Printer Friendly

The roles of semantic sense and form-meaning connection in translation priming.

For long time there has been long-standing debate about whether two languages access to a common or two separate conceptual systems in bilingual memory organization literature. The earliest bilingual-memory theories on the organization of a bilingual's two languages have stated that two languages are part of either a segregated (Kolers, 1963; Scarborough, Gerard, & Cortese, 1984) or an integrated memory structure (de Groot & Nas, 1991; Kirsner, Smith, Lockhart, King, & Jain, 1984; Meyer & Ruddy, 1974). Later, researchers proposed that bilinguals had separate lexical stores for each language and a shared conceptual store (Kroll & Stewart, 1994; Potter, So, Von Eckardt, & Feldman, 1984). Kroll and Stewart (1994) have assumed that there are two levels, lexical level and conceptual level, in bilingual mental representation. The connections between the first language Ll and the second language L2 at the lexical level and a direct access from the form (i.e., the lexical level) to the meaning (i.e., the conceptual level), and the connection strength between two lexicons and conception can be stronger or weaker than others (i.e., asymmetric); therefore, bilinguals translate words faster from L2 to Ll than from the L1 to L2. Cumulative support for these ideas led to the well-established revised hierarchical model (RHM). This model has dominated in the bilingual literature for more than 15 years and still proves to be very helpful by many recent researchers investigating bilinguals with high and low proficiency (e.g., Basnight-Brown & Altarriba, 2007; Dimitropoulou, Dunabeitia, & Carreiras, 2011; Dunabeitia, Perea, & Carreiras, 2010). Further development in bilingual representation and processing models was the sense model, proposed by Finkbeiner, Forster, Nicol, and Nakamura (2004). This model also assumes that there are two levels, a lexical level and a conceptual level, in bilingual mental representation, but translation priming depends on the ratio from the prime to unprimed senses associated with the target. A high ratio leads to robust priming, whereas a low ratio triggers no priming. The sense model was further tested and proved to be useful by recent researches (Wang & Forster, 2010). The main role of all these bilingual models is to offer a sound explanation for bilingual representation and processing phenomenon. A successful model must be able to account for all the phenomena in bilingual processing.

In monolingual processing, there is little doubt that the presentation of lexical items quite automatically triggers a whole series of well-learned cognitive processes, resulting in efficient comprehension of the presented items. Even though limited number of meanings are shared widely across speakers, and speakers vary individually in knowing all the means, such efficiency and the actual amount of meaning (i.e., the range of the semantic senses, including the "core meanings" that most people would know and "peripheral meanings" that vary across individuals) that is activated by these processes are assumed to be relatively stable compared to bilingual processing. In bilingual processing, however, these cannot be taken for granted. The typical scenario is that the bilingual is more proficient with the more familiar L1 and less so with the relatively unfamiliar L2. Would the two languages differ in the efficiency of semantic activation because of differential proficiencies? For instance, are Ll lexical items more closely connected to their underlying meaning than the corresponding L2 items? Do L1 items simply possess richer meaning? These are important questions to address in bilingual processing.

One observation of these issues is the famous translation priming asymmetry in lexical decision, which is that Ll primes significantly facilitate lexical judgments about their translations (i.e., targets) in the L2, while L2 primes affect the processing of Ll targets to a much smaller extent (de Groot & Nas, 1991; Gollan, Forster, & Frost, 1997; Grainger & Frenck-Mestre, 1998; Jiang, 1999; Jiang & Forster, 2001; Keatly, Spinks, & de Gelder, 1994; Sanchez-Casas, Davis, & Garcia-Albea, 1992; Williams, 1994). Such asymmetry occurs with Ll and L2s sharing the same alphabet (e.g., Dutch and French, both using the Roman alphabet), as well as those adopting writing systems that are quite distinct (e.g., Chinese and English). Priming asymmetry is significant because it points to disparity in how word forms and meaning are associated and organized between the bilingual's stronger and weaker languages. Issues surrounding this are directly relevant to the organization of the bilingual's two lexicons.

The said disparity may take two forms. First, LI items prime better than L2 items because the former are usually richer in meaning, which is due to their high familiarity. This is captured by the sense model, which stipulates that most words are polysemous (i.e., having various semantic senses) and that the range of senses a word can possess is language specific, depending on proficiency (Finkbeiner et al., 2004). Ll words are from a more familiar language and thus on average they possess more senses (i.e., are semantically richer) than their L2 translations, in which only the core meanings are learned, remembered, and used.

The second possibility is that Ll items prime better than L2 items because the former are more strongly or closely linked to their underlying meaning than the latter. This notion is based on the expectation that lexical access is simply more efficient in the native Ll than in the nonnative L2 in terms of both meaning activation speed and richness of semantic association after lexical access. In other words, Ll items may be more strongly connected to their underlying concepts than L2 items, over and above the sheer number of senses they possess, which plays a critical role in the sense model. Item-meaning connection strength as a factor for priming asymmetry is neatly captured by the revised hierarchical model (Kroll & Stewart, 1994; Kroll & Tokowicz, 2001).

According to the sense model, translations are usually not exact because they may cover rather different senses (usages, shades of meanings) in the respective languages, resulting in asymmetry in cross-language priming. For instance, according to WordNet 1.6 (Fellbaum, 1998), the English word black has 21 senses, denoting a type of humor, a calamitous day on Wall Street (i.e., Black Monday), and so on, whereas its Japanese equivalent, kuroi, can refer to people who are evil-minded as well as to those who are well tanned or guilty of a crime. Thus, translation equivalents have language-specific shades of meaning beyond their shared, core senses.

The sense model further stipulates that the number of senses available depends on how many usages and shades of meanings a word has and how knowledgeable the individual is about those usages (Finkbeiner et al., 2004). Since bilinguals are typically more proficient and knowledgeable in their Ll than L2, Ll primes are expected to activate a larger number of senses than corresponding L2 targets, and thus priming, which is based on semantic activation, should be more obvious in the LI-prime--L2-target than the L2-prime--L1-target direction.

In Figure 1, the shared semantic senses that define translation equivalency and the language-specific senses are in dark and light grey (or white), respectively. Because bilinguals have better knowledge about the senses associated with Ll than L2 words, Ll semantic representation should be richer in terms of number of senses than the corresponding L2 representation. Therefore, the proportion of L2 senses primed by an L1 prime is close to 1/1, giving rise to robust Ll--L2 priming, whereas the proportion of Ll senses primed by an L2 prime is much smaller, resulting in weak L2--L1 priming or no L2--L1 priming at all.

[FIGURE 1 OMITTED]

The sense model can actually be tested within a language: Many-sense primes would facilitate judgments about few-sense targets as in cross-language L1--L2 priming, while few-sense primes may not promote processing many-sense targets as in cross-language L2--L1 priming. This pattern was indeed shown within a language, using the masked priming paradigm (Finkbeiner et al., 2004). Some studies have also demonstrated that lexical decision speed is sensitive to the number of senses associated with the target (Rodd, Gaskell, & Marslen-Wilson, 2002). Others argue that semantic priming reflects the efficacy of prime words to preactivate the senses of target words (Cree, McRae, & McNorgan, 1999).

On the other hand, the revised hierarchical model (Kroll & Stewart, 1994; Kroll & Tokowicz, 2001) stipulates that the connection strength between lexical form and semantic representation is weaker in the L2 than Ll. As a result, Li primes are well capable of activating the conceptual nodes shared by their L2 translations, giving rise to preactivation of these translations and thus priming. But this cannot always be assumed for L2 primes, which are less strongly linked to the underlying conceptual nodes. These results in weaker activation of the conceptual nodes and consequently a lack of L2-to-L1 priming (see Figure 2).

[FIGURE 2 OMITTED]

This study aimed to establish that both the number of semantic senses and strength of form-meaning connection are involved in determining the Chinese--English bilingual's lexical decision performance in a translation priming paradigm. Experiments 1 and 2 tested the role of semantic sense in masked translation priming. Experiment 1 was a strong test of the sense model. Items having many senses in the L2 but only a few senses in the Ll were selected. The many-sense L2 words were then used to prime their few-sense Ll equivalent targets. If the sense model is accurate about the determining role of number of senses, irrespective of what language is in question, the classic priming asymmetry (i.e., priming occurs in L1--L2 direction but not vice versa) would be reversed by using L2 many-sense primes and Ll few-sense targets. Many-sense Li and their few-sense L2 translations were used in Experiment 2; the usual asymmetry effect should be observed. Experiment 3 examined the effect of form-meaning connection strength in cross-language priming, assuming that when the number of senses is controlled, semantic priming would be primarily determined by item-meaning connection strength. Single-sense items in the Ll and L2 were used as both primes and targets. If the differential strength of the Ll and L2 form-meaning connections is a factor for cross-language priming over and above the number of senses, the classic priming asymmetry would be observed using these single-sense items. This is because Li items are assumed to be more strongly associated with meaning than are L2 items, and hence the former may serve as more effective primes. Experiment 4 further investigated the effect of form-meaning connection strength by using few-sense Ll primes and many-sense L2 targets in cross-language priming. If the RHM accounts for the bilingual translation priming, and above the sense model, there should be significant masked translation priming observed. Outcome of this research serves as a contribution to the bilingual research literature about the organizations of the bilingual's lexicons in terms of how lexical forms are stored in relation to their underlying meaning and their roles in masked translation priming.

Experiment 1

In Experiment 1, many-sense English (L2) words were used to prime the participants' lexical decisions on their few-sense Chinese (L1) equivalents. The sense model predicts significant priming, although such priming contradicts the classic lack of priming using L2 primes and L1 targets, because rich semantic representation is activated by the many-sense L2 prime that facilitates the processing of the relatively semantically "lean" Ll target.

Method

Participants. Thirty 4-year English majors at the South China Normal University participated in Experiment 1. All the participants were Chinese speakers using English as their second language, having been learning and using it for at least 13 years. As English majors, they were receiving professional English training and had passed the eighth (highest) grade of the National Test for English Majors at the time of testing. The participants are therefore considered fluent bilinguals. They received a small sum of money for their participation. Informed consent was appropriately obtained.

Materials. Ten highly proficient bilinguals majoring in English were asked to translate 150 English word items from the material list into Chinese. Another 10 English majors were asked to translate the resulting Chinese items back into English. Out of the 150 items, 98 were back-translated into the original English words, agreed upon by all the translators and back-translators. We formed 130 pairs into transition equivalents list. From these 130 items, 30 pairs of translation equivalents were selected to contain many-sense English and few-sense Chinese equivalents. The 30 pairs were then given to 20 fourth-year English majors who were asked to write out all the senses of the items. Twenty-four pairs containing English items having no fewer than six senses (e.g., black) together with Chinese equivalents having no more than two senses (e.g., [??]) were finally selected. These pairs were used as the positive items in the translation trials, requiring ayes lexical decision response. The senses (i.e., objective senses from the Oxford Advanced Learner's Dictionary and subjective senses reported by subjects) were matched for the translation (Mo = 19.5 and Ms = 7.3, respectively) and unrelated primes (Mo = 18 and Ms = 6.5, respectively). There was no significant difference of the number of sense (for both objective and subjective) between the translation and unrelated primes (all ps > .1). Because number of senses was assumed to be the most important factor for any observed asymmetry in this experiment, it was critical that the main experimental participants agreed with the English student judges on the contrast of number of senses between the two items groups. After the experiment, the main participants were therefore presented with all the items that they had encountered in the experiment and asked to write out all the senses they knew about the items. None of the Chinese and English items was seen by any of the participants as possessing more than two and fewer than six senses, respectively. The contrast in number of senses was therefore maintained for every participant throughout the experiment.

Twenty-four Chinese nonwords were constructed as targets for the negative items requiring a no response. An additional 72 English words were selected as primes for the control (unrelated) trials and the negative trials involving Chinese nonword targets. All English items had a word length between three and seven letters ([M.sub.L2] = 4) and Chinese items with a word length between one and three characters ([M.sub.L1] = 2), English word frequencies were obtained from the Corpus of Contemporary American English Frequency Database (http://www.wordfrequency.info), and Chinese word frequencies were obtained from the SUBTLEX-CH database (Cai & Brysbaert, 2010). Mean frequencies for the translation primes, unrelated primes, and targets were 390.4, 399.8, and 72.6, respectively. Such are occurrence frequencies per million. Paired samples t tests showed no significant difference of mean frequencies between the translation/unrelated primes and the targets and between the translation equivalent and unrelated primes (all ps > .1). What is more important, the overall frequencies for all primes and targets were relatively high in their own language-specific corpus, and the primes and targets were also evaluated by subjects as frequently used words. Construction of the Chinese nonwords was modeled after the method used by Finkbeiner et al. (2004) to derive Japanese nonwords. Chinese lexical syllables were randomly combined into two 2-syllable nonwords. No target or prime items were repeated in the experiment. No cognates were involved (see Appendix A).
Appendix A: Experimental Items Used in Experiments 1 and 4

Word A  word  Word A  Word  Word A  Word  Word A   Word
           B             B             B              B

hand    [??]  black   [??]  post    [??]  cross    [??]

run     [??]  drink   [??]  play    [??]  cool     [??]

point   [??]  dress   [??]  form    [??]  bear     [??]

hit     [??]  paper   [??]  back    [??]  matter   [??]

have    [??]  fast    [??]  mark    [??]  express  [??]

head    [??]  judge   [??]  stand   [??]  hard     [??]


Design and procedure. On each trial, the participant was required to decide as quickly and accurately as possible whether the Chinese target, closely following an English prime, was a real word. Eight practice trials preceded the main test trials, which were randomly presented. Trial presentation was done on a personal computer and controlled by the E-Prime software. Participants received written experimental instruction and gave informed consent before testing.

Each trial consisted of the following sequence of events, adapted from Grainger and Frenck-Mestre (1998) and also from Finkbeiner et al. (2004). A forward mask (#####) appeared at the center of the computer screen for 500 ms, immediately followed by an English prime word in lowercase letters, appearing for 50 ms. Then a backward mask followed for 150 ms, and the Chinese target appeared for 500 ms, during which the participant made a lexical decision on it. The forward and backward masks were in different font types and sizes, Arial Black 18 and Times New Roman 11, respectively. The targets were divided into two lists to rotate against the primes across participants so that the targets had equal chances to appear in the different experimental conditions.

Results and Discussion

Mean response times from correct responses were calculated. Data from three participants were discarded because their overall accuracies were lower than 80%. For each individual participant, response times that were 2 SDs above or below his or her own mean were excluded from further analyses. All the response times from incorrect trials were removed. These procedures resulted in the removal of 1.9% of the response times. The overall error rate was 1.7%; error rates did not differ between the two experimental conditions and thus they were not further analyzed.

Mean response times were 625 ms (SD = 101.9 ms) and 673 ms (SD = 105.0 ms) in the translation equivalent (i.e., the prime and target were translation equivalents) and unrelated condition (i.e., the prime and target were semantically unrelated), respectively. An ANOVA showed that the facilitation priming effect was significant, F1(1, 26) = 30.1, p < .01; F2(1, 23) = 17.258, p < .01. Response accuracies did not differ between the priming conditions.

The findings of Experiment 1 support the sense model, which stipulates that the ratio between the senses of the prime and those of the target is a critical factor for cross-language priming asymmetry, assuming that L2 items usually have fewer available senses than the corresponding Ll items. A many-sense prime would facilitate the processing of a target more than would a few-sense prime, irrespective of whether the prime and target are in an L1 or L2. The present finding that many-sense L2 primes effectively primed few-sense Ll targets provides strong evidence for the hypothesis, as the classic asymmetry is reversed in accordance with the prime-target sense ratio.

Experiment 2

In Experiment 2, we attempted to replicate the usual asymmetry effect using many-sense Ll (Llmany) and few-sense L2 (L2few) items. If the effects generated in Experiment 1 had been due to the unusual L1-L2 sense ratio (i.e., few-sense Ll vs. many-sense L2 items), as predicted by the sense model, the usual asymmetry effect should be replicated in Experiment 2, with the sense ratio restored to a more normal level (i.e., L1many vs. L2few items). That is, there will be significant priming in Llmany-L2few direction, but not vice versa.

Method

Participants. Fifty-six English majors at the South China Normal University participated in Experiment 2. They were very similar to the Experiment 1 participants in terms of general socioeconomic, linguistic, and educational background. Participants received a small sum of money for their participation, and informed consent was appropriately obtained. We randomly assigned 28 of the participants to Experiment 2a (Llmany-L2few), and the rest were randomly assigned to Experiment 2b (L2few-Llmany). In the control condition, unrelated prime and targets pairs were used.

Materials. The experimental stimuli used in the Llmany-L2few priming condition consisted of 96 word pairs, 24 of which were chosen from the rest of 100 critical translation equivalent items developed in Experiment 1, making up the positive word pairs (e.g., [??]-eye). All the English items had a word length between three and eight ([M.sub.L2] = 5) and Chinese items had one character ([M.sub.L1] = 1). Most of the polysemous Chinese words are one-character words; therefore, we selected one-character Chinese words as our experimental materials. Twenty-four nonwords were chosen as targets, making up the negative (nonword) items. An additional 72 Chinese words (or English words in the L2few-Llmany priming condition) were selected to serve as primes in the control trials. They were semantically unrelated to their targets, and were matched with the critical primes for frequency and word length. English word frequencies were obtained from the Corpus of Contemporary American English Frequency Database (http://www.wordfrequency.info), and Chinese word frequencies were obtained from the SUBTLEX-CH database (Cai & Brysbaert, 2010). In Experiment 2a, mean frequencies for the translation equivalent primes, unrelated primes, and targets were 789.0, 760.1, and 91.7, respectively. Such were occurrence frequencies per million. The mean frequencies between the translation/unrelated primes and the targets did not differ significantly (both ps > .05). The translation equivalent and unrelated primes did not differ significantly in occurrence frequency (p> .1). What is more important, the overall frequencies for all primes and targets were relatively high in their own language-specific corpus, and all the words were also evaluated by subjects as frequently used words. Eight pairs of materials were chosen as the practice items before the experiment. The English nonwords were generated by the ARC Nonword Database (http://www.maccs.mq.edu.auk-nwdb/). In the L2few-L1 many priming condition in Experiment 2b, the materials and design were identical to those in the Llmany-L2few condition in Experiment 2a but were arranged in the opposite direction (i.e., few-sense L2 primes and many-sense L1 targets). Mean frequencies for the translation equivalent primes, unrelated primes, and targets were 91.7, 92.1, and 789.0, respectively. Such were occurrence frequencies per million. The translation equivalent and unrelated primes did not differ in occurrence frequency (p> .1). Even though there was difference between the translation/unrelated primes and the targets, it did not differ significantly (both ps > .05). The overall frequencies for all primes and targets were relatively high in their own language-specific corpus, and all the words were also evaluated by subjects as frequently used words. The design of Chinese nonwords followed that in Experiment I.

Because the control of item sense was critical to the design of this experiment, we used both updated dictionaries and proficient bilinguals to ascertain the number of senses each test item possessed. From the 100 items adopted from Experiment 1, we selected 35 pairs to contain many-sense Chinese items and few-sense English translation equivalents. We then gave these 35 pairs to 20 fourth-year English majors, who were asked to write out all the senses of the items. In addition to asking them to write out all the senses they knew about the items, we asked the English majors to evaluate how common and familiar these items were. Eventually, we selected 24 out of the 35 pairs containing Chinese items having no fewer than six senses, together with English equivalents having no more than two senses. The senses (i.e., the objective senses from Xiandai Hanyu Cidian (Dictionary editing office, 2001) and the subjective senses reported by subjects) were matched in Ll in Experiment 2a for the translation (M = 11.25 and M = 6.5), and unrelated primes (M = 12 for objective and M = 7.5 for subjective). In Experiment 2b, we also matched the senses in L2 for translation primes (M = 4 for objective and M = 2 for subjective) and for unrelated primes (M = 4.5 for objective and M = 1.5 for subjective). There was no significant difference of the number of sense (objective and subjective) between translation and unrelated primes in both Experiment 2a and 2b (all ps > .1).

As a manipulation check, after the experiment, we asked all the participants to write out the senses of the items that they had encountered in the experiment. They all agreed that the many-sense items picked by the English majors indeed had more than six senses and few-sense items picked by them had indeed only one or two senses. Hence, manipulation was deemed successful. These 24 pairs were used in both the L1-to-L2 and L2-to-L1 priming conditions in Experiments 2a and 2b, respectively, which constituted a between-subject factor. We selected additional word items for the control (unrelated) trials. We matched all the word items for length and occurrence frequency and selected additional word items for the control (unrelated) trials (see Appendix B).
Appendix B: Experimental Items Used in Experiment 2

Word  Word B   Word  Word B  Word  Word B  Word   Word B
A                 A             A             A

[??]  soil     [??]  shout   [??]  mouth   [??]  accurate

[??]  install  [??]  grow    [??]  noodle  [??]  smart

[??]  meeting  [??]  minute  [??]  hair    [??]  spring

[??]  sky      [??]  light   [??]  bag     [??]  body

[??]  flower   [??]  number  [??]  eye     [??]  picture

[??]  letter   [??]  cover   [??]  finger  [??]  drip


Design and procedure. The design and procedure were identical to those in Experiment 1, except that the 56 participants were randomly assigned into two groups, working with Llmany--L2few and L2few--Llmany priming experiments, respectively.

Results and Discussion

Experiment 2a. Mean response times for the correct responses were calculated. The data from one participant whose overall accuracy was lower than 80% were discarded. Also, responses more than 2 SDs away from the respective participant's mean were excluded from all subsequent analyses. These trimming procedures resulted in the removal of 4.2% of all the response times. Mean response times were 668 ms for targets preceded by translation equivalent primes and 745 ms for targets preceded by unrelated primes. An ANOVA showed that this facilitation effect of 77 ms was significant, F1(1, 26) = 12.5, p < .01; F2(1, 23) = 12.500, p < .01. The present result that Ll primes facilitate the recognition of L2 translation equivalent targets is consistent with previous findings from Dutch--English, Hebrew--English, and Japanese--English bilinguals (de Groot & Nas, 1991; Finkbeiner et al., 2004; Gollan et al., 1997; Jiang, 1999; Jiang & Forster, 2001; Keatly et al., 1994; Williams, 1994).

Experiment 2b. Mean response times for the correct responses were calculated. The same data trimming procedure used in Experiment 2a was employed. The data from two participants whose accuracies were lower than 80% were discarded. Responses more than 2 SDs away from the respective mean of each participant were excluded from subsequent analyses. This procedure resulted in the removal of 5% of all the reaction times. Mean response times were 629 ms for targets preceded by translation equivalent primes and 659 ms for targets preceded by unrelated primes. An ANOVA showed that this facilitation effect of 30 ms was insignificant, F1(1, 25) = 1.7, p> .05; F2(1, 23) = 1.786, p> .05.

The data of Experiment 2 were submitted to a 2 (priming direction, between-subject) x 2 (prime-target relatedness, within-subject) mixed-designed ANOVA. There was a significant main effect of the prime-target relatedness, F(1, 47) = 4.7, p < .05. There was a significant interaction, F(1, 47) = 6.2, p < .05. Overall, the participants responded faster in the translation equivalent than in the control condition. Hence, the usual translation priming asymmetry was replicated in Experiment 2, using many-sense Ll and few-sense L2 items (Finkbeiner et al., 2004; Gollan et al., 1997; Grainger & Frenck-Mestre, 1998; Jiang, 1999; Jiang & Forster, 2001; Keatly et al., 1994; Sanchez-Casas et al., 1992; Williams, 1994).(1)(2)

But is number of senses the only factor behind translation priming asymmetry? What would happen if the number of senses is controlled across the two languages? If Ll lexical forms are simply more strongly connected to their semantic representations than L2 lexical forms, regardless of the number of senses each of these items may have, then the usual asymmetry may be observed after the number of senses is controlled. Therefore, Ll primes should be more effective than L2 primes in activating a rich semantic representation on which priming is based, giving rise to cross-language priming asymmetry. Note that this effect is independent of the relative numbers of senses belonging to the Ll and L2 items in question. Furthermore, note that if the senses of both Ll and L2 were changed to be only one sense, there should be symmetric priming in both directions. Otherwise, the sense model may not account for this phenomenon but RHM, which has assumed that asymmetric form-meaning connection strength between the two lexicons and their shared conception causing asymmetric translation priming. Experiment 3 was designed to investigate and demonstrate an effect of form-meaning connection strength in bilingual priming, controlling for the relative numbers of senses of the items from the two languages.

Experiment 3

Only single-sense primes and targets were used in Experiment 3 so that any asymmetry in priming could not be attributed to prime-target sense ratio. Differential form-meaning connection strengths associated with the L1 and L2 items could therefore be evoked to explain any observed asymmetry.

Method

Participants. Fifty-eight English majors at the South China Normal University participated in Experiment 3. They were very similar to the participants in Experiment 2 in terms of general socioeconomic, educational, and other characteristics. Participants received a small sum of money for their participation, and informed consent was appropriately obtained. We randomly assigned 29 participants to Experiment 3a (single-sense Ll to single-sense L2 priming) and the rest of them were randomly assigned to Experiment 3b (single-sense L2 to single-sense L1). In the control condition, unrelated prime and targets pairs were used.

Materials. We selected 35 pairs of translation equivalents, including only single-sense items, from the rest of the pool of 65 critical-translation-equivalent pairs created before Experiment 1, which were mainly revised from the materials used in Finkbeiner et al.'s study (2004). The 35 pairs used in Experiment 3 were regarded and selected as single-sense items because both the Chinese and English translation equivalents were listed in all the major dictionaries in the respective languages as having one sense only, and 20 native Chinese-speaking university students majoring in English all agreed that they were single-sense words. In addition to asking them to write out all the senses they knew about the items, we asked the English majors to evaluate how common and familiar these items were. Eventually we selected 24 out of the 35 pairs of translations because they were frequently used single-sense items (e.g., [??]-movie). All Chinese words had a word length ranging from one to three characters (M[L.sub.1] = 2), and English words had a word length ranging from three to eight letters (M[L.sub.2] = 5). These 24 pairs were used in both the L1-to-L2 and L2-to-L1 priming directions, which constituted a between-subject factor. Additional word items were selected for the control (unrelated) trials. The senses were matched in Ll in Experiment 3a for translation and unrelated primes (all words have only one sense for both objective and subjective meanings) and also matched in L2 in Experiment 3b for translation and unrelated primes (all words have only one sense for both objective and subjective meanings). Frequencies of English words were obtained from the Corpus of Contemporary American English Frequency Database (http://www.wordfrequency.info), and frequencies of Chinese words were obtained from the SUBTLEX-CH database (Cal & Brysbaert, 2010). In Experiment 3a, mean frequencies for the translation-equivalent primes, unrelated primes, and targets were 37.4, 38.7, and 37.6, respectively. Such are occurrence frequencies per million. The translation primes/unrelated primes and targets did not differ in occurrence frequency. The translation equivalent and unrelated primes also did not differ in occurrence frequency (all [p.sup.s] Greater than .1). In Experiment 3b, mean frequencies for the translation equivalent primes, unrelated primes, and targets were 37.6, 37.5, and 37.4, respectively. Such were occurrence frequencies per million. The translation equivalent and unrelated primes did not significantly differ in occurrence frequency (p Greater than .1). Additional word items were selected for the control (unrelated) trials. English and Chinese nonwords were used in the negative trials requiring a no lexical decision response. The English nonwords were generated by the ARC Nonword Database (http://www.maccs.mq.edu.au/~nwdb/), whereas the Chinese nonwords were adapted from Experiment 1. Nonword items were matched for length. No items, words or nonwords, were repeated in the experiment. No cognates were used.

Because the selection of single-sense items was critical to the design of this experiment, we used both updated dictionaries and proficient bilinguals to ascertain that each test item carried only one sense. We asked English majors to write out the meanings of the items because they were knowledgeable in English, and if they could produce only one sense for a particular English item then it was likely to be single-sense to average participants, as well. Nevertheless, we acknowledge that, in addition to the dictionaries, we relied on the subjective judgments of a relatively small group of raters (n = 20) in the selection of single-sense items, which therefore might not be absolutely precise. But note that what was important was the participants' subjective understanding of the items rather than the items' actual objective meanings. As a manipulation check, after the experiment we asked the main participants to write out all the senses of the single-sense items that they had encountered in the experiment. They all agreed that the single-sense items picked by the English majors indeed had only one sense to them (see Appendix C).
Appendix C: Experimental Items Used in Experiment 3

Word A    Word   Word A   Word  Word A   Word  Word A  Word
             B               B              B             B

oxygen    [??]  cottage   [??]  oar      [??]  lawyer  [??]

movie     [??]  pebble    [??]  wheat    [??]  lawn    [??]

sofa      [??]  ankle     [??]  lung     [??]  bird    [??]

skull     [??]  kitten    [??]  poverty  [??]  gale    [??]

umbrella  [??]  mosquito  [??]  fist     [??]  creek   [??]

teacher   [??]  lengthy   [??]  beer     [??]  wrist   [??]


Design and procedure. The design and procedure were identical to those of Experiment 1, except that the 58 participants were randomly assigned to two experimental

Results and Discussion

Mean response times from correct responses were calculated. The data-trimming procedures used in Experiment 1 were again employed. The data from seven participants whose overall accuracies were below 80% were discarded, leaving 24 and 27 participants in the singleL1--singleL2 and singleL2--singleLl priming groups, respectively. For each individual participant, response times more than 2 SDs above or below the respective individual's overall mean were excluded from further analyses. Such trimming procedures resulted in the removal of 0.6% of the response times. Overall error rate was 0.3%; error rates from the different experimental conditions did not differ and were thus not further analyzed. Mean response times in the L1-to-L2 condition were 728 ms (SD = 277.2 ms) for the targets preceded by their translation equivalents as primes, and 803 ms (SD = 236.8 ms) for those preceded by semantically unrelated primes. In the L2-to-L1 condition, mean response times were 757 ms (SD = 166.4 ms) for the targets preceded by translation equivalents and 753 ms (SD = 223.3 ms) for those preceded by unrelated primes. The data were submitted to a 2 (priming direction, between-subject) x 2 (prime-target relatedness, within-subject) mixed-designed ANOVA. The significant effects were the prime-target relatedness main effect, F(1, 49) = 4.7, p = .03, and the interaction, F(1, 49) = 5.8, p = .02. Overall, the participants responded faster in the translation-equivalent condition than in the control condition, but this effect depended on priming direction. Follow-up analyses showed that the relatedness simple effect was significant in the L1--L2, F1(1, 23) = 8.4, p < .01; F2(1, 23) = 6.88, p < .01, but not in the L2--L1 priming condition, F1(1, 26) = .03, p> .05; F2(1, 23) = .118, p> .05. Therefore, the classic cross-language priming asymmetry was observed with single-sense primes and targets. Experiment 3 found that there was robust translation priming from one-sense Ll prime to one-sense L2 prime, but not vice versa. The sense model failed to explain this phenomenon. According to the sense model, there should have been symmetric translation priming due to the symmetric sense ratio in both L1--L2 and L2--L1 direction. In this situation, the asymmetry could not be attributed to the relative numbers of senses activated by the prime and target; differential strengths of connection between lexical form and meaning across the two languages were evoked as an account for the asymmetry.

Experiment 4

Does the form-meaning connection strength play an important role in asymmetric translation priming? Which model accounts for bilingual asymmetric translation priming, the sense model or the RHM? Or both? Experiment 4 further investigated the role of connection strength between lexical form and meaning by exploring priming effect in few-sense L1 to many-sense L2 (L1few--L2many) direction in lexical decision. If RHM plays a more important role in this situation, there should be robust priming in L1few--L2many direction. If the sense model plays a more important role, significant translation priming will not be observed in this direction.

Method

Participants. We recruited an additional 30 English majors from South China Normal University for the experiment. The requirements for participants and the gender structure are identical to those in Experiment 1.

Materials and design. The content and extraction method of the experimental materials were identical to those in Experiment 1, except the obvious difference that the priming direction was reverse (e.g., X'0, [??]--black). Primes were few-sense Ll words, and targets were many-sense L2 words. The design and the structure of the material were identical to those in Experiment 1. The English items had a word length of between three and seven ([M.sub.L2] = 4.5) and the Chinese items had a word length of between one and three ([M.sub.L1] = 1.8). The senses (i.e., the objective senses from Xiandai Hanyu Cidian (Dictionary editing office, 2001) and the subjective senses reported by subjects were matched for the translation ([M.sub.o] = 1.8 and Ms = 1, respectively), and unrelated primes ([M.sub.o] = 1.6 and Ms = 1, respectively). There was no significant difference between the senses for the translation and unrelated primes ([p.sup.s] Greater than .1). Mean frequencies for the translation equivalent primes, unrelated primes, and targets were 72.6, 72.2, and 390.4, respectively. Such are occurrence frequencies per million. The translation/unrelated primes did not differ significantly from the targets in frequency. The translation equivalent and unrelated primes also did not differ significantly in occurrence frequency (all [p.sup.s] Greater than .1). Frequency of English words were obtained from the Corpus of Contemporary American English Frequency Database (http://www.wordfrequency.info), and frequency of Chinese words were obtained from the SUBTLEX-CH database (Cai & Brysbaert, 2010; see Appendix A).

Design and procedure. The design and procedure was largely identical to that of Experiment 1.

Results and discussion. Mean response times for correct responses were calculated. Data of three participants whose accuracy was below 80% and responses more than 2.0 SD above or below each participant's mean were excluded from the analyses. The same outlier criterion was used as those in the previous experiments. Trimming resulted in removal of 0.2% of the response times. Mean response times were 568 ms for targets (e.g., [??]) preceded by a translation prime (e.g., black) and 598 ms for targets preceded by an unrelated prime. An ANOVA showed that this facilitation effect of 30 ms was significantly reliable, F1(1, 26) = 16.425, p Less than .001; F2(1, 23) = 17.258, p Less than .001. The average accuracy rate didn't differ significantly (1.3%). As was seen from the result, the response under translation-priming condition was faster than that under unrelated condition, and the evidence for priming effect in terms of response times was not at the cost of a decrease of accuracy rate.

The finding showed that Ll few-sense-prime can facilitate the decision time on L2 many-sense-target. Combined with the finding that there was no priming in L2few--Limany (Experiment 2b), L2single--L1single direction (Experiment 3b), but robust priming in Llsingle--L2single direction (Experiment 3a), it is believed that even when Ll has only few senses as a prime, its strong link from its lexical form to its meaning is able to in favor of a robust translation priming. Therefore RHM accounts for this bilingual processing phenomenon.

General Discussion

This study highlights relative numbers of senses and differential strengths of form-meaning connections between Ll and L2 primes and targets as critical factors for bilingual priming asymmetry in lexical decision. Asymmetry refers to stronger priming resulting from Ll-primes and L2-targets than to L2-primes and Ll-targets in lexical decision. According to the sense model, the number of senses made available by presenting a word depends on the individual's knowledge about the language. Because the bilingual is more proficient in Ll than in L2, an Ll prime typically activates many senses that effectively cover all the corresponding senses of its translation equivalent used as a target in L1--L2 priming, leading to a large ratio from the prime to unprimed senses relevant to the target. This results in pronounced priming. On the other hand, in L2--L1 priming, an L2 prime activates only a few senses (leading to a small senses ratio from the prime to the target), which may not give rise to sufficient semantic overlap with its many-sense L1 equivalent for priming to occur (Finkbeiner et al., 2004).

But when the number of senses is controlled across priming conditions, what other factors may affect the relative priming magnitudes using Ll versus L2 primes and targets? RHM has proposed that because L1 is the more proficient and familiar language, its lexical items may be more strongly connected to their underlying meaning. Using Ll primes may therefore result in more efficient semantic activation than using L2 primes in terms of activation speed and richness of semantic association after lexical access. We thus predicted that Ll-L2 priming would be more pronounced than L2-L1 priming with number of senses controlled, and RHM can give an explanation.

In Experiment 1 we examined translation equivalent priming using many-sense L2 primes and few-sense Ll targets. The sense model predicts significant priming in this situation even though L2-to-L1 priming has seldom been observed previously. In Experiment 2 we used many-sense Ll and few-sense L2 items so as to mimic what usually happens with the typical bilingual. Translation priming asymmetry should be observed. In Experiment 3, only single-sense primes and targets were used. If the connection strength between lexical forms and their meaning is important, the usual pattern of priming asymmetry would be observed even with single-sense primes and targets. In Experiment 4 we examined translation-equivalent priming using few-sense Ll primes and many-sense L2 targets. According to the sense model, there should not be reliable priming in this direction, except that the strong connection strength between lexical form and its meaning can counterbalance the lack of high ratio of from the primed to unprimed senses associated to the targets.

The results from these experiments are positive, indicating that the bilingual processes lexical items from the two languages rather differently. L1 items are more readily interpreted into their rich meaning than is their L2 counterparts, which are supported by a semantic representation that is not as rich. These differences manifest themselves as asymmetry in translation priming. Note that either of these factors constitutes a sufficient condition for priming. In Experiment 1, the effect of many-sense primes and few-sense targets actually overrode form-meaning connection strength, leading to an observable priming effect despite the fact that L2 materials were used as primes. This finding was consistent with some previous studies that found symmetric robust priming in L1-L2 and L2-L1 directions using highly proficient L2 bilinguals (e.g., Basnght-Brown & Altarriba, 2007; Dimitropoulou et al., 2011; Dufiabeitia, et al., 2010). RHM could also explain the result of Experiment 1 that when participants know many semantic senses of L2, they are actually proficient in L2, hence their strong connection strength between L2 lexical form and meaning may trigger the robust priming in L2many-Llfew direction. In Experiment 2, the usual asymmetry was replicated. In Experiment 3, when the sense factor was controlled, form-meaning connection strength alone gave rise to the usual asymmetry. Experiment 4 showed that Llfew was still able to prime L2many because of the earlier-formed stronger connection between Ll form and the shared conceptual meaning.' Hence, the whole finding of the present study has suggested that the sense model accounts for some phenomena in bilingual lexical processing, but not for all (e.g., Experiment 4), and the strength of association between lexical items and underlying meaning is critical and essential in translation priming. The sense ratio proposed by the sense model can compensate and counterbalance the connection strength proposed by the RHM in bilingual priming.

What we have found more interesting is the additional discovery of the importance of the role that the semantic senses of the prime plays in translation priming observed in the present research. If the prime of a specific language has more semantic senses than the target (i.e., no matter L1 or L2), hence a large ratio of the primed to the unprimed senses of the target causes significant translation priming (e.g., the robust L2many-Llfew priming in Experiment 1; and L1many-L2few priming in Experiment 2a; and Llmany-L1few priming in Finkbeiner et al.'s study, 2004). The sense model theretore accounts tor tnese phenomena. If the prime has very few semantic senses compared to the target, a small ratio from the primed to the unprimed senses of target will not trigger significant priming (e.g., no L2single-Lisingle priming in Experiment 3b; no L2few-Llmany priming in Experiment 2b). In this situation, no robust priming can be obtained except that there is strong form-meaning connection strength, which can counterbalance the lack of semantic senses in translation priming asymmetry. Therefore, no priming occurred in the L2single-Llsingle and L2few-Llmany direction, but strong priming occurred in the Llfew-L2many and Llsingle-L2single direction. The RHM therefore accounts for these phenomena. Our findings have suggested that the factor of form (i.e., form-meaning connection strength) and the factor of semantic sense both play important roles in translation priming asymmetry. These two factors interact, counterbalance, and facilitate each other in translation priming.

Another discovery is that the common dimension that appears to run through the two explanations is language proficiency. Ll words are more effective primes than L2 words because the former carry more senses, which in turn is due to high proficiency. That Ll items are more closely connected to meaning than are L2 items is also due to participants' high proficiency. If participants have high proficiency in L2, knowing many meanings of L2 words (i.e., L2 becomes dominant language), the large sense ratio will trigger significant translation priming in the L2many-to-L1few direction; therefore, they can translate from L2 to Ll very fast, too. If participants have low proficiency in L2, they will not be able to translate from L2 to Ll fast (i.e., robust priming will not be obtained). Our findings were consistent with the previous findings (e.g., Grainger & Frenck-Mestre, 1998; Kroll & Stewart, 1990, 1994, 2001) and some recent studies have demonstrated that L2 proficiency to be a determining factor in causing directional differences in cross-language translation (e.g., Basnght-Brown & Altarriba, 2007; Dimitropoulou et al., 2011; Dufiabeitia et al., 2010). In Basnght-Brown and Altarriba's study (2007), the bilinguals could experience a dominance shift where their L2 can actually act as their Ll. Therefore, an interesting issue to be addressed in future research is whether there are processing differences between the two languages in the highly balanced bilingual who has acquired both languages in a natural speech environment, as adult bilingualism usually develops from a formal classroom setting relying on written materials. The interplay between learning context and nature of processing provides valuable information for building more sophisticated models of bilingual processing.

References

BASNGHT-BROWN, D. M., & ALTARRIBA, J. (2007). Differences in semantic and translation priming across languages: The role of language direction and language dominance. Memory & Cognition, 35, 953-965. doi: 10.3758/BF03193468

CAI, Q., & BRYSBAERT, M. (2010). SUBTLEX-CH: Chinese word and character frequencies based on film subtitles [Data file]. Plos ONE, 5(6), el0729. doi:10.1371/journal .pone.0010729

CORPUS OF CONTEMPORARY AMERICAN ENGLISH FREQUENCY DATABASe. Retrieved from http://www.wordfrequency.info

CREE, G., MCRAE, K., & MCNORGAN, C. (1999). An attractor model of lexical conceptual processing: Simulating semantic priming. Cognitive Science, 23(3), 371-414. doi:10.1207/s15516709cog2303_4

DE GROOT, A. M. B., & NAS, G. L. J. (1991). Lexical representation of cognates and noncognates in compound bilinguals. Journal of Memory and Language, 30, 90-123. doi:10.1016/0749-596X

DICTIONARY EDITING OFFICE, LINGUISTICS RESEARCH CENTER, CHINA SOCIAL SCIENCE FACULTY. (2001). Xiandai Hanyu Cidian. Beijing: Shang Wu Yin Shu Guan.

DIMITROPOULOU, M.. DUNABEITIA, J.A., & CARRE1RAS, M. (2011) Masked translation priming effects with low proficient bilinguals. Memory and Cognition, 39, 260-275.

DUNABEITIA, J. A., PEREA, M., & CARREIRAS, M. (2010). Masked translation priming effects with highly proficient simultaneous bilinguals. Experimental Psychology, 57, 98-107. doi: 10.1027/1618-3169/a000013

ENGLISH NON-WORDS [ARC NON-WORD DATABASE]. Retrieved from http://www.maccs.mq.edu.au/-nwdb/ FELLBAUM, C. (Ed.). (1998). WordNet: An Electronic Lexical Database. Proceedings of 11th Eurographics Workshop on Rendering.

F1NKBEINER, M., FORSTER, K., NICOL, J., & NAKAMURA, K. (2004). The role of polysemy in masked semantic and translation priming. Journal of Memory and Language, 51, 1-22. doi:10.1016/j.jm1.2004.01.004

GOLLAN, T., FORSTER, K. I., & FROST, R. (1997). Translation priming with different alphabets: Masked priming with cognates and noncognates in Hebrew-English bilinguals. Journal of Experimental Psychology: Learning, Memory and Cognition, 23, 1122-1139. doi:10.1037/0278-7393.23.5.1122

GRAINGER, J., & FRENCK-MESTRE, C. (1998). Masked priming by translation equivalents in proficient bilinguals. Language and Cognitive Processes, 13, 601-623. doi: 10.1080/016909698386393

JIANG, N. (1999). Testing processing explanations for the asymmetry in masked cross-language priming. Bilingualism: Language and Cognition, 2, 59-75. doi:10.1017/S1366728999000152

JIANG, N., & FORSTER, K. I. (2001). Cross-language priming asymmetries in lexical decision and episodic recognition. Journal of Memory and Language, 44, 32-51. doi:10.1006/jmla.2000.2737

KEATLY, C. W., SPIN KS, J. A., & DE GELDER, B. (1994). Asymmetrical cross-language priming effects. Memory & Cognition, 22, 70-84. doi:10.3758/BF03202763

KROLL, J. F., & DE GROOT, A. M. B. (1997). Lexical and conceptual memory in the bilingual: Mapping for to meaning in two languages. In A. M. B. de Groot & J. F. Kroll (Eds.), Tutorials in bilingualism: Psycholinguistic perspectives (pp. 169199). Hillsdale, NJ: Erlbaum.

KIRSNER, K., SMITH, M. C., LOCKHART, R. S., KING, M.L., & JAIN, M. (1984). The bilingual lexicon: Language-specific units in an integrated network. Journal of Verbal Learning and Verbal Behavior, 23, 519-539. doi: 10.1016/S0022-5371(84)90336-0

KOLERS. (1963). Interlingual word associations. Journal of Verbal Learning and Verbal Behavior, 2, 291-300. doi: 10.1016/S0022-5371(63)80097-3

KROLL, J. F., & STEWART, E. (1994). Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language, 33, 149-174. doi:10.1006/jmla.1994.1008

KROLL, .J. F., & TOKOWICZ, N. (2001). The development of conceptual representation for words in a second language. In J. L. Nicol (Ed.), One mind, two languages: Bilingual language processing (pp. 49-71). Malden, MA: Blackwell.

MEYER, D. E., & RUDDY, M. G. (1974), Bilingual word recognition: Organization and retrieval of alternative lexical codes. Eastern Psychological Association Meeting.

POTTER, M. C., SO, K. F., VON ECKARDT, B., & FELDMAN, L. B. (1984). Lexical and conceptual representation in beginning and proficient bilinguals. Journal of Verbal Learning & Verbal Behavior, 23, 23-38. doi:10.1016/S0022-5371(84)90489-4

ROOD, J., GASKELL, G., & N1ARSLEN-WILSON, W. (2002). Making sense of semantic ambiguity: Semantic competition in lexical access. Journal of Memory and Langauge, 46, 245-266. doi:10.1006/jmla.2001.2810

SANCHEZ-CASAS, R. M., DAVIS, C. W., & GARCIA-ALBEA, J. E. (1992). bilingual lexical processing: Exploring the cognate/non-cognate distinction. European Journal of Cognitive Psychology, 4, 293-310. doi:10.1080/09541449208406189

SCARBOROUGH, D. L., GERARD, L., & CORTESE, C. (1984). Independence of lexical access in bilingual word recognition. Journal of Verbal Learning and Verbal Behavior, 23, 84-99. doi:10.1016/S0022-5371(84)90519-X

WANG, X., & FORSTER, K. I. (2010). Masked translation priming in semantic categorization: Testing the sense model. Bilingualism: Language and Cognition, 10(3), 1-14. doi: 10.1017/S1366728909990502

WILLIAMS, J. N. (1994). The relationship between word meanings in the first and second language: Evidence for a common, but restricted, semantic code. European Journal of Cognitive Psychology, 6, 195-220. doi:10.1080/09541449408520143

(1.) Some may be concerned that the current pattern of findings could be distorted by the differences in the number of target characters between Experiment 1 (many-sense L2 prime & few-sense Ll target) and Experiment 2 (few-sense L2 prime & more-sense Ll target). We separately analyzed the translation priming for one-, two-, and three-character targets in Experiment 1, and the analysis showed the 23 interaction between the factor of relatedness and the factor of number of characters in Experiment 1 was not significant (p> .05), which means the number of characters in a word does not matter in our study.

(2.) Even though we think it is unreasonable to compare word frequency of targets in Chinese and English (since they are different languages, in which Chinese words are formed by single characters and English words are formed by letter strings), to exclude the possible argument that the present asymmetric priming in Exp. 2 was caused by asymmetric word frequencies between the L2few targets in Exp. 2a and the Ll many targets in Exp. 211, we did extra analyses. A t test showed no significant difference between raw word frequency of L2few targets and that of Llmany targets (p> .05) but significance between their log-transformed forms, t(23) = 4.109, p < .05. However, the correlation between the log-transformed form of word frequencies and size of priming was very low (r = .12 and r = .16, respectively; both ps > .1). Furthermore, ANCOVA found the same robust priming in Exp. 2a but no priming in Exp. 2b after controlling the log-transformed word frequencies of all targets as covariates and no significant effect for the word frequency in both priming directions (both ps > .1). These analyses have demonstrated that the asymmetric priming in Exp. 2 was not caused by the word frequency difference.

(3.) Some may be concerned that the priming in Exp. 1 was influenced by the low frequency of the targets. We thought that even though the frequency of the target was relatively lower than the frequency of the prime it was not low by itself. We separated the data into two groups (one with high frequencies and one with low frequencies), and the further analysis showed no significant difference between these two groups (p Greater than . OS), indicating that our result was not influenced by the frequency level. What, is more, the robust priming in Llmany-L2few in Exp. 2, in which L2few targets are with higher frequency, has also demonstrated that the frequency difference is not a cause of the priming in our study.

Xueying Luo

South China Normal University

Him Cheung

The Chinese University of Hong Kong

David Bel and Li Li

South Chino Normal University

Lin Chen

Sun Yat-sen University

Lei Mo

South China Normal University

Xueying Luo, College of International Culture, South China Normal University; Him Cheung, Department of Psychology, The Chinese University of Hong Kong; David Bel, Foreign Languages Faculty, South China Normal University; Lei Mo, Department of Psychology, South China Normal University; Li Li, College of International Culture, South China Normal University; Lin Chen, School of Chinese as a Second Language, Sun Yat-sen University.

This research was supported by National Basic Research Program of China (973 Program) (2012CB720704), National Social Science Foundation of China (11CYY023), and National Natural Science Foundation of China (31200785).

Correspondence concerning this article should be addressed to Lei Mo, Department of Psychology, South China Normal University, Guangzhou, China, 510631. E-mail: molei@scnu.edu.cn
COPYRIGHT 2013 The Psychological Record
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2013 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Luo, Xueying; Cheung, Him; Bel, David; Li, Li; Chen, Lin; Mo, Lei
Publication:The Psychological Record
Article Type:Report
Geographic Code:9CHIN
Date:Jan 1, 2013
Words:9183
Previous Article:Cultural influences on number preferences: Christmas and grading systems.
Next Article:Cognitive defusion versus thought distraction in the mitigation of learned helplessness.
Topics:

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters