Reading Fluency in a Multilingual Context: The Impact of Text Complexity on Fluency Across Language Groups in Grade 3 Students

Literacy acquisition is considered one of the core outcomes of primary education. Reading development initially focuses on phonological awareness and technical skills but becomes increasingly automatic, allowing the student to shift attention to the meaning of what is read (LaBerge & Samuels, 1974). That is, reading words and texts initially mainly relies on decoding skills (non-lexical route) but is increasingly facilitated by word recognition (Coltheart, 2006; Coltheart et al., 2001). Reading development may however be affected by the language background of the student (in relation to the language of instruction) as well as the multilingual context in which (s)he grows up. This becomes increasingly relevant as Europe has experienced distinct waves of migration from both outside and within EU member states (Massey et al., 1998) and several countries (e.g., Switzerland, Belgium, Luxembourg) have more than one official language. The consideration of home language in relation to language of instruction becomes especially important when diagnosing learning disorders as international classification systems explicitly state that language proficiency must be ruled out as a possible underlying cause.

In Luxembourg, multilingualism is omnipresent in both society and the education system. Language varies across school levels - starting with Luxembourgish in preschool, transferring to German in Grade 1 and gradually introducing French from Grade 3, whereby roughly two thirds of the student population in primary education not speaking Luxembourgish or German at home (MENJE, 2024) and hence are acquiring literacy in their 2nd language (Ugen et al., 2021). Data from the National School Monitoring Programme (Épreuves Standardisées, ÉpStan; Martin et al., 2014) consistently demonstrates that students´ language backgrounds have a significant impact on their scholastic performance, whereby already in Grade 3 significant differences can be found in the performance levels between students who do and those who do not speak Luxembourgish or German at home (Hoffmann et al., 2018).

In text reading, different levels of complexity can be present, which in turn affect reading fluency, i.e., the ability to read accurately within a certain amount of time. For example, word frequency has been consistently identified as a robust predictor of word recognition, which in turn is reflected by higher fluency (Brysbaert et al., 2018). Word type may also impact fluency, as nouns are generally easier to read as they mostly refer to concrete things, whereas verbs are represented in different tenses and forms and adjectives often add complexity to both the sentence structure and its content. In addition, text length can impact fluency as reading increasingly longer texts requires continued focus and attention, and increases cognitive load and comprehension demands.

These levels of complexity may interact with the language background of the students as, for example, the effect of word frequency and word type may be accentuated in students with lower levels of language proficiency. Similarly, reading in a language other than the home language may increase cognitive load and reading may become less fluid, especially when text length increases. This interaction may be important when interpreting differences in reading fluency between students with different language backgrounds and becomes particularly relevant in the context of diagnosing a learning disorder and providing tailored support to students experiencing learning difficulties.

Methodology, Methods, Research Instruments or Sources Used
To investigate the relationship between reading fluency, text complexity and student language background, we administered three texts of increasing complexity to approx. 750 students in Grade 3 (age range 7-10 years). Students were selected from schools in the 15 different school districts of the country to ensure good representation, especially as regional differences in family affluency and percentage of students with immigration background in school have been found (Helfer et al., 2015). The data was collected as part of a larger project concerning the development of a test battery to diagnose specific learning disorders in a multilingual context (Romanovska et al., 2024).
Three texts were construed in which text length and complexity was systematically varied. Texts concerned themes (e.g., animals, friendship) and settings (e.g., at home, in nature) matching the experiences of students aged 7-10. The complexity of texts was varied by text length (60-90-120 words, respectively) and word type (only in Text 3 adjectives were introduced). Other variations in text complexity were kept to a minimum (e.g., all texts were written in active voice and mainly in present tense). Word frequency for nouns, verbs and adjectives was established by using ChildLex, a lexical database based on 500 German children’s books (Schroeder et al., 2015). Based on the absolute frequency per word type, we computed frequency classes, which are considered more reliable. Although we aimed to use mainly frequent words, adjectives generally occur less often in children´s books than nouns and verbs, hence Text 3 also contained more words that are less familiar to children in the targeted age range. In addition, socio-demographic student information was collected via a questionnaire, including age, gender and language background. All materials were pre-tested before students in the current sample were tested in their own school during normal school hours by trained administrators.
Text fluency was computed for each text by dividing the number of accurately read words by time in minutes. The sample was divided in 4 language groups: Luxembourgish/German (42%), Portuguese (19%), French (16%) and other (23%).

Conclusions, Expected Outcomes or Findings
Preliminary results of 3x4 mixed-method analyses of variance, with text as within-subject variable and language group as between-subjects variable reveal main effects for text (fluency decreases as text complexity increases) and language group (differences are observed between the four language groups). The significant text x language interaction indicates that differences in reading fluency for the three texts vary per language group. Post hoc analyses revealed that when text complexity is only varied by length (i.e. Text 1 vs. Text 2), reading fluency of students that speak Luxembourgish or German at home remains stable, whereas it decreases for students that speak Portuguese or other languages at home. When text complexity is also varied by word type (i.e., by adding adjectives in Text 3), reading fluency decreases in all groups regardless of their home language. Additional analyses will be conducted to further analyze the effect of word type, especially in consideration of the associated variation in word frequency.
The preliminary results confirm that home language influences reading fluency and that students that do not speak Luxembourgish or German at home may experience more challenges in reaching expected attainment levels. Although differences between language groups become less pronounced with increased text complexity (as indicated by a reduction in effect size), differences in fluency between language groups are observed for all texts. In addition, for students not speaking the language of instruction at home, fluency was already affected by varying text length, whereas for students that do speak the language of instruction at home, fluency was only affected by increasing text complexity by varying word type. These findings not only add to the state of the art regarding the effect of multilingualism on reading fluency but also have implications for educational practice regarding the identification of learning difficulties or diagnosing learning disorder in multilingual contexts.