Halliday & Matthiessen (2014: 369-70):
When a given nominal group refers to a member or members of a class – i.e. to a particular thing or set of things, the full ranges of specific and non-specific determiners can serve as Deictic and the values within the specific and non-specific number systems are quite distinct. Thus, when they refer to members of the class of elephant, an elephant, the elephants and elephants are not interchangeable; for example:
After four months, on New Year’s Eve, Russell reportedly proposed while on an elephant in India.
Do you suppose I’m going to find an elephant walking about the streets of London?
An African elephant lay dying alongside a well-traveled trail. Researchers noted that 38 elephants made a total of 56 visits to the dying elephant – including six visits by her mother and sister. After the elephant died, 54 individuals made 73 visits to her corpse – none by her mother and sister.
Here we could not replace an elephant in Russell reportedly proposed while on an elephant in India with the elephant or elephants. These alternatives would have quite different meanings. Typically, the member of a class is introduced by a non-specific reference, e.g. an African elephant, and once it has been introduced into the discourse, it is referred to again anaphorically by specific nominal groups, as with the dying elephant, the elephant (or, pronominally, it).
Blogger Comments:
Note that Matthiessen is using 'refer' here in two distinct senses: ideational denotation and textual reference, without acknowledging the distinction. This is not helpful, given the confusion of the two senses in Martin (1992).
To be clear, the determiner an in an African elephant realises non-specific deixis, not non-specific reference (the). The indefinite article is not a reference item (Halliday 1985: 295; Halliday 1994: 313; Halliday & Matthiessen 2004: 556; Halliday & Matthiessen 2014: 629). In confusing deixis with reference, Matthiessen is following Martin (1992).