--> Europe's Journal of Psychology ejop.psychopen.eu | 1841-0413 Research Reports Unconscious Plagiarism in Recall: Attribution to the Self, but not for Self-Relevant Reasons

Previous research has shown that if people improve other’s ideas, they subsequently unconsciously plagiarise them at a dramatically higher rate than if they imagine them, or simply hear them again. It has been claimed that this occurs because improvement resembles the process of generation, and that these are confused during retrieval. However, an alternate possibility is tested here: plagiarism may increases because improvement increases personal relevance of the ideas. Two studies were conducted in which there was an initial generation phase, followed by an elaboration phase in which participants imagined the previous ideas, improved them for their own use, or improved them for an older adult’s use. One week later, participants attempted to recall their own ideas, and generated new solutions to the previous problems. In both studies, improvement of doubled the rate of subsequent plagiarism in the recall own task, but this effect was not mediated by whether people improved ideas for their own use, of for use by someone else. Improvement had no effect on plagiarism in the generate-new task. These studies therefore rule out personal relevance, or personal semantics as the source of the improvement effect in unconscious plagiarism.

The creation and ownership of ideas is core to the professional activity of academic researchers in all disciplines.
As teachers, we strive to teach our students the importance of citing previous work where appropriate, and as researchers we endeavour to be scrupulous in acknowledging the work that has gone before.However, in order to publish and make a contribution to the literature, we must demonstrate originality in our methods, findings, or ideas.But how do we know if we have been genuinely novel, and that we haven't failed to acknowledge the work of others?Fortunately, the peer-reviewed publication system offers some protection, and so consequently public disputes about the origin of ideas are rather rare.However, in our experience, private disputes and resentments about the source of an idea are rather more common.It is rare that someone works entirely alone, without discussing ideas with colleagues or students, and even such academic isolates they will have read the work of others or heard talks at conferences.Given the broad literature on source memory errors it is inevitable that there will be occasions on which we are mistaken about the source of our ideas and risk unconsciously plagiarising others.
The experimental literature on unconscious plagiarism began with Brown and Murphy (1989), who developed the now-standard three-stage paradigm.Initially, groups of participants take turns to generate solutions to a problem.Subsequently, participants attempt to recall their own solutions, avoiding those generated by others (the recall-own --> Europe's Journal of Psychology ejop.psychopen.eu| 1841-0413 task).Finally, participants attempt to generate new solutions avoiding those mentioned in the first session (the generate-new task).In both tasks, Brown and Murphy (1989) reported significant rates of plagiarism.That is, when attempting to recall their own ideas people recalled the ideas of others, and when generating new ideas they plagiarised old ideas, including their own.Subsequent research has explored many of the factors that exacerbate or mitigate the magnitude of these plagiarism errors.However, rather than review that literature here we refer the interested reader to a recent review (Perfect & Stark, 2008a).Instead, we focus here on one finding that has emerged from this literature -the idea-improvement effect.
In a series of studies, we have investigated the impact of different forms of elaboration of ideas during the retention interval between the initial idea generation and the subsequent recall-own and generate-new tests (Perfect, Field, & Jones, 2009;Perfect & Stark, 2008b;Stark, Perfect, & Newstead, 2005;Stark & Perfect, 2006, 2007, 2008).
Our rationale for investigating the impact of thinking about the ideas during the retention interval was that real-world cases of plagiarism are unlikely to have resulted from a single exposure to an idea, without the accused having given considerable thought to the development of the idea under dispute.In our studies we have contrasted two forms of idea elaboration: idea imagery, which requires participants to form mental images of the ideas, and idea-improvement which requires participants to improve the ideas that they heard previously.Across many studies, the results have been remarkably consistent.Compared to control, both forms of elaboration improve correct recall equally, whilst neither affects the rate of plagiarism on the generate-new task.Idea imagery also has no impact upon the rate of plagiarism on the recall-own task.However, idea-improvement dramatically increases rates of plagiarism in the recall-own task.
In absolute terms, improving ideas once can result in three times the base rate level of plagiarism (e.g.Stark et al., 2005) whilst improving ideas twice can result in up to 13 times the number of plagiarised ideas (Stark & Perfect, 2008).This idea-improvement effect survives financial inducement to avoid plagiarism (Stark, Perfect, & Newstead, 2005), and the use of a source-monitoring test in place of a recall test (Stark & Perfect, 2007).It is also magnified by increasing the delay between idea generation and final test (Stark & Perfect, 2007).In contrast, neither repetition, nor delay impact upon the rate of plagiarism following imagery (Stark & Perfect, 2007, 2008).
We have sought to explain the differential effect of imagery and idea-improvement on plagiarism rates in terms of the source-monitoring framework (Johnson, Hashtroudi, & Lindsay, 1993).We have argued that improving ideas requires a process of generation, similar to the original phase of idea generation itself.Thus, when participants try to recall an original idea they may rely on records of generative processes as a cue to source, and so confuse original ideas with improved ideas.In contrast, imagery elaboration creates records of visual (or other sensory) records, which are readily distinguished from originally generated ideas.However, both generative processes, and imagery processes are cues to oldness, and so both forms of idea elaboration are equally useful cues in the generate-new task, where the focus is on avoiding familiar responses.Thus, we have argued that a source monitoring account can explain the differential outcomes of the two forms of elaboration across the recall-own and generate-new tasks.
Whilst we have favoured a process-oriented account of the idea-improvement effect, we also acknowledge that improvement and imagery result in very different ideas.Imagining an idea does not alter the fundamental nature of the idea itself: it merely changes a verbal proposition into a visual image.However, improving an idea alters the nature of the proposition itself.Thus, one possibility is that people plagiarise not because of the process applied to those ideas, but because the content of those ideas has changed.Unconscious Plagiarism in Recall 276 We have previously tried to address the issue of memory content in two ways, and we introduce a third in the present work.The first such attempt looked at whether it was the amount of information about an idea that was the cause of plagiarism.In all our earlier studies, participants generated 3 improvements to an idea.Thus, improved ideas were considerably richer in detail than unimproved ideas, and thus we reasoned that it might be this richness that led to plagiarism in the recall-own task.We explored this issue by having participants imagine ideas that had been improved by someone else (Stark & Perfect, 2006).In this manner, the content of improved and imagined ideas was matched.However, there was no evidence that imagining these already-improved ideas led to plagiarism at any greater rate than control items that were neither imagined nor improved in the interval.
Another possibility is that it is not the richness of the ideas that causes plagiarism, but rather the valence, or quality.If participants are genuine in their attempts to improve the ideas then perhaps they later plagiarise those ideas because the ideas are somehow better.Perfect and Stark (2008b) explored this possibility.Participants generated ideas in the standard manner.However, prior to the elaboration phase, participants were led to believe that their ideas had been evaluated by a panel of judges.They were told that some ideas were judged excellent, and required no further elaboration.Other ideas, judged very good required one improvement, whilst ideas judged merely good required three improvements.Finally, ideas judged to be satisfactory did not merit further improvement.
In fact, the assignment of the original ideas to these ratings was entirely arbitrary.Thus, in this design we manipulated both the perceived quality of the ideas and the amount of improvement they received.What we found was that for recall-own plagiarism the number of improvements predicted the rate of plagiarism, whilst rated quality of the ideas had no impact.
This was not because the rated quality of the ideas had no impact.For the generate-new task, plagiarism was influenced by rated quality but not by number of improvements.This pattern was replicated by Perfect, Field, and Jones (2009).They had participants generate ideas with a partner who was an expert or a novice in the topic under discussion.In fact, the partner was a confederate who generated identical ideas whatever their putative expertise.Subsequently, there was an improvement phase for half the ideas, before participants recalled their own ideas, or generated new ones.Source credibility had no impact upon plagiarism in the recall-own task, but was associated with generate-new plagiarism.Conversely, improvement had no impact on generate-new plagiarism, but increased recall-own plagiarism.Thus, the pattern across these studies, with respect to the idea-improvement effect on recall-own plagiarism is that it is not caused by the perceived quality of the ideas.
Here we focus on a third potential outcome of idea-improvement, namely self-relevance.One possibility is that when asked to improve an idea, participants make improvements that they would wish, in order to fit their own aesthetic sense or personal semantics.Imagine a participant asked to improve the idea of using a brick as a doorstop.They might think that the brick could be carved, painted, and mounted on cork to prevent scratching the floor.But what shape would they mentally carve it, what colour would they mentally paint it and where would they mentally place it?It is plausible that participants might imagine a brick-doorstop in their own home, and improve it to fit their own taste or decor.Later, when they attempt to recall the original ideas, thinking of the doorstop in the context of their home, with personally relevant details, might lead them to erroneously conclude that the idea must have been their own.Thus, it may not be the process of improvement that leads to plagiarism, but the self-relevance of the outcome.
We present two studies investigating the role of self-relevance in unconscious plagiarism.Experiment 1 utilises the same task at generation as we have used in the majority of our previous work, namely the Alternate Uses Test (Christensen, Guilford, Merrifield, & Wilson, 1960).This task requires participants to generate alternate uses for common objects such as a shoe, a paperclip or a brick.Experiment 2 replaced the Alternate Uses Test with a more real-world task based on the lives of our undergraduate participants, but was in all other respects identical to the first experiment, and so we report them together.The use of real-world problem solving tasks has successfully generated unconscious plagiarism in a number of studies (Marsh, Landau, & Hicks, 1997;Perfect, Field, & Jones, 2009;Perfect, Defeldre, Elliman, & Dehon, 2011).In both experiments, participants were asked to generate ways of improving the ideas either for themselves, or an older adult.
The aim of each study was to determine whether the idea-improvement effect is the result of the process of generation of improvements to ideas, or occurs because improved ideas are more self-relevant to the participants.
A difficulty in testing this hypothesis is that we cannot know what aspects of the idea are under consideration when a person improves an idea, or when they mistakenly recall the idea as their own.For this reason, we rejected use of self-report scales to determine the self-relevance of improved or imagined ideas.Instead we altered the nature of the improvement instructions to focus on either self-relevant improvement, or other-relevant improvement.
In the self-improvement instructions, we asked participants to improve the ideas presented to them so that they were more likely to use the idea in their lives.In the other-relevant instructions, we asked participants to improve the idea for everyday use by an older adult.We chose this group as the "other" category because it is a well-understood category, but not one we thought our younger participants would identify with.The effects of these two forms of idea-improvement (self-relevant and other-relevant) were contrasted with idea imagery and a no-elaboration control condition, as in our previous studies.Our expectation was that we would replicate the idea-improvement effect, and that the extent to which plagiarism of ideas subject to self-and other-relevant improvement differ would reflect the role played by self-relevance in the effect.

Method Participants
Experiment 1: Thirty undergraduate students were tested for the idea generation phase.However, 3 participants failed to attend the second testing session and so only 27 participants completed the experiment.
Experiment 2: Thirty-two undergraduate students were tested the initial generation phase.However, three participants failed to attend the second testing session and so only 29 participants completed the experiment.
Additionally, 1 participant failed to complete the generate-new phase in the second session, and so data for only 28 participants were analysed for that task.
For both experiments, participants were undergraduates from the University of Plymouth and participated either to achieve partial fulfilment of a course requirement or for payment of £8.

Procedure
Experiment 1: Participants were tested in groups of 4. Members of each group were told they would be presented with object names, and they would have to think of novel uses for those items.As an example, they were told that they might hear an object like a newspaper, and a novel use might be making it into a fan.The experimenter then read out the first object name (brick, shoe, paper-clip or button depending upon counterbalancing condition), and participants were instructed individually by the experimenter to generate an idea for the group to hear.The order that participants were asked for their idea was randomised, such that participants did not know when they would have to speak.Participants were told to listen to the ideas from the group, to avoid repeating previous ideas.Once Unconscious Plagiarism in Recall each participant had provided an idea, participants were asked for their next idea, again in randomised order.
This process was repeated until each participant had generated 4 ideas for each of the 4 objects.
The elaboration phase immediately followed the generation phase.Of the previously generated ideas, a quarter (one idea from each participant, from each category) was then subject to the following condition treatments.For the self-relevant idea-improvements participants had to write down 3 ways that the ideas could be improved such that they would be more likely to use the idea in their own lives.For the other-relevant idea-improvements participants had to write down 3 ways that the ideas could be improved such that an older adult (65 years or older) would be more likely to use the idea in their daily lives.For imagery items, participants rated how easy it was to imagine the idea, and how effective that idea would be.Control ideas were not re-presented at this stage.The order that participants performed these tasks was counterbalanced.This task completed the first session, which lasted approximately 60 minutes.
One week later, participants returned as a group to complete the recall-own and generate-new tasks, in that order.
In the recall-own phase, participants were given a response sheet with the 4 object name cues on, with 4 blank spaces under each.Participants were asked to recall their own ideas for each object that they had generated during the first session.They were instructed not to guess, and that they could leave blank spaces if they could not remember all their ideas.
Immediately following the recall task, participants were given a second response sheet, with the same object name cues, again with 4 blank spaces under each.For this task, participants were asked to generate 4 new ideas for each object, with instructions not to use any of the ideas from the previous session, from any of the objects.
For this task, participants were required to generate 4 ideas.
Experiment 2: The procedure was identical to Experiment 1, except that the Alternate Uses Task was replaced with a task of generating ideas to enhance: 1) their University course, 2) their University's facilities, 3) The city centre and 4) their student life.Participants were presented with the first topic and instructed one at a time to share their idea with the group.During the elaboration phase, we replaced the term "older adult" with "mature student aged over 65", to ensure the relevance of the improvement task to older adults.

Results
In each study, there were three dependent measures: level of correct recall of own ideas, plagiarism of other's ideas in the recall-own task, and plagiarism of ideas in the generate-new task.Inspection of the data for these measures in Table 1 reveals the standard pattern we have reported previously for both experiments.For recall, all forms of elaboration improved correct responding relative to control.For the generate-new task, there was an overall effect of elaboration on plagiarism, with control ideas plagiarised more often than other ideas, which did not differ.However, for the recall-own task, plagiarism was inflated following idea-improvement, as in our previous studies.Below we report the statistical analyses for both experiments.

Correct Recall
Experiment 1: A within-subjects ANOVA a significant main effect of elaboration status F(3,78)=11.52, p<.001, MSE = 9.52.Sidak-adjusted tests for multiple comparisons revealed that ideas that were subject to any type of elaboration (imagery, self-or other-relevant improvement) were recalled at a significantly higher rate than control, but the different forms of elaboration did not differ from one another.Table 1.
Mean rates of correct recall and plagiarism within the recall-own (RO) & generate-new (GN) phases for control ideas, ideas that were subject to imagery-elaboration, self-relevant and other-relevant elaboration.Experiment 2: Ideas that were subject to elaboration of any kind to any (imagery, self-or other-relevant improvement) were recalled at a numerically higher rate than control but a within-subjects ANOVA revealed that this difference did not reach statistical significance F(3,84)=2.12,p=.10, MSE 2.79.

Unconscious Plagiarism in the Recall-Own Task
In this task, participants were required to remember as many of their own initial ideas as possible, but recall was not forced.Unconscious plagiarism occurred when participants recalled someone else's idea as their own.A plagiarised idea was only counted once.
Experiment 1: A within subjects ANOVA revealed that there was a significant main effect of elaboration status on plagiarism in the recall-own task, F(3, 78)=9.40,p<.001, MSE = 9.25.Sidak-adjusted multiple comparisons across conditions were revealed that the rate of plagiarism following self-or other-relevant improvements ideas did not differ, but both exceeded the rate seen in the imagery and control conditions, which did not differ.
Experiment 2: A within subjects ANOVA of plagiarism errors in the recall task revealed that there was a significant main effect of elaboration status F(3, 84)=6.52,p<.001, MSE = 8.70.Mean plagiarism levels across conditions were compared using Sidak-adjusted multiple comparisons.These revealed that the rate that improved ideas, whether self-or other-relevant, were plagiarised at a greater rate than imagery or control ideas.No other comparison was significant.

Unconscious Plagiarism in the Generate-New Task
Experiment 1: A within subjects ANOVA was conducted on plagiarism errors in the generate-new task revealed a significant main effect of elaboration status on plagiarism errors F(3,75)=6.19,p<.05 MSE = 3.52.Sidak-adjusted multiple comparisons demonstrated that plagiarism of control ideas significantly exceeded the rate for imagined ideas and other-relevant improved ideas, but not self-relevant improved ideas.There was no evidence of a difference between self-relevant and other-relevant improvement on generate-new plagiarism.
Experiment 2: A within-subjects ANOVA revealed that elaboration status did not affect the rate of plagiarised errors, F<1.