Halliday & Matthiessen (1999: 265-6):
Processes do resemble other processes, but they share different features with different others; no single line-up is dominant enough to form the basis for permanent hyponymy. For example, if we consider a small subset of the words expressing verbal processes offer, tell, promise, threaten, recommend, warn:(1) offer, promise, threaten have the feature 'offer'; tell, recommend, warn have the feature 'command'(2) offer, tell are neutral in orientation; promise, threaten, recommend, warn have the feature 'oriented to addressee'(3) within the addressee-oriented, promise, recommend have the feature 'desirable', threaten, warn have the feature 'undesirable'.(4) offer, promise, recommend take direct participant ('propose to give ... to Receiver'; 'propose that Receiver should obtain ...').(5) tell, warn take circumstance of Matter 'about...'.
Processes thus have much less potential than participants for being characterised and taxonomised. For example, with a process like decide we can add a circumstance to it, saying he decided quickly or he decided on the spur of the moment; but if we want to identify the occasion as unique we have to say this decision, the previous decision, the only good decision he ever made. We can say his absurd decision but not he decided absurdly — at least not in the same sense, since absurdly could only characterise the figure (how he carried out the act of deciding), not the quality of the process of deciding as such.