Martin (1992: 131):
The discourse function of Halliday's entity Range (1985: 134) is less determinate and these present another borderline case of nominal groups which may or may not be best interpreted as realising a participant: cross the field, climb the mountain, play the piano etc. This indeterminacy however generates a prediction that Ranges are less likely to provide referents for presuming nominal groups than Goals:
PROCESS ^ RANGE PROCESS ^ GOAL play tennis play the puck drive the car drive the ball serve dinner serve the ball make a move make a doll take a dive take your friend fake an illness fake a Picasso etc.
 To be clear, Martin has defined 'participant' in terms of nominal groups functioning as the Agent or Medium of a clause. This definition already excludes Range: entity. To interpret these as possibly realising participants is inconsistent with Martin's definition of participant.
To be clear, in SFL theory, Ranges are participants, as shown by their potential to serve as Subject.
 On the one hand, this is a non-sequitur, since any purported indeterminacy of a category is independent of its frequency of use. On the other hand, the prediction itself is merely a bare assertion until it is supported or falsified by corpus evidence.
 To be clear, tennis, move and dive are Range: process, not Range: entity.
 To be clear, the car and dinner are here both Goal, not Range: entity. Each is the Medium through which the Process unfolds.