Studying Greek Word Order: Comparing Classical & Koine

No, this post isn’t going to provide a comparison of how word order in Classical Greek differs from that of Koine Greek. Rather, I want to compare how such studies have been done in Classical Greek and Koine, specifically with regard to their use of statistics:

Ivan Shing Chung Kwong (2006), in his study of word order in Luke writes the following:

The statistical result of the study contributes to the figure of relative positions between the three main constituents in two aspects: unmarked word order patterns and tendencies of certain word order patterns–a very high percentage of occurrences of a certain word order pattern indicates an unmarked order (a very regular/typical order patter, e.g., subject precedes the complement in a clause [as stated below]); as relatively high percentage of occurrences of a certain word order pattern indicates a certain degree of tendency of having such a word order (the percentage of a certain pattern is not as high as the unmarked ones, but it is still relatively high to demonstrate a certain degree of tendency of having such a word order pattern, e.g., subject tends to precede its  predicate in independent clauses [as stated below]).

Kwong (2006, 45).

Now compare that with what Helma Dik says in her first book on Greek word order in Herodotus:

We have seen 48 instances in this section, many of which discussed in the previous sections of this chapter. Throughout, I have refrained from giving any statistics of the ordering patterns found, mainly because I do not think that statistics on the order of, for instance, the three core constituents would be very illuminating. In any case, the number of instances examined here is so small, that we cannot even expect ‘significant’ results in the technical sense. Be that as it may, Table 4.1 presents the distribution over the various ordering patterns as introduced in section 4.1.

Table 4.1 Ordering patterns with two arguments expressed

Pattern Instances Total
A1-A2-P 1.66.3; 1.188.1; 4.160.2; 6.7 4
A2-A1-P 3.1.1 1
A1-P-A2 1.161; 3.25.2; 4.173; 5.14.2; 6.28.1; 6.108.4 [3.19.3; 3.52.7; 3.151.1; 4.80.2] 6
A2-P-A1 3.44.1 1
P-A1-A2 1.73.1; 3.47.1 2
P-A2-A1 1.166.1 1

What conclusions can we draw from this distribution? At first sight A1-P-A2 [e.g. Subject-V-Object] appears to be the preferred pattern. If we include 3.19.3, 3.52.7, 3.151.1, and 4.80.2, this pattern covers more than half of the total number of instances. Does this make A1-P-A2 the unmarked order and should we proceed to consider only the marked instances? Clearly this would be a complicated procedure; with five different ‘marked’ patterns, there is a lot left to explain. In the preceding chapter I have tried to show an alternative way of handling the data. From this it appears that we need not describe one pattern as the preferred or unmarked pattern, but that differences in pragmatic function assignment can account for the different orderings. In such a description, we would not regard A1-P-A2 as unmarked, but explain its frequency from the fact that it is often the first argument that has Topic function and, therefore, ends up in initial position, and that in continuous narrative the predicate is the primary candidate for Focus function. Ironically, if we wanted to describe one pattern as unmarked in this description, it would be P-A1-A2, for it is very simple to describe the five other patterns in terms of ‘deviations’ (resulting from pragmatic function assignment) from this pattern.

In the clauses in which only one argument is present, we can see a similar phenomenon. A1-P is the ‘statistically unmarked’ option; I have argued that in A1-P the first argument is pragmatically marked.

Dik (1995, 93-4).

Kwong and Dik’s perspectives on word order, here, are polar opposites of each other. Dik is highly critical of the sort of statistics that Kwong has based his entire study upon. And, indeed, given her critical perspective on this matter, Kwong’s comments about Dik’s work, in his literature review, are rather striking:

The method and approach of Dik’s work is strange and largely disappointing. . . . Dik’s attempt to study the area pragmatically on the theory of Topic and Focus is a good try, but she fails to have a reasonable methodology and thus her work cannot give a satisfactory figure on Herodotus’ pragmatic word order (Kwong, 20).

Kwong’s perspective is simply incredible and out of step with the entirety of the field. The Bryn Mawr review of Dik’s monograph was quite positive, the exact opposite of Kwong’s review. The only explanation I can think of consists of a combination of (1) Kwong simply not understanding Dik’s analysis and (2) extremely high value that he places on statistics in his own study (the few books that he speaks positively of in his lit review tend to be statistical). The vast expanse between the two approaches, one statistical and the other wholly pragmatic seems to have also introduced a gap in comprehension.

Dik’s approach does far more justice to the data itself and is significantly closer to reality for Greek word order. In the end, Kwong’s work on word order will be viewed as idiosyncratic, which is unfortunate, because there is some good stuff in the book, particularly on semantic chaining and there’s also some good discussion of topic change in chapter 8. But the thrust of his argument on word order patterns is marred by a poor methodology. And it is not simply a reliance on statistics or placing too high a value upon that. Rather, it is Kwong’s interpretation of those statistics. He attributes discourse/pragmatic significance to the frequencies themselves without methodological justification.