Transitive Inference Dependent on Difficulty and Informedness

1
Transitive Inference Dependent on Difficulty and
Informedness Anna Berg, Anthony Greene,
Kelli Pellmann Department of Psychology,
University of Wisconsin, Milwaukee
University of Wisconsin- Milwaukee
Introduction The transitive inference (TI) task
consists of premise pairs which may be encoded
discretely or as a linear hierarchy. By viewing
presentations of two stimuli wherein one is
correct, participants are trained on the premise
pairs (i.e. AB, BC, CD, DE where means
choose over). Note that context determines the
correct choice (e.g. C is correct in the context
of D but incorrect in the context of B). A
seemingly efficient way to encode the premise
pairs is to assemble them in a superordinate
hierarchy (ABCDE). Formation of the
hierarchy is tested with inferential pairs such
as B?D. Correctly selecting B over D may
indicate a contextually learned relationship, or
may be consistent with an anchoring end item
approach, or symbolic distance effect. Previous
research (Greene, Spellman, Dusek, Eichenbaum,
Levy 2001) shows that a 5-item TI task may be
solved when participants are not aware of the
hierarchy. Increasing task difficulty may
necessitate the use of conscious strategies. The
present experiments are the first to compare TI
performance across three levels of difficulty
with participants either informed or uninformed,
in order to better understand the breadth of the
TI task in humans. Experiment The experiment
consisted of three separate difficulty levels
utilizing 5, 6, or 7 items to be assembled in a
superordinate hierarchy. Each difficulty level
contained an informed and an uninformed
group.
An examination of the individual participants
data show no systematic relationship between
performance on any transitive pairs. For example
in TI6 those who correctly solved BD were not
more likely to correctly solve the easier BE
problem (t .618, p .510). Some participants
perform well on the beginning transitive pairs,
but not the end transitive pairs and vice versa.
The variability is so great, that it is difficult
at this point to forcefully argue for any single
method of learning the task. Contextual
transitivity predicts performance across
transitive pairs, while the symbolic distance
effect predicts performance only on end
transitive pairs. It should be noted that both
theories predict relative weaker performance on
inner transitive pairs.
Participants were in either the informed or
uninformed condition for their level of
difficulty. Participants viewed pairs of items
adjacent in the hierarchy (e.g., A?B and B?C) in
five training blocks, and were then tested on
non-adjacent items (e.g., A?E, B?D). Pair
presentation order in training was
counterbalanced across subjects. Hiragana
symbols were used as figures. Hiragana figures
used in hierarchy F Y U O Q n v A
B C D E F G Example of
screens viewed by a participant Results Diffi
culty and Informedness Informed vs. uninformed
performance was found to be dependent on the
condition of difficulty, F 5.099, p 0.008 for
the interaction. Participants in the 5 item
condition performed similarly in both the
informed and uninformed conditions, F 1, p
n.s., however, in the 6 item task, participants
performed differently on informed vs. uninformed
conditions, F 19.631, p task, the difference approaches significance. F
3.289, p 0.080. Awareness While
participants may become serendipitously aware of
the task, serendipitous awareness did not occur
more frequently in easier conditions.
For the subjects in the uninformed condition,
serendipitous awareness was not correlated to
performance on the transitive pairs (TI5 r
0.354, p 0.259 TI6 r -0.300, p 0.154
TI7 r -0.173, p 0.419. This finding is
consistent with previous literature for the five
item (TI5) task. Transitivity as a Function of
Difficulty Planned comparisons reveal that the
most distant inferential pairs show the most
consistent performance.
Specifically, in TI7 BF shows the highest
performance of all transitive pairs (F 7.966, p
.008). Though not significant, BE appears to
show high performance, in TI6 as well. This
appears to support the symbolic distance effect
(Wynne, 1995), which is taken by some as evidence
that inference is mediated by transfer of value
of end items. However, if value transfer were
occurring, one would predict that those
participants who solved the more difficult inner
pairs would necessarily have solved the outer
pairs as well.
F Y
Incorrect

• Conclusion
• Replicating previous literature (Greene, et al,
  2001), transitive performance is not dependent on
  serendipitous awareness.
• Transitive Performance is dependent on the level
  of difficulty whether in the informed or
  uninformed condition.
• Transitive pairs closer to end items showed
  higher performance and smaller variability than
  those in the middle of the hierarchy.
• Further research is required to fully understand
  transitivity in the absence of awareness, and the
  underlying mechanisms which may explain such
  transitive performance.

Performance in fifth block of training on premise
pairs from participants scoring 63 or higher on
all premise pairs (training pairs) in the block
preceding test.
Performance at test from participants scoring 63
or higher on all premise pairs (training pairs)
in the block preceding test.
Informedness effects performance in the six and
seven item tasks