Empirical proof: I use an iGesture Numpad, a MultiTouch tablet with a virtual numeric keypad. Works great as a replacement for a mouse with tons of functions no mouse could offer. But I can't enter numeric data worth a damn on it, even though I normally have decent 10-key data entry speed.
Well...a real test would be for you to type on a flat surface and measure errors. I suppose when I build a multi touch surface I can conduct that experiment.
Well the closest thing we have now to typing on a flat surface is the Cellulon Laser Projection Keyboard. Anyone could go out and buy one of these and see if it’s any easier or more accurate to type with than a traditional keyboard.
On this whole tactile feedback debate, I think that it is still necessary not because we would make lots of mistakes, but because our fingers would get tired. Try tapping your finger on the table for five minutes. It starts to get very tiresome very quickly. Without the travel of traditional keyboard keys, we would end up in situation where technology prevails over ergonomic considerations.
On the touch typing issue, well, I’m not sure that it’s the feel of the keys that lets you know you’re on the home row, or you’ve pressed the right key, I think it’s a combination of momentarily glancing down at the keyboard when you rest your hands, and, especially on a notebook, peripheral vision. Right now typing on my MacBook, I can see the keyboard at the bottom of my field of vision. I can’t read the letters, but it doesn’t matter because I recognise and know the shapes of the keys and the position of the letters in relation to the edge of the keymap.
Obviously, this may not apply when using a desktop Mac, because the keyboard and screen are at right angles to each other. That said, I find it harder to touch type when my MacBook’s plugged into my Pro Keyboard, is this because I can't see the keys in my peripheral vision, or is it because the keyboard is designed differently (heavier keys, greater travel and so on)? Perhaps it’s a combination of factors.
All I know is until someone invents a totally new input method that obsolesces the keyboard (perhaps handwriting recognition will one day become accurate and usable), I want one with traditional plastic moving keys that gives me tactile feedback and the satisfying feeling that I’ve pressed a key.
Think of it this way: pushing buttons is more fun than tapping a surface.
You do realize that most people who don't automatically know where the home position is without looking or feeling around will have to take a second to figure out where it is, and then from there have everything memorized... Well, then the same holds true for a multitouch keyboard.
The point is folks can find the home position by feeling around. Arguably you can say that you can always look down to find home. This does NOT help you self correct when you are drifting off center.
Quote:
Note that I am not talking about multitouch keyboards as though I think they are necessary secondary input devices. Wherever possible, I much prefer the idea of direct imitation of physical object manipulation as in the Jeff Han demonstrations of multitouch.
They are secondary input devices WHEN you are manipulating objects. They are PRIMARY input devices when you need to write a document. I am a big proponent of multitouch interfaces but you need to also accept the limitations. No technology is a panacea and even a paradigm shift would not eliminate the need to type large amounts of text in the near to mid-term future.
Quote:
3. While multitouch keyboards aren't the optimal input device, they are a type of input device that will be desired for some time until better solutions arise.
And physical keyboards will also be desired for some time and multitouch keyboards only used when limited typing is required.
Quote:
4. Multiple dimensions of feedback possible with multitouch greatly offset the loss of tactile feedback that some claim to be the issue.
A multitouch display does NOT preclude the use of a physical keyboard.
Quote:
5. Newer technologies such as speech recognition, speech to text, and tactile simulation on touchscreen elements will further offset such concerns in the near future.
Speech recognition is only BARELY closer to usable in the last 20 years (seriously, I started as an intern at a HCI lab and speech recognition is what we were playing with in the 80s) and has its own limitations. Tactile simulation would be wonderful but one that is either transparent or a display seems to be quite a ways off. NEITHER are near future.
Well the closest thing we have now to typing on a flat surface is the Cellulon Laser Projection Keyboard. Anyone could go out and buy one of these and see if it?s any easier or more accurate to type with than a traditional keyboard.
I will likely be building a MT interface later this year. One of our experiments could be error rates. The problem will be sample size.
Quote:
On this whole tactile feedback debate, I think that it is still necessary not because we would make lots of mistakes, but because our fingers would get tired. Try tapping your finger on the table for five minutes. It starts to get very tiresome very quickly. Without the travel of traditional keyboard keys, we would end up in situation where technology prevails over ergonomic considerations.
This I don't think will be as much an issue if we start tapping lighter.
Quote:
All I know is until someone invents a totally new input method that obsolesces the keyboard (perhaps handwriting recognition will one day become accurate and usable), I want one with traditional plastic moving keys that gives me tactile feedback and the satisfying feeling that I?ve pressed a key.
Think of it this way: pushing buttons is more fun than tapping a surface.
Heh...yes, I wouldn't worry...the keyboard will be around until working voice recognition happens and even then there are still issues in an office environment in terms of noise levels.
Dragon is okay but I wouldn't want to use it given I can type reasonably well. My office mate uses it to create memos and I have a good idea of his speed and accuracy.
As for the necessity of a 'tactile' feel, it could be replicated by simply bubbling the MT overlays. Certainly, the technology has the potential of showing on screen if our fingers are in the right postion. More importantly, I suspect that there will be a multitude of third-party solutions, virtual keyboard designs, and why not the ability to design your own.
Yes, and I've been trying to figure out a "bubbling" mechanism for a short bit. Can't be reasonably be done with FITR which we've been playing with. Doable with a capacitance array or membrane for which we don't have a multi-touch surface.
I went ahead and googled for a study and cornell has done one (there are probably more but this seemed recent and I'm at home and can't do a real lit search). With a multi-touch membrane keyboard it had lower average WPM and lower accuracy. Learning effects were linear.
Subject size is small but given there are only two dependent variables its within the mimimal set for statistically significant experiment. I'm not the statistician in my group but I believe this to be the case off the cuff.
Thanks for the reference to the Cornell study. I am a statstician and I can say that the study has too little statistical power to draw any reasonable conclusions. There was a "hint" (not significant) of improvement in speed and accracy with increased usage in the MT part of the study. Previous studies from the same lab suggested that speed and accuracy increases (up to 63%) are observed with increased usage of MT devices. These observations were anecdotal and not in any way statistically significant.
One of the more interesting points gleaned from the reference is that Dr. Alan Hedge (the authors' thesis advisor) published on MT with FingerWorks founders Wayne Westermann and John Elias. Also, further Googling of Hedge and the author indicate that the Cornell research seems to have ended shortly after that study coinciding with the acquistion/demise of FW by Apple.
Finally, for those interested in the FW/Westermann/Elias saga, where they are, do/did they work for Apple, etc. check out this old link
Thanks for the reference to the Cornell study. I am a statstician and I can say that the study has too little statistical power to draw any reasonable conclusions.
Hmmm...many of the HSI/HCI experiments I've read has sample sizes in the same range. We had 32 subjects for 4 independent variables which required 8 person cells x 3 cells for a ANOVA analysis (hmm...just said analysis twice...whatever). The number of participants required is based on the effect size you're trying to measure if I understand correctly.
Did the authors calculate the statistical power of the study to reject the null hypothesis (NH) prior to their studies?Did they state the confidence level at which the NH was being tested? If the study does not state what the NH is and what the power of the study is, then I would be suspicious of any conclusions drawn. Small sample size studies are notorious for drawing incorrect statistical inferences.
To further clarify my point, you might set up a study whose null hypothesis (NH) is that " Typing on a standard keyboard is faster than on an MT keyboard". With a small sample size, the data might fail to reject the NH but only at a very low confidence level. Stated differently, you might be able to say at the end of the study that "I am 20% confident that typing on a standard keyboard is faster than on an MT keyboard".
Did the authors calculate the statistical power of the study to reject the null hypothesis (NH) prior to their studies?Did they state the confidence level at which the NH was being tested? If the study does not state what the NH is and what the power of the study is, then I would be suspicious of any conclusions drawn. Small sample size studies are notorious for drawing incorrect statistical inferences.
Heh...I've seen it as much as you have. Its a powerpoint presentation of an experiment that looks like might be for a thesis. I did not see statistical power in my brief glance (I jumped to the conclusion slide ) but presumably this will appear in the paper or thesis or my opinion of cornell will have dropped quite a few points.
Yes, small sample sizes are sub-optimal sometimes leading to incorrect conclusions. This is traded against the value of the research, the budget and rigor. Our paper was for a conference with all the associated implications (ie not that rigorous, low budget, etc). Even so, I'd guess the study cost us on the order of $40K to run for a mere 32 subjects. HCI studies can be manpower intensive. We had 3 observers for every 2 sets of subjects. Each session was a couple hours and we had gaps between subjects. Add lab time, analysis time, experiment design time...
But hey...its data that supports my position so I think its grand. So what that its only marginally better than gut instinct? Look at the pretty charts!
One more point. Power is also directly related to the size of the real effect. So, perhaps in your studies, you had a large real effect (difference in what your measuring between groups) and despite a low sample size, the power (the probability that the results of the study will allow rejection of the NH if the independent variable has a real effect) of your study was "high". Your statistician can calculate all of this for you. The best practice is to design your study beforehand by asking the question--how many subjects do I need if I see a real effect of "magnitude x" to get a power (probability) of say greater than or equal to 0.70? 0.80? 0.90?
I do a lot of work connected with clinical trials of new drugs in humans where it is absolutely necessary to calculate the power of any study beforehand so that you can prove to the FDA and the scientific community that a drug really does work.
I think sometimes people make serious comments without thinking them through. Why would Apple make a touch screen iMac? Why? Think through that question and try and prove to me that they will. As I said on various other sites where similar stories have arisen, Apple may make multi-touch tablets and touch panels for retail, but I still think the way forward for the home consumer Mac's is multi-touch touch screen keyboards that replace both existing keyboard and mouse in one unit. The dock could replace the F-keys if you liked, the on screen cursor would move if one dragged one finger across the display. The keboard could do QWERTY or DVORAK or WHATEVER YOU WANT, and could display any message like you have 1 new Mail etc. It's the way forward I'm sure of it.
Perhaps a thin, flexible, wireless and easy to clean and store touch pad that behaves as a liaison to your touch screen and which encompasses vastly more than the the power to perform the duties of a traditional keyboard?
As your fingers approach the touch pad, their movement is visible, if you choose, on screen (yes, even your TV screen ...), for those that need visual markers for their fingers?
Manipulate what you see on your screen without actually having to lean over to reach out to it, because the touch pad is naturally where your fingers are accustomed to reaching to - comfortably right in front of you, wherever you may choose to be?
Comments
Well...a real test would be for you to type on a flat surface and measure errors. I suppose when I build a multi touch surface I can conduct that experiment.
Well the closest thing we have now to typing on a flat surface is the Cellulon Laser Projection Keyboard. Anyone could go out and buy one of these and see if it’s any easier or more accurate to type with than a traditional keyboard.
On this whole tactile feedback debate, I think that it is still necessary not because we would make lots of mistakes, but because our fingers would get tired. Try tapping your finger on the table for five minutes. It starts to get very tiresome very quickly. Without the travel of traditional keyboard keys, we would end up in situation where technology prevails over ergonomic considerations.
On the touch typing issue, well, I’m not sure that it’s the feel of the keys that lets you know you’re on the home row, or you’ve pressed the right key, I think it’s a combination of momentarily glancing down at the keyboard when you rest your hands, and, especially on a notebook, peripheral vision. Right now typing on my MacBook, I can see the keyboard at the bottom of my field of vision. I can’t read the letters, but it doesn’t matter because I recognise and know the shapes of the keys and the position of the letters in relation to the edge of the keymap.
Obviously, this may not apply when using a desktop Mac, because the keyboard and screen are at right angles to each other. That said, I find it harder to touch type when my MacBook’s plugged into my Pro Keyboard, is this because I can't see the keys in my peripheral vision, or is it because the keyboard is designed differently (heavier keys, greater travel and so on)? Perhaps it’s a combination of factors.
All I know is until someone invents a totally new input method that obsolesces the keyboard (perhaps handwriting recognition will one day become accurate and usable), I want one with traditional plastic moving keys that gives me tactile feedback and the satisfying feeling that I’ve pressed a key.
Think of it this way: pushing buttons is more fun than tapping a surface.
You do realize that most people who don't automatically know where the home position is without looking or feeling around will have to take a second to figure out where it is, and then from there have everything memorized... Well, then the same holds true for a multitouch keyboard.
The point is folks can find the home position by feeling around. Arguably you can say that you can always look down to find home. This does NOT help you self correct when you are drifting off center.
Note that I am not talking about multitouch keyboards as though I think they are necessary secondary input devices. Wherever possible, I much prefer the idea of direct imitation of physical object manipulation as in the Jeff Han demonstrations of multitouch.
They are secondary input devices WHEN you are manipulating objects. They are PRIMARY input devices when you need to write a document. I am a big proponent of multitouch interfaces but you need to also accept the limitations. No technology is a panacea and even a paradigm shift would not eliminate the need to type large amounts of text in the near to mid-term future.
3. While multitouch keyboards aren't the optimal input device, they are a type of input device that will be desired for some time until better solutions arise.
And physical keyboards will also be desired for some time and multitouch keyboards only used when limited typing is required.
4. Multiple dimensions of feedback possible with multitouch greatly offset the loss of tactile feedback that some claim to be the issue.
A multitouch display does NOT preclude the use of a physical keyboard.
5. Newer technologies such as speech recognition, speech to text, and tactile simulation on touchscreen elements will further offset such concerns in the near future.
Speech recognition is only BARELY closer to usable in the last 20 years (seriously, I started as an intern at a HCI lab and speech recognition is what we were playing with in the 80s) and has its own limitations. Tactile simulation would be wonderful but one that is either transparent or a display seems to be quite a ways off. NEITHER are near future.
Vinea
Well the closest thing we have now to typing on a flat surface is the Cellulon Laser Projection Keyboard. Anyone could go out and buy one of these and see if it?s any easier or more accurate to type with than a traditional keyboard.
I will likely be building a MT interface later this year. One of our experiments could be error rates. The problem will be sample size.
On this whole tactile feedback debate, I think that it is still necessary not because we would make lots of mistakes, but because our fingers would get tired. Try tapping your finger on the table for five minutes. It starts to get very tiresome very quickly. Without the travel of traditional keyboard keys, we would end up in situation where technology prevails over ergonomic considerations.
This I don't think will be as much an issue if we start tapping lighter.
All I know is until someone invents a totally new input method that obsolesces the keyboard (perhaps handwriting recognition will one day become accurate and usable), I want one with traditional plastic moving keys that gives me tactile feedback and the satisfying feeling that I?ve pressed a key.
Think of it this way: pushing buttons is more fun than tapping a surface.
Heh...yes, I wouldn't worry...the keyboard will be around until working voice recognition happens and even then there are still issues in an office environment in terms of noise levels.
Dragon is okay but I wouldn't want to use it given I can type reasonably well. My office mate uses it to create memos and I have a good idea of his speed and accuracy.
Vinea
As for the necessity of a 'tactile' feel, it could be replicated by simply bubbling the MT overlays. Certainly, the technology has the potential of showing on screen if our fingers are in the right postion. More importantly, I suspect that there will be a multitude of third-party solutions, virtual keyboard designs, and why not the ability to design your own.
Yes, and I've been trying to figure out a "bubbling" mechanism for a short bit. Can't be reasonably be done with FITR which we've been playing with. Doable with a capacitance array or membrane for which we don't have a multi-touch surface.
I went ahead and googled for a study and cornell has done one (there are probably more but this seemed recent and I'm at home and can't do a real lit search). With a multi-touch membrane keyboard it had lower average WPM and lower accuracy. Learning effects were linear.
http://ergo.human.cornell.edu/Grads/JTSdefense1.pdf
Subject size is small but given there are only two dependent variables its within the mimimal set for statistically significant experiment. I'm not the statistician in my group but I believe this to be the case off the cuff.
But the results support my assertions.
Vinea
Thanks for the reference to the Cornell study. I am a statstician and I can say that the study has too little statistical power to draw any reasonable conclusions. There was a "hint" (not significant) of improvement in speed and accracy with increased usage in the MT part of the study. Previous studies from the same lab suggested that speed and accuracy increases (up to 63%) are observed with increased usage of MT devices. These observations were anecdotal and not in any way statistically significant.
One of the more interesting points gleaned from the reference is that Dr. Alan Hedge (the authors' thesis advisor) published on MT with FingerWorks founders Wayne Westermann and John Elias. Also, further Googling of Hedge and the author indicate that the Cornell research seems to have ended shortly after that study coinciding with the acquistion/demise of FW by Apple.
Finally, for those interested in the FW/Westermann/Elias saga, where they are, do/did they work for Apple, etc. check out this old link
http://fingerfans.dreamhosters.com/f...opic.php?t=223
Vinea,
Thanks for the reference to the Cornell study. I am a statstician and I can say that the study has too little statistical power to draw any reasonable conclusions.
Hmmm...many of the HSI/HCI experiments I've read has sample sizes in the same range. We had 32 subjects for 4 independent variables which required 8 person cells x 3 cells for a ANOVA analysis (hmm...just said analysis twice...whatever). The number of participants required is based on the effect size you're trying to measure if I understand correctly.
Vinea
To further clarify my point, you might set up a study whose null hypothesis (NH) is that " Typing on a standard keyboard is faster than on an MT keyboard". With a small sample size, the data might fail to reject the NH but only at a very low confidence level. Stated differently, you might be able to say at the end of the study that "I am 20% confident that typing on a standard keyboard is faster than on an MT keyboard".
Did the authors calculate the statistical power of the study to reject the null hypothesis (NH) prior to their studies?Did they state the confidence level at which the NH was being tested? If the study does not state what the NH is and what the power of the study is, then I would be suspicious of any conclusions drawn. Small sample size studies are notorious for drawing incorrect statistical inferences.
Heh...I've seen it as much as you have. Its a powerpoint presentation of an experiment that looks like might be for a thesis. I did not see statistical power in my brief glance (I jumped to the conclusion slide
Yes, small sample sizes are sub-optimal sometimes leading to incorrect conclusions. This is traded against the value of the research, the budget and rigor. Our paper was for a conference with all the associated implications (ie not that rigorous, low budget, etc). Even so, I'd guess the study cost us on the order of $40K to run for a mere 32 subjects. HCI studies can be manpower intensive. We had 3 observers for every 2 sets of subjects. Each session was a couple hours and we had gaps between subjects. Add lab time, analysis time, experiment design time...
But hey...its data that supports my position so I think its grand.
Vinea
One more point. Power is also directly related to the size of the real effect. So, perhaps in your studies, you had a large real effect (difference in what your measuring between groups) and despite a low sample size, the power (the probability that the results of the study will allow rejection of the NH if the independent variable has a real effect) of your study was "high". Your statistician can calculate all of this for you. The best practice is to design your study beforehand by asking the question--how many subjects do I need if I see a real effect of "magnitude x" to get a power (probability) of say greater than or equal to 0.70? 0.80? 0.90?
I do a lot of work connected with clinical trials of new drugs in humans where it is absolutely necessary to calculate the power of any study beforehand so that you can prove to the FDA and the scientific community that a drug really does work.
Hope this is helpful.
I think sometimes people make serious comments without thinking them through. Why would Apple make a touch screen iMac? Why? Think through that question and try and prove to me that they will. As I said on various other sites where similar stories have arisen, Apple may make multi-touch tablets and touch panels for retail, but I still think the way forward for the home consumer Mac's is multi-touch touch screen keyboards that replace both existing keyboard and mouse in one unit. The dock could replace the F-keys if you liked, the on screen cursor would move if one dragged one finger across the display. The keboard could do QWERTY or DVORAK or WHATEVER YOU WANT, and could display any message like you have 1 new Mail etc. It's the way forward I'm sure of it.
Perhaps a thin, flexible, wireless and easy to clean and store touch pad that behaves as a liaison to your touch screen and which encompasses vastly more than the the power to perform the duties of a traditional keyboard?
As your fingers approach the touch pad, their movement is visible, if you choose, on screen (yes, even your TV screen ...), for those that need visual markers for their fingers?
Manipulate what you see on your screen without actually having to lean over to reach out to it, because the touch pad is naturally where your fingers are accustomed to reaching to - comfortably right in front of you, wherever you may choose to be?
Would you like that?