When looking at the Wikipedia article about "loudness", I'm presented with the following picture:
From my understanding, this represents the threshold for how loud a certain sound has to be to be perceived. If there only would be one curve, I would have no problem understanding the diagram. However, why are there several curves depicted? What do they represent? I can't wrap my head around this.
Each red line represents a loudness level (measured in phons). That is, for a given red line, every point on that line is perceived as equally loud.
This is a way of representing a 3d graph on a 2d medium. The x and y axes are sound frequency and sound pressure level (decibels), respectively. The z axis is a subjective measure, "Loudness". Now, we have no way of actually representing this z axis on a computer screen, so there are a couple of solutions. One is using some sort of flash applet to render the 3d graph in a way that allows the user to rotate it. Another is using a 45-degree angle view of the 3d graph.
The method used here is much simpler to produce; instead of going through all of the grief involved in 3d rendering, we'll display slices of the 3d graph and label how far up or down those slices would be in 3d.
Anyway, that was very long-winded. The point of this graph is to show that neither frequency nor sound pressure level is enough to determine loudness on its own; the combination of both of those factors as well as the anatomy of the ear determine loudness.
You are looking at the Fletcher-Munson equal loudness curves.
The point, which no other answers have explained yet, is that this is a graph of how the human ear responds to sounds and loudness. The human ear and brain do not perceive a linear increase in sound pressure levels in a linear fashion, and the human ear and brain's perception of the volume of a given area of the frequency spectrum, in relation to all other areas, is different at different sound pressure levels.
When I studied recording studio engineering in college, this graph was used to impress upon us that music recordings must be made, mixed and mastered with the human engineer listening at a moderately loud volume level (80 decibels or so), because if one were to listen at too low a level, one's ears would not give an accurate reflection of how the music would sound to a listener playing it back on a car radio, stereo system, or what have you. For example, if recording and mixing were done with the engineer listening to a very low volume, one would mix in too much bass to over-compensate for a perceived lack of bass in the recording. Then when consumers played the recording back on their stereo systems, there would be much too much bass coming out of their stereo system.
The ELCs illustrate how loud tones at every frequency need to be sound 'equally loud' - as loud as each other - to the ears of thousands of test subjects averaged together.
Using the graph curves, it suggests that we are most sensitive in the upper-midrange region, where reside the frequencies most critical to speech intelligibility. Conversely, less sensitive from 8kHz and higher frequencies, and MUCH less sensitive to lower midrange, mid-bass and deep-bass sounds(500Hz and down).
In fact, flip one of the curves in the middle upside down, and you'll have a good idea of the averaged frequency response of human ears. It's no wonder that more accurate(flatter) headphones and speakers sound so unappealing to many people, especially to average consumers who instead flock to headphones from Monster(Beats by Dre) etc.
Those are designer phones that cost up to twice as much as more accurate alternatives, and tend to have a more 'V'-shaped or smiley curve response. While initially appealing, that effect grows fatiguing after a while and does not serve the music being consumed through them.
Succinctly, it is better to seek out flatter headphones(Sony, Shure, and A-T all offer models for between U$100-130), and use the EQ built in to your device or the controls on your home listening system to accommodate your hearing preferences.
