Introduction

Reading interest plays a crucial role in motivating students to engage with expository texts for learning (Ainley, 2017; Liebfreund, 2021; Renninger, Hidi, 2019; Renninger et al., 2023; Schiefele, 2009; Schiefele, Löweke, 2017; Schraw, Lehman, 2001; Silvia, 2006). It motivates readers to spend more time on the text, put more cognitive efforts into processing, and use effective reading strategies (Clinton-Lisell, 2022; Duke et al., 2011; Fulmer et al., 2015; Springer et al., 2017).

Several factors that induce reading interest have been identified over the years. They include novelty, coherence, concreteness, emotiveness, vividness (i.e., imagery), simple vocabulary, prior knowledge, ease of comprehension, personal interests (as stable personal characteristics), and engaging themes (e.g., death, war, sex, etc.) (Ainley, 2017: 7–8; Schiefele, 2009: 199; Wade, 2001; see also the “long laundry list” in Silvia, 2006: 78). Silvia’s (2006) appraisal-based model suggests that reading interest stems from the appraisals of novelty-complexity and reader’s ability to understand the text. Despite this theoretical progress, empirical studies on the specific linguistic features that incite interest are limited.

Previous studies examined the contribution of linguistic units towards interest. They focused on concrete words, notably nouns (Mikk, Kukemelk, 2010; Sadoski, Paivio, 2013); but other units were only mentioned (e.g., “imagery and descriptive language”, “simple vocabulary”, “personal words”, “appealing words”; Mikk, 2000: 257–266; Silvia, 2006: 78; Wade, 2001). Some works addressed the impact of genre-specific features related to content and text complexity (Friedrich, Heise, 2022; Golke, Wittwer, 2024; Lepper et al., 2021; Liebfreund, 2021; Schiefele, Löweke, 2017; Shulman et al., 2020; van der Sluis et al., 2014). A number of studies identified discourse structures that provoke reader’s interest: e. g., seductive details, problem solving, attribution, contextualisation, concrete elaboration, positive ratings, figurative representation, and so on (Bermejo-Berros et al., 2022; Choi, 2006; Hidi, Baird, 1988; Hoeken, van Vliet, 2000; Kasper et al., 2018; Mikk, 2000: 247–256; Phan, Tin, 2022; Pinoliad, 2021; Renninger et al., 2019; Renninger et al., 2023; Shin et al., 2016; Wade, 2001). However, a consistent list of such structures has not been designed.

These studies have important practical implications, because they outlined general writing techniques to make text more interesting. But they draw on limited experimental reading materials which often neglect the role of specific text structures in constructing reading interest (see Piotrovskaya, Trushchelev 2022). That is why the interest-evoking potential of different text structures remains unclear.

The current study extends previous research by examining three discourse expository strategies – contextualization, problem solving, and concrete elaboration. By exploring their contribution to the interestingness of expository texts, the study seeks to experimentally assess an association between these strategies and reading interest. The primary focus is on linguistic structures embedded within these strategies. To investigate such structures, the study employs linguapragmatic methods of text analysis and gives a linguapragmatic description of experimental materials. The text materials for the study were drawn from the most widely used (Russian) school textbooks on Biology (n = 9), Geography (n = 6), History (n = 8), Physics (n = 9), the Russian Language (n = 8), and Social Sciences (n = 8). All the textbooks were written for comprehensive school students in grades 7–9 and issued in the last 10 years. The study analysed only expository texts (textbook sections); task texts were not extracted from the textbooks. The total size of the expository texts exceeded 2 million tokens.

Methodology

Interest-evoking strategies

Linguapragmatics investigates how linguistic resources realise strategies to achieve communicative goals, particularly those aimed at provoking readers’ feelings (e.g., Piotrovskaya and Trushchelev, 2021a; Scott, 2021; Wharton et al., 2021). In this sense, the text can be analysed as a system in which linguistic units embody interest-evoking strategies.

Utilising linguapragmatic methods, Piotrovskaya and Trushchelev (2021a; 2022) have identified three ways of using linguistic units to craft interest-evoking strategies: expression of dialogicity (using units that encode participants’ positions and communicative actions), concretisation (using units that encode concrete and perceptual things and actions), and manifestation of emotions (using units that express or describe emotional states). Building on the findings about the use of linguistic units, the current study inspects three expository strategies: contextualization, problem solving, and concrete elaboration. This focus is motivated by two factors: (1) these strategies are frequently mentioned by psychologists, and (2) these strategies are often used to popularise knowledge (e.g., see studies on various expository genres: Lewin et al. 2001; Meyer and Rey, 2011; Parodi, 2014; van Dijk and Atienza, 2011; Clinton and Walkington, 2019).

The following sections offer a linguopragmatic description of the strategies employed in school textbooks.

Contextualisation

Contextualisation entails the use of linguistic signs to express, actualise, and make relevant various aspects of communicative settings (Fetzer, 2021). For Leckie-Tarry (1991), contextualisation refers to the degree to which a text “is embedded in the activities immediately surrounding it” (111). From a psychological perspective, contextualisation is associated with personally involving and culturally relevant information, which can evoke interest (Clinton, Walkington, 2019; Hidi, Baird, 1988; Renninger et al., 2019; Renninger, Hidi, 2022; Shin et al., 2016; Wade, 2001).

The main way of contextualisation is to engage with readers and manage reading processes, by using dialogicity units that encode participants’ positions and communicative actions:

(1) Присмотримся ещё раз к характеру и формам взаимоотношений в группах. Рассмотрим в качестве примера коллектив промышленного предприятия. [‘Let us look again at the nature and forms of relationships in groups. Take the collective of an industrial enterprise as an example.’]

(2) Каждый из нас не только обладает правами, но и обязан соблюдать права других людей. Наверное, тысогласишься с тем, что основа соблюдения прав другого человека — внимание к его потребностям, понимание его интересов. Следовательно, твоя свобода заключается в возможности делать всё, что не приносит вреда другому. [‘Each of us has not only rights, but also a responsibility to respect the rights of other people. You would probably agree that the basis of respect for the rights of another person is paying attention to his/her needs and understanding his/her interests. Consequently, your freedom consists in being able to do everything that does not harm another person.’]

(3) 21 февраля 1613 года собор избрал на царство шестнадцатилетнего Михаила Фёдоровича Романова <…> Почему же выбор пал именно на Михаила? [‘On February 21, 1613, the Zemsky Sobor elected the sixteen-year-old Mikhail F. Romanov to the throne <...> Why did the choice fall on Mikhail?’]

The first passage includes personal verb forms (e.g., присмотримся [let us look-PRS.1PL]), imperatives (e.g., рассмотрим [take]), and cognitive and speech verbs. The second passage features personal pronouns (e.g., нас [us]), a personal verb form (e.g., согласишься [agree-PRS.2SG]), and modality markers (e.g., наверное [probably]). The third passage includes interrogatives.

Dialogicity units also encompass evaluative and emotive markers, progressive tense markers, vocatives, and colloquial language units (see Bondi, 2018; Hyland, 2014; Makkonen-Craig, 2014; Qin, Uccelli, 2019). These units can construct dialogic patterns, including pseudo-dialogue patterns:

 (4) А теперь внимание! Обладатели суперпамяти — два шага вперёд! Вопрос по материалу прошлого года. Как мы называем средний многолетний тип погоды, характерный для данной местности? Правильно, это климат. [‘And now, attention! Supermemory owners, [take] two steps forward! [There is] a question on last year’s material. What do we call the average perennial type of weather characteristic of a given area? [That’s] right, it’s a climate.’][1]

Contextualisation can also represent culturally/personally relevant situations the reader is (potentially) involved in. This is realised by units that give reference to the reader and to personally relevant entities. Consider the following passage:

(5) С проявлениями экономики ты встречаешься ежедневно: слышишь разговорыдома и на улице о ценах на товары, узнаёшь о размерах зарплатыродителей, читаешь в газете о налогах, участвуешь в ремонтешкольной мебели, покупаешь в магазине продукты. [‘Youencounter economics every day: youhearconversations at home and in the street about the price of goods, you learn about your parents' salaries, you read about taxes in the newspaper, you participate in the repairof school furniture, you buygoods in a shop.’]

To refer to the reader, the passage employs dialogicity units: personal pronouns (ты [you]) and personal verb form (e.g., слышишь [hear-PRS.2SG]). While these units indicate the reader, they do not manage a direct interaction between the authors and the potential reader. The passage also employs concrete words given in bold to refer to components of the reader’s every-day life and actualise the personally relevant context.

Problem solving

Problem solving is a specific way of contextualisation, because dialogicity units usually realise this strategy (Piotrovskaya and Trushchelev, 2021b, in Russian). Psychologists describe this strategy as “text manipulation” that induces “a need on the reader’s part to resolve some incomplete understanding of new information” (Hidi, Baird, 1988: 470; see also Markey, Loewenstein, 2014; Mikk, 2000: 247). The linguapragmatic studies split the problem-solving process into discursive steps: (1) a fact presentation, (2) a problem statement, (3) a problem solution (Makkonen-Craig, 2014; Piotrovskaya and Trushchelev, 2021b, in Russian).

Consider three passages, which are presented step-by-step below by the cases.

The first step conceptualises a background of the problem — a relevant fact the reader can be familiar with. As a rule, this step is realised by an assertion:

(6a) Когда человек идёт по рыхлому снегу в сапогах, валенках или ботинках, он проваливается. [‘When a person walks on loose snow in high, felt or field boots, s/he falls through.’]

(7a) Вам хорошо известно, что основным источником тепла на Земле является Солнце. [‘You are well aware that the main source of heat for Earth is the Sun.’]

The second step elaborates the text; the background is subsequently followed by a foreground that conceptualises a problem associated with knowledge gap. In most cases, this function is performed by expository questions — “questions whose answers the speaker regards as relevant to the hearer” (Sperber, Wilson, 1998: 252):

(6b) Почему же на лыжах можно идти по снегу, не проваливаясь? [‘Why is it possible to ski walk on the snow without falling through?’]

Expository questions provide grounding for the problem-solving process by stating students’ information gap and expressing their desire to learn something novel things. In addition, authors might use complementary linguistic resources, such as conditional or concessive meanings, to challenge students’ knowledge and model a contradiction:

(7b) Каким же образом передаётся тепло от Солнца? Ведь Земля находится от него на расстоянии 15 ∙ 10⁷ км.[‘How is heat transferred from the sun? After all, the Earth is located at a distance of 15 ∙ 10⁷ km from it.’]

There are other contextualisation units intended to set the problem:

(8a) Человек толкает тележку с силой F = 40 Н. При этом тележка движется со скоростью v = 0,5 м/с. [‘A man is pushing a cart with the Force (F) of 40N. The cart is moving with the Speed (v) of 0.5 m/s.’]

(8b) Можно ли по этому описанию ситуации найти развиваемую человеком мощность? На первый взгляд, нельзя: ведь неизвестны ни совершённая человеком работа, ни время, в течение которого она была совершена. [‘Is it possible to find the power developed by a man using these settings?At the first glance it seems impossible: neither the Work done by a man is known nor the Time during which the work was done.’]

In this case, the (a) sentence conceptualises a background of the problem. To set a problem and model a contradiction, the (b) part employs a guess question — a question for which the speaker knows the answer and the hearer could make a guess (see Wilson, Sperber, 2012: 222) — as well as modality units that represent the most likely way of students’ thinking.

The third step presents a solution (in fact, the answer to a question). In textbooks, it, for the most part, just conveys the piece of true knowledge:

(6c) Когда человек становится на лыжи, значительно уменьшается сила давления, приходящаяся на единицу площади соприкосновения со снегом. [‘When a person gets on a ski, the pressure force per unit area in contact with the snow is significantly reduced.’]

Expository texts may employ dialogicity units to contextualise a solution process by presenting interaction between authors and potential readers:

(7c) Как известно, в вакууме перенос энергии путём теплопроводности невозможен. Не может происходить и за счёт конвекции. Следовательно, существует ещё один вид теплопередачи. Изучимэтотвидтеплопередачиспомощью опыта. [‘It is well known that energy can be transferred neither by heat conduction nor by convection in a vacuum. Consequently, there is another type of heat transfer. Let’s examine this type of heat transfer with an experiment.’][2]

(8c) Но не будем сдаваться и введём время сами. [‘But let’s not give up and bring in the Time by ourselves.’]

Concrete elaboration

Concrete elaboration brings about detalisation of propositional contents (see Bermejo-Berros et al., 2022; Choi, 2006; Hidi, Baird, 1988; Mikk, 2000: 247; Wade, 2001). First of all, it is performed by means of concrete language units, which refer to physical things, observable qualities and literal actions (see Lievers et al., 2021). The following passage is a case in point:

(9) Любил парнишкапострелять из рогатки. Сначала целился в банку, потом в птичку, а затем в человека — пошутить хотел. Выстрелилкамешком — и попалслучайно в глаз. [‘A young guy likes to shoot with a slingshot. At first, he aimed at a can, then at a bird, and then at a person, just for fun. He shota little stone — and accidentally hitthe eye.’]

The strategy can be realised by perspectivation units, which express perceptual and mental meaning (see Graumann, Kallmeyer, 2002):

(10) Сегодня с помощью телескопа можно наблюдать только следы этих взрывов: они видны как гигантские облака. [‘Today, only the traces of these explosions can be observed with a telescope: they are seen as giant clouds.’]

Referring to reader’s perceptual and mental actions, perspectivation units can provide contextualisation:

(11) Вы, очевидно, наблюдали летом в лужах на сырой дороге, в придорожных колеях или прудах, а при сильном освещении и в аквариумах цветение воды. [‘You have obviously watched water blooms in puddles on wet roads, in roadside ruts or in ponds during the summertime, as well as in aquariums under strong light.’]

Also, there are complementary units for concrete elaboration: figurative language, emotive words, tense–aspect–mood markers, modifiers, markers of actual countability, and locative or temporal deictic markers (Lievers et al., 2021; Piotrovskaya, Trushchelev, 2022).

In general, concrete elaboration provides a reference to concrete things, attributes, actions, characters, personal positions, feelings, locations, and temporary parameters. For example, by employing concrete elaboration, Geography textbooks might represent a localised situation of visual perception, History textbooks might construct narrative tension, and Physics textbooks might give real-life analogies (see Piotrovskaya, Trushchelev, 2022; Trushchelev, 2022).

Experimental study design

The experiment assessed the contribution of the interest-evoking strategies to the text interestingness, which was treated as a text-based variable. Three sets of text passages were prepared:

(1) four passages (M_size = 181 tokens; SD = 26) were drawn from Social Science textbooks; they employed contextualisation,

(2) four passages (M_size = 167 tokens; SD = 26) were drawn from Physics textbooks; they employed problem solving,

(3) four passages (M_size = 210 tokens; SD = 20) were drawn from History textbooks; they employed concrete elaboration.

Obviously, set 1 was designed to examine contextualization, set 2 — to examine problem solving, and set 3 — to examine concrete elaboration.

The choice of a school subject depended on linguapragmatic factors.

First of all, the passages within a set differed in the number of interest-evoking units discussed above. Briefly, within each set, Passage 1 employed no units; Passage 2 employed 1–2 unit(s); Passage 3 employed 3–4 units; Passage 4 employed more than 4 units. Table 1 gives a detailed description of the strategies.

Within a set, the passages covered similar topics (see Table 1). Set 1 (Social Science textbooks) provided information about illegal actions; set 2 (Physics textbooks) contained information about molecules; and set 3 (History textbooks) described Russian riots. The uniformity of topics limits their effect on participants’ reading interest.

The passages include only the most common terms specific to a school subject. It was crucial to select texts that participants could understand without extra reference materials.

A total of 386 eighth-grade students from 16 classes in 6 secondary comprehensive schools participated in the study. All schools were located in St. Petersburg, Russia. Experiment sessions were conducted by the author in the schools during class hours. The participation was determined by teacher approval and students’ own willingness. Students’ average age was 14 years, and 183 (47.4%) of them were female.

The students were randomly divided into three groups: (1) 127 students (61 of them were female) read the passages in Social Science (set 1), (2) 124 students (56 of them were female) read the passages in Physics (set 2), (3) 135 students (66 of them were female) read the passages in History (set 3). Thus, three independent samples were formed. Dividing into groups did not correspond to classes and schools. All the passage sets were evenly and randomly distributed among students. Moreover, each participant received four passages in a random order.

Working with a complete passage set, each participant experienced the impact of only one interest-evoking strategy. This allowed for a more precise measurement of the linguistic contribution to text interestingness (the previous studies showed that participants tend to rate texts in relation to each other; Piotrovskaya, Trushchelev, 2022; Trushchelev, 2023).

The study used Likert-type scales to indirectly measure text-based variables, text interestingness and its key predictors. The predictors included stable individual interest in a school subject and subjective view of text characteristics: novelty, complexity, comprehensibility, and originality. Such scales have been validated for psycholinguistic studies (see Friedrich, Heise, 2022; List et al., 2018; Piotrovskaya, Trushchelev, 2022; Sadoski, Paivio, 2013; Sorokin, 1985; Trushchelev, 2023).

The study employed paper materials. The participants’ responses were anonymous, and there was no time constraint during the experiment.

Before reading, participants rated their individual interest in the school subject. This was done to assess the dependence of reading interest on text-based factors rather than personal ones (see Renninger et al., 2019). The instruction included a request to indicate the level of participant’s interest in the subject. To rate the level of interest, a table of seven emoticons, shown in Figure 1, was used. By measuring affective preferences, such emoji-anchored scale captures the level of individual interest (see, e.g., Phan et al., 2019).

Figure 1. The scale for rating individual interest

After reading each passage, participants expressed their attitude towards it using a 7-point scale, with 4 being ‘neutral’. Reading situational interest was assessed using the scale uninteresting (1) — interesting (7). Predictors of interest were assessed using the following scales: familiar (1) — novel (7), easy (1) — complex (7), incomprehensible (1) — comprehensible (7), ordinary (1) — original (7). These scales represented text-based variables, which were treated as the subjective ratings of text-based features. The text-based variables were labelled ‘interestingness’, ‘novelty’, ‘complexity’, ‘comprehensibility’, and ‘originality’.

Thus, the data to be analysed was purely quantitative. The descriptive statistics of the data are given in the following section in Table 3.

The data analysis was carried out by using Statistical Package for Social Sciences (SPSS) and PSYCH package for R. Rating reliability was assessed first. Reliability coefficients McDonald’s omega for the quality ratings taken from separate samples ranged from (.78) to (.93). The distribution of some ratings did not follow a normal distribution (the Kolmogorov–Smirnov test). In further analysis, the following non-parametric tests were used: the Wilcoxon matched pairs signed rank test, the Friedman test, Spearman’s rank correlation coefficient, the ordinal logistic regression, Pearson’s chi-squared test (χ²). The level of significance was set at .05.

Results

Reading interest

According to the descriptive statistics, the central tendency of interestingness appears as median and mean. Their values are given in Table 2 (see also Table 3).

Note. M = mean; Med = median; W = the Wilcoxon matched pairs signed rank test (comparing with the previous passage); neg/pos = negative/positive ranks are greater; * = statistical significance (p < .05).

As the central values revealed, in each of the samples, passages appear to differ from each other in their level of interestingness. To estimate the significance of variation across ratings, dependent distributions of interestingness were compared within themselves, using the Friedman test for homogeneity. The test values were 88.39 (p < .001) for the Sample 1, 132.86 (p < .001) for the Sample 2, 50.96 (p < .001) for the Sample 3. Hence, the variation in each sample was not random: it was contingent upon a regulated factor, i.e., a change in the text stimuli.

To establish significant differences between dependent distributions and to identify trends towards an increase in interestingness, the Wilcoxon matched pairs signed rank test was applied. The test values are presented in Table 2 above. Generally speaking, they supported the central tendency; only one non-significant value (p > 0.05) was found in each of the samples: there were no differences between Passage 1 and Passage 2 in Sample 1, Passage 2 and Passage 3 in Sample 2, Passage 2 and Passage 4 in Sample 3 (W = -1.43; p = .152). The average ranges between the mean values, significantly different from each other, were: (.73) for the Sample 1, (.97) for the Sample 2, and (.52) for Sample 3. As a result, the passages can be ordered with regard to interestingness, as Figure 2 illustrates.

Figure 2. The increasing trends in interestingness

It appears from Figure 2 that the increasing trends can be represented as follows: (1) contextualisation: Passage 1 & Passage 2 → Passage 3 → Passage 4, (2) problem solving: Passage 1 → Passage 2 & Passage 3 → Passage 4, (3) concrete elaboration: Passage 1 → Passage 2 & Passage 4 → Passage 3.

Reading interest and predictors

To establish the dependence of interestingness on the predictors — individual interest, novelty, complexity, comprehensibility, and originality — Spearman’s rank correlation coefficient was applied. For each passage, it assessed the relationships between the interestingness rating and each of the predictor ratings individually. The full statistics is presented in Table 3.

Note. M = mean; SD = standard deviation; r_s = Spearman’s rank correlation (with interestingness); * = statistical significance (p < .05).

The correlation models include significant values, which can mark the contribution of predictors to interestingness. The models contain patterns of recurring significant values: individual interest and originality for Sample 1; individual interest, comprehensibility, and originality for Sample 2; individual interest, complexity, and originality for Sample 3. However, there are no patterns of significant values which would be specific to the passages according to the trends in interestingness. In addition, in each sample, there are no great differences between the correlation values for passages; and none of them show strong correlation (i.e., exceeds the value of .50).

The correlation models show only the strength and direction of the relationship between the variables, but they do not provide a measure of the unique contribution of each predictor and predictors’ shared impact. Ordinal logistic regression was used for this purpose. A regression model was built for each of the passages. It estimated the relationship between interestingness as a target (dependent) variable and individual interest, novelty, complexity, comprehensibility, and originality as a set of predictor variables (covariates). The regression models are represented in Table 4 below. For each model, the χ² goodness-of-fit tests and test of parallel lines showed that the data was suitable for analysis (p > .05).

Note. R²_pseudo= pseudo R²-value (overall variance explained): R²₁ = Cox & Snell R², R²₂ = Nagelkerke R²; MFI = model fitting information; Thresholds = the boundaries between points of the target variable: 1 = (1) vs. (2), 2 = (2) vs. (3), 3 = (3) vs. (4), 4 = (4) vs. (5), 5 = (5) vs. (6), 6 = (6) vs. (7); p₁ – p₅= predictors (coefficients): p₁– individual interest, p₂ – novelty, p₃– complexity, p₄– comprehensibility, p₅– originality; * = statistical significance (p < .05).

Each of the regression models includes a significant overall variance explained (pseudo R²-values), thresholds that indicate significant boundaries between points of the interestingness scale, and a number of weighting predictors (p_n). It might seem that none of the regression models satisfactorily elucidates the variation in the target variable, since none of the pseudo R²-values exceeds the value of .40. However, it is important to take into account the field of research. In applied linguistics, “…R2 values in the realm of .20 (or below) and .50 (and above) might be considered as indicative of generally small and large, respectively, in terms of the percent of explained variance they represent” (Plonsky, Ghanbar, 2018: 728). That is, the models are suitable for explaining as least trends in constructing reading interest.

Discussion

The results suggest a positive impact of the strategies on students’ reading interest. For each sample, the variation across interestingness hinged on a change in the text stimuli. The thresholds derived from the regressions demonstrate that the increasing trends in interestingness are to some extent correlated with significant boundaries between ratings: the higher the interestingness of a passage, the less significant the boundaries between the low points of the interestingness scale.

The central values — mean and median — do not conceptualise any passage as uninteresting (see Table 2 and Table 3). So, the medians for the least interesting passages (Passage 1 in each of the samples) were all 4. A distribution of the three rating groups [1, 2, 3 (uninteresting)], [4 (neutral)] and [5, 6, 7 (interesting)] per the least interesting passage does not deviate significantly from a uniform distribution (χ² = 2.3, 1.7, 5.6; p > .05). The independent distributions of interestingness ratings per the least interesting passage do not differ (χ² = 2.41, 5.91, 7.17; p > .05). In addition, the thresholds set significant boundaries predominantly between interestingness ratings above 4 (neutral).

These outcomes allow us to make two suggestions. First, disinterest (perhaps, boredom) is a separate dimension of the reading experience, and it may stem from different discourse-based sources than those that generate interest. Second, experiencing (dis)interest is not a part of students’ genre expectations about textbooks. Simply put, students do not anticipate whether a school textbook will be interesting or not. That is why the different samples conceptualised the neutral point of 4 as a basis for rating interestingness and rated the passages without interest-evoking characteristics (Passage 1 in each of the samples) uniformly and identically. The suggestions illuminate the interest-evoking potential of the strategies: they provide such discourse-based sources that can provoke a specific dimension of reading experience — reading interest.

The increasing trends in interestingness (see Fig. 2) suggest that the quantitative characteristics of the strategies — a number and variety of linguistic units — benefit text-based interest (see also Piotrovskaya, Trushchelev, 2022: 69). These results seem to indicate the discursive influence of salience, which determines the intensity of interest (Markey, Loewenstein, 2014). Let us make some assumptions in this regard. Sample 1, which concerned contextualisation, rated Passage 1 and Passage 2 identically (W = ‑.56). Hence, the rare inclusion of contextualisation units, such as hortative verbs or personal pronouns (see Table 1), does not increase the salience of contextualisation. Sample 2, which concerned problem solving, rated Passage 2 and Passage 3 identically (W = ‑.40). These passages set the problem, but did not provide further contextualisation and did not present solution process (see Table 1). Hence, the salience of problem solving could be increased by expressing the discursive steps. These outcomes provide additional evidence for dividing the problem-solving process into two discursive steps. Sample 3, which concerned concrete elaboration, rated Passage 2 and Passage 4 identically (W = ‑1.43), and rated Passage 3 higher than Passage 4 (W = -2.48). This trend reveals that the salience of interest-evoking units does not guarantee an increase in reading interest. It is possible that concrete elaboration sets up such detalisation of Passage 4 that this expository text takes on narrative characteristics (see Table 1). Such a salient genre transformation could be irrelevant to learning contexts, thus reducing reading interest.

Employing interest-evoking units correlated with the explanatory power of a regression model. All the regression models, which were built for the passages employing interest-evoking strategies (Passages 2, 3, and 4 in each of the samples), explain more variation in interestingness.

Although the comparing models do not fully explain the variability in interestingness (see Table 3 and Table 4), they do point to two predictors that have a stable significant impact, individual interest and originality.

Individual interest has the greatest regression weight (M = .335) and correlation (M = .29) with interestingness. The explanatory power of individual interest does not depend on interest-evoking units. Thus, the person-based independent factor appears to be the most prominent predictor of participants’ interest. More specific dimensions of individual interests — such as the particular area of subject-specific knowledge or the text topic — can have a greater impact on interest. If so, the explanatory power of individual interest could be greater.

Originality has a slightly lower regression weight and correlation with interestingness. Its impact reveals that reading interest is linked to the reader’s experience and expectations about text characteristics. This finding supports the notion that interest is triggered by a violation of expectations (see Markey, Loewenstein, 2014). In contrast to individual interest, the explanatory power of originality tends to increase along with the number of interest-evoking units (as with the trends in interestingness). The stable significant impact of originality indicates that interest-evoking strategies increase reading interest by presenting knowledge in unusual and unexpected ways.

Novelty, which is considered a driver of interest, made the least significant contribution to interestingness (M_regr = 0.101; M_correl = .09). These outcomes support the assumption that textbook content is expectedly novel for students, and, therefore, the contribution of novelty to text-based interest is reduced (Piotrovskaya, Trushchelev, 2022: 69).

Comprehensibility and complexity have an impact on interestingness in many of the models. Specifically, 10 of the correlation models indicate at least a weak correlation, and 8 of the regression models assign weights comparable to those of individual interest and originality. Moreover, these variables significantly increase the explanatory power of the regression models (according to Plonsky, Ghanbar, 2018). It seems that their contribution to participants’ interest is not universal. Rather, their contribution depends on the strategies: as shown by the largest regression weights, comprehensibility is related to problem solving, and complexity — to concrete elaboration. Furthermore, comprehensibility has the largest correlation with interestingness for Sample 2 (M = .305); and complexity — for Sample 3 (M = .375).

It appears that text-based predictors do not capture reading interest. So, originality, even when combined with comprehensibility and complexity, is not sufficient to explain the variability in interestingness. The full appraisal model of interest, which includes appraisals of novelty-complexity and the ability to understand (see Silvia, 2006), is reflected in the correlation values for Passage 4 of Sample 2 and Passage 2 of Sample 3. But the values indicate only a minor impact (see Table 3). Two regression models, specifically for Passage 2 of Sample 2 and Passage 3 of Sample 3, incorporate weighting predictors that partially represent the appraisal model. But they do not dramatically enhance the explanatory power. Hence, participants’ ratings of text characteristics were not a comprehensive source of interestingness. Potential text-based triggers might be associated with other factors. In particular, participants’ interest might be influenced by personal relevance (see Connelly, 2011; Clinton, Walkington, 2019). This is buttressed by the average ranges between the interestingness means. The shortest significant ranges were found in Sample 3. This sample did not deal with contextualisation means, which introduce personally involving content. In contrast, the significant ranges for other samples, which dealt with a vast class of contextualisation units, are larger.

The role of individual interest points to a crucial position of person-based predictors in the model of reading interest. Such predictors may be related to contextual factors, such as work mode, specific topic interests, and learning goals (see Markey, Loewenstein, 2014). Person-based predictors may provide more significant triggers to induce reading interest. At the same time, the text-based impact established in this study makes it clear that the interest-evoking strategies can influence reading interest. So, they can be used at least to maintain interest previously caused.

Conclusion

The present article augments current research into reading interest by delivering findings on the strategies for increasing students’ arousing interest in an expository text. Systematically varying textbook passages with respect to three strategies — contextualisation, problem solving, and concrete elaboration — has resulted in text-specific findings on students’ interest. The results demonstrate that these strategies directly influence students’ reading interest. The interest-evoking potential of the linguistic structures embedded within these strategies has been assessed. Originality in linguistic presentation emerges as a consistent factor in fostering interest, surpassing the effects of content novelty, comprehensibility, and complexity. However, the study also reveals that text structures alone are insufficient to trigger all subjective appraisals (predictors) that lead to interest. Thus, while the interest-evoking strategies can effectively sustain students’ interest, they do not fully account for all factors contributing to it. The findings further suggest that students’ perceptions of a textbook do not include an anticipation of its level of interest; disinterest, as a separate dimension of the reading experience, may arise from discourse-based sources distinct from those that generate interest.