Traditional indices estimate text difficulty by measuring sentence length. Here, in considering the effect of syntax, we have a similar problem to that of vocabulary; correlation and causation.
"As Klop (1978) points out: It may seem surprising that counts of the two simple variables of word length and sentence length are sufficient to make relatively good predictions of readability. No argument that they cause ease or difficulty is intended: they are merely good indices of difficulty. Consequently, altering word or sentence length, of themselves, can provide no assurance of improving readability. How to achieve more readable writing is another and much more complex endeavour.......” (Davies, 1984:188-9)
It is hardly surprising that sentence length correlates with difficulty. Apart from memory considerations, longer sentences are likely to contain more complex structures such as coordination and subordination (Beaman 1984).
As long ago as 1964 Coleman pointed this out:
"It is almost certain that sentence (or clause) length can predict readability only because it is correlated with more fundamental predictors of syntactic complexity such as nesting, transformational complexity, and others (Miller & Chomsky, 1963)." (Coleman 1964:190)
If it is indeed true that certain structures do cause more difficulty than others, what we would like to know is which they are and why they cause difficulty. Chomsky, C (1969) provides a theoretical consideration of linguistic complexity. She postulates that difficulty of interpretation of the grammatical relations which hold among the words and phrases of a sentence is increased by the presence of four conditions:
"(A) The true grammatical relations which hold among the words in a [sentence] are not expressed directly in its surface structure.
(B) The syntactic structure associated with a particular word is at variance with a general pattern in the language.
(C) A conflict exists between two of the potential syntactic structures associated with a particular verb.
(D) Restrictions on a grammatical operation apply under certain limited conditions only" (Chomsky, 1969:6,7)
However, it is difficult to see how this can be translated into a workable (i.e. fast and simple) model for assessing readability. It may also be true that learners from different L1 backgrounds find different structures more or less difficult, but for the moment we will think about the question in general terms.
Coleman (1964) found that transforming nominalizations, adjectivalizations and passives to their active verb transforms improved comprehensibility of texts. Others found similar results (Bhatia, 1984) and it has long been an article of faith that structures such as nominalizations are difficult to decode (Klare, 1985; Price, 1984). Berman uses the notion of "heaviness" to describe items which may cause processing difficulties.
"Note that the notion of 'heaviness' is not a straightforward function of linear length in any simple terms. Rather, the problem seems to concern the amount and depth of information which the reader must store in memory in moving from one construct to the next, and how hard the transition becomes as a result. And in fact, in the experiment [..], students said things like: 'I was so busy working on this part of the sentence, I forgot it was connected with something else'
Heaviness may also occur where the basic NVN or 'kernel' structure is violated by a process such as nominalization - there are grounds for believing that nominalizations are often more complex than corresponding sentences with simple verbs or adjectives." (Berman, 1984:142-3)
However, experimental rewriting of EST texts by Strother & Ulijn (1987) showed that syntactic simplification had no significant effect on comprehension (but see criticism by Coady (1987:101-103) of their statistical analysis)
Ulijn & Strother (1990) conducted similar experiments which largely confirmed their previous results.
"…at advanced levels, syntactic simplification into a more common register does not really increase readability." (Ulijn & Strother, 1990:49)
Their restriction of these findings to advanced levels is probably suitably cautious:
"It may also be true that syntactic simplification may have an effect on groups other than those tested in this experiment, such as limited proficiency language users." (Ulijn & Strother, 1990:49)
However Berman seems to suggest that syntax complexity is an important factor and that
“efficient FL readers must rely in part on syntactic devices to get at text meaning.” (Berman, 1984:153)
Cooper found that:
"….practised readers are not distinguished clearly from unpractised readers by their ability to understand the meaning carried by syntax." (1984:130)
Laufer in her review of the literature stated:
"Laufer and Sim (1985a, 1985b) found that, in interpreting texts, students tend to regard words as the main landmarks of meaning. Background knowledge is relied on to a lesser extent, and syntax is almost disregarded." (Laufer, 1997(a):21)
And Ulijn and Kempen suggested that L1/L2 contrasts were not a problem as far as syntax was concerned.
"Under normal conditions reading comprehension is little dependent on a syntactic analysis of the text’s sentences. It follows that second language reading comprehension is possible without mastery of the second language’ syntax. Usually the reader’s conceptual knowledge will compensate for the lack of knowledge about linguistic contrasts between L1 and L2.” (Uliyn and Kempen (1976:499) cited in Alderson, 1984:12)
Perhaps we could say that low proficiency L2 reading does not constitute "normal conditions" and that complex syntax may cause problems for less proficient readers. Chaudron (1983:437) found that "Learners with relatively low English proficiency tended to have poorer recall ability on the syntactically more complex structures." Similarly, Barnett found in her experiments that "recall increases according to level of syntactic proficiency" (1986:346).
So in examining readability (especially if syntactic simplification is envisaged) the proficiency of the intended readership has to be taken into account.
Reading purposes may also have to be taken into account. While reading for gist may not require a thorough syntactic analysis,
"…if our aim includes students' acquisition of specific information accurately and in detail - which is the case with much scientific or technical material - exact appreciation of syntactic components of each sentence remains an important aim" (Berman 1984:146)
One problem with trying to simplify texts syntactically (whether for experimental purposes or to facilitate readability) is that it is difficult to change a text on one level without changing it on another. Simply breaking long sentences up into smaller ones while it may affect conventional readability indices does not make a text more readable as Blau discovered:
"..short, primarily simple sentences typical of low readability levels of the version 1 passage actually are an obstacle to comprehension. Choppy, unnatural sentences are difficult to read and the relationships and meaning revealed by the formation of complex sentences are apparently lost. Readers do indeed seem to benefit from the information regarding relationships that is revealed by complex sentences." (Blau 1982:525)
Thus making sentences shorter will necessarily increase the number of sentences thereby perhaps increasing the number of clausal relations which have to be inferred or explicitly signalled. The distance between anaphors and their antecedents may also increase rendering anaphor resolution more difficult.
So if we assume that for some readers (low ability or non-proficient L2 learners) syntactic complexity does cause problems. What syntactic features may affect readability?
Some have questioned the widely accepted notion that subordination is a cause of complexity: Beaman claimed her investigation showed "..that the basic assumption of many linguists in the past, i.e. that subordination implies complexity, is false. The evaluation of syntactic complexity is simply more complex than that." (1984:79,80). Similarly, Schleppegrell: "Linguistic complexity, although used as an indicator of language skill or of higher levels of linguistic development, is a construct which is not yet well defined. Sentence-level indicators, such as the use of subordinate clauses, are not adequate as measures of linguistic complexity." (1992:129)
It is also a common assumption that written language is more complex than spoken. Beaman has demonstrated that this is not the case. But one difference between written and spoken language which may cause difficulties is the increased lexical density manifested through non-finite subordinate clauses and complex noun phrases. (Halliday, 1979, cited in Beaman, 1984:50).
Many students learn English for scientific and technical purposes. The genres involved are often far different from much of what the students have previously been exposed to either in their own language or in the texts which have been used in their previous language learning activities. These genres are often characterised by nominalization of processes and the use of complex noun phrases (Bloor & Bloor, 1995:222). And it may be the case that any syntactic device to pack more propositional content into fewer words (e.g. nominalization, use of verbal nouns (Rutherford 1987:50,51)) contributes to processing difficulties.
But it is one thing to pick out particular difficulties which individual learners may have with specific genres, it is another to say why they have these difficulties or point out an underlying reason why certain features cause difficulties in general. The problem with simply looking at the text is that we assume that all sources of difficulty lie with the text itself. While this may be a reasonable approach in the investigation of complexity as far as native speakers are concerned, it does not take into account extra difficulties that L2 learner may experience because of their developing (but as yet undeveloped L2) competence or, perhaps, because of L1/L2 contrastive difficulties. It might be, for example, that if we could plot the course of the learner's developing interlanguage, we could make some predictions about syntactic features which might cause difficulties.
The assumption, complexity = difficulty, may have been debunked as far as native speakers are concerned but it may by the case that for L2 learners, complexity just adds to processing constraints, compounding any other problems the reader has. It is difficult to hold a great deal in working memory once you come across a difficulty and stop to fathom it out.