Positivity Bias and Cultural Differences in Acquiring Haihao in Chinese as a Second Language

Abstract

This study examines how Chinese as a Second Language (CSL) learners acquire the Chinese stance marker haihao with a focus on type and saliency. A total of 56 participants took part in the research, including 28 English-speaking CSL learners and 28 native Chinese speakers. The study utilized two evaluation judgment tasks. Results showed that participants categorized haihao into two simplified groups, guided by the economy principle and a positivity bias. English-speaking learners, influenced by a stronger positivity bias, tended to select more positive options, while Chinese participants favored slightly negative ones. Saliency improved the accuracy of recognizing negative haihao among American learners and low positive haihao among Chinese participants, though it was less effective for ambiguous expressions. These findings highlight how cultural differences and language saliency impact the interpretation of stance markers, offering insights for improving CSL teaching strategies.

1. Introduction

Vocabulary plays a crucial role in language acquisition, as noted by Wilkins (1972), serving as the foundation for learning a new language. Researchers consistently highlight vocabulary as the basis for acquiring both native and foreign languages. Recent studies further confirm this foundational role, emphasizing the challenges and strategies for vocabulary acquisition in second language learners, including Chinese as a Second Language (CSL) learners (Zhu and Wang 2022). Shen (2009) reinforces this by stating that “learning words is the key to picking up the language”, especially in the context of CSL (p. 74).

However, the process of building and structuring this foundational vocabulary is often challenging for CSL learners due to the complexities of Chinese semantics. One key challenge is the prevalence of polysemous words—words that have multiple meanings—which can confuse learners. For example, the Chinese word haihao “still good” is commonly used but rarely analyzed. Although it literally means “still good”, haihao can express a range of meanings, from slightly positive to negative (Chen and Zhao 2022). This raises two important questions: How positive is “still good”, and how can it also mean “not good”?

(1)	a.	Ta	suiran	lao	le,	danshi	jiyili		haihao.
		he	although	old	asp	but	memory		nevertheless good
	“Despite being old, his memory remains good”.
	b.	Zhe	bu	dianying	wo	juede	haihao,	feichang	wuliao.
		this	not	movie	I	think	not good	very	boring
	“I think this movie is not good; it’s very boring”.

In (1a), haihao conveys its literal meaning of “still good”, implying a generally positive evaluation. In contrast, in (1b), it expresses a negative evaluation, meaning the movie is “not good”, despite its literal interpretation. This contrast illustrates the polysemous nature of haihao, prompting this study’s exploration into how CSL learners interpret such words with multiple meanings (Chai and Bao 2023).

A key aspect of haihao’s usage is its function as a stance marker—a linguistic expression used to indicate the speaker’s attitude, evaluation, or perspective toward a proposition (Biber and Finegan 1989). Stance markers allow speakers to subtly convey their feelings about the information being presented, such as approval, doubt, or neutrality. In this study, haihao functions as a stance marker that can signal either a mildly positive evaluation, as in (1a), or a negative assessment, as in (1b). Understanding how learners acquire stance markers is crucial for grasping the intricacies of CSL vocabulary acquisition (Zhou and Goh 2023).

One useful framework for analyzing such challenges is Prototype Theory (Rosch 1975), which suggests that people categorize information by organizing it into categories with core and peripheral meanings. In this context, the literal meaning of haihao in (1a) can be seen as its “core meaning”, while the more negative meaning in (1b) is a “peripheral meaning”. Previous studies have shown that core meanings are processed more quickly than peripheral ones (Liang 2014), with learners focusing first on central or prototypical examples when forming categories (Ashby and Maddox 2005). This view aligns with recent research showing the relationship between the difficulty of acquisition and the prototypicality of meanings (Chen and Zhao 2022).

In addition to prototypicality, L1 transfer—where a learner’s first language influences second language (L2) acquisition—is another crucial factor (Ellis 1994). The presence or absence of a similar word in the learner’s L1 can significantly impact their understanding of polysemous words in L2. Positive L1 transfer occurs when similarities between L1 and L2 facilitate learning, while negative transfer happens when differences hinder it (Lado 1957). Research on CSL has found both positive and negative effects of L1 transfer on vocabulary acquisition (Chai and Bao 2023).

Despite its common usage, there is little research on the acquisition of haihao among CSL learners. The difficulty learners face in interpreting its various meanings, and how prototypicality and L1 transfer influence this process, remain underexplored. This study aims to fill this gap by addressing the following research questions:

(1)

What is the ranking of difficulty among different types of the polysemous attitude marker haihao for CSL learners?

(2)

How do contextual hints influence CSL learners’ judgments of the polysemous attitude marker haihao?

2. Previous Studies of Haihao in Mandarin Chinese

The examination of the Chinese evaluative polysemous construction haihao presents an intriguing linguistic inquiry, with studies debating whether it should be regarded as a cohesive phrase rather than a mere combination of the words hai and hao.

Over the course of linguistic exploration, the Mandarin adverb hai has attracted significant attention for its diverse meanings. Early analyses by Chao (1968) identified three distinct uses of hai: as a temporal adverb, a degree adverb, and an evaluative adverb. Similarly, Li and Thompson (1989) categorized its meanings as “still/even”, “also”, and “moderately”, with the latter aligning with Chao’s evaluative adverb category. More recently, Hsieh and Wang (2020) expanded this categorization for CSL (Chinese as a Second Language) learners, identifying five meanings of hai: temporal hai, additive hai, comparative hai, moderate hai, and counter-expectation hai. The counter-expectation meaning is particularly relevant to the current study’s focus on the stance marker haihao, as it arises when an event contradicts the speaker’s expectations, with hai underscoring the disparity between reality and expectation (Liu 2000; Wu 2009). Similarly, the Mandarin word hao, meaning “good”, has been widely examined by scholars. Lü (1980) categorized hao as an interjection that conveys a variety of moods, while Miracle (1991) explored its use in social actions, topic transitions, and activity closures.

While there is a substantial body of work on hai and hao as individual words, there has been less focus on whether the combination haihao should be considered a single lexical item. This combination warrants attention due to its inseparable use in Mandarin discourse. In many cases, hai and hao can appear together without forming a fixed meaning, as seen in (2), where the literal meanings “still” and “good” apply.

(2)	Wo	juede	xiansheng	de	piqi	hai	(suan)	hao.
	I	think	husband	poss	temper	still	(sort of)	good
	“I think the husband’s temper is still good” (PPT Corpus, Marriage, 2019/8/29).

However, in cases where haihao conveys meanings beyond the literal combination of “still” and “good”, it functions as a genuine compound, exhibiting rigidity that prevents insertion of other words without disrupting its cohesive meaning. Consider the following examples:

(3)	Haihao/	*Hai	hen	hao	yijing	xuehui	youyong	le.
	fortunately	still	very	good	already	learn	swim	asp
	“Fortunately, I’ve already learned how to swim”.
	“*It’s still very good that I’ve already learned how to swim” (Sinica Corpus, 4.0).

In (3), haihao in its sentence-initial position functions as “fortunately”, a meaning that cannot be derived by simply adding hai “still” and hao “good”. The inability to insert hen “very” between hai and hao in (3) demonstrates how breaking the integrity of haihao distorts its intended meaning, and underscores that haihao functions as a cohesive phrase. Treating it as a unified expression is essential to preserving its unique evaluative meaning, which often transcends the literal meanings of its individual components.

To clarify why haihao functions as a stance marker, it is essential to examine its role in conveying the speaker’s evaluative judgment or attitude. Stance markers are linguistic elements that encode the speaker’s assessment or affective stance toward the proposition (Biber and Finegan 1989). In this regard, haihao often conveys a positive or mitigated evaluation, particularly in contexts where the speaker expresses relief, satisfaction, or acceptance of a situation. For instance, in (3), where haihao means “fortunately”, it reflects the speaker’s positive assessment of having learned how to swim. The use of haihao here indicates that the speaker’s initial expectations may not have been met, yet they ultimately adopt a favorable stance toward the outcome. This evaluative function is a hallmark of stance markers, as they provide insight into the speaker’s subjective perspective.

Furthermore, the cohesive nature of haihao as a compound phrase reinforces its role as a stance marker, since the fixed combination of hai and hao conveys meanings that are not merely additive but evaluative in nature. This evaluative meaning often signals the speaker’s attitude toward a deviation from expected outcomes, thereby solidifying haihao’s classification as a stance marker.

Thus, the structural unity of haihao as a distinct lexical item is emphasized across different sentence positions. This cohesiveness supports the argument for recognizing haihao as a single linguistic entity rather than a combination of independent words. This perspective aligns with broader linguistic studies, such as Chao’s (1968) exploration of Chinese grammatical principles, which highlights the interdependency of components within phrases to convey specific meanings.

3. Empirical Studies of Chinese Polysemous Words Acquisition

This section examines the dynamics between Chinese polysemous words and second language acquisition focusing on three key factors: prototypicality, L1 transfer, and contextual hints. The empirical studies reviewed here highlight the role of Chinese polysemous words in language learning, contributing to a deeper understanding of how these factors influence the interpretation and acquisition of meanings in a second language.

3.1. Prototypicality

Several studies have examined the challenges that polysemous words present to L2 learners (Evans and Tyler 2004), with particular focus on whether acquisition depends primarily on prototypicality (Kellerman 1978, 1979). Despite progress, gaps remain in understanding polysemy acquisition in Chinese. Notable studies by Liang (2014) and Chen and Wang (2020) emphasize the role of prototypicality in acquiring polysemous verbs in Mandarin.

Liang (2014) explored the acquisition of the polysemous term shang (“to go up”) among Chinese L1 speakers and L2 learners. His study demonstrated that prototypicality strongly predicts acquisition patterns, as concrete meanings are consistently acquired before abstract or figurative ones by CSL learners. Similarly, Chen and Wang (2020) analyzed the polysemous verb da (“to hit”), finding that learners first acquire concrete, prototypical meanings before tackling more abstract senses. This work underscores the importance of prototypicality in early vocabulary acquisition.

Expanding on this foundation, Zhang et al. (2023) investigated the role of prototypicality within the context of Chinese Resultative Verb Compounds (RVCs) and proposed a prototype-based Dual Route Processing Model (DRPM). This study examined whether prototypical components and whole-word frequency affect RVC acquisition among L2 learners. Their findings, based on quasi-experimental research with Malaysian secondary school students, suggest that prototypicality is a stronger predictor than frequency in learning RVCs. This research highlights the pedagogical value of emphasizing prototypical components in instructed second language acquisition. Together, these studies demonstrate how prototypicality significantly influences the learning trajectory of polysemous and compound structures in Chinese, offering critical insights for teaching strategies in CSL and CFL contexts.

3.2. L1 Semantics Transfer

Adult L2 learners often rely on their existing semantic system when acquiring a second language, raising questions about the effectiveness of semantic restructuring in this process. A common initial strategy is mapping new L2 words onto familiar L1 meanings (Ellis and Schmidt 1997). However, successful L2 development depends on the learner’s ability to fully restructure these concepts to align with the new language.

Jiang (2004) found that while L2 learners initially rely on L1 concepts, they often struggle to fully develop L2-specific meanings, resulting in incomplete semantic development. Similarly, Yuan (2012) demonstrated that L1 transfer can influence the syntax–discourse interface in Chinese “daodi… wh-questions”, challenging the belief that this interface remains unaffected by L1 influence. Furthermore, Krish and Oh (2020) identified L1 interference in Chinese L2 speech acts, causing miscommunication in areas such as compliments and requests.

Building on these studies, the current research anticipates positive L1 transfer, predicting that participants whose L1 shares similarities with haihao perform better in acquiring this polysemous word.

3.3. Contextual Hints

Contextual hints play a crucial role in language acquisition, particularly when they are salient and prominently noticeable, significantly enhancing learners’ comprehension. Previous studies have examined the impact of strong contextual cues on second language acquisition. Kroll et al. (2013) explored how salient cues affect L2 vocabulary learning, focusing on English speakers acquiring Dutch words. Their experiments demonstrated that while semantic cues alone did not bypass L1 interference, the introduction of perceptually salient cues helped reduce L1 reliance and improved L2 vocabulary acquisition. This finding challenges the belief that all bilinguals mediate L2 through conceptual means, particularly in the early stages of learning, and highlights the potential of salient perceptual stimuli to foster more effective L2 processing.

Similarly, Cai and Lee (2010) investigated the influence of contextual clues on learners’ strategies and knowledge sources during L2 listening comprehension. They found that learners flexibly used local, global, and extra-textual cues to infer unfamiliar word meanings, utilizing both linguistic and non-linguistic knowledge sources. This flexibility in strategy contrasted with previous studies that suggested a stronger reliance on local linguistic cues in reading comprehension. Building on this, Zhang and Peng (2024) investigated the impact of multimodal reading strategies, such as integrating visual and auditory inputs, on comprehension and memory retention among CSL learners. Their study showed that combining modalities, like reading while listening, greatly enhanced learners’ understanding and recall, highlighting the critical role of contextual and perceptual cues in facilitating effective language acquisition.

Drawing from these insights, the current study anticipates that salient contextual hints enhances CSL learners’ acquisition of haihao. By incorporating strong contextual cues, it is expected that learners’ accuracy in interpreting the four types of haihao will improve, leading to better overall comprehension and learning outcomes.

To conclude, the acquisition of Chinese polysemous words in second language learners is shaped by prototypicality, L1 transfer, and contextual hints. Prototypicality provides a cognitive foundation, with learners acquiring concrete, frequently used meanings before abstract ones. L1 transfer, while often a barrier, can also act as a scaffold when typological similarities exist between L1 and L2. Finally, salient contextual hints play a critical role in reducing L1 reliance and fostering better comprehension. Together, these factors highlight the importance of targeted, evidence-based teaching strategies to optimize second language acquisition of Chinese polysemous words.

4. Research Design

4.1. Participants

To address these questions, 56 participants were recruited and divided into two distinct groups. The first group comprised 28 learners of Chinese as a Second Language (CSL) enrolled in the Mandarin Training Center at a public university in northern Taiwan. The second group consisted of 28 native Chinese speakers. Within the CSL group, participants were selected from the B1 and B2 levels of the Common European Framework of Reference (CEFR) to ensure a normal distribution. These CSL participants, hailing from English-speaking countries such as the United States, England, Singapore, and the Philippines, all shared English as their first language. The native Chinese speaker group consisted of 28 individuals without any identified language, hearing, or sight issues. Gender and proficiency were not considered variables in this study.

4.2. Tasks and Materials

An evaluation judgment task was chosen as the primary instrument for this study. This task requires participants to assess and categorize their interpretations of the target word, haihao, based on specific prompts. Before conducting the task, we introduce an analytical framework to guide the analysis. Previous research (e.g., Munnich et al. 2013) has demonstrated the effectiveness of judgment tasks in assessing L2 acquisition, particularly in capturing learners’ comprehension data. This method allows us to identify which category of haihao learners understand or interpret, providing insights into their grasp of polysemous meanings. Additionally, the use of both weak and strong saliency hints in the task enables us to explore how varying levels of contextual cues impact comprehension, offering a better understanding of learners’ interpretation patterns. This approach is essential for addressing the study’s research questions related to the classification of haihao.

4.2.1. An Analytical Framework

According to Hyland (2005), attitude markers are a type of metadiscourse marker used to express the writer’s or speaker’s attitude or opinion toward the proposition. In this study, the primary focus is on haihao, which serves as an attitude marker by expressing an evaluative perspective. Biq (2004) identified three distinct meanings of evaluative haihao by examining its use in spoken and written discourse. Building on Biq’s research, this study introduces a revised framework with four types of evaluative haihao, each outlined below.

Type 1: Haihao₁ (Low Positive)

The first type of evaluative haihao means “nevertheless good” or “just okay”. This interpretation aligns closely with the literal meanings of its parts, hai and hao, placing the evaluation on the positive side of the spectrum. Haihao₁ is used in sentences where positive remarks either precede or follow its use. It acts as a transitional signal after negative comments, effectively shifting the overall assessment to a positive one by rejecting or downplaying the preceding negative remark, as shown in the example below.

(4)	Ta	sui	yi	lao,	raner	jiyili	haihao,
	he	though	already	old	however	memory	nevertheless good
	“Despite his advanced age, his memory is still rather good”,
	neng	qingchu	jide	duonian	qian	fasheng	de	shi.
	can	clearly	recall	many years	ago	occur	DE	events
	“enabling him to vividly recall events from many years ago”.

In (4), the speaker provides a positive evaluation by stating that the elderly man’s memory is still intact. The phrase qingchu jide “clearly recall” highlights the elderly man’s ability to remember events from many years ago, suggesting that his memory is haihao “still rather good”, aligning with the positive end of the evaluation spectrum.

Type 2: Haihao₂ (Average)

Speakers use haihao₂ to convey neutrality or indifference, indicating neither a clearly positive nor negative evaluation. It can follow ambiguous text, providing a middle-ground perspective. For instance,

(5)	A:	Dajia	juede	zhongji	lanjiezhan	haokan	ma?
		everyone	feel	final	intercept station	good watch	part
		“Do you all think Ultimate Interception Station is good?”
	B:	Wo	juede	haihao	puputongtong.
		I	feel	so-so	ordinary
		“I think it’s so-so, quite ordinary” (PTT Corpus, Movie, 2022/6/8).

In this example, Person B uses haihao₂ to describe the movie as neither particularly good nor bad, suggesting an average evaluation. The phrase puputongtong “ordinary, so-so” underscores this neutral stance.

While sentiment analysis usually focuses on positive and negative expressions, neutral words like haihao₂ can also have significant sentiment weight, often overlooked in analysis. Methods such as lexicon-based and statistical techniques are beginning to address this by considering neutrality scores (Taboada et al. 2011). Studies by Colhon et al. (2017) and Kuhlmann et al. (2017) emphasize the importance of neutral words, showing that context affects their evaluation. Understanding neutral words and their sentiment polarity is crucial for accurate opinion mining. This study further explores the role of neutral words like haihao₂.

Type 3: Haihao₃ (Low Negative)

Haihao₃ expresses a judgment of “worse than average but not so bad” and appears on the negative side of the evaluative scale, often integrating more negative elements. Although the literal meaning of hao is “good”, haihao can imply negativity when used with negative expressions. Haihao₃ typically appears in sentences with mild negative comments and conveys moderate negativity without being too extreme. It also acts as a transitional signal following positive remarks, effectively shifting the overall evaluation to a negative one by downplaying the prior positive comment. Research on semantic prosody shows that the semantic profiles of words depend on their typical collocates (Hauser and Schwarz 2023). For example, set in carries a negative connotation due to its association with words like rot, decay, malaise, and despair (Louw 1993). This form of haihao offers a subtle way to convey mild negativity without expressing complete disapproval or strong criticism.

(6)	Quanqu	dajiyue	de	yinliang	dou	taida	le,
	whole song	percussion	poss	volume	all	too big	part
	“The volume of the entire percussion composition is too loud”,
	tingbujian	zhuxuanlu,	zhongjiande	jiyue	bufen,
	cannot hear	main melody	middle	percussion	part
	“the main melody cannot be heard, as for the middle part of the percussion”,
	qixian	haihao,	houlai	jiu	youxie	luandiao.
	at first	not so bad	later	then	somewhat	mess up
	“it was not too bad at first, but it then became somewhat messed up”.
	(Sinica Corpus, 4.0)

In (6), the speaker assesses the performance with a negative tone, starting with the critique that the composition’s volume is taida, “too loud”. Within this negative context, the speaker distinguishes between different performance aspects. Initially, the percussion is described as haihao, “worse than neutral but not too bad”, which then shifts to luandiao, “messed up”. Compared to the luandiao comment, the initial use of haihao indicates a less negative yet still critical evaluation.

Type 4: Haihao₄ (Negative)

The fourth type of haihao1 suggests “nothing in particular” or “nothing remarkable”, but it actually means “not good” or “poor”, indicating negation or rejection. Therefore, haihao₄ is classified as “negative” to differentiate it from “positive” (cf. Biq’s (2004) “negation”). It is used in sentences with strong negative comments either before or after it, and these comments are more negative (e.g., dou bu tai shou “all not too familiar” as in (7)) than those associated with haihao₃.

(7)	A:	Nimen	banshang	youmeiyou	fasheng	yixie	bijiao
		your	class	have not have	happen	some	relatively
		youqude	shiqing?
		interesting	matter
		“Is there anything interesting happening in your class?”
	B:	Women	ban,	qishi	haihao	ye,	dajia	dou	bu	tai	shou.
		our	class	actually	not good	part	everyone	all	not	too	familiar
		“Well, in our class, not really. We are not that familiar with each other yet”
		(Biq 2004, p. 19).

Speakers use haihao₄ to emphasize negation or rejection beyond mere indifference. It can precede negative comments to intensify rejection, indicating strong disapproval or unwillingness. Alternatively, it may follow strong negative remarks to reinforce the speaker’s clear rejection of a particular proposition or evaluation. Negation has been explored over the years (Hossain et al. 2022). Research on the syntax and semantics of adjectival modification in Chinese provides insights into how adjectives and adverbs convey rejection and negation (Grano 2012; Paul 2021).

4.2.2. The Evaluation Task

This study employed an evaluation judgement task designed with two distinct versions: Evaluation Judgment Task-I (EJT-I) and Evaluation Judgment Task-II (EJT-II). EJT-I included weak hints, promoting a more unbiased interpretation, while EJT-II featured strong contextual hints to guide participants more directly. The dual-version task was specifically crafted to ensure comprehensive and unbiased data collection. Each version of the task revolved around narratives involving the main characters Rosie and Jennie, which were created by the researchers. Both versions unfolded through eight scenarios per task, with each scenario corresponding to one of four types of evaluative haihao identified in the study: low positive, average, low negative, or negative. Participants answered questions related to each scenario. In EJT-I, participants were asked to interpret haihao using only weak saliency hints, while in EJT-II, scenarios provided strong saliency hints to assist with evaluations. Overall, the evaluation task consisted of twenty questions: eight with weak saliency hints and two filler questions in EJT-I, and eight with strong saliency hints and two filler questions in EJT-II. To enhance clarity and reduce ambiguity, images accompanied each test item.

In the experiment, each participant saw animations created by Microsoft Office PowerPoint 2021 on a computer screen while listening to a recorded story containing the test items. After presenting the story, the experimenter asked the participants a question and required them to select an answer. An example question from EJT-I is given in

Table 1. An Example Question of EJT-I.

The Participants Saw:	The Participants Heard:
Scene 1	Jennie wen: “Nabu dianying shibushi hen haokan ya?” “Jennie asked, “Is that movie good?””
Scene 2	Mickey shuo: “Qishi haihao yeh, qingjie hen bu jingcai”. “Mickey said: “Actually, it’s not good. The plot is very unexciting”.”
Scene 3	Qingwen Mickey hui xiang zai kanyici nabu dianying ma? “Would Mickey want to watch the movie again?”

, while an example question from EJT-II is presented in

Table 2. An Example Question of EJT-II.

The Participants Saw:	The Participants Heard:
Scene 1	Jennie wen: “Nabu dianying shibushi hen haokan ya?” “Jennie asked, “Is that movie good?””
Scene 2	Mickey kandao yiban jiu shuizhaole, ta shuo: “Qishi haihao yeh, qingjie hen bu jingcai”. “Mickey fell asleep halfway through watching it, he said: “Actually, it’s not good. The plot is very unexciting”.”
Scene 3	Qingwen Mickey hui xiang zai kanyici nabu dianying ma? “Would Mickey want to watch the movie again?”

.

4.3. Procedure

Participants were briefed on the research objectives and expectations before the study and provided informed consent. The study involved two evaluative judgment tasks, EJT-I and EJT-II, with a 5-min break between them. Participants were instructed to respond based on intuition. Each task was introduced by the experimenter using animations and pre-recorded stories featuring characters named Rosie and Jennie, emphasizing the evaluative term haihao. Each story lasted around 15 min, making the total duration of both tasks about 40 min. The study was conducted in a quiet room at NTNU’s main library, with all responses audio-recorded for transcription. EJT-I included weak hints, whereas EJT-II had strong hints. Participants selected the best interpretation of haihao from four options for each scenario. A chi-square analysis assessed the frequency of correct responses, providing insights into participants’ understanding and highlighting significant performance differences between the tasks.

5. Results and Discussion

5.1. Type Effect

Table 3. Two Language Groups’ Accuracy Rates of the Four Haihao Types.

Type Group	Type 1		Type 2		Type 3		Type 4
	f	%	f	%	f	%	f	%
American	83	74.11	87	77.68	59	52.68	52	46.43
Chinese	91	81.25	90	80.36	63	56.25	65	58.04

presents the accuracy rates of American and Chinese participants across the four haihao types.

In the American group, the highest accuracy rate was observed in recognizing Type 2 haihao, reaching 77.68%. This reflects a strong comprehension of “average” haihao among the American participants. Type 1 haihao followed closely, with a slightly lower accuracy rate of 74.11%, indicating a somewhat reduced ability to recognize “low positive” haihao. There was a significant drop in accuracy for Type 3 haihao, which fell to 52.68%, suggesting difficulties in interpreting “low negative” haihao. The lowest accuracy rate for the American group was recorded for Type 4 haihao, at 46.43%, indicating a level of difficulty similar to Type 3 for this “negative” haihao type. Chi-square tests within the American group revealed a significant difference in performance across haihao types, with a p-value of 0.0094**. The accuracy ranking for the American group was Type 2 = Type 1 > Type 3 = Type 4, with the best performance for “average” haihao.

The Chinese group exhibited a similar trend, with lower performance in Type 3 and Type 4 compared to Type 1 and Type 2. However, the Chinese group achieved a higher accuracy rate for Type 1 (81.25%) than the Americans (74.11%). Chinese participants consistently showed higher accuracy rates across all haihao types compared to their American counterparts (Type 1: 81.25% vs. 74.11%; Type 2: 80.36% vs. 77.68%; Type 3: 56.25% vs. 52.68%; Type 4: 58.04% vs. 46.43%). The chi-square tests for the Chinese group revealed a significant discrepancy in performance across the four haihao types, with a p-value of 0.0431*. The accuracy order for the Chinese group was Type 1 = Type 2 > Type 4 = Type 3.

A chi-square test revealed no significant difference in performance between the American and Chinese groups in accuracy rates across the four types, with a p-value of 0.586. Both American and Chinese participants tended to group Type 1 and Type 2 together, and Type 3 and Type 4 together, indicating a dichotomous interpretation. Chinese participants generally showed slightly higher accuracy rates. P-values indicated no significant difference in behavior between the two groups for any haihao types (Type 1: 0.567; Type 2: 0.831; Type 3: 0.732; Type 4: 0.256).

The post hoc comparison in

Table 4. Post Hoc Comparison of Two Groups’ Performance Across Four Haihao Types.

Group	Type	Type 1	Type 2	Type 3	Type 4
American	Type 1	-	1	0.0163 *	0.0007 ***
	Type 2	1	-	0.0022 **	6.3 × 10−5 ****
	Type 3	0.0163 *	0.0022 **	-	1
	Type 4	0.0007 ***	6.3 × 10−5 ****	1	-
Chinese	Type 1	-	1	0.0015 **	0.0038 **
	Type 2	1	-	0.0027 **	0.0066 **
	Type 3	0.0015 **	0.0027 **	-	1
	Type 4	0.0038 **	0.0066 **	1	-

Notes. * p < 0.05, indicating statistical significance at the 5% level. ** p < 0.01, indicating statistical significance at the 1% level. *** p < 0.001, indicating high statistical significance at the 0.1% level. **** p < 0.0001, indicating extremely high statistical significance at the 0.01% level.

sheds light on the distinct tendencies of American and Chinese participants in differentiating between the four haihao types in both EJTs. The p-values provide insights into the statistical significance of the observed differences in accuracy rates between specific pairs of haihao types for each language group.

For American participants, significant differences in Type 1 vs. Type 3 (p < 0.05), Type 1 vs. Type 4 (p < 0.01), Type 2 vs. Type 3 (p < 0.01), and Type 2 vs. Type 4 (p < 0.01) indicated their accuracy in discriminating between low positive and low negative, low positive and negative, average and low negative, as well as average and negative haihao expressions. However, the absence of significant differences in Type 1 vs. Type 2 (p > 0.05) and Type 3 vs. Type 4 (p > 0.05) suggested potential challenges in accurately distinguishing between low positive and average and low negative and negative haihao expressions among American participants. This indicated that American participants perceived Types 1 and 2 as being in the same category, while Types 3 and 4 were seen as another category.

Similarly, for Chinese participants, significant differences observed in Type 1 vs. Type 3 (p < 0.01), Type 1 vs. Type 4 (p < 0.01), Type 2 vs. Type 3 (p < 0.01), and Type 2 vs. Type 4 (p < 0.01) underscored their ability to differentiate between low positive and low negative, low positive and negative, average and low negative, as well as average and negative haihao expressions. The absence of significant differences in Type 1 vs. Type 2 (p > 0.05) and Type 3 vs. Type 4 (p > 0.05) suggested potential similarities in recognizing low positive and average and low negative and negative haihao expressions, indicating that Chinese participants also regarded Types 1 and 2 as the same category, while Types 3 and 4 as another category.

Table 5. Distribution of Errors with Type 1 to Type 4.

Type	Group	Option A		Option B		Option C		Option D
		f	%	f	%	f	%	f	%
Type 1	American	-	-	28	25	1	0.89	0	0
	Chinese	-	-	20	17.86	1	0.89	0	0
Type 2	American	15	13.39	-	-	10	8.93	0	0
	Chinese	6	5.36	-	-	16	14.29	0	0
Type 3	American	1	0.89	8	7.14	-	-	44	39.29
	Chinese	3	2.68	14	12.50	-	-	32	28.57
Type 4	American	3	2.68	14	12.50	43	38.39	-	-
	Chinese	0	0	8	7.14	39	34.82	-	-

presents a comprehensive analysis of error distribution data pertaining to Types 1, 2, 3, and 4 questions within both the American and Chinese groups.

An analysis of participant responses across different haihao types uncovered interesting patterns in how American and Chinese participants approached these questions. In Type 1 questions, where the expected response was “low positive”, participants from both groups tended to choose the “average” haihao option. This suggests a general preference for responses that are one degree more negative than anticipated. For Type 2 questions, which anticipated an “average” haihao response, Chinese participants predominantly selected the “low negative” option. This choice reflects a consistent tendency to opt for a response that is one degree more negative than expected. By contrast, American participants preferred the “low positive” option, deviating one degree less negative than anticipated. This contrast highlights differing response tendencies between the two groups. Type 3 questions, where the expected response was “low negative”, showed a similar pattern for both groups. Participants favored “negative” responses, again selecting an option that is one degree more negative than expected. In Type 4 questions, where the expected response was “negative”, participants mostly chose the “low negative” option. This choice was likely influenced by the absence of options more negative than the expected answer. Overall, this indicates that participants have a consistent inclination to select responses that deviate by one degree from what is anticipated.

As shown above, there is a noticeable tendency among both groups to simplify the haihao expressions into broader categories. The following discussion explores the cognitive, linguistic, and cultural factors that may explain these trends.

Both American and Chinese participants tended to select haihao responses that were slightly more negative than expected. However, Chinese participants consistently chose more negative options when interpreting the “average” type, while American participants leaned toward more positive interpretations. These findings reveal cognitive and linguistic factors influencing the categorization of haihao, with both groups simplifying the expressions into two main categories. This aligns with the principle of economy in information processing, cognitive biases, and linguistic theories such as L1 transfer and Prototype Theory.

Participants grouped haihao into two broad categories: Types 1 and 2 in one and Types 3 and 4 in another. This simplification contrasts with Biq’s (2004) suggestion that distinctions among low positive, average, low negative, and negative meanings of haihao can be easily identified using linguistic databases. The principle of economy, which emphasizes minimizing effort while achieving efficient communication (Zipf 1949), likely explains this tendency. By grouping similar meanings, participants reduced complexity and enhanced comprehension.

Positivity bias also influenced the categorization of haihao. This cognitive bias, the tendency to favor positive over neutral or negative information, explains why participants associated average haihao with more positive interpretations. Americans in particular exhibited a stronger positivity bias, likely due to cultural norms regarding emotional expression, consistent with previous research (Heine et al. 1999).

Error analysis revealed subtle differences in the interpretation of average haihao between American and Chinese participants, attributable to variations in positivity bias. While both groups tended toward slightly more negative responses than expected, Chinese participants favored more negative options, whereas American participants leaned toward slightly more positive ones. These differences are consistent with research suggesting Western cultures exhibit a stronger positivity bias than Eastern cultures (Heine et al. 1999; Salili 1996). Despite the universal positivity bias, these cultural differences highlight varying cognitive and emotional tendencies between the two groups. Americans were more inclined to interpret average haihao positively compared to their Chinese counterparts, likely due to cultural norms surrounding emotional expression rather than language proficiency. The similarity in accuracy rates between American and Chinese participants further supports the notion that language proficiency did not significantly influence the results. Instead, cultural frameworks and emotional regulation strategies appear to have played a more central role in shaping participants’ interpretations. While L1 transfer may still affect second language acquisition (Ellis 1994), in this case, cultural attitudes toward positivity bias provide a more plausible explanation for the observed differences.

Consistent with Prototype Theory (Rosch 1975), second language learners tend to prioritize central, prototypical meanings when confronted with multiple meanings of a polysemous word (Liang 2014). As expected, participants were more accurate in interpreting the prototypical types of haihao (Types 1 and 2), which reflect the core meaning of “still good”. The less prototypical types (Types 3 and 4) were more challenging, as they represent more abstract meanings. Although the differences in accuracy rates were not drastic, the overall pattern of higher accuracy for positive types and lower accuracy for negative types supports Prototype Theory.

The merging of low negative haihao with negative haihao into one category can also be attributed to the principle of economy. When the semantic differences between these types are minimal, it becomes cognitively efficient to merge them, reducing redundancy and simplifying interpretation.

In summary, these findings highlight the interplay between the economy principle, positivity bias, L1 transfer, and Prototype Theory in interpreting haihao. The tendency to simplify haihao expressions into two categories reflects a natural inclination toward efficiency in both communication and cognition. This analysis provides valuable insights into language acquisition theory and pedagogy, emphasizing the influence of cognitive biases, linguistic principles, and cultural variations on the interpretation and categorization of polysemous expressions.

5.2. Saliency Effect

Table 6. Two Language Groups’ Accuracy Rates: Weak vs. Strong Saliency Hints.

Hints	Type Group	Type 1		Type 2		Type 3		Type 4
		f	%	f	%	f	%	f	%
Weak	American	41	73.21	46	82.14	32	57.14	16	28.57
	Chinese	38	67.86	44	78.57	31	55.36	30	53.57
Strong	American	42	75	41	73.21	27	48.21	36	64.29
	Chinese	53	94.64	46	82.14	32	57.14	35	62.50

shows the accuracy rates observed in the EJTs for the American and Chinese language groups.

In scenarios characterized by weak saliency hints, a consistent accuracy trend emerged across haihao types for both groups, with Type 2 (average haihao) exhibiting the highest accuracy rates (American: 82.14%; Chinese: 78.57%). Notably, the Chinese group significantly outperformed the American group in Type 4 accuracy (American: 28.57%; Chinese: 53.57%), while the accuracy rates for the other three types remained relatively comparable between the two groups.

Upon the introduction of strong saliency cues in EJT II, the dynamics shifted, with Type 1 (low positive haihao) emerging as the most accurately identified type for both groups. In this context, the Chinese group notably outperformed the American group (American: 75%; Chinese: 94.64%). Type 4 (negative haihao) demonstrated enhanced accuracy for both groups, with the American group slightly surpassing the Chinese group in accuracy rates (64.29% vs. 62.50%). However, Types 2 and 3 did not exhibit significant improvements in accuracy rates for either group, despite the presence of strong saliency hints.

Table 7. Within Type p-Value Comparison: Weak vs. Strong Saliency Hints.

Type Group	Type 1	Type 2	Type 3	Type 4
American	0.883	0.474	0.384	2.10 × 10−4 ***
Chinese	0.0356 *	0.778	0.886	0.407

Notes. * p < 0.05, indicating statistical significance at the 5% level. *** p < 0.001, indicating high statistical significance at the 0.1% level.

provides a comprehensive within-type p-value comparison, highlighting the impact of saliency hints on the accuracy rates of haihao judgments in the EJTs for both American and Chinese participants.

The first notable distinction emerged in Type 1 judgments. American participants exhibited a non-significant p-value of 0.883, indicating consistent judgments irrespective of saliency hints. By contrast, Chinese participants demonstrated a significant difference with a p-value of 0.0356 *. This suggests that for Chinese participants, the introduction of strong saliency hints triggers a recognition of the Type 1 “low positive” meaning, resulting in a notable improvement in accuracy rates.

The second significant disparity was observed in Type 4 judgments among American participants, as evidenced by a highly significant p-value of 2.10 × 10⁻⁴ ***. This indicates that the presence of strong saliency hints significantly influenced their understanding of Type 4 “negative” haihao. Conversely, Chinese participants displayed a non-significant p-value of 0.407 for Type 4, suggesting that the addition of strong saliency hints did not significantly impact their judgments.

In Type 2 judgments, both American and Chinese participants showed non-significant p-values (American: 0.474, Chinese: 0.778), indicating a lack of significant differences in their judgments with or without saliency hints. Similarly, in Type 3, both groups displayed non-significant p-values (American: 0.384, Chinese: 0.886), suggesting consistent judgments across saliency conditions for this haihao type.

Starting with the types that showed improved accuracy rates when strong saliency hints were provided, within-type p-value comparisons revealed significant effects on the comprehension of Type 4 haihao among American participants (p < 0.01) and Type 1 among Chinese participants (p < 0.05). Initially, the accuracy rate for Chinese Type 1 with weak saliency hints was 67.86%, which was similar to that of the Americans at 73.21%. However, when strong saliency hints were introduced, the accuracy of the Chinese group significantly increased to 94.64%, while the American group only saw a slight increase to 75%.

Conversely, for American Type 4, the accuracy rate with weak saliency hints was 28.57%, significantly lower than that of the Chinese group at 53.57%. Yet, introducing strong saliency hints notably improved the accuracy of the American group to 64.29%, whereas the Chinese group also improved, though not significantly, to an accuracy rate of 62.5%. These findings suggest that strong saliency hints aid in conceptual processing in L2 vocabulary acquisition and help resolve linguistic ambiguity. This aligns with prior research by Kroll et al. (2013) on conceptual cues in L2 vocabulary acquisition and Cai and Lee’s (2010) study on the effect of contextual clues on processing unfamiliar words in L2 listening comprehension. The significant improvement in accuracy rates for negative haihao among the American group and low positive haihao among the Chinese group after the addition of strong saliency hints provides further support for this claim.

Further analysis of lexical bundles within the task questions may explain the performance disparities across the four haihao types under conditions with weak and strong saliency hints. For the Chinese group, Type 1 haihao, characterized by high accuracy rates, likely benefited from strong saliency hints containing explicit lexical cues like hen (e.g., hen shihhe “very suitable”, hen youchengjiougan “very fulfilling”). This could be because Chinese participants as native speakers are sensitive to using hen as an intensifier to amplify the intended meaning, which in this case is the low positive aspect of haihao. The saliency hints included in the study are all disyllabic, contrary to the usage of hen as the default semantically vacuous modifier with monosyllabic adjectives (e.g., hen hao “very good”, hen gao “very tall”), as argued by studies such as Grano (2012) and Zhang (2015). Since the saliency hints in the current study were disyllabic, the use of “hen” intensified the low positive meaning of haihao, and Chinese participants were sensitive to interpreting it as an intensifier; in other words, hen successfully intensified the low positive meaning for Chinese participants.

On the contrary, American participants did not display such sensitivity, which might be due to the weaker degree of intensification that the corresponding English word “very” holds compared to “hen”, as noted by Leffel et al. (2019). They found that stronger intensifiers like “extremely” are more likely than “very” to generate positive implicatures for vague predicates. For instance, “John is not extremely tall” should imply that John is tall to a greater degree than “John is not very tall”. Their claim is based on Bennett and Goodman (2018)’s Study 2, which quantitatively examined a wide range of intensifiers and found “extremely” to be a stronger intensifier than “very”.

Type 4 haihao for the American group showed marked improvement with strong saliency hints featuring lexical cues of bu (e.g., bu taishou “not too familiar”). Unlike Chinese participants who already achieved an accuracy rate of 53.57% without strong saliency hints, American participants struggled with this least prototypical meaning of haihao, which denotes the opposite of the literal meaning of “good”. According to Li (2011), English CSL learners can correctly interpret bu as a negation marker once their Chinese proficiency reaches an intermediate level, explaining why the recruited CEFR B1 and B2 level American participants benefited greatly from the addition of bu. As for why Chinese participants did not see a significant improvement in their accuracy rate of negative Type 4 even with the addition of the saliency hint bu, besides the already comparatively high accuracy rate under the weak saliency hint condition (53.57%), another plausible reason might lie in the choice of words for the phrases with bu included in the questions. Type 4 questions with added saliency hints were EJT-II Questions 5 and 9. In Question 5, the saliency hint was bu jingcai “not exciting”, a clear combination of bu “not” that negates the positive adjective jingcai “exciting”. Therefore, the accuracy rate for this question was high at 75%. However, Question 9 exhibited a lower accuracy rate of 50%; with the saliency hint given as bu taishou “not too familiar”, the appearance of tai “too” may have confused the negativity intended with bu. Compared to reading it as “not familiar” in a negative meaning, participants might have interpreted it as a low positive meaning of “familiar, but not too much”. This could explain why Chinese participants did not achieve a high enough accuracy rate for Question 28, which in turn led to a lower average accuracy rate for Type 4.

The inclusion of strong saliency hints did not alleviate the complexity of differentiating Types 2 and 3 haihao, which overlapped the ambiguous boundary between positive and negative meanings. This finding is consistent with Cheng and Chen’s (2023) study on Chinese frequency adverb acquisition, which indicates that learners struggle with typical adverbs that share vagueness in meanings. The unclear and indefinite meanings of Types 2 and 3 haihao posed difficulties for interpretation, consistent with Hoffner et al.’s (1990) suggestion that learners acquire absolute positives and negatives before comprehending other meanings in an ordered semantic set. Thus, strong saliency hints may cause more confusion in participants’ comprehension and do not significantly aid their interpretations of Types 2 and 3 haihao.

Overall, the observed differences in response to strong saliency hints underscored the complex nature of haihao interpretation and the varying effectiveness of saliency hints across different linguistic contexts. While saliency hints proved beneficial for extreme types such as Types 1 and 4, their impact was limited in disambiguating more ambiguous expressions, as in Types 2 and 3, highlighting the interplay between salient lexical hints and the accuracy of interpretation.

6. Conclusions

This study explored the difficulty ranking of polysemous attitude marker haihao and the role of contextual cues in influencing CSL learners’ judgments. In addressing the first research question, we found that both American and Chinese participants tended to simplify haihao expressions into two broader categories instead of the anticipated four distinct types. This merging, particularly between Types 1 and 2 and Types 3 and 4, highlights participants’ cognitive preference for efficiency, aligning with the principle of economy. Additionally, cultural differences in positivity bias impacted judgments, particularly for Type 2 expressions, where American learners exhibited stronger positivity bias compared to their Chinese counterparts. These findings suggest that CSL learners may rank the difficulty of haihao types not only based on linguistic complexity but also on culturally influenced cognitive biases.

For the second research question, regarding the influence of contextual saliency cues, the study found that strong saliency cues significantly improved accuracy for specific types of haihao expressions, such as negative expressions for Americans and low positive ones for Chinese participants. However, the effect was limited for more ambiguous expressions, reflecting the challenges of interpreting haihao in complex contexts. These insights underscore the importance of contextual clues in shaping learners’ interpretations but also suggest that certain ambiguous expressions resist even the aid of strong cues.

The present findings on haihao interpretation provide valuable pedagogical insights for language teaching and cross-cultural communication. Learners’ tendency to simplify haihao expressions into broader categories underscores the importance of addressing cognitive biases in language comprehension. Instructional strategies should emphasize the subtle distinctions among haihao types to improve interpretative accuracy and foster a nuanced understanding of language use. Incorporating cultural sensitivity training is vital for helping learners navigate variations in interpretation, such as the positivity biases observed among different cultural groups. Activities that encourage reflection on cultural assumptions and diverse perspectives can play a key role in overcoming such biases.

Additionally, teachers can enhance students’ interpretative skills by integrating storytelling, visual aids, and real-life examples to highlight saliency cues, making ambiguous expressions more accessible and meaningful. Drawing from Prototype Theory, instructional approaches should leverage prototypical examples to help learners build robust mental models of language structures, facilitating comprehension and interpretation. Real-life applications and context-rich scenarios can further emphasize the role of saliency in guiding learners toward accurate interpretations.2

Moreover, language instruction should explore potential positive transfer effects by drawing parallels between students’ first languages and the target language, enabling learners to connect familiar cognitive frameworks to new linguistic contexts. By integrating these insights, teachers can design more effective teaching methodologies that address both the cognitive and cultural dimensions of language learning, ultimately fostering a deeper and more holistic understanding of cross-cultural communication.

While the study provides useful insights for language teaching, it also faces certain limitations. First, the participant pool primarily consisted of L1 speakers of English and Chinese, which may have influenced the observed patterns of interpretation. Future research could benefit from a more diverse set of participants, including those from other language backgrounds, to explore whether these findings are generalizable across learners with different L1s. Additionally, the study’s provision of contextual cues was limited, which may have restricted the full exploration of their potential effects. Future research should include more extended and varied contextual saliency cues to better assess their influence. Finally, variations in participants’ CSL proficiency levels were not accounted for, which could affect how haihao is interpreted. Further studies should investigate how different proficiency levels interact with these findings to ensure broader applicability of the results. By addressing these limitations, future research can deepen our understanding of the cognitive and cultural factors influencing the interpretation of polysemous markers like haihao, thereby improving language teaching strategies and cross-cultural communication.