暗微幽光

Greek Nonword Reading

This presentation deeply connects with the orthography-to-phonology correspondence issue which has been broadly investigaged in English literatures but infrequently cared in the other lanaguages. Protopapas was interested in a graphophonemic case of Greek, his native language, that might be stored as a form of cateograical rule in Greek readers’ mind. In addition to 95% of feedforward consistent graphophonemic correspondences, the /i/ in the syllable strucute CiV could be pronunceable or read palatally. Protopapas attempt to confirm wheather the the pronunciation of nonwords reflect the regular rules compromised the real words. This observation would be evident when the nonwords were generally pronunced as /i/ or palatal in consideration of the neighborhood aspects. In the final part, they failed to isolate the regular reading of CiV words and proposed an argument to rethink the theoretical implications of DRC model.

In Protopapas’ corpus survey, there are 79,825 CiV words (by type and by token) which could be pronunced /i/ or palatal. Because it is impossible to figure out which pronunciation pattern is “regular”, he decide the major pronuciation type of a CiV sequence by the type and token frequency (2:1).  The principle of DRC would predict both kinds of pronuciations would be the “regular reading” of CiV words: the reading pattern will be consistent with the default pronunciation of source words or the majority pronunciation. His study firstly classifies the CiV words into four groups according to the source words pronunciation ( /i/ or palatal) and the group majority pronuciation (/i/ versus palatal). Then ecah word generates two pronunceable nonwords: one had one letter modified and the other had several letters modified (In his presented case, the letter sequence mapping CiV keeps contant). Therefore, the final stimuli list had 8 groups of 20 nonwords and 40 unrecognizable nonwords from the CiV words without clear major pronunciation pattern (conflict between type and token).

The primary results are the response rates of /i/ for these critical nonwords. The analysis showed no effect of the number of replaced letters, but this factor interact with the other two factors. For the nonwords similar to the source words, the response rate of /i/ is related to the source word pronunciation rather than the group majority pronunciation. However, this tendency was reversed for the nonwords dissimilar to the source words. This means that the pronuncing Greek words may have no determinatic influence from the GPC rules.

I leave three questions and comments for his study:

Q1: why did not he analyze the effect of source word frequency? Is that because the error rates are the only data for his analysis?

Q2: Their analysis focused on the pronunciation of /i/. Did they consider the consistency between the pronunciation pattern and the source word pronunciation?

Q3: Reaction times of reading these nonwords have no difference among these conditions. Why did not they design an experiment for the source words? This may offer a clear picture for their interpretation.

This study also generate some ideas I can test in the Chinese study:

1.  The tendency to read nonwords aloud might rely on the clearest orthography-phonology mapping aspects embedded in the word forms.

2. There will be a solid argument about nonword reading if we have a acceptable findings about real words reading.

3. The theoretical thinking of Chinese character reading should consider the aspects about the phonetics.