And now we come to an important moment indeed in the development of Michael Mann's directorial personality: his first experiment with film vocabulary. In this latter part of his career, he's become famous - infamous, in some circles - for being an arch-stylist, which is a simpler way of saying that he likes to bend the classic rules of storytelling and visual representation as far as possible, sometimes making something quite remarkable and sometimes breaking the movie in the process. And while there is the slightest tendency towards experimentation in The Keep, with its unusual reliance on unusually close medium shots, that's really not the kind of thing that I'm referring to.

"But what of Thief?" Ay, Thief is indeed not a normal movie at all, but it's not in any sense an "experiment". Poetically abstracted crime thrillers are rare, but hardly as new an innovation as 1981; at a minimum, Terrence Malick had gone to a very similar well eight years earlier in making Badlands, to say nothing of the French films that had mined the same territory since at least the 1930s. No, Thief may be extraordinary and unexpected in its visual elegance, but it is not at its heart an attempt to break any rules; it's just working in a very peculiar emotional register. (The odd mix of slow and fast motion might count, except that the director's cut featuring that oddness was years after the fact).

Thus come we to 1986 and Manhunter, adapted from Thomas Harris's novel Red Dragon, and we find Mann was not interested in just making a procedural about the FBI agent tracking down a serial killer - although it works, sometimes tremendously well, on that level - he wanted to see what you could do with editing. Although to what end, I have absolutely no idea. For Manhunter isn't just the first true example of Mann going crazy with rule-bending, it's the first example of him doing so in a way that ultimately detracts from the film. But I'll get back to the editing.

The film's story follows a retired FBI profiler, Will Graham (William Petersen), who is called back into the line of duty by Director Jack Crawford (Dennis Farina) to help track down a family-murdering psychopath nicknamed "The Tooth Fairy". This killer is a nasty piece of work, and Graham is finding it hard to allow himself to go into the dark mental place where he does his best profiling, so for aid - and, we're invited to assume, to help reacquaint himself with the notion of pure, insane evil - he visits the prison cell of his greatest capture and most psychologically debilitating case, the very case that drove him into retirement. That most wicked of killers is a psychiatrist who killed his young female patients, a certain Dr. Hannibal Lecktor (Brian Cox - and yes, "Lecktor" is thus spelt).

Post-1991, there's a big elephant in the room whenever anyone wants to talk about Manhunter, of course: Jonathan Demme's Oscar-winning The Silence of the Lambs, adapted from Harris's sequel to Red Dragon, and boasting an extremely similar plot: someone outside the FBI mainstream (retiree/trainee) is tasked to help find an unusually crafty serial killer, and refers to Dr. Lecktor/Lecter's unique insight to solve the case, while finding it much too difficult to keep the charismatic psycho from getting into his or her head. It's not a fair comparison to Manhunter at all, simply because the movies' aims aren't the same, and if Silence comes infinitely closer to batting 1.000, that's at least partially because it's facing much easier pitches( some would say that, at a minimum, Cox's approach to Lecktor is more successful than Anthony Hopkins's; these people are best ignored, although Cox undoubtedly gives a perfectly great performance). But, to be as frank as possible, Manhunter isn't nearly as good as Silence, and that's just that. Then again, post-2002, the better point to make is that Manhunter is lifetimes better than Brett Ratner's Red Dragon.

None of this matter back in 1986, and none of it will concern us any further now. Let us instead concern ourselves with Manhunter and Mann, and how the film represented both a tremendous leap back to his safety zone after the befuddling, ill-advised The Keep, and another bold step towards becoming one of America's most prominent art-action directors.

As a story, this is all right squarely in the director's wheelhouse. Graham is a quintessential Mann protagonist (though performed by Petersen without much distinction), torn between two duties: he knows on the one hand that he is uniquely qualified to stop the Tooth Fairy, but he also knows that returning to the FBI means risking his soul and sanity, and possibly endangering his wife Molly (Kim Greist, showing that her stilted and dull performance in Brazil in fact represented her best work) and son Kevin (David Seaman, showing that Mann is one of those filmmakers who just doesn't have a natural rapport with child actors). Perhaps Mann stacks the deck a bit more than he usually does; I give away nothing by stating that Graham chooses the FBI (there'd be no movie, elsewise), and it ultimately costs him very little, to judge from the warm and shiny closing scene that was probably necessary to make the movie at all marketable, but just flat-out doesn't fit the rest of the plot.

Endings, though, are not what Mann is ever about: he is a director obsessed with process and rising action. On those counts, Manhunter is an unabashed triumph: the scenes following Graham and the other agents going about their FBI business are of the highest imaginable quality. The most thrilling sequence in the movie is about nothing more exotic than watching a group of men try to quickly analyze a letter that Lecktor received from the Tooth Fairy before the prisoner has a chance to discover what's going on. It's a tiny slice of race-against-the-clock perfection, showing all of the circa 1986 state of the art tools that the FBI had at its disposal, and it plays at the same peculiar confluence of documentary realism and visionary abstraction that made the heist scenes in Thief so phenomenal.

Unfortunately, and owing more I suspect to the original novel than to Mann's own inclinations, a hefty portion of the film's running time, particularly in the second half, is not spent with Graham and the FBI at all, but with Francis Dollarhyde (Tom Noonan), the reclusive loner with a fixation on William Blake's "Red Dragon" series of paintings, who we already know the first time we see him puttering around his apartment to be none other than the Tooth Fairy himself. Let me first say this: I do not agree with the sometimes-forwarded argument that Noonan was miscast. He's not much at all like the character is described in Harris's novel, but as far as movie killers go, he does a damn fine job. He's a bit slimy and creepy and altogether discomfiting to watch, and you can tell in every shift of his limbs or quick twitch of his eyes that this is a dude who has some serious issues. The problem is thus not in Noonan, but in the film itself and Mann's treatment of the Dollarhyde subplot that it's such a useless waste of running time. Simply put, this is not a film about the mind of a serial killer, but about the man who tries to dig his way into the mind of a serial killer - a small and terribly important distinction. The Dollarhyde scenes are just plain "wrong", not because they are inherently bad but because they don't function with the rest of the film.

However, they do provide Mann with the canvas for some wanton experimentation, and here's where I get back to the editing. Throughout the film, but most frequently in the Dollarhyde sequences, Mann does something strange: he snips out a couple of frames, so that the action flicker forward. It reads very much like there's a flaw in the movie, but it happens often enough that it cannot be anything but a deliberate choice. My best guess is that this was meant to make those scenes where it occurs more unsettling, but whatever the motivation, it turns out to be a terrible idea: it's distracting as hell, and pulls the viewer screaming out of the movie in a way that Mann ordinarily doesn't try for. The cutting throughout is really quite peculiar and, I'm tempted to say, bad: never as pushy as those little snips, but absolutely contrary to Hollywood conventions without having the glow of divine, mad inspiration found in a Godard film, nor any kind of apparent purpose at all. The contest between Graham and Dollarhyde at the end is a particular awful example of editing gone amok for no reason other than because the editor and director wanted to do something just to be different.

Closely related, especially to that scene, is the generally unsuccessful score with music by The Reds and Michael Rubini, and a healthy chunk of pre-existing music. Gone are the gorgeous Tangerine Dream synths that made Thief and The Keep sound so interesting; replaced by something similar but much worse, a bit bigger than an '80s cop show but subscribing to the same mentality. But the score is far better than the soundtrack, featuring too many pieces by a band called Shriekback, a sex scene with a moody '80s ballad that is humorously ill-advised, and an appearance by Iron Butterfly's "In-A-Gadda-Da-Vida" that is positively awe-inspiring in its inspidity.

Far better for Mann when he just sticks to things that are simple and effective and well-done, such as the cinematography by Dante Spinotti before he became famous, in his first collaboration with the director: it's not as well-shot as Thief - nor even The Keep - but it's a fine-looking work, with great atmospheric night shots. Indeed, outside of the story structure and the bizarre editing, there's not much wrong with Manhunter at all; but those are some awfully big things. Generally speaking, Mann's later experiment with imagery would be more successfully than his botched attempt to play with editing patterns, but as it stands, Manhunter is at the very worst an intriguing misfire, and often much better than that, especially as a key step in the filmmaker's evolution.