Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Sequence learning under incidental conditions [poster]
1. Psychology
F.
Yeates,
F.
W.
Jones,
A.
J.
Wills,
R.
P.
McLaren
&
I.
P.
L.
McLaren
Contact
Fayme
Yeates:
fy212@exeter.ac.uk
Introduc8on
In
general,
Group
1
show
faster
sequence
learning
than
Group
2
Sequence
learning
studies
employing
serial
reac8on
Propor8on
of
Error
Difference
Scores
for
Experimental
and
Control
Par8cipants
for
Different
Sequences
over
Training
8me
(SRT)
tasks
provide
evidence
for
a
disPncPon
Blocks
under
Incidental
Condi8ons
between
explicit
knowledge
and
implicit
Group
1
0.15
Group
2
0.15
performance.
This
disPncPon
is
not
uncontested,
Condi8on
Experimental
F(1,26)
=
31.99
however,
and
there
are
those
that
argue
against
it
Control
Inconsistent
Errors
-‐
Consistent
Errors
Inconsistent
Errors
-‐
Consistent
Errors
p
<
.001
(e.g.
Mitchell,
De
Houwer
&
Lovibond,
2009).
0.1
Condi8on
*
Block
0.1
F(16,416)
=
3.87
Jones
and
McLaren
(2009)
suggested
that
evidence
p
<
.001
for
dissociable
learning
processes
would
be
found
if
0.05
0.05
learning
of
the
same
sequences
was
shown
to
be
qualitaPvely
different
under
incidental
and
intenPonal
condiPons,
whilst
controlling
for
the
0
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
sequenPal
effects
that
may
have
contaminated
Condi8on
earlier
SRT
research.
Block
F(1,29)
=
2.03
-‐0.05
-‐0.05
Block
p
>
.1
Jones
&
McLaren
(2009)
trained
parPcipants
on
Experimental
Condi8on
*
Block
sequences
constructed
from
triplets
following
an
F(16,464)
=
.72
-‐0.1
-‐0.1
Control
p
>
.7
exclusive-‐or
rule.
Different
triplet
subsequences
were
learnt
to
different
extents,
and
the
paern
of
Over
training
blocks,
experimental
subjects
in
Group
1
improved
their
RTs
and
error
rates
compared
to
controls
learning
was
different
again
for
parPcipants
that
on
trials
predicted
by
their
Group
rule
–
that
the
current
sPmulus
appears
in
the
opposite
locaPon
to
the
one
were
instructed
to
look
for
sequences
and
apply
a
two
trials
back
-‐
compared
to
those
that
were
inconsistent
with
this
rule.
strategy
(intenPonal)
compared
to
those
who
were
instructed
to
simply
respond
as
quickly
and
At
test
Group
1
show
be/er
learning
than
Group
2
accurately
as
possible
(incidental).
These
effects
Reac8on
Times
and
Propor8on
of
Error
Difference
Scores
for
Experimental
and
Control
Par8cipants
for
Group
1
and
2
over
were
explained
by
hypothesis-‐tes8ng
and
Training
Blocks
under
Incidental
Condi8ons
Reac8on
Times
0.1
Errors
associa8ve
learning
accounts,
respecPvely.
The
last
50
was
based
on
modelling
using
the
Augmented
SRN.
*
*
40
0.08
Our
aim
was
to
further
invesPgate
the
nature
of
*
Inconsistent
errors
-‐
Consistent
errors
30
0.06
sequence
learning
in
SRT
tasks
under
Jones
and
Inconsistent
RTs
-‐
Consistent
RTs
McLaren’s
(2009)
incidental
condiPons.
20
0.04
10
*
0.02
*
Method
64
parPcipants
took
part
in
a
two-‐choice
SRT
task
0
0
over
two
sessions.
ParPcipants
were
either
Group
1
Group
2
Group
1
Group
2
-‐10
-‐0.02
instructed
to
respond
as
quickly
and
accurately
as
possible
to
the
sPmuli
(a
circle
filling
on
the
led
or
-‐20
-‐0.04
right
of
the
screen).
Responses
were
spaPally
compaPble
key
presses.
-‐30
Experimental
-‐0.06
Experimental
-‐40
Control
-‐0.08
Control
Training
Sequences
Experimental
parPcipants
perform
beer
than
controls
at
test.
Consistent
with
the
training
data,
both
reacPon
Group
1
Group
2
Control
Pme
and
error
data
show
a
greater
difference
between
experimental
and
control
when
learning
Group
1
RRL
RRR
RRL
RRR
sequences.
Over
acquisiPon
and
at
test,
Group
1
demonstrates
more
learning
than
Group
2.
*
p
<
0.01
RLL
RLR
RLL
RLR
LLR
LLL
LLR
LLL
Modelling
Incidental
Learning
LRR
LRL
LRR
LRL
Mean
Squared
Error
Difference
Scores
for
Experimental
and
Control
Simula8ons
for
Jones
&
McLaren
(2009)
Group
1
and
2
in
Test
for
two
versions
of
the
Augmented
SRN
found
that
incidental
Training:
Experimental
parPcipants
were
trained
on
0.1
AugSRN
0.15
AugSRN
(-‐
Fast
Learn
Rate)
performance
on
their
one
of
two
sequences,
where
the
first
sPmulus
task
was
successfully
Inconsistent
MSE
-‐
Consistent
MSE
0.08
locaPon
in
a
triplet
would
predict
the
third.
modelled
by
the
Inconsistent
MSE
-‐
Consistent
MSE
0.1
Therefore,
the
sPmulus
locaPon
of
any
trial
could
be
0.06
augmented
simple
in
the
opposite
locaPon
(Group
1)
or
the
same
recurrent
network
0.04
locaPon
(Group
2)
as
the
sPmulus
locaPon
two
trials
0.05
(AugSRN;
Cleeremans
&
back.
Experimental
subjects
could
predict
the
0.02
McClelland,
1991)
–
locaPon
of
the
current
trial
on
two
thirds
of
training
giving
further
weight
to
0
0
trials.
Control
groups
were
trained
on
an
associaPve
account
of
Group
1
Group
2
Group
1
Group
2
pseudorandom
sequences
made
up
of
an
equal
-‐0.02
Experimental
Experimental
performance
in
this
number
of
all
of
the
possible
triplets.
Thus,
they
-‐0.04
Control
-‐0.05
Control
condiPon.
could
not
predict
the
locaPon
of
the
current
trial
on
Difference
Difference
the
basis
of
the
locaPon
of
the
trial
two
An
augmented
SRN
with
the
same
parameters
as
Jones
&
McLaren
(2009)
does
not
perform
in
the
same
way
as
presentaPons
previously.
Both
groups
received
the
humans,
showing
more
learning
in
Group
2.
When
the
fast
weights
of
the
augmented
SRN
were
disabled,
this
same
amounts
of
sPmuli
in
each
trial
posiPon,
and
paern
reversed
and
beer
simulated
human
learning.
extraneous
sequenPal
effects
were
controlled
for.
Conclusions
Test:
All
subjects
received
five
test
blocks
of
We’ve
found
that
aTer
taking
performance
effects
into
account,
people
are
be/er
at
learning
pseudorandom
sequences
as
described
for
controls
some
types
of
sequences
over
other
sequences.
This
is
not
easily
modelled
by
the
Augmented
above
ader
the
training
phase.
SRN.
Cleeremans,
A.,
&
McClelland,
J.
L.
(1991).
Learning
the
structure
of
event
Jones,
F.
W.,
&
McLaren,
I.
P.
L.
(2009).
Human
sequence
learning
under
incidental
Mitchell,
C.
J.,
De
Houwer,
J.,
&
Lovibond,
P.
F.
(2009).
The
proposiPonal
nature
of
sequences.
Journal
of
Experimental
Psychology:
General,
120,
235-‐253.
and
intenPonal
condiPons.
Journal
of
Experimental
Psychology:
Animal
Behavior
human
associaPve
learning.
Behavioral
and
Brain
Sciences,
32,
183-‐246.
Elman,
J.
L.
(1990).
Finding
structure
in
Pme.
CogniCve
Science,
14,
179-‐211.
Processes,
35(4),
538-‐553.