The book begins with a review of research into dictionary use. A number of experimental design problems are discussed, in particular the unreliability of questionnaire responses, and the need for detailed accounts of individual dictionary consultations whilst sampling in numbers sufficient to represent specified populations. The experiments reported in subsequent chapters investigate issues raised in the review. The first two studies find that dictionary use during a reading comprehension test affected completion speed but not test scores. The apparent failure of dictionary use to improve comprehension is attributed to the test itself, the dictionaries, and the users' choice of look-up words. The ability of users to interpret dictionary entries is investigated in three further studies which use computers to gather data on large numbers of individual consultations. The findings indicate that there is little difference between three major EFL dictionaries in terms of speed of consultation and overall productive success. They also indicate that Malaysian ESL subjects, who have higher vocabulary scores, are slower in their reading and less successfull in their interpretation of entries than Portuguese EFL subjects. Finally, the findings suggest that overall productive success is unaffected by the presence or absence of examples. The experimental findings lead to the conclusion that dictionary consultation is a process in which users match pre-existing beliefs about word meaning and behaviour against segments in the dictionary entry. Such segments are often selected because they are familiar-sounding and conceptually accessible, but may contain only incomplete or non-essential information. Where pre-existing beliefs and dictionary information conflict, dictionary information is sometimes overridden. Thus word knowledge acquired from a single consultation is often insufficient to ensure productive success. Although it is probably inevitable that word knowledge will be acquired slowly, through multiple encounters, modifications to the dictionary entry and the training of users might help to avoid serious misinterpretation of dictionary information.