoculture plots at a former arable field site, (1) DNAs were amplified separately by nested PCR and each amplicon was cloned separately; (2) DNAs were amplified separately by nested PCR, 1 mul of each amplicon was pooled, and a single cloning was made from the resulting amplicons mix; and (3) DNAs were pooled and the single amplicon derived from the nested PCR was cloned. Based on these three different methods, 109 nuclear ribosomal internal transcribed spacer sequences were obtained. Methods 1 and 2 enabled the detection of almost similar levels of arbuscular mycorrhizal fungal diversity. However, method 1 was expensive and time-consuming as much more cloning had to be done. Method 3 was completely biased by preferential amplification of nontarget organisms, which were only detected in low frequencies by the other methods.