Psychological laboratory experiment and analysis

Mock Research Experiment: The Generation Effect

This research study plans to look at and analyze the generation effect. The generation effect is the finding that self-generated stimuli will be both recalled and recognize with better performance than for read stimuli. This effect has been demonstrated primarily with words, yet this study uses pictorial stimuli within the experiment as opposed to using words. Because so many studies have been done focusing on the use of words, I decided to compose a study This that examines pictures within the generation effect and whether or not this effect influenced subjects memory. Memory was tested through different methods such as recall, recognition and also two different source monitoring tasks. The data obtained seemed to support the notion that the generation effect can and does occur for pictures.

Introduction

The phenomenon of the generation effect is one which I have long found very interesting. The effect can be replicated for many different generated responses such as incongruous items, effortful items, distinct items and elaborate generated responses. However, all of my knowledge of experiments on the generation effect focused on the use of words. Thus, I thought it would be both interesting and fun to compose a study of the generation effect the relies on pictorial stimuli.

This experiement shows that indeed a generation effect can be achieved through pictorial stimuli and not just from word stimuli. This effect can be attributed to the extra cognitive operations required to make the self-generated names and features of the pictorial stimuli (Kinjo 2000). Although very little research has concerned pictures and the generation effect, I am not surprised to see the generation effect manifest itself in pictorial stimuli. I believe the extra sensory information involved in pictorial stimuli allows for greater mental stimulation for pictures and thus leads to better retention for these items. And I believe that each of these experiments showed that sensory information and not just semantics are important features of the generation effect.

Experiment

Questions:

The question this experiment sought to answeris: Does the generation effect occur for pictures? Based on previous reports of the generation effect for pictorial stimuli, the I would hy potehizethat indeed a generation effect would occur. Subjects in the experiment would identify pictures with moderately incomplete images and other complete images. Subjects were then tested for their memory of the studied pictures in free recall, recognition and a source-monitoring task. The experiment also raised the question of which of the three testing methods, would a larger generation effect occur. Based on previous findings, it was also hypothesized that the effect would be bigger for recognition than recall. The experiment also sought to determine how positioning of the tests may or may not alter the generation effect. For example, would the effect be increased when the source-monitoring task was given first. Each of these questions were specifically identified within this experiment to reach answers that may be interpolated to a broader level.

Based on these questions this experiment could then look to a more universal level and study what factors affect the generation effect for pictures and whether these factors differed for pictures compared to words. The generation effect for pictures may after all be attributed to one of two factors: greater activation of the picture's sensory features or by greater activation of the picture's semantic features (Huffman 2000).

To see the difference between tests and pictures, test stimuli were changed from pictures to complete words. This change in stimuli should indeed eliminate the sensory match that masked the generation effect in the name test condition of the first experiment. This however, was not the only change in the experiment. A second source-monitoring task was added to the experiment. This task was a success/failure task where subjects were asked to determine if they had successfully identified the target item as correct in the study condition. Through these two changes in experimental design, many new questions are raised. First, it is possible that through comparison between the magnitude of the generation effect across experiments to determine whether sensory or semantic features had greater influence over the generation effect. Furthermore, through comparisons between each of the source-monitoring tasks, it may be possible for the experimenters to determine the source of the generation effect. Other questions asked in the experiment at a more specific level are would the generated pictures be recognized better than the named ones during the recognition task. In the complete/incomplete task one question is whether subjects would remember incompleteness better than completeness. Answers to these questions should hopefully provide the experimenters with answers to some of the underlying issues around the generation effect such as its source and also the factors that influence its strength.

Method:

To conduct this experiment and achieve potential solutions to the questions raised by this experiment, the experiment adhered to the following procedure. The subjects for the experiment were 24 undergraduates who were participating in the experiment to fulfill a course requirement. Each subject sat in front of a 256 x 256 pixel screen of an Apple Macintosh Classic computer. To promote subjects' motivation, subjects were informed that the student with the highest score at the end of the experiment would receive a $25 reward. Subjects were also told the experiment was on how people identify pictures and were never informed that their memory would later be tested. During the study, the subjects were presented with 80 experimental pictures at one of two levels of fragmentation. Half of the pictures were moderately incomplete (generation study condition) and the other half were complete (naming condition). The pictures were simple black and white drawings of common objects and animals. The incomplete pictures of the generation condition had about 75% of the picture deleted. Each stimulus appeared on the screen for two seconds and then erased. Subjects were then asked to type the name of each picture and were then told if they were correct or incorrect. In addition, following each answer whether incorrect or correct, the stimulus appeared again with the correct name for two more seconds. Subjects received 3 points for each correct answer. Following the eighty trials, a distractor task appeared where subjects were asked to judge whether each of 22 pairs of two visual texture patterns were the same of different. This task took about 5 minutes.

The testing phase followed this distractor task. Half of the subjects received the free recall test and then the recognition and source monitoring task (which was determining whether the target had appeared as complete or incomplete during study). The other half of the subjects received testing in the reverse order. In free recall, subjects were asked to type as many names of studied picture as possible. In the recognition task subjects were instructed to press a key on the keyboard if they had seen the item during study or press a different key if they did not. Finally in the complete/incomplete source-monitoring task, subjects pressed a key for items they had seen as complete, another for items they saw as incomplete and another key if the target was a new item. For the recognition and source-monitoring tasks, the eighty items were also mixed with 40 additional items that had not been presented earlier and of these forty half were complete and half were incomplete (Niles 2000). Also, of the 40 generated pictures from the study condition, during testing, half of these pictures were displayed as moderately incomplete as in the study condition and half were instead viewed as complete images. It is also important to note that the subjects' accuracy and reaction time were all recorded by the computer.

In the second phase, the pictorial stimuli of the testing phase were also replaced with words. These words were the names of the pictures that were displayed during the study phase. There was no free recall in the second experiment. Subjects were again tested with a recognition task and the same complete/incomplete source-monitoring task. Also, an additional source monitoring task was added to the testing. This task was a success/failure task, where students had to answer the question of whether or not they had correctly identified the target earlier in the study phase of the experiment. Subjects again used the keyboard to produce their responses and the computer once again recorded accuracy and reaction time.

Results:

The first data from the experiment focuses on identification rates of pictures during the study condition. The identification rates during study for generated pictures was 51% and 92% for named pictures. And after analyzing the data from the test condition, indeed my hypothesis was supported, a generation effect can and did occur for pictures. Nowhere was this more evident than in the recall test. The proportion of recall scores for generated pictures was.57 and was a smaller.48 for named pictures. Also it was found that the generation effect and its magnitude held stagnant whether the recall was administered before or after the recognition and source-monitoring task. The effect was also the same no matter whether the subjects had correctly identified the generated picture during the study period.

For the recognition task, false alarm rates differed between the two test conditions. There was a false alarm rate of.13 in the generate test condition and.07 in the name test condition. To analyze this data, recognition performance was calculated by subtracting false alarm rates to new items from the hit rates to old items. These numbers are known as the proportion of recognition (Pr) and their data supported the generation effect. Recognition was better for generated items than it was for named items. However, this generation effect was only found when the generated item was reinstated during the test.

For the complete/incomplete task, again recognition performance was again calculated using the same equation as in the recognition task for Pr. However, in this experiment there were two different types of false alarms. Subjects could classify a new item as complete or incomplete instead of new. It was found that subjects tended to call new items incomplete. Again, the results were similar. There was a generation effect found that contributed to enhanced recall for generated items, however, this effect was only found when the generation context was the method of test recall.

Like in the first experiment, the first set of data to analyze is the identification rates during the test study which were 55% correct for generated pictures and 97% correct for named pictures. This data is also very comparable to the identification rates of the first experiment.

The word phase test had similar findings. Again following the same equations as in the previous phase, Pr was used as a measure to determine the data and recognition of memory. The data from this experiment is lengthy and there are several important results from this data. Firstly, it showed superior memory for generated pictures over named pictures in both the recognition task and complete/incomplete task, however, no superiority was reported in the success/failure task. Secondly it was found that the magnitude of the overall generation effect was bigger in the complete/incomplete judgment task than in the recognition task. No generation effect was illustrated in the success/failure task from this data. Third, analysis of each of the tasks differed greatly. For the recognition task, accuracy for correctly generated pictures was higher than accuracy for correctly named pictures. For the complete/incomplete task, accuracy for not generated pictures was higher than accuracy for correctly generated pictures. And for the success/failure task, accuracy was higher for correctly generated pictures than for named pictures and not generated pictures (Richards 2000). Lastly, it is also important to note that the magnitude of the generation effect between phases one and two was compared for both the recognition and complete/incomplete tasks. The data from the Pr values showed that indeed there was a significantly larger generation effect in the first experiment than in the second.

Each of these results offers its own unique insight into the generation effect. More importantly when these results are interpreted to the next level, a more focused analysis can be made of the generation effect and also on its effect for pictorial stimuli

Discussion:

The substantial finding from experiment one illustrates the notion that indeed the generation effect does occur for pictures. A significant generation effect was demonstrated through the free recall testing. Also, it was observed that indeed this generation effect held constant regardless of test order. Furthermore, it was also discovered that the generation effect was obtained in both the recognition and the source-monitoring task. However, this only occurred when the generation context of the study condition was replicated in the test condition. Therefore, this result does not concur with previous experiments which showed a greater generation effect for recognition rather than recall. Since recognition performance was higher for the name/name condition (the latter referring to the study condition and the former to test condition) than it was for the generate/generate condition it can be interpreted that there is some difference in the generation effect between words and pictures, at least for recognition. A possible explanation for this maybe that pictures have a superior sensory code than words which thus gives a sensory match advantage that would be helpful in the recognition task. For words, the subject would only be able to rely on the semantic elaboration of the generation effect. This large recognition advantage in the name/name condition may mask generation effect. Finally, another finding from the experiment showed that the complete/incomplete memory task was more difficult than the recognition task. This effect could be demonstrated both by increased accuracy and decreased reaction time for the recognition task. It is possible to attribute this occurs because this task requires subjects to focus on the encoded sensory features of the pictures. And so in conclusion, the important universal interpretation of this experiment is the appearance of a generation effect for picture stimuli and also the potential masking of this generation effect from the sensory features of the pictures which occurred in the name/name condition. An important question thus to consider is if indeed this masking could be eliminated (Medger 2001).