Higher-Level Cognition - Computational Explorations in Cognitive Neuroscience - page 381

135−

Open a training log using View , TRAIN_LOG , and
the press Train on the control panel. You might also
want to turn off the Display on the network to speed
things up.
We next explore the differential weight strengths for
the two pathways that develop as a result of training.
Although the observed weight differences are not huge,
they are enough to produce the behavioral effects of
word reading being dominant over color naming.

, !

130−

Color Naming

125−

120−

115−

Word Reading

110−

0 1 2

Control Conflict Congruent

Use r.wt , and click on the output units — note that
the word reading pathway is stronger. Similarly, press
on the g and G hidden units in sequence, and note that
the G (word reading) hidden unit has stronger weights
overall.

, !

Figure 11.6: GraphLog results for the Stroop task in the
model. Compare this pattern of results with those for people
as shown in figure 11.3.

Question 11.1 (a) Report the weights for the g and G
hidden units from their respective input, PFC, and out-
put units (you need only report the gr output weight).
(b) At which layers in the network are the differences
greatest? (c) Can you explain this in terms of the error
signals as they propagate through the network?

You may need to press the Init button on the graph
log to rescale it to better fit the display.
If you compare this with the human data shown in
figure 11.3, you will see that the model reproduces all
of the important characteristics of the human data as
described previously: interference in the conflict condi-
tion of color naming, the imperviousness of word read-
ing to different conditions, and the overall slowing of
color naming.
Now, we can single-step through the testing events to
get a better sense of what is going on.

, !

Basic Stroop Task

Now, let's test the network on the Stroop task. First, we
will view the testing events.

Press View , WORDREAD_EVENTS .
You should see the control, conflict, and congruent
conditions for word reading, all of which have the word
reading PFC task unit clamped ( wr ). All patterns have
the R word unit active. Control does not have any active
color units, conflict adds the g (green) color unit active,
and congruent adds the r (red) color unit.

First, be sure your Display is on for the network,
and that you are viewing act (activations).

, !

Then, do

, !

StepTest .
Each StepTest of the process will advance one
step through the three conditions of word reading in or-
der (control, conflict, congruent) followed by the same
for color naming. For the word reading conditions,
you should observe that the corresponding word read-
ing hidden unit is rapidly activated, and that this then
activates the corresponding output unit, with little ef-
fect of the color pathway inputs. The critical condition
of interest is the conflict color naming condition.

Iconify the word reading environment, and then
press View , COLORNAME_EVENTS to observe a similar
pattern of inputs for color naming. Iconify that environ-
ment before continuing.
Now, we can actually test the network.

, !

Iconify the training log, and then open a testing log
using View , TEST_LOG . Then press TEST .
You will see the reaction times (cycles of settling)
plotted in the upper graph log, which should resemble
figure 11.6. The data points are plotted at X-axis values
of 0, 1, and 2, corresponding to the Control, Conflict,
and Congruent conditions, respectively.

, !

Question 11.2 (a) Describe what happens in the net-
work during the conflict color naming condition, pay-
ing particular attention to the activations of the hidden
units. (b) Explain how this leads to the observed slow-
ing of reaction time (settling).