O n e o f the m ost striking findings o f the N ijm egen studies o f how D utch learners com pensate for lexical problem s while speaking English (B ongaerts & Poulisse, 1989; K ellerm an, 1991; Poulisse, 1990) was the pow erful effect o f task in the selection o f particu lar strategy types. Moreover, the realization o f the lea rn ers’ com pen­ satory strategies (CpS), tha t is, th e ir syntactic form , the am o u n t of in form ation included in them , an d the m odality (speech, gesture) used to encode them , also tu rn e d o u t to be task-related. In m ore con tro lled tasks (reference to ‘difficult’ real-world objects and novel abstract shapes in the absence o f an in te rlocu to r), learners generally w ent to g rea t lengths to solve the ir lexical problem s, using extensive analytic (descriptive) strategies in the object refer­ ence task o r a com bination o f such strategies and holistic strate­ gies (which exp lo re hierarchical relations betw een lexical items, for exam ple, co-hyponymy o r hypernym y) in the abstract shapes task. In m ore naturalistic tasks (retelling in English a b rie f story originally heard in D utch and inform al conversation with a native speaker o f English), learners te n d e d to use a variety o f sho rte r strategies, includ ing n o t only less detailed analytic strategies, bu t also a fair n u m b er o f holistic and transfer-based strategies.