I edited the questions and added an example. As you can see, VERB is not coded for the C level of SENTENCE_TYPE.

I have 240 subjects and 24 verbs. In total, each student rated 36 sentences.

Thank you! I understand that anova() is a good way to decide if terms improve the model. I was just confused by some papers I read which seemed to exclusively rely on the p-values next to the coefficients. I am still a little confused by the meaning of the p-values next to the coefficients. I added a question about the difference in meaning when using treatment coding instead of sum coding.