Tan, L. H., & Perfetti, C. A. (1997). Visual Chinese character recognition: Does phonological information mediate access to meaning? Journal of Memory and Language 37, 41-57.
The authors introduce research done in the past on English reading that provided evidence of phonological information mediating access to meaning. What does this mean?
Prime | Pattern mask | |
1 | beach | sand |
2 | beech | sand |
So, taking the words used in the above example, in one instance the word “beach” would be used to “prime” for the word “sand”, meaning that it would be shown first in an effort to measure whether any effect is had on the reaction to the second word, “sand.” This effect, if there is one, is called a “priming effect”, or just “priming.” In a study done by Lesch and Pollatsek, phonologically mediated priming (in terms of the above example, that would mean priming by “beech”) was found to have an effect when the prime word was exposed for 50ms, followed by a pattern mask of 200 ms, but was absent when the prime was exposed for 200ms, followed by a pattern mask of 50ms. When the prime was shown for 50ms, the priming effect was the same for the word actually having semantic association to the target and a homonym of that word.
This tells us that, in English, phonology does provide access to meaning, and that this process is extremely rapid. It also tells us that there is a point (some time before 200ms) at which the reader does a kind of “spell check” on a word, at which point homonyms with different orthography are no longer confused for each other.
That’s the background given to us on what has been found for English reading. The authors’ experiments concern Chinese reading.
In their study, the authors used three types of primes: “(a) A prime was synonymous but neither visually nor phonologically similar to a target; (b) a prime was homophonic with the synonym prime but had no graphic, phonological, or semantic similarity to the target; (c) a prime had no graphic, phonological, or semantic similarity to the target” (44). Here’s an example of what that would look like:
target | synonym | Homophone of synonym | control |
年 | 岁 | 碎 | 抗 |
宽 | 阔 | 括 | 卡 |
They offer several possible predictions based on their model of the mental representation of Chinese lexical items. These are: “First, if only the visual-orthographic route controls access to character meaning, synonym primes but not homophones of synonym primes will accelerate target recognition. If the phonology connection mediates access to the meanings of all characters, both synonym primes and homophones of synonym primes should facilitate character processing, and their facilitation effects should not be different. …homophone density should control whether phonological information can be successfully used to access meaning” (44-45).
In their first experiment, they use a stimulus onset asynchrony (SOA) of 129ms. They note three significant results: “First, there was significant priming by homophones of synonyms, but, as predicted, this effect was dependent on homophone density. Second, synonym priming was more effective than priming by homophones of synonyms across all conditions of homophone density. Third, the effect of synonym priming was less when primes had many homophones than when they had few homophones” (47).
This is all quite exciting, really. The logic of their predictions seemed totally sound, yet the results don’t quite fit into any of them. Phonological priming is found, but, unlike with English, priming is stronger for synonyms than for homophones of synonyms. On top of this, homophone density actually affects the priming of synonym primes! This, as the authors point out, “suggests that semantic priming itself is partly mediated by phonological forms” (49. An interesting and exciting finding, as far as I’m concerned.
On to experiment 2. As I already discussed, the research on English showed that homophone-to-synonym priming disappeared after the prime was shown for 220 ms. So, for experiment 2A , the SOA was set to 243 ms. The findings were the same as those in experiment 1.
In experiment 2B, the SOA was upped to 500ms. The results of this experiment showed that the “spell check” or “verification stage” has been completed so that not only is there no priming for homophones of synonyms, but homophone density also had no effect on synonym priming.
So, this study has shed some light on Chinese character recognition and has also brought up a few questions. It is still, for example, unclear how the relationship between the phonologic, orthographic and semantic aspects of a character should best be represented. The authors suggest:
“What will work in Chinese is what works in alphabetic systems: Not ‘look up’ of semantics from phonology, but rather convergence of word identity from the combination of orthographic and phonological identities. Mediation is not moving from orthography to phonology in order to carry out some other operation. It is the convergence of information from an orthographic form with its associated phonological forms such that a unique identity, with associated meanings, is established” (54).
They also mention in their conclusion that the difference of priming for synonyms and homophones of synonyms “suggests (but does not prove) that there are direct links from orthography to semantics as well” (54). This certainly seems to be different than the findings of the studies on English reading, and more research is needed (and may very well have already been done – this article is from 1997) to get at what this really looks like for Chinese.
