Norman J. Slamecka has developed experiments to investigate The Generation Effect phenomenon, which suggests that information might be better remembered / recalled if this information is generated. For example, if an individual is subjected into generating the word “party” from APRTY, the person will be more likely to recall the words instead of if that was merely read.
There has been a lot of research regarding this phenomenon involving the generation effect on high and low frequency words, and this experiment will attempt to measure and discuss if there is a generation effect for Low Frequency Words.
Our prediction is that there is a generation effect on Low Frequency words, and the results suggest, as per our prediction that there is in fact a generation effect on Low Frequency words rejecting the null hypothesis.
There has been a lot of research and discussions regarding the Generation Effect triggered by Slamecka’s research “The Generation Effect: Delineation of a Phenomenon – Norman J. Slamecka & Peter Graf – 1978”.
First it is necessary to understand the definition of the “Generation Effect”, which can be simply explained as “the retention advantage for self-produced over read items”
Slamecka was driven to investigate the Generation Effect based on the “informally expressed sentiment that there is an especial advantage to active or effortful involvement of the person in the learning process”.
Slamecka claimed at the time that there was not enough research in the field to explain to what extent this general notion had an observational or an unreliable, casual base. As a result of the lack of literature to discuss and explain this question, and in order to understand this phenomenon, Norman J. Slamecka & Peter Graf carried out an experiment to compare memory for generated and read words.
In the experiment Slamecka observed that there was a much superior performance in recalling generated words compared to read words, and they concluded that the generation effect exists and, since then many psychologists have argued on this subject and developed many experiments and researches, although some are more relevant to this experiment than others.
In 1985 James S. Nairne, Constance Pusen and Robert L. Winder Jr. have published a very renowned journal discussing the Representation in the mental lexicon: Implications for theories of the generation effect.
Their findings suggest that representation is not sufficient, although necessary condition for the generation effect, and that we may also consider how much is the cue in the memory system related to the generated word.
The advantage of retention of generated over read words has been reported several times in numerous studies. However a very influential research by Tyler, Hertel, McCallum & Ellis suggested that there is a cognitive effort in memory, which can be defined as “the engaged proportion of limited capacity”. Their hypothesis was that the subjects could recall words better as a result of high effort degree than a low effort.
Furthermore, Nairne & Pulsen (1985) suggested that low frequency words, like non words, could be better recalled if those were generated.
Following those researches, we have designed an experiment to try and measure if there is in a fact a generation effect for low frequency words, and if there is a significant difference between high and low frequency words. The results show that the participants recalled significantly more generated words than read on both conditions, low and high frequency words, which suggests that there is a generation effect between read and generated words, which will be explored on this research.
158 undergraduate psychology students were invited to undertake this experiment, in which they were randomly presented with two types of stimuli, low and high frequency words, and were asked to recall as many words as possible a few seconds later.
The results suggests that there is a significant generation effect on both conditions, however what this experiment is attempting to measure is if the generation effect is greater on low frequency words compared to the generation effect on high frequency words.
This experiment was designed to measure the generation effect between groups for low and high frequency words amongst undergraduate psychology students at Birkbeck College in London.
Based on previous researches, it is expected that there is a significant generation effect for Low & High frequency words.
Subjects: The subjects were 158 undergraduate students of the second year of Psychology at Birkbeck College with mixed age, gender and nationality. All participants voluntarily accepted the task to undertake this experiment, being randomly assigned to two different conditions of stimuli.
88 subjects had English as a second language and 67 subjects had English as first language, and therefore minor spelling mistakes have been allowed on data collection.
Design: Between groups was used to execute this experiment. One group was presented with a list of low frequency words and amongst those words there was a mixture of read and generated words, whilst the second group was presented with a list of high frequency words, which contained also both read and generated words. It means that low frequency words are classed as words that are not used frequently on daily tasks and routines, while high frequency words are frequently used.
In total fifty-six words were used, two lists of twenty-eight concrete nouns, one for high frequency and the second for low frequency words were used for this experiment, which were extracted from the Kucera & Francis word count (1967).
The lists have been further split in four other groups: High Frequency 1, High Frequency 2, Low Frequency 1 & Low Frequency 2.
High Frequency 1 & 2 had 28 words each with both generated and read words. Group 1 for High Frequency words had 14 words with the first two letters reversed and underlined (e.g., OBY). On group 2 for High Frequency words had the remaining 14 words that did not have the letters reversed on group 1 were used on group two as generated words, also with the first two letters reversed and underlined (e.g., OHRSE). (See Appendix A)
The same process has been applied for Low Frequency words which have also been split between 2 groups of 14 low frequency words.
Low Frequency 1 had 13 words in which the first two letters have been reversed and underlined (e.g., EKG), whilst Low Frequency 2 had 14 words that did not have the first two letters reversed on Low Frequency 1 (e.g., HTORN). (See Appendix A)
A table in which the students could mark their results for data collection.
Procedure: Subjects were presented randomly with the four different lists and the test was conducted without a training session. The experimenter was instructed to time the subjects who had to copy the words rearranging the letters correctly to form the correct word when necessary based on the time determined by the experimenter.
Once the subjects completed the first task of copying the words, subjects have been instructed to flip the sheets and try to recall the most possible amount of words from the list assigned to each of them.
After the experiment was done, participants were asked to mark their lists as LF for Low Frequency words list, and HF for High Frequency words lists.
After the experiment, a sheet with a table has been passed around the subjected so they could record their results in order to collect the data for the experiment.
Results: The results have shown that there is a generation effect on both Low & High Frequency words. Figure 1 demonstrates that there is a significant generation effect on both conditions.
Figure 1. Mean Values for recalled read and generated words in both conditions, low and high frequency words.
Paired Samples Tests have been used to calculate and compare the means on this experiment, and the results have demonstrated that on the “Low Frequency” condition the participants recalled significantly more “Generated” words (M=3.38, SD=1.63, SE=0.19) compared to the condition of “Read” words . (M=1.74, SD= 1.66, SE=0.19)= t(77)=-7.16, p<0.01, r=0.40.
The results also found that the generation
effect was significantly higher with the High
frequency condition for “Generated” words
(M=5.49, SD=2.00, SE=0.22) compared to the
“Read” words condition (M=2.74, SD=2.26,
SE=0.25), t(79)=-9.34, p<0.01, r=0.52.
Figure 2. Difference between Means of High & Low Frequency words.
An Independent T-Test was used in order to assess if the difference of means between both Low & High Frequency conditions is significant. The results indicated that the participants in the High Frequency condition were able to recall significantly more words (M=2.73, SD=2.610, SE=0.29) compared to the Low Frequency condition participants (M=1.64, SD=2.03, SE=0.23).
Equal variances were assumed (p=0.264, i.e. p>0. 05). A significant difference of the means was recorded t(156)=2.91, p>0.05, r=0.05.
As predicted, the amount of recalled words was greater for generated words compared to the read words
The results of this experiment are very encouraging in supporting that our predictions were correct, that there is a generation effect on Low Frequency words.
However, what is important to observe is that the generation effect on High Frequency words was significantly higher compared to the Low Frequency words, which leads us to question if there is a familiarity factor is directly linked to the advantage of free recall.