Prsentation PowerPoint

1
12345678901234567890123456789012345678901234567890
12345678901234567890123456789012345678901234567890
12345678901234567890123456789012345678901234567890
12345678901234567890123456789012345678901234567890
123456789012345678901234567890123456789
Effect of position information on the Attentional
Blink
Dominic Charbonneau Denis Cousineau. Université
de Montréal Dominic.charbonneau_at_umontreal.ca
12345678901234567890123456789012345678901234567890
12345678901234567890123456789012345678901234567890
12345678901234567890123456789012345678901234567890
12345678901234567890123456789012345678901234567890
123456789012345678901234567890123456789
Discussion Even though the subjects possessed
information to control the processing, they
cannot influence the treatment of the targets.
With these results, Shapiro et al. assertion
remains true, it appears that any target
containing pattern information produces an AB
(p.368), thus the AB phenomenon seems to be
ballistic. A strange but strong effects were
revealed. In the second condition of the second
experiment, the information about the positions
of the targets made the lag-1 sparing effect
disappear. An explanation is that the subject
knowing the position of the targets, with
practice, tried to focus on these positions,
creating a sort of window open only for the
duration of this target. Thus the entry of the
second target with the first, as some model
proposes, couldnt occur. Our results are in
favour of a system which takes time to start such
as the inhibition model or the two-stage model
(see Table 1). With practice, subjects learned to
synchronize better their temporal processing.
Abstract The Attentional Blink effect is a drop
in performance to report a second target within a
RSVP between lags 2 to 5. According to many
models, the processing order is fixed (ballistic
hypothesis), creating unavoidable bottleneck. In
the first condition, we fixed the distance
between the two targets, informing the subject on
the relative position of the second target. In
the second condition, we fixed both targets
position informing the subject on their absolute
position. Results show no reduction of the blink
effect, supporting the hypothesis of a ballistic
treatment in the AB. The only difference between
the two conditions is the absence of
lag-1-sparing in the second condition. This leads
to the paradoxical observation that trained
subjects on a specific condition are worst than
subjects trained in more variable conditions.
We can see that a common hypothesis of all these
models is the limitation of attentional
resources. Therefore, in the same way Kawahara
(2003) did, it is tempting to try to help the
subject to allocate these resources more
efficiently. One method is to give more
information on the task demands.
Experiment 1 In the first experiment, we
informed the subject of the position of the first
target. With this information, the subject could
prepare to the arrival of the first target and
save some attentional resources for the second
target.
Schema 2 procedure of experiment 1

• Method
• A RSVP of 16 digits, 14 white and two red (the
  targets)
• 12 subjects had to report the two digit at the
  end of the RSVP
• First condition
• replicate the AB effect.
• Second condition
• subject were informed of the position of the
  first target
• they knew that this position would be constant
  for all 400 trials.

To explore the limits of attention, many
researchers uses the Attentional Blink (AB)
effect described for the first time by Raymond et
al. (1992). The Attentional Blink effect is a
drop in performance to report a second target
within a RSVP between lags 2 to 5. This effect is
shown in Figure 1.

Inter-stimuli interval 100 ms
5

Model A common explanation of the AB effect
states that two stages are necessary for this
task an encoding stage where most of what we see
is encoded in cognitive terms. Within this stage
is found the universal detector (UD) of the task
demands, and a second stage memory. We used
these two stages as a framework to integrate all
models discussed previously to create an overall
portrait of the situation. Our goal is to change
these qualitative models in a quantitative one.
Afterwards, we will be able to see what aspects
of actual models are essential to this
phenomenon.
4

T1
3
Many models have been proposed to explain this
particular phenomenon. A summary is shown in
Table 1. Raymond et al. (1992) postulated an
inhibition model. According to them, when the
first target enters the encoding stage, an
inhibition mechanism starts and inhibits all
input trying to be processed. After the first
target is processed, following items can enter
the encoding stage. Since this inhibition system
takes time to start, if the following item is
immediately after the first, it can be processed
with the first target which results in the lag-1
sparing effect.
Figure 1 Control condition.

8
Lag
Lag-1 sparing

1

T2
6
T1 position either fix -or- variable (control
condition)
In part A of Schema 3,The world sends photon to
the retina which are submitted to the encoding
stage. It encodes most of the stimuli (Ward,
Duncan Shapiro, 1996). The UD looks for the
important feature of the task so that the target
can go on to memory. As shown by the coloured
lines, we placed the four discussed models. The
inhibition mode affects the input before it
enters the encoding stage (yellow). The
Similarity model affect the homunculus who
determines which targets should pass (blue). The
two stage model closes a door after stage one
(like a bridge in green) and the bottleneck model
stops input before entering the memory stage
(purple). Many effect found in the literature can
be place in this framework. The masking would
effect the encoding stage, whereas the task
switching as Visser Bishof Di Lollo (1999)
showed, affects the homunculus.
Schema 3 Framework for quantitative model
Subject to masking
A)
Figure 2 Control (1) and T1 fixed (2).
Results An ANOVA analysis showed a significant
effect of lag indicating the presence of an
Attentional Blink effect. But no significant
effect of the condition was found and the
interaction between the lag and the condition was
also not significant. Therefore, our manipulation
did not help the subject in his task.
Requires time and central attention
Shapiro et al.(1994) postulated a similarity
model. In this model, participants makes a
template of the to-be-reported targets according
to task demands. When presented to the RSVP,
participants allocate more or less weight to the
stimuli based on the fitting with the template.
Since a lot of weight is allocate to T1 and its
following item, less is available for T2 until T1
is processed. This excess of weight for T1 is
also responsible for the lag-1 sparing since the
item following the first target would benefit of
this weight. This hypothesis rise from the
findings of Raymond et al. (1992) about the
importance of the 1 item. The more the stimulus
corresponds to the feature template, the more
weight it gets , the more it will be recalled.
Chun Potter (1995) postulated that incoming
targets have to pass through a pre-attentive
stage which decides if the item has to be
processed. If so, it passes to the second stage
where it is processed. When a first item enters
the second stage, this models closes slowly an
attentional gate making other items lost by
masking of following items or time. When this
second stages is free, what is left of the second
targets can enter. Lag-1 sparing is explained in
this model by the slow closing of the gate
permitting a T11 item to enter. Jolicoeur
(1998) proposed a bottleneck model. This model
explains this AB effect by postulating that the
limited attentional resources acts as a
bottleneck. The second target would be waiting in
the pre-attentional stage (A2) that the first
target exits the bottleneck stage (B1) to be
treated. This model does not explain clearly the
occurrence of lag-1 sparing. This last model is
shown in Schema 1.
W O R L D
Encoding stage
Memory
Retina
UD
Chance
Can be switched
Experiment 2 In the second experiment, we fixed
the distance between the targets in the first
condition and the positions of both targets in
the second condition. By this manipulation, we
wanted to look at a possible benefit resulting
from information on the second target. This
information could be relative to the position of
the first target or absolute.
B)
Lag-1
T1 
Lag 1
Lag 2
Lag 3
Lag 4
T1 in memory
T2 in memory
UD ON

• Method
• The method is the same has in experiment 1 except
  what follows.
• First condition the lag was known to the
  subjects.
• Second condition both T1 and T2 positions were
  known to the subjects.

Accumulator ready to restart
UD Going OFF
Schema 1 The bottleneck model
A1
B1
C1
Results An ANOVA showed a significant
interaction between the condition and the lag.
The simple effects showed a significant effect at
lag 1 indicating that at lag 1, the second
condition is significantly lower than the first
condition. The lag-1 sparing effect has almost
vanished in the second condition.
Figure 3 Lag fixed (1) vs both T1 and T2 fixed
(2).
In part B, we can see our framework in a more
quantitative way. The UD acts as an accumulator
which permits targets to enter the memory. The
length of the box to enter memory determines the
accuracy of the target it is on. In the example
above, if T2 is at lag 2, it will be poorly
recalled because the accumulator is not ready to
restart. However, if T2 is at lag 1, it will
partly enter memory along with T1 explaining the
lag-1sparing effect. The model is probabilistic
permitting a smooth curve. This framework opens
the door to an integrative view of the literature
on the AB effect, and to quantitative tests of
the models.
Slack
A2
B2
C2
Table 1 Summary of AB models

Chance