Reading and Writing

, Volume 23, Issue 3, pp 311–336

Cross-language transfer of insight into the structure of compound words


  • Jie Zhang
    • Center for the Study of ReadingUniversity of Illinois at Urbana-Champaign
    • Center for the Study of ReadingUniversity of Illinois at Urbana-Champaign
  • Hong Li
    • Department of PsychologyBeijing Normal University
  • Qiong Dong
    • Department of PsychologyBeijing Normal University
  • Xinchun Wu
    • Department of PsychologyBeijing Normal University
  • Yan Zhang
    • Department of PsychologyBeijing Normal University

DOI: 10.1007/s11145-009-9205-7

Cite this article as:
Zhang, J., Anderson, R.C., Li, H. et al. Read Writ (2010) 23: 311. doi:10.1007/s11145-009-9205-7


Cross-language transfer of awareness of the structure of compound words was investigated among native speakers of Chinese who were learning English as a second language. Chinese fifth graders received instruction in the morphology of four types of compound words in either Chinese or English. They then completed both the Chinese and English versions of a compound word structure analogy task. Compared with children who received no instruction, children who received instruction in Chinese were able to transfer knowledge they had acquired of compound types in Chinese to comparable types in English. Reverse transfer from English to Chinese was found among children with high reading proficiency.


Morphological awarenessCross-language transferCompound morphologyReading proficiency


This study investigated cross-language transfer of awareness of the structure of compound words. Chinese children studying English as a second (or foreign) language received a lesson on the structure of compounds in either Chinese or English. Of course, it was expected that the lesson would improve awareness of compound morphology in the language studied. But, the interesting question was this: Would studying Chinese compound words improve awareness of the structure of comparable compound words in English, and would studying English compounds improve awareness of the structure of comparable Chinese compounds? If the answer is yes, this would be among the first experimental demonstrations of cross-language transfer of an aspect of metalinguistic awareness.

Most existing studies of cross-language transfer are correlational and do not provide a strong warrant for the casual connection entailed by the concept of transfer. The best studies have included statistical controls for other factors that might explain a correlation between a facet of language ability in L1 and L2, controls for such factors as family socioeconomic status, parent level of education, IQ, and vocabulary knowledge and other facets of language ability. Still, even the best studies have fallen short of proving causation. Kuo and Anderson (2008) expressed the issue this way: “There is excessive reliance on correlational methods in research on metalinguistic awareness and second language learning. Analysis of covariance, hierarchical regression analysis, and structural equation modeling provide useful tools for weighing the relative importance of factors, but it is an illusion that a causal model can be established by statistical legerdemain.” (p. 59)

Most studies of cross-language transfer have focused on phonological awareness (e.g., Chow, McBride-Chang, & Burgess, 2005; Durgunoglu, Nagy, & Hancin-Bhatt, 1993; Gottardo, Yan, Siegel, & Wade-Wooley, 2001; Lindsey, Manis, & Bailey, 2003; Verhoeven, 1994), whereas transfer of morphological awareness has received less attention. Among cross-language studies of morphological awareness, most have investigated inflectional or derivational morphology (e.g., Fowler, Feldman, Andjelkovic, & Oney, 2003), probably because inflection and derivation are productive word formation processes in Western languages and most language research has been done in the West. In contrast with English and other Western languages, Chinese is highly productive in compound words (Packard, 2000) and contains relatively few inflectional and derivational words.

Among the few attempts to investigate understanding of compound morphology, the Noun+Noun compound, which is frequent and productive in most languages, has been most widely studied (e.g., Ku & Anderson, 2003; McBride-Chang et al., 2005; Nagy, Berninger, & Abbott, 2006; Nicoladis, 2003). Other types of compound word have been largely overlooked.

Structure of compounds in Chinese and English

Basic compounds consist of two constituents (e.g., classroom, cranberry). The constituents may be free [e.g., room in classroom] or bound [e.g., rasp- in raspberry]. Most compounds follow the right-head rule, for instance, that a soap dish is a kind of dish whereas dish soap is a kind of soap. Exceptions are the ubiquitous Verb+Particle compounds, such as melt down and walk on, which are clearly left-headed.

The structure of compound words can be usefully described in terms of the form class of the constituents (Packard, 2000; Selkirk, 1982). Four types of compound nouns frequent in both Chinese and English are Noun+Noun [e.g., 电脑, electricity-brain, ‘computer’; dog house, baseball], Noun+Verb [e.g., 壁挂, wall-hang, ‘wall hanging’;haircut, track meet], Verb+Noun [e.g., 睡袋, sleep-bag, ‘sleeping bag’;cry baby, playground]. Types of compound verbs in both Chinese and English include Noun+Verb [e.g., 瓜分, melon-divide, ‘to partition’;to role play, to baby sit], Adjective+Verb [e.g., 干洗, to dry clean; to direct dial, to second guess], and Verb+Particle [e.g., 穿过, to run through; to knock out, to show off].

Word structures vary in productivity and transparency, that is, how readily words can be formed according to a structure and how evident the meanings are of the resulting complex words. English has a productive and transparent derivational morphology [e.g., words formed with -tion, -ly, -ness]. However, overall, the compound morphology of English is less productive and transparent than the compound morphology of Chinese (Packard, 2000).

In terms of the productivity of each compound structure in English, Bauer (1987) concluded that Noun+Noun compounds comprise the most productive class of compound words. The Verb+Particle compound is made up of phrasal verbs and nouns and is also highly productive. The Noun+Verb compound has limited productivity but it is more productive than the Verb+Noun (object) compound, which has its origins in Romance languages such as French and Latin, and is regarded as no longer productive in modern English (Bauer & Renouf, 2001).

Overall, research on the productivity of different types of compounds in Chinese parallels the research in English. Yuan and Huang (1998) analyzed a corpus of 40,958 two-character Chinese words and reported that the proportions of Noun+Noun, Verb+Noun, and Noun+Verb compounds were .308, .195, and .022, respectively. Analysis of the data in Huang (1997, p. 264) obtained from the dictionary Guoyu Ribao Cidian yielded comparable figures of .293, .203 and .034, respectively.

To summarize, Noun+Noun and Verb+Particle are productive structures within both Chinese and English. Noun+Verb and Verb+Noun are less productive in each language, with the interesting difference that Verb+Noun is more productive than Noun+Verb in Chinese whereas the reverse is true in English.

Understanding compound words and learning to read Chinese and English

Morphological awareness has long been recognized as an important factor in children’s vocabulary growth and reading development (Carlisle, 1995, 2000; Nagy & Anderson, 1984; Tyler & Nagy, 1990; White, Power, & White, 1989). Previous research on morphological awareness has focused on inflectional and derivational morphology. Recently, the importance of compound morphology in learning to read has also been documented in several languages (e.g., Ku & Anderson, 2003; Li, Anderson, Nagy, & Zhang, 2002; McBride-Chang, Shu, Zhou, Wat, & Wagner, 2003; McBride-Chang et al., 2005; Nagy, Berninger, Abbott, Vaughan, & Vermeulen, 2003; Nagy, Berninger, & Abbott, 2006).

Nagy et al. (2003, 2006) found that understanding the structure of compound words (e.g., Which is a better name for a bee that lives in the grass, a grass bee or bee grass? Expected answer: grass bee) made a unique contribution to reading comprehension above and beyond vocabulary knowledge for at-risk second grade children, as well as for upper elementary and middle school students. In a study with kindergarten and second-grade children in Hong Kong, McBride-Chang et al. (2003) reported that morphological awareness was a unique predictor of Chinese character recognition after controlling for phonological awareness and vocabulary. McBride-Chang and her collaborators employed a morphology construction task. A sample item is, “There is a home appliance that is used to wash clothes. It is called a washing machine [literally, ‘wash clothes machine,’ 洗衣机]. If a home appliance could be used to wash shoes, what would you call that?” The expected answer is wash shoes machine, 洗鞋机.

However, previous findings have been mainly restricted to knowledge of simple Noun+Noun compounds. In a cross-language study of implicit knowledge of compound structures, Zhang, Anderson, Packard, Wu, and Tang (2007) extended the investigation of compound structures to additional types, including Verb+Noun, Noun+Verb, and Verb+Particle in addition to Noun+Noun compounds. Chinese and American second-, fourth-, sixth-graders, and college undergraduates completed a compound structure analogy task in their native language, assessing awareness of the structure of the four types of compounds, half involving familiar compounds and half involving novel compounds. A sample familiar word item is Step-up: [a] step-ladder, [b] step-over, [c] step-father. From a structural point of view [b] is the best answer because step-up and step-over share the Verb+Particle structure. As predicted from the hypothesis that productivity and transparency influence compound structure awareness, Chinese participants performed significantly better than American participants at every age on every type of compound. Both language groups performed better on compound types that are more productive in their respective languages.

Because of the greater productivity and transparency of the compound morphology in Chinese, it has been hypothesized that compound morphology is more important in learning to read Chinese than in learning to read English (Anderson & Li, 2005; Nagy & Anderson, 1999). Ku and Anderson’s (2003) cross-language study only weakly supported this hypothesis, but the hypothesis was clearly supported in a recent cross-language study by McBride-Chang et al. (2005) that compared the contributions of phonological awareness and morphological awareness to word recognition in second graders from Beijing, Hong Kong, Korea, and the United States. As anticipated, morphological awareness was more important for reading Chinese, phonological awareness was more important for reading English, and both phonological and morphological awareness contributed to reading Korean.

Cross-language transfer of morphological awareness

Evidence of cross-language transfer of phonological awareness has not only been found across alphabetic languages (e.g., Durgunoglu et al., 1993; Lindsey et al., 2003; Verhoeven, 1994), but also between alphabetic languages and non-alphabetic languages, such as Chinese (Chow et al., 2005; Gottardo et al., 2001; although see Bialystok, McBride-Chang, & Luk, 2005). While results are not entirely consistent from study to study, the typical finding is that children who are more sensitive to phonological structure in L1 perform better in L2 word reading.

Only a few studies have addressed the transfer of morphological and lexical skills (e.g., Hancin-Bhatt & Nagy, 1994; McBride-Chang, Cheung, Chow, Chow, & Choi, 2006; Nicoladis, 1999, 2002, 2003; Wang, Cheng, & Chen, 2006). Hancin-Bhatt and Nagy (1994) were among the first to investigate morphological and lexical transfer across languages. Cognates (words similar both in form and meaning) in Spanish and English offer the possibility of morphological and lexical transfer. Spanish–English bilingual children of fourth-, sixth-, and eighth-grade were asked to translate low-frequency English words with cognates or noncognates into Spanish. Results showed that children’s cognate knowledge increased with age above and beyond any increase in their vocabulary knowledge in Spanish and English. Students recognized cognate stems in suffixed words (e.g., amicably) more easily than noncognate stems in suffixed words (e.g., shortly). These results suggest a role for cross-language transfer in learning derivational or inflectional morphology.

As illustrated by French–English bilingual children’s compound acquisition, cross-language transfer of compound awareness occurs when ambiguous structure exists in two languages (Nicoladis, 2002, 2003). Verb+Noun (e.g., taille-crayon, sharpen-pencil, meaning ‘pencil-sharpener’) compounds are productive in French, but few Verb+Noun compounds (e.g., pickpocket) exist in English. Instead, Noun+Verb+er compounds (e.g., bus driver) are more common in English. Such ambiguous morphological structure seems to serve as a catalyst for cross-linguistic transfer between French and English. In a compound production task (naming pictures of novel creations, e.g., sun juggler), Nicoladis (2003) found that French–English bilingual children were more likely to produce ungrammatical Verb+Noun compounds in English than their English monolingual peers. However, no differences between monolinguals and bilinguals were found in a compound comprehension task. These results seem to suggest that cross-linguistic transfer of compound morphology is a language production phenomenon and structural ambiguity predicts when morphological transfer occurs.

The extent of cross-language transfer depends on the degree of L1 and L2 language proficiency, as well as the similarity between the two languages and writing systems (Bialystok et al., 2005). Cummins’ (1979) threshold hypothesis says that a minimum level of proficiency in L1 and L2 must be reached in order for bilingualism to have benefits. To date, a vast number of studies have provided evidence suggesting transfer from the stronger language to the weaker language.

However, few prior studies have investigated whether there are distinct patterns of transfer for bilinguals varying in language proficiency. One exception came from Proctor, August, Carlo, and Snow’s study (2006) involving fourth grade Spanish–English bilinguals. Results showed that more fluent English readers benefited more from Spanish vocabulary knowledge than their less fluent counterparts in English reading comprehension. Students who were the speediest (75th percentile) English word readers were likely to benefit from increased Spanish vocabulary knowledge, whereas the students at lower fluency levels (25th percentile) were unlikely to show any transfer. Such an interaction between Spanish vocabulary and English fluency underscores the impact of L2 proficiency on the nature of cross-language transfer.

The present study will examine the effect of both L1 and L2 reading proficiency on the bidirectional transfer of knowledge of compound morphology in Chinese and English. Because of the high correlation between L1 and L2 basic reading processes (Geva, Wade-Wooley, & Shany, 1997), supporting Cummins’ (1991) notion of a unitary construct of language proficiency, there is reason to believe that language proficiency in L1 and L2 represent common skills required in the process of recognizing and transferring insights about the structure of compounds across languages.

So far evidence about transfer of morphological concepts between Chinese and English has been contradictory. In a study with Chinese immigrant children of first to fourth grade in the US, Wang et al. (2006) investigated cross-language transfer of derivational and compound morphological awareness in Chinese and English. Derivational morphology was assessed with an oral cloze test (e.g., My uncle is a __ [farm]. Answer: farmer) and compound morphology was assessed with the same novel compound production task used by Nagy et al. (2003). Results indicated that awareness of English compound morphology uniquely contributed to character reading and reading comprehension in Chinese after controlling for Chinese predictors, such as age, grade, vocabulary, and phonological awareness. Thus, reverse transfer of compound morphological awareness from L2 (English) to L1 (Chinese) was observed, but not forward transfer from L1 (Chinese) to L2 (English). This pattern of results was explained by the speculation that these early bilinguals tended to use L2 strategies in processing L1 because of the rapid increase in L2 proficiency of immigrant children.

In a study with 6–7 years old Chinese children in Hong Kong, McBride-Chang et al. (2006) investigated the association of Chinese morphological awareness and vocabulary knowledge in Chinese and English. Chinese morphological awareness was assessed with two measures: receptive morphological awareness and morphological construction. In the receptive task, children were orally presented the target word: 斑牛 (striped cow), and were asked to select one of the four pictures: correct answer, relational distracter: 斑马 (striped horse), categorical distracter: 牛 (cow), and an unrelated alternative. The morphological construction task was similar to the one in McBride-Chang et al.’s (2003) study. Inconsistent with Wang et al.’s (2006) findings, McBride-Chang et al. (2006) found that Chinese morphological awareness explained unique variance in Chinese vocabulary but not in English vocabulary, thus indicating no transfer of compound awareness for Chinese ESL children.

The discrepancies in these two studies may result from several factors. First, the possibility of transfer, and direction of transfer if it occurs, may have been influenced by the different second language environment of Chinese children in Hong Kong and in the US. Compared to Chinese immigrant children in the US, who are immersed in English, Hong Kong children are learning English as a second language, and have less exposure to English at home or school. Second, different test batteries were used in two studies. Wang et al. (2006) used character reading and reading comprehension as dependent variables whereas vocabulary was used as dependent variable and only Chinese morphological awareness was measured in McBride-Chang et al.’s (2006) study. Given the higher productivity of compound morphology in Chinese, transfer of insight into compound morphology from L1 (Chinese) to L2 (English) would be expected. In contrast, reverse transfer of awareness of derivational morphology from English to Chinese might be expected because of the lower productivity of derivational morphology in Chinese. However, no reverse transfer of derivational morphology, but instead reverse transfer of compound morphology, was found in Wang et al.’s (2006) study. This pattern of results may be due to the fact that Chinese derivational morphology is not productive and bears little similarity to English derivational morphology.

Hypotheses about transfer of insight into compound words

The present study investigated transfer of compound structure awareness of fifth grade Chinese ESL children. Three hypotheses were formulated. First, transfer of compound structure awareness from L1 (Chinese) to L2 (English) was expected for children of all language proficiency levels. Second, reverse transfer of compound structure awareness from L2 (English) to L1 (Chinese) was considered unlikely because of the children’s limited English language proficiency. However, we anticipated that children with high language proficiency receiving the English intervention might show transfer to Chinese, whereas transfer was considered implausible for the less proficient children. Third, based on our previous findings, we anticipated that the relative productivity of compound structures within a language would influence children’s understanding of compounds. As shown in the Zhang et al. (2007) study, children were expected to perform better on Noun+Noun and Verb+Particle than Noun+Verb and Verb+Noun compounds in both Chinese and English.



The participants were 168 second-semester fifth graders from a large elementary school in Tianjin, a city in north China. The average age of these children was about 11 years old. Four classes described by the principal as comparable participated in the study. One class (N = 42) was randomly assigned to receive the Chinese compound morphology intervention, another class (N = 43) received the English compound morphology intervention, and the other two classes (N = 83) served as controls.

The children had had approximately four and a half years of English instruction at the time of study. During these years, children received four English classes per week, with each class lasting for 40 min. Oral language activities predominated in first and second grade. Reading and writing instruction began in the third grade. The quality of English teaching in ordinary public elementary schools in China is less than optimum. Most English teachers are graduates of a local normal university with a bachelor degree or teaching certificate in English. Typically they speak imperfect English. Chinese children do not have a favorable English learning environment outside of school. Children seldom speak English or hear English. Their parents are unlikely to speak English unless they are highly educated. Based on an analysis of the school English curriculum, over four and one half years of instruction, the fifth grade children participating in the current study had been introduced to less than 1,000 English words. Their oral English proficiency was very limited.

Parents of the participating children were speakers of a dialect of Chinese close to Mandarin and had at least a primary education or higher. According to the teachers, the occupations of the students’ parents varied, including factory workers, company employees, business people, and local government officials. Overall, the household income of the children’s families was reported to be average or somewhat above the local average.

Reading proficiency measures

Two reading proficiency measures in both Chinese and English were administered to all the participants before the intervention. In addition, the children’s regular Chinese and English teachers rated each student’s Chinese or English language proficiency into three categories: high, medium, and low, respectively.

Curriculum-based word reading tasks were individually administered to assess children’s word decoding ability. The Chinese task contained 60 Chinese words, 36 single-character words and 24 multi-character words. The English task consisted of 60 English words including root words, inflections, derivatives, and compounds. Among the 60 words for each language, 20 were selected from the vocabulary taught in preceding grades, 20 from vocabulary introduced in the current grade, and 20 from vocabulary introduced in following grades. The 60 words were arranged in order of estimated difficulty, from easy to more difficult. The task terminated if a student failed to correctly read six consecutive words. The Spearman-Brown odd–even reliabilities of the Chinese and English word reading tasks were .72 and .87, respectively.

Group-administered cloze tasks were employed to assess children’s basic reading comprehension. The English cloze task was created from age-appropriate local extracurricular English reading materials. The Chinese cloze task was selected from Zhang’s (2004) test package. The task for each language consisted of three 10-item passages. The Spearman-Brown odd–even reliabilities of the Chinese and English cloze tasks were .67 and .74.


A 45-min intervention on Chinese or English compound morphology was implemented 1 day after the pretests. The aim of the intervention was to promote explicit understanding of the structure of Chinese or English compound words. The lesson plans for the Chinese and English interventions were developed and refined by four of the authors who are Chinese–English bilinguals. All the compound words used in the interventions were familiar to the children from oral language, based on ratings made by the children’s regular teachers.

To ensure the comparability of the interventions, one of the authors, a native speaker of Chinese and a fluent speaker of English who has 3-year’s elementary school teaching experience, acted as the instructor for both the Chinese and English interventions. Compound morphology might have been challenging for the regular teachers since it is not part of the curriculum in either Chinese or English language arts classes. Furthermore, it would have been difficult to ensure that English and Chinese interventions were comparable if they had been implemented by different teachers. The interventions were rehearsed twice by the experimental instructor with groups of graduate students. The lesson plans were improved based on feedback from these audiences. The instructor warmed up and established rapport with students in the two experimental classes 1 day before the intervention.

The Chinese and the English intervention followed the same design. First, the instructor introduced the form classes of morpheme constituents of compound words: noun, verb, and particle. Four types of compound structures were then illustrated: Noun+Noun (e.g., 电灯, electricity light, ‘light’; bookshop), Verb+Particle (e.g., 跳上, ‘jump onto’; blow up), Verb+Noun (e.g., 担心, carry-heart, ‘worry’; jump rope), and Noun+Verb (e.g., 日落, ‘sunset’; snowfall). Considering the abstractness of particles, more examples of Verb+Particle compounds were given. Because of the children’s limited English vocabulary, some of the English items were possible compounds rather than actual English compounds.

Next, children explored the structure of compound words in four activities: compound structure match, compound structure oddity, compound classification, and compound production. In the compound structure match activity, students were presented two columns of compound words with each of four words in a column representing a distinct type of compound structure. The orders of word structure in the two columns were scrambled. Students were asked to match the structure of the compounds in the two columns by drawing lines between the items with the same structure.

In the compound structure oddity activity, children were presented 12 arrays of three words (e.g., school bag, hand bag, wash bag; dog-house, farm-house, move-house), and were asked to determine which word had a different structure from the other two. The compound oddity task was a whole-class activity. Students were individually called on to make choices. Compound word classification was a small-group activity. Children were divided into small groups with four to five students in each group. Each group was given 20 compound word cards and a piece of cardboard divided into four blank columns with each column representing one structure. Children were asked to categorize the 20 words into the four structural categories and stick each word card onto the appropriate column. Group representatives reported their classifications to the whole class and received feedback from the teacher and other students.

The compound production task was a whole-class activity in which children were asked to generate as many compound words containing a certain root word as possible. For example, given the root word ball, children made up a series of Noun+Noun compounds such as basketball, snowball, meatball, football, iceball, and volleyball. To make the activity easier and more fun, novel compounds were allowed provided meanings of the novel compounds were imaginable (e.g., kick ball, throw ball,…). Finally, the instructor briefly summarized the four types of compound structures taught in the lesson.

To make the intervention more efficient and interesting, attractive PowerPoint slides and word cards for the group activity were prepared in advance. According to classroom observers and the instructor, students were highly engaged in the activities and the Chinese intervention was more accessible than the English intervention. The instructor felt rushed during the English intervention because explanations of activities often had to be repeated in Chinese due to the children’s limited oral English proficiency.

Measure of word structure awareness

The word structure analogy task was a refined version of the one employed by Zhang et al. (2007). It assesses the ability to make compound structure analogies involving four types of compounds that were targeted in the interventions: Noun+Noun, Verb+Particle, Noun+Verb, and Verb+Noun. Each item consisted of a compound target followed by three choices. The task was to circle the choice that was most similar to the target word. An example is the target word handcar and the choices a. handout, b. handbook, and c. handover. The answer is handbook because it has the same Noun+Noun structure as handcar. For each structure, four arrays of familiar compounds (e.g., football, a. snowball, b. jump ball, c. play ball) and four arrays of novel compounds (e.g., candy-door, a. candy-wash, b. candy-road, c. candy-dance) sharing one morpheme constituent were developed. As shown in Zhang et al.’s (2007) study using similar items, the mean proportion correct of fourth grade Chinese children on the word structure analogy task is quite high, above .70. To increase task difficulty, another four arrays of familiar compounds (e.g., rain coat, a. blow out, b. turn down, c. ball game) and four arrays of novel compounds (e.g., orange taxi, a. wait-panda, b. save-candle, c. potato-train) with distinct morpheme constituents were developed for each structure. Altogether there were 32 familiar word items and 32 novel word items in both the Chinese and English versions of the task. See the Appendices A and B for the instructions and stimuli.

Items were designed to require a decision based on the structure of the compound, not merely semantic overlap. Items were constructed, judged, and revised in an iterative process. All the familiar words were judged by teachers to be familiar to participants. All of the familiar compounds were transparent, in the sense that the constituent morphemes contributed in an obvious way to the meaning of the compound. Novel compounds were plausible but non-occurring or very low in frequency. Both familiar compounds and novel compounds were made up of two familiar morphemes and did not overlap with the words used as examples during either intervention. Verb+Noun and Noun+Verb structures are relatively unproductive in English and the compounds that do occur are often low in frequency or contain constituents low in frequency (e.g., whetstone, browbeat). To generate enough English Verb+Noun and Noun+Verb compounds familiar as a whole with familiar constituents, it was necessary to include a few Verb+Noun and Noun+Verb phrases whose status as compounds is dubious (e.g., cook rice, car wash). The Chinese and English tasks were made as close to parallel as possible, as judged by those of the authors who are Chinese–English bilinguals.

Four days after the intervention, all participants were administered both the Chinese and English versions of the word structure analogy task. The order of the familiar word task and novel word task was counterbalanced within each class. Within tasks, items representing the different types of compounds were intermixed. Instructions told students to check the option that was “most similar” to the target; we did not say “similar in structure” because that would have put the control group at a disadvantage. Instructions for both tasks were in Chinese; however, the items in the English task appeared only in English. The Spearman-Brown odd–even reliabilities of overall performance on the Chinese and English word structure analogy tasks were .94 and .93, respectively.


Table 1 displays mean proportions correct on the four reading proficiency pretests by intervention condition. MANOVA analysis showed no significant difference among the three conditions on the pretests, Wilks’ Lambda = .89, F (8,324) = 2.32, p > .05, ή2 = .01. It is noteworthy that there were large variations in English word reading, which may be due to the individual differences in English literacy activities outside of school. Some children get private afterschool or weekend English instruction, whereas others do not.
Table 1

Mean proportions (SDs) correct on reading proficiency measures


Chinese (N = 42)

English (N = 43)

Control (N = 83)

Chinese word reading




English word reading




Chinese cloze




English cloze




Table 2 presents the means and standard deviations of proportions correct on the compound structure analogy posttest by condition, compound structure, and word type. Tests of normality showed that the performance of Chinese intervention group on the Chinese posttest was negatively skewed and the performance of all three groups on the English posttest was positively skewed. To correct skewness, an arcsine transformation was applied to the Chinese and English compound analogy posttest scores. The tests of significance reported later are based on transformed scores.
Table 2

Mean proportions (SDs) correct on compound analogy posttests


Chinese analogy

English analogy











































































ANCOVA analyses were conducted to investigate forward transfer from L1 (Chinese) to L2 (English) and reverse transfer from L2 (English) to L1 (Chinese). The scores on the four reading proficiency pretests were covariates. For the Chinese compound analogy posttest, the analysis showed a significant intervention effect, F (2, 161) = 26.60, p = .00, ή2 = 0.25. A pre-planned contrast, showed that the Chinese intervention group performed significantly better than the control group, t (79) = 6.89, p = .00, Cohen’s d = 1.55, which documents gains from the Chinese compound morphology intervention. The difference between the English intervention group and the control group on the Chinese posttest trended positive, but a second pre-planned contrast revealed that the difference was not significant, t (69) = 1.67, p = .09, Cohen’s d = .40.

For the English compound analogy posttest, ANCOVA analysis also showed a significant intervention effect, F (2,161) = 5.77, p = .004, ή2 = .067. A pre-planned contrast showed that the English intervention group performed better than the control group, t (61) = 2.10, p = .04, Cohen’s d = .53. This indicates that the children learned something about the structure of English compounds and were able to apply what they had learned to new English compounds. The Chinese intervention group also performed significantly better than the control group on the English analogy test, t (64) = 2.92, p = .00, Cohen’s d = .73. This indicates transfer of knowledge of compound morphology from L1 (Chinese) to L2 (English).

As shown in Table 3, children’s teachers’ ratings of language ability, the reading proficiency pretests, and the compound morphology posttests were significantly intercorrelated. To determine if there is a single construct of reading proficiency underlying the four pretests, principal component analysis was conducted. Only one principal component had an eigenvalue above 1.0 which suggests a single underlying factor. The first principal component explained a total of 54.39% the variance in the four pretests. The loadings of English word reading, Chinese word reading, English cloze, and Chinese cloze on the principal component were .84, .74, .70, and .65, respectively. Principal component scores, the best linear combination of the four pretests, were employed as an overall index of reading proficiency in subsequent analyses.
Table 3

Correlations among reading proficiency measures and compound analogy posttests








1. Teacher’s rating







2. Chinese word reading







3. English word reading






4. Chinese cloze





5. English cloze




6. Chinese compound analogy



7. English compound analogy


N = 168. All correlations significant, p < .01. Correlations with teacher’s ratings, Spearman’s rho

Hierarchical regression analyses were run to explore whether the interventions were differentially effective depending on the level of reading proficiency of the children. Two analyses were completed in which the dependent variables were scores on the Chinese or the English compound analogy posttest. In each analysis, reading proficiency principal component scores were entered first, followed by two contrasts (Contrast 1: Chinese vs. Control, Contrast 2: English vs. Control), and two interactions (Reading Proficiency × Contrast 1, Reading Proficiency × Contrast 2) entered last.

Table 4 displays the results of the hierarchical regression analyses. For the Chinese compound analogy posttest, consistent with the previous ANCOVA, Chinese vs. Control was significant, which confirms significant gains from the Chinese intervention on knowledge of the structure of Chinese compounds. The contrast English vs. Control fell short of being significant (p = .06), again confirming the ANCOVA.
Table 4

Hierarchical regression analysis of performance on the Chinese and English compound analogy posttests



Chinese analogy

English analogy

R2 change



R2 change




Reading proficiency








Chinese vs. Control







English vs. Control






Reading proficiency × Chinese vs. Control







Reading proficiency × English vs. Control





Both the Reading Proficiency × Chinese vs. Control interaction and the Reading Proficiency × English vs. Control interaction were significant in the analysis of the Chinese analogy posttest. The interactions are diagramed in Fig. 1. The x-axis represents reading proficiency principal component scores, and the y-axis represents the predicted scores calculated from the regression equations for each intervention group. The interaction plot shows children with lower reading proficiency gained more from Chinese intervention than did children with higher reading proficiency. Interestingly, reverse transfer from English to Chinese was observed for the higher proficiency children, but not for the lower proficiency children. As shown in Fig. 1, among children with higher reading proficiency, those in the English intervention group performed better than those in the control group on Chinese compound analogy posttest, but among children with lower reading proficiency in the English, those in the intervention group performed slightly worse than those in the control group.
Fig. 1

Chinese compound structure analogy performance as a function of reading proficiency and intervention

For the English compound analogy posttest, consistent with the ANCOVA, the contrast of Chinese vs. Control was significant, confirming positive transfer. English vs. Control was not significant, which seems inconsistent with the pre-planned contrast following the ANCOVA. The seeming inconsistency is attributable to the significant interaction of Reading Proficiency × English vs. Control. Figure 2 shows that children with high reading proficiency who received the English intervention performed better than high proficiency readers in the control group, but children with lower reading proficiency did not benefit from the English intervention. In other words, high proficiency children demonstrated reverse transfer whereas the low proficiency children did not. A Reading Proficiency × Chinese vs. Control interaction was not found, which suggests that receiving the Chinese intervention resulted in positive transfer for children of all reading proficiency levels.
Fig. 2

English compound structure analogy performance as a function of reading proficiency and intervention

In alternative analyses, separate reading proficiency indexes in Chinese or English were computed by adding the Z scores of word reading and cloze in Chinese or English, respectively. Hierarchical regression analyses parallel to the ones described already were completed with Chinese and English reading proficiency indexes entered first, followed by the two contrasts, and then the four interactions. For both Chinese and English compound structure analogy posttest performance, the regression results were virtually identical to the results using the single reading proficiency construct as the predictor. The fact that the results were the same supports the idea of a single unified construct of language proficiency.

To investigate the compound structure effect and word type effect, a three-way repeated-measure analysis of variance was conducted in which Intervention (Chinese, English, Control) was a between-participants factor and Compound Structure (Noun+Noun, Verb+Particle, Noun+Verb, Verb+Noun) and Word Type (Familiar, Novel) were within-participant factors. For the Chinese compound structure analogy posttest, results showed both a significant compound structure effect, F (3, 495) = 23.56, p = .00, ή2 = .13, and a significant word type effect, F (1, 165) = 10.68, p = .001, ή2 = .06. Post hoc multiple comparisons showed Noun+Noun = Verb+Particle > Noun+Verb = Verb+Noun. Children performed significantly better on novel words than on familiar words.

For the English compound analogy posttest, results also showed a significant structure effect, F (3, 495) = 8.17, p = .00, ή2 = .05, and a word type effect, F (1, 165) = 8.37, p = .00, ή2 = .05. Post hoc multiple comparisons showed Verb+Particle > Noun+Noun = Verb+Noun > Noun+Verb. Similarly, Children performed significantly better on novel words than on familiar words. This pattern of results was generally consistent with our previous cross-language study (Zhang et al., 2007) in terms of better performance on Noun+Noun and Verb+Particle than on Noun+Verb and Verb+Noun and better performance on novel words than on familiar words.


The current study provides evidence for cross-language transfer between two typologically distant languages of insight into the structure of compound words. Not only was significant forward transfer from L1 (Chinese) to L2 (English) observed for children of all levels of ability, but also reverse transfer from L2 (English) to L1 (Chinese) was found for high proficiency children. Chinese fifth grade ESL children could transfer their knowledge of compound types in Chinese to comparable types in English after a short but intensive intervention on the structure of Chinese compound words. Interestingly, high proficiency children could transfer their knowledge of compound morphology from English to Chinese after a comparable intervention on compound words in English.

This study goes beyond previous studies of cross-language transfer in several ways. It is among the first studies to demonstrate cross-language transfer using a quasi-experimental approach. Two types of studies on cross-language transfer are common in the literature: cross-sectional (e.g., Durgunoglu et al., 1993) and longitudinal (e.g., Comeau, Cormier, Grandmaison, & Lacroix, 1999; Lindsey et al., 2003). Typically, transfer is considered to have been demonstrated when there is a unique contribution of a specific literacy component in L1 (at time 1) to language outcome variables in L2 (at time 2), after controlling for other literacy-related factors using regression techniques. Such correlational studies cannot provide full justification for a causal inference and the mechanism to link the specific literacy component with the outcome variables is often obscure. This quasi-experimental study supports a causal inference: After being explicitly taught Chinese or English compound morphology, Chinese fifth grade ESL children were able to apply the knowledge they had acquired about the structure of compound words in one language to analyze the structure of compound words in the other language. The mechanism of transfer is straightforward in the current study as cross-language transfer obviously stemmed from compound structures shared between Chinese and English.

Second, cross-language transfer has been mainly reported at the phonological level. The present study adds evidence of transfer on the functional or meaning level. Transfer is more likely to be observed on abstract, decontextualized tasks that demand a high level of metalinguistic awareness (Bialystok, 2001). The compound structural analogy task used in the present study falls into this category. In a comprehensive longitudinal study, Verhoeven (1994) investigated the language interdependency of L1 and L2 for 6–7 years old Turkish–Dutch bilinguals in the Netherlands. Moderate to strong interdependence was observed for a variety of metalinguistic and literacy skills including pragmatic skills (as assessed by linguistic complexity in elicited speech), phonological skills (as assessed by phoneme discrimination), word reading efficiency, and reading comprehension; however, no clear evidence of inter-language dependence was obtained at the lexical (as assessed by receptive and productive vocabulary tests), morphosyntactic, or syntactic levels (as assessed by a sentence imitation task). In line with Hancin-Bhatt and Nagy’s (1994) and Wang et al.’s (2006) studies, the current study provides somewhat different view than Verhoeven’s (1994) about the domains of possible transfer across languages.

Bidirectional influences between languages at the morphological and lexical level have been documented in naturalistic research. For instance, in a recent study, Marian and Kaushanskaya (2007) elicited narratives in Russian and in English from Russian–English bilinguals. They observed influences of each language upon the other. For example, the use of the word table in “I was very hungry and was happy to see the table” was coded as a semantic transfer from Russian to English because in Russian, the word table (stol) can mean either the actual table or the food on the table (p. 375). This is transfer in a local and concrete sense, in that it hinges on the meaning of a single word. The present study shows transfer in a more general and abstract sense, in that transfer hinged on an understanding of structural relations encompassing several types of word formation.

Not only did the present study provide the evidence of bidirectional cross-language transfer of compound morphological awareness, but we were able to delineate some of the circumstances under which transfer occurs. As Durgunoglu et al. (1993) pointed out, cross-language transfer is meaningful only when a study delineates what literacy skills transfer and under what conditions such transfer happens. Strong forward transfer from L1 to L2 was observed among children of all levels of reading proficiency, whereas reverse transfer from L2 to L1 occurred only among children with higher reading proficiency. Consistent with previous studies of cross-language transfer, which most often involved children with higher L1 proficiency than L2 proficiency, this study demonstrates strong forward transfer from L1 (Chinese) to L2 (English). It is reasonable that for second language learners with relatively little experience in L2, like the ESL children in the current study, forward transfer from L1 to L2 is the most likely pattern because learners tend to rely on L1 strategies in processing L2.

Especially interesting is the fact that this study provides the evidence of reverse transfer from the weaker language L2 to the stronger language L1 among higher proficiency children. This result can be explained in terms of the threshold hypothesis (Cummins, 1979). That is, to realize the benefits of bilingualism, children must reach a threshold level of linguistic competence. Consistent with Proctor et al.’s (2006) findings, the current study identified a limit on reverse transfer of compound morphological awareness from L2 (English) to L1 (Chinese). Reverse transfer occurred only in higher proficiency children but not in lower proficiency children. Low proficiency children got no benefit from the English compound morphology intervention; they performed no better than low proficiency children in the control group on the English posttest, so it is not surprising that they showed no generalization to Chinese. In contrast, low proficiency children gained significantly from the Chinese compound morphology intervention, and showed generalization to English.

The present study supports the idea that unified language competence underlying both L1 and L2 reading proficiency, rather than L2 proficiency only, determines the extent and direction of cross-language transfer. Previous research suggests that children with higher L2 proficiency tend to transfer morphological and lexical strategies between languages. Wang et al.’s (2006) finding of reverse transfer of compound morphological awareness from L2 (English) to L1 (Chinese) was attributed to the rapid development of L2 in the English immersion environment for the early bilinguals, so that L2 had become the children’s dominant language. In the current study, even fifth grade ESL children with limited L2 (English) proficiency (compared to L1 (Chinese) proficiency) showed reverse transfer. Therefore, it seems possible for high proficiency L2 learners to show a reverse transfer from a weaker L2 to a stronger L1, provided L2 instruction rationalizes a connection between structures in the two languages.

In line with Cummins’ (1991) notion of a common underlying construct of language proficiency, in the present study a single factor accounted for word reading and passage comprehension performance across Chinese and English. Further, similar regression results were obtained when using a language proficiency composite as the predictor as when using separate measures of L1 and L2 proficiency. Our findings are in accord with those of Mehta, Foorman, Branum-Martin, and Taylor (2005), who obtained evidence consistent with the idea of a unitary construct of literacy in a longitudinal study of English-speaking children followed from Grade 1 through Grade 4. A multi-level confirmatory factor analysis showed that variability in word reading, passage comprehension, and spelling could be adequately explained in terms of a single literacy factor.

As in a previous cross-language study (Zhang et al., 2007), on both Chinese and English compound structure posttests, children performed better on Noun+Noun and Verb+Particle compounds and performed less well on Noun+Verb and Verb+Noun compounds. These results can be explained in terms of the theory that the productivity of word formation processes influences the ease or difficulty of structural analysis of complex words. Thus, as expected, children performed better on compound types that are more productive in the languages they know.

Children performed significantly better on novel compound words than on familiar compound words on both the Chinese and English compound analogy posttests, again replicating Zhang et al. (2007). An explanation for this result is that frequent compounds have been lexicalized and do not invite decomposition into constituents. Psycholinguistic evidence suggests that in both English and Chinese familiar lexicalized compounds tend to be processed as whole units, whereas low frequency or novel compounds are more likely to be decomposed into constituent morphemes (Libben, 2006; Sandra, 1990; Taft & Zhu, 1995; van Jaarsveld & Rattink, 1988; Zhang & Peng, 1992). Morpheme decomposition is essential in order to make compound structure analogies. If familiar compounds have been learned as a whole, this would explain why children performed better on novel compounds than on familiar compounds in the current study.

One limitation of the current study is that, because of the limited English of Chinese ESL fifth graders, the English compound morphology intervention did not work as well as the Chinese intervention, especially for low proficiency children. It should be cautioned that the evidence for cross-language transfer obtained in this study was based on a brief intervention. It remains to be seen whether the transfer was a long lasting or a fragile phenomenon, as other studies suggest that children must be immersed in learning and using a second language for an extended period of time in order for cross-language benefits to accrue (Carlisle, Beeman, Davis, & Spharim, 1999). Also, worth pursuing in future studies is whether receiving instruction in morphology could improve Chinese children’s ability to segment and spell English words, as was shown in a study of native English-speaking children’s spelling (Nunes, Bryant, & Olsson, 2003).

To recapitulate, this study demonstrates cross-language transfer of compound morphological awareness in Chinese ESL children. The quasi-experimental approach goes beyond previous naturalistic and correlational studies as it allows the inference that an intervention that increased knowledge of morphology in one language caused the change in performance in the other language. The findings extend the available evidence about cross-language transfer of morphological awareness from unidirectional transfer to bidirectional transfer. The study showed that whether cross-language transfer occurs depends upon reading/language proficiency. Forward transfer from L1 to L2 occurs among children of all levels of ability. Reverse transfer from L2 to L1 was observed among high proficiency children, but not among low proficiency children. Finally, this study has implications for effective L2 instruction for Chinese ESL children: Teachers should capitalize on structural similarities between L1 and L2 to facilitate cross-language transfer.

Copyright information

© Springer Science+Business Media B.V. 2009