As Bill notes, the Earth's orbit is not circular.
As Kepler deduced about 400 years ago, all orbits are ellipses, with the Sun at one of the foci - hence the Earth's distance from the Sun varies throughout the year. He also realized that the speed of a planet in its orbit is not constant, but when the planet is closer to the sun, it moves faster and vice-versa.
However, the spin rate of the Earth - i.e. the length of the day - is basically constant. (I say basically, because at the millisecond level, the length of the day does vary.) This is the length of the day that you measure with respect to the stars or something far, far away from the Earth.
A "day" as most of us think of it, however, is tied to the Sun. i.e. It's a "day" from one noon to the next, say.
But, since where the Sun is in the sky depends on both the rotational (daily spin) position AND orbital location of the Earth, the time of local solar noon changes systematically throughout the year - the curvature of the Earth's orbit has to be taken into account and added to (or subtracted from, if we are getting behind) the Earth's rotational position.
This effect shifts the daylight part of the day ahead or behind. This results in the analemma effect (the figure 8 often drawn on a globe) which notes how far behind or ahead the sun is. The amount of behindedness or aheadedness is with respect to a fictitious Sun that would be observed if the Earth was in a circular orbit.
To a small extent, the shift is fast enough that part of it occurs during the day, shifting sunrise and sunset by slightly different amounts. As also noted, the tilt of the Earth's axis, which affects where the Sun appears to rise and set, also comes into play.
See here for some more details.
See
also .
Edited: 13 Dec 2011, 3:15 p.m.