The phrase "unreliable narrator" shows up in nearly every student essay on The Turn of the Screw, often as a label rather than an argument. Students write that the governess cannot be trusted and then move on without explaining what in the text makes her suspect. A good rubric separates students who simply use the term from students who demonstrate the concept with evidence.

Grading this concept is tricky because the novella never settles the question. A student who argues the governess is deluded and a student who argues she is truthful can both earn top marks if their reasoning holds up. The rubric therefore has to reward the quality of the case rather than the side chosen, which is a shift that many teachers need to state explicitly to students.

Clear criteria also protect teachers from their own preferences. It is easy to reward essays that match the reading you find most convincing and quietly penalize the other camp. Writing the standard down before grading, and applying it with the same steps each time, keeps the evaluation anchored to the student's argument instead of the grader's taste.

What Strong Evidence of Unreliability Looks Like

Strong essays point to specific moments where the governess's perception outruns the facts. She sees Quint at the tower before she has heard his name, she insists on her certainty even when the housekeeper cannot see what she sees, and she interprets Miles's silences as confirmation. The best students show how these moments create a pattern and then explain why James might have built the story that way.

• Identifies at least two specific passages where the governess's certainty exceeds her evidence
• Distinguishes between what the governess observes and what she concludes
• Considers how the Douglas frame adds a second layer of mediation
• Addresses counter-evidence, such as the housekeeper's partial confirmation
• Explains the effect of the ambiguity on a reader instead of only asserting it

Common Weaknesses Worth Naming in Feedback

The most common weakness is assertion without support, such as calling the governess hysterical with no quoted scene behind the claim. Another frequent problem is anachronistic diagnosis, where students apply modern psychological labels without considering how a Victorian reader would have understood her behavior. A third is treating unreliability as proof that nothing supernatural happens, when the text leaves both options open.

Effective comments name the gap and suggest a repair. Instead of writing "needs more evidence," a teacher might write that the claim about her delusion would be stronger if the student compared her description of the figure at the window with what the housekeeper actually confirms. That kind of comment points to a specific move the student can make in revision.

Using AI to Apply the Rubric Consistently

When teachers load a rubric like this into an AI grading tool, the software can scan each essay for the criteria and flag where evidence is missing. It can note that a student quotes the lake scene but never connects it to a claim about perception, or that an essay mentions the frame narrative once and never returns to it. These flags give the teacher a map of each paper before reading begins.

The teacher still decides how much a subtle argument deserves, but the repetitive checking is handled. Over a set of essays, this produces more uniform scoring and more specific comments than most people can sustain manually. Students benefit because the feedback they receive is tied to the same stated criteria that the whole class was given.