Panel 4: We see Jan in close-up, over Gary's left shoulder.
Presumably, that image happened to pop into the writer's head at that stage in his writing -- and, against all common sense, against all the rules of decency, against every civilized fiber of his being, he wrote it into the script.
The problem, of course, is that visual narrative must be built chronologically, step-by-step, each panel relating sensibly to the one preceding it. So the odds against our writer's piece of ad-libbed direction actually fitting into our shot list are staggering. We'd be as likely to improvise a suitable line of dialogue to inject into the middle of his scene as he is to invent a suitable visual to inject into ours. With one stroke, his piece of direction has inadvertently destroyed whatever sequence of visual narrative we have constructed until that point.
Two obvious solutions to this problem occur to the inexperienced artist:
For one, why not just ignore the writer's direction and go on with our own plans? Unfortunately, in comics, most writers have greater authority than most artists, so the artist isn't often in the position to ignore directions in the script. And the writer tends to see every word of his script (not just the dialogue) as "my baby," so to ignore any of it is to insult His Artistry -- even if what you're ignoring is some pedestrian visual idea he included without planning or reflection. Now, most writers are willing to "compromise" by letting you depart occasionally from their visual direction, but even then they'll keep score, and if you stray too often they'll throw your past departures in your face. In essence, every time you ignore a writer's direction, you're borrowing a cup of sugar. You may be safe the first few times, but eventually the writer will put his foot down. "I think I've been flexible QUITE ENOUGH with you, Buster." And then you're sunk.
The second obvious solution would be to construct each scene around its occasional piece of visual direction. If, in a three page scene, the writer has specified that the fourth panel of page 2 needs an over-the-left-shoulder shot, why not accommodate him by building the scene in either direction from that shot? Well, one problem with this approach is that even that single shot may be incompatible with the rest of the scene. The writer may essentially be asking you to construct a jet engine around a Tootsie roll center. Another problem is that, two panels later, the writer may include yet another piece of direction, which will no doubt be incompatible with the previous piece of direction, and with whatever visual narrative you might construct around that. Even if you are ingenious enough to design a dessert recipe around Tabasco sauce, God help you when you are then ordered to include anchovies.
So. There are really only two workable solutions to the problem of visual direction (apart from chucking a mainstream career altogether and writing your own darn comics):
The first -- and by far the most preferable -- solution is for the writer to MIND HIS OWN BUSINESS and limit his writing to plotting, characterization, and dialogue, leaving any visual narrative to trained professionals who will actually think it through. Suggestions are always welcome ("I'm picturing a close-up, here, but do what you think best"), but the astute writer will remember that the artist is juggling countless details that are beyond the writer's purview, and will allow that process to happen unhindered.
The second workable solution is for the writer to actually thumbnail the scene himself, beginning to end, to see if his direction actually makes sense in the broader scheme of things. Once he has drawn the scene out, hopefully he'll notice any problems in his direction, and adjust it accordingly. This isn't a great solution, since even problem-free direction may be uninspired in the hands of a non-artist, and in any case it waltzes all over the artist's toes. But some writers (such as Alan Moore) have been known to make it work. To this end, here's a list of pitfalls for the self-thumbnailing writer to watch out for:
*FIRST SPEAKER ON WRONG SIDE
In comics, it's usually a good idea for each speech balloon to appear above the head of its speaker. This helps readers keep a straight idea of who's saying what. To this end, the first speaker in a scene is usually placed on the left, so that the resulting reading order throughout the scene will maintain this balloon-above-its-speaker convention. The writer should be sensitive to this convention when deciding the speaking order and point of view of each panel.
*BROKEN 180 RULE
This rule states that the readers' point of view shouldn't jump across the line of action in a scene. The line of action is a lateral space through which the scene's primary action occurs: if two characters are conversing or fighting, the line of action extends through them both; if a character is running toward a mailbox, the line of action passes through him and the mailbox; if a character is staring at a street sign, the line passes through him and the sign. The readers' POV can move anywhere in a semi-circle (180 degrees) around the line of action, but to cross that line from one shot to the next is to break the 180 degree rule.
This rule exists because readers construct imaginary space for each scene to occur in, and they pretend to occupy that space as they read. This imaginary space is generally built around an axis created by the line of action (due to the action's prominence in the scene), so the reader will relate everything to that line. If the reader is suddenly 'teleported' to the unfamiliar side of that line, his imaginary space will lose its axis, and he'll be disoriented. He may continue to understand the scene's events, but his sense of the scene's space will begin to disintegrate, along with its reality.
Another reason the 180 rule is important to comics is that it supports the balloon-above-head convention described above. If we always see the characters in a scene from the same side, it's easier to keep their speech balloons straight.
If the line of action must be crossed, the visual storyteller can 'pivot' the line by creating a new line, such as by a character at one end of the line shifting her attention -- which constitutes the line -- to a new person or object, or by a character at one end of the line moving across the reader's fixed POV.
The storyteller can also pivot on the line, with a bridging POV shot (facing from a character or object at one end of the line to the character or object at the other end of the line), followed by the new shot from the far side of the line.
While this second solution works fine on film, it sometimes poses problems for comics. Which leads us to...
In film, each shot is seen by itself, with nothing on either side of it. As a result, film can get away with shots that pose problems in comics. Since prior and subsequent comic panels remain in the readers' peripheral vision, oddities can occur there which are invisible to writers who only imagine their stories in filmic terms. For example, suppose a script says the following:
Panel 1: Square panel. We see Jan in profile from her left side, knees bent, hands on her knees, smiling. In front of her we glimpse the bars of a baby's crib.
Panel 2 : Same size panel, same angle, same pose. Jan is reaching off-panel into the crib.
The artist reads this description and immediately emails the writer, begging to alter it. The writer can't understand why the artist is being such a prima donna. "I'm picturing the scene in my head and it works perfectly," thinks the writer. Unfortunately, he's picturing it in filmic terms, one shot before the other:
On the other hand, the artist is picturing the sequence in comics terms...
...and the problem becomes obvious.
A similar problem occurs when we switch POV -- even along an axis, in accord with the 180 rule. Back to the script:
Panel 3: Same angle. Phone rings in background.
Panel 4: Opposite angle, same pose. We see Jan from where the phone is as she turns her head toward the sound.
Notice what happens when we cut to Jan's reaction. In filmic terms, it looks fine:
But in comic terms...
...it's as though she grew a twin, or is standing by a mirror.
Opportunities for such tangencies are common, and they multiply when visual direction is being given by someone who hasn't thumbnailed the scene.
*WIDE EVEN PANELS, TALL ODD PANELS
Consider the following direction:
Panel 2: a broad panel, stretching clear across the page.
Panel 5: a tall panel, stretching two or three tiers.
The comic artist groans when he sees direction like this, because it fails to consider how space is divided on a comics page. In the first example, no room is left for panel 1. (That is, unless panel 1 is also stretched across the page. But a panel that wide is a cumbersome waste of space, unless its subject is uniquely suited to that shape.) Similar trouble would result if panel 4 had to stretch across the page: panels 1 through 3 would have to be narrow, and panel 5 would have to be broad.
Likewise, a tall panel on the right usually means you must stack panels on the left, and stacking on the left causes the reader confusion: "Do I read the panel below, or the panel to the right?"
So panels' shapes and sizes shouldn't be dictated unless the writer has thumbnailed the page to determine how they'll fit.
*MUTUALLY EXCLUSIVE EMOTIONS IN ONE PANEL
I've even seen this from Eisner Award winning writers -- a character's two opposing emotional notes are summed up in one panel:
Jan: "Man, I'm bored...I've been sitting here all day with nothing to -- oh NO, I forgot to deposit that CHECK!"
It looks fine at the script level, and it would look fine on film... but how is an artist supposed to draw this panel? Either Jan's distress looks like boredom, or her boredom looks frantic, or you're stuck with that corny Archie Andrews dual-face convention:
Bleh. The dual-face is admittedly workable in cartoony contexts like the one above, but when you're dealing with subtler shades of emotion in a more realistic context, it's atrocious.
As a rule: there should only be 1 emotional beat per panel. Split dialogue into multiple panels if it communicates mutually exclusive emotions.
*INCITING ELEMENT ON WRONG SIDE OF REACTION
When a character reacts to a noise or a piece of dialogue or action, the inciting element should appear on the left side of the character, or above the character, to reflect our reading order. It's disorienting to see a character wince and THEN read what bothered her. So writers should thumbnail where the dialogue or inciting element is likely to be placed in the panel, ensuring there's enough room on either side of the reaction for any pre-reaction and/or post-reaction business or dialogue.
What NOT to do:
(Narrow panel. We see Gary and Jan from the waist up.)
GARY (elbowing Jan in the arm): "Better bring up those grades; I don't want to attend the prom alone when your parents ground their prodigal daughter!"
JAN: "Ow! Cut it out; I just bought a sleeveless prom dress, and I don't want any bruises to ruin it!"
Jan's "Ow" should occur to the right of -- or below -- Gary's elbowing, but a waist-up shot probably won't grant us enough space down there for her ensuing dialogue. Also, we don't want to see Jan's facial reaction to the elbowing before it has occurred. So, ideally, this panel would show us the couple from their knees to their shoulders, with Gary's dialogue to the upper left of his elbowing, Jan's to the lower right of it. The elbowing is the panel's key visual element, so that takes precedence over the faces -- one of which we don't want to see anyway, since it would precontextualize Jan's reaction.
(There's room for a certain amount of achronology in comics, since the readers must imbue the still images with flexible timeframes to carry the story comfortably. But generally speaking, the more chronological sense a panel makes, the stronger its spell over the reader.)
Usually when an image pops suddenly and fully formed into a writer's head, the ease of its conception is due to one thing: it is a visual cliche. The fastest idea is the most obvious idea, and the most obvious idea is the most tired idea.
Unfortunately, the writer often interprets ease of conception as a sign that he has been struck with a blessing from Heaven, an inspired creative vision which Must Be Recorded. He fails to suspect that this dime-a-dozen image has simply been engraved on his cortex by years of comics-reading and TV-watching. Pitching such images to an artist would be tantamount to the artist suggesting phrases like "See you in hell!," or,"Don't look now, but we've got company!," yet the writer -- who would never dream of using such cliched dialogue -- happily burdens his artists with one visual cliche after another.
Cliches are hard to identify in the abstract; usually it's just a case of "I've seen that a zillion times." But it's a safe bet that if an image occurs to you without a moment's reflection, it's a cliche. You, the writer, may not recall having seen it before, but you aren't paid to recall such things. Likely your artist will be able to identify it: "Oh yeah...the Blade Runner 'elevator' shot -- still doing good business on the straight-to-video market." Defer to his nose; it always knows.
Another hint that your brilliant image is a cliche is if the panel would look good all by itself on a T-shirt. A shot that ably survives decontextualization probably isn't strong in context. This includes close-up & straight-on shots of dramatic reactions, key characters or objects occupying the center of a panel, characters seen in full-figure when they're not doing anything, and any shot with the moon in the background, a ceiling fan or a stripper's legs in the foreground, or the shadows of a window frame or Venetian blind falling on anything in the middle ground.
Also cliched: any facial expression that the man on the street would automatically associate with a given emotion. Don't dictate to the artist that the angry character must grimace or the happy character must smile; allow him to explore less conventional ways to visually communicate those emotions.
It often behooves the artist to leave out background detail. This may be a scheduling consideration (he won't have time to draw a second angle of that shuttle launch after he's spent the afternoon drawing the first one), or a dramatic consideration (he doesn't want background details to interfere with what's happening in the foreground). The writer isn't obligated to consider the artist's schedule -- a page is a page; sometimes the artist gets off easy, and sometimes he doesn't. But dramatic considerations should concern the writer, and he should avoid writing scenes like this:
Panel 1: Our couple is sitting near the front row of a theater, and looking between them we can see a magnificent play unfolding on the stage. Let's make it CATS.
Panel 2: Same angle. Gary and Jan turn to look at each other; Jan's eyes are wet with tears.
Panel 2 is not going to come across well if we can still see all those dancers cavorting around in cat costumes between our principals. Better to let the artist shift to an angle which disincludes the stage.
Also bad, but in the opposite way:
Panel 6: close-up on Gary and Jan as they and everyone else in the theater get up and file out.
It's hard to cram background details (such as everyone filing out) around the edges of two principals in close-up. Unless the close-up is Absolutely Mandatory for some reason, let's pull back for a comfortable shot of everyone filing out. The reader can surely stand to be several yards from our beloved heroes for a few seconds.
*CHEATING OUT FACIAL EXPRESSIONS
"Cheating out" is dramaspeak for artificially turning a character's face to the audience to show the audience her expression. It's occasionally useful, but it can become a crutch in the hands of storytellers who don't trust context or body language to communicate a character's inner life. To wit, borrowing from the scenario above:
Panel 3: We see Jan staring wide-eyed as the lead cat begins his dance.
The artist could cheat Jan's face to the side here, to give us a glimpse of her eyes, but this will look fake, as though she's looking to one side of what she's supposed to be seeing. Better to just indicate her stare through her body language, from behind. Perhaps she's sitting erect, leaning slightly forward, gripping Gary's hand, etc.
If we must see her eyes, that should be worth an extra panel, one which doesn't attempt to include the onstage dancer.
More could be said, but these tips should be a good starting point.