Expectation-based Processing: Chinese classifiers
General idea - Definite or enumerated Chinese NPs are preceded by classifiers that depend on the semantic class of the intended meaning of the noun. For example, the Chinese word for "telephone" (dian4hua4) can either refer to an individual phone, a type of phone, or a phoning event (ie a phone call), depending on what type of classifier is paired with the noun. Whether an NP evokes an individual, kind, or event seems grammaticized in the Chinese classifier system. (See work by Ahrens FMI.)
Crucially, there also seem to be cases where producers have a choice between two classifiers for the same noun without causing much meaning difference (eg the WEI/GE option for people Ns). In cases like this, one of the possible classifiers is more restrictive (WEI teacher) than the other (GE teacher), in the sense that the latter is a more GENERAL classifier, while the former can be paired only with a more limited set of nouns (people, in this case). Since the classifiers precede the noun, we hypothesize that comprehenders can integrate the information provided by classifiers during incremental comprehension to hone in on the noun they expect. This should lead to increased processing speed for specific classifiers (WEI) compared to general classifiers (GE) when comprehenders reach the noun.
For more detail continue to the [wiki:/HlpLab/Projects/ChineseClassifiers/LabLog lab log]
Design - This self-paced reading study crosses classifier (WEI or GE) with presence or absence of an intervening relative clause (RC or NORC). All critical nouns are people nouns, which are biased to appear with WEI in Google search results. We chose to use the Linger software to present items.
Critical Items - 24 locative-inversion sentences with people nouns were written as critical items. Here is an example item from the study: 公园里有一个/位(所有人都注意到了的)尼姑在打坐。 park-inside there is one GE/WEI (everyone is paying attention to DE) nun meditating. In the park there is a nun (to whom everyone is paying attention) meditating.
We specifically selected Chinese people nouns for which the specific classifier WEI seems informative; in other words, we used nouns that appear more frequently with WEI than GE in Google searches.
We used locative-inversion sentences to minimize any expectation for a person noun conveyed by the sentence material aside from the classifier. Object-extracted relative clauses (ORCs) were used for this same reason, since both locative-inversions and ORCs were relatively easy to write with neutral contexts, and we took care to do this. Sentences never ended with the critical noun, because we did not want wrap-up effects to confound our results.
In the course of the study, participants see the following type of critical items: 6 Locative ORC GE 6 Locative ORC WEI 6 Locative no-RC GE 6 Locative no-RC WEI
Filler Items - 66 filler items of various kinds were included to counter-balance the critical items. Because locative-inversion structures are infrequent, 54 of the fillers did not have this construction. We also included 6 fillers that have locative-inversions that are not followed by an ORC, and have an inanimate critical noun (with GE). This is because we did not want to train participants that locative-inversions will be paired with people nouns.
We added 18 fillers with subject-extracted relative clauses (SRCs) modifying people nouns so that participants see a total of 18 ORCs and 18 SRCs during the course of the study. Participants see 54 sentences with no RCs at all, reflecting somewhat the fact that RCs are not ubiquitous.
In addition to the ORC fillers, 6 additional fillers have GE followed by an inanimate noun. These reflect the fact of natural language that GE is more generic. We did not want to re-train comprehenders to feel that GE is a specific classifier for people only.
We included 12 fillers that have inanimate nouns with their corresponding specific classifiers (eg "TIAO fish"). This is so WEI is not the only specific classifier participants encounter in the course of the study.
Here is a summary of the types of filler items participants see: 12 no-loc no-RC no-CL inanimate proper noun 12 no-loc no-RC spec inanimate noun 12 no-loc SRC WEI person noun 6 no-loc SRC WEI person noun 6 Loc ORC GE inanimate noun 6 Loc no-RC GE inanimate noun 12 no-loc no-RC no-CL person noun
Total Item Summary - 36 Locatives, 48 no-Loc 18 ORC, 18 SRC, 54 no-RC 24 GE, 30 WEI, 12 other specific, 24 No-CL 54 person, 36 inanimate
Comprehension Questions - All 90 items were followed by a comprehension (Y/N) question. Feedback is given to participants. This allows us to encourage careful reading, as well as to exclude data from participants who do not understand the material. Our questions are counter-balanced to refer to different parts of the sentence, so that participants do not learn that one part of the sentence is more "important" to answering questions than others.
Critical items have the following type of questions: 1. Was Noun(12.5% wrong) in LOC(12.5% wrong)? 2. Did Noun(12.5% wrong) do Verb(12.5% wrong)? Half the answers are Y and half are N. The N answers are wrong in one of the 4 ways mentioned.
For fillers w/o LOC or RC(36): Did X verb Y? (if Y is inanimate, 25% verb wrong; if Y is animate, 25% exchange X and Y)
For fillers with RCs(24): Ask questions about RC content. Wrong ones have the subject (the head noun) and object reversed, if the object in the RC is animate. Otherwise, the verb is wrong.
For fillers with LOC but w/o RCs(6): Was N in LOC? (one object wrong, one LOC wrong, one both object and LOC wrong)