Input devices are a classic topic in human-computer interaction.

Continuous actions can be divided into two classes(Gibson, 1979).

In the HCI context, performative actions have been studied quantitatively. Fitts law (1954) can be considered the "Law of Pointing". Pointing was considered as a discrete act performed using a continuous device, or like sending a message through a communication channel (in Shannon's terms). Pointing as a discrete act has been exploited in Graphical User Interfaces as well as in speech-gesture interaction (Bolt 1980). That this is commonly believed to be the fundamental paradigm for "intuitive" interaction is testified by its wide use in the movie Minority Report. However, since the nineties some people started to look at continuous gestures in GUIs, like crossing targets or traversing a hierarchical cascading menus, and tried to model them as limits of sequences of discrete gestures. The steering law was derived. Some other people looked at what is in between starting position and final target, and discovered that there are kinematic patterns in goal-directed movements (see Figure 1, from Bootsma, Fernandez, and Mottet, 2004). So, there is much more behind point-like acts. See for example preparatory movements in drumming.

Before industrial revolution, most of human actions in the world were essentially continuous. Supposedly, continuous actions and gestures are more "natural" than triggers. Naturalness here means that control is left to the human manipulator rather than transferred to some machinery. According to the tightness of sensory feedback to the handle, control can be more or less direct/physical. Example: sailing using the tiller or the wheel to control the rudder. In the latter there is a decoupling that allows application of smaller forces. Any potentiometer relies on a (abstract) mental model: a map. It is not a very natural kind of interaction. Examples: audio mixer, fires in the kitchen. Our physical (mechanical) actions control changes in non-mechanical (acoustic, thermal, electromagnetic, etc.) energy whose display is displaced from the locus of action.

Triggers can elicit sustained feedback, and the perceived behavior is sometimes that of autonomous life. Conversely, Enaction is based on motor skills. In the closed loop between perception and action the cause-effect relation breaks down.

When continuous action is relevant, it is important to understand the "more subtle characteristics about the way a movement is done with respect to inner intention. The difference between punching someone in anger and reaching for a glass is slight in terms of body organization - both rely on extension of the arm." ( Laban movement analysis is a useful tool). Continuous action conveys an emotional content which can be organized in terms of kinematic variables and patterns. "Etymologically, the word emotion is a composite formed from two Latin words. e(x)/out, outward + motio/movement, action, gesture."

Token+Constraint systems for tangible interaction with digital information (Tangible User Interfaces - TUI) have no notion of function embodied in the objects. They rather stand on: (i) objects as representations (of information), and (ii) physical manipulation. In TUIs, tokens are container+control. Constraints are used to reduce dimensionality and guide the actions.

Shneiderman's principles of direct manipulation:

• Continuous (persistent) representation of the object of interest
• Physical actions or labelled buttons instead of complex syntax
• Rapid incremental reversible operations whose impact on the object of interest is immediately visible.

Indeed, the direct manipulation of GUIs has a level of indirection. Feedback is not where the action is. Conversely, embodied interaction (à la Dourish) tends to be direct and physical.

A disembodied interface, as most of existing interfaces are, gives a schizoid perception and action in the world. This is related to the concept of schizophonia, coined by Murray Schafer. He has been reported to say that "It shouldn't be allowed to have sounds without knowing where they come from, so that you can destroy the source if you don't like it". Indeed, distruction seems to be the most compelling outcome of large-scale marketing of a partially-embodied interface such as the Nintendo Wii remote.

The controllable dimensions are many. Sensors and actuators should be considered together. How can we tackle this complexity? One possibility is to think in terms of basic phenomena, constructively. So, we should look for fundamental interaction gestalts (Svanaes, Understanding Interactivity, 1999) that we exploit in "natural" interactions. Examples in perception to be inspired from:

• figure/ground (in vision and in music)
• good continuation, streaming
• transparency (à la Albers o à la Metelli)
• pre-attentive features (Healey)

Interaction gestalts may result from abstraction of actual interactions, in the spirit of the Ramsauer drawing tutor (after Franinovic and Visell, at the HGKZ workshop on Sound in Interaction).

How can we proceed in the interactive realm? By constructing workbenches or design patterns. A tentative list of workbenches:

• Equilibrium (as in the the ballancer)
• Effort, resistance
• Warm/Cold
• Affectionate touch (attractive, repellent)
• Ensembles of objects (patterns, waves, activity)
• textures (resistant to spatial or temporal inversion) and their continuous exploration
• extending into depth, proximity, perspective
• constraints (visual, mechanical, acoustic)
• volume inflation/deflation

Specific problems should help finding more gestalts. Any hints from designers?

Bellotti et al raised five questions for sensing-based interaction.

• When I address a system, how does it know I am addressing it?
• When I ask a system to do something how do I know it is attending?
• When I issue a command (such as save, execute delete), how does the system know what it relates to?
• How do I know the system understands my command and is correctly executing my intended action?
• How do I recover from mistakes?

Another problem is how to provide invisible affordances. If my tangible object is a bottle, can it emit a sound that makes me thirsty and induces me to drink from it?

Kandinsky said "Line is a track made by the moving point". We may say that "Sound is a pressure signal made by interactions with and between objects".