1. Children’s Cognitive
Research Lab
CCRL
Children’s Cognitive
Research Lab
CCRL
Stability of Reasoning: Measuring Degree of Order in
Adults’ Predictions in Sinking Objects
Ramon Castillo, Samantha Hinds, Ashlee Kromski and Faculty Mentor, Heidi Kloos
OVERVIEW
INTRODUCTION
METHOD
Participants
• 104 undergraduate Psychology students at U.C.
• 25 participants in each condition
ACKNOWLEDGEMENTS
REFERENCES
DISCUSSION
Materials
• 12 unique transparent jars (large, medium, small)
• Aluminum disks (1 cm high, 4 cm in diameter, 43 g
each)
Procedure
• DirectRT Precision Timing Software (Version 2012) was used to administer 90 trials during the pre-test, 180- trials during
training, and 90 trials during the post-test yielding 360 trials total.
• Participants were asked to choose which of the two jars would sink faster in the sink-faster condition. And in the sink-slower
condition, the task was to choose which of the two objects would sink slower. Both phases consisted of a series of prediction
trials, the difference being that the participant was given feedback during training, but not during actual testing (pre & post-test).
Bartlett, F. C. (1932). Remembering. Cambridge
University Press: Cambridge, England.
Castillo, R. D. & Kloos, H. (2013). Can a Flow
Network Approach Shed Light on Children’s
Problem Solving? Ecological Psychology, 25(3)
281-292.
Kohn, A. S. (1993). Preschoolers’ reasoning about
density: will it float? Child Development 64,
1637–1650.
RESULTS
•Accuracy: •Amount of Explained Variance:
•Degree of Order:•Robustness:
The experimental phase consisted of eight segments of 90
trials each, the middle four segments consisting of feedback
trials. In a 2 by 2 between-group design, we varied the task
instructions and the distribution of trials to explore
variability in performance. Specifically, participants had to
determine either the faster sinking object (sink-faster
instruction) or the slower one (sink-slower instruction). And
the different types of trials were either evenly distributed
within a segment (even-distribution condition) or there was a
higher number of critical versus non-critical trials (uneven-
distribution condition).
• Stability is essential for us to understand our
surroundings. We have the ability to search our minds for
past patterns of stability, which can help us predict events
in current situations. These predictive reasoning’s are
known as representations, beliefs, schemas, strategies or
goals. The existence of these entities confirmed in
numerous studies, ranging back to the earliest research in
cognition (e.g., Bartlett, 1932).
• Mental structures that give rise to stable predictions in a
reasoning task are different in many ways from the
processes that give rise to stable eco-systems. However,
there are also important similarities between the two
systems (Castillo & Kloos 2013). The stability of eco-
systems depends on the flow of energy among multiple
species, none of which makes up the stability of the system
alone. Stability comes from the holistic configuration of
interdependent components. Similarly, stable mental
structures are likely to be holistic configurations that
depend on interacting elements of knowledge that change
their nature as a result of a larger whole.
• Given the similarities between eco-systems and mental
structures, we propose to apply eco-dynamic matures of
network stability to describe stability in mental structures.
• The goal of the current study was to apply these
macroscopic measures of stability and change to observe
the stability of adults’ predictions in a simple reasoning
task.
• The specific task involved adults’ speculations about how
different objects behave when submerged in water.
Reasoning in regard to sinking objects has been studied
numerous times (e.g., Inhelder & Piaget, 1958; Smith et
al., 1985; Kloos & Somerville, 2001; Kohn, 1993), and
performance is typically attributed to incorrect beliefs. This
leads to the studying of stability of performance, especially
in the context of corrective feedback.
• The idea outlined by these experiments is quite simple. When
the percentage of variance explained was high during the
pretest segments, this percentage was attributable to a common
pattern of response, clearly identifiable and stable in the group
of participants. Under the effect of feedback, this common
structure was diluted and this change was perceived as a
reduction of the explained variance. Finally, when a new
configuration of responses emerged, the explained variance
increased again. Under feedback training participants generated
different dimensional organization depending on if they were
predicting whether the object would sink faster or slower.
• Our study is the first to investigate the degree to which this
more traditional measure of amount of explained variance
coincides with eco-dynamic measures of degree of order and
robustness. We found that these measures could track the
fluctuation observed in the first dimension extracted by
CATPCA: The degree of order and robustness had a positive
correlation with the Eigen value of the first dimension. This
means that every time a structure emerges, levels of order and
system robustness are high too.
• The measures of degree of order and robustness come with far
less conceptual baggage than traditional concepts of beliefs,
representations, or the like. This is because they capture
stability without depending on auxiliary hypotheses and
constructs to explain a system’s memory or its drive to survive.
The analogy with eco-systems forces us to think of a different
kind of explanation for system robustness, one that moves away
from the trap of computer metaphors, pointing towards a theory
that could capture the essence of biological systems.
• The eco-dynamic measures allow us to leave behind questions
about the exact content of stable structure. They incorporate
informational measures of uncertainty and average mutual
information that focus on the nature of the system network, not
the details of the components that make up the network. Thus,
eco-dynamical measures allow us to incorporate heterogeneity
of mental structures into explanations of cognitive phenomena,
without trapping the discussion in what exactly it is that adult
know or not know.
We would also like to thank Michael Richardson,
Catherine Schneider as well as all participants who took
part in this study.