Sentence Complexity

Managing the complexity of sentences means managing the difficulty of comprehension.

Complexity and Clarity

Writers generally acknowledge that our primary aim is to be understood. We recognise that it is not the reader’s responsibility to understand; it is our responsibility to be clear. We also know that clarity is a complex issue, and that achieving clarity means achieving control over many aspects of language-based communication.

Aspects to be controlled include:

Sufficiency of information (providing needed information, but not restating what the reader already knows);

Relevance of terminology (using technical terms the reader would use, but not adopting inappropriate jargon);

Appropriateness of expression (using linguistic structures that reflect information relationships).

The last of these is itself complex. The language we use does more than reflect relationships between elements of information; it also determines relationships between the information and the reader. Thus, quite apart from issues of vocabulary, our choice of forms of expression determines the tone and style of our writing, as well as its clarity.

It is this last issue — how our choice of forms of expression affects the clarity of our writing — that I focus on here. Specifically, I want to consider the link between structural complexity and comprehensibility.

The nature of Sentence Complexity

By using the phrase "structural complexity", I may appear to have imposed a restriction on the subject. However, in fact all I have done is make the obvious manifest. In language, all complexity is structural. Complex words are complex precisely because they combine multiple elements of meaning at one level; complex expressions are complex precisely because they combine multiple elements of meaning at a different level. Let me illustrate, first with a word.

Dispossession — not a particularly complex word, but sufficiently so — combines the central idea of possess (a verb meaning own, have, hold) with two extra ideas: we have turned the action into a "thing", so that we can talk about it, by adding -ion; and we have changed the meaning into "removal of that thing", by the addition of dis-. Notice that the first of those changes is done for grammatical reasons, while the second is "semantically motivated".

Now consider these two sentences:

Set the master distribution panel control switch to the "standby" position.

The pressure on the inlet side of the valve is indicated by Gauge 3, which is to the right of the main control valve.

Again, neither example is particularly complex, but will serve for illustration.

In each of these cases, we have a basic idea that is being elaborated in some way. In (1), we have extra detail answering the question "which switch?" In (2), we have extra detail answering the questions "what pressure?" and "where is the gauge?" Contrast these examples with sentences lacking the elaboration:

Set the switch to the "standby" position.

The pressure is indicated by Gauge 3.

We are immediately struck by the need for the extra detail: the simplified sentences on their own tell us very little. We can also see that the extra information could be added in separate sentences:

The control switch is on the master distribution panel. Set the switch to the "standby" position.

Check the pressure on the inlet side of the main control valve. The pressure is indicated by Gauge 3. The gauge is to the right of the valve.

(We have had to "invent" a purpose in (6) by using check in the first sentence. A larger extract from the original text might have made this unnecessary — or, at least, legitimate; as it is, this is not relevant to the question of complexity.)

Interestingly, (5) seems to be just as acceptable as (1), but (6) seems a stilted and simplistic version of (2). Even more interestingly, we can see that (1) and (2) are complex in ways that are somehow different. In both cases, we have the option of adding meaning — adding semantic content — either by adding additional sentences or by adding complexity to the basic sentence. Thus, as with words, we have the ability to adopt largely semantic measures (adding semantic units, or sentences) or essentially grammatical measures (manipulating the grammatical complexity of the existing sentence). In one case, there seems little to choose between the two; in the other case, the essentially grammatical measure seems preferable.

Forms of Complexity

What we have seen so far is that we can increase the complexity of information in a sentence in two ways: we can add sentences, or we can add to the complexity of the original sentence. Closer examination of our original examples shows more.

In (1), the added complexity is in the string of nouns-as-adjectives master distribution panel control switch. In (2), the added complexity is in the prepositional phrase on the inlet side of the valve and the relative clause which is to the right of the main control valve. Note further that on the inlet side of the valve contains another prepositional phrase of the valve, and which is to the right of the main control valve contains to the right of the main control valve, which in turn contains of the main control valve.

These two forms of "elaboration" can now be seen as quite different. In (1), we see extra detail being piled on at a single level — changing the noun switch to the extended noun phrase master distribution panel control switch. In (2), we see extra detail being built up at multiple levels, with elements containing other elements. The first of these is a form of parataxis, or linear extension; the second is a form of hypotaxis, or extension of hierarchical depth.

When we rewrite (1) as (5), we are creating a new hierarchical layer at the level of sentence group, or paragraph. (A paragraph is not a grammatical unit, but it is certainly a unit in which language structures reflect information structures.) When we rewrite (2) as (6), we are stripping one hierarchical layer within the sentence, then adding it back at the paragraph level; we have made no change to the hypotaxis, but by adding sentences we have increased the number of units at that level — we have increased the paratactic complexity. So changing (1) to (5) replaces paratactic complexity with hypotactic complexity, and the result appears acceptable; changing (2) to (6) replaces hypotactic complexity with paratactic complexity, and the result appears unsatisfactory.

We must now ask: is this apparent difference real? Can we account for the difference, finding a basis for relating complexity to clarity? I believe that we can.

The effect of complexity on comprehension

Before exploring the specific mechanisms in detail, we must consider the way in which readers (or hearers) process language. We read or hear one word at a time, but we do not process a text purely in that linear fashion. Language works as it does because of a principle called constituency.

Hierarchy in language structure

Constituency refers to the fact that linguistic elements can be combined into structures which can in turn be combined into larger structures; those larger structures can be combined into larger structures again; and so on, more or less ad infinitum.

We can perform a "constituent analysis" of sentence (2) to show how this works. In such an analysis, we use square brackets to delimit each syntactic unit:

[The pressure [on [the inlet side [of [the valve]]]]] [is indicated [by [Gauge 3, [which [is [to [the right [of [the main control valve]]]]]]]]]

Within this sentence, we can see that the simple noun valve (first occurrence) is in a noun phrase the valve; this is part of the prepositional phrase of the valve; that’s part of the noun phrase the inlet side of the valve, which is part of the prepositional phrase on the inlet side of the valve; the whole thing is part of the noun phrase The pressure on the inlet side of the valve, which functions as the subject of the main clause. That simple noun is buried at the bottom of a pile five layers deep. In its second occurrence at the end of the sentence, valve is nine layers deep! We can see that (2) is, indeed, complex, yet we have no problem understanding it. Why not?

Processing hierarchical information

The point about constituency is that, as we read or hear a sequence of words, we reconstruct the larger structural combinations in our heads. The amount of processing involved in deconstructing an idea into words, then reconstructing the words into the original idea, is quite remarkable, given that we do it in "real time". (So remarkable, in fact, that we can hardly be surprised when communication fails; perhaps we should be surprised that it ever succeeds!)

Sentence (2) is 24 words long. That is (just) within the bounds of what is normally considered acceptable length. This does not mean that we can remember all 24 words with little difficulty: it means that we can process them, building intermediate structures "on the fly", so that we achieve an understanding of the overall idea — with little difficulty. In (2), the reader must reassemble concepts through first five and then nine iterations. This is a tall order, but it seems to be entirely manageable.

At the same time, we know that some sentences are difficult to process. What is it about (2) that makes its complexity acceptable? The answer lies in what happens when we have processed the last element in a structure — when we reach a closing bracket in (7), for example.

Closure

As we encounter a word, one of several things can happen. The word might be the first in a new structure, or it might not. If it is the first, then we must start forming a new concept, but we must also decide how the new structure relates to any existing but incomplete structures. If the new structure is at a deeper level of nesting, then we must open up a new sub-concept within the current concept being developed. If it is at a shallower level of nesting, then we can complete development of at least the most recent concept or sub-concept.

Once we have taken appropriate action with concepts already under development, we can start processing the new structure, developing a new concept or sub-concept. Initially, this involves merely setting up some expectations — for example, we know that the is always followed by a noun or an adjective.

If the word we have just encountered is not the first in a new structure, then we use it to refine and clarify our expectations. For example, at the start of a sentence, we know that a noun is usually followed by a preposition or a verb.

Thus, in (2), The tells us that we are at the start of a noun phrase; pressure satisfies our expectation of a noun; on opens up a new nested structure, so we hold open the noun phrase and start developing the new sub-concept. This new sub-concept is not distinct from, but is embedded in, the already open structure. So we go on, until we reach the verb is, when we know that the noun phrase functioning as subject is complete. We can now "wrap up" our concept of the subject, and get on with finding out what the writer has to say about it.

Until we encounter a signal telling us that a structure is complete, then, we keep it open. We are prepared to keep developing the concept until we receive such a signal, at which point the concept is closed.

Managing complexity

The relationship of closure to clarity should now be clear. Technical communication is, above all, about helping our readers to acquire structured information — or, to put it another way, to develop information structures in their own minds. Informational or conceptual structures map directly to linguistic structures; clear writing is a clear reflection of information structures, and encourages the orderly development of structured conceptions.

Maximising Closure

As can be seen from the discussion of closure, the demands placed on the reader’s processing capacity increase with linguistic complexity. However, hypotactic (hierarchical) complexity allows a progressive development of a basic concept at increasing depth of detail. Paratactic (linear) complexity, on the other hand, requires that the reader hold multiple concepts under development. Consider the next example:

Gauge 3, to the right of the main control valve, and gauge 4, above and to the right of gauge 3, should show equal pressures.

Sentence (8) is only one word longer than (2), but is considerably harder to manage. The reader must build a complete concept about gauge 3, but keep it available until the concept about gauge 4 is also complete. These two concepts together form a larger concept, which is the subject of the whole sentence. We thus have a closed concept (gauge 3 with its details) within an open concept (the subject of the sentence). This is an example of paratactic complexity, where the closure of a concept does not close the higher-level concept. In a bracketed constituency analysis, we would see a closing bracket followed by an opening bracket below the topmost level (ie, other than between subject and predicate, where it is not a problem).

In (7), we do not see any such evidence of paratactic complexity. On the contrary, we see multiple and complete closure after each appearance of valve — five closing brackets in succession, and then nine in succession. As the lowest-level structure is closed, all the higher-level structures containing it are also closed.

We can now see that the two forms of complexity — paratactic and hypotactic — are quite different in their effects. At first glance, parataxis appears simpler than hypotaxis; so, indeed, it is at any one level. However, language cannot function without hypotaxis, so parataxis can only happen at one of a number of hierarchical levels. At the level of the complete sentence, parataxis is neither problematic nor avoidable. At lower levels, it conflicts with the objective of maximal closure, and is to be avoided.