Tom Bartlett wrote to sys-func and Sysfling on 21 March 2015 at 18:53 about the analysis provided by Alice Caffarel-Cayron:
1 Do you want this jacketThis is a possibility, agreed. But in general, I'm very dubious about using ellipsis as a way of explaining coordination, especially if the posited ellipted material is substantial (as here). It begins to look very like underlying deep structure.... And it can be used to explain just about anything, not only troublesome examples. For example, why not analyse the following coordinations as ellipsis instead?
+ 2 or [do you want] to try another
I saw Tom and MaryWhy only invoke ellipsis when we are in theoretical trouble, as a kind of verbum ex machina? What Alice's alternative/agnate examples show, in my opinion, is the functional similarity between the two strings following "Do you want" - both encode a desired change of state. As either way [of] representing the change of state can structurally follow from "would you like" then the speaker can "choose" to coordinate them even though one option is an ngp and the other a to-clause. Got to trust the data!
I would like a bat and a ball.
I saw Tom and I saw Mary
I would like a bat and I would like a ball.
 Interpersonally, the clauses I saw Tom and Mary and I would like a bat and a ball each enact a single proposition, whereas the clause complexes I saw Tom and I saw Mary and I would like a bat and I would like a ball each enact two propositions. Logically, the distinction is between two nominal groups related by paratactic extension, and two clauses related by paratactic extension.
 Ellipsis is not invoked to get out of theoretical trouble; it is a theoretical means of systematically explaining the textual choices that speakers and writers use in creating cohesion in their texts.
 A grammatical analysis of Bartlett's invented example Do you want this jacket or to try another? can be found here. Both clauses can be interpreted as a material process, in which the Process is realised by a hypotactic verbal group complex of desiderative projection. See Halliday & Matthiessen (2004: 515-9).
 Only forms of the same rank can be related tactically and logico-semantically. A nominal group can only be related logically to a clause if the clause is rank-shifted so as to be functioning at group rank.
 Theorists encode the theory by reference to the data, and text analysts decode the data by reference to the theory. In each case, the linguist is a cognitively projecting Assigner participating in an identifying process. Trust, on the other hand, is a matter of faith.