A Contrastive Study of Heteroglossia in the Reasoning of Criminal Judgments of Chinese Mainland and Hong Kong

Within the framework of the Engagement System of Appraisal Theory, this thesis investigates the heteroglossia of the reasoning of criminal judgments of Chinese mainland and Hong Kong and examines their similarities and differences in the employment of heteroglossic engagement resources and underlying causes. The reasoning of 20 criminal judgments of Chinese mainland and Hong Kong produced upon second instance for the same cause of action are collected and built as two separate corpora to carry out the study. It is found that judges of both Chinese mainland and Hong Kong employ various heteroglossic engagement resources to locate position and negotiate with other voices while proceeding with reasoning. Furthermore, they share some similarities in the selection of subtypes of engagement resources, which is attributed to the fact that they hold similar communicative purposes in the reasoning of judicial judgments; while the differences can be interpreted from the distinct legal doctrines in the mainland and Hong Kong and the textual structure of the reasoning of criminal judgments.


Introduction
The reasoning of judgment refers to the judge's justification of his decision (Golding, 2001). As the core of criminal judgments, reasoning serves a variety of functions such as promoting the cause of social peace and the supervision of judiciary (e.g., Liu, 2015;Golding, 2001). Furthermore, many of the functions that court serves require reasoned decision (Golding, 2001). According to Golding (2001), a justification is directed to the audience, say, perhaps the first person to whom the justification is directed is the losing litigant; and to this may be added all other people whose interests might be adversely affected by the result , and therefore, it can be viewed as attempts at rational persuasion, an interaction between the judge and the putative audience. Leenes and Lodder (1994) also propose that legal reasoning should be analyzed as a dialogue game since reasoning is a highly significant feature of judicial argumentation. All of the above statements imply that reasoning is some kind of dialogue between the judge and putative audience of the case, with a pronounced overtone of heteroglossia. More specifically, the process of legal reasoning is highly heteroglossic.
In recent years, with the advance of judicial reform in Chinese mainland, the quality of criminal judgments has been constantly improving, but there are still some problems remaining to be solved especially in the reasoning (Wang, 2005), which affects the progress of judicial justice and fairness to a great extent. While in Hong Kong, part of common law system, judges are required to give full and forceful theoretical proof of their decisions, providing objective and complete basis for the trial of subsequent cases of the similar nature. However, the contrastive study of the reasoning of criminal judgments of Chinese mainland and Hong Kong remains underexplored. Furthermore, to date, linguists have associated legal reasoning with semantics, structuralist, style, rhetoric, semiotics and systemic functional grammar (e.g., Stamper, 1991;Neumann, 2005;Dong, 2010;Mazzi, 2010;Manzin, 2012;Verenich, 2012;Huisman & Blackshield, 2014), while its inherent heteroglossic nature is underrepresented.
The above problems call for the present research, a contrastive study of the heteroglossia in the reasoning of criminal judgments of Chinese mainland and Hong Kong, within the framework of engagement system of appraisal theory. To be specific, the author will investigate how judges manipulate heteroglossic engagement resources to express their stance and negotiate with the other voices in the reasoning. Additionally, the Figure 1. Heteroglossic engagement system (Martin & White, 2005)

Research Questions
The present study makes a contrastive study of the heteroglossia of the reasoning of criminal judgments of Chinese mainland and Hong Kong from addressing the following three questions.
• How are heteroglossic engagement resources distributed in the reasoning of criminal judgments of the mainland and Hong Kong respectively?
• How are these resources realized? What are the specific effects of these engaging strategies on negotiation between judges and other voices?
• What are the similarities and differences between the distribution of these resources in the reasoning of criminal judgements of the mainland and that of Hong Kong? What are the reasons for these similarities and differences?

Data Collection
The reasoning of 20 criminal judgments is collected from http://wenshu.court.gov.cn/ (Chinese mainland) and https://www.judiciary.hk/zh/home/index.html (Hong Kong) and constructed as two separate corpora, namely, the Chinese mainland dataset and Hong Kong dataset; each of them includes the reasoning part of 10 criminal judgments (please see Appendix A for more details). The authenticity of such data is guaranteed as the judgments are uploaded by the courts at different levels for the supervision of exercise of justice. In order to enhance the comparability of the two datasets and the rigorousness of the study, the author selects the reasoning of criminal judgments for the same cause of action, say, drug trafficking, with the length being neither too short or too long, specifically, ranging from 430 words/characters to 597 words/characters. In total, there are 5024 characters and 4774 words respectively in Chinese mainland dataset and Hong Kong dataset. Additionally, it should be noted that all the selected data is the reasoning part of criminal judgments produced upon second instance, for the reason that its heteroglossic feature is more explicit as it also enters dialogue with the previous opinions of the lower court.

Annotation Procedures
Linguistic software UAM Corpus Tool 3.3 is responsible for the annotation of heteroglossic engagement items. According to Taboada and Carretaro (2012, p. 278), as appraisal theory is a flexible interpretive theory, the annotation of appraisal resources is complicated and subjective. Therefore, the author takes two measures to reduce subjectivity and enhance the credibility and replicability of the research, namely, formulating guidelines for annotation, receiving annotator training and taking inter-coder agreement test.

Guidelines for Annotation
• Hold a naturalized reader position. Martin (2000, p. 162) claims that the evaluation of evocations rests on the institutional position from which one is reading. Accordingly, the present research holds a naturalized reading position to guarantee a consistent reading of implied engagement.
• Conduct double coding where different interpretations are possible. Double coding contributes to reduce errors and brings forth deeper understanding of engagement items, both implicit and explicit.

Inter-Coder Agreement Test
Interpretation of engagement resources may vary from annotators regardless of the well-defined explanatory guidelines, for the fact that appraisal is defined as a flexible interpretive theory instead of a definitive model applicable to any discourse. While the annotation is credible if agreement on the categories assigned to units can be shown by coders, to a degree determined by the aims of the study (Artstein & Poesio, 2008, p. 557), which calls for the inter-coder agreement test to produce a standard reference model to annotate the entire corpus.

Analytical Approaches
Both quantitative as well as qualitative approaches are adopted in this study.
The quantitative method is employed to explore the frequency and distribution of each sub-category in heteroglossic engagement system. It aims to respond to the first question proposed in section 3.1, that is, how engagement resources are distributed in reasoning of criminal judgment of mainland and Hong Kong respectively. Following the quantitative findings, qualitative approach attempts to identify the similarities and differences between the two corpora in the selection of heteroglossic engagement resources and investigate the reasons, to answer the third question.

Descriptive Statistics of Heteroglossic Engagement Resources
The reasoning of criminal judgment of the mainland and Hong Kong is processed by UAM Corpus Tool 3.3 which is responsible for the annotation and calculation of heteroglossic engagement resources. After careful annotation of the items, the author obtains the statistics of engagement resources in the two corpora, both global and local.

Global Descriptive Statistics
Global statistics demonstrates the percentage of each subcategory of the whole engagement system which amounts to 100%. The statistical results of the heteroglossic engagement resources in reasoning of criminal judgment of mainland and Hong Kong are shown in Table B1 (see Appendix B- Table 1) from global means.
As for local statistics, it shows the propensity to prefer one resource to the others in each subsystem which account for 100%. Besides the demonstration of the descriptive statistical results by choosing the option 'global counting', the author extends the statistics to local statistics in order to display and compare the frequency of heteroglossic engagement resource in each subtype in the reasoning of criminal judgments of the mainland and Hong Kong, and hence the Table B2 (please see the Appendix B- Table 2). Table B1 and Table B2 reveal a fact that heteroglossic engagement resources are widely distributed in the reasoning of criminal judgment of the mainland and Hong Kong. In this section, examples will be taken from the corpora to explore how judges of Chinese mainland and Hong Kong employ various engagement resources to proceed with reasoning.

Contract (Dialogic Contraction)
Dialogic contraction refers to the utterance that acts to challenge, fend off or restrict the dialogically alternative positions and voices by dint of some locutions, including Disclaim and Proclaim. From Table B1 and Table B2, it can be easily seen that Disclaim occurs more frequently than Proclaim, which reveals that judges of both Chinese mainland and Hong Kong tend to present their views by contracting dialogic space.

Disclaim
Disclaim concerns the textual voice positions itself as at odds with, or rejecting some contrary position (Yu, 2017), which can be achieved by Deny and Counter by which alternative stance is revoked to be deemed as eligible or to be overtly refuted or substituted. Deny contains negation. By virtue of Deny, a writer/speaker explicitly denies alternative point of view through negation. An alternative viewpoint is acknowledged and rejected, overtly disaligning the author/speaker with the explicit or implicit position holder. Besides, Deny is used to contract dialogic space to show the speaker/writer's authority. In Table B1, it can be seen that judges of both mainland and Hong Kong adopt a great number of Deny resources, which can be exemplified as follows.
1) 卢志吼及其辩护人所提意见，经查理由均不能成立，不予采纳。 (Upon investigation, the submissions of Zhihou Lu and his counsel cannot be justified and shall not be accepted.) (Sample 2) 2) As the applicant realistically acknowledges, there is no basis under existing sentencing guidelines upon which this Court could say that the starting point of 20 years' imprisonment was manifestly excessive for the quantity of dangerous drugs concerned. Nor in our view can one say that the one-year enhancement was improper. (sample 18) In the above samples, the denials such as 不能 (cannot), 不 (not), no and nor are employed. By the adoption of ijel.ccsenet.org International Journal of English Linguistics Vol. 10, No. 5;2020 Deny resources, judges first introduce alternative positive voices into the reasoning and acknowledge it and eventually reject it. In sample 2, the denial is adopted twice to invoke and express the proposition that the submission of Zhihou Lu and his counsel cannot be justified as responding to the potential audience who hold the opposite opinion that the their opinions are strong and persuasive, and project the proposition that the submission cannot be accepted as responding to the claim that the defense opinions of the appellant or counsel shall be accepted, from which it can be also seen that the negation is indeed dialogic and heteroglossic. In sample 18, by virtue of not and nor, the judges contract the dialogic space and suppress opposite opinions to voice out their own views, without undermining their authority in the communicative context.
Counter contains concession and counter expectation (Yu, 2017). It covers the resources which indicate the current proposition as replacing or supplanting and thus counter an expected proposition (Martin & White, 2005). By means of such resource, judge does not reject the alternative voices, instead, he respects these voices. In the reasoning of criminal judgment, judges also use Counter to foreground their own opinions, or it can be said that, present and signalize their opinions by countering and replacing the prior expected views.

3) 其虽具有立功表现，但根据本案的具体情节，不足以对其从轻处罚。
(Although the appellant has rendered meritorious service, it is insufficient for a lenient punishment in view of the specific circumstance of the case.) (Sample 5) 4) Be that as it may, the fact remains that, while the starting point for sentence in respect of 824 grammes of narcotic would have merited 21 years' imprisonment (21 years and 1 month to be more exact) under the guidelines in Abdallah, the enhancement for the international factor of bringing the drugs across the border into Hong Kong ought to have been more than 18 months' imprisonment, in accordance with the judgment in HKSAR v Chung Ping Kun[12]: yet, the judge enhanced the sentence for this factor by only 12 months' imprisonment, thus taking the notional sentence after trial to 22 years' imprisonment.
It can be found from the above samples that a variety of Counter resources are distributed in the reasoning of criminal judgments of mainland and Hong Kong, such as 虽 (although) and yet. By using Counter resources, the judges counter the expected proposition and make their opinions more prominent. For instance, in sample 5, with the contribution of Counter resource 虽 (although), the judge replaces the expected proposition that 'the meritorious service necessarily deserves a lighter sentence' and project the opinion that 'the meritorious performance of the appellant is insufficient to receive a lenient punishment in accordance with the specific circumstance of the case'. Likewise, the judge in sample 15 employs yet introducing his viewpoints to supplant the putative audience's expectation that the judge in this case ought to enhance sentence for bringing the drugs across the border into Hong Kong by 18 months' imprisonment.

Proclaim
By Proclaim, the speaker/writer represents the proposition as highly warrantable and thus suppresses or rules out alternative position rather than rejecting a different voice. It can be clearly seen from Table B1 that Proclaim accounts for 34.19% in the reasoning of criminal judgments of the mainland and 13.30% in that of Hong Kong. According to Martin and White (2005), Proclaim exerts its functions to narrow down dialogic space of alternative positions by virtue of Concur, Pronounce and Endorse.
Judges view the proposition as shared with or generally accepted by people while using Concur. They can also ask rhetorical questions whose answer is obvious or self-evident. In these circumstances, the dialogic space is thus contracted since the interpersonal cost of anyone who attempts to put an opposite proposition would otherwise be raised. From Table B1, it can be seen that Concur is not prevailing in the reasoning and even no item of Concur is found in Chinese mainland dataset. 5) Clearly, the third count was completely different in nature from the first two counts and did add to the overall culpability of the applicant. (Sample 16) In criminal judgments, judges employ Concur resources to indicate that they share the same opinion with or have the same knowledge as those assumed to be the hearers/readers of the criminal judgments, or suggest that position presented in the reasoning is generally acknowledged or well-known to close down the dialogic space. By virtue of clearly in sample 16, the judge renders his opinion an overtone of 'objectivity' and increases the interpersonal cost for any view contradicting the 'objective' proposition, persuading the audience to accept his opinion.
Pronounce covers the resources that indicate the speaker's/writer's emphasis or apparent intervention. According to Table B1, significant difference is found in the frequency of Pronounce resources in the two datasets, specifically, the former occupies 17.95% while the latter only takes up 2.66% in the whole global statistics ijel.ccsenet.org International Journal of English Linguistics Vol. 10, No. 5;2020 respectively. In the reasoning of judicial judgments, judges tend to use expressions such as our 本院认为 (court contends that)…, 判决如下 (we hereby decide as follows)…, the fact remains that… to introduce their positions and assume the responsibility for their utterance. The interpersonal cost for any propositions challenging their opinions is therefore increased. Additionally, intensifiers such as indeed is also employed to enhance the warrantability of the proposition against dialogic alternatives.
6) 本院认为，上诉人郭红天、郭红丹、苑某、原审被告人赵博、武某某、候某某贩卖毒品之行为均已构 成贩卖毒品罪。(Our court contends that the drug trafficking committed by appellants Hongtian Guo, Hongdan Guo, Yuan and the accused in the original trial Bo Zhao, Wu and Hou, constitutes the crime of drug trafficking.) (Sample 7) 7) Be that as it may, the fact remains that, while the starting point for sentence in respect of 824 grammes of narcotic would have merited 21 years' imprisonment (21 years and 1 month to be more exact) under the guidelines in Abdallah, the enhancement for the international factor of bringing the drugs across the border into Hong Kong ought to have been more than 18 months' imprisonment, in accordance with the judgment in HKSAR v Chung Ping Kun. (Sample 15) 8) Indeed, in all of the cases to which the applicant has drawn our attention, the defendants received a one-year enhancement of sentence for more or less similar quantities of heroin narcotic brought across the border into Hong Kong.
In the judgments of Chinese mainland, the opinions of the judges are typically introduced by Pronounce resources such as 本院认为 (our court contends) in sample 7. In this way, the judges assume responsibility for the proposition and strongly restrict voice diversity. In sample 15, the fact remains that is taken to reinforce the credibility of the proposition against different voices. In addition, adverb indeed which is used to emphasize a statement or description is also employed in the reasoning. By virtue of the above resources, the authorial voice is set to narrow down the dialogic space and strongly against the dialogic possibilities.
Endorse refers to those formulations that attribute propositions to external sources and frame these propositions as maximally correct, viable and undeniable, that is to say, the author strongly endorse the value of the proposition. In criminal judgments, Endorse is adopted frequently. In the explanation of the reasons for sentences, judges tend to employ Endorse resources to show that their opinions are supported by laws, regulations, evidence or precedents and hence it is warrantable and should not be doubted by the parties and other putative audience of the judgments. 11) It is usually the case that an accused who has given evidence against a co-accused at trial would receive a 50% reduction to his sentence (incorporating the one third discount for a plea of guilty In sample 8, the judge refers to law to support his decisions as they are of high authority and play a crucial role in sentencing. While judges practicing civil law system tend to cite precedents in sentencing as sample 15, to demonstrate the fairness of the court decisions. Besides, evidence such as the report produced upon investigation also helps to enhance the reliability of the judge's opinions. With the support of such Endorsement resource, the judge shows that the decision is made based on law, precedent or accessible evidence rather than an arbitrary act of the judge, making the judge's view more authoritative and reliable and hence restricting alternative voices. More specifically, the judge's employment of Endorse strategy contributes to the alliance with audience and promotes their acceptance of the judgments.

Expand (Dialogic Expansion)
Dialogic expansion refers to those linguistics resources allowing for dialogically alternative voices and positions by dint of some locutions. Such resources, in the reasoning, allow the heteroglossic negotiation between the judges and audience and make the heteroglossic diversity possible.

Entertain
Entertain means that the author/speaker represents the position as but one of many possible positions or voices, that is, the speaker/writer allows for alternatives and entertains those dialogic possibilities. It is usually represented by modal words and expressions such as it seems…., perhaps… and expository questions.
12) 公安机关侦破报告证实，第三次毒品交易确系民警让吸毒人员段某电话联系邵某某购买冰毒，但上诉 人邵某某对此并不知情，其前往约定地点的主观意愿是实施毒品交易，故该次毒品交易应计入其贩毒的 次数，但考虑存在特情引诱的情形，在量刑上可酌情从轻处罚。(The investigation report of public security organ proves that the third drug trafficking was indeed the Duan's calling Shao for the purchase of Ice under the instruction of police. However, Zhao, unaware of the situation, proceeded to the agreed place with the subjective will to traffic in drug and thereupon, this transaction should be counted in the number of drug trafficking, but in consideration of the special lure in this case, the penalty may be lighter as appropriate.) (Sample 9) 13) It seems that following his arrest, the applicant told the police that he was expecting a telephone call from a friend and he indicated his willingness to participate in a controlled delivery operation. (Sample 14) 14) The problem with a strictly arithmetical application of sentencing guidelines is that they can sometimes overlook or obscure features which go to aggravate the starting point for sentence. (Sample 19) In the process of reasoning, modal verbs of low and medium value such as 可 (may), and can are employed by the judges in Chinese mainland and Hong Kong, to demonstrate negotiability and make their proposition projected as contentious. It seems that in sample 9 also opens up the dialogic space for different positions. It can be seen from the above examples that in using such resources, judges respect the fact that there may be some alternatives different from the authorial voice. In the above examples, the use of Entertain resources by judges not only expresses the views of the court, but also leaves room for other views and avoids arbitrariness, so as to demonstrate the fairness of the decisions of the court and makes them more acceptable.

Attribute
Martin and White (2005) explore linguistic phenomenon that dissociate propositions from the author and assign them to others as Attribute. Attribute is typically realized through reporting. Besides, it should be noted that a degree of overlap is found in the lexical items of entertainent and of attribution, despite the fact that these instances are easily disambiguated by the subject of the construct (Read & Carroll, 2012). A proposition can be attributed through Acknowledge or Distance.
Acknowledge does not explicitly imply the indication as to where the voice of the speaker/writer stands related to the proposition. It is usually represented by indirect speech, to be specific, it reports the utterance and opinions of external voice by reporting words (or their nominalizations) or phrases. In reasoning, judges adopt Acknowledge resource strategically to report the opinions from others, bringing in the external voices to render the argument of the judges a hue of objectivity and impartiality. According to Table B1, 5 instances (occupy 3.14 % in the whole heteroglossic engagement system) and 50 instances (take up 27.62% in the whole system) of Acknowledge are identified respectively in Chinese mainland and Hong Kong dataset. It can be found from the above samples that the implicit engagement resource 意见 (submission) and formulations such as said are distributed in the reasoning. By virtue of such kind of reporting words or their nominalizations, the judges bring the external voice into the current text in an indirect manner so as to make their production impersonal and objective. Besides, the relationship between judges and potential readers can be also adjusted. For instance, in sample 10 and sample 17, with the dedication of the process of reporting, the judges assign their authorial responsibility to external voice and avert conflict with the potential readers holding the opposite position. That is, the judges are establishing alignment with them.
An author can Distance themselves overtly from a reported proposition and evades responsibilities for the proposition. Different from Acknowledge, Distance resources are employed to present the authorial voice as refusing to assume the responsibility for the quoted/reported while open up dialogic space to a maximum extent. Distance is scarcely employed in both datasets. Only 3 instance and 7 instances are identified respectively in the mainland and Hong Kong dataset. The employment of the words such as claim, assert and scare quotes can be ijel.ccsenet.org International Journal of English Linguistics Vol. 10, No. 5; 2020 adopted to achieve Distance. As Distance serves to separate or detach the authorial voice from the reported or quoted material, judges, in this way, overtly detach himself from the responsibility for what have been reported or quoted to a maximum extent. Distance is conveyed through the expressions used to separate the authorial voice from the textual voice, including reporting verb like claim, rumor and myth and phrases like it's rumored that…, X claimed to have shown that…etc.
17) 关于上诉人邵某某所提"第三次贩毒是在公安人员的引诱下完成的， 不应计入犯罪的次数"的辩解意见， 经查……(With respect to Shao's submission that "the third drug trafficking was completed under the lure of public security personnel and should not be counted in the number of crimes", upon investigation…) (Sample 9) 18) D1 claims that he and his family are at personal risk because he testified against D2 who was a close family friend in Sri Lanka until the time of this offence. (Sample 12) In virtue of reporting words or scare quotes, the above samples overtly mark the judges' voices as disengaged from the reported and cited propositions to show their disalignment and negotiability. In sample 9 and 12, the judges directly quote the appellant's submissions, distancing from the external voice while presenting their neutrality and refraining from responsibility for the credibility of the propositions.

Comparison of Heteroglossic Engagement Resources in the Two Corpora
From the above analysis of the distribution and realization of heteroglossic engagement resources, it ican be concluded that heteroglossic engagement resources are employed widely in the reasoning of criminal judgment of both Chinese mainland and Hong Kong. In this section, the author will compare the two corpora to examine their similarities and differences in the employment of heteroglossic engagement resources and discuss the reasons.

Similarities and Discussion
It can be learnt from the above analysis of the distribution as well as realization of heteroglossic engagement resources in the two corpora that judges of both Chinese mainland and Hong Kong adopt various heteroglossic engagement resources to locate their position and negotiate with other voices, in order to make their decisions justified and accepted. Following the above analyses, this section will move on to the similarities and difference in respect of the selection of heteroglossic engagement items between the two datasets and attempt to discuss the reasons. Table B2 demonstrate that judges of both mainland and Hong Kong have preference for Disclaim rather than Proclaim, to be specific, 52.38% versus 47.42% and 72.83% versus 27.17% respectively within the Contract-Type. The employment of Disclaim resources contributes to the clear statement of judges' viewpoints or attitudes against other alternative voices; besides, it helps to persuade the putative audience to accept their opinions in the sentencing or even support their statements by increasing the interpersonal cost for any views against their opinions. The intention of demonstrating judicial authority also leads to the prevailing distribution of Disclaim items which close down dialogic space to a maximum extent.

Prefer Deny to Counter within Disclaim
Within the Disclaim-Type, judges of both Chinese mainland and Hong Kong are inclined to employ Deny resources instead of Counter resources (70.45% versus 29.55% and 79.10% versus 20.90% respectively in the mainland dataset and Hong Kong dataset). Rather than posing a proposition by countering an expected opinion, the judges tend to bring forth a firmer and more prominent attitude and stance in reasoning by overtly rejecting the alternative voices. This raises the interpersonal cost to a maximum degree for anyone who intent on contradicting the judges and the authority of the court is shown in this way.

Prefer Acknowledge to Distance within Attribute
Within the Attribute-Type, Acknowledge resources are more favored by judges than Distance resources, accounting for 62.50% versus 37.50% and 75.00% versus 25.00% respectively in the mainland and Hong Kong dataset. With the contribution of Acknowledge resources, they manage to present their impartiality and neutrality in the convincingness-required reasoning. On the contrary, Distance resources, which indicate the overt detachment of the judges from the textual voices, are less adopted.

Reasons for Similarities
From the above analysis, it can be concluded that the reasoning of criminal judgments of Chinese mainland and Hong Kong share some similarities in the selection of engagement resources. Such similar tendency can be ijel.ccsenet.org International Journal of English Linguistics Vol. 10, No. 5;2020 attributed to the similar communicative goals of reasoning of judicial judgments.
Despite the fact that the mainland of China and Hong Kong belong to different legal systems, the reasoning of judicial judgments is dealt with based on the similar purposes. To be specific, the primary goal of reasoning is to justify the judge's decision and persuade the putative audience to accept or even support his opinions (Golding, 2001). Therefore, Disclaim, particularly Deny resources are prevailing in the reasoning of criminal judgments of both datasets, to narrow down the dialogic space for different voices and demonstrate the judicial authority. While they also employ substantial Acknowledge resources to open up the dialogic space to avoid arbitrariness, show their neutrality as well as impartiality, and avert conflict with the audience taking the opposing positions. That is, the judges are establishing alignment with the putative audience of the judgments.

Differences and Discussion
Differences can also be found between the reasoning of criminal judgments of mainland and Hong Kong from Table B2.

Differences within Heteroglossic-Type
As presented in Table B2, judges of the mainland tend to adopt Contract rather than Expansion with the percentage of 71.79% versus 28.21%, while their Hong Kong counterparts prefer Expand than Contract with the proportion of 51.06% versus 48.94%. The ratio of resources for dialogic expansion is slightly higher than those for dialogic contraction in the mainland dataset. It suggests that, on the whole, judges of the mainland are inclined to narrow the dialogic space to suppress the alternative voices. By contrast, larger dialogic space and more alternative possibilities are allowed for the putative audience of criminal judgments of Hong Kong.

Differences within Proclaim-Type
Differences can also be found in the selection of the subtypes of Proclaim. Judges of Chinese mainland are apt to use Pronounces resources instead of Endorse and Concur resources with the percentage of 52.50% versus 47.50% and 0.00% within the Proclaim-Type. With the preference of Pronounces resources, they interpolate themselves into the discourse as overtly liable for the utterances and show the power of the judges. In contrast, Endorse resources are popular with judges of Hong Kong than Pronounce resources and Concur resources (68.00% versus 20.00% and 12.00%). Endorse is a common engaging strategy in the reasoning of criminal judgments in Hong Kong belonging common law system which requires the reference to precedents in sentencing. By virtue of Endorse resources, the judges cite relevant precedents to support their views.

Reasons for Differences
As analyzed above, differences are identified in the reasoning of criminal judgments of the mainland and Hong Kong. The judges of mainland tend to contract the dialogic space while their Hong Kong counterparts do the opposite. Besides, within Disclaim-Type, Pronounce items are mostly widely distributed in the reasoning of criminal judgments of the mainland while Hong Kong judges prefer Endorse. The present study ascribes these differences to the following reasons.
Legal doctrine contributes to their different tendencies in selecting engaging strategies. Differences are found in the legal doctrine of the mainland and Hong Kong belonging to different legal systems, which influence the power of the judges and ultimately affect their reasoning of judicial judgments. In respect of the trial grade system, Chinese mainland conducts 'two-tiered system' in criminal procedures (Bi, 2005). That is, a criminal case will be finalized and terminated after being tried by two people's courts at different levels; the people's procuratorate cannot protest against and the parties cannot appeal against the final judgment or ruling made by the people's court of second instance (Chen, 1999). This doctrine in the judicial practice of the mainland, to some extent, gives judges, particularly those hearing the cases on appeal, a highly centralized power. Under such mechanism, judges tend to choose those resources that contract the dialogic space to clarify the court's position and demonstrate the authority of court decisions. By contrast, Hong Kong, practicing three-tiered trial system (Guo & Zhu, 2000) and case law system which lays emphasis on the judicial influence and binding force of the precedent on the later cases (Fisher, 2019), makes the power of the judges hearing the case of second instance rather restricted and limited. Under the influence of the case law system, the appellate judges may act as 'law-maker' when deciding cases and thus must take full account of the judicial significance of the current judgment to the subsequent cases, as their judicial opinions may be set as legal rules which will have an impact on the judgment of future cases of the similar nature (Ghai, 2009). As a result, they strive to balance the relationship between power and equality and avoid arbitrariness to realize objectivity and rationality of their reasoning. This can be seen from their general manipulation of the dialogic space in reasoning, specifically, the resources for dialogic expansion in the reasoning of criminal judgments of Hong Kong is slightly higher than those for dialogic contraction. What's more, the citation of the precedent(s) to support the judges' decision(s) also leads to the wider distribution of Endorse than Pronounce and Concur within Proclaim-Type in the reasoning of criminal judgments of Hong Kong.
The textual structure of the reasoning of judicial judgments also plays a part in leading to their differences. The reasoning of criminal judgments of Chinese mainland follows a standard textual structure (Jiao, 2016). According to the Form of Criminal Procedure Document of The Court (1999), the typical pattern is, 'our court contends…, pursuant to (the legal basis of the judgments) …, we hereby decide as follows: …' (Zhou, 2000). Such conventional phrases, to a degree, drive up the frequency of Pronounce in the Chinese mainland dataset. By contrast, no standard model is identified in the reasoning of criminal judges of Hong Kong.

Conclusion
The present research is concerned with the contrastive study of the heteroglossia in the reasoning of the mainland of China and Hong Kong. It is found that judges of mainland and Hong Kong give preference in the choice of some subtypes of engagement resources. They employ more Disclaim than Proclaim, more Deny than Counter and more Acknowledge than Distance. The similarities in question can be interpreted from the purpose of reasoning of criminal judgments, that is, to justify their decisions of the case and persuade the putative audience to accept and even support their judgments. Differences are also found in their manipulation of engaging strategies. Judges of Chinese mainland incline to choose Contract rather than Expand resources to narrow down the dialogic space to suppress the alternative voices. On the contrary, larger dialogic space and more alternative voices are allowed for the putative audience of criminal judgments of Hong Kong. This difference can be explained from the distinct legal doctrines in the mainland and Hong Kong and the textual structure of the reasoning of criminal judgment.
Although lots of work has done on this study, it is subject to some limitations. Firstly, discrepancy is inevitable in the statistic due to cultural differences. Secondly, absolute objectivity of this study is almost impossible to attain despite that great effort has been exerted to reduce subjectivity. Since there is a lack of clear-cut boundary between different subtypes of heteroglossic engagement system. Last, a larger corpus is needed to enhance the validity of the research findings.