Yes, it could be done but only by the use of a program or sound scheme, not by the human ear. When processed, sounds can be understood and reveal the text being typed. Computer programs could analyse what you type in the same way they analyse speech. They’d know what you type just by hearing you type it. The software detects the differences in keys and keystrokes.

Theoretically, yes. A computer program could determine the length of the words you type by the distinctive sound of the spacebar indicating a new word. Pauses in your typing could indicate new sentences. These phenoma are only two of the many variables in any SAMPLE the program needs to determine your typign patterns: You see, with enough samples, any A.I. program that listened to you type for long enough, would eventually be able to more and more accurately guess what you've typed. However, this is not possible without giving the program some form of 'key'. The same way Dragon Naturally Speaking (a program that types what you speak into a microphone in realtime) requires users to read a 20 minute essay so the program can "train" on the idiosyncrasies of your voice, this typing-listener program would need something of the same sort. It would listen to the way you type certain keys, the delay between keys (timed and compared of course could reveal the distance your fingers travel to the keys). This is all compiled into a data set that is compared against your 'key' and then again against an english dictionary. Again, all in theory. The amount of samples this sort of neural network would require would be astronomical! And, then again, what would the applications of this program be? Surely, any programmer talented enough to create an A.I. system that can determine a user's typing through sound is talented enough to hack into the computer and the get document in the first place! *EDIT* "Keyboards are unique and sound different from each other. So no to both the ear and program... " Sure. The question was asked simply and implied that a person standing outside a room could possibly train themselves to hear what was being typed. I agree that is impossible. However, in the spirit of having fun with our minds, and using our imaginations, I concieved the previous hyopthetical. Saying that keyboards are unique is confusing, since I clearly described that the program would need 'training' or many samples. This assumes that you are doing all that is usually necessary of scientific experiments, be it have a good handle on your controls, and try to maintain good standards. Anyone knows that if you were working on this program and switched the keyboards it would basically be as if you started the experiment over again. In fact, I'm glad you pointed this out; the uniqueness of the keys is the only thing that makes this possible! If you had spilled a coke at some point and your A key was stuck, the program would quite easily (relative to the other keys) pick that one out from the rest. Compared against the documents you are using to train the program, it would take the LEAST amount of samples to pick out the MOST unique keys! Did you know that is how a touchtone phone works? With SOUND!

Not with the human ear alone.

I'm going to say that it isn't possible because even if you do figure out when I press my space bar (I guess I can give you that, since the space bar IS a much bigger key than other keys on the keyboard so it makes a different shound...), you are still left with all the letters of the alphabet, numbers, punctuation marks to contend with. If you disassemble a keyboard, you will see that while each key is encoded with a different signal so that the computer knows which keys you have pressed, the keys themselves are not programmed to sound differently like your touch tone phone. The ability to extract any data beyond the length of a message off the sounds of a keyboards clicking sounds seems pretty far fetched when you take into account that not all words are spelled correctly or even fully spelled out. Pauses in between typing can be the person thinking about what to type next (hey not all of us think in complete sentences lol). There are simply too many technical obstacles, the differences of sounds between one key and an adjacent key is so miniscule that their distinguishing qualities are lost in the ambient noise. If there were such "programs" I'd like to visit some of these links because I really doubt this can be done w/o some serious ultra sensitive listening hardware and some way to draw out the sounds of each keypress individually, even when they are pressed almost at the same time..

I at first would think the answer would be no, but in thinking about the question, there may be some ways in which it might be possible. The first might be the strangest. A blind person may be able to pick up the differences of the hand and finger movements and may be able to detect keys that are typed the most, like "A"s etc. Those keys are weaker and make a different sound, to say the letter Z. Some may be able to pick up on what is being typed, as there is a short break between words. I guess certain keys make the same sound, so if the "E" or the "T" are been hit. I think at the end of the day though, different people and different computers and keyboards would make it fairly hard to figure out what one was typing, but I'm sure the FBI has an answer!

By a human listening? No. By a computer listening? Definitely. Not so much the difference between the keys, but the clicking pattern itself could be analyzed such that the application could find out which words the user typed. In fact, as far as I remember, there are such applications.

Actually you can,,,I was able to relize most of the words that my brother was typing on the keyboard just by listening,,,it might work better with slow typers though

The answer is 'yes'. It can be done. It is much easier when monitoring a machine that produces a louder noise and whose keys are tied to a specific mechanical chain of events, right up to the letter striking the paper. I'm talking about a manual typewriter or teletype, for example. On a computer keyboard the task is more difficult, but still possible. You may not think the keys sound different, but they do - the sounds can be recorded, then analysed by computer to determine what keys were pressed. As another contributer stated, the solution is to prevent the sound from being heard outside an uncontrolled area (and allowing recording devices into such an area). When processing extremely sensitive data, this may be necessary.

That seems impossible to me. I don't know how people can do that unless they are psycic.