A Musical Pathway Using a small number of sound emitters, could you create a room where certain nodes emitted particular tones, but no meaningful sound was heard anywhere else. 
So, for example, by walking down a certain path, you could hear the tones for "Mary Had a Little Lamb." Is there a generalized algorithm to make particular paths for particular tone sets?
 A: You could use audio spotlights to make the sounds audible only along certain beams. This should be close enough to let people walk around in the room and only hear certain notes at certain places.
These audio spotlights use intense ultrasound that makes the air along its path a nonlinear acoustic medium, so modulation in the audioble frequency range can be exctracted.
For more information see:
http://youtu.be/8wJ5Eff7hx0?t=48s
http://en.wikipedia.org/wiki/Sound_from_ultrasound
A: Wavefield Synthesis can do this but not with a small number of emitters, uses a massive array of speakers create a field effect, and phase alteration can reposition sounds within that field.
http://en.wikipedia.org/wiki/Wave_field_synthesis
Downside is it's not perfect, it takes a big array of speakers placed very accurately, works best on higher frequencies. I've experienced the setup in the Technical University of Berlin, you should contact them, but it's 2400+ speakers embedded in the walls of the room so not exactly portable. I heard a few demonstrations, some using speech and some using musical tones and timbres. It worked about 70% of the time, but it's quite fragile.
http://www.four-audio.de/en/references/wave-field-synthesis.html
I've heard of people doing portable wavefield with about 40 speakers (still expensive given the quality of speaker required) but no reports on its reliability.
A: Lower frequencies tend to dissipate in all directions, while higher frequencies tend to be "directed". (For example, you can place your subwoofer anywhere in the room, as the sound waves will propagate in all directions, while your other speakers are more "directed" because they reproduce higher frequencies). 
See this Wikipedia article on sound localization for more info.
If you want directional sound, you have a couple of options:


*

*Control the direct (non-reflected) sound using directional speakers (Example: Maestro directional speakers.

*Control the reverberant (reflected) sound by using room shape, and specifically placed absorptive and reflective surfaces.
A: Let´s assume you could get speakers which are 100% directional, thus ensuring that nodes would be perfectly located around the room. You would still have to rule out if diffusion occurred when the standing waves are formed.
Then, let´s assume all the walls are totally smooth and reflective and provide no diffusion at all, but their absorption coefficient is enough for standing waves, and therefore modes, to be formed in the room.
Also we have two ears, and we do not receive the pressure waves in both at the same time or place. Let´s assume for the sake of fun that you cover up one of them and you do the whole path listening with only one ear. Let`s also assume that we only use single tone notes, and do not use the whole spectrum which would mess up our perfect little modes.
Now, Mary had a Little Lamb starts with an E4, which has a fundamental frequency of $329.628 [Hz]$. In a normal room, we can assume sound speed to be around $343 [m/s]$, resulting in a wavelength of about $1.04[m]$ for our E4.
Now, depending on where we place our speaker, we can produce standing waves of one entire wavelength, or half a wavelength. But the amplitude with which we will hear the tone, will vary sinusoidally along the standing wave.
If we place each speaker across the wall, but with different distances depending on the tone wavelength (thus standing wave) to be generated, we could mark the spots where you would get zero pressure, and the spots where you would get maximum pressure. Between them, perceived amplitude will vary in a sinusoidal fashion. You could try with little jumps, providing the person walking has good balance and can land on the exact spots marked as antinodes.  But then again, you would have to have a speaker for each tone, and if you want to do this for an entire melody in a pathway you would need as many speakers as notes you have in the melody.
So in essence, this cannot be done with no sound anywhere else. It can be done with sinusoidal varying sound between spots, and using a large quantity of speakers.
A: I suppose you could use destructive interference and set up speakers in just the right positions for it to work, but I also assume the calculations needed to achieve it would be complicated (luckily, you're not asking for that).
It should be possible in theory
A: While I'm not sure the exact the answer to this, but there is a room in Grand Central where 2 points are connected acoustically so that you can hear each other clearly (despite being far) if you stand in the corners.
 http://www.sonicwonders.org/?p=426
A: Here is a TED talk on a sound laser type device. It probably could be fashioned to do what you are suggesting. http://www.ted.com/talks/woody_norris_invents_amazing_things.html
