This report summarizes the results of the field workshop of the IAVCEI Commission on Tephra Hazard Modelling that was carried out in Salcedo, Ecuador (January 16-18, 2006) with the main objective of assessing the best way to characterize the largest clasts of tephra deposits. Defining the largest clasts of a tephra deposit is necessary for the compilation of isopleths maps, which are important for two main reasons: i) determination of column height when no direct observations are available (e.g. Carey and Sparks 1986 and Pyle 1989) and ii) definition of eruptive style (e.g. Pyle 1989). In particular, the determination of the column height is extremely valuable because it represents a critical input of tephra models and because it is used to derive information on the mass discharge rate and the duration of eruptions (i.e., ratio between erupted mass and mass eruption rate). Nonetheless, our field exercise has shown the dependence of the results on different averaging and sampling techniques used, confirming the need of a standardized strategy, and that the characterization of the population of the largest clasts that fell at a given distance from the vent is more appropriate than the definition of a maximum clast. Recommendations on the selection of sampling area, collection strategy, choice of clast typology and clast characterization (i.e., axis measurement and averaging technique) are given based on a thorough investigation of two outcrops at different distance from the vent. First, specified-area sections should be preferred to unspecified-area sections when possible (ideally 0.5m2 and a flat paleotopography). Second, in order to avoid large discrepancies from the assumptions of sphere considered in most empirical models (e.g. Carey and Sparks 1986), a clast should be characterized based on the geometric mean of its three axes taken perpendicularly between each other with the approximation of the minimum ellipsoid (lithics should also be preferred to pumices when present). Finally, the method of the 50th percentile of a population of 20 clasts was found as the best way to assess the largest clasts because it has the advantages of: i) eliminating the problem of outlier identification based on a rigorous statistical approach, ii) offering a more reliable reproducibility of the characterization of a given outcrop than the measurement of a small population of large clasts (e.g. 3 or 5), iii) reducing analysis time in the field by requiring the measurement of only one clast (i.e., the smallest of the 11 largest clasts). In addition, the underestimation of values is in the same order of magnitude of the differences due to the choice of the collection strategy, sampled volume and averaging technique and can also be corrected when compiling the isopleth map. Further investigations on the stability of the discrepancy between 50th percentile of a 20-clast population and the largest clasts found at a given outcrop should be carried out. Finally, the survivor-function data should also be calibrated with the method of Carey and Sparks (1986) in order to correct for the discrepancies with the 3- and 5-clast populations typically used.