The original scale – Reading Anxiety Scale (Zbornik & Wallbrown, 1991) –, consists of a 45-item questionnaire (in a Likert 5-point scale) that evaluates three domains of anxiety for reading: “fear of curiosity”, “aggression” and “independence”. The original scale was developed for fourth, fifth and sixth grade students (Elementary School, referring to the North American schooling system) – from 9 to 12 years of age, L1. Coefficient alpha reliability estimates for the English language version of the RAS were obtained for the 45 items of the RAS and each 15-item subdomain. The alpha coefficient for the entire scale was 0.88. Alpha coefficients for the RAS sub-domains were 0.72 for fear of curiosity, 0.75 for fear of aggression and 0.78 for fear of independence. A 0.88 alpha indicates that the RAS items are highly intercorrelated (Zbornik 1988; Zbornik & Wallbrown 1991).

Three samples were included in this study. The first sample consisted of 20 typical developed children from nine to twelve years old (five children from each age: 9, 10, 11 and 12 years old), students from public and private schools ofPorto Alegre, Brazil. The second sample consisted of 12 children with learning disabilities attending a school-clinic inBelo Horizonte, Brazil, who exhibited learning disabilities. Those children were 9 to 15 years old (M = 11.75 ;SD = 2.45), 75% (N = 9), female, attending the 3^rd to 9^th grade in public (N = 8) and private (N = 4) schools.

Finally, the third sample included 88 students, attending 2^nd to 4^th grade, 8 to 10 years old (M = 8.9;SD = 0.72), 57.3% female, attending public (N = 60) and private (N = 28) schools from two cities inRio Grande do Sul, Brazil. All participants were Brazilian Portuguese native speakers. Exclusion criteria (just for the third sample) were: neurological or psychiatric disorders; uncorrected hearing or visual impairments (reported by parents/guardians) and performance below the 25^th percentile in the Raven’s Colored Progressive Matrices test - Special Raven’s Scale. (Angelini, Alves, Custódio, Duarte, & Duarte, 1999). Children were assessed after their parents or guardians signed the Informed Consent Form.

This study was approved by the Ethics Committee ofUniversidade Federal do Rio Grande do Sul (UFRGS) Institute of Psychology, under protocol 1.632.099/2016. For the three samples, the schools were selected by convenience sampling and the students from each grade were randomly selected. Data collection took place in one individual section for the first sample, and two collective sessions for the second and third samples, with no re-testing or longitudinal follow-up. Children from the third sample responded to Raven’s Colored Progressive Matrices test - Special Raven’s Scale (Angelini et al., 1999) in the first collective sessions, which lasted approximately 20min. In another 45-min section, children were evaluated using the Multidimensional Anxiety Scale for Children and the Reading Anxiety Scale.

First, non-verbal intelligence was assessed by Raven’s Colored Progressive Matrices Test - Special Raven’s Scale (Angelini et al., 1999). Then, general anxiety was evaluated by the Multidimensional Anxiety Scale for Children (MASC) (March, Parker, Sullivan, Stallings, & Conners, 1997; March, Sulivan, & Parker, 1999). MASC evaluates four main factors: physical symptoms, social anxiety, avoidance of harm and separation anxiety. It consists of 39 items and its overall score approaches the diagnosis of generalized anxiety disorder. Children were instructed to respond based on how they had been feeling during the past two weeks. The items are rated on a 4-point scale, ranging from zero (it is never true about me) to three (often true about me). The MASC was validated for the Brazilian population by Asbahr (2004). Its total score was calculated by summing up all the items and each subscale was calculated summing up its relevant items. Afterwards, math anxiety was evaluated using Math Anxiety Questionnaire (MAQ). The MAQ was developed by Thomas and Dowker (2000) and the Brazilian Portuguese version of the Questionnaire was developed and standardized by Wood et al. (2012). The Brazilian version of the MAQ contains 24 items that can be answered by children individually or in groups within 5 to 10 minutes. For the total score, scores of all items of the scale were added up. Each subscale (Affective and Cognitive) was analyzed by adding up its relevant items.

A subsample of the 88 children (N = 52) was evaluated on reading skills. The ability to read words was tested using the Isolated Words Reading (Leitura de Palavras Isoladas [LPI]) (Salles, Piccolo, & Mina, 2017). In the LPI testing, participants were instructed to read aloud 60 stimuli (40 words and 20 pseudowords). The Reading Words Fluency task (Tarefa de Fluência de Leitura [TFL]) (Justi & Roazzi, 2012) consists of 55 regular words, which the participant is instructed to read 30 seconds. The score on this task was calculated by adding up the number of words read correctly during the given period. Finally, textual reading fluency (Basso, Piccolo, Corso, Mina, & Salles, 2018) was tested using AFLeT, a test in which the student is instructed to read a text at his pace, without stopping or interrupting the reading. Subsequently, the number of words in the text, the reading time, the words read correctly, and the reading speed were reviewed using a composite score.

Finally, reading anxiety was evaluated using the Reading Anxiety Scale (RAS) (Zbornik, 1988; Zbornik & Wallbrown, 1991) – short version for Brazilian Portuguese. The examiner first explained to the children that the RAS is a test to investigate how children feel about reading, but that there were no right or wrong answers. Next, the examiner read the statements to the children who answered how they agreed with the statement by circling (strongly agree, agree, not sure, disagree, strongly disagree) an alternative in the sheet provided.

Two main analyses were performed for this study, to meet the following objectives: (1) to reduce the number of items of the original RAS, and (2) to obtain evidence of the validity of the short version. In order to reduce the items from the original 45-item scale, we performed an Exploratory Factorial Analysis and Confirmatory Factorial Analysis, using principal axis factoring and Promax and the Kaiser normalization rotation method, in order to verify if the new factorial structure was similar to the original one. Finally, we investigated the psychometric properties of RAS 20-item using (1) alpha coefficients to measure internal consistency; and (2) correlations between the RAS-20 and the MASC and RAS-20 and MAQ for the third sample, and with the reading tasks (LPI, TFL, and AFLeT) in order to test convergent validity of the short version. Analyses were performed using the IBM^®SPSS^® Software (version 25).

Modifications were restricted to verbal tense or terms replacement, corrected to facilitate children’s comprehension. First, the evaluation of the translations’ synthesis by three bilingual referees pointed out that the instructions and structure were adequate for the Brazilian population. However, some verbs had to be modified to obtain more understandable verbal tenses: (1) item (If no one helped me, I would fail reading) “Se ninguém me ajudar, é possível que eu não consiga ler” was changed to “Se ninguém me ajudasse, eu possivelmente não conseguiria ler”; and (2) item (I don’t like to look at books) “Eu não gosto de olhar livros” was changed to “Eu não gosto de olhar para livros”. Just one aspect of the layout was pointed out: in the original scale, children were asked to write their first name in the sheet upper right hand corner, but we changed to the upper left corner, since in Brazil the heading is usually placed on the left.

In the review of the 45-item instrument by the target audience (first sample, 20 typically developed children), the respondents agreed that the instructions and the response scale were understandable to them. However, they pointed out the similarity between two of the questions (questions number 5 “I can read, but I don’t understand what I have read” and 14 “I can’t read, but I don’t feel like it”). Nonetheless, the authors have agreed to keep both questions, given that the items were referring to different aspects of the reading anxiety construct.

After back-translation, in order to produce the idiomatic and conceptual equivalence, some terms in the instructions were replaced by other more frequently used in Brazilian Portuguese: (1) (you disagree) was wrongly translated to “você não concorda” and were changed to “discorda” to make the expression consistent with the scale; and (2) (while you read it) “enquanto você a lê no questionário” the terms “a” and “no” were added because the articles were in the original version, but were not included in the translation.

Then, after back-translation and the agreement of the author of the original scale, the provisional forward translated questionnaire was pilot-tested and administered to the first sample (N = 12) of this study. After the evaluation, many of the participants (66%) stated that they found it difficult to complete the questionnaire because it was too long and some questions seemed to be repeated. Taking into account that statement, we decided to review the factorial analysis of the original scale (Zbornik, 1988) and develop a short version, based on the original analysis.

Reviewed the original scale factorial analysis of the Zbornik (1988) study, it was noticed that some items were not contributing to the scale adequacy. The sample from the original study (Zbornik, 1988) was composed of 436 children, from 9 to 12 years of age, attending parochial schools, L1, in the U.S. After examining the factorial structure of the original 45-item instrument and excluding items with low loadings in each factor (values below 0.40), a shorter version of the scale (in English) was proposed by the authors. After excluding those items, the RAS short version maintained 20 items. For this proposed shorter version in English, no published psychometric studies are available for RAS so far. Although we have followed the 5 steps to adapt the original RAS, we opted to use the shorter version, considering that a scale with 20 items is more feasible to be used in studies (and in clinics) than a 45-item questionnaire.

The descriptive analyzes for the third sample are reported on Table 1. There were no significant differences in the scores among 2^nd (N = 14), 3^rd (N = 38), and 4^th (N = 36) grades (one-way ANOVA). The results for the total sample are shown below.

For the Brazilian Portuguese version, results from the Exploratory Factorial Analysis (EFA) using the third sample (N = 88) – the eigenvalue criteria (higher than 1), screen plot, and parallel analyses – revealed that the three-factor model proposed by Zbornik and Wallbrown (1991) was supported. Nonetheless, the Confirmatory Factorial Analysis (CFA) results showed that the items loaded in the three dimensions were different from those of the original scale (Table 2). CFA showed good model fit indices, according to Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy = 0.75 and the Bartlett’s test of sphericity (χ² (210) = 626.386,p < 0.001).