Since I understand that your website core goal is to help people learn language - You should get straight to the point of "Language they wish to learn" with options provided (in a carousel, drop-down etc.), Ideally rather than to ask users the language they understand firstly. I believe users where you tested are confused between these two questions and needs a little of cognitive load applied when they are not really attentive of the question asked.
If your website goal is to make users learn languages - then probably your focus question and first should be on that.
But - how do you currently show these phrases - I mean in which language? Because a Polish guy may want to read this first question in his language of choice rather than to be in English. Ideally there exist a problem where the Polish guy might have traveled to Russia and thus even tracking through the IP will show the geo-location but not the preference of language (unless its overridden by browser by setting it to convert all languages to Polish). To connect this issue -
I would recommend clubbing your both questions as one - "I like to learn German", "Me gusta aprender aleman", "J'aime apprendre l'allemand"....the following in "English", "Spanish", "French".
Now that if the user chose "I like to learn German" it means that the user understands English and would prefer to learn German. So the main prevalent languages are assembled on priority on the top half and moves to lesser consequent languages to the bottom.
Since you have solved the first level problem, now you cannot definitely keep writing "I like to learn Spanish or Arabic", so next to "German" there could be a drop-down/mega-drop-down that quickly list out the language you need the user the choose such as "Arabic, French....."
So it means the user do not need to say what they know and what they need to learn - but simply say what they wanted to learn. It puts lot of cognitive load to answer questions, but by showing preferred or language of choice the eyes can quickly read it amongst the noise.
Designing the content part -
Now the problem of language choice/learnability is addressed - the way to arrange the entire content - "I like to learn German or Foreign language with a drop-down" (with multiple languages) are showcased in some rational way (alphabetical).
The other way to address the content is to have an option for the lesser relevant languages with a country of residence/world map on the nav bar. By clicking Angola - you may show options for Portuguese and Bantu alone in languages that mean "I like to learn German/Foreign language".
These options may have a higher probability to sell with your users!