I am mixed on motion controls, but this is almost entirely based upon my personal philosophy regarding controls in general rather than regarding motion controls in general. As such, I'm going to outline that philosophy before I get into why I like some motion controls and dislike other motion controls.
Controls are, at their base, a method of player input in order to interact with the game. The key term here is "input" - a control scheme and a game's use thereof should always involve player input into the action being performed. A button press/motion (or a combination thereof) should always correspond to a choice being made by the player - a choice made to use a certain attack, a choice made to use a certain item, a choice made to move in a certain direction, a choice made to speak to a given NPC, and so on. The control scheme is the interface that the player uses to make his choices within the game. Now, the goal of most games - I feel - should be to reduce that interface, to minimize it. A game wants the player to make a choice within the game world (the choice to swing their sword) rather than a choice within the real world (the choice to press the B button). The larger and more convoluted that interface becomes, the greater the gap between the choice within the game world and the choice within the real world becomes. Fighting games, a genre I am not a fan of, are particularly bad at this: with the incredibly complex combo system that many of them feature, the game feels like a system of button presses rather than a game in which one is actually fighting. The combo list is too intimidating, too memorization-based, and as such the decision in the player's mind becomes less one of what attack to use and more one of "okay, so the combo is Z-A-Down, right?" That kills the experience of the game.
Thus the best control schemes are those that will reduce the distance between the player and the game: the ones that, rather than forming a barrier between the player and the game, bring them closer. The best control scheme, as I have said in the past, is one that is "invisible": it's so natural that players aren't consciously thinking of their button presses or motions, they're thinking of the actions they're performing. This control scheme is ideal, but not practical. Almost every game is going to have an even partially visible control scheme, if not an obstructive one. The two best examples I can think of are Super Meat Boy and Journey. In the case of the former, the controls are incredibly smooth and tight, and the game itself so brutally punishing: any split second decision being made will result in death, so the player is quickly trained in such a way that the controller almost vanishes, and the game becomes about planning a path rather than honing reaction time and button presses. In the case of the latter, movement is such a large part of the game, and movement is perhaps the most natural feeling aspect of any control scheme; where Journey earns its control scheme credit is in the smooth and fluid motion of its camera, controlled either by the right stick or by the tilt of the Dualshock. It's a very natural system of controlling, and works very well. Both games, however, have their shortcoming. The time it takes to learn reaction time and the nuance of the control in Super Meat Boy is noted, and is a brief barrier between the player and the game, if one that erodes over time. The hover and chime buttons (X and O respectively) in Journey are much less natural than the movement and camera options, and while neither has the real sense of being a barrier, they are nonetheless, by virtue of being held as the action progresses, more visible.
So in summation: a control scheme should be invisible, self-evident, and unobstructive, as to reduce the distance between the player and the game and amplify the illusion that the player is performing the actions on screen. This kind of control scheme is ideal, if unreachable.
It may sound like motion controls, by virtue of the emulation of the action being an inherent part of performing the action in game, are a perfect candidate for my ideal control scheme. However, I would argue that most motion controls are not. There are a variety that can help a control scheme approach the ideal scheme, but there are also varieties that can remove it further, and create an additional barrier between the player and the game. In particular, there are three varieties that I see often repeated and would like to discuss. I'll categorize them as good, mixed, and bad.
I've already mentioned an example of the good variety: the camera movement in Journey. This is in the good variety of motion controls precisely because they are unobtrusive. The controller, a standard button-based controller, does not have to be reoriented in order to use the motion control: it has to be tilted, which is the motion necessary to move the camera. During the tilt, access to the buttons continues, allow seamless integration of camera control with other actions (movement, hover, and chime). Additionally, the physical exertion necessary to execute this control is negligible; it's an incredibly minor motion of the wrists. This type of motion control is one that does not create barriers because the action requires no setup, impairs no other functions, and requires minimal physical exertion.
The mixed variety is perhaps the most common, and can be expressed quite simply: aiming motion control. In virtually any game with motion control options that features shooting mechanics, motion control is used to aim the shooting device. These are mixed because there are good things about them, and bad. In the good column, there is the increased accuracy of aiming and the variable speed of aiming. I can move a Wii Remote, for instance, across the screen in a second, or in five. I can completely control the speed at which I adjust my aim, which is preferable to the more often fixed speed of using a control stick to move an aiming cursor. Aiming with a remote or a wand is usually much quicker and accurate, and as such helps to reduce the interface between the player and the game. Physical exertion is typically minor, though I wouldn't say minimal; for the best accuracy, it is usually best to move one's entire arm rather than just the wrist. It's not a strenuous motion, but it's more taxing than a simple wrist tilt, and raises player awareness of the controller. The bad part of this motion control variety is that it often involves re-positioning the wand or remote. Perhaps I am alone in this, but I tend to hold my controllers in a very relaxed fashion, so when an aiming segment would arise - for example, using the bow in Twilight Princess - I was often greeted with a "Point the Wii Remote at the screen" message. This not only involves forcing the player to move their controller into position before executing motion controls, but it also directly alerts them to the fact that "Hey, you are using a controller!" This is contrary to the stated goal of controls to be invisible, and thus is a bad thing. But the increased accuracy and speed of aiming is still a very noted upgrade over the cumbersome control stick aiming, and as such this variety of motion controls fits squarely in a mixed category.
Of mention are gyroscope controls, which rest somewhere in between good and mixed - while the motion required is usually more in line with the mixed variety, it rarely involves actually re-positioning the device. I can't place them confidently in either category, and they aren't widespread enough to merit their own, but it's worth noting.
Now, the bad variety. I know a lot of people really liked it, but I would say that Skyward Sword is the best example of the bad variety of motion controls, which I call "total motion". That's not the best moniker, as there are still a number of button-based controls in Skyward Sword, but I use it because it represents my ultimate gripe with the bad variety: that the fact that the game is motion controlled is constantly put into the spotlight. Let's talk first about the most significant aspect of the motion controls: swordplay. An attractive feature for many, swordplay in Skyward Sword is completely contrary to my ideal control scheme. A swing in one of the eight cardinal directions (as well as a thrust or spin attack) requires a very deliberate action. The offending term here is "deliberate": because the player has to be very precise and decisive with their swings, the mental decision becomes not one of swinging a sword, but one of swinging a Wii Remote. The distinction may seem negligible, but because the motion control must be precise to avoid a poor swing that registers in a way that does not reflect the mental decision (this false registration of the swing being perhaps the worst possible way a control scheme can create a barrier) the player is aware the entire time that they are swinging a Wii Remote. The act of controlling is both physically and mentally active, rather than physically active but mentally passive. No degree of control acuity can remove that.
Skyward Sword also commits a motion control sin that I term "forced input". Basically, forced input involves the player executing an action that is required to proceed and initiated for them, rather than the player choosing to initiate and execute an action that allows them to advance. As I said at the beginning of this post, controls are inputs, and each control action should correspond to a mental decision. But being forced by the game to both approach a sword and then slowly pull it out of its pedestal is not a mental decision; it's a guided cutscene that is based not in reaction or reflex, but in simple tedium. It's an unnecessary act, but one the game forces you to perform regardless. Skyward Sword is littered with these sorts of moments, and I for the life of me cannot determine why these moments exist in the game, except to call attention to the motion controls, which is contrary to the ideal control scheme.
This is not to say that all of Skyward Sword's control scheme is bad - far from it, it's a very functional and acceptable control scheme. But it is far from ideal, as its implementation of motion controls creates conscious barriers between the player and the game that, for me, and I am sure for many others, damage the experience of the game.
So, overall, I have a very mixed opinion of motion controls. It clearly depends on the implementation, and I cannot make a blanket judgment because each implementation is different. I said in the above poll that I "dislike" them primarily because in most cases when people say "motion controls" they are referring to the mixed or bad variety. But I am open to motion controls, so long as they work in the service of the ideal control scheme rather than raising barriers.