RIM has done something very interesting with the touchscreen on their new BlackBerry Storm. As Phone Scoop describes it (and everything I write here is based on media reports — I have no insider knowledge):
The big news about the display is the touch technology. First, it's
capacitive, like the iPhone and the HTC G1. This alone is great news,
but the real innovation is the true tactile feedback. The whole screen
is essentially one giant physical button that you can press down for a
gratifyingly real "click" action. There are no haptics or funny tricks
here; just good ol' tactile feedback, and it works very well.
As you would expect, the click action lends a great deal to the
experience of pressing virtual buttons and text entry. However, it also
enables a whole new dimension (literally) in touch interaction.
Capacitive touch technology requires no pressure; it responds to the
very lightest touch. This means the Storm has two distinct ways to
press the screen. A light touch is just a "touch", while a more
forceful press results in a "click". This makes the Storm the first
phone we know of with what could be considered a "3D" touch screen.
The part above that I highlighted in italics is what I find most interesting. From an HCI-theory perspective, what RIM has done is transform the touchscreen from a two-state to a three-state input device, according to Bill Buxton's three-state model of interaction. (I am assuming that you can drag your finger while pressing on the Storm; if you can't do that then I'm wrong and it's not a three-state device.) That model is described in Buxton's 1990 paper from the Interact conference: A three-state model of graphical input (pdf). (He also describes it in chapter 4 of his more recent book-in-progress on input devices.)
The three states in Buxton's model can be described roughly as "out of range," "tracking," and "dragging," and he shows how this state transition model is very useful in understanding the differences between all sorts of input devices, including mice, joysticks, tablets, and touchscreens. The picture below (copied from his paper) shows the three states as they apply to a graphics tablet with a stylus, which is a three-state device.
In comparison, a basic mouse provides only states 1 and 2, a touchpad provides states 0 and 1, and a touchscreen provides states 0 and 2 (for the reasons why please read the Buxton paper). For touch devices you need to add more complexity to get the extra state transitions needed for more complex actions. This is done typically by adding external buttons or recognizing actions like double taps or tap-and-hold, and usually these additions aren't very satisfying or intuitive.
RIM isn't the first to extend touch interfaces this way. Scott Mackenzie and Aleks Oniszczak published a paper at CHI 1998 describing a very similar technique applied to a touchpad (see A Comparison of Three Selection Techniques for Touchpads). They basically put a button underneath a touchpad so that the device could transition between the three states of Buxton's model in an intuitive way. Pressing harder allows you to drag or rubber-band in a simpler way than pressing an external button or doing a "lift and tap." Their study showed that their "tactile touchpad" gave the best performance of the three methods on a standard Fitts pointing test and people also liked it better. There were some quirks about their prototype that made it less than perfect, but in general the idea was a winner. So why didn't it take off? Why, 10 years later, don't we all have tactile touchpads on our laptops? PC companies and touchpad makers (like, ahem, Synaptics) seem to have missed an opportunity there.
There are similar ideas in the iPod's clickwheel and in touchscreens with selective haptic feedback, but none of those extend the device as much as these two examples do.