Plain Language in the Czech Republic, the UK, and the US: An Analysis of Tax Return Instructions
Abstract
7
Full text
Plain Language in the Czech Republic, the UK, and the US: An Analysis of Tax Return Instructions Eva Dvořáková ABSTRACT: Over the past 50 years, plain language has been an important topic in English-speaking countries, finding its way particularly into legal and official documents. Curiously, plain language has never been awidely discussed topic in the Czech Republic. It would therefore be interesting to see whether Czech tax return instructions (which are, after all, addressed to the wide public) display any plain language features, in comparison with their British and US counterparts. The present contribution aims to compare tax return instructions from the UK, US, and the Czech Republic in terms of their readability (using readability measures), sentence length, passive constructions, certain personal pronouns (you and we), nominalisations (with the -ion suffix), and lexical richness (type-token ratio). For the purposes of the analysis, the official English translation of the Czech tax return instructions has been used so that the comparisons can be made within one language. The results have confirmed the initial hypothesis that the Czech tax return instructions are not compliant with plain language rules. ABSTRAKT: Srozumitelný jazyk vČeské republice, ve Spojeném království avUSA: analýza pokynů kvyplnění daňového přiznání Vposledních padesáti letech se vanglo-americkém prostředí věnuje velká pozornost tzv.„plain language“ (srozumitelnému jazyku), ato zejména vsouvislosti správnickými aúředními dokumenty. VČeské republice toto téma doposud velký zájem nevzbudilo. Bylo by proto zajímavé zjistit, zda české pokyny kvyplnění daňového přiznání (které jsou určeny široké veřejnosti) dodržují určitá standardní pravidla pro srozumitelný jazyk formulovaná vanglicky mluvících zemích azda je jazyk těchto pokynů srovnatelný spodobnými texty ze Spojeného království aUSA. Cílem této studie je tedy porovnat pokyny kvyplnění daňového přiznání zČR, Spojeného království aUSA pokud jde ojejich čitelnost (spoužitím měřítek čitelnosti), průměrnou délku vět, použití trpného rodu, osobní zájmena (vy amy), nominalizace (anglická slova spříponou -ion) alexikální bohatost. Vzhledem kodlišnému charakteru češtiny aangličtiny byl pro český text použit jeho oficiální překlad do angličtiny, aby bylo srovnání provedeno vrámci jednoho jazyka. Výsledky analýzy potvrzují původní hypotézu, že české pokyny nejsou vsouladu spravidly pro srozumitelný jazyk (plain language). KEY WORDS / KLÍČOVÁ SLOVA plain language, plain English, readability measures, lexical richness, sentence length srozumitelný jazyk, měřítka čitelnosti, lexikální bohatost, délka vět 1. INTRODUCTION Recent decades have witnessed aconsiderable interest in plain language across the world, but in particular in English-speaking countries. The critics of traditional OPEN ACCESS
8 STUDIE ZAPLIKOVANÉ LINGVISTIKY 1/2021 unintelligible texts have focused primarily on legal and official documents, preferring ease of comprehension over traditional and often ritualistic ways of writing. Numerous authors (e.g., Mellinkoff, 1963; Tiersma, 2000; Butt, 2001) have devoted their lifetime work to promoting the plain language agenda and defending it against the conservative advocates of the traditional practices. Campaigners in various countries have exercised substantial pressure on the governmental agencies to adopt amore user-friendly approach in their communication with the public. In some cases, plain language has been incorporated through legislation, for example through the Civil Procedure Rules 1998 (the UK) or the Plain Writing Act of 2010 (the US). These efforts have helped to transform legislative guidelines, jury instructions, and administrative documents— such as the instructions for filling in atax return. In the Czech Republic, similar efforts have been very modest and plain language has never become awidely discussed topic. For instance, the Czech Legislative Drafting Rules (Legislativní pravidla vlády ČR) only mention the requirement that the laws should be drafted in clear language (Article 2 (2) (d)). But since ordinary citizens without any legal background are directly affected by legal and administrative documents, they have aright to understand. While many researchers have explored plain English in legal texts, to the best of our knowledge little is known about the Czech plain language efforts, and whether such efforts have yielded any tangible benefits, for example by making official documents widely comprehensible. This paper aims to compare the instructions for filling in atax return from three different countries and analyse them in terms of their readability. However, objective as they are, readability scores give merely an indication of the ease of comprehension. The paper will therefore examine anumber of other characteristic features of plain language texts: the length of sentences, passive constructions, nominalisations, and personal pronouns (you and we). Last but not least, lexical richness will be ana lysed to verify the hypothesis that plain language texts tend to be lexically less diverse. The paper proceeds as follows. Section 2 provides aliterature review, summarising major research activities undertaken in this area. Section 3 outlines the methodology and the details of the corpora compiled. Section 4 sets out the hypotheses, and Sections 5 and 6 present the main findings and the discussion thereof. 2. LITERATURE REVIEW Abundant guidance is now available for the drafters of official documents, typically including the following recommendations:1 — Keep your sentences short — Prefer active verbs — Use you and we — Use words that are appropriate for the reader 1 Plain English Campaign. (2020). How to Write in Plain English. Available from www.plainenglish.co.uk/how-to-write-in-plain-english.html. OPEN ACCESS
EVA DVOřáKOVá 9 — Don’t be afraid to give instructions — Avoid nominalisations — Use lists where appropriate Yet this trend has not been embraced by everybody. In his seminal work, Peter Butt mentioned ajudge who used the expression “language of the pop songs” (Butt, 2001, p.83) to refer to aplain English provision in the Australian Corporations Law. Other critics argue that it is often not the language that causes problems, but rather the complexity of legal concepts. For example, Phillips (2003, p.43) pointed out that when the expression “beyond all reasonable doubt” was replaced with “sure” in the jury instructions, the jurors were confused just the same. Yet another argument is that readers often like complexity in language because it allows them to perceive the author as competent, credible, and communicative. According to Dillard and Pfau (2002, as cited in Jarvis, 2013, p.48), “when the complexity of amessage does not exceed listeners’ ability to comprehend the message, ‘listeners prefer complexity because it is interesting, and lexical diversity should be preferred because it represents more complex lexical choiceʼ”. Lexical richness was explored by Cvrček & Chlumská (2015) in connection with the topic of “translation universals”. The authors found support for the hypothesis that “non-translated texts tend to be lexically richer” and proposed anew method for measuring lexical richness, namely zTTR (later developed into zqTTR),2 which compares the values for agiven text with referential values (taking the text size and type into consideration). The original method (TTR), “the most widely used technique to examine and compare the lexical richness of two or more texts or corpora” (ibid., 2015, p.315), is often deemed unsatisfactory due to its excessive dependence on corpus size. In the US, book publishers, educators, and the military have sought to specifically target their readers and tailor their messages to the reading skills of their audience. For that purpose, they have promoted the development of objective ways for assessing readability— the readability formulas. Although such formula scores should be considered as “rough guides” rather than “highly accurate values” (DuBay, 2004, p.19), they do offer some objective guidance. DuBay offers acomprehensive overview of readability principles and the most important measures of readability (according to his paper, there are now about 200 readability formulas). He uses George Klare’s definition of readability as “the ease of understanding or comprehension due to the style of writing” (ibid., 2004, p.3). According to DuBay, “the average adult in the U.S. reads at the 7th-grade level” and materials for the public should be written at the fifthor sixth-grade reading level (ibid., 2004, p.1). Thus, publishers and authorities consciously seek to adapt their written materials for aparticular readability score. Most readability formulas rely on the assumption that textual difficulty is caused by the complexity of sentences and the difficulty of vocabulary (ibid., 2004, p.19). Complexity of sentences is usually associated with the length of sentences. In terms 2 Now included in an online tool: https://www.korpus.cz/calc/. OPEN ACCESS
10 STUDIE ZAPLIKOVANÉ LINGVISTIKY 1/2021 of sentence length, DuBay cites L.A.Sherman (ibid., 2004, p.10), who showed how sentences became shorter over time. While in the Pre-Elizabethan times the average sentence length was 50 words per sentence, in our time it is only 20 words. Sherman further suggested that “[o]ver time, written language becomes more efficient by becoming more like spoken language” (ibid., 2004, p.11). While it is relatively easy to measure sentence length (and indeed, most formulas take this variable into account), the difficulty of vocabulary is much harder to assess. Rudolf Flesch, for example, originally toyed with the idea of measuring the use of affixes, but later concentrated on the number of syllables (ibid., 2004, p.20). Similarly, several other researchers have used sentence length and the number of syllables as the main criterion for measuring readability (e.g., the Gunning fog index). Arather different approach to vocabulary was taken by Dale and Chall (ibid., 2004, p.23), who compiled alist of 3000 easy words and focused on the “difficult” words not included in the list. This is arguably amore precise measure, given the fact that such words as “interesting” are composed of many syllables, but they are considered “easy” (and are indeed included in the Dale-Chall Word list).3 Yet even sentence length is not areliable predictor of difficulty. In their study of readability formulas applied to Czech, Šlerka & Smolík (2010) found that judging by the readability measures, Dášeňka stands out as the most difficult text of the analysed Čapek’s works.4 This is apparently due to the fact that narrative sentences tend to be long (but not necessarily difficult to understand). The argument for using sentence length in the readability formulas has to do with human memory, because requiring areader to remember along sentence places greater demands on his/her memory, thus making reading more difficult. An interesting perspective was offered by Long & Christensen (2011), who attempted to answer the question of whether the readability of an appellate brief affected the judge’s final decision. Their conclusion was negative. Readability (determined by the Flesch and Flesch-Kincaid measure) had no impact on the outcome of the cases. The analysis raised some important issues, for instance the idea that readability may not be an inherent quality of the text itself, but rather an issue connected with the reader. In other words, the readability of appellate briefs poses no apparent difficulties to “the highly educated audience of appellate judges and justices” (Long & Christensen, 2011, p.147) simply because they are the target audience. In their conclusion, the authors recommend that novice lawyers use the available Microsoft Word tools to ensure that the readability of their text falls well within the readability scores conventionally found in their profession. This is seemingly adifferent piece of advice to what plain language advocates normally recommend, namely to strive for the greatest possible ease of comprehension. Yet it could be argued that this concern about one’s audience is what both the plainers and Long & Christensen (2011) have in common: atext should fit the audience. Of course, this idea is not new. As DuBay 3 https://www.readabilityformulas.com/articles/dale-chall-readability-word-list.php 4 “Extrémem je vtomto případě Dášeňka, která vychází jako prakticky nejobtížnější text izpoužitého korpusu děl Karla Čapka.” [In this respect, the book Dášeňka is an extreme case because it turned out to be the most difficult text from the whole corpus of Čapek’s works.] (Šlerka & Smolík, 2010, p.42). OPEN ACCESS
EVA DVOřáKOVá 11 quoted L.A.Sherman: “The universally best style is not athing of form merely, but must regard the expectations of the reader as to the spirit and occasion of what is written.” (DuBay, 2004, p.11). 3. METHODOLOGY Since the purpose of this study is to inquire whether the Czech official documents follow certain plain language principles, we were faced with the dilemma of whether to use the original text written in Czech (CS/Ori) or the English translation thereof. Both approaches are likely to yield slightly distorted results. When comparing the original Czech text (CS/Ori) with the original English ones, it is necessary to take into account the different features of both languages. For example, personal pronouns are likely to occur less frequently in the Czech language because they can be expressed morphologically through inflected verbs. Likewise, for nominalisations amuch more complicated method would be necessary to make the results comparable across languages. As regards the readability formulas, the different morphological features of the different languages could also affect the results. On the other hand, when using atranslated text (CS/Tra), the analysis is undertaken within one language, but the results could be distorted because of certain translation issues (for example, the skills of the translator or the possible reduction of lexical richness which can occur in translations). After carefully considering both approaches, we decided to remain within one language (English), and use the translated text (CS/Tra), which relatively faithfully corresponds to the Czech original. The four corpora used consist of the following documents: CS/Ori POKYNY kvyplnění přiznání kdani zpříjmů fyzických osob za zdaňovací období (kalendářní rok) 2019 9 160 tokens CS/Tra INSTRUCTIONS for filling out the income-tax return by individuals for the taxable period (calendar year) 2014 (English translation) 9 603 tokens UK Tax Return notes— HM Revenue and Customs, Tax year 6April 2018 to 5April 2019 (2018–19) 11 198 tokens US IRS–1040— Instructions 2018 Tax Year 109 908 tokens Table 1:Texts used. As the work on this research started in 2019, we used the most recent documents available at that time (UK and US instructions for 2018). The Czech tax website offered the instructions for 2019. The official English translation was only available for the year 2014. Yet the different time periods probably make no difference to the outcome of the research, because these documents usually do not undergo radical changes as to the language used. Some of the results (sentence length, passive constructions, readability) were relatively easy to obtain by using the Microsoft Word option “Show readability statisOPEN ACCESS
12 STUDIE ZAPLIKOVANÉ LINGVISTIKY 1/2021 tics” in File/Options/Proofing. Others (you/we, nominalisations) had to be obtained through corpora analysed in SketchEngine. Lexical richness was explored by using Corpus Calculator.5 4. HYPOTHESES Since everybody has to pay taxes, and instructions should offer clear guidance, we chose the tax return instructions to illustrate to what extent the English-speaking countries have incorporated plain language principles into the official documents produced by government agencies, and whether asimilar trend can be observed in the Czech language. In order to determine how “plain” the language is, we focused on the following criteria (Table 2): — length of sentences — passive constructions — personal pronouns (you, we) — nominalisations (-ion) — lexical richness — readability Judging by the lay-out, one has the impression that the texts from the US and the UK are in harmony with the plain language principles. By contrast, the CS/Tra text appears to be rather dense and impenetrable. It is therefore presumed that the CS/Tra text will have longer sentences, more passive constructions, fewer personal pronouns (you, we), more nominalisations, and will be found more “difficult” in terms of readability scores. For nominalisations, it was decided to focus on the -ion suffix, which is often cited as aclassic example of nominalisation.6 When it comes to lexical richness, the “plainer” texts from the UK and the US could be less diverse than the CS/Tra text, but at the same time the translation process could have deprived the original Czech text of its lexical richness. Hypotheses sentence length passive you/we nominalisations: -ion readability lexical richness UK short few more few easy less diverse US short few more few easy less diverse CS/Tra long more few more difficult more/less diverse? Table 2:Hypotheses. 5 https://www.korpus.cz/calc/ 6 “As aresult, the trend in present-day English is to use, for deverbal nominalisations, either the appropriate variant of -ion or conversion (mainly for unaffixed verbs).” (Štekauer & Lieber, 2005, p.438). OPEN ACCESS
EVA DVOřáKOVá 13 5. RESULTS Table 3 below shows the results summarised for the three corpora. sentence length (words per sentence) passive you (ipm) we (ipm) nominalisations: -ion (ipm) readability (Flesch) UK 17.3 6.4% 49 116 5 269 17 771 fairly difficult US 12.9 9.9% 30 998 644 16 747 fairly difficult CS/Tra 33.9 25.5% 11 975 046 027 difficult Table 3:Summarised results. In terms of sentence length, the CS/Tra clearly stands out as the text with the longest sentences, considerably exceeding the recommended sentence length of 20 words. The UK and the US texts have substantially shorter sentences. As an example, we can compare the initial sentences from the respective documents: CS/Tra: “‘Instructions for filling out the income-tax return by individuals for the taxable period (calendar year) 2014ʼ no.25 5405/1 MFin 5405/1— model no.22 (hereinafter ‘Instructionsʼ) are instructions for filling out the form ‘The Income Tax Return by individuals pursuant to Act no.586/1992 Coll., on Income Tax, as amended, for the taxable period (calendar year) 2014ʼ no.25 5405 MFin 5405— model no.21 (hereinafter tax return).” (68 words) UK: “These notes will help you to fill in your paper tax return.” (12 words) US: “Form 1040 has been redesigned.” (5 words) As regards passive constructions, the analysis shows that the CS/Tra text has the highest proportion of passive constructions. Quite frequently, where the CS/Tra text uses the passive, the UK and US instructions prefer the active voice, as in the following examples: CS/Tra: “Only rows and fields with white background are to be filled out by the taxpayer (those with pink background are used by the tax administrator).” UK: “Fill in the ‘Trusts etc’ pages if you were: abeneficiary of atrust (not a‘bare’ trust) or settlement…” US: “You must fill in and attach Schedule B if the total is over $1,500 or you received, as anominee, ordinary dividends that actually belong to someone else.” With regard to the use of you (Table 4), the query [lemma="you|your|yours|yourselves|yourself "] yielded the highest number of hits in the UK text, followed by the US OPEN ACCESS
14 STUDIE ZAPLIKOVANÉ LINGVISTIKY 1/2021 text. After comparing the results for the CS/Tra text with the results for the UK and the US texts for statistical significance (Chi-squared test), it becomes clear that the difference is statistically significant. you raw frequency ipm UK 550 49 116 US 3 415 30 998 CS/Tra 115 11 975 Table 4:Results for you. For we (Table 5), the query [lemma="we|us|our|ourselves"] was used, and the results follow asimilar pattern. While the highest number of instances was found in the UK text, no instances were found in CS/Tra. The CS/Tra text apparently uses the words tax office where the UK and US documents use we, as in the following examples: CS/Tra: “To Tax Office in, at, for— fill in the official name of the relevant Tax Office (the Tax Administrator), in whose jurisdiction you permanently reside at the time of the filing.” UK: “We’ll confirm in your tax calculation if you’re aScottish taxpayer and will pay Scottish Income Tax.” US: “If you have paid too much, we will send you arefund.” we raw frequency ipm UK 59 5 269 US 71 644 CS/Tra 0 0 Table 5:Results for we. In order to compare the words most frequently associated with nominalisations, the query [lemma=".*ion"] was used. Table 6 shows that by far the greatest number of -ion nominalisations were found in the CS/Tra text. The value for CS/Tra is statistically significant with respect to the other two corpora. The most frequent nouns with the -ion suffix were as follows: CS/Tra: section, subsection, administration, pension, information UK: pension, information, contribution, deduction, remuneration US: instruction, information, deduction, distribution, section OPEN ACCESS
EVA DVOřáKOVá 15 -ion raw frequency ipm UK 199 17 771 US 1845 16 747 CS/Tra 442 46 027 Table 6:Results for nominalisations -ion. In order to assess readability, 3 different readability formulas were used (Table 7). The Flesch and Flesch-Kincaid formulas (included in Microsoft Word) analyse the text in terms of sentence length and the number of syllables in aword. The Flesch formula uses aconversion table for the scores (the lower the score, the more difficult the text), while the Flesch-Kincaid provides the number of years of schooling needed for aparticular text. The New Dale-Chall formula7 measures sentence length and the number of “hard” words that are not included in its list of 3000 familiar words. The final score indicates the (school) grade level. As it was not possible to analyse the whole texts, three random samples were created for the New Dale-Chall formula. The readability scores show that the UK and US texts are more readable than the CS/Tra text, according to all three readability formulas. Readability Flesch Flesch-Kincaid (grade) New Dale-Chall (3samples) UK 58.7 (Fairly difficult) 9.3 9/10; 11/12; 9/10 US 55 (Fairly difficult) 8.7 9/10; 7/8; 9/10 CS/Tra 37.3 (Difficult) 16.4 13/15; 11/12; 11/12 Table 7:Results for readability. Lexical richness of the texts (Table 8) was measured using the Corpus Calculator.8 After entering the data (types and tokens) into the tool, the programme creates abell curve (normal distribution) with the mean representing the expected value for this type of corpus, and the vertical line (the shorter line) indicating the observed value. The selected referential corpus was “written— non-fiction, InterCorp v2011”.9 The results show that the US text is the least diverse, while the UK text is the closest to the mean and thus the richest. Nonetheless, all three texts are lexically less diverse than expected. Here it is interesting to subject the CS/Ori text to the same analysis. The results show that this text is the most diverse of all four, with the observed value almost corresponding to the mean. This seems to confirm the findings from the literature that translation leads to adecrease in lexical diversity. At the same time, the 7 https://readabilityformulas.com/new-dale-chall-readability-formula.php. 8 https://www.korpus.cz/calc/ 9 Klégr, A., Kubánek, M., Malá, M., Rohrauer, L., Šaldová, P., & Vavřín, M. (2018). Korpus InterCorp— angličtina, verze 11 z19.10.2018. Ústav Českého národního korpusu FF UK, Praha. Available from http://www.korpus.cz. OPEN ACCESS