When the line extensions are applied correctly, their purpose is to show us where our convergences are off. The analysis is one part of the battle, and applying it correctly is very important. Once we know they've been applied correctly (extending away from the viewer), we can trust in what they're telling us. From there, the second half of the battle begins - we have to consciously look at the pattern of behaviour, identify what we're doing (or more accurately, not doing) that results in our lines converging inconsistently.

In your previous submission, the majority of your boxes' lines were extended incorrectly, resulting in at least one set of lines going off in the wrong direction, and making the analysis unreliable. In this submission, your results are much better, but there are still some that are off. Here I have gone over your work, emphasizing what appears to be your starting Y and testing whether your lines were extended in the right direction (similarly to the technique I showed you last time to identify the correct direction). You are correct far more often than before, but there are still some mistakes. The mistakes fall into two categories:

• When the test shows 1 set of lines being extended in the wrong direction, this shows me that you're not thinking back to that original Y and following its arms to determine your correct direction. Perhaps you got a little overconfident. Unfortunately the solution for this is to keep going back to that Y method-based technique for determining the correct direction, to go through the steps in the diagram I provided before each and every time.

• There are a few others - on page 4 and 5 - where all the lines of a given box are extended in the wrong direction. This tend to coincide with you putting your hatching on the wrong side of the box - placing it on one of the faces that points away from the viewer. The box itself may be correct, with lines converging in the wrong direction, but you end up thinking about the box backwards. Correcting this goes back to first ensuring that you are using the Y method - specifically so that the center point of the Y represents the corner of the box that is closest to the viewer (perhaps you constructed those boxes backwards, starting with a Y representing the farther corner instead). Then, you need to make sure that your hatching is placed on one of the planes touching the original Y. Once the box's main edges have been drawn, you may not entirely remember what your original Y was, so you'll have to make a particular point to keep this in mind and not forget, in order to place that hatching on the correct side.

These are the points you need to keep in mind in order to ensure that your analysis is correct, reliable, and trustworthy. Once that is confirmed, it really comes back to what you're thinking about when deciding how to add a new edge to your box (and how to orient it). We have to think about not only the edge we're drawing, but also the three other edges that are meant to converge with it to a single shared VP - including those that haven't yet been drawn.

That means that as you're deciding on the angle for a single edge, you're actively thinking about 4 edges in total. Those that have been drawn are already locked in place, and so those that are yet to be drawn are the only ones whose orientation can be adjusted so all 4 converge together.

Another point that may be useful is to remember that each arm of the Y belongs to a different set, and points to the vanishing point to that set. If that arm were extended infinitely, then the vanishing point would exist somewhere along its length. From there, the second line you add to the set will pin down the specific location for your VP - it will be where these two lines connect.

No matter how you orient that second line, as long as it converges towards the corresponding arm of the original Y as it moves away from the viewer, it will not be incorrect. It is the third and fourth lines where the mistakes occur, where the lines can be converging in the correct direction, but where they fail to meet at a single point. In fact, given that we're not perfect machines, it's guaranteed that we won't do this perfectly, but the purpose of doing this exercise over and over is to bring those margins of error down, and to improve our overall accuracy.

Since you still do have some issues with your analysis, we are going to have to continue ensuring that you can do that correctly every time. Since this is a matter of the steps you take in determining the correct direction, and not a matter of skill/mileage, this is something I want to see you perform consistently before I let you move forwards.