The ecological interpretation of ordinations is easiest when there is a linear relationship between the site variables and the ordination axes. There are methods, such as detrending, which were claimed to remedy the horseshoe- or arc effect in ordinations. "Flexible shortest path adjustment" (FSPA) is a possible alternative. It is suspected that this method may improve the interpretability of ordinations, by reduction of the non-linearity between vegetation and environment in the ordination projection. The efficiency of this is measured by computing the stress between the ordination and the original resemblance matrices. Ecological interpretability is evaluated first by correlating the ordination axes with environmental data. Alternatively, Mantel-correlograms are computed for the vegetation and environmental data. In one example, using 63 relevés of a statistical sample from a transition bog gradient, principal coordinates analysis combined with FSPA yielded an extremely efficient ordination and improved considerably the linearity. The results were further improved when noise was removed from the data by secondary analyses. The efficiency of FSPA is confirmed in a second example, where 71 centroid records of Swiss forest types with very high species turnover rate were ordinated. The problem of interpretation in the application of the Mantel-statistics in non-stationary systems is discussed. It is concluded that FSPA and Mantel-statistics, when used with care, are very helpful to interpret biological and environmental relationships.