Random error equally likely to be plus or minus, and if of small dimensions, reasonably likely to be compensated by further errors. In contrast, systematic or biased errors all fall on the same side of correct measurement and may therefore accumulate and produce serious discrepancies.