Communication is a lot like music; and like music, communication has multiple dimensions.
A word is a roughly one-dimensional message--a single concept, a single signal, a single channel of information. It's like one musical note: There's meaning, but only one meaning.
Now consider a sentence. A sentence is a string of single words--like a string of notes making up a melody. The melody has more meaning than the sum of the meanings of the notes that make up the tune; similarly, the sentence has more meaning than the individual words. Each word, by its presence and position, modifies the meanings of all the other words in the sentence. A sentence is a two-dimensional signal made up of words and word order.
So far we've been working with just words; but there's a lot more to communication. Let's add more: Volume and tempo are present in both music and in speech. In the sentence, they provide rhythm and emphasis. Now we have two more channels and two more dimensions. At this point, it becomes impossible to model communication all at once as a physical signal because we have run out of dimensions (it's still possible with some math tricks, but I shan't bore you with them). Communication, in the form of sentences, with volume and tempo, is still only auditory; but we have once again refined the meaning of the message by adding more dimensions to the linguistic signal.
For you math types, let me remind you of the effect of adding more dimensions to something: The amount of information you carry doesn't increase linearly--it increases exponentially. In the physical dimensions, a ten-unit line stretches to a hundred-unit square, to a thousand-unit cube, to a ten-thousand-unit hypercube, still ten units on a side just like the line. A sentence with volume and tempo is exponentially more complex than a single word.
So far, all we have is a single linguistic melody, such as you might play on a flute; but there's no harmony yet. Let's add yet more channels to the sentence: Eye contact. Facial expression. Body posture. Distance from the audience. Gestures.
In music, any composition written for a choir, an ensemble, or an instrument capable of multiple simultaneous pitches (like a piano or a guitar), usually has multiple notes being played at the same time. Most of these "extra" notes support the main melody, reinforcing it and making it more meaningful. If you like music, you probably know that a G reinforced by a C and an E has a very different musical meaning from a G reinforced by a C and an E-flat; the first combination is a major chord and the second is a minor chord. They sound very different; yet the G is the same melody note either way. The meaning depends not on the melody, but on the harmony lines.
Body language, the visual harmony to the auditory melody, adds a great deal of meaning to communication. Most people use this visual harmony very instinctively. If you've ever observed someone talking on the phone, you may have noticed that they tend to use gestures, facial expressions, and body postures just as they would when they talk face-to-face.
Are we done describing communication? Not hardly. So far, we've only described what happens when one person talks. What about a conversation? Well, in music, as you may have noticed, it's not uncommon to have more than one identifiable melody. These melodies may be simultaneous, with different rhythms; they may alternately become prominent and then fade into the background; they may even sound as though they are in conflict.
Like a song with multiple melodies, a conversation combines the meanings of more than one linguistic melody; each melody takes meaning from the other melodies. You might hear repeated themes or modifications of old themes; you might hear contrasting themes; you might hear minor themes that support dominant ones.
A conversation is exponentially more complex than a single person speaking because the multiple melodies elaborate on each other, creating yet more meaning.
Surely we're done now, aren't we? We've got words, sentences, body language, and conversational reciprocity. But--no, we're still not done. Communication is more than just what you say, what you do, and how you respond to others. Conversations don't occur in a vacuum; they take place against a complex backdrop that includes things like generational identity, gender, social position, relationships, and culture. A conversation is filtered through a shared framework that goes all the way from the smallest social group to the largest society. You might say that these are conversational genres--the way there are musical genres, from classic rock 'n' roll to Gregorian chant. You expect certain things from a musical genre; there are patterns that tend to be followed, like the banjo-and-guitar twang of country music or the almost entirely rhythmic style of rap. Communicating between far-flung cultures can be as difficult as combining musical genres (and yet, may I note that it is possible).
Autistic people often find it harder to access one or more of the dimensions of communication. All autistic people, by definition, have problems learning one or more of these dimensions, from single words to the cultural context. However, knowing one needn't be a prerequisite for learning others; for example, someone may learn the meaning of rhythm and pitch before they understand the meaning of words, or learn to read facial expressions without realizing that they are embellishing spoken words. Some people understand the visual dimension much better than the auditory one, or vice versa; or they have a cognitive style that lets them process a limited number of channels at once--the many variations of the "I can make eye contact, but only if I don't listen" problem.
In general, my style of communication makes heavy use of words and sentences. By using very precise words, I improve my ability to communicate, similar to the way that someone with an extensive vocal range of, say, five octaves, is able to sing a much more complex melody than someone who has a range of only two octaves. I may have a talent for written communication precisely because I have a weakness in the aspects of speech which do not involve words; words and sentences are all that's involved in writing, and I've learned to use them to compensate for the non-verbals at which I am much more clumsy.
Neurotypical conversation often takes advantage of the multiple-channel, multiple-dimension nature of communication to convey a message using a relatively narrow range of words and a small array of simplistic sentence structures. With all the extra dimensions, NTs often don't need to be precise with words; the combinations of all the other signals are carrying enough of the message to fill in the gaps. But this can be problematic to the autistic, who often needs to be told things precisely so that he can gather the information that is carried by the channels to which he has problems "tuning in".
When communicating NT-to-autistic, I think it is important that both sides try to take stock of which channels are most available and which are easiest to use. Emphasizing these channels should make it easier to send the desired message.