1. Conclusion
• In education systems where time and resources are precious, it could be incredibly beneficial to be able to predict how much
different measures will improve based on working memory skills so that interventions could be targeted to more specific groups.
• Currently the data suggest that Working Memory improvement on the training tasks might be predictive of improvement on Dot
Counting Accuracy (WM), Reading Fluency, and Auditory Word Attack.
Working Memory Training: Predicting Transfer
Chelsea M. Parlett 1 , Masha R. Jones 1, Maria Jesus Maraver 2, Maria Teresa Bajo 2, Carlos J. Gomez-Ariza 3, Jacky Au 1 4, Martin Buschkuehl 4, Susanne M. Jaeggi 1
1 University of California – Irvine, 2 University of Granada, 3 University of Jaen, 4 MIND Research Institute
References
MethodsAbstract
Literacy is important, and children with low reading skills often suffer in more subjects than
just English. Especially in the information rich culture of the 21st century, the ability to read
and comprehend large amounts of data is critical for success Children with dyslexia account
for approximately 17% of American youth. Reading comprehension involves considerable
working memory (WM), and WM deficits are among the major underlying factors driving
reading difficulties in dyslexia. Our ongoing study is a pre/posttest randomized controlled
trial with 4th and 5th grade dyslexic and ND students, randomly assigned to the WMT or
control condition. Children train on three computerized WMT games or on control games for
10 sessions lasting 15 minutes each and take pre- and posttests of WM and reading ability.
Preliminary results suggest that training works better for high skill readers than low skill
readers. Our hope is that this would become something that is easily accessible online to kids
everywhere. This could also allow us to gather larger amounts of data in order to look at more
relationships between reading and working memory.
• Klingberg, T. (2010). Training and plasticity of working memory. Trends Cogn Sci. 14(7):317-24.
(2) Jaeggi S. M., Buschkuehl M., Shah P., Jonides J. (2014). The role of individual differences in
cognitive training and transfer. Mem. Cogn. 42 464–480.
• Authors’ note: MB and JA are employed at MIND Research Ins8tute, whose interest is related to this work. SMJ
has an indirect financial interest in MIND Research Ins8tute.
§ Training gains were calculated for the Working Memory Training (WMT) group by subtracting the average of the first two rounds from the average of the last two rounds (all games were averaged together).
§ A linear regression was then fit with 6 of our relevant outcome measures.
§ Future models could also take into account starting level since there could be a ceiling on Working Memory improvement.
§ Training improvement may have different predictive power for different types of measures. For example, Spelling ability is not expected to improve with WMT, and in fact there is very little predictive power of WMT gains on
Spelling improvement.
§ It is also of interest that Dot Counting and Updating are near transfer measures whereas reading fluency (RF) and Reading Efficiency (RE) are far transfer measures. What kind of measures are more predicted by WMT
improvement?
§ Further analysis could also help us see whether other factors such as gender, age, or SES could improve the predictive power of WMT improvement on different measures.
§ 34 students (4th and 5th grade) trained for 15 minutes every day on three
different working memory tasks (right) for 10 days each
§ Before and aLer training, subjects took a baOery of working memory and reading
tests to measure improvement.
§ Data Collec8on is ongoing.
§ Transfer of working memory training to reading ability will be assessed upon
comple8on of data collec8on.
Subject Training
Improvement
Transfer to non-
trained tasks
(near and far)
Prediction models
Contact
cparlett@uci.edu || m.jones@uci.edu || mjmaraver@ugr.es
Dot Coun8ng
P = 0.0075
Dot Coun8ng is a Span Task that
is measured by having subjects
count and remember sets of
dots.
r2=.23
Reading Fluency
P = 0.0214
Reading Fluency measures
Reading Comprehension and
speed.
r2 = 0.21
Auditory Word
AOack P = 0.077
Auditory Word AOack is
adapted from a sub-measure of
the Woodcock Johnson test that
measures non-word decoding.
r2 = .10
Upda8ng
P = 0.2872
For the upda8ng task, subjects
remember the 3 largest numbers
in an audio list. This requires
them to con8nually update
which numbers are the largest.
r2 = 0.05
Spelling
P = 0.5062
Our spelling task measured
subjects’ ability to spell
progressively harder words. It is
not expected to improve as a
result of Working Memory
training.
r2 = 0.02
Reading Efficiency
P = 0.7763
Reading Efficiency is measured
on complex sentences (Hits/Hits
RT)
r2 = 0.003
Far Transfer
Near Transfer
Control Transfer