This had the unfortunate consequence that, after Seltzer went out of business later in the 1920s, and his edition of Kangaroo went out of print, the only text of Kangaroo available in bookshops around the world was Lawrence's penultimate version of the novel, with the longer Secker ending, but without his final editing-revisions.

In his new CUP edition, Dr Steele decided, rightly, that the out-of-print Seltzer or American variant was (to use Dr Roberts' terminology) "the text he wanted".

In making that choice, Steele also adopted the Seltzer "broken attachments, broken" ending, despite textual evidence that it may not have been the one that Lawrence originally wrote, and ultimately wanted, and despite its "missing" full-stop.

As Steele himself pointed out, the Seltzer ending was just one of several alternative endings with which Lawrence could have concluded the text of his Australian novel.

As soon as I saw the CUP-edition ending, and read Dr Steele's explanation of how Lawrence had (he argued) chosen to end the text of Kangaroo with that truncated Seltzer conclusion, I realised something was wrong.

For I, too, had concluded that this "broken attachments, broken" ending was the one that Lawrence had wanted. But I had come to that conclusion via a very different route to the one Steele followed in the CUP edition.

I realised straight away that one of us had to be mistaken.

One of the "Berg" typescripts, copies of which were in my possession, was clearly the Seltzer setting-text. On what I had assumed was its last page, Lawrence had written "End of Kangaroo", underneath an incomplete sentence that ended, to cite the exact words and punctuation of that typescript:

broken attachments, broken

- ie, those were the final typewritten words on the last page of that particular typescript, and setting-text (note the absence of a full-stop).

Back in 1977 I had decided that this Seltzer "broken attachments, broken" ending, though incomplete, was what Lawrence had finally settled on.

I had arrived at this conclusion because of what I had come to believe was the serendipitous character of the novel - the fact that it had been constructed of odd bits and pieces that Lawrence had picked up, in Sydney and Thirroul, in the course of putting the text together.

My theory was that some earlier accident had caused the text to be cut at that point - 375 words short of Somers's departure from Australia - but that, given the way Lawrence had used casual or "accidental" material in the novel, he had decided to leave it that way, such an "accident" being of a piece with the happenstance nature of the rest of his undertaking.

Indeed, I could not see why else Lawrence would have left the text in that incomplete state - in mid-sentence.

Also, as both Steele and I knew, Lawrence corrected the proofs of the text of Kangaroo in New Jersey in July 1923, and had apparently left the shorter, incomplete Seltzer ending intact. (It is normal textual practice to invest the last text the author worked on with ultimate authority.)

Besides, those words "End of Kangaroo" - in Lawrence's own handwriting - seemed to put the matter beyond doubt, despite the unfinished sentence immediately above them, with its "missing" full-stop.

However, Dr Steele's explanation for choosing the Seltzer "broken attachments, broken" ending was quite different to mine.

His reasoning was that Lawrence had made, not - as I had come to believe - an "accidental" decision to leave the novel there, but rather that he had made a deliberate decision to sever the text at that incomplete point.

(Rather reprehensively, however, he omitted to mention, in his new CUP Kangaroo edition, that the text he had chosen now ended with an unfinished sentence. Moreover, he made no effort to explain - as he was duty-bound to do - why in his CUP ending he had added a non- authorial full-stop.)