Can You Trust Your Memory When Remembering Lists Of Words?
The purpose of this experiment was to determine whether true and false memories have the same qualitative characteristics. The experimental hypothesis was that the proportion of “remember” (R) responses would be significantly higher for genuinely studied items than for lures.
There were 39 participants comprised of 29 females and 10 males. All participants were Level 2 psychology students at the University of Aberdeen, Scotland, aged 19-31 with a mean age of 21.2 years.
During the study phase of the experiment, participants were shown a sequential list of 120 words taken from Roediger and McDermott (1995), with each word being displayed for 4 seconds and asked to commit those words to memory.
After the study phase, participants were then asked to count backwards for two minutes to prevent rehearsal in working memory. During the testing phase, participants were given a protocol sheet containing 16 studied words, 8 related unstudied words (lures) and 16 unstudied unrelated distracter words and asked to record a remember (R), know (K) or not seen (N) response for each word.
The results from this experiment revealed a statistically significant higher proportion of R responses for studied words than for lures, thereby allowing for the experimental hypothesis to be accepted. On the basis of these findings, it was concluded that true and false memories do differ in their qualitative characteristics.
Memory can be defined as the ability to encode, store and retrieve information. There are however, different types of memory which can affect what type of information is encoded, how long it is stored for and when that information can be retrieved.
Sensory memory for example, which can include both echoic (auditory) and iconic (visual) memory, can be retrieved milliseconds (iconic) or seconds (echoic) after exposure to a stimulus.
Short term memory allows for a slightly longer retrieval time, typically seconds to minutes, whilst long term memory, which includes both declarative and nondeclarative memory, can allow for retrieval of information days or even years after stimulus exposure.
The ability of long term memory to store and retrieve information for such long periods of time is crucial for individual development, as essentially, people are just the sum total of the past experiences that they have had in life.
These experiences, which are stored as episodic memories, allow a person to reflect upon their life and use what they have learned to help them accomplish future tasks and activities.
However, the retrieval of episodic memories may not always yield entirely truthful or accurate information, as sometimes memories may become distorted producing “false memories” which can result in an event being recalled differently to how it originally happened or recalling an event that did not happen at all (Payne, Neuschatz, Lampinen and Lynee, 1997).
Bartlett (1932) for example, found that when subjects were asked to repeatedly recall a story which they had previously read, errors in the recollected events occurred with each recall attempt suggesting that memory somehow became distorted over time.
Although Bartlett’s results have never been replicated by independent researchers, they have still been useful for distinguishing between reconstructive and reproductive memory, as it is with reconstructive memory that distortions are most likely to occur as an individual fills in, or reconstructs, gaps in their episodic memories.
Furthermore, evidence obtained by separate researchers seems to suggest that the more closely related a stimulus is to a previously remembered event, the more likely such distortions are to occur during the reconstructive memory process.
One such example can be found with Underwood (1965) who discovered that when asked to identify which words had been displayed on a previously memorised list of words, subjects were more likely to make a false identification with words that were associatively related to those on the original list.
In a similar experiment carried out by Deese (1959), subjects were given lists of 12 words to remember that were associated with a particular target word. For example, if the target word was “needle”, the list words included related words such as “thread”, “sewing” and “thimble”.
Interestingly, Deese found that some lists resulted in subjects recalling the target word during the testing phase of the experiment even though the target word had not been previously presented to them.
Expanding upon Deese’s original findings, Roediger and McDermott (1995) used both recall and recognition tests on subjects with surprising results. During the recall tests for example, subjects had a high probability of correctly recalling items positioned at the start and end of a list.
However, for items in the middle of the list, subjects had a higher probability of recalling the non presented target word which the words on the list were associatively related to.
These findings seem to suggest that certain memories are more “resistant” to reconstructive errors, whilst other forms of memory, perhaps those which are less firmly encoded, are more vulnerable to reconstructive intrusion errors.
During the recognition phase of the experiment, subjects were then asked to identify which words they had previously seen from a list containing studied words, non presented related target words and non presented unrelated distracter words.
Items which subjects felt that they had previously studied were labelled as “old” whilst words which they felt that they had not previously studied were labelled as “new”.
For words labelled as old, subjects then had to state whether they specifically remembered that word as having been studied or just knew, non specifically, that they had studied that word.
Roediger and McDermott found that subjects recognised around 80% of studied words with the large majority being labelled as “remembered”. For non presented target words, lures, the recognition rate was also around 80% with subjects reporting a “remember” response almost as frequently as they had done with the genuinely studied words. It should be noted however, that for unrelated distracter words, the recognition rate was around 15% with the vast majority being labelled as “know”.
What these findings seem to imply is that within an episodic memory there are certain pieces of information that are firmly encoded (i.e. such as items at the start and end of a list) that provide a framework, or general gist, of what occurred in that episode. This framework appears to be fairly robust, allowing an individual to remember the source of that memory.
However, within this episodic framework are gaps which the individual, as Bartlett suggested, must fill in or reconstruct by themselves. It is during this reconstructive process that information which appears to fit within that particular framework, such as information which is similar or associated to it, may become incorporated within it potentially resulting in the intrusion of a false memory constructed of information which was never originally a part of that particular episode.
Gallo, Roberts and Seamon (1997) have shown that the occurrence of such false memories may be reduced, although not eliminated entirely, if subjects are forewarned about confusing lures with studied words.
This suggests that it may be possible to reduce the occurrence of false memories by altering how information is presented to an individual, and possibly, by also altering how subjects encode that information.
The aim of this experiment was to determine whether true and false memories have the same qualitative characteristics by analysing “remember” (R) responses for studied words and unstudied related lure words.
The experimental hypothesis is that the proportion of R responses will be significantly higher for genuinely studied items than for lures.
This experiment was of a within subjects design as all participants were tested under the same conditions in a single group. The independent variable which was manipulated was the presentation of lures and studied words. The dependent variable which was measured was the number of R responses for the lure and studied items.
39 participants comprised of 29 females and 10 males aged 19-31 with a mean age of 21.2 years took part in this experiment. All participants were Level 2 psychology students at the University of Aberdeen, Scotland.
During the study phase of the experiment participants were shown a list of 120 words taken from Roediger & McDermott (1995). During the testing phase of the experiment, participants were shown a list of 16 studied words, 8 lures and 16 unstudied and unrelated words.
A Microsoft PowerPoint presentation was used to display the list of words onto a projector screen via an overhead projector.
Participants were given a score sheet (see appendix) to record their responses to each of the 40 words presented to them during the testing phase of the experiment.
The initial phase of the experiment began with a study phase which involved presenting a list of 120 words, one word at a time for 4 seconds per word, to all the participants.
Once all 120 words had been presented, participants were then instructed to count out loud, backwards, in sevens from 517 for 2 minutes to prevent rehearsal of the studied items in working memory.
For the testing phase of the experiment, participants were shown 40 words comprised of 16 studied words, 8 related unstudied lures and 16 unrelated and unstudied words.
For each word, participants recorded their response as “R” (indicating that the participant had remembered specific information from the study event), “K” (indicating that the participant knew that they studied the word but does not remember anything) or “N” (indicating that the participant definitely did not study the word) onto the provided score sheet.
At the end of the testing phase, all 39 score sheets were then collected by the experimenter from the participants.
Table 1 – Mean % R responses for lure and studied items in main data group.
% Mean Score
Figure 1 – Bar graph to show mean % R responses for lure and studied items in main data group.
As shown by table 1 and figure 1, the mean % R response in the main data group was 65.9 for the studied items and 39.1 for the lure items. This shows that participants recorded more R responses for studied words than for lures. These results differ by a score of 26.8.
Table 2 – Mean % R responses for lure and studied items in class data group.
% Mean Score
Figure 2 – Bar graph to show mean % R responses for lure and studied items in class data group.
As shown by table 2 and figure 2, the mean % R response in the class data group was 69.8 for the studied items and 41.2 for the lure items. This shows that participants recorded more R responses for studied words than for lures. These results differ by a score of 28.6.
The mean % R response for studied items in the main group data differs from the mean % R response in the class group data by a score of 3.90.
The mean % R response for lure items in the main group data differs from the mean % R response in the class group data by a score of 2.10.
Figure 3 – Scatterplot showing difference scores for each participant in the main data group. Solid line shows 0 reference point, dashed line shows mean reference point.
Figure 4– Scatterplot showing difference scores for each participant in the class data group. Solid line shows 0 reference point, dashed line shows mean reference point.
Figure 3 and 4 show scatterplots displaying the difference scores (%R studied – %R lure) obtained by participants in the main and class group data sets respectively. The figures show that both the main and class group data sets each had 5 difference scores below the reference line of zero. Scores above the zero point reference line indicate participants who scored a higher percentage of “remember” responses for studied words than for lures.
The main group data set had 18 difference scores above, and 21 difference scores below, the mean reference line of 26.8. In contrast, the class group data set had 21 difference scores above, and 23 difference scores below, the mean reference line of 28.6. The mean difference scores for the main and class data sets, as shown by the dashed mean difference line, differ by a score of 1.80.
A 1-tailed paired-samples t test was carried out to determine whether the difference between the Rstudied and Rlure mean scores was statistically significant.
Rstudied items produced a mean score of 65.9 with a standard deviation (SD) of 17.8 (SD = 17.8) and Rlure items produced a mean score of 39.1 (SD = 23.0). A 1-tailed paired samples t test showed that this difference was statistically significant:
t (38) = 6.18, p < .001
As the p value is less than the critical value of .05, it can be said that there is a statistically significant difference between these results and therefore that true and false memories do differ in their qualitative characteristics.
On the basis of this data, the proposed experimental hypothesis that the proportion of R responses will be significantly higher for genuinely studied items than for lures can be accepted, as the 1-tailed paired samples t test suggests that the obtained results are unlikely to have occurred by chance.
The aim of this experiment was to determine whether true and false memories have the same qualitative characteristics.
As shown by the mean % R responses in the main data group, Rstudied items achieved a mean score of 65.9 and Rlure items a mean score of 39.1. The recorded p value of .000 was less than the critical value of .05 indicating that these results were statistically significant. The data obtained from this experiment therefore provides statistically significant support for the experimental hypothesis allowing for it to be accepted.
On the basis of these results, one may state that subjects in this experiment were able to differentiate the qualitative characteristics of true and false episodic memories. This is in contrast to the results obtained by Roediger and McDermott (1995), who found that subjects were not able to differentiate between true and false memories reporting very similar “remember” responses for studied words and unstudied target words.
A possible explanation which may account for these findings can be found in the way that this experiment was conducted, as subjects were forewarned that during the testing phase they would be presented with both studied and unstudied words. Having this prior knowledge may have somehow primed the participants to become more aware of the words that they were presented with during the study phase of the experiment.
This in turn may have lead to a stronger encoding of those words into episodic memory and a subsequent greater resistance to false memory intrusion by the unstudied lure words.
In this sense, there is some similarity with Roediger and McDermott’s results, as they discovered that words at the start and end of a list tended to have the highest recall rates and were the most resistant to false memory intrusions.
However, the fact that subjects did still provide a 39.1% “remember” response for unstudied lure words in this experiment, shows that although prior knowledge of events can reduce the occurrence of false memories, it does not eliminate them completely.
Results obtained by Gallo et al. (1997) do seem to support these assertions, as Gallo discovered that groups who were warned about confusing studied and unstudied words tended to form less false memories than those who were not forewarned.
McDermott and Roediger (1998) provide further support for these findings, as their results also showed that false memories could be reduced when subjects were warned about lures and given a practice trial.
If one takes into account the results obtained by Roediger and McDermott (1995) which showed an almost equal number of “remember” responses for both studied words and unstudied lure words, and then the subsequent reduction of false memories with prior knowledge as shown by Gallo et al. (1997) and the results obtained by this experiment, one may naturally question as to why false memories are not eliminated entirely?
The answer to this may lie in how an individual encodes information into memory. Winograd, Peluso and Glover (1998) for example, found that participants who formed vivid mental images when presented with words during the study phase of the experiment were more likely to suffer from false memories during the testing phase.
One account given for these findings by Winograd et al. (1998), is that individuals who do form vivid mental images may in fact suffer from source memory confusions which impede upon their ability to differentiate between internal and external events.
Another possible cause of source confusions are implicit associative responses (Dodson, Koutstaal and Schacter, 2000), which occur when a participant thinks of the target word through a process of association made with the related studied words during the study phase of the experiment.
As a result, during the testing phase of the experiment, the participant may then incorrectly “remember” a lure as being part of the original set of studied words. Evidence showing that presenting more related words tends to increase this false memory effect (McDermott, 1997), does seem to support the idea that word associations can indeed lead to source confusions and thus false memories.
The results of Winograd et al. (1998) and Dodson et al (2000) suggest that false memories are in fact an inherent part of the encoding and retrieval process of memory.
The pervasiveness of these false memories however, appears to be influenced by how an individual attempts to remember a piece of information and the type of information they are remembering.
Following this line of reasoning, one could therefore expect different individuals to display differing levels of false memories which is something that could be investigated in future research.
This experiment exposed participants to 16 studied, 8 lures and 16 distracter words during the testing phase. Further research could also investigate the effect of having an equal number of studied and lure words, as the higher proportion of studied words in comparison to lure words, may in fact have lead to the higher percentage of “remember” responses for the studied words seen in this experiment.
The effect of different memory encoding strategies could also be investigated. One could for example, compare the occurrence of false memories in groups who used wrote repetition, word associations, rhymes or method of loci memorisation techniques.
This may then help to cast light on what sort of memory encoding techniques tend to maximise false memory intrusions and which techniques can be used to minimise false memory intrusions.
Between the study and testing phase of this experiment, subjects were asked to count backwards for a period of 2 minutes. It would be interesting to investigate what effect extending or shortening this period would have to determine what sort of an influence time has on false memory intrusions.
Finally, it is worth noting that participants in this experiment had a mean age of 21.2 years and were all Level 2 psychology students. As a result, it could be argued that these participants were not a truly representative sample of the population as a whole, and therefore, that this study lacked ecological validity.
The results of this experiment showed a statistically significant higher proportion of “remember” response rates given by participants for studied items than for lures.
As a result, the proposed experimental hypothesis that the proportion of R responses will be significantly higher for genuinely studied items than for lures, can be accepted. On the basis of these findings, true and false memories do indeed appear to differ in their qualitative characteristics.
Bartlett, F.C. (1932). Remembering: A study in experimental and social psychology. Cambridge, England: University Press.
Deese, J. (1959). On the prediction of occurrence of particular verbal intrusions in immediate recall. Journal of Experimental Psychology 58, 17-22.
Dodson, C.S., Koutstaal, W., & Schacter, D.L. (2000). Escape from illusion: Reducing false memories. Trends in Cognitive Sciences 4, 391-397.
Gallo, D.A., Roberts, M.J., & Seamon, J.G. (1997). Remembering words not presented in lists: Can we avoid creating false memories? Psychonomic Bulletin & Review 4, 271-276.
McDermott, K.B. (1997). Priming on perceptual implicit memory tests can be achieved through presentation of associates. Psychonomic Bulletin & Review 4, 582–586.
McDermott, K.B., & Roediger, H.L. (1998). Attempting to avoid illusory memories: Robust false recognition of associates persists under conditions of explicit warnings and immediate testing. Journal of Memory and Language 39, 508–520.
Payne, D.G., Neuschatz, J.S., Lampinen, J.M., & Lynee, S.J. (1997). Compelling memory illusions: The qualitative characteristics of false memories. Current Directions in Psychological Science 6, 56-60.
Roediger, H.L., & McDermott, K.B. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory & Cognition 21, 803-814.
Underwood, B.J. (1965). False recognition produced by implicit verbal responses. Journal of Experimental Psychology 70, 122-129.
Winograd, E., Peluso, J.P., & Glover, T.A. (1998). Individual differences in susceptibility to memory illusions. Applied Cognitive Psychology 12, 5–27.