Endel P

1
ECVP 2013, 25-29 August, Bremen, Germany
Crowding by a single bar
Endel Põder University of Tartu, Estonia E-mail
endel.poder_at_ut.ee
Background Visual crowding does not affect much
the detection of the presence of simple visual
features but perturbs heavily combining them into
recognizable objects. Still, the crowding effects
have been rarely related to general pattern
recognition mechanisms.
Examples of stimuli
Example of display

• Methods
• In many aspects similar to Dakin et al (2010).
• Observers had to identify the orientation (4AFC)
  of a rotated T presented briefly (60 ms) at a
  peripheral location (eccentricity 6 deg).
  Adjacent to the target, a single bar was
  presented. The bar was either horizontal or
  vertical, and located in a random direction
  (0-360 deg) from the target.
• Orientation and position of the crowding bar and
  observers responses are expressed relative to
  the upright orientation of the target. The data
  are pooled across absolute orientations of the
  target.

Response panel
Results of the experiment
Very strong and regular effects of the crowding
bar on the identification of target orientation.
Wrong answers are evoked by a global stimulus
configuration that resembles a correspondingly
oriented target. Only rough relative positions of
features matter, exact metrical relations are not
important.
Modeling Follows the ideas of the standard
model (e.g. Riesenhuber Poggio, 1999) local
feature detectors, spatial pooling of their
outputs, combining the results into second-order
features. Assumes three kinds of simple features
horizontal and vertical bars plus a low-pass blob
at the target position. Second-order features are
combinations of the oriented and un-oriented
features (e.g. presence of a horizontal bar above
the object center). Particular combinations of
second-order features provide evidence for each
target orientation. The hypothesis with maximum
support was selected as the response on a given
trial.
Results of a simulation (points experiment,
lines model)
Supposed spatial arrangement of receptive fields

• References
• Dakin, S. C., Cass, J., Greenwood, J. A., Bex,
  P. J. (2010). Probabilistic, positional averaging
  predicts object-level crowding effects with
  letter-like stimuli. Journal of Vision,
  10(10)14, 116.
• Pasupathy, A., Connor, C. E. (2001). Shape
  representation in area V4 position-specific
  tuning for boundary conformation. Journal of
  Neurophysiology, 86, 25052519.
• Riesenhuber, M., Poggio, T. (1999).
  Hierarchical models of object recognition in
  cortex. Nature Neuroscience, 2, 10191025.

• Conclusions
• The results are broadly consistent with the
  standard model.
• Crowding in visual periphery is caused by large
  pooling regions that include information from
  surrounding objects.
• This study supports the idea that visual system
  uses second-order features that encode the
  presence of simple visual features in some
  approximate position relative to candidate object
  center (Pasupathy Connor, 2001).