Reading Pinyin activates sublexcial character orthography for skilled Chinese readers

Abstract

How do skilled Chinese readers, accustomed to characters, process Pinyin, a phonemic transcription of Chinese? Does the orthography of Chinese characters become activated? In four experiments, native speakers first made a meaning judgment on a two-syllable word written in Pinyin. Immediately following, they responded to a character whose orthography sometimes was related to the character corresponding to the Pinyin. In Experiments 1 and 3, participant named the color of the presented characters; there was an interference effect when the presented characters included phonetic radicals that were part of the character corresponding to the Pinyin. In Experiments 2 and 4, participants named the character; naming times were affected if either the semantic or phonetic radical was shared with the character corresponding to the Pinyin. The results indicate that access to lexical representations in Chinese is centered on the orthographic character, even when the input is Pinyin.

All writing systems establish links between their orthographic forms and the phonological and meaning units of their language. However, writing systems vary in how (and at what level) their orthographies produce activation of phonological and semantic information. In alphabetic languages, the orthography to phonology link is established most strongly at the phoneme level. Because Chinese characters do not correspond to phonemes but to meaning-bearing morpheme-syllables, they allow strong connections to meaning, while also producing automatic activation of syllabic level phonology (Perfetti & Tan, 1998; Perfetti & Zhang, 1991; Zhou & Marslen-Wilson, 1999, 2000).

If skilled Chinese readers develop strong orthography to meaning connections in character reading, what happens when they read alphabetic Pinyin? In China, Pinyin is used in beginning reading to establish a connection between printed and spoken language and to support learning the pronunciation of characters and in learning to read Chinese as a foreign language. Pinyin reading by adults provides an interesting test of how reading experience affects orthographic pathways to meaning during skilled reading. One possibility is that, for an adult reader, Pinyin functions mainly as a path to pronunciation and, through the spoken language, provides a link to meaning. However, because the character is the skilled reader’s primary orthography, there is an alternative possibility: The character has become the gateway to meaning, while also activating phonology (Perfetti, Liu, & Tan, 2005; Perfetti & Tan, 1998; Perfetti & Zhang, 1991). Thus, it may function as an orthographic mediator of meaning access, even when the reader encounters Pinyin. On this possibility, with increasing experience with characters and low use of Pinyin, reading Pinyin comes to activate Chinese orthography, i.e. the character that corresponds to the word written in Pinyin. In an extension of the usual formulation of an indirect path to meaning in alphabetic reading—the alphabetic string to pronunciation to meaning—the Chinese skilled reader’s indirect pathway could become alphabetic string to pronunciation to character to meaning.

Recent studies have provided some evidence for the orthographic activation in Pinyin reading (Chen, Perfetti, & Leng, 2017; Chen, Zhong, Leng & Mo, 2014). Using a color judgment task, these studies presented a two-syllable word written in Pinyin (e.g., bēi zi, meaning ‘cup’; characters corresponding to the Pinyin word: 杯子) that required participants to make a semantic judgment (deciding whether the word referred to a man-made object). Following their response, participants were shown a target character (怀) and required to judge the color of that character. The colors and characters were unrelated in meaning, pronunciation and orthography. Responses in the color judgment task were slowed when the colored character (怀) was orthographically related to the character that corresponds to the Pinyin word (杯). This interference effect, which was found for both native Chinese readers (Chen et al., 2014) and highly experienced learners of Chinese (Chen et al., 2017), implies that when Pinyin is read for meaning, activation of the character form draws attention resources away from color processing (MacLeod, 1991).

These studies provide evidence that reading Pinyin activates the character corresponding to the Pinyin. However, normal character recognition includes use of sublexical character components that give clues to pronunciation and meaning (Chen & Weekes, 2004; Ding, Peng, & Taft, 2004; Feldman & Siok, 1999; Lee, Tsai, Chiu, Tzeng, & Hung, 2006; Lee, Tsai, Huang, Hung, & Tzeng, 2006; Su & Weekes, 2007; Tsang & Chen, 2009; Tsang, Wu, Ng, & Chen, 2017; Zhou, Peng, Zheng, Su & Wang, 2013). The role of these sublexical components in the character mediation effect is our focus here.

Eighty-one percent Chinese characters are compound characters consisting of phonetic and semantic radicals (Chen, Allport, & Marshall, 1996), which can be activated early in character reading. Important for our experiments are the facts that the semantic radical does not always provide a valid cue to the meaning of the character and the phonetic radical does not always provide the pronunciation the character. If the orthographic structure of an activated character is involved in reading Pinyin, we should see evidence for this by manipulating components of the target character in the color judgment task. More broadly, varying both the phonetic and semantic radicals allows us examine the activation of sublexical constituents (orthographic, phonological, and semantic) of character during Pinyin reading.

The experiments we report here test the role of orthographic structure and radical function in two tasks that participants perform when they see the target character. One task requires a silent color judgment response and the other requires naming the character. We manipulate the subcomponent (radical) information of the target character in relation to the character corresponding to the Pinyin presented in the Pinyin reading task. In Experiments 1 and 2, the manipulation centers on the phonetic radical, which could be shared between the Pinyin character and the target character, designated as O+ for shared radical. The shared radical could be a valid cue to the character’s pronunciation (P+) or could be an invalid cue to the character’s pronunciation (P-). Thus, the target character and the character corresponding to the Pinyin word varied in their conditions of radical sharing: They (a) share a phonetic radical that is a valid cue to phonology (O+P+), creating identical pronunciations between the implicit Pinyin character and the explicitly presented target character (for example, 珠 and蛛 have same pronunciation “zhū” as their shared phonetic radical “朱” , whose pronunciation is also “zhū”); (b) share a phonetic radical that is an invalid cue to one of the characters, thus sharing a radical without the characters being phonetically related (O+P-, for example, 柱 and 往 share the same phonetic radical “主” (pronunciation of the radical is zhǔ), but have different pronunciation “zhù” and “wǎnɡ”); or (c) share neither a phonetic radical nor phonology, serving as a baseline (O-P-, for example,胡 and鸥 do not share phonetic radical and have different pronunciation “hú” and “ōu”).

Experiment 1 required a judgment of the target character’s color, involving the identity of the character only implicitly; Experiment 2 required the explicit naming of the target character. The color judgment task allows us to make a direct comparison with previous studies using the same task, which found an orthographic activation on character level (Chen et al., 2014; 2017). Including the naming task allows a test of whether orthographic activation during Pinyin reading can be detected by tasks that have different word processing requirements—one (character naming) that requires explicitly orthographic and phonological processing, compared with one (color judgment) that does not require explicit orthographic processing.

If the phonetic radical is activated during Pinyin processing as a strictly orthographic object (i.e. without phonology), the color judgment or character naming latency of target character in O+P+ and O+P- conditions should be comparable, both producing interference relative to the baseline condition. The assumption is that in the color judgment task, the orthographic activation of the character attracts attention from color naming, thus increasing latency (MacLeod, 1991). In the character naming task, competition from similar orthography produces inhibition during character identification (Perfetti, Liu, & Tan, 2005).

If the phonetic radical’s functional value is also activated as part of the character that is activated during Pinyin processing, then the color judgment task and the character naming task would generate different predictions. In the color judgment task, the increased target character activation from sharing same phonetic radical value should produce an increase in color judgment latency (Klein, 1964; Warren, 1974). Color judgment latency in the O+P+ condition should be longer than the O+P- condition, and the O+P- condition should be longer than the O-P- condition. In the naming task, the shared phonetic radical value in the O+P+ condition would reduce the interference from the orthographic similarity in the O+P- condition and produce shorter character naming latency.

In Experiments 3 and 4, we manipulated the semantic radical through three conditions analogous to those for the phonetic radical in Experiments 1 and 2: the target character and the character corresponding to the Pinyin word (a) share a semantic radical that is a valid cue to meaning, creating related meanings between the two characters (O+S+, for example, 沙and 海share the same semantic radical氵[meaning is related to water] and meaning related, meaning of the characters are “sand” and “sea”, respectively); (b) share a semantic radical that is an invalid cue to one of the characters, creating different meanings between the two characters (O+S-, for example, 镜 and 镇 share the same semantic radical钅[meaning is related to metal] but the meaning of the characters are different, “mirror” and “town”, respectively); (c) share neither a radical nor a meaning (O-S-, for example, 椅 and 钢 shared neither a semantic radical nor meaning: chair-steel).

Experiment 1

Method

Participants.

Thirty-two Chinese graduate students (average age 24.6, SD=2.4) from the University of Pittsburgh and Carnegie Mellon University were paid for their participation in the experiment. There were 10 males and 22 females. All participants were Chinese native speakers and familiar with Pinyin, which they had learned in the first grade of elementary school. All had completed their college study in China before graduate study in the US and were bilingual in Chinese and English.

Stimuli and design.

The stimuli included 60 pairs of words (in Pinyin form) and characters. Most Chinese syllables (which are mapped to characters) are homophones; any individual syllable generally corresponds to several different characters. To avoid ambiguity, the Pinyin words in current study had two syllables, thus corresponding to two characters, and did not share pronunciation with any other Chinese words. Thus, each Pinyin word was unique, associated with only one two-character word. For example, the Pinyin “bēi zi” corresponds exclusively to the two-character word “杯子”. All the two-character words had a postfix zi (子), which indicates the word is a noun, as their second character. This manipulation promotes focus on the first character, which was used to manipulate the relations between the implicit character associated with the Pinyin and the character presented for color naming and minimizes interference from activation of the second character.

Table 1 provides examples of the three conditions: In O+P+ condition, the first character corresponding to the first syllable written in Pinyin (珠, pronunciation, zhū) shared a phonetic radical (朱, pronunciation, zhū) and had the same pronunciation as the color-target character (蛛) (pronunciation, also zhū)¹. In the O+P- condition, the first character corresponding to the first syllable of the Pinyin word (柱, pronunciation is zhù) shared the same phonetic radical (主, pronunciation is zhǔ) with the target character (往) (pronunciation is wǎnɡ) but with different character pronunciation. In O-P- condition, the first character of the Pinyin word and the target character shared neither radicals (胡-鸥) nor pronunciation (pronunciation is hú and ōu, repetitively). There were 20 pairs of stimuli in each condition. All stimuli are listed in Appendix A table A1–A3.

Table 1.

Example stimuli of each condition in Experiment 1

Condition	Pinyin Word		Target Character
O+P+	zhū zi (Pinyin) meaning: bead	珠子(corresponding characters of Pinyin)	蛛 (Pinyin: zhū) (meaning: spider)
O+P−	dù zi (Pinyin) meaning: belly	肚子(corresponding characters of Pinyin)	社 (Pinyin: shè) (meaning: agency)
O−P− (baseline condition)	hú zi (Pinyin) meaning: beard	胡子(corresponding characters of Pinyin)	鸥 (Pinyin: ōu) (meaning: gull)

Word frequency and character frequency were based on the frequency dictionary of Cai and Brysbaert (2010). Means and standard deviations of these properties are presented in Appendix C, Table C1. Neither the word frequency nor stroke number of the corresponding word for the Pinyin word were significantly different across conditions, Fs<1. The conditions did not significantly differ in target character frequency, F<1, or stroke number for the target character, F<1.

Procedure.

Participants were tested individually in a quiet experimental lab, seated approximately 50 cm from a computer screen. On each trial, a fixation signal (a black “+”) was first presented at the center of the computer screen for 500ms, followed by the appearance of a Pinyin word. Each Pinyin word was displayed approximately 5cm wide × 2cm high. For the Pinyin word, participants performed a semantic judgment by pressing a designated button on keyboard only when the Pinyin word referred to a person (such as hái zi ‘kid’ or zhí zi ‘nephew’) and making no response to all other Pinyin words. 20 of the 100 words referred to a person, and thus required a response within 1500ms, after which the word disappeared. The remaining 80 no-response trials included the 60 critical trials (20 for each condition) and 20 fillers, in which the target character and the character corresponding to the Pinyin word did not share a radical, a pronunciation or a meaning. The colored target character was then presented, and participants had 2000ms to respond by making a keyboard response whether the color was blue or red. A 1000ms blank followed each trial. Participants were instructed to respond as quickly and accurately as possible, and they received four practice trials.

The target results were the reaction times in color judgment trials on critical trials, all of which followed no-response Pinyin words (i.e., those that did not refer to people).

Results and Discussion

Only trials that produced accurate responses for both semantic judgment and color judgment were included in the analysis of color judgment decision times. Data from three participants were excluded because they did not reach the accuracy criterion of 90% color judgment accuracy. This criterion was informed by the results of Tanenhaus et al. (1980), who reported that colors could be named at a high average accuracy of 98.7%, so accuracy below 90% is likely to reflect a misunderstanding of the task.

The decisions times for the color judgment of target characters are shown in Table 2, along with results from the semantic judgment task for Pinyin words. The key evidence for an orthographic effect is the comparison of the O+P- condition with the O-P- baseline condition. Planned pairwise comparisons showed that color judgment latencies in the O+P- condition were longer than in the O-P- condition, thus demonstrating interference from the shared radical and suggesting that the phonetic radical was activated during Pinyin processing. Color judgment responses of O+P+ condition did not differ from the O-P- condition or the O+P- condition. This intermediate status of O+P+ makes the role of phonology difficult to assess. Thus, there is evidence for radical-produced interference when the pronunciations were not the same across characters, while same pronunciations produced an unclear effect, with means more similar to O+P- than to O-P-. The statistical tests described below are consistent with this description of the pattern of results.

Table 2.

Means and standard deviations (in parentheses) of semantic judgment accuracy, color judgment accuracy and reaction time (ms) in Experiment 1 and semantic judgment accuracy, naming accuracy and reaction time (ms) in Experiment 2.

Conditions	Experiment 1			Experiment 2
	Semantic judgment	Color judgment		Semantic judgment	Naming
	Accuracy	Accuracy	RTs	Accuracy	Accuracy	RTs
O+P+	.98 (.04)	.96 (.05)	711 (150)	.95 (.07)	.97(.06)	671 (93)
O+P−	.97 (.04)	.97 (.04)	724 (152)	.94 (.07)	.98(.04)	772 (126)
O−P−	.99 (.02)	.97 (.04)	686 (119)	.97 (.04)	.98(.05)	666 (116)

We analyzed the data using mixed-effects modeling (Baayen, Davidson, & Bates, 2008) with color judgment RTs as the dependent variable. The models were implemented in the lme4 and lmerTest packages in R (Bates, Maechler, Bolker, & Walker, 2015; Kuznetsova, Brockhoff, & Christensen, 2017). The base model included the variable we manipulated—Condition—as the fixed effect and intercepts for participants and items as random effects. Model comparisons using the likelihood-ratio test determined whether the final model included the random participant slope or random item slope for the manipulated variable. The final model had Condition as a fixed factor, and random intercepts for participant and item and a by-participant slope for Condition as random effects.

To determine whether there was a main effect of Condition, we compared a full model with Condition as a fixed effect against a reduced model without Condition. The two models showed a difference, although not at the standard level of significance, χ²=4.47, p=.11. Planned pairwise comparisons were carried out in the final model to test the condition effects. The planned pairwise comparisons showed that the O+P- condition produced longer RTs than the O-P- condition, estimate=38.92 ms, SE=18.12, t=2.15, p=.036. The O+P+ and O-P- conditions did not differ significantly, estimate=23.66 ms, SE=17.91, t=1.32, p=.19, nor did the O+P+ condition significantly differ from the O+P- condition, estimate=−15.26 ms, SE=16.88, t=−0.9, p=.37. Appendix D presents the details of the model.

Neither Pinyin semantic judgment accuracy nor target color judgment accuracy showed a Condition effect, χ²=1.61, p=0.45; χ²=1.53, p=.46, respectively. The final logistic models for Pinyin semantic judgment accuracy and color judgment accuracy had Condition as a fixed factor, and participant and item intercepts as random effects.

In summary, Experiment 1 found that when the implicit character and the explicitly presented characters shared phonetic radicals, interference emerged, suggesting that the phonetic radical has been activated in Pinyin reading. An effect of the phonological value of the phonetic radical was not detected in the color judgment task based on the non-significant difference between the O+P+ and O+P- conditions. Experiment 2, which employed the explicit naming task, may provide a more sensitive test for this phonological effect.

Experiment 2

Method

Participants.

Twenty-four Chinese undergraduates (10 males, 14 females; average age 20.4, SD=1.18) from South China Normal University were paid for their participation in the experiment. All participants were familiar with Pinyin, which they had learned from first grade of elementary school. All were bilingual in English.

Stimuli, design and procedure.

The stimuli and design were the same as in Experiment 1: the shared properties of a phonetic radical were manipulated. The procedure was adopted from Experiment 1, with the major change that participants were asked to name the target character instead of making a color judgment. A microphone recorded participants’ oral response.

Results

Semantic judgment accuracies of Pinyin words and naming accuracies and latencies of target characters across the three conditions are in Table 2. Only trials that produced accurate responses for both semantic judgment and character naming were included in the analysis of naming times. The naming latency of O+P- condition was longer than O-P- condition. The O+P+ condition was not different from the O-P- baseline and produced shorter naming latencies than the O+P- condition. The results indicate that the presence of a shared phonetic radical produced interference when the character pronunciations were not shared. This further suggests that the phonetic radical and its phonological value were accessed during Pinyin reading. The statistics that support these conclusions are reported below.

Naming times for target characters differed among the three conditions, χ²=19.35, p<.001. Planned pairwise comparisons in the final model showed that character naming latency in the O+P- condition was longer than the baseline condition, estimate=107.7 ms, SE=24.96, t=4.32, p<.001. O+P+ naming latencies did not differ from the baseline, estimate=8.5 ms, SE=24.97, t=0.34, p=.74, and were shorter than in the O+P- condition, estimate=−99.2 ms, SE=25.03, t=−3.96, p<.001. Neither semantic judgment accuracy of the Pinyin word nor character naming accuracy differed across three conditions, χ²=2.62, p=.27; χ²=0.66, p=.72. The final models for naming accuracy, naming RTs and semantic judgment accuracy had Condition as a fixed factor, and participant and item intercepts as random effects.

Consistent with Experiment 1, Experiment 2 confirmed that the phonetic radical and its phonological value were both activated during Pinyin reading. Interference emerged if the characters shared phonetic radical but had different pronunciations. The interference disappeared when the phonetic radical was congruent with the character pronunciation. One thing to note is that the students residing in China (Experiment 2) would have more recent use of characters than the Chinese students who had come to the U.S. (Experiment 1). However, the key finding is that phonetic radical is activated across these two different populations regardless of the character experience.

Experiment 3

Experiment 3 used the color judgment task to test whether the activation of the semantic radical of an implicit character occurs during Pinyin reading. We manipulate the implicit and explicit target characters so that a shared semantic radical can be identical in meaning (O+S+) or unrelated in meaning (O+S-), analogous to the possibilities for a phonetic radical.

Method

Participants.

Thirty-two Chinese graduate students (18 females, 14 males) with an average age of 25.2 (SD=3.1) from the same university population as Experiment 1 participated in the experiment. None had been in the earlier experiment.

Stimuli, design and procedure.

Three conditions presented three types of relations between the first of two characters that correspond to the presented Pinyin word and the color target character: (a) shared semantic radical with related meanings (O+S+); (b) shared semantic radical with different meanings (O+S-); (c) neither shared semantic radical nor shared character meaning (O-S-).

In each condition, there were 20 pairs of two-character Pinyin words and their corresponding (unpresented) characters. Again, the words were presented in Pinyin. In the O+S+ condition, the character (沙) that corresponds to the first syllable of the Pinyin word and the color-target character (海) shared the semantic radical (氵, meaning related to water), and the character meanings were related (sand-sea). In the O+S- condition, the first character of the Pinyin word (镜) and the color-target character (镇) shared the semantic radical (钅, meaning related to metal) but the two characters are not related in meaning; the first character means “mirror” while the second character means “town”. In the O-S- condition (the baseline condition), the first character of the Pinyin word (椅) and the color-target character (钢) shared neither a semantic radical nor meaning (chair-steel). An example from each condition is in Table 3². Corresponding word frequency for Pinyin word, target character frequency, stroke number for word and target character were controlled across conditions: Fs<1. For more details on frequency and stroke number, please see Appendix C Table C2.

Table 3.

Example stimuli of each condition in Experiment 2

Condition	Pinyin Word		Target Character
O+S+	shā zi (Pinyin) meaning: sand	沙子(corresponding characters of Pinyin)	海 (meaning: sea)
O+S−	jìnɡ zi (Pinyin) meaning: mirror	镜子(corresponding characters of Pinyin)	镇 (meaning: town)
O-S- (baseline condition)	yǐ zi (Pinyin) meaning: chair	椅子(corresponding characters of Pinyin)	钢 (meaning: steel)

The procedure was the same as in Experiment 1. Participants were required to make a semantic judgment on a Pinyin word and then to make a color judgment of a character that followed. If the semantic radical is activated as an orthographic object in Pinyin processing, we expect the color judgment latency of the target character in O+S+ and O+S- conditions to be longer than in O-S- condition. If the semantic value of the semantic radical is also activated during Pinyin reading, the color judgment latency of the target character in O+S+ condition would be longer than O+S- condition because of increased character activation from related meanings.

Results and Discussion

Two participants were excluded in the color judgments RTs because their color judgment accuracy was lower than 90%, so the results reported below are based on data from 30 participants.

Results of the semantic judgment of Pinyin words and the color judgment of color-target characters in the three conditions are in Table 4. The mixed-effects model comparison showed a difference, although not at the standard level of significance, χ²=3.49, p=.17. The two shared radical conditions (O+) showed longer mean decision times than the baseline condition, and the 33ms difference between O+S+ and O-S- is associated with SE=17.41, t=1.90, p=.06 in a planned pairwise comparison. The O+S- condition did not differ significantly from the O-S- condition, estimate= 16.87 ms, SE=17.40, t=0.97, p=.34. The O+S+ and O+S- condition did not differ, estimate= 16.17 ms, SE=17.39, t=0.93, p=.36.

Table 4.

Means and standard deviations (in parentheses) of semantic judgment accuracy, color judgment accuracy and reaction time (ms) in Experiment 3 and semantic judgment accuracy, naming accuracy and reaction time (ms) in Experiment 4.

Conditions	Experiment 3			Experiment 4
	Semantic judgment	Color judgment		Semantic judgment	Naming
	Accuracy	Accuracy	RTs	Accuracy	Accuracy	RTs
O+S+	.97 (.04)	.97 (.05)	676 (121)	.97 (.04)	.97(.06)	669 (133)
O+S−	.98 (.04)	.96 (.05)	659 (117)	.94 (.05)	.98(.04)	718 (144)
O−S−	.97 (.04)	.98 (.04)	643 (104)	.93 (.04)	.98(.05)	675 (123)

On the Pinyin semantic judgment task, accuracy did not differ significantly among the three conditions, χ²=0.02, p=.99. For color judgment, accuracy also did not significantly differ across the three conditions, χ²=2.64, p=.27. The final models for color judgment RTs, semantic judgment accuracy and color judgment accuracy all had Condition as a fixed factor, and participant and item intercepts as random effects.

Thus, although the mean color judgment times suggest a radical-based interference effect, especially for a semantic radical that shared its meaning across the implicit and explicitly present character, which is consistent with the prediction that increased word activation produces an increase in color judgment latency, these differences in means fell short of statistical reliability. The weakness of the semantic radical effect—present in means but not in the mixed effects model—may reflect the role of attention, which is explicitly directed to the color judgment task. This possibility leads to the prediction that effects of a semantic radical might emerge in an explicit word processing task. Experiment 4 tested this prediction by employing a naming task on the target character.

Experiment 4

Method

Participants.

Twenty-four Chinese undergraduates from South China Normal University (17 females, 7 males), average age 20 (SD=2.1), were paid for their participation in the experiment. All participants were familiar with Pinyin and had started to learn Pinyin from first grade of elementary school. All of them were bilingual in English. None had been in the earlier experiment.

Stimuli, design and procedure.

The stimuli and design were the same as in Experiment 3. The procedure was the same as in Experiment 2: Participants were required to make a meaning decision on a Pinyin word and then to name a character that followed. The key comparisons involved the effects of radical and meaning overlap between the implicit character corresponding to the Pinyin and the character to be named, O+S+; O+S-; O-S-.

Results and Discussion

The semantic judgment accuracies of Pinyin words and the naming accuracies and latencies of target characters in the three conditions are displayed in Table 4. Only trials that produced accurate responses for both semantic judgment and character naming were included in the analysis of naming times. For target character naming, the mean latency for the O+S- condition was longer than for the baseline O-S-, suggesting that a shared semantic radical interfered with target naming when its meaning was different from the meaning of the character. Latencies for the O+S+ condition did not differ from those for the O-S- baseline and were shorter than those of O+S-. The interference produced by a radical when the meanings were not shared across characters disappeared when the meanings were shared. Thus, unlike the interference effect in the color judgment task, which requires the presented character to be ignored, an explicit requirement to name the character brings about an effect of semantics, suggesting that meaning was activated by the Pinyin meaning judgment task. The following paragraphs report the statistics that support these conclusions.

The three conditions produced differences in target naming latencies, χ²=6.20, p=.045. Planned pairwise comparisons showed that character naming latency in O+S- condition was marginally longer than in the baseline (O-S-) condition, estimate=40.77 ms, SE=21, t=1.94, p=.057. The O+S+ condition produced shorter naming latencies than O+S- condition, estimate=−50.26 ms, SE=20.81, t=−2.42, p=.02, but did not differ from the O-S- baseline, estimate=−9.49 ms, SE=20.91, t=−0.45, p=.65.

Neither semantic judgment accuracy nor character naming accuracy showed significant differences across the three conditions, χ²=0.81, p=.67; χ²=2.20, p=.33. The final models for naming accuracy, naming RTs and semantic judgment accuracy had Condition as a fixed factor, and participant and item intercepts as random effects.

Experiment 4 found that both semantic orthographic units and their meaning values were activated in an explicit character naming task. When the implicit and explicit characters shared semantic radicals, interference emerged if the characters were unrelated in meaning. However, there was no interference when the semantic radical was congruent with the character meaning. The results indicate that the meaning of the semantic radical was activated and contributed to the interference effect in character naming.

General Discussion

Our experiments examined reading Pinyin for meaning to observe the extent to which such reading caused the activation of character orthography, including its sublexical orthographic constituents and their phonological and semantic values. We found evidence that sublexical orthographic, phonological, and semantic information were all activated during reading Pinyin for meaning. The sublexical orthographic (radical) activation was observed in both an explicit character naming task (Experiments 2 and 4) and an implicit color judgment task (Experiment 1).

The interference effect emerged only when the functional values of the radical were incongruent with the pronunciation or meaning of the character; i.e., when the phonetic radical was incongruent with character level pronunciation and when the semantic radical was incongruent with character level meaning. No interference effect was found when the shared radicals were congruent with the meaning or pronunciation of the character. The results indicate that pronunciation and meaning of sublexical orthographic units contribute to the interference effect in an explicit character naming task. In the next section, we consider what this pattern of results implies for reading processes.

Which character-centered constituents are activated in reading Pinyin for meaning?

The results indicate that when Chinese adults read a word written in Pinyin for meaning, this leads to an automatic activation of the corresponding Chinese orthographic unit: the character. As part of this process, the sublexically orthographic, phonological and semantic constituents of character are also activated.

The congruence of character pronunciations and meanings in relation to that of their phonetic and semantic radicals emerged as important when a character was named. The finding that orthographic, phonological, and semantic constituents are activated during Pinyin reading mimics results from primed character naming (Feldman et al, 1997, 1999; Perfetti & Tan, 1998; Zhou & Marslen-Wilson, 1999). For example, orthographic, phonological, and semantic priming effects have been observed to emerge rapidly over a few milliseconds of character processing (Perfetti & Tan, 1998), with orthographic effects emerging first. Thus, although Pinyin reading seems to be quite different from character reading, we find that reading Pinyin produces implicit character reading processes.

The fact that reading a Pinyin spelling of a Chinese word activates its corresponding character and its sublexical radicals has general implications for the role of characters in Chinese reading. With sufficient experience in Chinese reading, the character becomes the basic point of access to the written Chinese lexicon. Its central role in skilled Chinese reading causes it be a mediator to meaning—a gateway to the printed lexicon--even when a reader reads the less common Pinyin. At a more general level--in any language--learners come to acquire interconnected lexical constituents (orthography, phonology and semantics) that allow written words to be read for their meaning and pronunciation. Although reading shares both common procedures and neural pathways across different writing systems (Bolger, Perfetti, & Schneider, 2005; Perfetti, Cao & Booth, 2013), visuo-orthographic information appears to play a more important role in Chinese reading compared with alphabetic writing (Li, Shu, McBride-Chang, Liu, & Peng, 2012; Siok & Fletcher, 2001; Sun, Yang, Desroches, Liu, & Peng, 2011). This study adds the conclusion that reading Pinyin also brings access to the character-based lexicon that serves typical Chinese reading. In addition, Pinyin can strengthen not only the phonological representation but also the association between phonology and character orthography (Chen, Zhou, & Wang, 2016; Guan, et al., 2011; Lin et al., 2010; McBride-Chang, Bialystok, Chong, & Li, 2004; Shu, Peng, & McBride-Chang, 2008; Zhu, Liu, Ding, & Peng, 2009), which would facilitate the orthographic activation in Pinyin reading.

The character effect in Pinyin reading is also observed in spoken Chinese word processing across a range of different tasks, including lexical decision task, semantic relatedness judgment task and picture-word interference paradigm (Qu & Damian, 2017; Zhang, Chen, Weekes, & Yang, 2009; Zhang & Weekes, 2009; Zou et al., 2012), although some studies found that character activation occurs only in spoken language tasks that rely heavily on orthographic information (Bi, Wei, Jassen, & Han, 2009; Cao et al., 2011; Zhang & Damian, 2012). It appears that literate adults have established connections between spoken words and their corresponding characters and that these connections can be activated at least under some conditions. For example, in an auditory lexical decision task, Zou et al. (2012) manipulated the relation between the first syllable of two-syllable primes and targets. The first syllable of the primes and target corresponded to the same character in half the trials, but without sharing either pronunciation or meaning; i.e., the shared character was a homograph. Decision times were faster when the first syllable of the prime and target corresponded to the same character than when they corresponded to different characters. Because the prime words and target words were always unrelated in meaning, this effect was due to the syllable-character correspondence—an implicit character orthography effect. Thus, access to the lexicon for meaning, either through the spoken syllable itself or through the Pinyin that spells it, seems to produce activation of the character. The orthographic effect in Pinyin reading and auditory Chinese word processing is consistent with the bimodal interactive activation model of word recognition (Grainger et al., 2003), if we make an additional assumption that activation processes can operate on syllabic units. The model proposes a bidirectional connection between orthographic codes and phonological codes. This allows orthographic representations to be activated from auditory input and phonological representations to be activated from visual input. Reading Pinyin activates phonological codes, which can activate the orthographic representations of Chinese words.

The semantic radical effect was found only in the explicit naming task and not in the implicit color judgment task. Because both tasks were preceded by the Pinyin meaning task, we assume that the Pinyin-to-character process occurred in both cases. These differing results indicate that radical activation may play a greater role when one has to name a character. The color judgment task is able to detect whole character activation in Pinyin reading (Chen et al., 2014, 2017), but may be less sensitive to sublexical radical activation.

Facilitation or Inhibition?

Priming effects in explicit primed naming and lexical decision can show facilitation, whereas the Pinyin-initiated effects here were inhibitory. One reason for this may be that primed naming effects are time-dependent. For example, studies of Ding et al. (2004), Perfetti & Tan (1998) and Weekes, Chen & Lin (1998) found facilitative priming from similar orthography at 43ms SOA and 50ms, respectively; however, with increasing SOA, the prime form produced competition, and thus inhibition of target naming (Perfetti & Tan 1998). Similarly, Zhou and Marslen-Willson (1999) found that naming target characters that shared radicals with primes became inhibitory with longer SOAs, both 100ms and 200ms compared with 57ms SOA; Feldman and Siok (1999) found inhibition when primes and targets shared semantic radicals at 243ms SOA. Leck, Weekes and Chen (1995) also found an interference effect in a semantic categorization paradigm when the characters are visually similar to the targets. Thus, the emergence of facilitative or inhibitory effects depends on the duration of prime processing that occurs prior to the appearance of a target. In the present naming studies, the SOA between the Pinyin word and the target character was response dependent. Participants had 1500ms to respond prior to the appearance of the character, with response times averaging around 700ms. This is well into the range where inhibition would be expected in a priming task. Thus, there is congruence between primed naming results and the present Pinyin results. In addition, the facilitative priming effect for congruent phonology and meaning in Pinyin reading is also consistent with character reading (Perfetti, Liu, & Tan, 2005), where the facilitated phonological and semantic priming effect persists with long SOA.

Boundary Conditions

An important point to note is the boundary conditions on the Pinyin character priming effect. In one of their experiments, Chen et al. (2014) found no evidence of character activation (i.e., no interference effect) when participants read Pinyin silently (with no meaning judgment) at an SOA of 300ms. The key boundary condition is meaning judgments require the reader to access a lexical entry that connects the word form to meaning, something they could do well when required, as indicated by their high accuracy rates on Pinyin semantic judgments. Reading for meaning, something that is only implicit—and thus variable—without a specific meaning task. Without explicit target instructions (e.g., does the word refer to a person?), reading Pinyin silently is an unreliable path to meaning, because of the large number of homophones (one syllable generally maps to an average 11 characters without regard to tone). Readers come to rely on the character-to-meaning connection to select or verify a meaning in combination with the phonological activation that can occur in character reading. Thus, in the Pinyin semantic judgment task, the activation of the character by a meaning category (e.g., kind of person) secures a character-meaning connection.

The SOAs between experiments are also different, 300ms for Pinyin reading task and 1500ms for meaning judgment. Although it is possible that an SOA longer than 300ms is needed for the effect, 300ms should be sufficient, based on a result of Fu, Chen, Smith, Iverson & Mathews (2002), who found that reading two-syllable Pinyin words took 3 times longer than reading two-character words. Since single character priming can occur at around 50ms (Chen & Shu, 2001; Perfetti & Tan, 1998), extrapolation suggests 150ms for a Pinyin syllable and 300ms for a two-syllable Pinyin word.

Conclusion

We find evidence that reading a word written in Pinyin, an alphabetic spelling of Chinese words, leads to the activation of the corresponding character. This activation involves the character as an orthographic unit, its component semantic and phonetic radicals, and their meaning and pronunciation. That these effects are initiated by the reading of an alphabetically written word reflects the centrality of the character in the Chinese orthographic lexicon.

Appendix A. Stimuli of Experiment 1 and 2

Table A1.

Stimuli of O+P+ condition

Pinyin	corresponding characters	meaning	phonetic radical	Pronunciation of radical	target	Pinyin of target	meaning of target
zhū zi	珠子	bead	朱	zhū	蛛	zhū	spider
dié zi	碟子	saucer	枼	yè	谍	dié	spy
nǎo zi	脑子	brain	㐫	n/a	恼	nǎo	angry
páo zi	袍子	robe	包	bāo	炮	pào	cannon
dīnɡ zi	钉子	nail	丁	dīnɡ	盯	dīnɡ	stare
ɡōu zi	钩子	hook	勾	ɡōu	沟	ɡōu	ditch
wén zi	蚊子	mosquito	文	wén	纹	wén	stripe
ɡǎo zi	稿子	manuscript	高	ɡǎo	搞	ɡǎo	do
hóu zi	猴子	monkey	侯	hóu	喉	hóu	throat
yǐ zi	椅子	chair	奇	qí	倚	yǐ	lean
yā zi	鸭子	duck	甲	jiǎ	押	yā	detain
lán zi	篮子	basket	监	jiān	蓝	lán	blue
dǎn zi	胆子	courage	旦	dàn	担	dān	undertake
jìnɡ zi	镜子	mirror	竟	jìnɡ	境	jìnɡ	condition
jiǎn zi	剪子	scissor	前	qián	煎	jiān	fry
zhǒnɡ zi	种子	seed	中	zhōng	肿	zhǒnɡ	swell
chánɡ zi	肠子	gut	怀	n/a	畅	chànɡ	fluent
lún zi	轮子	wheel	仑	lún	伦	lún	ethic
jiǎo zi	饺子	dumpling	交	jiāo	较	jiào	compare
qún zi	裙子	skirt	君	jūn	群	qún	group

Table A2.

Stimuli of O+P- condition

Pinyin	corresponding characters	meaning	phonetic radical	Pronunciation of radical	target	Pinyin of target	meaning of target
zhuī zi	锥子	awl	隹	zhuī	淮	huái	Huai River
zhù zi	柱子	pillar	主	zhǔ	往	wǎnɡ	toward
shénɡ zi	绳子	rope	黾	mǐn	蝇	yínɡ	fly
bǎ zi	靶子	target	巴	bǎ	肥	féi	fat
pāi zi	拍子	racket	白	bái	伯	bó	uncle
bēi zi	杯子	cup	不	bù	怀	huái	bosom
bǎn zi	板子	board	反	fǎn	皈	ɡuī	convert
chuí zi	锤子	hammer	垂	chuí	唾	tuò	spit
cūn zi	村子	village	寸	cùn	肘	zhǒu	arm
ɡé zi	格子	plaid	各	ɡé	洛	luò	Luo River
shū zi	梳子	comb	㐬	n/a	流	liú	flow
jià zi	架子	shelf	加	jiā	贺	hè	congratulate
méi zi	梅子	plum	每	měi	侮	wǔ	humiliate
niè zi	镊子	tweezers	聂	niè	摄	shè	absorb
shā zi	沙子	sand	少	shǎo	钞	chāo	money
shì zi	柿子	persimmon	市	shì	沛	pèi	abundant
dù zi	肚子	stomach	土	tǔ	社	shè	agency
xiù zi	袖子	sleeve	由	yóu	抽	chōu	draw
xié zi	鞋子	shoes	圭	guī	娃	wá	child
wà zi	袜子	socks	末	mò	沫	mò	foam

Table A3.

Stimuli of O-P- condition

Pinyin	corresponding characters		meaning	phonetic radical	Pronunciation of radical	target	Pinyin of target	meaning of target
kòu zi		扣子	button	口	kǒu	际	jì	boarder
zhuì zi		坠子	pendant	队	duì	贷	dài	loan
yín zi		银子	silver	艮	gèn	扭	niǔ	twist
hú zi		胡子	beard	古	gǔ	鸥	ōu	gull
yǐnɡ zi		影子	shadow	景	jǐng	款	kuǎn	fund
xiānɡ zi		箱子	suitcase	相	xiāng	弯	wān	bent
táo zi		桃子	peach	兆	zhào	纯	chún	pure
pén zi		盆子	basin	分	fēn	奢	shē	luxurious
shī zi		狮子	lion	师	shī	惋	wǎn	regret
sǎnɡ zi		嗓子	throat	桑	sāng	材	cái	material
liàn zi		链子	chain	连	lián	糊	hú	paste
mào zi		帽子	hat	冒	mào	炸	zhà	blast
pínɡ zi		瓶子	bottle	并	bìng	顽	wán	naughty
chǎn zi		铲子	spade	产	chǎn	税	shuì	tax
tī zi		梯子	ladder	弟	dì	验	yàn	test
lú zi		炉子	stove	户	hù	份	fèn	part
kù zi		裤子	trousers	库	kù	殖	zhí	breed
ɡùn zi		棍子	stick	昆	kūn	味	wèi	smell
lǐnɡ zi		领子	collar	令	lìng	欣	xīn	joyful
biān zi		鞭子	whip	便	biàn	微	wēi	tiny

Appendix B. Stimuli of Experiment 3 and 4

Table B1.

Stimuli of O+S+ condition

Pinyin	corresponding characters	meaning of Pinyin	semantic radical	meaning of semantic radical	target	Pinyin of target	meaning of target
zhù zi	柱子	pillar	木	tree	根	ɡēn	root
táo zi	桃子	peach	木	tree	核	hé	core
chénɡ zi	橙子	orange	木	tree	桔	jié	tangerine
lǐ zi	李子	prune	木	tree	杏	xìnɡ	apricot
ɡuì zi	柜子	cabinet	木	tree	橱	chú	closet
lín zi	林子	forest	木	tree	树	shù	tree
nǎo zi	脑子	brain	月	organ	胸	xiōnɡ	breast
chánɡ zi	肠子	gut	月	organ	肝	fù	liver
chuí zi	锤子	hammer	钅	metal	铁	tiě	iron
chǎn zi	铲子	spade	钅	metal	锹	qiāo	shovel
yín zi	银子	silver	钅	metal	铜	tónɡ	copper
dīnɡ zi	钉子	nail	钅	metal	针	zhēn	needle
kù zi	裤子	trousers	衤	clothes	褂	ɡuà	upper garment
xiù zi	袖子	sleeve	衤	clothes	衫	shān	shirt
dàn zi	担子	burden	扌	hand	挑	tiāo	carry
shā zi	沙子	sand	氵	water	海	hǎi	sea
hé zi	盒子	box	皿	utensil	盆	pén	basin
wén zi	蚊子	mosquito	虫	insect	蜂	fēnɡ	bee
lán zi	篮子	basket	⺮	bamboo	筐	kuānɡ	crate
dí zi	笛子	flute	⺮	bamboo	箫	xiāo	vertical flute

Table B2.

Stimuli of O+S- condition

Pinyin	corresponding characters	meaning	semantic radical	meaning of semantic radical	target	Pinyin of target	meaning of target
yòu zi	柚子	grapefruit	木	tree	权	quán	power
shì zi	柿子	persimmon	木	tree	村	cūn	village
méi zi	梅子	plum	木	tree	标	biāo	sign
bǎn zi	板子	board	木	tree	杭	hánɡ	Hang
bànɡ zi	棒子	bar	木	tree	枕	zhěn	pillow
ɡùn zi	棍子	stick	木	tree	极	jí	pole
bó zi	脖子	neck	月	organ	腊	là	cured meat
dù zi	肚子	stomach	月	organ	胶	jiāo	glue
jìnɡ zi	镜子	mirror	钅	metal	镇	zhèn	town
zhuī zi	锥子	awl	钅	metal	钧	jūn	an ancient unit of weight
niè zi	镊子	tweezers	钅	metal	钞	chāo	money
zhuó zi	镯子	bracelet	钅	metal	锦	jǐn	brocade
qún zi	裙子	skirt	衤	clothes	褐	hè	brown
páo zi	袍子	robe	衤	clothes	初	chū	beginning
pāi zi	拍子	racket	扌	hand	挺	tǐnɡ	endure
chí zi	池子	sink	氵	water	汝	rǔ	you
ɡài zi	盖子	lid	皿	utensil	盐	yán	salt
xiē zi	蝎子	scorpion	虫	insect	虹	hónɡ	rainbow
xiānɡ zi	箱子	suitcase	⺮	bamboo	策	cè	strategy
shāi zi	筛子	sifter	⺮	bamboo	篇	piān	a piece of writing

Table B3.

Stimuli of O-S- condition

Pinyin	corresponding characters	meaning	semantic radical	meaning of semantic radical	target	Pinyin of target	meaning of target
yǐ zi	椅子	chair	木	tree	钢	ɡānɡ	steel
tī zi	梯子	ladder	木	tree	键	jiàn	key
zhuō zi	桌子	desk	木	tree	盔	kuī	helmet
yē zi	椰子	coconut	木	tree	袄	ǎo	coat
shū zi	梳子	comb	木	tree	换	huàn	change
jià zi	架子	shelf	木	tree	符	fú	symbol
dǎn zi	胆子	courage	月	organ	松	sōnɡ	pine
bǎnɡ zi	膀子	arm	月	organ	锁	suǒ	lock
qián zi	钳子	pliers	钅	metal	桶	tǒnɡ	bucket
liàn zi	链子	chain	钅	metal	材	cái	material
ɡōu zi	钩子	hook	钅	metal	虾	xiā	shrimp
ɡuō zi	锅子	wok	钅	metal	脚	jiǎo	foot
wà zi	袜子	socks	衤	clothes	汤	tānɡ	soup
bèi zi	被子	quilt	衤	clothes	桩	zhuānɡ	stake
kòu zi	扣子	button	扌	hand	肤	fū	skin
zhā zi	渣子	dreg	氵	water	锯	jù	saw
pán zi	盘子	plate	皿	utensil	筝	zhēnɡ	kite
é zi	蛾子	moth	虫	insect	袱	fú	wrapper
kuài zi	筷子	chopsticks	⺮	bamboo	梨	lí	pear
lǒu zi	篓子	crate	⺮	bamboo	柴	chái	firewood

Appendix C

Table C1.

Characteristics of stimuli (standard deviations in parentheses) in Experiment 1 and 2

Condition	Corresponding word for Pinyin		Target character
	frequency	stroke	frequency	stroke
O+P+	12(18)	14(3)	81(107)	10(3)
O+P−	10(12)	13(3)	78(87)	9(2)
O−P−	13(17)	14(3)	79(90)	10(2)

Table C2.

Characteristics of stimuli (standard deviations in parentheses) in Experiment 3 and 4

Condition	Corresponding word for Pinyin		Target character
	frequency	stroke	frequency	stroke
O+S+	12 (19)	13 (3)	69 (112)	10(3)
O+S−	10 (15)	15 (3)	65(81)	10(3)
O−S−	8 (13)	14 (2)	56(65)	10(2)

Appendix D

Table D1.

Fixed effect estimates and pair-wise comparisons for color judgment reaction times in Experiment 1.

	Estimate	SE	t	p
Intercept (O-P- baseline)	687.26	23.42	29.35	<.001
O+P+ vs. O-P-	23.66	17.91	1.32	0.19
O+P− vs. O−P−	38.92	18.12	2.15	<.05
O+P+ vs. O+P−	−15.26	16.88	−0.90	0.37

Note. SE = standard error.

Table D2.

Summary of random item and participant effects for color judgment reaction times in Experiment 1.

		Correlations
Random effect	SD	1	2
Item
1. Intercept	39.96	—
Participant
1. Intercept (O-P- baseline)	108.74	—
2. O+P+ vs. O-P-	33.18	> .99	—
3. O+P- vs. O-P-	36.37	> .99	> .99
Residual error	184.08

Note. SD = standard deviation.

Table D3.

Fixed effect estimates and pair-wise comparisons for naming times in Experiment 2.

	Estimate	SE	t	p
Intercept (O-P- baseline)	665.48	27.20	24.46	<.001
O+P+ vs. O-P-	8.50	24.97	0.34	0.74
O+P− vs. O−P−	107.70	24.96	4.32	<.001
O+P+ vs. O+P−	−99.20	25.03	−3.96	<.001

Table D4.

Summary of random item and participant effects for naming times in Experiment 2.

Random effect	SD
Item
Intercept	67.71
Participant
Intercept	101.59
Residual error	191.52

Table D5.

Fixed effect estimates and pair-wise comparisons for color judgment reaction times in Experiment 3.

	Estimate	SE	t	p
Intercept (O-S- baseline)	643.65	22.60	28.48	<.001
O+S+ vs. O−S−	33.03	17.41	1.90	0.06
O+S− vs. O−S−	16.87	17.40	0.97	0.34
O+S+ vs. O+S−	16.17	17.39	0.93	0.36

Table D6.

Summary of random item and participant effects for color judgment reaction times in Experiment 3.

Random effect	SD
Item
Intercept	41.37
Participant
Intercept	103.79
Residual error	192.40

Table D7.

Fixed effect estimates and pair-wise comparisons for naming times in Experiment 4.

	Estimate	SE	t	p
Intercept (O-S- baseline)	678.28	29.90	22.69	<.001
O+S+ vs. O−S−	−9.49	20.91	−0.45	0.65
O+S− vs. O−S−	40.77	21.00	1.94	0.057
O+S+ vs. O+S−	−50.26	20.81	−2.42	<.05

Table D8.

Summary of random item and participant effects for naming times in Experiment 4.

Random effect	SD
Item
Intercept	55.0
Participant
Intercept	126.9
Residual error	168.8