I am working with a set of vowel lengths generated by non-native speakers of English and I want to find out whether they are consistent with vowel lengths generated by native speakers of English according to an established standard.

My data therefore looks like the following:

enter image description here

and so on.

I want to find out whether these values are significantly different than the established average vowel length for native speakers. That is, I need a statistical test which will determine if the data set is significantly different from one value.

What statistical tests would be appropriate for this?

  • $\begingroup$ The explanation of your dataset is very unclear. It is just an image of a column with numbers, but what do those numbers mean? $\endgroup$ Commented Aug 29, 2023 at 13:53
  • $\begingroup$ Also note: statistical tests are only a mathematical model, and they may not necessarily represent the statistical significance of your experiment. It is also very important to describe the process of gathering the data and take it into account. For example: People from different regions will speak with different speeds and will have different vowel lengths. Your selection of native speakers and non-native speakers might not need to be representing a representative random sample from the entire populations of the two groups, and it can be that your sample biases certain regions. $\endgroup$ Commented Aug 29, 2023 at 14:08
  • $\begingroup$ Is every participant pronouncing the same vowel? Do all the vowels have the same average pronunciation length (by native speakers)? $\endgroup$
    – dipetkov
    Commented Aug 29, 2023 at 19:39

2 Answers 2


A one-sample t test would allow you to compare your sample mean to the established average for native speakers.

  • 1
    $\begingroup$ Just make sure the established mean is really established. If it has any estimation error in it (e.g., if it came from a non-huge sample) this error has to be taken into account in the analysis. Also, statistical "significance" is overblown. I'd rather see an uncertainty interval for the difference in means. $\endgroup$ Commented Aug 29, 2023 at 11:41

If you want to test if the mean is different, then you can use a one sample t-test.

If you want to test something else, please let us know.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.