I would say that they definitely "seem" accurate to me. Which, I'm assuming is the goal. But, someone with more mileage behind the figure in perspective, in regards to anatomical shapes, might have a better eye. Would you say these are in true orthographic view? The skeleton is still in perspective and foreshortened in a few areas in both frames. You could try and draw the skeleton in a similar but basic pose in a pure profile view. Maybe just having the skeleton squatting with their arms out. You wouldn't even need to draw details of bone structure, of course That way, you can draw the lines necessary to transfer to a foreshortened image. It's a bit of more work, but worth a try - maybe? One thing is for sure though, the form seems a bit hard to read because you've drawn through everything. So like in the frame where they are facing away, the pelvis area is hard to read accurately. But, it's cool and I think it reads, just maybe beyond my abilities to nit-pick!

Lot of awesome replies already, but I did want to mention something on line quality. I think overall, your line quality tends to lean towards feeling a bit scared and uncommittable. In the longer sketches, I think this becomes more apparent. As other's have said, focusing too much on the contours of the form can lead down a not-so-desirable path in gesture drawing. In these drawings, I feel when you get caught up in the contour, your commitment and confidence of strokes becomes lessened - you make more smaller, corrective strokes. Some artists call this "scratchy". You said in a thread below you were doing DrawABox, which has good exercises to help with that! What I think too could help is to slow down, ghost the line-of-action enough to where you can be more confident in where it's going. In the longer sketches, I wouldn't approach it as "ah, I can do more detail now", but rather "ah, I've got more time to do the same thing but better"!

I wouldn't worry about any conventional horizon lines or any vanishing points for that matter - if you are drawing from these sorts of references, and just working on the figures themselves (that is to say, you aren't drawing them as a part of bigger scene). The photos have the figures so close to the picture plane in deep space that any real convergence of any curved or straight line would be pretty minimal IF you are plotting a vanishing point. Plus, you would be tracking which contour and cross-contour lines are parallel and converging, and which are not. With an organic form, that's pretty hard and overkill imo. What could be useful is working on foreshortening. Which is just a form staying at the same distance from the picture plan, but changing orientation. https://youtu.be/R60e9_ofV68?t=600 This whole video is great, but Marshal Vandruff show's what I'm talking about here pretty nicely and concisely.

I know that the pelvis's front plane tilts down in normal standing posture, and the front plane of the rib cage tilts upwards. But, when the model is in these twisted positions, I can't seem see how they should slope.

Here's the reference photos as they appear;

My immediate reaction to seeing these; cool rendering! Did you use a reference, like a photo or proko's vids, or are they imagined? The jaw construction on the lip study seems a little "elongated" vertically. I think this would work in caricature, but sense I don't have much else to reference in the construction (i.e. I don't have a nose, eyes, or face shape to see how they are pushed in relationship to the lips), it just feels like lips that are stretched in a vacuum. The vertical narrowness of the perceived mandible shows this too, I would maybe push that out a bit. The eye ball itself has a good sense of rounded form. However, the iris and pupil don't seem to map onto the rounded surface exactly, in perspective. Maybe work on circles, ellipses, and stability of the stroke of the circle? I guess this would depend on where the person is looking. One big thing about the eye study though, is that I think you've got some anatomical features in reverse, or didn't include eye lashes on one side of the eye. The inner part of the tear duct usually is on the opposite side of the where the most noticeable eye lashes fan out. The nose stands out to me. It feels like the bulb at the tip of the nose is actually turning to the figures left side, and down, rather than being exactly in the middle. It also feels a little elongated. Again, don't know if it's the reference, but it feels like the position of the nose isn't lining up with the ghosted figures head in deep space. I want to think that it should be turned more inward, and upward, so that we are mostly seeing underneath the nose in a more triangular fashion.

Hi @Samuel Eli, i went through Marshall's perspective course recently as well. The first thing i noticed was that the box you drew is not a cube, the length sides are larger than the height of the box. This means that if you want to draw a perfect circle in one of the side planes the circle will touch the top and bottom of the rectangle, but it will not touch the left and right side of the rectangle. So i think the construction of the ellipse is fine. Try doing it on a more perfect cube in perspective and it will work out better. In the beginning of Marshall's classes he eye-balled/estimated a perfect square in perspective, in the last class he actually provides a method to draw perfect squares in perspective (Planned Projection). Hope this helps

One of the things Scott Robertson talks about in his book is using ellipses to estimate a cube. In other words, draw the ellipses first, get them placed, and then draw the cube around them. Your ellipses look like they are in correct perspective. They are just too far apart.