In common language, as well as in knowledge–based systems, the truth of a proposition can be evaluated in a qualitative manner using adverbs usually represented on a scale of symbolic degrees. To combine or aggregate such symbolic degrees, we may need scales of different precision levels. We propose to model small variations inside a degree scale using linguistic modifiers in a symbolic framework. We formally define such modifiers and we distinguish three families: reinforcing, weakening and central modifiers. We also introduce the original notion of intensity rate associated to a linguistic degree on a scale base. After that, we propose a generalization of our modifiers, so we obtain more sophisticated tools. These have been used, in particular, in an application on colorimetry that allows us to alter the color slightly. The importance of our work is that most of our linguistic modifiers verify some interesting properties on their intensity rate: notably, they assume a certain order relation.