By "that," of course, I mean, "accomplish that strange graphical flatness without sacrificing a sense of volume and verisimilitude." This is what that trixy sonofabitch Ingres did that is so mystifying.
Consider his 1836 study of the rather anemic Madame Victor Baltard and her daughter Paule:

Look carefully at the face. There are few darks, very few. She is lit entirely; there are few halftones, and the only full shadows are cast shadows at the lower eyelid on the right, the bottom of the nose, and the lips. In short, she is the next thing to being a line drawing only, like the architectural detail on the left. Yet we have no trouble interpreting her as three-dimensional, fully formed.
This reads naturally, but in my experience it is a deeply unnatural way to draw, and never arises by accident. So how did Ingres pull it off?
The effect is even more pronounced in the bottom figure in his 1814 study for La Grand Odalisque:

Consider how much of the figure is blank paper! Ingres has somehow completely indicated the mass and form of a figure without drawing anything at all!
Well, perhaps you are lazy like me, and you have admired this effect in Ingres drawings and never gotten around to analyzing it enough to replicate the technique. That's a safe short-cut, until the day arrives when you suddenly wish to use the technique and don't have recourse to a book of Ingres. That fateful day arrived for me last Monday.
I was at Spring St., where the beautiful Claudia was modeling. Having been out of town or otherwise occupied quite a lot recently, my drawing was rusty, and I was not particularly pleased with most of what I was doing. For the final 40-minute pose of the evening, Claudia took a reclining pose. And I thought to myself, "Holy shit, it's Ingres lighting! I can do an Ingres drawing!" And a second later I mentally wailed, "But I never figured out how!" So I was in one of those delightful high-wire situations where you have to solve a complex problem on the fly to meet an opportunity which will never come around, in quite the form presented, ever again.
Re-deriving Ingres' pictorial principles as I drew, I came up with this:
Let's leave aside for a moment any questions of quality and discuss this from a purely formal perspective, because I think I pretty much sorted out how Ingres was pulling it off. The system depends on a set of internal and external variables being set to values which complement each other - external variables being those pertaining to the subject, and internal those that pertain to the drawing technique itself.
EXTERNAL VARIABLES
1. The lighting must be frontal, so that the brightest planes are those perpendicular to the line from viewer to object. Planes advancing toward the viewer or receding from the viewer turn away from the light as well, resulting in darkening, as can be seen in the central foot in this drawing:
The front of the big toe, and the bottom of the gap between the big toe and the second toe are bright, but the facing sides of the big toe and second toe are dark, because they are turning away from the light. Similarly, the edges of the foot darken more than its frontal plane, as they too dip away from the viewer, and the source of light.
In fact, this lighting scenario produces something rarely seen in reality: an actual outline. Because views of rounded objects always terminate in edges turning away from the viewer, the dimming of turning planes results in a distinctly dark edge. Here reality merges with line drawing, and a scenario occurs in which an object can be drawn largely with contour lines and still retain realism.
2. However, the light must not be on exactly the same axis as the viewer - the viewer should not be sitting beside a spotlight. There must be an offset, producing mild shadowing on one side of the figure. The result of this effect is well-pronounced in this drawing:
In this instance, Ingres has placed the lighting to his left, producing halftones of recession on the left, but true cast shadows begin to emerge on the right.
This offset is not important pictorially, but rather cognitively. The eye is attracted to even lighting, but it slides off of pure even lighting: equal dimming on the left and right sides of the object makes the object unfocused. A weighting of the dimming to one side anchors the object, producing a greater impression of form. You will see this offset again and again in Ingres, and in the case of the Claudia drawing, the offset was as much as 50 degrees - but it still didn't read as true lighting from the side.
3. The primary object must be pale in color. This is lighting for marble, alabaster, and white people. Why? Because the technique depends on high contrast between the local color of the object and the half-tones, cast shadows, and edges.
Claudia's Armenian, with a kind of bronze skin tone and olive undertone, but for the purposes of making an Ingres drawing, she is of sufficient honkitude to work. Especially if you're drawing on tan paper, as I was. You can see in my drawing that much of the variation between brightest-brights and moderate brights results from variation in local color, not in lighting. Which is to say, her breasts and pubic bone are brighter than her belly, not because the light there was brighter, but because the skin there is lighter (Claudia, perhaps, spent a great deal of time gallavanting around Cape Cod in a bikini this summer). Accurate reflection of variations in local color is insanely difficult in monochromatic drawings of chiaroscuro lighting situations, but relatively simple to do when using Ingres lighting.
4. Soft light. It is important that the light softly model the object. Hard edges to shadowed areas are apart from the purpose here, as well as spotting of light around dimmer regions. The light should be broad and diffuse because, as will be explained below, the drawing technique itself heightens contrast.
INTERNAL VARIABLES
Constructing a situation in which it is possible to draw an Ingres-type drawing is insufficient to actually draw an Ingres-type drawing. You and I, walking into the studio at Spring St., would certainly have looked at Claudia in that pose and immediately thought, "Ingres." But an innocent bystander might well have missed the resemblance, and this would have been a valid interpretation of the scene. In fact, Claudia looked nothing like an Ingres drawing. As much as Ingres' drawing technique is grounded in a certain configuration of externals, it is also a wildly stylized technique. It only looks realistic. Let's consider some of the distortions involved in translating even a well-suited scene into an Ingres drawing:
1. Nonlinear gamma
Gamma, as used in visual technologies, is the name for a particular mathematical formulation of the relationship between original object brightness and representation of object brightness in the reproducing medium (computer monitor, movie screen, inkjet print, paper and pencil drawing). Here's a recent photograph of me in front of some building somewhere:

This image's gamma, represented here by Photoshop's CURVES tool, is linear - input brightness is strictly proportional with output brightness, producing an ordinary tonal range with lots of intermediates (greys, half-tones).
Now here's the same picture with a little gamma modification:

In this case, the gamma has been altered so that output brightness remains at zero for a few degrees of input brightness up from pure black. Likewise, output brightness goes to pure white before input brightness reaches there. So a lot of darkish regions have gone to black, and a lot of brightish regions have gone to white. Also, the graph curves, so that there is a rapid transition from black to white, with fewer intermediate values. But notice that the curve is not an even curve - it has a bulge in the top half of the graph. This bulge drags halftones toward lightness, and clusters them in the bright range.
You will notice that this image looks much more similar to an Ingres drawing than does the first version: it is composed of flat bright regions, falling off abruptly toward darkness, with little in the way of middle values.
It doesn't take Photoshop to accomplish this gamma distortion. All it takes is knowing what you're trying to accomplish (or, in my case, figuring it out quickly while sweating bullets at Spring St.). Consider the Claudia drawing again:

You think those shadows under her butt were that light in person? Fuck no. They were pretty dark. But the halftones were dragged toward brightness by the nonlinear gamma of the representation, and they wound up pretty light. On the other hand, the shadow beneath her neck was just a little darker - so it started to fall off the cliff of that steep curve toward really dark.
2. Finicky line
This technique lives and dies by line - choice of line, and quality of line. It is a simplifying technique, and the line must be consistent with that. Extraneous lines must be eliminated, and remaining lines must be traced out beautifully. Whether or not you think I hit any beautiful lines in this drawing, the principle stands: that because line is so important to the paradigm, the character of the line must be clear, specific, and well-executed for a drawing inside the paradigm to succeed. Of the 40 minutes I had for this pose, I spent about 12 on the lines alone, an unusually high proportion for me (I'm a partisan of the let's-wing-it faction of art).
3. Precision of form
Line must be executed brilliantly because it is so explicit in the Ingres drawing. Form must be executed brilliantly, paradoxically, because it is not. What I mean is, the tonal compression of this technique eliminates much of the information about form available to the eye through halftone rendition. This means that the Ingres drawing depends unusually heavily on the information-completion procedures of the visual brain. Therefore, the half-destroyed traces of form found in the drawing must correspond unusually precisely with those information-completion procedures in order to invoke them correctly. Practically speaking, it means that if you want to use the Ingres model to depict a figure, you have to really know the jesus out of your anatomy.
A failure to know the jesus out of anatomy is amply demonstrated by a different, and enormous, body of work. The Ingres technique is closely related to a second technique: pencil drawings by not-very-talented beginning artists copied from pictures in Playboy magazine. I'm not going to provide any examples here, because I'm not in the business of trashing innocent dilettantes, but Hef long ago figured out the same thing Ingres did: diffuse frontal lighting looks great on the figure. So when people who are clearly never going to be functional artists take it into their heads that they're going to draw nudes, and go to the obvious source for the non-serious art student, they stumble immediately on Ingres lighting. And invariably, they soon figure out to blend their tones by smearing their pencil marks with their thumbs. Also invariably, they incorrectly invoke the form-completion software of the brain, because they don't understand a thing about anatomy.
4. The well-placed dark point
Here we begin to depart from the representational altogether and enter into the purely compositional elements of the technique. Look at Ingres's 1815 drawing Lady Harriet Mary and Catherine Caroline Montagu:

This time around, consider the darkest points in the picture: the curves where the hats meet the (caucasian) girls' heads. A couple spots where the taller (blindingly white) girl's shawl meets her shoulders, the bow under the shorter (pigmentally challenged) girl's chin. And a few other points.
The function of these points is to organize and focus the composition, and round out its range of values from full white to full black. While these are necessary functions, it was not necessary that these particular points be chosen for the purpose. Any number of points could have served. Ingres chose these points for a reason described by art teachers as "wanting to lead the eye in a particular way, or emphasize certain structural features of the figures or narrative properties of the scene." I personally have never believed this kind of thing is so explicit for an artist, and prefer to phrase it that he chose these points because they felt right. Anyway, he had a lot of leeway with his choices. There is always a lot of leeway in this mode when choosing the needed points of maximum dark.
In the Claudia drawing, I placed my darkest darks where her raised near arm meets her side, at the deepest incurve of her waist, where her butt separates from a fold in the cloth underneath her - and on the lower curves of her farther breast and rib cage. Why up there? Who knows, it seemed like a good idea at the time.
5. Good sense
All this exposition has hopefully made a couple of things clear - a. that successful use of the Ingres technique depends on a higher skill level than most other branches of realist drawing, and b. that within the realm of realist drawing, it is perhaps the most subtly but radically expressionist of modes. Because it fools the eye into considering it realist, while partaking of such extreme stylization, it affords a very broad field for interpretation. Beyond all of the mechanical considerations covered above, and even beyond the semi-abstract consideration of the well-placed dark point, the technique depends on inspiration, on a well-formulated vision, on taste and style - on all the things that go to make up good sense. Consider again Ingres's more fully-rendered study for La Grande Odalisque:

It remains mostly white. But look at the majesty of curve, where he has found the outlines. Consider the eroticism of the selected shadows - her armpit, the lower curve of her breast, her butt, her thigh, the shadow of her arm, and the bases of her toes. Where Ingres's eye has snagged, your eye snags. Where he has adored a form, you will adore a form. What he craves, you too must crave. This image is not a representation of a thing that exists in the world. It is a conversation between Ingres and his model, and Ingres is doing most of the talking.
Now, I'd like to offer the usual caveat. I've talked up the skill involved in making an Ingres drawing, and I've talked some smack about people who do it wrong. I've offered a drawing of my own as an example of the technique. And the caveat is - I'm not making any claims of success. That's not for me to decide, and trust me, my judgment is harsher than yours anyway. I offered my drawing because this subject was on my mind while I drew it, and I learned as I drew, and I felt like I could illustrate many of the principles I'm discussing by reference to it. It is very possible for a picture to demonstrate a principle without also demonstrating it well. I'm still learning.