RESEARCH REPORT
Unskilled but Aware:
Reinterpreting Overconfidence in Low-Performing Students
Tyler M. Miller and Lisa Geraci
Texas A&M University
People are generally overconfident in their self-assessments and this overconfidence effect is greatest for
people of poorer abilities. For example, poor students predict that they will perform much better on
exams than they do. One explanation for this result is that poor performers in general are doubly cursed:
They lack knowledge of the material, and they lack awareness of the knowledge that they do and do not
possess. The current studies examined whether poor performers in the classroom are truly unaware of
their deficits by examining the relationship between students’ exam predictions and their confidence in
these predictions. Relative to high-performing students, the poorer students showed a greater overcon-
fidence effect (i.e., their predictions were greater than their performance), but they also reported lower
confidence in these predictions. Together, these results suggest that poor students are indeed unskilled but
that they may have some awareness of their lack of metacognitive knowledge.
Keywords: metacognition, overconfidence, prediction
When people make self-assessment errors, and they often do,
they are usually in the direction of overconfidence. For example,
people overestimate their reasoning ability, their ability to recog-
nize humor, and their knowledge of grammar (Kruger & Dunning,
1999), and they underestimate the time they need to complete tasks
(Buehler, Griffin, & Ross, 1994). In addition, most people believe
they have better than average leadership skills (Dunning, Heath, &
Suls, 2004), driving skills (Knouse, Bagwell, Barkley, & Murphy,
2005; A. F. Williams, 2003), and dating popularity (Preuss &
Alicke, 2009). Similarly, people are overconfident in their aca-
demic abilities. In the classroom, undergraduate students tend to
overestimate their performance on upcoming exams (cf. Hacker,
Bol, Horgan, & Rakow, 2000; Miller & Geraci, 2010). For exam-
ple, in Hacker et al. (2000), many students predicted that they
would earn scores more than 30% higher than their actual scores.
This overconfidence effect is greatest for people who score
below average compared with those who attain above average
scores (e.g., Bol, Hacker, O’Shea, & Allen, 2005; Burson, Larrick,
& Klayman, 2006; Hacker et al., 2000; Kelemen, Winningham, &
Weaver, 2007; Krueger & Mueller, 2002; Kruger & Dunning,
1999; Miller & Geraci, 2010; Nietfeld, Cao, & Osborne, 2005).
When undergraduate students are asked to predict their exam
scores, students with the higher scores have more accurate predic-
tions than do students with lower scores (Hacker et al., 2000).
Similarly, in the laboratory, students with high SAT scores are
more accurate than students with low SAT scores when asked
about their mastery of a set of Swahili–English word pairs (Kele-
men et al., 2007).
Several studies have confirmed that low performers are more
overconfident than high performers, but the reason for this greater
metacognitive inaccuracy is debated. Some have suggested that the
exaggerated overconfidence effect in low performers is the result
of a measurement artifact whereby low performers have room to
make predictions that are much higher than their level of perfor-
mance, whereas high performers do not (Krueger & Mueller,
2002). But perhaps the leading interpretation is that low perform-
ers are overconfident because they have a general deficit of meta-
cognitive insight. The double-curse account suggests that in addi-
tion to lacking knowledge of the material, poor students also lack
awareness of the knowledge that they do and do not possess.
According to recent characterizations, low-performing students are
“blissfully incompetent” (W. M. Williams, 2004) and “unskilled”
and “unaware” (Ehrlinger, Johnson, Banner, Dunning, & Kruger,
2008). Evidence for this characterization of low performers as
unskilled and unaware comes from studies showing that low
performers in particular predict that they will perform much better
than they actually do (the overconfidence effect) even in the face
of counterinformation (Kruger & Dunning, 1999).
It follows from the double-curse account that if low performers
are blissfully incompetent, then, in addition to making inaccurate
performance predictions, they would also be unduly confident in
these predictions. Indeed, Dunning (2005) likened low performers’
inflated self-assessments to brain damage (i.e., anosognosia) and
suggested that “people performing poorly cannot be expected to
recognize their ineptitude” and that “the ability to recognize the
depth of their inadequacies is beyond them” (p. 15).
The current study examines whether this characterization of low
performers is correct. Are low performers entirely unaware of their
This article was published Online First January 24, 2011.
Tyler M. Miller and Lisa Geraci, Department of Psychology, Texas
A&M University.
Correspondence concerning this article should be addressed to Tyler M.
Miller, Department of Psychology, Texas A&M University, College Sta-
tion, TX 77843. E-mail: milltyl@tamu.edu
Journal of Experimental Psychology: © 2011 American Psychological Association
Learning, Memory, and Cognition
2011, Vol. 37, No. 2, 502–506
0278-7393/11/$12.00 DOI: 10.1037/a0021802
502

deficits? To answer this question, we asked low- and high-
performing students to rate their confidence in their prior grade
predictions. If low performers are unaware of their metacognitive
deficits, then they should be at least as confident in their perfor-
mance predictions as high performers are in theirs. However, if
low performers have some awareness of their errors in judgment,
then they should be less confident than high performers in their
performance predictions.
To examine the nature of the overconfidence effect in low and
high performers, we highlight the distinction between two forms of
confidence. Herein, we refer to errors of overestimating one’s
ability (predicting that one will perform better than one does) as
functional overconfidence and errors of overcertainty (being
overly certain of one’s predictions) as subjective overconfidence.
As far as we are aware, only one study has examined subjective
confidence associated with predictions of performance (Dunlosky,
Serra, Matvey, & Rawson, 2005). In this study, participants made
judgments of learning (JOLs) to indicate the likelihood that they
would remember unrelated noun pairs. For each JOL, participants
made a second-order judgment (SOJ) indicating their confidence
in the JOL. Results showed that JOLs and SOJs were functionally
distinct from each other, displaying a U-shaped curvilinear rela-
tionship with higher SOJs at extreme JOLs. In addition, the curve
was asymmetrical, showing that SOJs associated with high JOLs
were much greater than SOJs associated with low JOLs.
In the current study, we used a methodology similar to that used
by Dunlosky et al. (2005) to examine the nature of the overcon-
fidence effect in low-performing students. College students in an
upper level cognitive psychology course were asked to predict
their exam scores and rate their confidence in their predictions. In
Study 1, students made letter grade predictions for the first exam
immediately before the exam, and in Study 2, they made percent-
age predictions immediately before the first and final exams. In
both studies, students rated their confidence in their predictions.
We predicted that low-performing students would show a greater
functional overconfidence effect than would high-performing stu-
dents in that low-performing students would predict that they
would receive disproportionately higher grades than they did. Of
interest was whether the low-performing students would also be
subjectively overconfident. According to the double-curse expla-
nation, subjective confidence should be just as high, if not higher,
for the low-performing students compared with the high-
performing students because the low-performing students are as-
sumed to lack awareness of their metacognitive deficit. If low-
performing students are less confident than high-performing
students in their grade predictions, then this would suggest that
low-performing students may not be unaware of their metacogni-
tive difficulties, as previously suggested.
Study 1
Method
Participants. Ninety-one students from a cognitive psychol-
ogy course at Texas A&M University participated in the study.
The students were largely junior and senior psychology majors.
Seventy-four percent of psychology majors at Texas A&M Uni-
versity are female, 72% are European American, 5% are African
American, 17% are Latino or Hispanic American, 5% are Asian
American, and 1% were in an “other” category.
Design and procedure. Immediately before the first exam,
participants recorded a letter grade prediction on the exam cover
sheet. These letter grades were converted to numeric values for the
analyses using the standard grading scale (i.e., scores between
100% and 90% received an A, between 89% and 80% received a
B, between 79% and 70% received a C, between 69% and 60%
received a D; and 59% or less received an F. The same grading
scale was used for both studies). For example, if a student pre-
dicted a B
, that score was converted to a numeric value of 88%
on the basis of the midpoint of the B
range. Students also rated
their confidence that their exam score prediction was correct using
a Likert-type scale ranging from 1  low confidence to 5  high
confidence. Students were given an incentive to be as accurate as
possible in these predictions: If their exam score prediction was
within the same letter grade equivalent as their actual performance,
they received an additional 2 percentage points on their exam
score.
Results
We divided students into quartiles on the basis of their exam
performance. We created difference scores for each student by
subtracting the actual score from the predicted score. As such,
positive numbers indicated overconfidence and negative numbers
indicated underconfidence (see Table 1). As expected, the omnibus
F test indicated significant differences in calibration by quartile,
F(3, 86)  25.23, MSE  30.92, p  .001, p2  .47. Tukey post
hoc tests indicated that students in the top two quartiles, Quartiles
3 and 4, were significantly more calibrated (less overconfident)
than were students in the bottom two quartiles, Quartiles 1 and 2
(ps  .05). Students in Quartile 2 were significantly more cali-
brated than were students in Quartile 1 (p  .05). Thus, results
Table 1
Mean Prediction, Grade, Difference, and Confidence Scores for
Studies 1 and 2 by Quartile
Quartile Prediction Grade Difference Confidence
Study 1 Exam 1
1 79.60 (1.31) 69.09 (0.88) 10.50 (1.45) 3.23 (0.11)
2 83.57 (1.27) 78.26 (0.54) 5.30 (1.06) 3.65 (0.18)
3 85.78 (1.06) 84.87 (0.41) 0.91 (1.00) 3.74 (0.11)
4 87.36 (0.86) 90.73 (0.50) 3.36 (1.14) 3.77 (0.11)
Study 2 Exam 1
1 79.10 (1.44) 66.83 (1.33) 12.28 (1.98) 3.24 (0.16)
2 84.45 (0.95) 80.14 (0.71) 4.31 (0.97) 3.55 (0.13)
3 86.83 (0.93) 88.62 (0.27) 1.79 (0.94) 3.66 (0.13)
4 89.09 (0.80) 96.00 (0.58) 6.91 (1.08) 3.75 (0.12)
Study 2 Final Exam
1 78.10 (1.74) 60.38 (0.90) 17.71 (2.12) 64.76 (4.48)
2 79.55 (1.72) 69.30 (0.47) 10.25 (1.71) 72.70 (4.51)
3 79.64 (1.59) 76.00 (0.46) 3.64 (1.51) 80.95 (2.17)
4 84.57 (1.58) 85.62 (0.84) 1.05 (1.51) 72.38 (3.96)
Note. Students in Studies 1 and 2 reported confidence on a scale of 1–5
for Exam 1 and students in Study 2 reported confidence on a scale of
0–100 for the final exam. Standard errors are shown in parentheses.
503UNSKILLED BUT AWARE