To follow up on Mark's post, below, on the bottomless fatuity of Robert Fisk: we gave a bunch of these items in -ess to the members of the American Heritage Usage Panel some years ago; Kristen Hanson and I reported some of the results in an LSA paper in 1988. What we found is that even then, the generally conservative and venerable writers and editors on the panel were bailing out on the suffix, retaining it only where it had a certain historical signficance. Here are the relevant paragraphs of the paper:
In one set of items, we asked the panelists to judge the acceptability of sentences containing a variety of words with the suffix -ess, including ambassadress, sculptress, waitress, stewardess, hostess, actress and seductress. Some of the results are shown in examples (23)-(30):
23. When the ambassadress arrives, please show her directly to my office. women: 13% men: 28% total: 23%
24. Georgia O'Keefe is not as well-known as a sculptress as she is as a painter. women: 15% men: 42% total: 34%
25. Mary Ann is such a charming hostess that her parties always go off smoothly. women: 74% men: 92% total: 86%
26. Mata Hari used her ability as a seductress to spy for the Germans. women: 69% men: 70% total: 69%
27. His only hope now is to marry an heiress. women: 92% men: 95% total: 94%
28. Mr. Bhutto's daughter and political heiress, Benazir Bhutto, returned to Pakistan in April. women: 28% men: 37% total: 34%
29. There are not very many good parts available for older actresses. women: 92% men: 98% total: 96%
30. He coaches British actors and actresses in the pronunciation of American English. women: 73% men: 83% total: 80%
It is notable that the large majority of the panelists have reservations about continuing to use at least some of these words in the traditional way. Only 23 percent of the panel and 28 percent of the men continue to accept the suffix in all words, judging by the number who allowed it in the least acceptable item, ambassadress. (This result is particularly striking when you consider that the average age of men on the panel is probably well over 60.) Most panelists have apparently taken into account the objections by some feminists that the suffix implies invidiously that there is a difference in social and occupational roles according as they are performed by a man or by a woman.
At the same time, we can infer that an even smaller proportion of panelists have categorically dropped from their lexicons all words containing the suffix: more than 90% accept the use of actress in (29) and heiress in (30). Instead, panelists have evaluated the relevance of this claim on a word-by-word basis. For occupational terms, the panel appears generally to have accepted the feminists' argument, as witness the low acceptability of sculptress and ambassadress, but they have made an exception for actress, presumably because acting is an occupation in which the parts one can play depend on one's sex. With terms denoting social roles, by contrast, the panel appears to have felt that the distinction was legitimate: hence 86 percent of the panel accepted the suffix in hostess, and 69 percent accepted it in seductress.
Note moreover that the panel made significant distinctions of acceptability among different uses of the same word. In (27) we see that 95 percent of the panel accepted heiress in a social context, while only 34 percent accepted it in a political context in (28). In (29) and (30) we see that the acceptability of actress tends to vary according as sex is relevant to the point at issue.
One other item that we didn't report in the paper involved heroine: not suprisingly, many more people were willing to use this of Elizabeth Bennet ("the heroine of Pride and Prejudice") than of Rosa Parks ("a heroine of the civil rights movement"). I am confident that if the questions were given to the current panel, an even larger proportion of the panelists would reject -ess in most contexts. Actually, my guess is that you'd want to ask about a different set of items.
PS. I wrote up a longer paper dealt in greater depth with the implications of these and other observations about the panel's opinions. It appeared in 1990 in The State of the Language, ed. by Christopher Ricks and Leonard Michaels, and turns out to be online at Google Books.