The use of information from flanking markers to estimate the position and size of the effect of a quantitative trait locus (QTL) lying between two markers is shown to be affected by QTLs lying in neighbouring regions of the chromosome. In some situations the effects of two QTLs lying outside the flanked region are reinforced in such a way that a 'ghost' QTL may be mistakenly identified as a real QTL. These problems are discussed in the framework of a backcross using a regression model as the analytical tool to present the theoretical results. Regression models that use information obtained from three or more nearby markers are shown to be useful in separating the effects of QTLs in neighbouring regions. A simulated data set exemplifies the problem and is analysed by the interval mapping method as well as by the regression model.