Terrestrial arthropod surveys and inventories frequently suffer from undersampling bias; common species are over-represented and rare species may be missed entirely. This study compared a rapid (3 days) and intense inventory of spiders from one hectare of a mature beech forest (Fagus sylvaticus) in Hestehaven, Denmark, comprising 8,710 adult spiders of 66 species to a previous, much more thorough, bi-weekly survey of two years duration from the same site that comprised 42,273 spiders (adult and juvenile) of 141 species. Non-parametric species richness estimators were used to assess the degree of undersampling bias in various data partitions. The current study used five experienced, four novice collectors, and five semi-quantitative collecting methods. Method and time of day strongly affected numbers of species and adults per sample. Collector experience affected numbers of species but not numbers of adults per sample. Despite the intensive collecting, number of adults per sample did not decrease over the course of the study. At the end of the sampling, 31 species were still rare in the sample (singletons or doubletons). Non-parametric richness estimators suggest that the actual richness of adult spiders in the study plot at this time of year and susceptible to the methods used was about 80 species. Species turnover between the two surveys (ca 23 years) was remarkably small: the two lists were 92% identical. The base-line study suggests that the rarity of 12 of the 31 rare species was artifactual (10 due to phenology, one to method, another to spatial edge effects). The rarity of the remainder is unexplained and by default is interpreted as undersampling bias.