B&K's studies concluded that the ideal steady state listening room response looks like this:
The film industry's "X Curve" shows a similar trend:
but for a much larger listening space and again this is steady state. Some of that was done to hide flaws in the recording technology of the day as well.
So what would I look for graphically in a speaker? A smooth, flat response across roughly the front 30 degrees horizontally with a smooth, falling treble response further off axis. I also look for the vertical polar response so I know how to best aim the speakers. Bass has been found to account for about 30% of listener preference--so get some subs. Resonances are important! They will color the sound you hear. If you can see them in the polar response, you will hear them. BTW, my polar response graphs posted on this site shouldn't really be trusted below 1kHz--they are low resolution and too near field for anything to be drawn from them. Wider dispersion speakers should probably be kept away from the boundaries (see references below) which means they will require a larger room (and they typically are incapable of high output to start with). My 2 favorite speakers I've had in the house and measured are the Mackie HR624 mk2 and the JBL 2325. They are somewhere in the middle dispersion wise and are free from resonances/diffraction artifacts from 1 kHz to 10 kHz. That's what I'd primarily look for, and then how much output is it capable of. Resonances become more audible as output is greater and motor nonlinearities will rear their ugly heads as heat builds up.
http://www.gedlee.com/downloads/AES06Gedlee_ll.pdf
Narrower dispersion speakers can probably be used in any room at a further than 2 meters/ 6 feet in a cross-fired manor. They are also typically large (Larger speakers naturally have less dispersion), efficient, and handle high volumes well. This makes them ideal for just about any situation except near field monitoring because every movement will result in a large difference in audible FR(don't tell recording engineers--many won't believe it). They are also the most difficult to build well d/t the large cones and waveguides involved. Just listen to PA speakers and ask yourself if you would want that in your living room. Dr. Geddes, Jean-Michel Le Cleac'h, and I'm sure many others have gone to great effort to improve this sort of speaker system and they are not cheap.
So my opinion would be to avoid the extremes b/c the middle of the road is affordable, easy to get a great spectral balance from, and capable of accomplishing the output required of home speakers. Broad dispersion will require better drivers to get great output and thus cost more for HT needs, but will give you more than enough for near field monitoring. Narrow generally will give you all the output you could ever want, but the less resonant/great sounding designs are costly.
Oh, read this as well:
http://seanolive.blogspot.com/2008/12/part-3-relationship-between-loudspeaker.html
Bass must ultimately be measured in room as the room becomes part of the source at low frequencies. The time domain is less important here and the frequency response is the dominant part of what we hear. That's not to say the time domain can be ignored, just that FR is the most important thing to pay attention to in the bass range. Good time performance/quick decay is the icing on the cake.
The in room, steady state midrange/treble response is not important either except that it should be rolled off to some degree. A smaller room should probably have less than a large room. We have great time domain hearing resolution and the anechoic response of the listening window will dominate our frequency perception and the reflections will add detail, space, source width, envelopment and timbre as well.
More recently Dr. Olive did a subjective evaluation of room correction software. The results seem much in line with older studies.
Dr. Olive's Study
This is where this post started:
http://dtmblabber.blogspot.com/2010/12/psychoacoustics.html
http://dtmblabber.blogspot.com/2010/12/above-modal-dominated-frequencies-300hz.html
Further reading that will cost some greenbacks on the subject:
http://www.aes.org/e-lib/browse.cfm?elib=6248
http://www.aes.org/e-lib/browse.cfm?elib=7673
