Two students sit the same assessment. Same questions, same clock, similar reparation. One of them is also carrying a quieter worry: that a weak result could be read as proof about an entire group.
That invisible extra load is at the heart of assessment equity in higher education, and it is the subject of Claude Steele’s Whistling Vivaldi (2010).
What “identity contingencies” means
Steele, a social psychologist, uses the phrase for the conditions a person must manage in a setting because of a social identity they carry.
His title comes from a story about the writer Brent Staples, who whistled Vivaldi and the Beatles in public so strangers would read him as an educated, refined person rather than through a stereotype.
Stereotype threat: two tasks at once
With Joshua Aronson, Steele named the best-known version of this idea: stereotype threat, the risk of confirming a negative stereotype about one’s group when that stereotype is known to the person and to others.
The claim is not that a student believes the stereotype. Simply being aware of it can change the experience of a task, because the student is answering the question and tracking how the answer might be judged.
The question shifts from “what is wrong with this student?” to “what is this situation asking of them?”
Steele also points to cues, such as how many people who share a student’s identity are in the room. That moves the explanation from the individual to the setting, which a university can examine and change. As one reviewer summarises his practical responses: feedback that signals confidence a student can meet high standards, chances to affirm a sense of self, and conversations across groups.
A small example: demographic questions before a test
Many institutions ask students for demographic details at the start of a test. Could that routine act as a cue?
A PLOS ONE replication by Matthew Inglis and Steven O’Hagan tested this in the UK Mathematics Trust’s Junior Mathematical Challenge and found no evidence of the effect. They concluded that moving such questions on an answer sheet is unlikely to be a silver bullet.
A 2025 AAPT poster by Jayson Nissen, Ben Van Dusen and Kevin Roberge describes a randomised trial with about 12,000 students on low-stakes physics assessments. Its abstract doesn’t report results, so it shows the question is being tested at scale, nothing more.
The takeaway is modest. Faculty can look at how an assessment’s purpose is framed, what feedback signals about expectations, and which cues, intended or not, tell a student whether they belong in the room.
The honest limit
The evidence needs care. A meta-analysis by Paulette Flore and Jelte Wicherts on girls’ maths and science performance found a small average effect, with signs of publication bias. A later preregistered replication by Flore, Mulder and Wicherts did not find it.
In 2024, Russell Warne and Ross Larsen published a protocol for a new meta-analysis of studies with African American participants. They note the idea is widely taught and that some findings haven’t held up in close replications.
Steele’s core insight, that the situation around an assessment shapes what it reveals, can be taken seriously without treating any single effect size as settled.
Why this matters now
Many universities are rethinking how they assess learning, including spoken and in-person formats. Every redesign changes the cues around an assessment, so each is a chance to ask how it feels as well as what it measures.
Socratic Metric AI is a voice-based oral assessment platform that invites students to explain their reasoning aloud, giving faculty more evidence of how a student thinks. Whether a spoken format feels more or less pressured to particular students is an open question each institution should examine with its own students. The platform doesn’t remove stereotype threat, and faculty remain the people who judge, interpret and decide.