To make a portrait you can benefit from some understanding of how people look at and recognize faces. There are two types of recognition systems: visual and emotional.
visual recognition of faces
Visual facial recognition occurs in a special, dedicated part of the brain: the fusiform gyrus in the right temporal lobe. Recognition of most other things occur elsewhere in the brain. So portraits go to a place in the brain that landscapes, still life paintings, etc do not go.
We have a great deal of information processing ("wetware") in our brains for facial recognition which probably includes facial expressions and body language as well. This is why figure drawing and portrait drawing are such a good exercises. Because we have this area of our brain dedicated to the task, we can tell very easily when there is a mistake. John Singer Sargent said that a portrait is "a likeness with a little something wrong around the mouth". I take this to mean that the brain's facial recognition ability can defeat even the best of us.
It is much harder to spot something a little wrong, for example, with birds or even rainbows. In fact, many of the great artists painted and drew birds very badly. If, however, you happen to be a serious bird watcher, bird identification probably goes to the so-called facial recognition area also (it is obviously not only for faces but for recognition tasks that are important to the person). Consequently, you will probably be able to draw birds better than non bird watchers because you can be more critical of your results using the "facial recognition" part of the brain. Reubens has a rainbow in a painting with the colors going the wrong way.
On the other hand, if you turn a portrait upside-down, the image is not recognized by the brain as a face and does not go to the face recognition area but to the general recognition area. This can be helpful sometimes as all artists know.
Some computer programs searching video tapes for faces do so by looking for the color orange as understood by the computer which includes a dark valued orange which happens to be how the computer sees African American skin color. However, it is thought that the brain recognizes a face as if it were a barcode by looking at the pattern of light bands on the face, just this coarse pattern. This is similar to the artist's understanding of color and value on the face: forehead light, middle of face darker and redder, and lower part of the face gray even in women. This set of three "bars" switches the image as it travels from the eye to the face recognition area.
The interesting thing about facial recognition is that some people are very good at it and some people are terrible at it. There are several tests for how good people are recognizing faces available on the internet (google "face recognition test"). At the very worst end of the spectrum there are people who are "face blind". This normal variation of abilities seems to run in families, appears to be a dominant gene, and more males have it than females. One estimate is that about 1 in 50 people have a significant degree of face blindness.
This can create a problem for an artist if his client has some degree of face blindness. For example, on tests of facial recognition, I do exceptionally well. My wife, on the other hand, is poor at it. The other day, I used what I thought was an "interesting" photograph to do a portrait of a friend. I showed it to my wife who couldn't recognize who it was. Before I got upset it occurred to me to show her the photograph. She couldn't recognize who it was from the photograph either. I have to be careful to make a portrait that the sitter can recognize. In this case a sketch or a photograph could be useful to test if that particular view reads well to the sitter.
There can also be face blindness as a result of disease. In this case it is usually a blood clot in the part of the brain responsible for facial recognition. In medicine it's called "prosopagnosia" from prosopan=face and agnosia = lack of knowledge --from Greek roots. There are thought to have been about 200 known cases with two of them people who were born with it. It is a very profound condition in which people cannot recognize photographs of themselves or their image in a mirror.
We learn about this other sort of facial recognition from some serious brain problems that some unfortunate people have called "delusional misidentification syndromes.". In Capgras syndrome the person thinks that someone well known to them has been replaced by an identical impostor. In this case the person recognizes the person very well, but doesn't believe that they are who they are. There are several variations of this syndrome. For example in mirror self-misidentication the person believes the image in the mirror is really someone else. In both cases there is no problem with visual facial recognition.
The explanation is that there is an emotional response to familiar faces that appears to be entirely separate from the visual recognition task. People who are totally face blind and who cannot say who a picture is supposed to represent, have a physiological response that can be measured by the type of responses that lie detectors use. For example, they might breath a little faster or their galvonic skin response might change when shown a familiar face.
The people suffering from delusional misidentification syndromes don't seem to have the physiological response to faces that the rest of us have. They recognize the face but it doesn't read right to them hence they have to explain it in peculiar ways like the person is a substitute for the real person, an impostor.
E. H. Gombrich , the art historian, has asked whether a portrait is supposed to look like the sitter or, instead, evoke the same response we have to the sitter in real life. It's an interesting question since many of us look at portraits whose subject we never knew and still like them. My wife, who I earlier mentioned is not particularly good at facial visual recognition, is very good at picking up emotional responses to portraits. The other day we were looking at a small Velasquez portrait, and she said, "Boy, I wouldn't want to cross that guy!" I could "see" or should I say feel what she meant.
Gombrich's question can be asked of just about every picture. Some artists are better at communicating the emotional response and some are better at communicating the visual response. Charles Burchfield, for example, clearly tried to paint landscapes that evoked the emotional response. See www.artcyclopedia.com/artists/burchfield_charles.html for a listing and links to specific paintings and drawings. It, obviously, isn't an either or situation, but there can be too much emphasis in what is said on the visual aspect in art instruction and not enough on the emotional response.
I can see a problem with Burchfield's approach. I happen to like his work very much, but I spent a great deal of my childhood in the woods getting the same feelings that Buchfield seems to be portraying. But, if the viewer has no experience of this, he doesn't find a center of interest in Burchfield's work. It's safer to have more visual recognition in a painting.