Conclusions

1
THE PREFERRED LEVEL OF FACE CATEGORIZATION
DEPENDS ON DISCRIMINABILITY
Chris DLauro1, James W. Tanaka2, and Tim
Curran1 1. University of Colorado, Boulder 2.
University of Victoria
Contact christopher.dlauro_at_colorado.edu with
questions or comments.
Experiment 2 Homogenous Faces
Results
Trial Structure
Reaction Time (RT)
Accuracy
This example shows a true,subordinate- level
trial for Julia Roberts.
Experiment 2 used a homogenous group of faces
(shown below). According to the differentiation
hypothesis, we would expect these faces NOT to
show a subordinate level shift.
subordinate level shift
Time
Sub
Basic
Super
Results
Accuracy
Reaction Time (RT)
Dogs
Diverse Faces
Diverse Faces
Homogenous Faces
Data matched results from Experiments 1 and 2.
The diverse face block showed the
subordinate-level shift, while the homogenous
face block did not. As it stands, one could
explain this data by arguing that the diverse
face group is simply better-known. How can we
exclude this explanation? By looking at the two
faces included in both blocks.
Hypothesis If the context of similar distractors
causes the basic level advantage, then even
faces should show a basic level advantage if the
distractor faces are similar enough.
basic level advantage
Between-block Results
Demi-Britney RTs by block, level
Dogs
Homogenous Faces
Britney Spears and Demi Moore pictures were
included in both the diverse and homogenous
blocks. In these trials, there was a significant
interaction of block and level (p.03), showing
an RT cost at the subordinate level in the
homogenous block only. Thus, the context of more
similar faces caused the subordinate level RT
cost, not some property intrinsic to the faces.

Experiment 1 Diverse Faces
Both dogs and faces showed the basic level
advantage in RT - matching our prior
predictions.
basic level advantage
In the Experiment 1, we replicated Tanaka (2001)
using the same faces and dogs as in the original
study.
Experiment 3 Blocks of Diverse and Homogenous
Faces
In Experiment 1, we demonstrated the expected
subordinate level shift for faces. In
Experiment 2, we showed that similar-looking
faces would NOT show the subordinate level shift
- in agreement with the differentiation
hypothesis. In Experiment 3, we aimed to
specifically show that the context of the other
faces is key for the subordinate level shift.
Subjects performed two blocks one with
homogenous faces (from Experiment 2), and one
with diverse faces (shown below). Britney Spears
and Demi Moore were included in both blocks.
subordinate level shift
Jerry Seinfeld
Marilyn Monroe
Hillary Clinton
George Bush,Sr.
Jay Leno
Cindy Crawford
Princess Diana
Bill Clinton
Basic
Sub
Diverse Faces
Homogenous Faces
Results
Conclusions
Reaction Time (RT)
Accuracy
Consistent with the differentiation hypothesis,
faces were identified less quickly among similar
faces and more quickly among a diverse group of
faces. Thus, categorization level depends on
stimulus discriminability, rather than merely
being an inherent property of different stimulus
types such as faces vs. non-face objects.
subordinate level shift

• Research Supported by
• Temporal Dynamics of Learning Center (NSF Grant
  SBE-0542013)
• James S. McDonnell Foundation

Diverse block faces are (top row, left to right)
Jack Nicholson, Tiger Woods, Martha Stewart
(bottom row) Harrison Ford, Tom Hanks, Oprah
Winfrey.

• References
• Murphy and Brownell (1985). J. of. Exp.
  Psych.Learn, 11, 70-84
• Rosch et al. (1976). Cog. Psych., 8, 382-439.
• Tanaka (2001). J. of. Exp. Psych.Gen., 130,
  534-543

• Data replicated the previous findings of Tanaka
  (2001).
• Subjects showed
• Subordinate level shift to faces, but not dogs
  (shown in red).
• High accuracy for dogs and faces.