When do honey bees use snapshots
during navigation?
By Frank Bartlett
Bees and wasps learn information
about visual landmarks near the goal
Edge orientation (Srinivasan et al., 1994)
Color (von Frisch, 1967; Cheng et al.,
1986)
Size (Cartwright & Collett, 1979;
Ronacher, 1998)
Spatial relationships among multiple
landmarks (Cartwright & Collett, 1983)
Departing Wasp
Returning Wasp
How is this information subsequently
used over successive visits?
Snapshot template matching (Cartwright
& Collett, 1983)
Niko Tinbergen (1938)
What is snapshot navigation?
After Cartwright & Collett, 1983
View of landmarks is
memorized from the goal
Upon return the bee
steers flight by
sequentially matching
her memory to the
environment
Experiments revealing the contents
of snapshot memories
Single Landmark
Landmark
Training
3 equidistant Landmarks
Training
Goal
Goal
Distance from landmark
Landmarks
From Cartwright & Collett, 1983
-- When a single landmark is present bees rely on retinal image size.
-- When multiple landmarks are available bees rely on the inter-landmark angles
(or the spaces between landmarks)
Testing the snapshot hypothesis in a
small scale arena environment
►
The snapshot hypothesis makes
accurate predictions about where
insects should spend their time
searching for the goal.
 Can we replicate these findings?
►
The hypothesis also generates
predictions of flight paths to the goal
from more distant locations.
 Do steering commands generated by
snapshot matching predict honey bee
flight behavior while en-route to a
familiar goal?
 This has not been tested explicitly.
Methods
Training: bees visit an initial landmark configuration (60+
visits)
Testing: track with original configuration and other landmark
manipulations
The camera records bee position and body axis orientation at
60 Hz.
Search distributions
Replication of Cartwright & Collett (1983)
2x
Training
-- When a single landmark is present bees rely on retinal image size.
-- When multiple landmarks are available bees rely on the inter-landmark angles
-- These results are consistent with previous studies
Flight paths to the goal location
Bee Flights
Model
F
E
The model predictions were generated in
Matlab based on the algorithm provided by
Cartwright & Collett (1983)
Bees appear to be attracted to the nearest landmark and use it as a beacon even
over very short distances
Conclusions
► Search
at the goal
 Consistent with previous findings
► Steering
from more remote locations
using template matching
 Bee flights not consistent with model
predictions
 Strong role of beacons
►Consistent
with other results (Fry & Wehner,
2005; Collett & Baron, 1994) but extended to
shorter distances and more complex arrays
►Beacon selection probably driven by visual
salience
Acknowledgments
NSF IGERT
Yoav Littman
Fred Dyer
Jenny Jones
Steven Fry
Lora Bramlett
Mike Mack
Kourtney Trudgen
Chris Speilburg
Lauren Davenport
Short range visual navigation in
flying hymenopterans
►
Bees and wasp learn
information about visual
landmarks near the goal
 Edge orientation (Srinivasan et al.;
1994)
 Color (von Frisch, 1967; Cheng et
al., 1986)
 Size (Cartwright & Collett, 1979;
Ronacher, 1998)
 Spatial relationships among
multiple landmarks (Cartwright &
Collett; 1983)
►
How is this information
represented and subsequently
used over successive visits?
 Snapshot template matching
(Cartrwright & Collett; 1983)
Niko Tinbergen (1938)
How is this information learned?
The turn-back-and-look
First Visit
Tenth Visit
From Lehrer, 1993
--Motion parallax cues allow bees to distinguish nearby landmarks from
distant landmarks (Lehrer, 1993)
--Believed to aid in the selection and learning of the landmarks near a goal
What is snapshot navigation?
After Cartwright & Collett, 1983
View of landmarks is
memorized from the goal
Bee sequentially
matches her memory
to the environment
upon return
-- Insect visual memory is thought to be comprised of a two dimensional
“snapshot” that encodes the retinotopic sizes and positions of landmarks and
the gaps between them.
Model predictions vs. Flight Behavior:
pushed off course
F
E
After their course is diverted by the novel landmark, bees again use the next
nearest landmark as a beacon to guide flight.
Finding the match
►
Near the goal bees prefer to
maintain a southern facing body
axis
 Snapshot is probably
anchored to the retina
(Collett & Baron, 1994)
►
Bees perform bouts of lateral
flight during their return to the
goal
 Probably to help bring their
memory into register with
their current view (Collett &
Reese, 1997)
Model predictions vs. flight Behavior:
middle landmark removed
F
E
Model predictions vs. Flight Behavior:
farthest landmark removed
F
E
Model predictions vs. Flight Behavior:
removed nearest landmark
F
E
Fixed body axis and scanning flights
2060 *
1830 *
1550
Bees preferred a southern
facing body axis orientation
during their first pass through
the goal region
Bees rarely performed lateral
scanning flights near the
landmark. Circling flights were
the norm.
Snapshot overview
►
Insects memorize a visual template or “snapshot” of
landmarks they experience at important locations of their
environment
►
The memory encodes the sizes and retinal locations of
landmarks
►
Insects sequentially match this template to the
environment upon return while maintaining consistent body
alignment
►
Lateral scanning movement may aid the matching process
Snapshot overview
►
Insects memorize a visual template or “snapshot” of
landmarks they experience at important locations of their
environment
►
The memory encodes the sizes and retinal locations of
landmarks
►
Insects sequentially match this template to the
environment upon return while maintaining consistent body
alignment
►
Lateral scanning movement may aid the matching process
Testing the snapshot hypothesis in a
small scale arena environment
►
Honey bee flight behavior during other visual navigation
experiments in our apparatus appeared inconsistent with
snapshot guidance.
►
Investigated elements of snapshot navigation in a carefully
controlled arena environment
 Snapshot predictions of search behavior near the goal location
(tested by Cartwright & Collett, 1979, 1982; Cheng, 1999)
 Predictions of flights paths to the goal from distances of up to two
meters (largely untested)
 Consistent body axis orientation near the goal (Collett & Baron,
1994)
 Lateral scanning flights near the goal (Collett & Rees, 1997)