Do some genres become more ‘complex’ than others?

1. Do some genres become more ‘complex’ than others? Javier Pérez-Guerra (jperez@uvigo.es) Ana E. Martínez-Insua (minsua@uvigo.es) Language Variation and Textual Categorisation Research Unit University of Vigo DGFS (Syntactic Variation and Emerging Genres), Siegen, Feb 2007

2. Introduction • McWhorter (2001:127): “it is a truism in linguistics in general that all languages are equally complex” Really? • cross-linguistic counter-hypothesis: Languages can be graded according to their complexity. • intra-linguistic counter-hypothesis (central assumption of this paper): Aspects within a language (genres or text types, historical stages, etc.) can be graded according to linguistic complexity.

3. Goal of this paper • exploration of linguistic complexity in two text types in the recent history of English • analysis of the unmarked (preverbal) subjects (external arguments) and unmarked (postverbal) objects (internal arguments) of declarative sentences in a corpus 1750-1990 corpus: ARCHER (British component) • periods: 1750-1799, 1850-1899, 1950-1990 • text types: - news (written-to-be-read/spoken): formal, written, public - letters: more informal, written~speech-based, public~private

4. The corpus Table 1: The corpus (word totals for the subjects and the objects)

5. The corpus Table 2: Pronominal and non-pronominal subjects (percentages per text type and period)

6. The corpus Table 3: Pronominal and non-pronominal objects (percentages per text type and period)

7. Assumptions and ‘hallmarks’ • garden-path approach: Even though the sentences can be ambiguous, the human parser interprets unambiguously the linguistic input. (ii) text types encode linguistic features and differ in complexity (Taavitsainen 2001:141). (iii) complexity is influenced by linguistic ‘circumstances’ and is not inherent to the clauses (Crain & Shankweiler’s 1988 Processing Deficit Hypothesis)

8. Assumptions and ‘hallmarks’ (iv) complexity as a relational (than-) notion (v) complexity as a relative notion: Frazier (1988:204): “there is no general unit of complexity (...) which would predict in ‘absolute’ terms the complexity of a sentence” => several metrics

9. Assumptions and ‘hallmarks’ • importance of the subject (external argument) and the object of a sentence (internal argument(s)) as far as the determination of complexity is concerned. E.g.: · Davison & Lutz (1985:60): “the high load of processing would occur in subject position” · Gibson (1998:27): “modifying the subject should cause an increase in the memory cost for predicting the matrix verb”

10. Concept of *complexity Uses of complexity this paper is not about: • complexity as linguistic richness: McWhorter (2001): “an area of grammar is more complex than the same area in another grammar to the extent that it encompasses more overt distinctions and/or rules” • complexity as linguistic explicitness or transparency: Rodhenburg (1996): I help him to write the paper. [more complex] I help him write the paper.

11. Concept of *complexity • conceptual complexity: Gibson’s (1998, 2000) Dependency Locality Theory: # of new referents (before our marker) has consequences for the ‘integration cost’ of a constituent • informative and/or cognitive complexity: giveness/newness, animacy, etc. • complexity as processing difficulty: McWhorter (2001): “all languages [in all periods] are acquired with ease by native learners” • derivational complexity

12. The metrics [lack of speakers’ intuitions in diachronic research => text/evidence-based methodology, independently of uses and users] • size/length: · Wasow (1997:81): grammatical weight implies “size of complexity” · Yaruss (1999:330): “attempts to separate length and complexity are somewhat artificial” • metric1: # of words of the subjects • metric2: # of words up to the ‘marker’ of the rightmost immediate constituent

13. The metrics ‘Markers’ • assumption: concept of ‘incrementality’: “the language processing system must very rapidly construct a syntactic analysis for a sentence fragment, assign it a semantic interpretation” (Pickering et al 2000:5) • concept: markers alone can characterise the syntactic status of the constituents to which they belong (~ Chomsky’s syntactic heads). The identification of the markers also relies on statistical information (Corley & Crocker 2000:137).

14. The metrics ‘Markers’: examples • Your Ladyship dares me to stop in my new work! (1751Richardson.X3) [determiner as the marker of the noun phrase] • a bust family for the children is like a solar system without a moon. (1951Durrell.X9) [preposition as the marker of the prepositional phrase] • Helen & Bill, by the way, send their fondest regards to you both. (1950Thomas.X9) [conjunction as the marker of the coordinating construction] • the humility which you laud in a character such as that of Macready has always to me a certain falseness about it – (1876Trollope.X6) [wh-proform as the marker of the wh-clause]

15. The metrics ‘Markers’: examples (cont.) • Nato’s first mission was now complete (1989TIM1.N9) [’s as the marker of the possessive phrase] • the apotheosis of Scobie – culminating for me in the shower of rockets from H.M.’s Navy – is sublimity. (1960Aldington.X9) [ing-form as the marker of the nexusless nonfinite clause] • The declaration of neutrality demanded by the Minister of France, might have been considered as superfluous [ed-form as the marker of the nexusless nonfinite clause] • pleasure-seekers are notoriously the most aggrieved and howling inhabitants of the universe, (1869Eliot.X6) [noun as the only element in the subject noun phrase]

16. The metrics • density: • metric3: number of immediate constituents • metric4: ratio of words per immediate constituent

17. The metrics • depth: • metric5: # of abstract (non-terminal nodes) in a ‘simple’ (non-derivational) syntactic analysis • assumption: few non-terminal nodes implies weak complexity • phrases (1) and (2) differ as far as complexity is concerned: • the spy with binaculars from Italy (‘the spy is from Italy’) • the spy with binaculars from Italy (‘the binaculars were made in Italy’)

18. The metrics (1) the spy with binaculars from Italy the spy from Italy with binaculars 3 non-terminal levels (Minimal Attachment, Frazier 1979)

19. The metrics (2) the spy with binaculars from Italy the spy with binaculars from Italy 4 non-terminal levels (Late Closure in Frazier 1979 or Recency in Gibson et al 1996)

20. The metrics • depth (cont.): • metric6: non-terminal-to-terminal ratio: amount of structure that is associated with the words of a constituent

21. The metrics • (lack of) efficiency: • metric7: ratio words-up-to-the marker / immediate constituents, inspired by Hawkins’ (1994) IC-to-word ratio • metric8: on-line IC-to-word ratio, based on Hawkins’ (1994)(aggregate of the partial divisions of the # of immediate constituents by the # of words of such a constituent (up to the marker))

22. Analysis of the data Table 4: Metrics for the subjects

23. Analysis of the data Table 5: Metrics for the objects

24. Preliminary results • metric1(no. of words) • objects are much longer than subjects, which accords with end-weight • subjects and objects are longer in the news => complexity (further research: only non-pronominal constituents); no diachronic variation • metric2 (no. of words till the marker) • the figures for subjects and objects are more alike => after the results of metric1, only the post-marker segment in the objects is longer (and not the whole constituent) • similar basic processing requirements in subjects and news • the material previous to the marker is only slightly longer in the news => somewhat more complex than the letters

25. Preliminary results • metric3 (no. of ICs) • similar results in subjects and objects • similar results in letters and news • metric4 (words/ICs) • the figure is considerably higher in objects; now… • if metric2 showed that the material previous to the marker is not longer • if metric3 showed that the number of ICs is similar in subjects and objects then onevery long post-marker IC in the objects accounts for the difference (see metric7) • the news contain even longer post-marker ICs (see metric7) => more lexical complexity

26. Preliminary results • metric5 (number of intermediate nodes) • the figure for the objects is higher than that for the subjects; in the light of metric1, the difference is proportional to the overall length of the objects (so, the post-marker material in the objects is not necessarily more complex, on syntactic grounds) • more non-terminal nodes in the news; in the light of metric1, the difference is proportional to the overall length of the constituents in the news (so, the news are not necessarily more complex on syntactic grounds) • metric6 (non-terminal/terminal nodes) • the statistical difference between subjects and objects is not significant; diachronic variation is not significant => no differences of syntactic complexity • the statistical difference between letters and news is not significant; diachronic variation is not significant => no differences as far as syntactic complexity is concerned

27. Preliminary results • metric7 (=metric4 up to the marker) • the differences between subjects and objects are not statistically significant; this corroborates the view that the pre-marker material is neither lexically nor syntactically more complex in the objects • the figures in the news are slightly higher than in the letters => not only the size (see metric2) but also the syntactic configuration of the pre-marker material are somewhat more complex in the news

28. Preliminary results • metric8 (on-line IC-to-word ratio) • the values for the objects are lower than for the subjects; this average ratio corroborates what metric4showed, namely that the number of words per IC was higher in the objects and that this was due exclusively to the post-marker IC • similar results of the letters and the news; if the ICs are larger in the news (metric4) and metric8 does not take into account the post-marker material, then the rightmost constituents in the news are particularly responsible for unbalancing the syntactic distribution of the constituents => (not syntactic) lexical complexity of news

29. Preliminary results • general diachronic tendency: weak drift towards more lexical complexity in the news in the latter periods (see metrics 1, 4, 5, 7)

30. Concluding? remarks • pilot-study => care! • hypothesis: text-types can be linguistically characterised and can be placed on a scale of complexity by investigating the (linguistic) complexity of the clausal constituents • minor diachronic differences between Late Modern and Present-Day English; only the news evince a weak drift towards more lexical complexity

31. Concluding? remarks • minor differences of syntactic complexity between (the subjects and the objects of) letters and news both synchronically and diachronically – the news have evinced a slightly higher degree of syntactic complexity (before the marker); subjects and objects differ in lexical (not syntactic) complexity (end-weight is correct) • the news contain constituents (subjects and objects) which are lexically more complex or dense (particularly after the marker)

32. Concluding? remarks • Beaman (1984:46): “spoken language is just as complex as written” • Halliday (1985:62): “each [sub-language] is complex in its own way. Written language displays one kind of complexity, spoken language another (...) the complexity of written language is lexical, while that of spoken language is grammatical”

33. Further research • more subjects, objects • more text types (Biber 1992:158: “[w]ritten registers differ widely among themselves in […] complexity, whereas spoken registers follow a single pattern with respect to their kinds of complexity”) • also marked (‘moved’, non-preverbal) subjects, subjects in passive sentences and (‘moved’, non-postverbal) objects • fine-grained syntactic analysis: • differences of adjuncts (modifiers) and arguments (complements) • differences of right- and left-adjunction/branching

34. Do some genres become more ‘complex’ than others? Javier Pérez-Guerra (jperez@uvigo.es) Ana E. Martínez-Insua (minsua@uvigo.es) Language Variation and Textual Characterisation Research Unit University of Vigo DGFS, Siegen, Feb 2007

35. References Beaman, Karen (1984) “Coordination and subordination revisited: syntactic complexity in spoken and written narrative discourse”. Ed. Deborah Tannen. Coherence in spoken and written discourse. Norwood: NK: Ablex (45-80). Biber, Douglas (1992) “On the complexity of discourse complexity: a multidimensional analysis”. Discourse Processes 15: 133-163. Corley, Steffan and Matthew W. Crocker (2000) “The modular statistical hypothesis: exploring lexical category ambiguity”. Eds. Matthew W. Crocker, Martin Pickering and Charles Clifton Jr. Architectures and mechanisms for language processing. Cambridge: Cambridge University Press (135-60). Crain, Stephen and Donald Shankweiler (1988) “Syntactic complexity and reading acquisition”. Eds. Alice Davison and Georgia M. Green. Linguistic complexity and text comprehension: readability issues reconsidered. Hillsdale, NJ: Lawrence Erlbaum (167-192). Davison, Alice and Richard Lutz (1985) “Measuring syntactic complexity relative to discourse context”. Eds. David R. Dowty, Lauri Karttunen and Arnold M. Zwicky. Natural language parsing. Psychological, computational, and theoretical perspectives. Cambridge: Cambridge University Press (26-66).

36. Frazier, Lyn (1979) On comprehending sentences: syntactic parsin strategies. Blooomington, In.: Indiana University Linguistics Club. Frazier, Lyn (1985) “Syntactic complexity”. Eds. David R. Dowty, Lauri Karttunen and Arnold M. Zwicky. Natural language parsing. Psychological, computational, and theoretical perspectives. Cambridge: Cambridge University Press (129-189). Frazier, Lyn (1988) “The study of linguistic complexity”. Eds. Alice Davison and Georgia M. Green. Linguistic complexity and text comprehension. Readability issues reconsidered. Hillsdale, NJ: Lawrence Erlbaum (193-221). Gibson, Edward, Neal J. Pearlmutter, Enriqueta Canseco-Gonzalez and Gregory Hickok (1996) “Recency preference in the human sentence processing mechanism”. Cognition 59: 23-59. Gibson, Edward (1998) “Linguistic complexity: locality of syntactic dependencies”. Cognition 68/1: 1-76. Gibson, Edward (2000) “The dependency locality theory: a distance-based theory of linguistic complexity”. Eds. Alec Marantz, Yasush Miyashita and Wayne O’Neil. Image, language, brain. Papers fron the First Mind Articulation Symposium. Cambridge, MA.: MIT (95-126). Hawkins, John A. (1994) A performance theory of order and constituency. Cambridge: Cambridge University Press. Hawkins, John A. (2004) Efficiency and complexity in grammars. Oxford: Oxford University Press.

37. McWhorter, John H. (2001) “The world’s simplest grammars are creole grammars”. Linguistic Typology 5: 125-166. Miller, George A. and Noam Chomsky (1963) “Finitary models of language users”. Eds. R. Duncan Luce, Robert R. Bush and Eugene Galanter. Handbook of mathematical psychology. Vol. 2. New York: Wiley (419-492). Pérez-Guerra, Javier and Ana E. Martínez-Insua (2006) “Subjects and complexity in the recent history of English”. Paper read at DELS, University of Manchester, April. Pickering, Martin J., Charles Clifton Jr. and Matthew W. Crocker (2000) “Architectures and mechanisms in sentence comprehension”. Eds. Matthew W. Crocker, Martin Pickering and Charles Clifton Jr. Architectures and mechanisms for language processing. Cambridge: Cambridge University Press (1-28). Rohdenburg, Günter (1996) “Cognitive complexity and increased grammatical explicitness in English”. Cognitive Linguistics 1/2: 149-82.Wasow, Thomas (1997) “Remarks on grammatical weight”. Language Variation and Change 9/1: 81-105. Yaruss, J. Scott (1999) “Utterance length, syntactic complexity, and childhood stuttering”. Journal of Speech, Language, and Hearing Research 42/2: 329-344.