The area of cognitive psychology which is interested in generation effect is quite wide. Many experiments were done but also many theories made on them. To widely examinee the whole subject about generation effect in free recall, it has to be explained what generation effect means. Very meaningful experiment was done by Slamecka and Graf (1978). According to their findings main theory about what the generation effect is has been created. Those researchers made 5 experiments in which generated words were compared to read words. What was founded is that performance in generated words was better than in the read ones. They hypothesize that information will be always better remembered if is generated than if is just read. It supports that generation effect is real. It also creates problem which is about interpretation of this hypothesis. The interest about generation effect rise when this experiment was published. Since this time this research was like a starting point for next studies.
There is another experiment which has impact on the generation effect. Jacoby (1978) wanted to find out when solution of the problem will be better remembered. Two experiments on 36 students were completed. What was discovered is that remembering solution suffered in the situation when was found by remembering the final solution. While remembered solution obtained by solving the problem was better remembered.
McElroy and Slamecka (1982) did experiment which obtained very consequential theories. When the experiment has been started it was believed that generation effect is due to innate differences in the process of generate and read. They did three experiments on 44 students. Firstly, it has been suggested that generation effect occurs only if during the study semantic memory in employ. Subsequently they did not find any generation effect with non-words. At the end they suggested lexical activation hypothesis. What means that for generation effect to occur the items have to be characterized in mental lexicon. However for the read condition generation emphasize lexical representation activation. This causes that memory performance is better.
Different outcome was obtained by Slamecka, Fevreiski and Jacobo (1983). They wanted to examine the generation effect when generation fails. What they finally hypothesize is that generation fails, because generations where incomplete. They were incomplete, because semantic features were processed instead of surface features.
In 1985 Nairne, Pusen and Widner used high frequency words, low frequency words and non-words for their experiment. What was first established is that generation effect occurs only for high frequency words. They did not believe in previous theory about lexicon representation, so according to their findings they reject lexical activation theory. Their findings showed that representation in the lexicon can not be condition which is sufficient for the generation effect. Nairne proposed different theory to replicate the one which was rejected by his team. Associative linkage hypothesis states that the fundamental is the fact that in the semantic networks are items which are related associatively with those items which are generated. According to this theory low frequency words have less associative links than high frequency words so they should show less of generation effect. Therefore, non-words do not have any representation in semantic network so they are not able to create generation effect at all. Summarizing, the generation effect in low frequency words is possible. Thus, it also means that this effect will be smaller in low frequency than with high frequency words.
As shown above generation effect has been widely researched what produce the biggest evidence for the importance of this phenomenon. In addition, conduction of this study was to enrich already existing findings or to gain new significant results. Experimenter wanted to find out if generation effect is possible with low frequency words. Another reason was to check the difference in the generation effect for low frequency and high frequency words. Experimenter stated two experimental hypotheses for the present study. Hypothesis that there will be difference in the generation effect between high and low frequency words during the free recall. Null hypothesis says that there will be no differences between generation effects in high and low frequency words. Second hypothesis stated that generation effect with low frequency words during the free recall is possible. Null hypothesis says that there will be no generation effect with low frequency words during the free recall.
In this research experimenter chosen combination of two designs which can be called “mixed design”. As a first repeated measures design was used in participant’s task. Stimulus list presented contained words which were read and those which were generated. In this case dependent variable was amount of remembered words in a free recall, while independent variable was either generated or read words. Second design used in this research was independent groups design. Experimenter divided two groups from the same participants as before. Allocation to those groups was made randomly. Each of the groups performed in two different conditions. One condition was list of low frequency words while second condition was list of high frequency words. The dependent variable in this case was score of recalled words when independent variable was high frequency words versus low frequency words. Dependent variable in both designs was measured by manipulation of independent variable.
Experimenter obtained 158 people as participants. All of those participants were psychology students from the Research Methods course at Birkbeck. Some of those students were BSc Psychology students year II and some of them MSc Psychology students. They were female and male participants. Verification of participant’s age was not done. Experimenter did not need that information, because was not interested in their influence on the study. All participants had different occupation. During the second part of experiment participants were randomly allocated in to two different groups. In group A was 80 people, in the group B was 78 participants. According to the total number of participants, the suggestion that the groups were almost equal is possible. Participation in this study was volunteer for all people. Also, they did not pay for anything such as participation or any materials used to conduct the research. When the experiment has been started the participants did not know what the purpose of this study was. In other words they were naA?ve about the purpose of the study. However debriefing and explanation how the first part of the study should look likes was done just before experiment begins.
Materials and Apparatus
To conduct experiment without problems and concerns experimenter prepared a number of materials and apparatus needed. The most important thing to carry out the study was to organize such a big classroom to fit all of the participants. Next thing was to prepare appropriate number of desks and adequate number of sits. Those procedures were done so experimenter concentrated on the rest materials needed. The person who was carrying on the study had a stopwatch and big felt-tip as a metronome (both needed during first part of experiment). As a first thing given to the students were sheets of paper one per person. All of them included list of equal number of words (28 words). There were two different lists of words given. Both types of stimulus lists contained 56 nouns which were taken from the Kucera and Francis (1967) word count (see Appendix 1 for examples of both stimuli lists). The apparatus consisted of a computer, projector and white board. As a material can be also, lecture slides about “variance” which was showed to participants during planned break in experiment. In the second part of experiment all participants were required to have pen and piece of paper to write recalled words. The last item given to the students was one sheet of document including a table to fill in by every each of participants.
When experiment began, short and fairly simple instructions were given. Participants were tested all together at once. Explanation not to turn out the sheet of paper given was done. Each sheet of paper included list of words on the left side of the page. The words were written in the column one under the other. The right hand side of the same sheet was an empty, which was for participant’s responses. It was clarified that when experiment will begin the participant’s task will be to simply copy the words red to the right side of the page. The words with underlined initial letters should be transformed by reverse those initial letters and then write on the right side of the page. The instructions advised to be accurate when completing the task. Participants knew that they should not look up on collogues sheets of paper until experiment finish. Total time for this task was 2.1 minutes. The time for copying each word was signalized by experiment’s metronome. Felt-tip produced signal every 4.5 second what meant that the time for one word was finished and participant has to go to copy next one. When 2.1 minutes passed and task was completed experimenter did 10 minutes break. For the duration of this break experimenter organized the research method lecture. After the speech a free recall test was done. Every participant had time to recall all words remembered from the previous sheet of words given. When that happened one document paper was passed to the participants and person after person were asked to fill it in. The document contained table with fallowing questions: In which group the person was. How many recalled words where generated and how many where words read. Language predispositions.
According to the experiment’s results very useful information was found. To have consistent knowledge about numerical part of the experiment it would be helpful to demonstrate the graph. Figure 1 presents mean for the recall words of high and low frequency words with division on read and generate words. As shown for the high frequency words, mean number of generated words recalled is 5.49 with standard deviation SD=2.00. Mean for the read words is 2.76 and SD=2.26. For the low frequency words, mean number of generated words recalled is 3.38 with SD=1.63. For the read words mean is 1.74 and SD=1.66. However when compare high frequency words with low frequency words the total number of words recalled is less for the low frequency words (400), than for the high frequency (660). (see Figure 1 and Table 1).
After descriptive statistics are described there is a time to take in to account inferential statistics. In this experiment two t-tests were done.
First dependent t-test was conducted to check if there will be generation effect with low frequency words. In this case score from both experimental conditions come from the same people because experimenter used repeated measures design. This test shows that scores from those conditions correlate. For the present study correlation coefficient yield to be quite strong r= .243, p< .05. The final results from this test report that the generated words were highly statistically significantly more likely to recall (M=3.38, SE=.183), than the words which were read (M= 1.74, SE= .188), t(77)= 7.15, p< .01, r=.64 (large effect size as it is above 0.5 which is essential finding).
Subsequently independent t-test was done to check the difference in the generation effect for the high and low frequency words. They were two groups: one 80 participants, and the other 78 participants. Lavene’s test in this case is non- significant because p=.264, p>.05. Variances of two samples are not different from each other, so they cannot cause any concerns. In this situation experimenter considered equal variances assumed. When interpret this test the conclusion came that there is highly statistically significant difference between high frequency words (M=2.72, SE=.29) and low frequency words (M= 1.64, SE=.23), t(156)=2.91, p<.005, r=.23 (small effect size as it is below 0.3 which is medium effect).
Table 1. Presentation of mean and standard deviation.
Figure 1. Mean recall for high and low frequency words.
Data collected in the results section confirmed both hypotheses. In the present study there is generation effect for low frequency words and there is a difference in the generation effect for low and for high frequency words. Both hypotheses proved highly statistical differences.
Those findings strongly support previous study: Slamecka & Graff (1978), McElroy & Slamecka (1982), Nairne, Pusen, Widner (1985), Slameka, Fevreiski and Jacobo (1983). It gives external reliability to those findings. In the present study the generation effect occur for high frequency words (Slamecka & Graff, 1978). However occur for the low frequency words as well, what was not found by Slamecka & Graff.
McElroy & Slamecka established lexical activation theory as an explanation of generation effect. They believed that mental representation of the item is required for the generation effect to occur. The present study produce extend of those findings, but on the other hand McElroy’s study does not say anything about generation effect for low frequency words. Nairne (1985) refused lexical activation theory and proposed new one called associative linkage theory as an explanation of generation effect. Subsequently in 1988, Nairne sustained his theory by another experiment which supports position that lexical representation is not necessary to produce generation effect (Nairne, Widner, 1988). Current experiment also supports those findings. Especially that Nairne find out the possibility of generation effect for low frequency words which was confirmed and supported by the current study. Another challenging outcome for the associative linkage hypothesis came out from the study made by Saint-Aubin (2005). Suggestion that high frequency words are better recalled was explained by the fact that they have long- term associative links. However when low frequency words (which have short- term associative links) were added to the list of high frequency words the recall was good for both types of the words. It creates additional finding which is essential for the generation effect phenomena. Another series of experiment made by Gardiner (1988) gives new idea about generation effect for low frequency and high frequency words. The results from this study explained that there is generation effect for the low frequency words the same as for high frequency words. The results of the present study support Gardiner’s findings. However there is no support for the next part of Gardiner’s conclusion which says about no difference between those two generation effects. As clarified by the second hypothesis of present study, there is a difference between generation effect for the high frequency and low frequency words. Perhaps high frequency words are more common words and this produces generation effect so that is why they are better remembered.
One of the strengths of this study is the methodological fact. Sample size of this study was quite large, as experiment obtained 158 participants. It means that generalization is possible. Moreover, it may be fairly representative in the general population what creates very strong point of this study and raise the external validity. In the contrast there is weakness about the people participating in the study. They were all undergraduate and postgraduate students what suggests that they were in similar age. Someone could say that this fact is causing mineralization of external validity. Also, the results would be slightly different, because of the narrow age range (Hultsch, 1969). A good improvement in this case would be to put attention on people’s age when obtaining participants to the study.
Another limitation associated with this experiment is taking in to consideration the fact about noise disturbance. It was found that noise can cause impaired ability to remember words. Noise produces disturbance of the information input in short- term memory (Salame, 1978). Good improvement of the study would be to take care about environmental factors during the study. The place where experiment was carried out should be controlled by the experimenter to organize quiet environment. There could be a problem with concentration and perception during free recall if it would take place in the noisy area. This could cause inappropriate responses.
Every psychological research has to take care of the participant’s well being during the experiment. In this study there was lack of precisely made debriefing. There was a part in the procedure section when experimenter explained what participants should do when experiment starts. However, experimenter should take into the account the fact that the information about free recall after 10 minutes break should be known by the participants. The appropriate information about the nature of the study should be acknowledged by the participants since experiment began. It would produce more trustworthy approach to the study.
In this particular area of interest Hendry’s (2005) experiments gave a suggestion about implication of word length on generation effect. Those experiments compared length of the words and generation effects. Single item recognition tasks were done the same as serial recall recognition tasks. What he discovered is that long words are worst recall than short ones on the serial recall tasks. However, better recognized are words which are long than those one which are short. It gives a suggestion about improvement of current study. Length of the words in both stimuli list was not taken in to consideration. It might be good idea to add it when replicate the study to obtain new reliably results.
To lead the experiment to the perfection it would be necessary to take in to consideration student’s language preferences. Van den Noort & Maurits (2006) tested language proficiency on multilinguals. It has been found that the number of languages spoken has impact on the tasks results (memory task). Good improvement of present study would be to state another hypothesis saying something about language preferences. Then, conduction of the parametric test might me useful.
As indicated above there are some strengths and some weaknesses of presented experiment. Nevertheless, it is suggested and explained how to improve the study by proposing suitable solution. In conclusion, all of the results from the current study support the findings from previous experiments. Moreover, both of the conclusions strongly support Nairne’s, Pusen’s and Widner’s (1985) associative linkage theory. It was found that low frequency words are easier to remember when they are generated rather than when just simply read. Also, the difference between generation effect in low frequency and high frequency words was discovered. Besides, this particular study is very reliable and useful because of the external validity. It is shown that there is a room for improvement in this research what might be useful for the next extensions in this particular topic. That could give opportunity to make replication of this research more valid and reliable.
1. Examples of stimulus lists.
High Freq 1 High Fre q2 Low Freq1 Low Freq2
BOY OBY THORN HTORN
WOMAN OWMAN EKG KEG
APRTY PART SHRIMP HSRIMP
NIGHT INGHT ROCHID ORCHID
CHILDREN HCILDREN QSUIRREL SQUIRREL
OHRSE HORSE UNN NUN
OBDY BODY CLAW LCAW
WATER AWTER BLUEBELL LBUEBELL
IRVER RIVER HWALE WHALE
KSY SKY PICKLE IPCKLE
USMMER SUMMER KITE IKTE
HCAIR CHAIR PAE APE
FATHER AFTHER BLOSSOM LBOSSOM
HOTEL OHTEL IZPPER ZIPPER
OMNEY MONEY KILT IKLT
ESA SEA SLIPPER LSIPPER
CHURCH HCURCH RAINBOW ARINBOW
SILAND ISLAND AHRP HARP
EYE YEE AGRBAGE GARBAGE
ROAD ORAD WEED EWED
ELTTER LETTER PUPPY UPPPY
CITY ICTY LEOPARD ELOPARD
OHUSE HOUSE OBULDER BOULDER
MOTHER OMTHER FLEA LFEA
UGN GUN AHRBOUR HARBOUR
CAR ACR USRF SURF
RGEEN GREEN ICDER CIDER
HAND AHND LPUM PLUM