At this broad level of analysis, then, we associate different Engagement subsystems (for example, Disclaim versus Proclaim versus Probabilise versus Attribute) with different degrees of dialogistic contraction/expansion. It is possible, however, to see this parameter of variation operating more narrowly and at a more delicate level of analysis, as a cline of variation operating within, rather than between, the Engagement sub-systems. Within Likelihood, for example, it is possible to identify different levels of force or intensity - thus, `This is possibly a bad idea' (low), `This is probably a bad idea' (median) and `This is definitely a bad idea' (high). Clearly these different options within Likelihood vary from more dialogistically expansive (possibly) to more dialogistic contracting (`definitely). This optionality is available with many of the Engagement subsystems.
I note in passing, that the more contractive values of Likelihood (e.g. I'm certain this is a bad idea., This is definitely a bad idea, This must be a bad idea) seem quite close in the their rhetorical functionality to values of Pronounce generally. I would certainly not want to argue that there is any major difference here in terms of dialogistic contraction/expansion. The difference is a relatively subtle one - a meta-discursivity distinction turning on whether the speaker indicates that the speaker expressing an assessment of high likelihood or whether the speaker interpolate themselves explicitly into the text as committed `sayer'.