There is light in the darkness. Three lights to be specific. Green. They hang in the blackness, connected by some invisible string and humming their inaudible song. These dots are the image of Splinter Cell, and they're not alone. The authoritative 'flick' of night-vision; Sam Fisher's distinctive crouch and wobble; the agent's grace as he hangs upside down from an overhead pipe. Splinter Cell is a franchise built on imagery. Chaos Theory shows this, because after three outings the series is showing the lack of substance behind the iconic images.

One moment captures the reason behind Chaos Theory's underachievement. It's in the third level, which consists of you raiding a bank vault. It should be classic stealth stuff, with lasers and booby traps and tension so palpable you could choke on it. But it's not. You drop down through the skylight on a rope, sliding through the soft twilight, a method of entry that promises excitement the level doesn't deliver. A moment spikes your anticipation though: when you've thieved from the vault the alarms go off. Security forces are on their way. Your mind races with thoughts of what is to come; taking on a whole squadron of armed officers with smoke grenades and headshots, rolling to avoid their gunfire and belting to the soundtrack of shouts and bullets. But none of that happens. You exit casually, and the threat remains just that.

That's the essential problem with Chaos Theory's Solo campaign – it's never as good as you want it to be. Or, with two dress rehearsals behind it, as it really should be.

There's little wrong with Chaos Theory in all honestly; it's just that so much of it fails to incite the excitement that it once did. We've spent two whole games entering rooms, shooting out lights, and grabbing enemies from behind; we want more this time.

Fisher walks the walk

Familiarity breeds contempt; and while games should theoretically be judged on their own merits, sequels should too be criticised as part of a series. CH is a credible, solid stealth game, and worthy of a purchase by any fan of the genre, but the Solo mode is largely disappointing because you get the distinct feeling of being here before. There are a few, admittedly worthwhile, additions: you can view a 3D map of the level now, to get a better idea of where you should be heading, and a 'soundometer' is a good companion to the lightometer. But more than anything, it's a case of more of the same.

The world is under threat from super-smart terrorists, and it's up to the CIA's favourite boy Sam Fisher to save it. So it means more exotic locations, more names that are difficult to pronounce, and more incomprehensible technojargon. And more of the gameplay niggles we hoped they'd amend for Chaos Theory, all of which prevent the game from rising to a higher plane of stealth.

These are things that should've been addressed by now. Linearity was a big sticking point for critics of the first two games, and it's back with a vengeance in Chaos Theory. Although many of the locations lend themselves well to a more open-ended approach, your passage through them usually ends up being a narrow one. The game attempts to alleviate this somewhat with secondary and opportunity objectives, broadening your focus in the missions, but these are usually perfunctory tasks pretty much en route anyway. Air ducts and the like break up the flow of movement well, but they sometimes feel almost tacked on and trivial. A tighter focus on the main objectives and more means for the player to approach it in their own way would've been preferable.

[pagebreak]

AI is where the game falters too. The enemies, whether they are mercenaries, security guards or soldiers, never put up a great challenge. Popping out lights and breaking glass bottles may get their attention for a minute or two, but they'll return to preset patrol routes and reaction mechanics afterwards. There is a lack of spontaneity in the game; it feels too rigid and mechanical. In a great stealth game enemies should adapt to new situations, become more cautious or trigger-happy when for no apparent reason all the lights have went off. Or at least notice when their mate beside them is divulging all the mission's secrets to some hidden person. The set of rules which govern Chaos Theory are all too transparent, meaning you can just work around them for the chance of a free ride.

More worrying is what has been removed from the game. The 'three alarms and you're out' mantra of the previous two is gone, with the apparent intention of reducing the frustration they created (a quicksave option has been added too). It's an admirable intention – to make the game more accessible – but it robs the game of most of its tension. The quicksave feature is welcome, allowing you to tackle situations with the comfort of a recent save, but a loose hand with alarms proves disastrous to the game's atmosphere. With poor opponents and no limits on alarms, there's little to stop you bolting through lit areas all guns blazing. And it is such gunfire which leaves holes in the game.

The alarm punishment system – whereby enemies becoming more armour-clad when you set them off – is hardly harsh enough. If you are to allow the player to continue the mission after numerous alarms, then you have to at least make it almost impossible to complete, whether it's by adding reinforcements or tightening up the AI. The threat of 'game over' was part of what the, admittedly flawed, previous games compelling. Without it the game's pitfalls become more obvious; the only repercussion of poor stealth being a telling off by Lambert.

A showy but impractical camera angle

The responsibility shifts then from the game to the player – it's up to you to make sure you act with razor-sharp stealth. If you really care about your mission success rating then the tension will come self-created, and most of my criticisms will become irrelevant. If you don't, then the missions become a detached, empty affair; unengaging until the more difficult later missions.

If the Solo campaign displays the game's intransigence then it's the Co-op where it finds a new dynamic. Taking on the role of two trainee Fishers, you and a significant other must work as partners in four designated Co-op missions. It's by far the most enjoyable part of the game, giving the action a newfound authenticity as you squabble over how to tackle a problem, work together to take out a guard or scream out to be revived quickly. The levels cater well to the players' duality, demanding friendly interaction and use of new double moves in their design. Success is rewarding because you actually accomplished something tangible together.

It's in multiplayer that Splinter Cell appears to have found its new lease a life (a great versus mode is available in the Xbox version, Live-enabled and all). The ghosts of the stilted Solo play linger, but they're forgotten quickly because you're actually having fun again.

Gormless cutscenes and a tedious Clancy plotline give Chaos Theory pretensions beyond simple stealth – and they're ones the gameplay falls short of justifying. Multiplayer, conversely, reinvigorates a tired stealth formula, even if it's only for four levels. If they intend to keep it as unaffecting as it currently is, one is tempted to suggest Ubisoft do away with singleplayer altogether for the sequel, and put resources into creating the apex of social stealth.

There, the future lies.