Manual error tagging of learner corpus data is time consuming and creates
a bottleneck in the analysis of learner corpora. This had led researchers to
apply techniques from the area of natural language processing to assist in the automatic analysis of such data. This chapter presents the novel application of a hybrid approach to the detection of spelling errors in learner data. The Variant Detector (VARD) software was developed to match historical spelling variants to modern equivalents with the intention of improving the accuracy and robustness of corpus linguistics techniques when applied to historical corpora. Here, we describe its application to detect spelling errors in written learner corpora consisting of 50,000 words from each of three learner backgrounds (French, German and Spanish).