Inhibition and Multilingualism

The purpose of this experiment was to gather data on a range of people, from monolingual to multilingual, and their scores on various tasks related to inhibition. The tasks were audio digit span, digit span, n-back task, language task, Simon task, and task switching.

From a cursory glance, the Simon Effect seemed to be statistically significant, so I created linear regression plots with various other variables.

Try the Simon Effect

Our Simon Effect Score was calculated by taking the average reaction time for incongruent trials minus the average reaction time for congruent trials in milliseconds (only including correct trials).

This graph is a representation of a linear regression graph (generated by python script) of the age of acquisition of a second language and the Simon Effect. Though the line seems almost flat, the graph is visually deceiving because the data is not evenly distributed on the x-axis. Despite that, the p-value given shows the data to be statistically significant, with 0.000047 < 0.05.

We then moved on to test if Simon Effect proficiency increased with age itself, without any multilingualism or monolingualism variable. There is a very clear downward slope, though the age range is quite small, and the 25-year-old may be an outlier.

This data is interesting but doesn't seem fit for a linear regression graph. This is another fascinating graph that seems to be statistically significant from the results but isn't visually convincing. This regression is not based on the Simon Effect, but rather a test called the audio digit span, which essentially tests how many numbers in a sequence a person can recall when the sequence is spoken to them.

This is the best graph we made. "proftotal" was calculated by adding together the proficiency score of speaking, reading, and writing in each individual's language. For example, if a person was monolingual but 10/10 proficient in speaking, reading, and writing that language, then they got a score of 30. This explains the high amount of data points at 30. The rest are the range of multilingual people and their varying proficiencies.

In the end, the lab is still working on finding convincing data at the moment. Some current theories would be to have a "fluency score" rather than just a total of user-reported proficiency, and also including different inhibition test scores in the regressions.