How the Computer Mouse Works - Computerphile

Used by millions every day we're dissecting mice today. Dr Steve Bagley takes us through the workings, starting with a ball-powered rodent.
/ computerphile
/ computer_phile
This video was filmed and edited by Sean Riley.
Computer Science at the University of Nottingham: bit.ly/nottsco...
Computerphile is a sister project to Brady Haran's Numberphile. More at www.bradyharan.com

[@MyAvitech]

Ahh Yes. The roller ball mouse. They worked great at sucking up any dust or hair on your desk, while conveniently storing it away for later disposal. It also had a feature where if you dropped the ball, it would seek out the dirtiest and most inaccessible corner of the room.

[@RMMLz]

That feeling when you'd open up the mouse and use a knife to clean the rods inside it... I miss it so much.

[@artiem5262]

Before we had mice, we had graphics tablets and light pens connected to computers. There is a fundamental difference between these older devices and the modern mouse. Those light pens and tablets were (for the most part) absolute position sensors, sending data to the computer on the position of the pen on the screen or on a purpose-built sensing surface. If you kept updating and tracking this position, you got motion. Modern computer mice are quite different -- they sense (and report) *relative* motion. The mouse moved left. It moved up. It moved left and down. Where? Not the problem the mouse needs to solve! Tracking that relative motion made so many things possible, like scrolling three feet to the side on a ten inch surface -- move the mouse to the side, pick it up, move it back, set it down, and move it some more -- relative motion lets you do that. So many other things fall out of going to relative motion. Calibration as to position and distance moved -- gone. Home position? Nope, put the initial mouse cursor where you want it in a window or on the screen, and track the relative motion. So much simpler! And our modern optical mice -- imagine going back in time to the early 1980's and sitting down with Englebart or Hovey, two (of many) mouse innovators, and telling them, Forget those quadrature wheels -- we're going to put a video camera and a processor inside the mouse. The video camera captures a magnified image of the desk surface many times per second, and feeds that image data to the processor. The processor does motion estimation, correlating successive video frames to derive motion between those frames, taking the output of the motion estimation algorithm and outputting that to the host the mouse is connected to as relative motion. Early 80's? They would have laughed! And I think they would have seen that yes, it would work -- but how large, how complex, how expensive in early 1980's technology would it be to build that system, a video camera feeding a processor doing motion estimation? PDP-11 sized? Not something that would fit in a mouse, run on very little power, and sell in huge quantities for $20 or under! Then tell them it will work better if you use a laser diode for illumination and interferometry techniques to do motion estimation. The technology, and the technological advances in some of these small and ubiquitous parts of our modern computers is amazing!

[@ForbinKid]

I had an Atari mouse problem on a weekend. Worked OK after cleaning and apart, but back on the desk no go. Bought a new (cheap) mouse and worked OK until the next weekend. Solution = "If your mouse doesn't work, close the house curtains". Yes, the sun shone through the plastic case for about 2 hours during the day.

[@brettbreet]

It was so satisfying to clean the mouse rollers and then have a *smoooth* and *renewed* device!

[@gordonrichardson2972]

A modern optical mouse has more processing power than a 1980 era desktop computer.

[@yonatanbeer3475]

This sounds iffy. Where do you get it from?

[@TheUglyGnome]

@@yonatanbeer3475 Simple: A desktop computer from '80s cannot process thousands of images/second.

[@tazogochitashvili6514]

@@yonatanbeer3475 I'd say that a high-end wireless one would probably be faster, as those have lights and "macro" software on them, as well as extra software to sync with the computer. It's not that it needs faster hardware, I'd just assume that whatever's the cheapest in that form-factor is probably just faster than whatever was around in the late 70s at least.

[@An.Individual]

this needs to be fact checked

[@tazogochitashvili6514]

@@An.Individual Totally, would be interesting to see, but I wouldn't be surprised if it was true.

[@Ziferten]

Oh man, I do NOT miss cleaning the hair and crap off those rollers.

[@famailiaanima]

Wym bro, just boil another egg!

[@mfx1]

I use a trackball so it's still a thing, not so bad though, as while there is gunk, there's no rollers anymore.

[@ELYESSS]

I miss removing and stealing them from school computer lab.

[@mfx1]

@@ELYESSS Our college glued the ball retaining ring in place to the only way to clean them was to completely dismantle them.

[@InventorZahran]

*picks up the mouse as if it were a microphone* "Computer? Hello, computer?" -Montgomery Scott, in Star Trek: The Voyage Home

[@haqvor]

Reminds me of the prank of rotating the SUN optical mouse pads so the pointer went in the opposite direction. Fun times! :)

[@erezra]

oooh, do an episode on Sprites! That's an awesome subject.

[@Gooberpatrol66]

I'm curious to know more about the particular algorithm optical mice use to compare images.

[@XxThePlaylistxX]

I wanna know more too. It sounds crazy that a mice has enough processing power in it to compare thousands of images and work out exactly how much and in exactly which direction a mouse has moved, while streaming it to the PC within a few milliseconds.

[@tilmanahr]

@@XxThePlaylistxX See my reply to Nathan’s comment. That sort of bit-wise comparisons of relatively small datasets is actually easy to do really quickly for DSPs.

[@tilmanahr]

@@0x90h I doubt it. At least in the early 2000s, neural networks and machine learning were relatively expensive and not very well established. The original patents cite bitwise comparisons between adjacent frames to work out the movement. That sort of thing is something simple DSPs can do very well, and very quickly. Remember, we’re talking 32x32 pixels or something. That isn’t exactly a whole lot of data, even by 2000 standards. It’s possible that some modern mice might use neural networks, but since simple algorithms running on dirt-cheap DSP chips do the job just fine, I doubt that is the usual solution.

[@NeatNit]

you really need to make a video describing how such comparisons are made. if I have two images, how do I measure their offset from one another?

[@KurtRichterCISSP]

Great vid! Just a suggestion: for the close-up work, have a camera on your desk facing downward towards a well illuminated workspace. Keep up the great work!

[@dewdop]

Lol

[@robintst]

Consequently, Nintendo adapted the mechanics of a ball mouse's optical encoders for the analog stick on the Nintendo 64 controller.

[@SeanBZA]

I still use an Intellimouse 1.1, does the job perfectly, and has been in use for a really long time.

[@MonochromeWench]

Mentions the third roller then tries to act if it doesn't exist. Could have just mentioned the third roller is on a spring and acts to push the ball against the other two rollers that are sensed.

[@ELYESSS]

He's a computer professor not a physics one XD

[@fabian999ification]

I've always wondered what the third roller was for.

[@Abrifq]

Ah yes, the hardboiled egg yolk in the mouse would be replaced every week. 😂😂😂 Memories

[@shanehebert396]

LOL... back in '91, I took my freshman Public Speaking class as a senior (someone said that it'd be a good idea to do that... 3 hour class, low burn, last semester before graduation to have five classes for 15 hours but not all five classes requiring a lot of work to complete). One of our speeches had to have props... most people used a poster that they drew. I had a poster that I drew (the night before it was due) that diagramed one of the sensors for a mechanical mouse (that's what people who had computers had back then) including the wheel. I also took the top off my Atari ST mouse (and cleaned it ;) ) and took it in to pass around so people could put their fingers on the rollers and move it and stuff. At the time, there were some Time Life books (I think they were) about "How Computers Work" or whatever and the TV commercial started out with a dude holding a mouse and saying "How does the mouse move the cursor?" rather dramatically. So, I started my speech that same way... I held up the mouse, had a similar look as the actor, leaned forward and said in what I thought was a TV voice, dramatically, "How does the mouse move the cursor?" and then said "Well... I'm going to tell you how it does that today." I handed the mouse to someone at the front of the class and asked them to play with it and then pass it on to the next person and then went into a description of how it all worked. Now... this wasn't entirely planned in advance. True to form, I had put it off due to work in other classes and literally the night before it was due, realized that it was due the next day and within about three or four minutes decided what I was going to do. I ran over to the student bookstore and bought the blank poster. Went back to my dorm room, drew the diagram in about, say, ten minutes after finding something that was circular and big enough to be seen well (I honestly think I used the bottom of a desk-side trash can), traced it all out, ran over it with a Sharpie to make it thick/dark. The speech itself? I totally adlibbed it on the fly while giving the speech in front of the class. I made an A ;)

[@klaxoncow]

And a happy X-mouse to you too!

[@patton72010]

"Inside the mouse you've got 2 rollers... well, 3 rollers but 2 rollers "

[@Computers0101]

2 encoder (sensing) rollers, and a tension roller that pushes the ball up against the encoder rollers to reduce slipping where the cursor doesn't move along with the mouse. so really three rollers but two rollers.

[@luv2stack]

Just what I needed to know before xmas

[@hamiltonianpathondodecahed5236]

Mice then: had huge rubbery b a l l s o f s t e e l Mice now: Disco light go brrrr

[@sayanghosh6996]

i saw the thumbnail and got excited and thought the video is about Xorg :(

[@HECKproductions]

always amazed at how simple but genius some of those old technologies were much more interesting than todays approach of having a billion sensors sending data to one huge IC that does all the work

[@Ovi0]

Instead of sending direction and velocity, why not send the change in x and y?

[@kc9scott]

If it periodically sends the changes in x and y at a fixed regular rate, that's the same as the velocity. If they're not at a fixed periodic rate, I'm sure each message gets timestamped in some way.

[@kc9scott]

And, I suspect that what's really sent IS the change in x and y.

[@FiNiTe_weeb]

@@kc9scott it is periodic, depends on polling rate, my G502 gets polled 1k times/sec, most office mice are 125 Hz afaik, and yea no point in sending velocity, it is almost certainly sending x and y deltas

[@flymypg]

The mouse that came with early Sun workstations used a pattered reflective mouse pad and simple optical sensors. It was spectacularly sensitive, with much higher XY resolution (close to 10x) than the competing mechanical ball or roller mice. The only hassle was that the mouse worked best when its orientation closely matched that of the pad, though significant rotation was tolerated. Still, it lead to a fair amount of mouse and pad rotation to get optimal performance. When I eventually had to give up my Sun workstation for a PC, I immediately found a PS/2 version of the exact same pad and mouse.

[@rewirestrike]

I enjoyed reminding myself of the hardware functions of the mouse. I am interested to learn more about the programming model between the operating system and the mouse. In particular, if there is any probability model that is used for a type of improvement for User Experience. Like a type of program that assists a user to navigate using the mouse.

[@drozcan]

Mouse only sends the movement delta at it's polling rate, positioning is done by the operating system and if you have a mouse with high polling rate, your pc utilizes more cpu to process.

[@VivekYadav-ds8oz]

Could you discuss the algorithm the mouse might use to calculate the shift between two images? Also, is an optical mouse powerful enough to capture and process this algorithm within itself or does it outsource it to the computer?

[@TheOzumat]

13:38 I loved my IntelliMouse Explorer!

[@pgrvloik]

Nice but I'd like to know how a mouse pointer is rendered too

[@anophelesnow3957]

I never knew. Steve Bagley, that was entertaining and informative. Thank you and merry Christmas. And thank you to the inventor of the ambidextrous wireless mouse.

[@simmo1024]

Perfect timing! I was wondering why the mouse on my soon-to-be-repaired Atari ST wasn't working properly and found it was in fact a Commodore mouse... Apparently, the switch flips the connections between pins 1 & 4. There is not one fitted inside the Commodore mouse, so I may have to fit one. But good to know!

[@pierreabbat6157]

What happens to the ball if the mouse is attacked by Cuterebra emasculator?

[@bhuwanbhandari6862]

Can we make a mouse which can work without placing on table? Can we design in such a way that it can just detect the non rigid surfaces like air?

[@prasanttwo281]

ooooh, I like the way you think, but I don't imagine it would be terribly comfortable to use I know there are some TV remotes that can do that in very precise ways using accelerometers (I think), so the technology definitely exists Phones have accelerometers too, so it could be interesting to write an app or something that can do that TV remote thing but with computers and a phone! I might try that some time

[@kewakl8891]

Glad that I don't need a ballmouse anymore...just like that ZIP Drive on the desk shelf.

[@DrSteveBagley]

There’s also a SyQuest EZ135 next to the ZIP drive (albeit without it’s case :-) )

[@shdon]

My very first mouse in 1988 was an optical mouse without a ball. Although the mouse was not branded as such, I think it was based on Mouse Systems Corporation's technology. It required a special mouse mat, which was made of reflective metal and had a very fine grid printed on it. After a few years of use, the grid was worn away in the centre of the pad and the mouse worked reliably only at the edges. For the first couple of years of using a PC, I never had to worry about cleaning any mouse rollers.

[@timseguine2]

As someone who sweats a lot, the first optical mice from Microsoft were a godsend. Might be exaggerating bad memories, but I feel like I had to clean the gunk off of rollerball mice every day for them to work right for me.

[@fep_ptcp883]

Roller ball mouse showed me the utility of a Bic pen lid

[@James100707]

How does the mouse know where the screen is ?

[@notsobob]

one of the biggest takeaways from this vid is that I now know how a video compression aglorithm works!

[@kartykeya7741]

Next video. Could u explain mechanical keyboard and modern keyboard

[@srn306x]

My cat keeps playing with my mouse... help.

[@canonpi]

😂

[@patrickmaurer2716]

I thunk it‘s broken. Cut it open and see if you can fix it ;)

[@Amonimus]

Wait, does that mean if the mouse is on completely mono-colored surface it won't work?

[@gordonrichardson2972]

It works on undulations (light and shadow) not colour.

[@kc9scott]

Light and shadow IS part of color. Amonimus is correct; when on that type of surface, it won't work.

[@RagHelen]

This time the video was shot in the arboreum.

[@YuanLiuTheDoc]

Pretty cool to know that the "mechanical" mouse had to use optical decoders. I had always assumed that they would use mechano-electric ones (i.e., tiny electric generators).

[@MegaKopfschmerzen]

Interesting, I thought they would use potentiometers.

[@Bellenchia]

The first few seconds are so british

[@pete3897]

Interesting fact - PIR sensors work in a similar way to the modern optical mouse with the image/frame comparison used to detect movement through a low resolution CCD

[@ecdhe]

Nice vintage Atari ST on the desk

[@steppenhenge]

trackball ftw

[@PiddeBas]

What about laser mice?

[@everythingisinfrared6029]

How much I would like to just receive the video frames from the optical mouse camera as images...

[@davidwillmore]

There are some mouse chips which allow that, but it is a bit of a hack as you can imagine.

[@ecdhe]

The first mice (made by Xerox) were actually optical. They were however very expensive. One of the things that Apple did for its Lisa and Mac in the early '80s is to find a company that could come up with the rollerball design - it required to clean up the mouse but was as reliable and *much* cheaper.

[@guilherme5094]

Nice!

[@ashwinalagiri-rajan1180]

what's the font in computerphile intro

[@sid5468]

whole comment section is underrated.

[@tayzzed]

Loving the small Christmas tree 🎄

[@billkendrick1]

Oh also, Atari's "Driving Controller", only officially used for Atari VCS/2600 "Indy 500" top-down racing game, used a similar technique to roller mice, but only in one dimension. They looked like paddle controllers (which were analog and used a potentiometer), but they were not restricted, and could spin in either direction forever. They sent a digital signal which needed to be decoded similar to mouse input, a few years later.

[@ikkombikom]

please allow auto-generated English subtitles. it would help a lot for non-native English viewers like me :)

[@stivassc]

I think youtube removed that feature recently :(

[@adrianpopagh]

Subtitles help me watch videos on mute while putting the kids to sleep...

[@ThePhragmites]

interesting. how about tablets like the Wacom, which I use for my photography editing work, or the surface pen?

[@UltraGaivalas]

OK, but how the mouse wheel work? When you scroll some content up & down, the mouse pointer remains in the same place - how?

[@grn1]

I don't think early mice had a mouse wheel or extra buttons (probably just right and left click) so they only needed 3 bytes. Modern mice send a lot more data and in an extendable format (so the mouse wheel if of course a different byte). Mechanically most mice wheels work very similarly to the trackballs but even simpler since they only need to turn one wheel instead of two. Realistically you don't even have to connect it to a seperate mechanical wheel you can just have the spokes (the part that connects the center of the wheel to the outside) use the stepped (or lightning shaped) design. Mine has textured dots on it so it may just read though dots (though I doubt it since there is a very distinct step to the movement). My mouse also has the ability to be pressed down (which is fairly common now and just requires a spring sensor underneath and someway either by hardware or software to tell it not to read other wheel movements). My mouse wheel can also be moved side to side (which is uncommon but probably not too difficult to implement).

[@jonathanhusni9484]

im looking for an optical sensor but with the microprocessor that decodes the signal integrated, does anyone know a module with this specifications?

[@vaidhyanathansm7625]

I love all your videos, but I feel difficult to comprehend because of the varied accents used by different speakers. It would be great if you could add subtitles in the video! Just a suggestion!

[@JogBird]

omg is that an Atari Mega ST ???

[@DrSteveBagley]

Just an STe

[@JeffBourke]

The first optical mice were terrible. You would play Counter Strike and try to do a 180° turn and you would just stand there and die.

[@whiterottenrabbit]

What about the Foodskey channel? Care to finally tell the 20k+ subscribers there that you abandoned Foodskey in favour of more profitable channels like Numberphile or Computerphile?

[@Reziac]

Wait.. my early (2004ish) optical mice preferred glass over everything else. At the time they were A4Tech-branded mice. In fact I used a glass mousepad right up until my last A4Tech mouse died. And since then I've seen other optical mice that prefer a slick metal or enamel surface. What gives?

[@Xylos144]

It turns out that mice can actually track on glass decently provided that there is either enough grease/grime on the surface, or if the glass has a bunch of bubbles or micropits near the surface that provide enough contrast to be matched between images for tracking. As long as the images can find the matching parts between themselves, it can do a decent job of tracking, though it will still tend to not be quite as accurate as using some kind of high-contrast gaming surface. One thing you can do as a test is get some glass cleaner and really clean off your mousepad. See if the mouse struggles to track. If it struggles, then if you just keep moving the mouse around, it ought to dirty itself from the grime on the little feet/sliders on the bottom of the mouse and then start tracking well. And if you want to help it along, you can smear your hand over the surface and add some of you own grime to help it out.

[@Pyrazahn]

With all the nostalgia hype nowadays, I wonder if there are people who still prefer ball mice over optical ones. I know I definitely don't miss having to interrupt my work and clean the rollers every few days because one of them collected enough dirt again and stopped turning properly.

[@jlp1528]

Cool video but then how do laser mice work?

[@vylbird8014]

Exactly the same as LED mice - the use of monochromatic coherent light gives a clearer image. Although in practice, a lot of 'laser' mice are really just plain LED mice with a lie on the box.

[@jlp1528]

@@vylbird8014 yeah, unfortunately, but I think they've cracked down on that. Gaming mice over 50 bucks usually have actual lasers as far as I know.

[@zacherynuk842]

The first Microsoft Intellimouse was wonderful. As highly regarded as cherry keyboards. Don't make like they used to.

[@arewenot]

as a software developer, this is one of the questions i thought i never want to know.

[@AgentM124]

Why didn't the ball mice use an electromotor/dynamo to generate a current instead of the optical signal?

[@vylbird8014]

Two reasons. Firstly, it wouldn't work at all at very low rates of movement - the current generated by slow movement would be too low to detect with affordable components. And secondly, it would require more complicated electronics to process the analog signal. The optical wheels produce a digital signal that can go straight into the logic circuits.

[@UberAlphaSirus]

Because it would be awful. Grey encoders apart from being good, they impart minor mechanical drag.

[@AgentM124]

@@vylbird8014 ah makes sense! Thanks for the explanation.

[@stevojohn]

Tear a bit off a post-it and put over a colleague's mouse sensor.

[@HShango]

I prefer a mouse (wireless mouse specifically, rechargeable versions) over a trackpad (trackpads are the worst).

[@tyrgoossens]

Trackballs forever.

[@TheUglyGnome]

I rather use pen tablet. So much better than any mouse.

[@thinboxdictator6720]

I like wired mouse over wireless. I think I understand why you would like wireless tho. For notebook,it does seem better to have wireless. On my PC , I'm using for some time now roccat nyth and I can't say anything against it.

[@joshjohnson8459]

You could have given a much better context surrounding the transmission.

[@davidgustavsson4000]

Mice were always a gimmick. It's great for games and graphic design, but to think of all the developer time that has gone into needlessly optimizing software for mouse operation, and all the user time wasted chasing down buttons and menus in programs that could have been keyboard based.

[@MintyLime703]

lol a "gimmick" Easily the most common modern method of navigating most computer environments, even moreso than the keyboard, and yet it's a "gimmick".

[@AcornElectron]

Well it’s been a weird year. Might as well click that scary af thumbnail from computerphile....... .

[@TechXSoftware]

You need a mouse with a 80,000 pulling rate!

[@simontay4851]

I find optical mice work best on a black surface.

[@okaro6595]

Logitech mouses do not have any visible light. They use infrared.

[@413x1s]

foo

[@SteveGouldinSpain]

Mice are for wimps! Famously, Bill Gates once took a side-swipe at Apple by saying "real men don't use mice". That was obviously before he thought better of it and came up with Microsoft Windows!

[@RelytM17]

this dude said "direction and velocity"... needs to go retake some entry level physics again lol

[@M2AProductions]

I think you may need to lol

[@RelytM17]

@@M2AProductions please explain. I'll wait KZbin professor

[@harishganesan3575]

Wait.. so an optical mouse uses image processing to detect movement !? Wth ! I am expecting future mouses to use machine learning to learn future moves now.

[@Pablud3S]

What? Why?

[@harishganesan3575]

@@Pablud3S i was almost joking. I was commenting about the computation it is doing for something like a mouse.

[@pushkarsoni8927]

Second

[@nejatulusal1475]

Hello

[@nejatulusal1475]

And first by the way

[@nicklaskaridis]

@@nejatulusal1475 actually correct

[@nejatulusal1475]

@@nicklaskaridis Its the first real first comment which says first

[@PrasadIndi]

Congratulations on this achievement.

[@spiliosmaderakis4494]

7th i guess

[@thePronto]

Reminded me while I like software, not hardware.

[@doge6824]

Sugondese

[@lumix99]

Dude this guy talking too fast :( My native language is not English so I cant understand him and why this video don't has auto subtitles :((((((((((((((((((((((((((

[@CarlosAlmeida1972]

I love your videos, so I'm asking you to please upload on alternative platforms. I'm deleting my KZbin account, because of the free speech restrictions. We should fight and defend free speech! I will miss your videos. Sorry for my unsubscribe!!!

[@HebaruSan]

I still need to clean my mouse wheel.

[@davidgillies620]

Those Sun optical mouse pads were a damn nuisance. They tended to wander off and you couldn't find one when you needed it (and they were surprisingly costly). At the university where I worked in the mid to late 90s they were glued down in shared computer labs. I was glad to see the back of them.

[@UberAlphaSirus]

Does it make me a nerd that I knew this already, I thought everyone knew this. Our school glued the ball in as all the kids used them in their catapults.

[@UberAlphaSirus]

​@spindletea Watch the video, how did they get the ball in or out, how was it rolled about?

[@UberAlphaSirus]

@spindletea Refer to the original post. I am not here to teach.

[@YTANDY100]

@spindletea he means glued the cover so you cant take out the ball :-)

[@alibulus4382]

please use automatic cc. i lilke your videous but i can only understand with reading cc :(

[@Computerphile]

We have an open ticket with KZbin support about this - the subtitles are enabled but not showing - very sorry, hope to resolve soon >Sean

[@latchdeadbolt]

Ah yes. I recall those times when clearing the ball fluff was a routine part of the work week.

[@DanBowkley]

What's the advantage of using direction and velocity vs just very quickly sending a string of up/down/left/right blips? I'd think that even with ancient 1200 baud serial it would still be plenty fast enough to work well.

[@okaro6595]

Why is that people resort to 80s terminology when talking about 80s? You would not use bauds to refer to USB speeds. Baud means signal changes per second. In the 80s it was widely misused to mean bits/second when talking about modems.

[@DanBowkley]

@@okaro6595 I was using baud to refer to the ancient serial technology of the last millennium.

[@ZipplyZane]

Would 1200 baud be enough when you have to send everything in discrete chunks instead of a continuous feed? That's what I assumed the problem was fro what he said--the serial nature of the connection.

[@neilbrookins8428]

I still have a ball mouse cleaner. It’s a ball covered in Velcro hooks and comes with isopropyl alcohol to put on it Then there is a pad with Velcro felt and you roll it around.

[@root42]

At Dr Bagley: you did not tell us why the Atari mouse went into only one direction horizontally! Please enlighten us!

[@gogl0l386]

Vim/evil mode with tiling window manager gang: eww wtf is that rodent?

[@cheaterman49]

:-) people be like "eww you use touchpad?" me be like "barely using a pointer at all"

[@aspirohk3558]

What about the software part? Like how does it(pointer) know it's above a file? Or start button?