Professional Documents
Culture Documents
Dylan Gration
The University of Western Australia
Memory of Processing Performance under Differing Conditions
Abstract
The study recites empirical evidence on levels of processing as well as examining the
difference between phonemic and semantic encoding, and the consequent effect this has
memory recall in general. It also looks at the general population’s accuracy of performance
estimation on memory recall tasks, and how most people in general will overestimate their
performance and actual abilities. The greater recall ability when information is encoding
via a semantic level of processing is shown in experiment’s results as well as the large
difference between estimated recall performance and actual recall performance. Implications
The recall of memory has been of keen interest for decades now among psychological
scholars, in particular the actual type of processing one engages in when recalling such
memories. Consequently, research has shown that there are levels of processing one engages
in, and this is particularly evident in short-term memory. The degree to which one is able to
recall certain memory has seen to have been established by the level of processing in which
information is encoded, rather than memory retrieval (Craik and Lockhart, 1972). This level
of processing ranges from superficial levels of encoding to far deeper levels of encoding and
Craik (2002) mentions this notion of encoding specificity, in particular the phonemics versus
semantics.
This research has shown that there is a greater potential for retrieval with deeper encoding,
and that a greater level of memory recall is also associated with a retrieval environment,
such as specific retrieval cues being present (Craik, 2002). Such deep encoding traces are
considered even more memorable if there is an equation of both encoding and retrieval at
a semantic level, and Craik (2002) states that Morris et al. data showed this combination
It is important to note that this level of processing however is not simply a set of levels
hierarchy in which the beginning of sensory encoding is of a sequential nature, with a transfer
Levels of processing are also attributed to recall performance among the general population.
While Craik has found that “cued recall was better when the target items had been presented
in a category judgment than when the targets had involved rhyme judgments” (Lewandowsky
& Hockley, 1987), most of the population seem to be quite unaware of the levels of
processing and how this in turn may effect their recall performance. This can be seen in
Memory of Processing Performance under Differing Conditions
the relationship of memory self-efficacy with memory task predictions, explored by Shaw
and Craik in 1989. In their experiment on performance prediction in memory tasks among
adults, the results exemplified the fact that the subjects’ predictions of their performance
even after being informed of the manipulations before the task began.
Hertzog and Dixon found other factors that generally affect one’s estimate of their
performance in memory tasks such as task appraisal, memory monitoring and again, memory
self-efficacy.
My study considers theoretical implications of the level of processing and global accuracy
predictions in memory, in light of empirical evidence stating that many are not necessarily
Hultsch, Butterfield et. Al plus Lovelace and Marsh are generally consistent in showing a
lack of age differences when predicting accuracy and in relation to memory self-efficacy.
My study instead aims to look at the performance predictions and actual performance
when participants are explicitly foreclosed of the type of encoding beforehand, which in
due relation affects the level of processing to be used by the participant. The study aims to
also discover where there are measurable differences between the level of recall between
phonemic and semantic encoding and if performance recall differs between the two types of
encoding, it also seeks to add to existing research that people generally overestimate their
performance.
I hypothesized that deeper levels of processing (and more elaboration when encoding) will
result in greater recall of to-be-remembered items when conducting the experiment. As such
I predicted that those participants in the semantic group will have better actual recall than
those in the phonemic group. I also hypothesized that most people won’t realize the impact
that levels of processing has on their recall performance and predicted that their will be
Memory of Processing Performance under Differing Conditions
no substantial difference in the estimates of their recall performance for the two differing
conditions. I lastly hypothesized that people usually overestimate their actual abilities of
memory recall and furthermore predicted that if people overestimate their memory capacity
in the same way they overestimate performance in other areas, then I expect estimates of
recall performance will be higher than actual recall performance among participants.
Method
For my study, the participants involved were the whole first year cohort of Psychology
students studying at UWA, which numbered 992 participants. No account was taken of age
or gender, but most are assumed to fall within the 17-21 years of age bracket and all are
experiment as a requirement of their PSYC1101 course. The materials used in the experiment
firstly required a computer as part of the task. The task was a program for the computer,
and was structured with 26 word pairs of stimuli. The stimuli in turn were written in blue
and were of size 14 font in Times New Roman. These stimuli were placed in the center of a
white background for the program. The actual procedure of the experiment was as follows.
these were appeared as “rhyming” and “category” on their computer screens. They were
then directed to the website address of the task on their computers where they were first
required to enter their name, then select the category that they had been randomly assigned to
The first phase of experiment also included six practice trials of stimuli, where they would
be required to state whether the two words of the ‘same’ category or ‘different’, they would
be told whether they were right or wrong after each answer and this section had a 30 second
time limit for responses to be made. After the practice, the first phase, learning phase, began
Memory of Processing Performance under Differing Conditions
with the 26 word pairs appearing on screen again needing to be stated whether they were
The second phase, or recall phase, of the experiment then asked participants to now
remember a corresponding word of each 26 stimuli pair as one had been taken away, while
participants were given as long as they’d like to respond but were required to enter at least
one word. If participants could not recall the exact spelling of a word, the program would
accept approximate spelling. After completing this phase, participants were finally presented
with their individual results of what they scored percentage wise for both phase one and two,
Results
Table 1
Means (SD) for the Phonemic and Semantic conditions on memory recall performance
Phonemics Semantics
Table 1’s mean for estimated recall are both very similar, while the actual recall presents
quite a large difference by more than 10% in comparison to the 2% difference in estimated
recall. Standard deviations are quite similar. The Cohen’s D was .61 for the semantic
Table 2
Mean SD
The mean of the estimated and actual recall are at almost a 30% difference, while the
standard deviation is relatively similar by about 2%. The Cohen’s d was calculated at 1.44 for
Discussion
The results of table 1 explicitly show that the semantic group recalled more word pairs than
in comparison to the phonemic group. This was quite a substantial difference between the
performances of the two groups as represented by the Cohen’s d figure of .61. This confirms
the hypothesis and adds to previous research (Craik, 2002) that deeper levels of processing do
The results of table 1 also confirm that there was no major difference between the conditions
on the estimated amount of word pairs participants believed they’d recall, even though
semantic encoding allows for greater recall. The Cohen’s d was .09 representing no
meaningful difference whatsoever between the two groups and supports they hypothesis, and
Table 2 again confirms the experiment’s hypothesis in that people generally overestimate
their actual abilities. There was a substantial 28% difference between people thought they
could recall and what actually could recall. This supports other empirical evidence focusing
on other predictions of performance such as that of Dunning, Johnson, Ehrlinger and Kruger
who asked sophomore college students to predict how well they’d do on their exam.
This experiment could be improved by allowing for a greater sample size of participants, and
possibly by the recording of ages to record memory recall performance differences between
young and older adults. This study I believe adds to existing empirical evidence (Chaik,
2002; Craik & Lockhart, 1972) that semantic encoding allows for greater memory recall.
Memory of Processing Performance under Differing Conditions
References
Craik & Tulving (1975), Depth of Processing and the Retention of Words in episodic
Processing, Journal of Experimental Psychology: General, 104(3) pp. 268-294
Hertzog, Dixon & Hultsch (1990), Relationships between Metamemory, Memory Predictions,
and Memory Task Performance in Adults, Psychology & Aging, 5(2) pp215-227