This thesis concerns alternative concatenating operations in extensive measurements and the degree to which concatenating operations are matter of convention. My arguments are directed against Ellis' claim that what prevents us from choosing alternative ways of combining extensive quantities is only convenience and simplicity and that the choice is not based on empirical reasons. My first argument is that, given certain relational theories of measurement, there can be no more than one concatenating operation per quantity; because combining operations are the only conclusive distinguishing characteristic among distinct extensive measurements and thus the only thing that can serve as an indicator for the identity of a quantity. Rectangular and collinear concatenation, for example, cannot both be used as a way of combining lengths. However, rectangular concatenation could be used to measure some other extensive quantity since it fulfills the necessary conditions for adequate numerical assignment. For those that adopt theories of measurement which can support the claim that dinches measure length, I demonstrate, in the second argument, that rectangular concatenation would be a bad alternative and thus no alternative at all. Implementation of rectangular concatenation requires a new science; a science compatible with dinches violates present invariance principles and a new set must be provided, if possible. It is not clear that there is a set of invariance principles compatible with dinches which also gives basis for a science that explains the phenomena as well as the old one. Moreover, new invariance principles would be unintuitive and would require that certain fundamental concepts change their meaning. For these reasons the choice between them is guided by a number of empirical reasons and is not only a matter of convention.