4. General Discussion

The two experiments reported in the present study produced several important findings. First, the experiments tracked the developmental trajectory of Chinese children’s homophone awareness from the beginning of kindergarten to Grade 3. Our results suggest that homophone awareness emerges in the kindergarten years and increase steadily with age and grade level. Second, the development of homophone awareness is influenced by both semantic relatedness and morpheme type. The relative importance of the two factors changes from the beginning of kindergarten to the third grade. It is the interplay of the two factors, together with children’s age, that determines the developmental trajectory of homophone awareness in Chinese children.

Results of Experiment 1 suggest that semantic relatedness is a predominant factor in the development of homophone awareness in the early years. A very strong effect of semantic relatedness was observed in kindergartners. Even children in the beginning kindergarten class, whose average age was 3;10, performed above chance level when the meanings of words were highly related. When semantic relatedness is high, however, we cannot claim with certainty that children are relying on morphological decomposition to perform correctly on the morphological judgment task because they may make a “yes” judgment simply based on wholeword meanings. In contrast, successful performance on semantically distant words is a robust indicator that conscious morphological decomposition is taking place. The kindergartners in our study, especially those in the beginning and intermediate classes, had difficulty detecting a common morpheme in the latter condition. The pronounced difference between the two levels of semantic relatedness in the kindergarten years, together with the lack of significant interaction between semantic relatedness and age, suggests that kindergartners (ages 3~5) are at the initial stages of the development of morphological awareness. For these young children, attention to whole-word meaning may serve as a springboard into awareness of morphemes.

The difference between the two levels of semantic relatedness was reduced in the early primary grades, as shown by the significant semantic relatedness by age interaction and the significant semantic relatedness by time interaction observed in Experiment 2. These developmental changes suggest that children begin to separate morpheme meaning from whole-word meaning between the first and second grade. Children in the second grade and above are more analytical of the morphological structure of words, and as a result, are less affected by wholeword meaning when asked to make decisions regarding the presence of a shared morpheme. Formal literacy instruction is likely to play a key role here. Second graders, who have already been receiving literacy instruction for more than a year, may be starting to figure out that in Chinese the same morpheme is always represented by the same character regardless of the meaning of words in which it appears. This insight likely leads to the understanding that two words can contain the same morpheme even when the word meanings themselves are not closely related.

Despite the observed progress, in Experiment 2, children’s performance in the low semantic relatedness condition was still worse than that in the high semantic relatedness condition. In fact, the difference between the two levels of semantic relatedness, although much reduced, remained significant when children entered third grade (the second graders at the beginning of Experiment 2 were in third grade at Time 2). Combining the results of Experiment 1 and Experiment 2, we find that semantic relatedness has an important impact on Chinese children’s homophone awareness from kindergarten to Grade 3. These findings parallel those on the importance of semantic relatedness in adults' processing of compounds reviewed above, although routine lexical processing and language awareness cannot be equated. Research has consistently demonstrated that phonological awareness develops early and has largely reached ceiling by Grade 3 among normally developing readers (Casalis & Louis-Alexandre, 2000; Chen et al., 2009; Li et al., 2002; Shu, McBride-Chang, Wu & Liu, 2006). In comparison, morphological awareness appears to develop more slowly (Kuo & Anderson, 2006). Future research needs to investigate whether the impact of semantic relatedness continues to decrease as children become more experienced readers.

Type of morpheme is another factor that influences the development of children’s homophone awareness. Children were better able to identify free morphemes than bound morphemes in both experiments reported in this study. This finding likely reflects how different types of morphemes are represented in the mental lexicon (Taft & Zhu, 1995; also Beauvillain, 1996; Forster & Azuma, 2000). A free morpheme is represented as a morpheme at the morphological level while also having a corresponding representation as an independent word at the lexical level. This increases the likelihood of identifying the morpheme as a separate unit of meaning. In contrast, a bound morpheme is always part of a compound word and therefore is only represented at the morphological level. Our study also offers some evidence that the impact of morpheme type changes with age. A marginally significant interaction between type of morpheme and age, observed in kindergarteners, seems to indicate that the awareness of free morphemes emerges before the awareness of bound morphemes. However, this finding should be interpreted with caution because the interaction was not significant at the .05 level. This trend disappeared in older children, suggesting that the awareness of both types of morpheme increases at similar rates in the early primary grades.

Levelt (1989) noted that speakers of agglutinative languages (such as Turkish or Hungarian) may be particularly likely to have separate representations of bound morphemes due to the enormous number of possible word forms in these languages. We would add that another reason might be the regularity of form with which morphemes tend to be realized in words in such languages. Separate representations of bound morphemes might also be expected in isolating languages, such as Chinese, because the identity of individual morphemes in polymorphemic words is not obscured by morphophonological changes. On the other hand, speakers of fusional languages, in which morphemes often have phonologically opaque allomorphs, may favour whole-word representations. Thus, awareness of bound morphemes may develop at an earlier age in Chinese than in inflecting languages. Future research needs to investigate whether the findings of the present study can be generalized to children whose first languages are typologically different.

Our study reveals that semantic relatedness and type of morpheme play different roles at different stages of the development of homophone awareness. As shown in Experiment 1, semantic relatedness is the most prominent factor in the early stages of development. Because kindergartners do not appear to have a deep and consistent explicit understanding of morphemes, it is not surprising that morpheme type has a more limited influence at this time. The relative importance of semantic relatedness and type of morpheme changes with children’s age/grade level. The results of Experiment 2 reveal that, in the early primary grades, the effect of semantic relatedness varies as a function of morpheme type. The effect of semantic relatedness on free morphemes virtually disappears when children enter Grade 3, indicating that free morphemes become increasingly consciously accessible to children as units of meaning. By contrast, the effect of semantic relatedness on bound morphemes is reduced but remains significant.

To summarize, the present study sheds light on the time course of the development of homophone awareness (and, by extension, morphological awareness) in Chinese children. As previously noted, given the constraints of the morphological judgement task, children’s homophone awareness is better represented by their performance in the low semantic relatedness condition, which is a robust indicator of the ability to identify morphemes as separate units of meaning. Combining the results in this condition from the two experiments, we demonstrate that awareness of morphemes begins to develop in the kindergarten years. In particular, awareness of free morphemes emerges between the beginning and the intermediate classes (ages 3;10 and 5;0), whereas awareness of bound morphemes emerges between the intermediate and senior classes (ages 5;0 and 5;10). The two types of homophone awareness continue to develop, in similar fashion, from kindergarten to early third grade (age 7;10). Awareness of free morphemes seems to be close to ceiling by early third grade. Lagging behind, awareness of bound morphemes is still developing at this time. It is reasonable to assume that awareness of bound morphemes will continue to develop past this age range, as the average performance is less than 70% in the early third grade.

Findings from a separate line of inquiry, word association research, seem to converge with our findings (Brown & Berko, 1960; S?derman, 1993; see also Aitchison, 2003; Meara, 1984; Pan & Uccelli, 2009). When asked to supply one or more words that come to mind when a target word is presented, individuals with developing mental lexicons tend to provide more associates that are related to the stimulus word phonologically (also known as clang associates) or which can collocate with it in discourse (also known as syntagmatic associates). Mature mental lexicons, in contrast, are characterized by a larger proportion of meaningbased (or paradigmatic) associations. This developmental change, commonly referred to as the syntagmatic-paradigmatic shift, takes place throughout the years of formal education. The change in response patterns appears to be the most dramatic around age 7 (Pan & Uccelli, 2009), which is approximately the age of our oldest participants. In addition, there is evidence that semantic categories are broadened and consolidated as the mental lexicon develops. At age 8, children are able to provide almost twice the number of examples of a semantic category compared to 5-year-olds, and the examples are more accurate (Nelson, 1974). The syntagmatic/paradigmatic shift likely contributes to the growth of morphological awareness, as good understanding of meaning is a sine qua non of understanding morphology.

Note that, although morphological competence is a necessary prerequisite for morphological awareness in naturalistic first language acquisition, our results show that awareness of word formation does not drastically lag behind the firm entrenchment of word formation processes in oral production (which takes place between ages 3 and 4, as mentioned above). This intimate relationship between word formation competence and awareness may be explained in light of the evidence that lexical storage and processing are linked with explicit memory, which is the only aspect of human long-term memory compatible with the concept of awareness (for recent reviews and models of the relationship between memory systems and language, see Paradis, 2004; Ullman, 2004; see also Levelt, 1989; Levelt, Roeloefs & Meyer, 1999). In addition, because these models associate inflectional morphology with implicit memory, which is not amenable to direct conscious inspection, the interesting question arises of whether awareness of inflectional morphology emerges later and reaches lower levels than awareness of word-formation morphology, despite the fact that inflectional morphology itself emerges much earlier in spontaneous language use. This issue merits future investigation.

Our study also makes a significant methodological contribution by providing a measure of homophone awareness that distinguishes between different levels of semantic relatedness. To the best of our knowledge, this is the first study that has evaluated children’s morphological awareness with semantically dissimilar words. Previous research (Li et al., 2002; McBride-Chang et al., 2003) only examined morphological awareness with words that are closely related in meaning. McBrideChang and colleagues (2003) reported that children’s performance on a morpheme identification task already reached ceiling in second grade and ceased to be a significant predictor of character reading. Because the target words in this task, for example, laamlui (boy and girl), are more closely related in meaning to the words containing the shared morpheme, laamhai (boy), than to the words containing homophones, laamkou (basketball), and laamcik (blue), children may choose the correct answers based on word meanings without having to access the morphemes. In fact, the finding of McBride-Chang and colleagues (2003) is similar to our finding in the high semantic relatedness condition. As we have demonstrated, children’s ability to identify morphemes develops more slowly when words are semantically distant. This ability may be predictive of character reading in Grade 2 and beyond.