The Use Of The Cloze Test In Reading Comprehension Assessment In Brazil: Post-Pandemic Challenges

The Cloze test has been widely used for 70 years to assess proficiency in understanding texts in different languages, both L1 and L2. The basis of the test is the systematic gapping of a text, and hit scores are associated with the participant’s degree of comprehension of the task. In Brazil, the Cloze procedure has been productive like an instrument for reading measures in L1. Through it, it is possible to distinguish reader profiles- frustrated, instructional, and independent- by checking the participant’s performance when filling in the gaps. Almeida’s survey (2023) shows that this procedure was adopted in 75 of the 345 studies on reading assessment over the last 20 years. In the post-pandemic period, when reading asymmetries are even greater, the adoption of the exact word criterion does not allow an appropriate diagnosis: post-pandemic applications carried out at the Colégio de Aplicação of the Federal University of Sergipe- April 2023- showed that there are students who are concentrated at the level of insufficient comprehension (Machado, Santos & Cruz 2019; Santos & Machado 2022). It endorses the necessity of stimulating the application of consolidated resources, such as the Cloze test, that may assist education professionals in improving their practices and students’ reading levels. This article analyzes the papers published in three virtual databases that applied the traditional Cloze test to assess reading comprehension. The criteria for analyzing these answers are based on Taylor’s (1953) exact answers initial proposal (Brown 1980; 2013), added to other assessment instruments used in the Psychology field. Therefore, this article carries out a systematic review whose method considers empirical studies on reading assessment with the Cloze test, published from 2009 to 2022, and applied to Elementary School students. We found that most of the papers adopted the completion of the exact word as the measurement criterion (Joly & Istome 2008), likewise the computerized tests (Joly & Piovezan 2012; Brito, Ribeiro & Seabra 2022) instead of considering familiarity (Oliveira et al. 2007) or grammatical class (Abreu et al. 2017) of the word. We highlight the need to adopt natural language techniques, with measures of lexical distances, which can help diagnose how much comprehension there is in low reading scenarios. These distances can be found through the “minimum number of insertions, deletions or substitutions of a single character needed to transform one word into the other” (Petroni & Serva 2010). They will be measured by following the same grammatical class or semantic field as the expected word.