The third level (semantic) is considered to be the deepest level of the three levels. It depends on the nature of the stimulus and the time available for processing. The more deeply information is processed the more likely it is to be retained. This experiment supports Craik and Lockhart because it supports the idea that the semantic level is the most important as it focuses on what kind of information is being processed. If the information is recognised (i.e. there is an existing schema) then information is more likely to be comprehended and therefore retained. This links to schema theory as giving someone a schema is giving them the ability to process information on a semantic level, it puts the information into context.
This investigation will focus on the work done by Bransford and Johnson. It's a partial replication of a study they conducted in 1972. Their study was based on the use of a schema on aiding comprehension and recall of a short passage. They conducted some prose passages, which could be difficult to understand in the absence of a context or schema. They then compared recall and comprehension between a group of participants who had been supplied with contextual information and a group who had not received this additional information. Their results clearly showed that the group given a context/schema to match the passage to comprehended and recalled significantly more than those not given a schema. This theory therefore supports the Top Down Processing theory.
This investigation however, will be recording only the level of comprehension of the passage and not the level of recall and will only use two conditions.
Aim
The overall aim of this experiment is in the interest of science, it is to test the reliability and the validity of Bransford and Johnson’s (1972) original study and to see if it still applies today.
The precise aim of this experiment is;
To investigate the use of a schema and its effects on comprehension of a short passage. Using partial replication of a 1972 Bransford and Johnson study on prerequisites for understanding.
A one-tailed hypothesis was used because from looking at Bransford and Johnson's research the investigator is confident that they can predict the direction of the results.
NULL HYPOTHESIS: there will be no significant difference in the level of comprehension between those given a schema and those not given a schema. Any difference will be due to chance factors alone.
EXPERIMENTAL HYPOTHESIS: the group given a schema before being read a short passage will comprehend significantly more than those not given a schema before reading a short passage.
Method
Design
The experimenter chose to conduct an experiment for many reasons. An experiment enables the experimenter to control the effect of the extraneous variables; this allows cause-and-effect relationships to be established. Conducting an experiment in a laboratory provides the experimenter with the opportunity to measure behaviour with a greater precision than would be possible in a natural environment. Using a laboratory also allows the researcher to simplify the sometimes-complex events of the natural world using experimental reductivism. Another reason the experimenter may have chosen to do an experiment is because the study is a partial replication of Bransford and Johnson (1972) and they chose to use an experiment.
The experimental design used was independent groups. This design was chosen because if the task has to be repeated, participants would have had practice the second time and there is also the risk that the participants would have guessed the aim. A disadvantage of using independent groups is the risk of participant variables; if the experimenter was aware of any of the participant’s memory skills he may be tempted to use this to sway the results. The experimenter used random allocation to overcome this potential problem. This involved participants picking at random a piece of paper from a hat. The piece of paper had either an A or a B on it, the letter indicated which group the participant was assigned to. The same number of A’s and B’s must be in the hat to make this method fair.
An extraneous variable the experimenter controlled for was the use of psychology students. This is a problem because the use of psychology students in the experiment could impair the results as they may try to guess the aim of the experiment or the schema this will increase the demand characteristics and their recall. A way of controlling for this would be to ask participants before the experiment whether they are or have been a psychology student. This control would work because anyone currently studying, or had studied psychology would not be allowed to take part I the experiment.
The difficulty of the passage could also have impaired results. It may not have been comprehendible with or without the schema, creating a floor effect. This was controlled for by conducting a pilot study. An extraneous variable which the experimenter did not control for was noise. Noise levels vary throughout the day in college. For example lunchtime had much more outside noise than lecture times. This was not controlled for.
The significance level was set at P≤0.05 in order to avoid type 1 and 2 errors.
The IV in this experiment was whether participants were given a schema to aid comprehension before hearing a short passage, or not given a schema to aid comprehension before hearing a short passage.
The DV in this experiment was how well the participants comprehended the passage by rating their comprehension on a likert scale.
Participants
The target population in this experiment was 16 to 19 year old students from Truro College in South West Cornwall. The sampling method used in this experiment was opportunity sampling. The experimenter chose to use opportunity sampling because it uses people who are willing, available, and suitable to take part and part of the target population. It is a very quick and easy method to use, although, a disadvantage of opportunity sampling is that the experimenter chooses the participants. Therefore not everyone in the target population has the same chance of being chosen as the researcher may have been sub-consciously biased and so may not choose a very representative sample as they may be tempted to choose people they think will perhaps perform better and allocate them to a certain condition. This was controlled for using random allocation.
Random sampling may have been better here because it means that every member of the target population has an equal chance of being chosen. A disadvantage of random sampling is that it is much more time consuming and the names pulled out of the hat may not wish to take part, also the investigator did not have access to all the names in the target population at Truro College
The ratio of male to female in this experiment was 13:17
All participants were tested at Truro College between the 1st August 2003 and 31st August 2003 between 9am and 4:15pm.
Task and materials
The first thing that the experimenter had to do was to design a mind map containing everything which was needed to carry out the experiment. The first to be done was the standardised instructions/brief (ref app 1 and 2). This was done on computer using Microsoft Word 2002, the font used was Tahoma size 12, bold and underlined for the title. It was printed on white A4 paper. This procedure was used for both sets of standardised instructions (Group A and Group B) (ref app 1 and 2) and the debriefing (ref app 3). When designing the content of the standardised instructions and the debriefing ethical issues were taken into consideration and it was made sure that the same thing was said to everyone. Once this was complete the experimenter then designed a passage about sunbathing and typed it up on Microsoft Word 2002 font size Tahoma size 12. The researcher designed a Likert scale (a scale from 1 to 7) asking the participants to rate their comprehension of the passage. 1 being not comprehended at all and 7 being completely comprehended. The researcher also printed off thirty data collection sheets consisting of a scale of 1 to 7 designed on Microsoft Publisher 2003 (ref app 8-38). Then the results table (ref app 39) was designed; this was done using Microsoft Publisher 2003 and selecting the table icon. The table consisted of two columns and fifteen rows in each column. Each column was labelled either condition A or condition B and the rows were numbered one to fifteen all in times Tahoma size 12 again printed on white A4 paper. Lastly the participant consent sheets (ref app 6-7). Once the experimenter had carried out a pilot study with three friends and made any necessary alterations. The experimenter’s lecturer than checked it met all ethical guidelines.
Ethics
There were four major ethical issues which had to be controlled for in this experiment.
The first of these was ‘consent v informed consent’. In psychology, researchers aim to inform participants of as much information about the experiment as possible in order to give them informed consent. However sometimes this is not possible as revealing the aim can cause demand characteristics which is where the participant behaves in a way they think the experimenter wants them to behave. Therefore the ‘real’ aim cannot be revealed to the participants at the start of the experiment. The experimenter must control for this by informing the participant of their right to withdraw at anytime during the experiment as stated in the standardised instructions (ref app) and debrief (ref app). This works as a control because if the participants feels uncomfortable at any stage of the experiment for any reason they can stop immediately and be given a full debrief (ref app).
The second ethical issue which had to be controlled for in this experiment is deception. Deception is the deliberate misleading of participants during an experiment either through telling them lies or by failing to tell them some important detail of the research. This is an ethical issue because as mentioned earlier participants should be fully informed of the aims of the experiment and what the experiment will involve. The control for deception is that at the end of the experiment the participant is fully debriefed where the real aim of the experiment is revealed to them and the reason for using deception. This control works as the participants then leaves fully aware of how the experiment works and exactly what they were being tested for.
The third major ethical issue which had to be controlled for was confidentiality. Confidentiality is where the participants have the right to expect that information gathered during the experiment will not be made public without their consent. This is an issue as it may cause the participant psychological stress and trauma if their results were to be made public. This is controlled for by assuring the participant in both the standardised instructions (ref app) and debrief (ref app) that “…all results will be held 100% confidential…” and that participants have the right to withdraw their data if they are unhappy with the final report. This control works as it assures the participant before and after the experiment that that their results will remain anonymous at all times.
The fourth main issue which had to be controlled foe in this experiment was psychological stress and trauma. It is a known fact that participants should leave after an experiment in the same psychological state as they arrived in. participants can become psychologically stressed during an experiment, especially if they think their results are abnormal. As a control for this participants are assured in the debrief that “From looking at your results I can tell you that they are perfectly normal for the group you were in.” and if they are still unhappy, as mentioned earlier they have the right to withdraw their data. This ensures that the participant suffers no psychological pain and trauma.
Procedure
The researcher stood outside the canteen asking random passers who fitted into the target population “Hi, my name is Vickie and I am a psychology student. I am conducting an experiment as part of my A Level coursework, would you like to take part?” those who agreed and confirmed that they were not themselves psychology students were asked to pick a piece of paper from a hat which had either an A or a B on it, randomly allocating them to a condition. The participant was then read the relevant standardised instructions/brief (ref app 1-2) which applied to the condition they were in. After checking that the participant understood what they had to do and that they were still happy to take part, the experiment could begin. The researcher gave the participant a data collection sheet (ref app 8-38) and a pen. The researcher then read the stimulus (ref app 5) to the participant and immediately asked them to rate their comprehension on a scale of 1 to 7 on the data collection sheet. Once this was complete the participant was then read the debrief (ref app 3-4) and allowed to ask any questions regarding the experiment. The experimenter then thanked the participant for taking part “Thank you very much for taking part in my study” and asked them to sign the participant signature sheet (ref app 6-7).
The experimenter then stood outside the canteen again and waited for the next participant to repeat the procedure to. This was done until the experimenter had thirty pieces of raw data.
Results
The raw data suggest that the participants given a schema comprehended more than those not given a schema.
Measures of central tendency.
The researcher chose to use measures of central tendency because they provide a single number or value to describe a set of raw data values. Central tendency provides a representative score for the experimenter to see what is going with the data. In this experiment the researcher chose to use the mean and the median. The mean was calculated to establish whether there was a normal distribution of scores as this is the first assumption of the parametric test. If the mean and the median are similar then it can be assumed that the data has a normal distribution of scores. The median is calculated because it is the most appropriate measure of central tendency. This is because the data is ordinal. The data is ordinal because there are no equal distances between the class intervals as the researcher used a subjective scale.
Table 1 – Measures of central tendency for comprehension of participants given or not given a schema.
The means show that the average level of comprehension for the participants given a schema is higher than the average level of comprehension of the participants not given a schema. As the means and medians are similar for both groups it can be assumed that there is a normal distribution of scores in each sample.
Measures of dispersion for comprehension.
Measures of dispersion are statistics which display the difference in spread or dispersion of the data. The researcher chose to use the range and the standard deviation in this experiment. The range was chosen because it is the most appropriate measure of dispersion as the date is ordinal. The range is the difference between the highest value and the lowest value in the set of raw data. It can however, be distorted by an extremely high or low value. The standard deviation was calculated because it is more sensitive than the range although it is not always as good with ordinal data. The standard deviation also represents the spread of scores.The standard deviation can indicate homogeneity of variance which is the second assumption of the parametric test.
Table 2 – Measures of dispersion for comprehension of a passage for participants given a schema or not given a schema.
The range is the same for both conditions, this shows that there is an even spread of scores for both participants given a schema and participants not given a schema. The standard deviation is a much more sensitive measure, it shows a more true representation of the spread of scores around the mean. It can be seen from the results that the standard deviations are not more than two away from each other, this indicates homogeneity of variance, the second assumption of the parametric test.
However, although two assumptions of the parametric test have been met, the data is ordinal and so the third cannot be met. Therefore the researcher carried out a non-parametric test.
Graphical representation of data.
The researcher chose to include two graphs because they provide a visual representation of data. In this experiment a frequency polygon and a bar chart were used. A frequency polygon was chosen because it highlights a normal distribution of scores. The frequency polygon also visually shows the range.
Graph 1 :
Results show that giving participants a schema before reading them a passage will enable them to comprehend more than participants not given a schema. This can be seen by the peak in Condition 1 (participants given a schema) is further to the right on the x-axis than the peak in condition 2 (participants not given a schema). It gives a clear representation of standard deviation and a visual range of scores.
Although the frequency polygon does highlight a normal distribution of scores, it is n he most appropriate form of graph for ordinal data. For this reason the researcher chose to also include a bar chart. A bar chart of medians is the most appropriate graphical representation of ordinal data.
Graph 2:
The graph shows that the participants given a schema comprehended more than the participants not given a schema. This is shown by the median for condition 1 (participants given a schema) being higher than the median for condition 2 (participants not given a schema).
Treatment of results.
Descriptive statistics such as the mean and the medians suggest that participants given a schema before hearing a short passage comprehend more than those not given a schema.
In order to see if this difference in comprehension is significant or not, the researcher has to apply inferential statistics.
Although normal distribution of results and homogeneity were established from the statistics, the level of data is only ordinal. The reason the level of data is ordinal is because there is no equal difference between the class intervals as a subjective test was used. So for this reason a non parametric test was applied.
The Mann Whitney U test was chosen for several reasons including;
- The Mann Whitney U test is a test of difference not a test of correlation.
- The type of design is independent groups and therefore ‘unrelated’.
- The level of data is ordinal.
- Parametric test cannot be carried out as the data is ordinal.
Table 3 – Treatment of inferential statistics to show levels of comprehension of a passage of participants given a schema or not given a schema.
1 tailed test
Cooligan (1999) states the decision rule:
“For any N1 and N2 the observed value of U is significant at a given level significance if it is equal to or less than the critical values shown.”
It can be seen that the critical value exceeds the observed value which allows for the rejection of the null hypothesis; that there would be no significant difference in the level of comprehension between those given a schema and those not given a schema. Any difference will be due to chance factors alone. Therefore the experimental hypothesis can be accepted; that the group given a schema before being read a short passage will comprehend significantly more than those given a schema before reading a short passage.
Discussion
Explanation of results.
The aim of this experiment was in the interest of science, to test the reliability and validity of Bransford and Johnson’s (1972) original study and see if schema theory is still relevant today. It is clear from the results of this experiment that the use of a schema will increase comprehension as it gives the information semantic meaning.
From looking at Bransford and Johnson’s (1972) original research this was the expected outcome of results. Their study showed that participants given a schema not only comprehended more than those not given a schema, they also recalled more information correctly. This suggests that comprehension has implications for recall.
On the basis of my results the null hypothesis; there will be no significant difference in the level of comprehension between those given a schema and those not given a schema. Any difference will be due to chance factors alone, can be rejected.
Therefore the experimental hypothesis; the group given a schema before being read a short passage will comprehend significantly more than those not given a schema before reading a short passage, can be accepted.
The researcher was confident that the results would follow Bransford and Johnson’s (1972) original research, allowing the experimenter to use a one tailed hypothesis
From looking at the results of this experiment the researcher is clearly able to conclude that participants given a schema before hearing a short passage comprehended significantly more than those not given a schema before being read a short passage.
This is supported by measures of central tendency (ref app 43-44) because the mean for the participants given a schema is much higher than the mean for the group not given a schema before hearing a short passage. This is also supported by the two graphs as the peaks on the frequency polygon clearly demonstrates the difference in means for those given a schema and those not given a schema.
Relationship to background material.
It is clear from the results that the findings in this study do support Bransford and Johnson’s (1972) original findings. Which in turn support the Top Down processing theory. It can be suggested that in this study participants not given a schema were attempting to attach this passage to an existing schema but failed to do this as the passage gave no real indication of what it was about. As participants were unable to match the passage to a schema they were unable to comprehend it. Participants given a schema were therefore able to comprehend the passage with no real difficulties. This study therefore supports the Top Down processing theory.
Table 4: The means from Bransford and Johnson’s original findings (1972)
It can be seen that the means of comprehension from Bransford and Johnson’s (1972) research are not so different to the means from this experiment.
Mean for group given a schema before;
Bransford and Johnson: 4.5
This experiment: 6.2
Mean for group not given a schema at all;
Bransford and Johnson: 2.2
This experiment: 1.9
The findings also support Craik and Lockhart’s level of processing theory (also a top down theory). They suggested that memory is a by-product of perceptual analysis and therefore the amount of rehearsal is not as important as the type of rehearsal. They proposed that we process information on three different levels;
- Structural/Shallow level – what does information sound like?
- Phonetic level – what does information sound like?
- Semantic level – what does information mean?
The third level (semantic) is considered to be the deepest level of the three levels. It depends on the nature of the stimulus and the time available for processing. The more deeply information is processed the more likely it is to be retained. This experiment supports Craik and Lockhart because it supports the idea that the semantic level is the most important as it focuses on what kind of information is being processed. if the information is recognised (i.e. there is an existing schema) then information is more likely to be comprehended and therefore retained. In this study the participants given a schema comprehended more because the information they were being given meant something to them. Whereas those not given a schema found the information being given to them meaningless and so they therefore comprehended less.
Findings in this experiment do not support Atkinson and Shiffrin’s multi-store model because it is a bottom up processing theory.
It suggests that rehearsal is more important to comprehension and recall than the type of information. Clearly in this experiment both conditions had the same level of rehearsal but those given a schema comprehended significantly more than those not given a schema. This suggests that the kind of information being processed is important and whether or not there is an existing schema. Therefore the researcher can reject Atkinson and Shiffrin’s multi-store model.
Limitations of Generalisation and suggestions for improvements:
A limitation of this experiment was the lack of control for background noise. The college has different levels of noise at different times of the day. For example at 9:30am students are mostly all in a lecture and therefore the noise level is fairly low. However at 10:30 everyone is on a break and moving around the college to their next lecture so it’s clear that the noise level will be much higher and therefore cause more of a distraction. The noise levels vary constantly throughout the day. This is a problem because it is much more difficult to concentrate in a noisy atmosphere than it is for them to concentrate in a quiet atmosphere. A lack of concentration is likely to effected comprehension levels in a negative way. This could have been controlled for by taking all participants to somewhere which remains reasonably quiet throughout the day, for example the library. The library is a silent study area so it is always fairly quiet which would allow the participants the concentration they needed and would allow all participants the same level of concentration.
Another major limitation of this experiment was the mark scheme used. This experiment used a simple Likert scale going from 1 to 7 with participants being asked to rate their level of comprehension on it. (1 being not understood at all and 7 being completely understood). However the problem with this is that it is a completely subjective scale in that there are no equal class intervals. A way of improving this would be to measure something that did not rely on a subjective element such as time or weight as they have set class intervals.
A third limitation which was not controlled for was the time of day and the day tested. It’s suggested that students learn more and in doing so comprehend more on a Monday morning than on a Friday afternoon. The participants in this experiment were tested at random times over a period of around four weeks so their alertness rate would have varied greatly. It is possible to control for this to a certain extent but it would take a ridiculously long time to test them all on the same week day ay exactly the same time e.g. 9:37am. But by testing all participants in the same time range say between 9am and 12 noon it would ensure to some extent that they were equally alert. This would give participants a more fair chance of performing to their full potential. However it is impossible to truly control for this in terms of factors such as how much sleep they had the night before and whether or not they had breakfast that morning etc and such other environmental factors beyond the experimenter’s control.
Another way which this experiment could be improved would be to check whether participants had already participated in a study of this nature. At Truro college psychology is a very popular subject and most students who are required to conduct an experiment have a target population of Truro college students as it is the most convenient option. Assuming that each psychology student tests between 30 and 40 people for their study it is probable that people will end up testing the same students especially as psychology students often won’t test other psychology students. If in fact a participant had already participated in another study of the same nature they would have had a debrief explaining the true aims of the experiment. This could lead to problems especially if the participant was in condition 2 (participants not given a schema) they would be aware that there is a schema to be guessed giving the participant an unfair advantage over the other participants. This in turn could affect the results. A simple way to control for this would be to check with the participant before they take part whether or not they had taken part in any psychology experiments and if so what did they involve. From this information the researcher could then decide whether or not the participant was suitable to take part or not.
Implications for results.
The results of this experiment have several implications for every day life especially as comprehension is seen to have implications for recall.
The findings in this study have implications for teaching methods in the classroom, in particular revision methods. Mind mapping has long been a popular revision method. The results in this experiment can attempt to explain why this is. Mind maps contain a word of phrase in the middle of a piece of paper and then all the information which needs to be recalled around the outside. The word or phrase in the middle can be said to be a schema for all the things around the edge. Therefore when a student draws a mind map everything they write around the phrase becomes part of that schema so in an exam all the student needs to remember is the word or phrase from in the middle and that will trigger other information from the schema. This is therefore support for mind maps being a successful form of revision.
Another real life implication which can be drawn from this study is the cognitive interviewing technique used for eye witness testimonies. The cognitive interviewing technique involves the witness mentally revisiting the scene of the crime, recalling the events of the crime backwards and telling what happened from different peoples positions. This helps to strengthen the schema making it more difficult for it to be distorted in future. It often triggers other pieces of vital information which had other wise been forgotten. The results of this experiment can therefore be seen as support for the cognitive interviewing technique.
Suggestions for further research.
A suggestion for further research would be to test the effectiveness of the cognitive interviewing technique as opposed to just straight forward questioning.
AIM: To test the effectiveness of the cognitive interviewing technique.
METHOD: would involve having two conditions with twenty participants in each condition. Condition 1 would be shown a video of a crime for example an armed robbery in a bank and would then be asked to write a statement of what happened. Condition 2 would be shown the same clip and then be interviewed using the cognitive approach before writing a statement of what happened.
RESULTS: would be expected to show that those interviewed using the cognitive approach before writing their statement would recall significantly more details correctly than those not interviewed using the cognitive approach before writing their statement.
CONCLUSION: from this the experimenter could conclude that interviewing eye witnesses using the cognitive approach does in fact help them to recall more information correctly than those not interviewed using the cognitive approach before writing their statement. This would entail the participants imagining themselves from a different position in the room and thinking what they were likely to have seen, and going over the series of events backwards to help jog their memory as to what happened.
Conclusion.
From looking at the results of this experiment it is clear that the participants given a schema before hearing a short passage comprehended significantly more than those not given a schema before hearing a short passage.
It is clear that giving people a context for new incoming information increases comprehension, and following Bransford and Johnson (1972) has implications for recall. This is supported by the measures of central tendency (ref app 43-44) and the clear difference in medians which is shown in the bar graph on page 13.
Therefore the researcher is able to reject the null hypothesis that there would be no significant difference in the level of comprehension between those given a schema and those not given a schema. Any difference will be due to chance factors alone. And accept the experimental hypothesis that the group given a schema before being read a short passage will comprehend significantly more than those given a schema before reading a short passage.
References
Allport, G., and Postman, L., The Psychology of Rumor (New York: Russell
And Russell,1947).
Atkinson, R.C., & Shiffrin, R.M., (1968) as cited in Pennington. D (2002)
Introducing psychology, approaches topics and methods, Edition 1
London: Hodder and Stoughton pp 264-265
Bartlett, F., (1932) as cited in Gross, R.D (1996) Psychology – The Science
of mind and behaviour edition 3. London: Hodder and Stoughton
Pp 302
John D. Bransford & Marcia K. Johnson (1972) as cited in Gross, R.D (1996)
Psychology – The Science of mind and behaviour edition 3
London: Hodder and Stoughton Pp 291
Cooligan, H., Introduction to Research Methods and Statistics in
Psychology edition 2. London: Hodder and Stoughton
Craik, F., & Lockhart, R.S., (1978) as cited in Gross, R.D. (1996)
Psychology – The Science of Mind and behaviour edition 3.
London: Hodder and Stoughton pp 290-291
Ebbinghaus, H., (1885) as cited in Gross, R.D. (1996)
Psychology – The Science of Mind and behaviour edition 3.
London: Hodder and Stoughton Pp 280
Rummelhart, D., (1980) as cited in Gross, R.D (1996) psychology – the
science of Mind and behaviour edition 3. London: Hodder and
Stoughton pp 345-346
Appendices
- Standardised instructions – condition 1
- Standardised instructions – condition 2
-
Debrief – condition 1
- Debrief – condition 2
- Stimulus
- Participant signature sheet 1
- Participant signature sheet 2
8. – 38. Data collection sheets
39. Results table
40. – 42. Individual contribution
43. Workings for measures of central tendency – condition 1
44. Workings for measures of central tendency – condition 2
45. Workings for measures of dispersion – condition 1
46. Workings for measures of dispersion – condition 2
47. – 49. Workings for Mann Whitney U test
50. Workings for frequency polygon.
