Semantic radicals’ semantic attachment to their composed phonograms

Abstract

In Chinese character processing studies, it is widely accepted that semantic radicals, whether character or non-character ones, can undergo semantic activation. However, there is a notable absence of studies dedicated to understanding the nature and operation of the semantic radicals’ semantic information. To address this gap, the present study employed a masked semantic priming paradigm combined with a part-of-speech decision task and a lexical decision task across three experiments. Experiment 1 was designed to examine the semantic autonomy and the semantic attachment of semantic radicals in transparent phonograms. Experiment 2 sought to further investigate the degree of semantic autonomy of semantic radicals in opaque phonograms. Experiment 3 was crafted to further probe into the presence of semantic attachment of semantic radicals in pseudo-characters. Results showed significant priming effects in both transparent and opaque phonogram conditions, with faster reaction times and higher accuracy for semantically related prime-target pairs. However, no such priming effect was observed in the pseudo-character condition, indicating that semantic radicals are not activated in non-lexical contexts. These findings suggest that semantic radicals were semantically activated when embedded in both transparent and opaque phonograms, but not when planted in pseudo-characters. The plausible account put forward is that semantic radicals stand on pars with their composed phonograms in possessing their own semantic information, but the former is semantically strongly attached to the latter, such that it cannot live without the latter’s semantic company.

Introduction

Chinese phonograms and semantic radicals

Chinese characters fall chiefly into two categories, simple characters and compound characters. A majority of compound characters are pictophonetic characters or phonograms that consist of semantic radicals and phonetic radicals [17, 29]. Semantic radicals usually occupy the left-most position of the composed phonograms and provide semantic cues to the entire characters, while phonetic radicals stand in the right-most position and indicate the pronunciation of the whole characters [30]. For example, the phonogram “烤” (/kao3/, to bake) contains the semantic radical “火” (/huo3/, fire) locating at the left side, which denotes the instrumental characteristics of the whole character, and the right-side phonetic radical “考” (/kao3/, to test) gives hints to the whole character’s pronunciation “kao3”. Depending on the degree of semantic correlation between phonograms and their embedded semantic radicals, namely, transparency, phonograms are divided into transparent and opaque ones [28]. Transparent phonograms are those which bear semantic connections with their embedded semantic radicals, while opaque phonograms are ones that have no semantic relationship with their embedded semantic radicals. Similarly, semantic radicals, based on their characterhood, are differentiated between character semantic radicals and non-character ones.

Semantic radical’s characterhood refers to the degree to which a semantic radical contributes meaning to a character and can stand alone as an independent character. For example, the radical “木” (meaning wood) has a high degree of characterhood because it can stand alone to mean “tree” and is also a component in many characters related to wood, such as “桶” (barrel), and “林” (forest). We will measure the semantic radical’s characterhood by analyzing the frequency of the radical’s occurrence as an independent character in a corpus of modern Chinese literature [14, 25].

The concept of phonogram transparency pertains to how clearly the phonetic element of a character corresponds to its actual pronunciation. To evaluate this transparency, we will examine the alignment between the phonogram's pronunciation and that of the character it is part of. The transparency is categorized on a spectrum from low to high, depending on the degree of pronunciation consistency among related characters. For instance, the phonogram “青” exhibits high transparency in characters like “请” (qǐng) and “清” (qīng), where the phonogram’s pronunciation closely matches that of the characters. In contrast, it demonstrates no transparency in “精” (jīng), where the pronunciation is notably different.

Current research on the semantic activation of semantic radicals

Currently, a predominant number of studies concentrate on the semantic activation of semantic radicals in Chinese phonogram processing, taking account of the possible variables like the semantic radical’s characterhood and the embedding phonogram’s transparency that may modulate the activation process.

For example, Flores d’Arcais, Saito, & Kawakami [9], Zhou et al. [29], Wang et al. [21], and Zou, Tsang, & Wu [31] examined the semantic activation of character semantic radicals in transparent, opaque, or both types of phonograms. Flores d’Arcais et al. [9] employed a semantic relatedness judgment task, where participants judged the semantic relatedness between a prime radical and its phonogram target. Although the transparency of the phonograms was not differentiated, results indicated that the semantic information of character semantic radicals could be activated, which could contribute to the recognition of composed phonograms. Zhou et al. [29] used a priming lexical decision task with a short stimulus onset asynchrony (SOA) of 57 ms, focusing on semantically opaque phonograms. They found that the meaning of character semantic radicals could be accessed within 100 ms in the processing of semantically opaque characters. Likewise, Wang et al. [21], who manipulated the frequency of character semantic radicals, a employed a lexical decision task to unveil that both semantically opaque and transparent characters, which contained high-frequency semantic radicals, elicited a smaller N400 than those containing low-frequency ones. This result indicated the semantic activation of the embedded semantic radicals, suggesting an easier semantic integration (for transparent phonograms) or suppression (for opaque phonograms) between the high-frequency semantic radicals and the whole characters. In addition, Zou, Tsang, and Wu [31] employed a masked priming lexical decision task and recorded ERPs to,investigate the temporal dynamics of character semantic radical processing. Results showed that character semantic radicals evoked an N400 and a late positive complex, suggestive of signifying semantic activation and extensive semantic processing. In another study by Tong et al. [18], a priming lexical decision task was used to explore the priming effects of character semantic radicals (e.g., the character soil semantic radical “土”,/tu3/). They classified the targets into five levels according to the semantic relatedness between the priming semantic radicals and the embedding target phonograms. The findings revealed a gradual faciliatory priming effect in phonogram recognition, underscoring the semantic activation of character semantic radicals and the graded priming effect of radicals in phonogram processing.

In the realm of phonogram processing studies, another number of studies were committed to investigating the semantic activation of non-character semantic radicals, which did not clearly distinguish between transparent and opaque phonograms. Wu et al. [24] examined the semantic activation of the non-character hand semantic radical “扌” with MRI measurement in a passive reading task. Results showed greater activity in the right medial frontal gyrus for characters containing hand radicals, as compared to those containing no hand radicals, suggesting an independent semantic activation of the non-character hand radicals that usually convey the “hand-related” semantic features. Employing a synonym judgment task, Hung et al. [12] used a synonym judgment task, presenting character pairs sharing or not sharing the same non-character semantic radical. Results showed that compared with the different semantic radical condition, synonyms sharing identical semantic radicals elicited reduced M450, while non-synonyms sharing semantic radicals elicited greater M450 especially in the middle frontal region, which suggested the sub-lexical semantic activation of non-character semantic radicals. Similarly, Wang et al. [19]employed a semantic judgment task in which participants judged the direction of body actions encoded in phonograms. The verbs contained non-character semantic radicals such as “扌” (indicating the “upward” direction) and “” (indicating the “downward” direction), which were also presented separately as primes. Results demonstrated that when the direction suggested by the character and the primed semantic radical were congruent (e.g., “扌” – “提”, to lift), response times were shorter as compared to incongruent conditions (e.g., “扌” – “捶”, to thump). This suggested that these non-character semantic radicals were semantically activated. Also, Zhang et al. [27] investigated how the N170 neural response adapted to sub-lexical semantic and phonological processing by manipulating both the semantic radicals and the semantic relatedness of four sequentially presented characters. Nearly all the semantic radicals were non-character ones. The results indicated that N170 exhibited heightened sensitivity to semantic processing, particularly in the right hemisphere during Chinese character reading, implying the semantic activation of non-character semantic radicals.

In addition, a body of research has also addressed the semantic activation of radicals without clearly differentiating their characterhood status, as evidenced in a range of behavioral [5, 8, 23], eye-tracking [20], and neuro-electrophysiology studies [26]. For instance, Feldman and Siok [8] used a priming lexical decision task and demonstrated semantic facilitation from radicals regardless of their character status. They suggested that radical meaning could influence lexical access even in the absence of overt transparency. Zhang et al. [26] used ERP recordings to examine character recognition and addressed radical-level processing.

Overall, the studies reviewed above collectively suggest that semantic radicals—whether character or non-character, and regardless of phonogram transparency—can undergo semantic activation during word recognition. However, previous research often lacked direct comparisons across radical types and transparency conditions, or failed to separate characterhood effects from semantic priming. Our study aims to fill this gap by systematically manipulating both the characterhood of semantic radicals and the transparency of embedding phonograms, while adopting a refined priming paradigm and well-controlled stimuli.

To sum up, researchers place great interest in the semantic activation of semantic radicals. The focus is on the presence of semantic activation of semantic radicals as well as on the modulating variables like transparency of phonograms and characterhood of semantic radicals. It is generally accepted that semantic radicals can be semantically activated across different conditions; however, a closer inspection reveals that the effects differ depending on the transparency of the host phonograms. While the majority of studies report facilitative effects in transparent phonograms, the findings for opaque phonograms are more mixed. For instance, Wang et al. [21] observed reduced N400 for opaque phonograms with high-frequency radicals, suggesting possible suppression costs for incongruent semantic information. This aligns with the view that semantic mismatch between radicals and phonograms may impose integration difficulty.. A seeming observation is that the semantic activation of semantic radicals plays a facilitatory role in the processing of composed transparent phonograms, but an uncertain (either facilitatory or inhibitory) role in the processing of composed opaque phonograms.

Despite these empirical efforts, the theoretical development concerning how semantic radicals contribute to phonogram comprehension remains relatively under-specified. Most prior work adopts the perspective of sub-lexical semantic priming but lacks a unified framework to account for conditions under which semantic radical activation facilitates or hinders processing. Our study builds on this empirical base and aims to clarify this theoretical issue by systematically manipulating both characterhood and transparency in a controlled priming design. In doing so, we attempt to move beyond descriptive findings and contribute to a more integrated understanding of sub-lexical semantic access.

The present study

Given the above description of Chinese semantic radical research, to our best knowledge, little known study is devoted to inquiring into the nature of the semantic information of semantic radicals, including the mechanism that governs the semantic activation of semantic radicals as well as the facilitatory or inhibitory effect resulting thereof. The specific questions to be raised are: Since both character and non-character semantic radicals can be semantically activated, where does their semantic information respectively come from? How does the semantic information of semantic radicals reside in the mind? In what way is it related to the composed phonogram’s semantic information? What are the hidden happenings that bring about the semantic radicals’ facilitatory or inhibitory effects on the processing of composed transparent and opaque phonograms? To address these questions, three experiments were conducted in the present study. Experiment 1 was designed to confirm semantic radicals’ semantic activation and thereby to testify to the semantic autonomy and semantic attachment of semantic radicals in the composed transparent phonograms. Experiment 2 was designed to confirm whether semantic radicals are activated and whether thet play a facilitatory or inhibitory role in the processing of the composed opaque phonograms, thereby testing their degree of semantic autonomy. Experiment 3, by planting semantic radicals in pseudo-characters, was intended to examine whether semantic radicals could still be semantically activated, in an attempt to unveil the presence of semantic attachment of semantic radicals in the composed phonograms, thereby elaborating the semantic attachment of semantic radicals to their composed phonograms as purported by Experiments 1 and 2.

To enhance the coherence of the study design, it is important to explicate the logical progression among the three experiments. Experiment 1 serves as the foundational step by investigating whether semantic radicals embedded in transparent phonograms can be independently and meaningfully activated during character processing—this speaks to the radical’s semantic autonomy and attachment in a context where the radical and the whole character share related meanings. Building on this, Experiment 2 examines opaque phonograms, where the meaning of the radical does not align with that of the host character, in order to explore whether the semantic activation of radicals still occurs and whether it facilitates or interferes with character processing. This comparison allows us to evaluate how semantic transparency modulates radical activation. Finally, Experiment 3 extends the inquiry by embedding semantic radicals into pseudo-characters that lack any real lexical semantics, testing whether radicals can be activated in the absence of a legitimate host character. This experimental sequence thus progresses from natural to opaque to artificial contexts, gradually isolating and probing the mechanism and conditions under which semantic radicals exert their influence on character processing.

Two central concepts underlying the present study are semantic autonomy and semantic attachment. Semantic autonomy refers to the radical’s capacity to activate its own meaning independently from the compound character it is embedded in. Semantic attachment, on the other hand, refers to the radical’s semantic activation being contingent upon or constrained by the semantic representation of the whole character. These concepts build upon the idea of multi-level semantic access in Chinese morphological processing [7, 16] and serve to explain the interaction between radical- and character-level processing.

Experiment 1

Method

Participants

A priori sample size calculations were conducted using G*Power 3.1 [6]. The analysis indicated that 28 participants would be required to detect a medium effect size (f = 0.25) (Cohen 3) with 80% power for a two-way within-subjects ANOVA (α = 0.05, number of groups = 1, number of measurements = 2*3 = 6, non-sphericity correction = 1). These calculations adhered to the guidelines for repeated measures wthin-subjects ANOVA (a priori) provided in the G*Power 3.1 Manual [10] and Brysbaert [1]. To account for possible exclusions, a total of forty participants who were native Mandarin speakers and resided in mainland China took part in this experiment (mean age = 25 years, SD = 2.9). They were all students from Sichuan International Studies University (SISU). All of them were right-handed, with the normal or corrected-to-normal vision. Before participation, participants were asked to provide written informed consents and got paid after the experiments.

Materials and design

The experiment adopted a 2 (prime picture: foot picture vs. control picture) × 3 (Target character: character foot semantic radical-embedded vs. non-character foot semantic radical-embedded vs. foot meaning-loaded only character) design. The prime pictures included a human foot picture, and a control picture which was made by distorting the human foot picture through the utilization of the software GIMP, such that it was no longer discernible.

As for the target characters, the first type was those that contained character foot semantic radicals and took on meaning related to human foot effector (e.g., the target phonogram “跳”/tiao4/, meaning “to jump”, contains the character foot semantic radical “”/zu2zi4pang2/and denotes the meaning of bodily motion executed by human foot effector). The second type contained non-character foot semantic radicals and took on meaning related to human foot effector (e.g., the target phonogram “追” (/zhui1/, meaning “to chase”, contains the non-character foot semantic radical “辶”/zou3zhi1pang2/and denotes the meaning of human foot effector-executed motion). Although (/zu2zi4pang2/) is technically a radical form and cannot function as a stand-alone character, it shares the same form and meaning as the independent character 足 (/zu2/), which denotes “foot”. In this study, we adopt a functional perspective that focuses on semantic contribution rather than character status. Therefore, is considered semantically equivalent to 足 in our classification, and is treated as a character semantic radical for experimental purposes. The third type was foot meaning-loaded only characters that bore close semantic relationship with human foot effector but contained neither character nor non-character foot semantic radicals (e.g., the target phonogram “绊”/ban4/, meaning “to stumble”, denotes the human foot effector-executed motion meaning but contains no foot semantic radical). The difference between the second type and the third type lies in whether the semantic radical itself conveys the foot-related meaning. For instance, the semantic radical ‘辶’ in ‘追’ (/zhui1/) conveys foot-related meanings such as walking, movement, or paths. In contrast, the semantic character ‘纟’ in ‘绊’ (/ban4/) does not carry foot-related meanings but instead is associated with thread, textiles, or twisting/tangling. Each type consisted of 10 Chinese phonograms (see Table 1 for sample materials). The three types of targets were matched in character structure, number of strokes (F(2,18) = 1.678, p = 0.215), part of speech (both verbs), and mean frequency (F(2,18) = 0.017, p = 0.983).

[IMAGE OMITTED: SEE PDF]

All target characters were carefully screened using authoritative references including the Xinhua Dictionary (11 th edition) and Handian online dictionary (https://www.zdic.net/hans/) to confirm their semantic association with “foot”. Then, 25 native Chinese speakers who did not participate in the main experiment rated the semantic relatedness of these characters to “foot” using a 7-point Likert scale (1 = not related at all, 7 = extremely related). The results demonstrated strong semantic associations (mean rating = 6.004, 6.152, and 5.932, for three types of target characters, respectively), confirming that all selected characters were appropriately related to the concept “foot”.

To address part-of-speech ambiguity in Mandarin, all target characters were selected based on frequency and consistency of their verb usage. A pretest was conducted with 20 native speakers who rated the likelihood of each character being a verb on a 5-point scale. Only items with an average score above 4 were retained as verb targets.

To minimize participants’ expectations and reduce strategic processing, 10 additional prime-target pairs involving noun targets were included as filler trials. These trials, originally referred to as “controls”, were not of central theoretical interest but served to balance the part-of-speech decision task and diversify the trial structure. In total, 80 prime-target pairs—including both critical and filler items—were presented across trials.

Procedure

Participants were seated approximately 50 cm from a computer in a comfortable chair. Each trial began with a fixation signal for 300 ms, followed by a blank screen for 300 ms. The prime picture was then presented for 500 ms, followed by another 300 ms blank screen. The target (a phonogram, either a verb or a noun) was presented until the participants had made a response. The next trial began after a random interval of 600 to 800 ms. Participants were instructed to determine whether the target character was a verb or not as quickly as possible by pressing the “F” or “J” keys. To reduce ambiguity and ensure clarity during the part-of-speech judgment task, labels indicating “NOUN” and “VERB” were affixed directly to the “F” and “J” keys on participants’ keyboards. The key-to-category assignments were counterbalanced across participants: half of the participants used the “F” key to indicate “noun” and the “J” key to indicate “verb”, while the other half had the opposite assignment. This mapping was explained in advance and practiced before the formal task began. Accuracy and reaction times were recorded. This experiment was conducted in the Key Laboratory of Foreign Language Learning and Cognitive Neuroscience in SISU.

The part of speech judgment task required participants to access the semantic content of the target character to determine its grammatical category. As such, any facilitation from the radical-level semantic activation—primed via picture stimuli—could be observed through faster or more accurate part-of-speech judgments.

Results

Data of 35 participants were included in the statistical analysis (5 participants were excluded due to low accuracy). The mean accuracy for the primed part-of-speech judgment task was 94.82%. The errors and reaction times data exceeding 2.5 SD (1.77%) from the mean RT of each condition were excluded. Table 2 and Fig. 1 exhibit the mean reaction times in Experiment 1.

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

A two-way repeated-measures ANOVA on RTs showed significant main effect of prime picture (F(1, 34) = 9.182, p = 0.005, MSE = 239,188.41, η2 = 0.213) and target character (F(2, 68) = 19.069, p < 0.001, MSE = 76,550.06, η2 = 0.359). The interaction between prime picture and target character was also significant, F(1.691, 57.479) = 4.133, p = 0.027, MSE = 10,902.057, η2 = 0.108. All effect sizes for ANOVA results are reported as partial η2, which is commonly used in repeated-measures designs to estimate the proportion of variance accounted for by a given factor, excluding variance explained by other factors. Greenhouse–Geisser correction was applied where necessary. Post-hoc pairwise comparisons used Bonferroni correction.

The simple effect analysis showed that the RTs to targets containing the character foot semantic radicals preceded by foot picture primes (586 ms) were significantly faster than those preceded by control picture primes (665 ms) (F(1, 34) = 9.275, p = 0.004). The RTs to targets containing the non-character foot semantic radicals preceded by foot picture primes (631 ms) were also significantly faster than those preceded by control picture primes (714 ms) (F(1, 34) = 12.165, p = 0.001). The difference between the RTs to foot meaning-loaded only target characters preceded by foot picture primes (668 ms) and to those preceded by control picture primes (709 ms) were marginally significant (F(1, 34) = 3.296, p = 0.078). Further analysis from paired samples t-tests showed that the RT differences (665 ms RT-control picture – 586 ms RT-foot picture = 79 ms) of target phonograms incorporating the character foot semantic radicals were significantly larger than the RT differences (709 ms RT-control picture – 668 ms RT-foot picture = 41 ms) of the foot meaning-loaded only target characters (t(34) = 3.145, p = 0.003, Cohen’s d = 1.08). The RT differences (714 ms RT-control picture – 631 ms RT-footpicture = 82 ms) of target phonograms incorporating the non-character foot semantic radicals were also significantly larger than the RT differences (709 ms RT-control picture – 668 ms RT-footpicture = 41 ms) of the foot meaning-loaded only target characters (t(34) = 2.332, p = 0.026, Cohen’s d = 0.8). These results demonstrated that the character and non-character foot semantic radicals embedded in the transparent phonograms were both semantically activated.

No significant difference was found between the RT differences (665 ms RT-control picture – 586 ms RT-foot picture = 79 ms) of target phonograms incorporating the character foot semantic radicals and those (714 ms RT-control picture – 631 ms RT-footpicture = 82 ms) incorporating the non-character foot semantic radical (t(34) = −0.189, p = 0.851, Cohen’s d = −0.065). This indicated that the strength of the semantic activation of character foot semantic radicals was the same as that of non-character foot ones.

Discussion

Results from Experiment 1 showed that transparent target phonograms incorporating both character and non-character foot semantic radicals exhibited larger priming effects than the foot meaning-loaded only target phonograms. This indicated that foot semantic radicals embedded in the transparent target phonograms, whether character or non-character ones, were semantically activated.This finding is consistent with previous findings [22, 27]. The observed priming effects in part-of-speech judgment tasks suggest that semantic radicals—particularly in transparent phonograms—may pre-activate relevant semantic content that supports the identification of the character as a verb, consistent with the idea of radical-level semantic contribution.

What mechanism may have enabled this activation event? Presumably, this activation may have stemmed from the foot semantic radicals’ retrieval of their own semantic information. This account appears especially solid given the fact that character foot semantic radicals, which can stand alone as simple characters, have their lexical-level semantic representation in storage [15]. Even for non-character foot semantic radicals, it is quite possible that they have developed their own semantic information from Chinese native speakers’ way of learning them in childhood schooling. In kindergarten, for example, the Chinese non-character foot semantic radical, “辶”, was frequently taught in conjunction with and supported by the character foot semantic radical “” [13]. There is little exaggeration that Chinese children have acquired the “” and “辶” as an intimate pair. In their mind, the two types of foot semantic radicals are equivalent counterparts, indistinguishable from each other. Put another way, both character and non-character foot semantic radicals can really possess their own semantic information.

Since the composed phonograms always make their presence with their own semantic information, the same thing seems to be true of the embedded character and non-character foot semantic radicals. It is unclear, however, how the semantic information of character and non-character foot semantic radicals resides in mind. On the one hand, the sub-lexical foot semantic radicals appear to have full sovereignty over their semantic information, so as to meet the needs of retrieving it whenever necessary. Put another way, the semantic information of character and non-character foot semantic radicals enjoys a high degree of autonomy. On the other hand, the sub-lexical foot semantic radicals have partial semantic overlap with their composed transparent phonograms. Their semantic information might be a part of the semantic information of their composed transparent phonograms. It is meant that the semantic information of character and non-character foot semantic radicals does not stand on its own feet. It finds its existence in the shelter of the semantic information of the semantic radicals’ embedding phonograms. In short, the semantic information of character and non-character foot semantic radicals seems to be in a special status. It is at once highly autonomous and at the same time largely ancillary to the semantic information of the composed transparent phonograms. In whatever case, the semantic information of character and non-character foot semantic radicals and that of their composed phonograms can be said to stay along with each other. They keep each other’s company more often than not.

Also, it is noteworthy that character foot semantic radicals exhibited the same magnitude of priming effects as non-character ones. This may be due to the identical manner in which Chinese users learn and internalize the semantic information of both types of semantic radicals [13].

Moreover, the observed interaction between semantic radical type and transparency further refines our understanding of radical-level semantic processing. In the case of transparent phonograms, character semantic radicals elicited slightly stronger priming effects compared to non-character radicals. This suggests that character radicals may benefit from a dual source of semantic support—both from their independent lexical representation and from their semantic congruency with the whole character. In contrast, although non-character radicals also showed facilitation, their effects were less sensitive to transparency, possibly due to their looser lexical representation and stronger dependency on the whole-character context. This interaction supports the view that semantic transparency modulates the degree to which different types of radicals contribute to word recognition, reflecting a graded interplay between radical-level semantic autonomy and character-level semantic integration.

As the present experiment is confined to transparent target phonograms wherein foot semantic radicals enjoy less semantic autonomy in that they are semantically embraced by the composed phonograms, it is interesting to make a close scrutiny over opaque phonograms wherein foot semantic radicals enjoy more semantic autonomy since they bear no semantic relevance to their composed phonograms. Toward this purpose, Experiment 2 was conducted.

Experiment 2

Method

Participants

The participants in Experiment 2 were identical to those recruited in Experiment 1. Although the same participant group was used across the three experiments, the two experiments were conducted one week apart, with mutually exclusive sets of materials to minimize carryover and fatigue effects.

Materials and design

Experiment 2 was totally identical to Experiment 1, except that the targets in Experiment 2 were opaque phonograms.

Specifically, the first type of target phonograms incorporated character foot semantic radicals, and the second type incorporated non-character foot semantic radicals. Both types of target phonograms possessed no semantic relationship with the foot semantic radical, be it character or non-character (e.g., the target phonograms “趴”/pa1/, meaning “to lie down”, and “邀”/yao1/, meaning “to invite”, contain the character (“”) and non-character (“辶”) foot semantic radicals respectively, but none of them denote meaning relevant to the bodily motion executed by human foot effector).

The third type was phonograms that contained neither character nor non-character foot semantic radicals and bore no meaning related to human foot effector (e.g., the target phonogram “游”/you2/, meaning “to swim”, contains no foot semantic radical, and bears no connection with the meaning of human foot effector-executed motion). Each type consisted of 10 Chinese phonograms (see Table 3 for sample materials), and the character structure, number of strokes (F(2,18) = 1.403, p = 0.271), part of speech (both verbs), and mean frequency (F(1.107,0.154) = 2.241, p = 0.168) were matched in the three types.

[IMAGE OMITTED: SEE PDF]

All target characters were carefully screened using authoritative references including the Xinhua Dictionary (11 th edition) and Handian online dictionary (https://www.zdic.net/hans/) to confirm their semantic association with “foot”. Then, 25 native Chinese speakers who did not participate in the main experiment rated the semantic relatedness of these characters to “foot” using a 7-point Likert scale (1 = not related at all, 7 = extremely related). The results demonstrated strong semantic associations (mean rating = 1.300, 1.284, and 1.236, for three types of target characters, respectively), confirming that all selected characters were not semantically related with the concept “foot”.

Two types of prime pictures were matched with 30 target characters and 10 noun controls (for the purpose of meeting the needs of the part-of-speech decision task), with a total of 80 prime-target pairs being fully used across trials.

Procedure

The procedure in Experiment 2 were identical to those used in Experiment 1. The two experiments stood one week apart.

Results

Data of 34 participants were included in the statistical analysis (Data of 6 participants were deleted due to their low accuracy). The mean accuracy for the part-of-speech judgment task was 93.31%. The errors and reaction times data exceeding 2.5 SD (1.77%) from the mean RT were excluded from further analyses. Table 4 and Fig. 2 display the mean reaction times.

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

A two-way repeated-measures ANOVA was performed on the RTs. Results showed that the main effect of prime picture was significant (F(1, 33) = 5.238, p = 0.029, MSE = 40,777.838, η2 = 0.137). The main effect of target phonogram was also significant (F(1.627, 53.706) = 6.733, p = 0.004, MSE = 24,807.9, η2 = 0.169). The interaction between prime picture and target phonogram was also significant, (F(2, 66) = 4.955, p = 0.014, MSE = 16,308.556, η2 = 0.131).

The simple effect analysis showed that the RTs to targets containing character foot semantic radical preceded by foot picture primes (608 ms) were significantly faster than those preceded by control picture primes (650 ms) (F(1, 33) = 5.697, p = 0.024). The RTs to targets containing non-character foot semantic radical preceded by foot picture primes (636 ms) were also significantly faster than those preceded by control picture primes (683 ms) (F(1, 33) = 11.384, p = 0.002). The difference between the RTs to targets containing non-foot semantic radicals preceded by foot picture primes (660 ms) and to those preceded by control picture primes (655 ms) were not significant (F(1, 33) = 0.087, p = 0.782). These results demonstrated that both character and non-character foot semantic radicals embedded in the opaque phonogram were semantically activated. It seems that the semantic activation of character foot semantic radicals exerted a facilitatory effect on the composed opaque phonograms’ processing.

Further analysis showed that when preceded by foot picture primes, the RTs to targets containing character foot semantic radicals were significantly faster (51 ms) than those to targets containing non-foot semantic radicals (p = 0.008). Also, the RTs to targets containing non-character foot semantic radicals were significantly faster (24 ms) than those to targets containing non-foot semantic radicals (p = 0.047). This pattern of results endorsed the above findings.

Discussion

Results from Experiment 2 showed that both character and non-character foot semantic radicals embedded in the opaque target phonograms were semantically activated. This finding is similar to that for transparent phonograms obtained in Experiment 1. The meaning is that, both character and non-character foot semantic radicals embedded in opaque phonograms also possess their own semantic information and that the semantic information stays along with that of their composed opaque phonograms, nothing unlike the behavior of that of character and non-character foot semantic radicals embedded in transparent phonograms. What is special is that in opaque phonograms, no semantic relevancy exists between the character and non-character foot semantic radicals and the composed phonograms. The semantic information of the sub-lexical foot semantic radicals and that of the composed phonograms stand apart from each other in the mind. There is a greater semantic autonomy entitled to character and non-character foot semantic radicals in opaque phonograms as compared to those in transparent phonograms. The character and non-character foot semantic radicals seem to be semantically independent of their composed opaque phonograms.

The experimental results also showed that the semantic activation of character and non-character foot semantic radicals facilitated their composed opaque phonograms’ recognition, which is consistent with Tong et al. [18]. This finding goes against Wang et al. [21] which reported inhibitory effects on the part of semantic radicals. A common account of the observed inhibitory effects is that the semantic radicals’ semantic information has to be repressed to pave the way for the retrieval of the semantic information of the composed phonograms, since the two sources of semantic information are in conflict. Regarding the reported faciliatory effects, no known account is put forward so far. A plausible explanation may be as follows: In processing a Chinese phonogram, three general steps are involved. First, the component radicals are identified. Second, the identified component radicals are integrated so as to identify the whole character. Third, the semantic (as well as orthographic and phonological) information of the whole character is retrieved so as to fulfill the processing task. With respect to the foot semantic radical-based opaque phonograms, as concerned in the present study, the foot semantic radical needs to be first identified so as to be integrated with the other radical to serve the purpose of identifying and orthographically, semantically, or phonologically retrieving the whole character. Based on the experimental design of the present study, the “foot” prime picture might have pre-activated the semantic information of the foot semantic radical. This might in turn have assisted with the identification of the foot semantic radical, presumably in a top-down manner, which finally led to a faster processing of the whole character. In phonogram processing studies, it is contentious whether semantic radicals exert facilitatory or inhibitory effects on the composed opaque phonograms. The present study seems to take the side of the facilitatory one.

The interaction between radical type and transparency observed across Experiments 1 and 2 also yields important implications. In opaque phonograms, the facilitation effects of character and non-character radicals were more comparable, and the difference between the two types diminished relative to that observed in transparent phonograms. This pattern suggests that when semantic overlap between the radical and the whole character is absent (i.e., in opaque characters), the radicals’ own semantic representations—especially for character radicals—operate more independently. The relative independence of radical-level processing becomes more salient under opacity, diminishing the advantage held by character radicals under transparency. Therefore, the interaction reflects how the semantic system flexibly adapts to the structural properties of the stimulus, shifting reliance between radical autonomy and character-level integration based on transparency.

Experiment 2 takes advantage of the foot semantic radical-based opaque phonograms, which, on the basis of Experiment 1, suggests that the semantic information of foot semantic radicals seems to stand and operate rather independently of the composed phonograms’ semantic information. A big wonder arising is: How should the foot semantic radical behave if it is displaced into a Chinese pseudo-character context? Its behavior may shed further light on the semantic relationship between the foot semantic radical and its composed phonogram. This exact wonder inspires Experiment 3.

Experiment 3

Method

Participants

The participants in Experiment 3 were identical to those recruited in Experiment 1 and 2. A priori sample size calculations were conducted using G*Power 3.1 (Faul et al., 2007). The analysis indicated that 24 participants would be required to detect a medium effect size (f = 0.25) [3] with 80% power for a two-way within-subjects ANOVA (α = 0.05, number of groups = 1, number of measurements = 2*2*2 = 8, non-sphericity correction = 1). These calculations adhered to the guidelines for repeated measures wthin-subjects ANOVA (a priori) provided in the G*Power 3.1 Manual [10] and Brysbaert [1]. To account for possible exclusions, a total of forty participants were recruited.

Although the same participant group was used across the three experiments, each experiment was conducted one week apart, with mutually exclusive sets of materials to minimize carryover and fatigue effects.

Materials and design

The experiment adopted a 2 (Type of semantic radicals: foot vs. control semantic radicals) × 2 (Characterhood of semantic radicals: character vs. non-character) × 2 (Position of semantic radicals in primes: left vs. right) design.

Primes in this experiment were all pseudo-characters built by manipulating the types of semantic radicals, the characterhood of semantic radicals, and the position of semantic radicals in primes, such that they resembled the left–right structured phonograms. The two types of semantic radicals were foot semantic radicals (i.e., and 辶) and control ones (i.e., 耳 and阝). The foot semantic radicals took on two forms, namely, the character foot semantic radical “” and the non-character foot semantic radical “辶”. The control (ear) semantic radicals, similarly, also took on the two forms, namely, the character ear semantic radical “耳” and the non-character ear semantic radical “阝”. Regarding the position of the semantic radicals in primes, they were placed either at the left side of the pseudo-phonogram primes or at the right side. All the pseudo-characters were of the left–right structures. Thirteen simple characters (e.g., “中”,/zhong1/, meaning “middle”) were selected to replace the role of phonetic radicals, such that the combination of the two types of radicals resembled maximally a Chinese genuine phonogram. In total, there were 8 groups of primes, each of which contained 13 pseudo-characters (see Table 5 for sample materials).

[IMAGE OMITTED: SEE PDF]

Thirteen genuine characters were chosen as targets, which were all semantically related to the foot semantic radicals, either character or non-character ones, that had helped to build the pseudo-phonogram primes. However, all the targets were semantically unrelated to the control (ear) semantic radicals, either character or non-character ones, that had performed the same building function.

As the target characters, which were used for a character decision task, were all genuine Chinese characters, 7 pseudo-characters were created by either adding strokes to or removing strokes from genuine Chinese characters. Eight groups of primes were matched with 20 targets (13 genuine characters and 7 pseudo-characters). In total, 160 prime-target pairs were utilized across trials.

Procedure

Each trial began with a fixation signal for 300 ms, followed by a blank screen for 300 ms. Then the pseudo-character prime was presented for 500 ms, followed by another 300 ms blank screen. The target (a character, either a genuine character or a pseudo one) was then presented and remained visible until participants had made a decision. An interval of 600 to 800 ms was set between the two consecutive trials. Participants were required to decide whether the target character was a genuine character or not as quickly as possible by pressing the “F” or “J” keys. The configuration for “pressing key” were also counterbalanced across participants, with half using “F” for “GENUINE CHARACTER” and “J” for “PSEUDO-CHARACTER”, and the other half vice versa. Accuracy and reaction times were recorded. This experiment was also conducted in the Key Laboratory of Foreign Language Learning and Cognitive Neuroscience in SISU.

Experiment 3 stood one week apart from Experiment 2. The experiments were always presented in the same order (Exp 1 → Exp 2 → Exp 3). While this fixed order raises concerns of practice effects, we attempted to mitigate this by spacing the experiments over three weeks and using non-overlapping materials.

Results

Data of 34 participants were included in the statistical analysis. The mean accuracy for the primed character decision task was 94.32%. The errors and data exceeding 2.5 SD (1.67%) from the mean RT of each condition were discarded. Table 6 and Fig. 3 display the mean reaction times.

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

A three-way repeated-measures ANOVA was conducted on the RTs. Results showed that the main effect of types of semantic radicals embedded was not significant (F(1, 31) = 1.070, p = 0.309, MSE = 251.183, η2 = 0.033). The main effect of characterhood of semantic radicals in primes was not significant (F(1, 31) = 0.041, p = 0.841, MSE = 20.908, η2 = 0.001). The main effect of position of semantic radicals was not significant (F(1, 31) = 0.229, p = 0.636, MSE = 76.104, η2 = 0.007). The interaction between types of semantic radicals embedded and characterhood of semantic radicals in primes was not significant, (F(1, 31) = 0.244, p = 0.625, MSE = 58.446, η2 = 0.008). The interaction between types of semantic radicals embedded and position of semantic radicals was not significant, (F(1, 31) = 0.387, p = 0.538, MSE = 109.647, η2 = 0.0.12). The interaction between characterhood of semantic radicals in primes and position of semantic radicals was also not significant, (F(1, 31) = 0.573, p = 0.455, MSE = 149.94, η2 = 0.018). The interaction between types of semantic radicals embedded, characterhood of semantic radicals in primes and position of semantic radicals was also not significant, (F(1, 31) = 0.402, p = 0.531, MSE = 155.75, η2 = 0.013).

The simple effect analysis showed that the RTs to targets preceded by pseudo-phonogram primes containing the character foot semantic radicals (515. ms) were not significantly faster than those preceded by pseudo-phonogram primes containing the character control (ear) semantic radicals (518 ms) (F(1, 31) = 1.071, p = 0.309). The RTs to targets preceded by pseudo-phonogram primes containing the non-character foot semantic radicals (516 ms) were not significantly faster than those preceded by pseudo-phonogram primes containing the non-character control (ear) semantic radicals (517 ms) (F(1, 31) = 0.155, p = 0.696). The RTs to targets preceded by pseudo-phonogram primes incorporating the character foot semantic radicals (516 ms) in the right position were not significantly faster than those preceded by pseudo-phonogram primes incorporating the non-character foot semantic radicals in the right position (516 ms) (F(1, 31) = 0.067, p = 0.797). The RTs to targets preceded by pseudo-phonogram primes incorporating the character foot semantic radicals (518 ms) in the left position were not significantly slower than those preceded by pseudo-phonogram primes incorporating the non-character foot semantic radicals in the left position (515 ms) (F(1, 31) = 0.426, p = 0.519). These results indicated that neither character nor non-character foot semantic radicals were semantically activated in pseudo-characters.

Discussion

Results from Experiment 3 showed that while the foot semantic radicals embedded in the prime pseudo-characters may have undergone semantic activation, such activation did not reach statistical significance, irrespective of their characterhood and position in the primes. This finding falls quite out of our expectation. Both pseudo-characters and opaque phonograms had embedded the same semantically unrelated foot semantic radicals. How does it come about that the same foot semantic radicals, when embedded in opaque phonograms, were semantically activated, but not when embedded in pseudo-characters? The only difference therein lies in that all the composed opaque phonograms were genuine Chinese characters which possessed their own semantic information, while all the composed pseudo-characters were artificial non-genuine Chinese characters which were possessed of no semantic information. It might appear that foot semantic radicals are not semantically self-sufficient. However, we must interpret this result with caution. The absence of statistically significant effects in this experiment does not constitute definitive evidence that semantic radicals cannot be activated without a real host character. Rather, the null findings should be taken as inconclusive under the specific task demands and stimulus conditions applied in the current design. The activation of their semantic information may be largely pivoted upon their composed characters’ semantic information. The presence of the composed characters’ semantic information seems to have a fate-deciding role in determining whether the foot semantic radicals’ semantic information can be activated or not. If it is present, the foot semantic radicals’ semantic information is well subject to activation. If not, it is not subject to activation. The composed characters’ semantic information seems to assume the function of awakening the foot semantic radicals’ semantic information which seems to be in a dormant state. Once being awakened, it gets out of its dormant state, taking on vitality and potency, ready to find an existence. Otherwise, it always remains in the inertia state, never be ready to exhibit its existence. Put briefly, foot semantic radicals’ semantic information is other-reliant, instead of self-sufficient. Things being such, this is a rather novel finding.

Importantly, this finding can also be understood within broader models of context-sensitive semantic processing. As proposed by Connell and Lynott [4] and Horchak and Garrido [11], semantic activation during language comprehension is not fixed but dynamically modulated by contextual relevance. In line with this view, our results tentatively suggest that the semantic activation of foot radicals depends on the semantic status of the host character, highlighting the flexible and layered nature of semantic representation during lexical access.

Alternatively, it is possible that other task paradigms or stimuli could reveal subtle or context-dependent activation effects that were not captured here. Therefore, while the current results suggest a potential reliance of radical semantic activation on the semantic content of the host character, they should not be overgeneralized. Further studies employing varied methods (e.g., neurophysiological measures, semantic priming with different timing parameters, or deeper semantic tasks) are needed to clarify whether and how semantic radicals can be activated independently in non-lexical contexts.

General discussion

The semantic activation of semantic radicals, either character or non-character ones, is a topic of great concern. Previous research has primarily focused on verifying the existence of sub-lexical semantic activation. No known inquiry is dedicated to examining the nature of semantic radicals’ semantic information as well as the mechanism that brings about the activation event. By conducting three experiments, the present study took an initiative to in this respect.

Experiment 1, which included transparent phonograms, found that the semantic information of both the embedded character and non-character foot semantic radicals was activated, much in alignment with previous studies (Wang et al., 2017; [27]. This finding indicates that foot semantic radicals seem to have great sovereignty in determining the activation of their semantic information, although they are semantically embraced or enveloped by the composed transparent phonograms. However, given the limitation that transparent phonograms bear close semantic relevance to their embedded foot semantic radicals, it is hard to determine the degree of sovereignty the foot semantic radicals have over their semantic information.

To overcome this deficiency, Experiment 2, which employed opaque phonograms, was performed. Opaque phonograms, by definition, are semantically unrelated to their embedded foot semantic radicals. This provides an advantage of examining the degree of the sovereignty foot semantic radicals possess over their semantic information. It was found that the semantic information of both the character and non-character foot semantic radicals which were embedded in opaque phonograms was also activated, no less than its activation when the foot semantic radicals were embedded in transparent phonograms. This finding demonstrates that the foot semantic radicals are entitled to a high autonomy in determining the activation of their semantic information, having little to do with the composed phonograms’ semantic information.

To further corroborate this speculation, Experiment 3, which employed pseudo-characters, was conducted. Pseudo-characters, by definition, not only bear no semantic relation to their embedded foot semantic radicals, but have no semantic information of their own and are thus nonsensical at all. This provides a special advantage of investigating the full right the foot semantic radicals gain over their semantic information. The observation was that the semantic information of both the character and non-character foot semantic radicals, once being displaced from the composed phonograms’ semantic information, failed to be activated. A possible account is that foot semantic radicals have no absolute sovereignty or autonomy over their semantic information. The semantic information of foot semantic radicals stays there but it is somehow plunged into a dormant state, needing the semantic awakening by the composed phonograms.

Based on the above findings, the general conclusion to be arrived at may be that foot semantic radicals, either character or non-character ones, are possessed of their own semantic information but are semantically strongly attached to the composed phonograms. When staying along with the composed phonograms’ semantic information, the foot semantic radicals’ semantic information is imbued with vitality, ready to act. When estranged from the composed phonograms’ semantic information, the foot semantic radicals’ semantic information immediately loses its potency, becoming inertia.

In addition, the present study found that the semantic activation of foot semantic radicals facilitated, rather than inhibited, the processing of the composed opaque phonograms. This finding provides endorsement to the “faciliatory side” [18]. The hidden mechanism may be that the pre-activation of the foot semantic radical’s semantic information had assisted with the retrieval of its orthographic information and thereby accelerated its orthographic identification and the whole character’s processing in a top-down manner.

Importantly, these findings also contribute to broader discussions of semantic processing by demonstrating how sub-lexical semantic features are dynamically modulated by context. Specifically, our results resonate with theories of parallel and context-sensitive semantic activation during language comprehension [4, 11], which emphasize that the semantic system can simultaneously maintain and process multiple, potentially conflicting representations depending on contextual support. Just as perceptual and conceptual features can be flexibly activated at the sentence level depending on context, the activation of semantic radicals in our study similarly appears to be contingent upon the semantic viability of the whole character. This suggests that semantic activation in Chinese orthography is not purely componential or bottom-up, but is tightly integrated with top-down contextual influences—aligning with interactive models of language processing. By drawing this parallel, we aim to situate our study within the broader cognitive framework of language comprehension, underscoring that the observed patterns are not unique to Chinese characters but may reflect general principles of how semantic systems operate across languages. In this way, our findings have theoretical implications beyond the domain of Chinese orthography, contributing to a unified view of flexible, context-sensitive semantic processing across scripts and modalities.

While this study provides valuable insights, certain limitations warrant consideration. First, while the lexical decision and semantic categorization tasks have been widely used to investigate implicit semantic processing, they may not provide direct evidence of semantic activation at the radical level. The absence of explicit semantic judgment tasks means our conclusions about semantic attachment remain tentative. Future studies incorporating more direct measures (e.g., semantic relatedness judgments or priming paradigms with explicit semantic tasks) would help clarify the nature of these effects. Second, it should be noted that although the present study focused on semantic radicals, the potential influence of phonetic radicals cannot be fully ruled out. Given that phonograms are inherently dual-component structures involving both semantic and phonetic radicals, the latter may influence semantic access, either by contributing phonological cues or by affecting radical competition. The properties of phonetic radicals (e.g., their consistency, frequency, and phonological transparency) were not systematically controlled in our experiments, which may have introduced uncontrolled variance in the observed effects. Therefore, the current conclusions regarding semantic radical processing should be interpreted with caution. Future studies are needed to explicitly control or orthogonally manipulate both semantic and phonetic radical features to further clarify their respective and interactive roles in Chinese lexical processing. Third, we acknowledge that not all Chinese characters share a uniform structural layout, primarily due to constraints imposed by radical positioning and character composition rules. Although efforts were made to balance character types across conditions, the inherent structural heterogeneity remains a limitation that may affect the generalizability of our findings. Finally, although we categorized characters with “辶” as transparent phonograms with non-character semantic radicals, some items (e.g., “逾”, “返”) may have relatively weak associations with foot-related meanings. This was partly due to constraints in balancing word frequency and stroke number across conditions. While we conducted semantic relatedness ratings to control for transparency (see Materials and Design), minor inconsistencies remain. Future research will refine material selection based on more fine-grained semantic norming.

Taken together, the present study has obtained very interesting and novel findings, well attaining its objectives. However, it may have raised more questions than it had attempted to answer. First, studies employing alternative experimental paradigms and techniques are needed to provide more solid and converging evidences. Second, it is tempting to confirm whether other types of semantic radicals would display a similar pattern of behavior.

Conclusion

In the realm of Chinese character processing studies, there is a widely-accepted belief that semantic radicals—regardless of whether they are characters or non-characters—can undergo semantic activation. However, few studies have explored the possible mechanism governing this activation as well as the nature of the semantic information of semantic radicals involved therein. The present study sought to address this gap by conducting three experiments that foreground three distinct types of Chinese logograms (i.e., transparent phonograms, opaque phonograms, and pseudo-characters) built from the Chinese character and non-character foot semantic radicals.

The results revealed that the foot semantic radicals embedded in both transparent and opaque phonograms were semantically activated. Quite out of our expectation, the foot semantic radicals embedded in pseudo-characters, which much resembled opaque phonograms in possessing no semantic information pertaining to semantic radicals, were not semantically activated. A plausible account is that foot semantic radicals stand on pars with their composed phonograms in possessing their own semantic information. The former has a great say in determining the retrieval and use of its own semantic information, but it is semantically highly attached to the latter, such that it cannot live without the latter’s semantic company. This interpretation also sheds light on the facilitatory effects of foot semantic radicals obtained for the opaque phonograms in the present study, which involves a controversial issue in the literature [2, 18].

In conclusion, the semantic activation of Chinese character and non-character semantic radicals may implicate the nature as well as the complex operation of the semantic information therein. Future research should further investigate the conditions under which radical meaning is retrieved, and how character-level and radical-level semantics interact during visual word recognition.

Data availability

The datasets and materials generated during the current study are available from the corresponding author on reasonable request.

References

Brysbaert M. How many participants do we have to include in properly powered experiments? A tutorial of power analysis with reference tables. J Cogn. 2019;2(1):16. https://doi.org/10.5334/joc.72.

Chen MJ, Weekes BS. Effects of semantic radicals on Chinese character categorization and character decision. Chin J Psychol. 2004;46(2–3):181–96.

Cohen J. Statistical power analysis for the behavioural sciences. New York: Academic Press; 1969.

Connell L, Lynott D. Is a bear white in the woods? Parallel representation of implied object color during language comprehension. Psychon Bull Rev. 2009;16:573–7.

Dang M, Zhang R, Wang X, Yang J. The interaction between phonological and semantic processing in reading Chinese characters. Front Psychol. 2018;9:2748. https://doi.org/10.3389/fpsyg.2018.02748.

Faul F, Erdfelder E, Lang AG, Buchner A. GPower 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences*. Behavior Research Methods. 2007;39(2):175–91. https://doi.org/10.3758/BF03193146.

Feldman LB, Perfetti CA. Morphological and phonological processing in reading Chinese. In: Wang J, Inhoff AW, Chen HC, editors. Reading Chinese script: a cognitive analysis. Mahwah, NJ: Lawrence Erlbaum Associates; 1999. p. 241–64.

Feldman LB, Siok WWT. Semantic radicals contribute to the visual identification of Chinese characters. J Mem Lang. 1999;40:559–76. https://doi.org/10.1006/jmla.1998.2629.

Flores d’Arcais GB, Saito H, Kawakami M. Phonological and semantic activation in reading kanji characters. J Exp Psychol Learn Mem Cogn. 1995;21:34–42. https://doi.org/10.1037/0278-7393.21.1.34.

G*Power 3.1 Manual. 2023. Available from: https://www.psychologie.hhu.de/fileadmin/redaktion/Fakultaeten/Mathematisch-Naturwissenschaftliche_Fakultaet/Psychologie/AAP/gpower/GPowerManual.pdf.

Horchak OV, Garrido MV. Language comprehenders are sensitive to multiple states of semantically similar objects. J Mem Lang. 2024;135:104478.

Hung YH, Hung DL, Tzeng OJL, Wu DH. Tracking the temporal dynamics of the processing of phonetic and semantic radicals in Chinese character recognition by MEG. J Neurolinguist. 2014;29:42–65. https://doi.org/10.1016/j.jneuroling.2013.12.003.

Kuo LJ, Li Y, Sadoski M, Kim TJ. Acquisition of Chinese characters: the effects of character properties and individual differences among learners. Contemp Educ Psychol. 2014;39(4):287–300.

Li R. An analysis of meaning illustration by pictophonetic characters applied in the teaching of Chinese as a foreign language. Appl Linguist. 2005;2:104–9.

Liu L, Jiang Y, Zhang Q, Li Z. Shengfuyu Yifuzai Xingshengzi Shibiezhongde Yingxiang Jizhi [The influential mechanism of semantic radicals and phonetic radicals in Chinese phonogram recognition]. Psychol Explor. 2022;42(3):219–25.

Packard JL. The Morphology of Chinese: A Linguistic and Cognitive Approach. Cambridge: Cambridge University Press; 2000.

Tan L, Hoosain R, Peng D. Role of early presemantic phonological code in Chinese character identification. J Exp Psychol Learn Mem Cogn. 1995;21(1):43–54.

Tong X, Xu M, Zhao J, Yu L. The graded priming effect of semantic radical on Chinese character recognition. Front Psychol. 2021;12:611066.

Wang B, Li Z, Wu L, Zhang J. Effects of embodied simulation on understanding Chinese body action verbs. Acta Psychol Sin. 2019;51(12):1291–305. https://doi.org/10.3724/sp.j.1041.2019.01291.

Wang J, Zhang J. The effects of category consistency and neighborhood size of the semantic radical on the semantic processing of Chinese character. Acta Psychol Sin. 2016;48(11):1390–400. https://doi.org/10.3724/sp.j.1041.2016.01390.

Wang X, Wu Y, Zhao S, Ni C, Zhao M. The effects of semantic radicals and phonetic radicals in Chinese phonogram recognition. Acta Psychol Sin. 2016;48(2):130–40.

Wang XS, Pei M, Wu Y, Su Y J. Semantic radicals contribute more than phonetic radicals to the recognition of Chinese phonograms: behavioral and ERP evidence in a factorial study. Frontiers in Psychology. 2017;8:2230. https://doi.org/10.3389/fpsyg.2017.02230.

Williams C, Bever T. Chinese character decoding: a semantic bias? Read Writ. 2010;23(5):589–605. https://doi.org/10.1007/s11145-010-9228-0.

Wu Q, Chan Y, Lavallee JP, Chen HC, Chang KE, Sung Y. Processing Chinese hand-radicals activates the medial frontal gyrus. Neural Regen Res. 2013;000(20):1837–43. https://doi.org/10.3969/j.issn.1673-5374.2013.20.002.

Xing H. A statistic analysis of components of the character entries in the HSK Graded Character. Chin Teach World. 2005;002:49–55.

Zhang H, Su I, Chen F, Ng ML, Wang L, Yan N. The time course of orthographic and semantic activation in Chinese character recognition: Evidence from an ERP study. Lang Cogn Neurosci. 2019;35(3):292–309. https://doi.org/10.1080/23273798.2019.1652762.

Zhang R, Wang Z, Wang X, Yang J. N170 adaptation effect of the sub-lexical phonological and semantic processing in Chinese character reading. Acta Psychol Sin. 2021;53(8):807.

Zhang X, Cai W, Dang M, Zhang R, Wang X, Yang J. The neural correlates of sub-lexical semantics and its integration with the lexical meaning in reading Chinese characters. J Neurolinguist. 2024;69:101176.

Zhou L, Peng G, Zheng H, Su I, Wang WSY. Sub-lexical phonological and semantic processing of semantic radicals: a primed naming study. Read Writ. 2013;26(6):967–89.

Zhou Y. Xiandai hanzizhong shengpangde biaoyin gongneng wenti [To what degree are the “phonetics” of present-day Chinese characters still phonetic?]. Zhongguo Yuwen. 1978;146:172–7 (in Chinese).

Zou Y, Tsang YK, Wu Y. Semantic radical activation in Chinese phonogram recognition: evidence from event-related potential recording. Neuroscience. 2019;417:24–34.