I am highly skeptical of the last paper you linked.
First, they make the classical mistake of concluding causation where there is only correlation. In particular, there is no discussion whatsoever about possible confounders, i.e. common factors that influence both programming ability and English proficiency. Some on top of my mind:
- Owning a computer with internet access at home
- Effort/ability from the individual students (reflected in overall grades in all subjects)
- Students' language ability in their mother tongue (can hint at students' difficulties in expressing their thoughts in a structured manner)
Second, their statistical analysis is inadequately simple. In particular, given that they only have 16 students with intermediate English knowledge, I would not be surprised if the difference in grades distribution was not statistically significant.
Third, quoting from the paper:
> In analyzing the data, the grade E for computer programming course was excluded because it did not reflect the relationship between English language proficiency and computer programming ability.
This is quite a red flag, they are essentially removing data that does not show the correlation they want! And nothing is said about these students: how many they were, their English proficiency, etc.
Finally, as anecdotal evidence, a sizable portion of my classmates were unable to come up with even the simplest algorithms, even when taught in and asked to use pseudo-code inspired from their mother tongue.