Introduction: In defense of Fry
Gareth Rees's characterization, in his answer, of Fry's summary of McGann's argument as "misleading if not outright tendentious" seems to overstate the case. To see why, it's necessary to understand the context of both McGann's original argument about Keats, and Fry's rebuttal.
The context of McGann's claims about "La Belle Dame Sans Merci"
As Gareth notes, McGann's reading of Keats forms part of a larger essay written in 1979 on the historical method in literary criticism. McGann's aim was to show that any sound analysis of a literary work requires taking into account social and historical conditions: not only those under which it was produced, but also those under which it has been and is being read. In making this argument, McGann himself was writing a polemic against the New Criticism. The New Critics held that a work of literature is a verbal icon, making meaning strictly through the interrelationship of the words arranged in the text. For the New Critics biographical, historical, or sociological information was irrelevant, and trying to use such information to understand a text was fundamentally mistaken, because such a reading made the text a historical artifact rather than a work of literature. For the New Critics, then, any reading of the work that relies on this "extrinsic" information is misguided, because the proper focus of criticism should be the "intrinsic" literary qualities of the text.
McGann agrees with the New Critics that literature has its own identity that cannot be reduced to its biographical, social, or historical context. He however says that the history of the text's production and reception is a necessary part of literary analysis. The most fundamental illustration of this claim is quite simple: in order to analyse a text, we need to be sure that we have in front of us the actual text. Establishing the actual text requires historical and biographical information: when was the poem written? When was it first published? If there are multiple versions of the text in circulation, how do we know that the text we are looking at is the correct one to analyse? McGann argues that the necessity of textual criticism, i.e., of establishing the definitive form of a text, shows how literary analysis is inseparable from a knowledge of the conditions of writing and reading.
What does McGann actually say about this poem?
To back up his argument with a concrete illustration, McGann discusses the textual history of "La Belle Dame Sans Merci". He says that this poem is more commonly read in its manuscript version (dating back to 1819, printed with Keats's collected works in 1848) rather than in the form of its first print appearance (in Leigh Hunt's periodical The Inquisitor in 1820). The 1820 version was the only publication of the poem in Keats's lifetime. McGann claims that since the 1820 print publication of the poem postdates its manuscript version, this 1820 version must represent the final form Keats gave to his poem. This print version, and not the 1819 version, is therefore the proper object of literary analysis.
McGann argues that the changes between the manuscript and the print versions show Keats adopting a more original, less conventional tone. The 1819 version indulges in romantic conventions, particularly those associated with a medieval setting, but those conventions are brought down to earth by the 1820 revisions. Instead of being about a knight seduced and betrayed by an evil spirit, the poem becomes one about a wight endlessly pining after the lady with whom he has had a one-night stand. McGann uses the more earthy reading made possible by the 1820 version to explain the criticism often directed at Keats and his mentor Hunt by the literary establishment of the period: they were seen as vulgar, a term one would find difficult to apply to Keats today.
It's worth mentioning that McGann would not agree with the question's characterization of the changes as "minimal" or Gareth's claim that they are "very minor": his own words are that the poem "differs greatly" between its two versions (p. 1000). These differences, which represent Keats's own revisions to conform with his intentions, have been overlooked, leading to a misreading of the poem. To understand the poem correctly, therefore, we have to see it in its historical context, beginning with its textual history and with the reception Keats's poems received in his day.
What is Fry's argument about McGann?
McGann agrees with the New Critics that the literary realm is not reducible to the historical, but argues that recognizing a text's historicity is a requirement for literary analysis. Fry, too, both partly agrees and partly disagrees with McGann. He lauds McGann's recognition of every text's historical nature:
McGann's general pronouncements about the historicity of texts, about the saturation of texts by the circumstances of their production and by social pressures, are bracing and worthwhile. The idea that a text just falls from a tree—if anybody ever had that idea—is plainly not a tenable one, and the opposite idea that a text emerges from a complex matrix of social and historical circumstances is certainly a good one. (p. 254)
He then proceeds to turn McGann's procedure against himself. Just as McGann explains why the critics of Keats's day accused him of vulgarity, Fry explains why McGann argues that the 1820 version of "La Belle Dame Sans Merci" should be seen as the authoritative one. And it is in fact the same reason in both cases: the 1820 version is not conventionally romantic. McGann argues that the changes between the 1819 and 1820 versions show Keats's irreverence to the contemporary literary establishment. A fatal seductress casting a spell on a knight was quite in keeping with both contemporary views of women's sexuality and prevalent poetic conventions. By contrast, a woman (not an elf or sprite, but an actual woman) engaged in mutual lovemaking with an ordinary wight was shocking as both social fact and poetic subject.
Fry points out that McGann, like Keats's contemporary critics, was invested in judging poetry by the social standards of his day. By 1979, the view that sexually active women must be fatal seductresses was outmoded, at least among hip literary types. Fry says McGann lets his desire to present Keats as pioneering, or egalitarian, dictate his argument. So rather than presenting a conventional and conventionally misogynist romance, Keats must present a sexual encounter and its unfortunate aftermath. Fry argues that this reading is just as colored by the sociology of McGann's day as the reading of Keats as "vulgar" was colored by his own times.
Fry goes on to show how McGann's reading is not well supported. It rests on the claim that the changes in 1820 print version were made by Keats himself. Fry disagrees:
who's to say that the 1848 text wasn't Keats's last thought? He was already ill when the Indicator text was published in 1820. He was close to the end of his ability to think clearly about his own work and to worry very much about how or whether it was published; and at the same time, we don't know when Brown received his version of the text. We can't suppose Brown just sat down and rewrote it, and if he didn't rewrite it, then Keats must have given it to him in the 1848 form. Who's to say that wasn't his last thought, if not his best? (p. 255)
Is Fry's argument against McGann sound?
The foundation of McGann's argument seems a bit shaky, because his assertion that the changes in the 1820 version of this poem represent a deliberate revision by Keats is unsupported by convincing evidence.
The weakness of the claim can be seen in McGann's discussion of "A Dream, After Reading Dante's Episode of Paolo and Francesca". This sonnet first appeared in print in the same issue of The Indicator as "La Belle Dame Sans Merci". Like the latter poem, this sonnet too has an 1819 manuscript version which was reprinted in 1848 as part of Keats's collected works; and McGann similarly argues that the 1820 version should be given preference. The version in The Indicator has:
Not unto Ida with its snow-cold skies,
Nor unto Tempe, where Jove grieved a day (ll. 7–8)
The printed version of 1848, based on Keats's manuscript, has:
Not to pure Ida with its snow-cold skies,
Nor unto Tempe, where Jove grieved a day.
The argument for refusing The Indicator reading in line 7 is a textual one: "'Not unto Ida' in 7 looks like a copyist's or printer's mistake based on the similar wording in the next line." Although the point is a good one, Keats's situation at the time of publication makes it somewhat less than conclusive, especially since Keats almost certainly read a proof of the sonnet before it was printed. (p. 1004)
The quote within the quote is from editor Jack Stillinger's notes to this sonnet in the standard modern edition of Keats's complete poems.
"Almost certainly"? McGann offers no evidence for this claim of near-certitude. Was it the case that in the early 1800s, periodicals routinely offered poets proof-texts of their work prior to publication? (Based on my own experience, periodicals don't offer proof-texts to writers even today; both online and in print, I've submitted work that has been edited and published without further input from me.) Would Hunt have offered Keats a look at the poems to get him to sign off on the forms in which they were published? Given the relationship between the two, I assume rather that Keats would have deferred to Hunt's editorial changes in rather the same way that Eliot deferred to Pound with regard to The Waste Land. Do we have anything substantiating the claim that the changes in the 1820 versions of these two poems were Keats's rather than Hunt's? No. So Fry is correct in pointing out the lack of a solid foundation to McGann's argument about these specific poems, even as he accepts McGann's larger claims about the historicity of texts.
Conclusion: Fry's goals, (a) substantive (b) pedagogical
Fry's goal is not to propose an alternative reading of "La Belle Dame Sans Merci" on the theory that the 1819/1848 version is the correct one. His goal is to question the very notion of an authoritative text: "who's to say?" His point is that faced with competing versions of a given text, unbiased literary analysis cannot rely on one version or the other as correct based on the author's intentions, or on our aesthetic or political preferences:
while there are all sorts of things to say about the scrupulous intelligence of Keats ... there is in fact no reason to admire his recorded opinions about women. McGann wants to infuse Keats with a pleasing political correctness. From our own historical horizon ... we don't want Keats to have thought in that demeaning way about women. But his letters and his other poems suggest that he did. Therefore, you might feel free to conclude, the 1848 text is the one he intended and preferred. Or rather, since both texts do exist and offer different sorts of interest, let's just have both of them, which the various editions after all do give us. (p. 256)
One might hold that Fry is being mean to McGann. But he is being polemical in the same way that McGann himself was polemical against the New Critics. At least in the discounting of the author's intentions as a locus of authority, Fry is on the side of the New Critics against McGann.
Finally, while McGann published his essay in a scholarly journal, Fry's book is a textbook, meant for readers who are new to literary analysis and likely would find it difficult to work their way through densely argued scholarly essays. This probably explains his somewhat flip tone. Setting that tone aside, his argument bears weight. Literary scholars today take for granted the importance of biographical, sociological, and historical detail to literary analysis, demonstrating the triumph of McGann's arguments over the New Critics. They also, post Derrida and Foucault, take for granted the indeterminacy of the text, the absence of a single authoritative source of meaning, and the text's entanglements with power structures. In its approach to McGann's essay, Fry's textbook is explaining these theoretical concepts in a practical way. His representation of McGann may verge on caricature, but I don't find it misleading.