Halliday & Matthiessen (1999: 170-1):
… participants are located at some point along a scale of individuation, ranging from most generalised (e.g., diamonds are forever) to most individuated (e.g., Elizabeth's diamonds were stolen). The Range element tends towards the generalised end of the scale. This is especially the case with those of the elaborating type, where the Range usually represents a general class; and it is always the case if the figure is one of being, with Range as Attribute. For example:Peter plays tennis (cf. is a tennis-player)Peter plays the piano (cf. is a pianist)His opinion is not important (cf. does not matter)As a corollary to this, when some element that has functioned as Range is carried through the discourse, being picked up either by a lexical repetition or by a pronominal reference, it is more likely to be being picked up as a class, rather than as individuated:Sharon plays tennis at the same time every other day .... Tennis is a wonderful game, but tennis-players tend to be very obsessive.Peter spends a lot of time at the piano ... It is a difficult instrument.Hence a form of reference such as the following is somewhat improbable:Peter used to play the piano; but he sold it.
Blogger Comments:
Again, this assumes that the reference of the piano is usually homophoric. There is nothing improbable about the following instance, where the Range is individuated, not generalised, and acknowledged as such by pronominal reference:
We've always had a range of keyboard instruments. Peter used to play the piano, but he sold it.
Other examples of individuated Ranges, acknowledged as such by pronominal reference:
We climbed the mountain yesterday. It was very steep in places.
We are playing the match tomorrow. It'll be a war of attrition.