Martin (1992: 108):
What is perhaps best treated as a hypothesis then is that English uses the unmarked realisations a and /∧sm/ when it does not wish to comment on the centrality of the participant being introduced. However, if it does wish to comment, it scales deictics along the following cline:
least central #||$ He asked any guy he met there. (pl. — any guys) He asked /some/ guy he met there. (pl. — guys) He asked one guy he met there. (pl. — /some/ guys) He asked this guy he met there. (pl. — these guys) most central
 This is not a valid conclusion from the premiss from which it is derived (p107):
Unpublished analyses undertaken in conjunction with this research showed that mass and plural nouns lacking an indefinite article (e.g. frogs, grass) provided referents for other nominal groups three times less often than when an indefinite article was present (e.g. some frogs, some grass). This misconstrues reference to referents as comment on participants.
 This confuses the "centrality" of participants (experiential metafunction) with potential referents (textual metafunction).
 This misconstrues reference to referents (textual metafunction) as the introduction of participants (experiential metafunction).
 On the one hand, this is a bare assertion unsupported by data or argument, while on the other hand, it confuses deixis with reference. Moreover, it mistakenly attributes reference functions to non-specific Deictics (any, some, one). The only genuine instance of reference (this/these) is a case of structural cataphora, in which 'the reference is resolved in the same nominal group where the reference item appears' and the Deictic 'is used to indicate that the Qualifier of a nominal group is to be taken as defining' (Halliday & Matthiessen (2014: 625).