Just for once, I'm ahead of the game, even if only by a little bit. Usually I write an overview of the Arthur C Clarke Award shortlist at the last minute (2009, 2010), which means I write with a clear sense of the general opinion of the nominated works, and have a benchmark against which to make my own judgements. This year most people seem to still be making their way through the books, and I'm not sure opinion has settled. One straw in the breeze is the competition at Torque Control to guess the winner, which suggests that Ian McDonald’s The Dervish House, Lauren Beukes’ Zoo City or Patrick Ness’ Monsters of Men would be the most popular choice. Another, of course, is the BSFA Awards ballot, on which McDonald and Beukes also appear, along with Tricia Sullivan’s Lightborn, which I have somehow perceived as gaining momentum over the last couple of weeks. That leaves the two Powers, Richard and Tim, on which I think it’s fair to say opinion is divided.

Declare is certainly the runt of the litter as far as I’m concerned; put me in the same camp as Adam Roberts and David Hebblethwaite. Leaving aside the fact that it's too centrally a fantasy novel to meet my interpretation of the Clarke's eligibility criteria, it is straightforwardly one of the worst books I've read in the last year. An obsessively explanatory, numbingly long-winded secret history and spy would-be thriller, Declare is the bore at the party who needs to tell you everything they know about their pet subject. Half of the book is flashbacks to the 1940s, all of which could have been excised entirely to leave a slightly less tediously earnest story with much less geographical and historical tourism. What limited interest I can muster has mainly to do with Powers' framing of encounters with the supernatural, religious and otherwise, as inherently shameful; but even this tips over into parody by the end, wherein it is revealed that left-wing philosophies are so unnatural they can only be sustained in the world by unholy pacts. (Wax Banks offers a more charitable reading here.) It doesn't help Declare that a couple of weeks after closing it for the last time I read Ken MacLeod's The Restoration Game, a relatively minor novel by its own light, but one that covers much the same territory as Declare with a lighter touch. If you must read a novel about an heir to an important bloodline who gets caught up in East-vs-West spy shenanigans focused on control of a fantastika macguffin that reveals the true nature of reality, may provide protection to certain countries, and happens to be located on a peak in the Caucasus mountains, read MacLeod's. At the very least, his book is shorter.

Formalities over with, we can move on to the shortlist proper, at which point the winnowing process gets rather more tricky. But since I have to knock all but one of them down:

Zoo City. The greatest thing Beukes' second novel has going for it is the intricate construction of an energetic alternate Johannesburg (a minute into the past, now -- it's set in March 2011), through which Beukes' ex-journalist protagonist-narrator Zinzi December can move. And move she does: Zinzi sees the heights and the slums, the right side of the tracks and the wrong, and demonstrates the ways in which her world and her life have been shaped by Beukes' most obvious speculative conceit, the Acquired Asymbiotic Familiarism that has given Zinzi a sloth for a companion. Zoo City is at its best when charting the ways in which this condition, somewhere between a disease and a conviction, has become a foundation for societal prejudice, and when exploring how this prejudice interacts with others with which we are more familiar. Meanwhile the narrative takes off with sufficient velocity that you're giddily whisked off your feet and carried along right to the end of the story. As John Clute put it, it doesn’t stop for breath until it’s done for good; it’s just a shame that when it is done for good, you're dumped somewhat unceremoniously at the exit. The ending is the novel’s main flaw: it’s a loss of poise, an unnecessarily neat resolution and a betrayal of the noirish story Zinzi had seemed to be living. If it wasn't for the fact that the journey matters so much more than the destination, the flaw might be fatal; as it is it's enough to create the sensation that the book is emptied-out when you're done, with little to discover on a return visit.

Generosity, and its ambiguous narrator, are defined by anxiety, a tone that seems to take us very far away from Zoo City but which, it becomes clear, is a response to the same globalising modernity that Zinzi confidently inhabits. In fact Generosity resonates with great swathes of contemporary genre sf. It's a future-oriented novel after the manner of Pattern Recognition; it’s suffused with the immanent spirit of genetic transhumanism; it’s a novel about speculative science after the manner of Gwyneth Jones' Life; and it agonises about the ability of fiction to grapple with modernity, imagining the evolution of "a new, post-genomic fiction [...] that grasps the interpenetrating loops of inheritance and upbringing" (a wish that, for me, anticipates a novel like The Windup Girl, which I'd say is the major omission from this shortlist). It does all of these things very well. But the actual story -- that the network of genes coding for proteins that predispose individuals to higher than average baseline levels of happiness is mapped and published, and that this drastically alters the life of a young woman whose genetic profile sits right out at the end of the predicted bell curve for this trait, and embodies a radiant generosity that enriches the lives of all those she knows -- is handled unevenly. Powers rigs the game by refusing to give us the woman's viewpoint for more than a couple of pages, clearly feeling that it's necessary to preserve the mystery of her experience for his readers as well as his characters; and arguably it is necessary for the ending he's working towards. But by making Thassadit Amzwar a female Algerian refugee in Chicago, Powers sets her up as unapproachably other, and narrative hand-wringing about appropriation and exoticisation (of which there is plenty) can't entirely offset the damage done by her "otherworldly glow"; and that ending itself is no great shakes, either, being a philosophical retreat from the science-fictional potential of the novel's premise to a place of meaningless comfort. I think James Wood, in his review, is characteristically blind to many of Generosity's virtues, but has the measure of its failings.

Lightborn. It's striking that with three books gone I haven’t yet discussed a novel that’s significantly set in the future. Generosity faces in that direction but doesn't step over the threshold and Declare looks to the past, while Zoo City and, now, Tricia Sullivan's novel are set in alternate near-pasts. Where the rationale in Zoo City is pretty clear, however, here the world is sufficiently different – the invented technology called Shine sufficiently embedded in the fabric of the place – that it seems it might just as well have been near-future sf. That it's not feels significant, and it colours how the book reads. We can’t get there from here. Nor is it the only ambiguity on offer. Lightborn is a slippery novel, harder for me to pin down than any of its companions on the shortlist, even having discussed it at some length last year. It’s a novel about growing up, or not, with young but complex protagonists and the potential emergence of a new social order. It’s a novel about wrestling with the culture you swim in, about literally embodying culture as something that can be wrestled with. It’s a lot of fun; you never quite know where the next sentence is going to take you. And what I admire most about Lightborn are its messiness and its ambition, the sense that Sullivan is reaching for something just beyond expression, and almost getting there. What knocks it out of the running is that almost, at least from the perspective of my single reading. More than any of the other books on the shortlist, I wouldn’t be surprised if the judges’ second read of this one revises their opinion, one way or the other.

Monsters of Men. No book on the shortlist, I think, has more moments of outright brilliance, or more moments of intense frustration. The trademarks of the previous two volumes of Chaos Walking are present and correct -- the driving first person narratives, the piercingly direct emotionalism, the provocative unpacking of sfnal conceits -- and to the mix are added the long-promised arrival of more human colonists, and an enlarged role for the alien Spackle. I’m delighted to see it on the shortlist; Monsters of Men is every bit as remorselessly readable as its predecessors, and the whole makes a story I desperately want to revisit and more fully comprehend. (I’d love to know how many of the judges had read the first two volumes before they read this one; I note, with some surprise, that I’ve heard from a few people who’ve read Monsters of Men as a result of the nomination and enjoyed it in isolation.) But this time around, even if the cliffhangers never quite stop, we know this is the end; this is where we get resolution, where the balls that have been thrown into the air must be dealt with. And while narratively I have no complaints, thematically I'm left feeling a little unsatisfied. The arguments about information and communication that have been developed during Chaos Walking have to do with Noise, the psychic broadcast infection that affects every male and female organism on New World except human women; and Ness's final word on those arguments comes in the form of assertion from a character positioned as a voice of authority. I don't mind the absence of an explanation per se, particularly since that final word puts Ness pretty much where I'd guessed, philosophically; what I mind is the absence of a demonstration. For a series that's been so viscerally about showing, it's jarring and problematic to be, on such an important point, merely told.

All of which leaves The Dervish House, and me feeling a little predictable. A large part of me would like to see Monsters of Men take the prize; a slightly smaller part, Lightborn, or perhaps Generosity. But in the end The Dervish House is the Condorcet winner from this shortlist, a more complete novel than any of its rivals. On a shortlist notable for narrative drive and distinctive voices, it has both; yet you could also say it has the energy of Monsters of Men, the vivid detail of Zoo City, the dazzling future-gaze of Generosity, the wit and science-fictional enthusiasm of Lightborn. It is written with the sort of rhythm, continuousness and specificity that made Brasyl and River of Gods so memorable, but is probably even better, by virtue of being a more controlled, lower-key story, one that never lets anything get in the way of exploring the depths of McDonald's self-consciously romantic Istanbul. (And on a shortlist with a number of disappointing endings, it has the most satisfying.) Let’s not say that it’s perfect: I could argue that that control becomes artifice, now and then. And if I tried, I could suggest that for a novel that is, as Nic Clarke wrote in these pages, about the mosaic of the many systems of the world, it isn't quite fragmented enough: imagine a version of The Dervish House that hands off its narrative between sixty perfectly sketched characters, rather than six explored at more conventional length, think how much more fully that would embody the argument McDonald is making. But I wouldn't want to make those arguments, because sentence to sentence, paragraph to paragraph, and page to page what McDonald has built here gave me more consistent pleasure than any other sf novel published in 2010, and for that, I would give him the award.