II.
Simultaneous Same-Different Discrimination
We initially trained pigeons to discriminate arrays of 16 identical icons from arrays
of 16 different icons (Young & Wasserman, 1997). These arrays actually represent two ends of a scale of display
variability,
Same displays involving minimal variability and Different displays involving maximal
variability. Pigeons were subsequently tested with novel arrays consisting of all new
icons (a transfer test) and novel arrays consisting of mixtures of same and different
icons. Through the systematic exploration of the pigeon's responding to mixture arrays, we
sought to determine the effective stimulus that was controlling its same-different report
responses.
Thirty-two highly distinguishable Macintosh icons were chosen as the total item pool;
these icons were randomly sorted into two sets of 16 icons each (Set 1 and Set 2), and
from these two 16-icon sets, 16-icon 
displays were constructed. For any given Same array,
a single icon from the appropriate set was randomly chosen and used to make up an array of
16 identical icons. For the Different arrays, all 16 of the icons of a set were used with
no repetitions. The 16 same or 16 different icons were randomly distributed over 25
locations arrayed in a 5 x 5 grid (Figure
2).
During training, pecks to the green button on Same trials and to the red button on
Different trials were correct and were reinforced with food; pecks to the red button on
Same trials and to the green button on Different trials were incorrect and were punished
by repetition of the trial until the correct response was made. Button color was reversed
for half of the subjects.
A Java applet is available on-line that demonstrates
the task (with you playing the part of the pigeon). There are some differences between the
on-line task and the actual task (e.g., the Java script alternates Same and Different
trials for demonstration purposes, whereas the original program used block randomization);
consult the original publications for procedural details.
In our group of four pigeons, acquisition of same-different responding was very rapid
and responding exhibited strong transfer to novel 16-item displays; accuracy on training
displays averaged 93% correct after 4,800 trials and 79% correct on new testing displays
created from 16 untrained icons. Relative to the 2-item displays used by prior
researchers, these 16-item displays were learned quickly, asymptotic learning was
superior, and discrimination of novel displays was strong. Although these findings were
important in their own right, we were interested in the pigeons' discriminative behavior
when they were faced with arrays that were intermediate in variability between Same and
Different.
Mixture Manipulations
The Same and Different arrays represent two ends of a continuum of variability.
Displays with some identical icons and some different icons would fall between these
endpoints. Through an examination of the pigeons' classification of displays with
intermediate levels of variability, we sought to determine the stimulus dimension being
used by our pigeons.
At the outset of these experiments, it was not clear how the pigeons would respond to
these Mixture displays. All of the Mixture displays included more than one kind of icon
thus making them nominally "different;" however, all also included repetitions
of icons which might generate a "same" classification. Additionally, it was
possible that the pigeons would show a graded response as the displays changed from being
Same to Different, thus evidencing a sensitivity to differing degrees of display
variability.
In Experiment 2 of Young and Wasserman (1997), we constructed three types of Mixture
arrays: 2 icons each repeated 8 times (2 icons x 8 copies), 4 icons each repeated 4 times
(4 icons x 4 copies), and 8 icons 
each repeated twice (8 icons x 2 copies). The Same and
the Different arrays thus represent each end of a continuum, where a Same array involves 1
icon repeated 16 times (1 icon x 16 copies) and a
Different array involves 16 icons
repeated once (16 icons x 1 copy). The accompanying Figure
3 illustrates the given points
along
the continuum of display types for one of the 16-icon sets. The
results of the experiment
are shown in Figure
4. The pigeons were
increasingly likely to choose the key associated with the Different arrays when the
Mixture array increased in variability.
In Experiment 3 of Young and Wasserman (1997), we introduced Mixture arrays that were
constructed by making a subset of the 16 icons identical to one another and the remaining
icons all different from the others. This method freed us from Experiment 2's requirement
that each repeated icon appear the same number of times, a constraint that limited the
number of possible Mixture arrays. The pigeons here observed five Mixture arrays: 2D/14S
(2 different and 14 same icons), 4D
/12S, 8D/8S, 12D/4S, and 14D/2S. These Mixture arrays
spanned the variability dimension from the Same arrays (0D/16S) to the Different arrays
(16D/0S). Representative arrays are shown in Figure
5. The res
ults of the experiment are shown in
Figure
6.
The pigeons were again increasingly likely to choose the key associated with the Different
arrays when the Mixture array increased in variability.
In order to unconfound display variability from the number of icon types in an array,
the Mixture displays in the fourth and final experiment of Young and Wasserman (1997) were
all created using four types of icons, each icon type chosen randomly on every trial from
the 16-icon set on which the pigeon was trained. The relative number of the four types of
icons was varied from those displays that we believed to involve the least variability to
those that we believed t
o involve the most variability: 13-1-1-1 (13 repetitions of one of
the icons and 1 each of the other three), 10-3-2-1, 7-5-3-1, and 4-4-4-4. Representative
arrays are shown in Figure
7. The results of the experiment
are shown in Figure
8. Given that the
number of icon types was constant (four) across the four mixture arrays, the pigeons'
sensitivity to the distribution of those icons indicates that it is the display
variability and not the number of icon types that is determining choice behavior in this
task.
The results of these experiments pointed to visual display
variability as the
key controlling dimension in our pigeons' discrimination of Same from Different displays.
Discriminative performance with arrays comprising intermediate degrees of
variability was systematically graded rather that sharply categorical. Arrays with
high degrees of variability were predominately categorized as "different,"
whereas arrays with lower degrees of variability were increasingly likely to be
categorized as "same."
Next section: An Information Theoretic Account of
Same-Different Discrimination
Figures 3, 4, 5, 6, 7,8, and 9 are adapted from:
Young, M. E., & Wasserman, E. A. (1997). Entropy detection by
pigeons: Response to mixed visual displays after same-different
discrimination training. Journal of Experimental Psychology:
Animal Behavior Processes, 23, 157-170.
