This is probably some kind of testing artifact. Table 1 shows that 64.3% of monolinguals were literate compared to 92.9% of bilinguals. Then in the supplemental

This is probably some kind of testing artifact. Table 1 shows that 64.3% of monolinguals were literate compared to 92.9% of bilinguals. Then in the supplemental methods we see that illiterates took a different test than literates. They defined cognitive impairment as a score of under mean - 2*SD measured from groups consisting of ~80% literates: >"The process of adaptation included culturally appropriate modifications of the original English version of ACE-R by examining each item for cultural relevance, translatability, comparable difficulty and adaptability with an aim of tapping the domain identical to the original version. Translations and back translations, and pilot testing were done based on standard procedures.3 ACE-R was adapted for the illiterate population by modifying literacy dependent items.4 Normative data and cut off values in the local languages- Telugu and Hindi were developed by administering the test to healthy controls. Telugu control group subjects consisted 907 literates and 199 illiterates, and Hindi control group subjects consisted 436 literates and 109 illiterates. Mean and SD for total scores and sub-scores in each domain were calculated across 4 age groups (20-39, 40-59, 60-79, 80 and above), and three education levels (>12 years, 4-12 years, 0-3 years), using standard formulae of Mean-2SD. Impairment in cognition was defined if the score on the ACE-R sub-domains was less than 2.00 SD below the mean level of age and education-matched norms." Then we have the arbitrary age groups and education levels which makes this look p-hacked, there needs to be some assessment of how much those arbitrary choices affect the conclusions.