Duplication and redundancy can increase the size of the code, make it hard to understand the many code variants, and cause maintenance headaches. The goal of avoiding redundancy has provided the impetus to investigations on software reuse, software refactoring, modularization, and parameterization. Even in the face of the ethic of avoiding redundancy, in practice software frequently contains many redundancies and duplications. For instance the technique of "code scavenging" is frequently used, and works by copying and then pasting code fragments, thereby creating so-called "clones" of duplicated or highly similar code.
Redundancies can also occur in various other ways, including because of missed reuse opportunities, purposeful duplication because of efficiency concerns, and duplication through parallel or forked development threads. Because redundancies frequently do exist in code, methods for detecting and removing them from software are needed in many contexts.
Over the past few decades, smatterings of research on these issues have contributed towards addressing the issue. Techniques for finding similar code and on removing duplication have been investigated in several specific areas such software reverse engineering, plagiarism in student programs, copyright infringement investigation, software evolution analysis, code compaction e. Common to all these research areas is the problems involved in understanding the redundancies and finding similar code, either within a software system, between versions of a system, or between different systems.
Although this research has progressed over decades, only recently has the pace of activity in this area picked up such that significant research momentum could be established. This seminar gathers leading scientists from all different areas related to software redundancy and young researchers ready to pick up the ball. Reflections, Conclusions, and Acknowledgments The remaining entries in this proceedings consist of one of three types of entries. The first are summaries of the keynote presentations. Following Kruschke ,. Deflections from the grand mean representing effect of condition were constrained to sum to zero across conditions.
Using the data sample mean and standard deviation to set parameters of the prior ensured that the prior distribution was scaled appropriately to the data Kruschke, To test for the possibility of lateral left versus right attentional bias, along with redundancy gains, we estimated RTs in two different ways. In the first case, to check for the possibility of lateral asymmetries in performance, data were coded such that two single-target conditions represented the left single-target and right single-target trials. Thus, any difference in RTs in the first case would signal that participants had tended to respond to targets in one location faster than the other.
In the second case, to provide a conservative estimate of the redundancy gain, data were coded such that the two single-target conditions represented the faster and slower mean single-target condition for each participant. Redundancy gain was defined as the difference between the shorter of the two single-target RTs and the redundant-target RT.
This method of measuring redundancy gains provides more conservative estimates than the alternative approach of comparing redundant-target RT to the mean of the single-target RTs cf. Each parameter estimate was based on four MCMC chains, run for burn-in steps, followed by , steps each. Chains were thinned to every fifth step in to reduce sample autocorrelation, leaving a total of 50, samples for analysis. Detection error rates were analyzed to ensure that participants had correctly followed instructions.
As a general rule, the capacity coefficient is robust against error rates of up to 0. In all experiments, RTs were only analyzed for correct target-present trials i. Inspection of the data suggested that participants generally complied with the instructions to respond to targets bimanually, making button presses with both thumbs in quick succession. Analyses were carried out using the RT for the faster of the two button presses for each trial.
The mean single-target RT provides a measure of baseline response speed independent of any redundancy gain. As noted above, redundancy gains were calculated by subtracting RT for the redundant-target condition from RT for the faster single-target condition left or right for each participant. Values equivalent to zero represent UCIP processing, positive values indicate super-capacity, and negative values represent limited capacity. In Experiment 1a, the participant-controlled cursor was invisible, and participants were told to ignore the movements of the red dot of the tracking task.
However, joystick movements were recorded. These data provided an estimate of chance-level tracking accuracy, suitable as a baseline against which to compare active tracking performance in the subsequent dual-task experiments. Performance was measured by calculating the RMSE in angular distance of the cursor position relative to the target position.
Mean RMSE was If participants followed instructions to ignore the tracking task in Experiment 1a and perform it in the subsequent dual-task experiments, RMSE should be smaller in the later experiments. The goal of Experiment 1a was to provide a baseline measure of processing efficiency before any secondary task load was added.
Resilience for redundant-target processing was highly limited, despite attention being wholly focused on the target detection task. Thus, within a standard distractor-present redundant-target task, the RT gains produced by redundant target presentation were smaller than predicted by statistical facilitation in a UCIP model. Experiment 1b replicated the procedure of Experiment 1a but with the addition of a central manual tracking task, to test whether concurrent task load reduced processing resilience.
As we aimed to match sample size from Experiment 1a, we ran participants until we had data for 25 participants who met the inclusion criteria for detection error rates.
None had participated in the previous experiment. The apparatus and stimuli were identical to those used in Experiment 1a, except that the cursor in the pursuit tracking task was made visible. In Experiment 1b, participants performed the peripheral target detection and manual tracking tasks concurrently. Participants were encouraged to maintain accuracy on both tasks, while also aiming to minimize RTs on the detection task. As in Experiment 1a, the task involved one block of tracking intervals, comprising one s practice interval followed by 20 s experimental intervals.
As with Experiment 1a, participants with false alarm or miss rates greater than 0. Data for three participants with excessive false alarm rates ranging 0. Thus, data suggest that participants in Experiment 1b engaged in the tracking task as instructed. To test the possibility of a tradeoff in performance between the target detection and tracking tasks, bivariate correlations were calculated between Rz scores and RMSE. The credible interval on the correlation included a value of 0.
Experiments 1a and 1b tested whether a manual tracking task impairs processing efficiency for redundant visual targets. Consistent with previous findings Eidels et al. More surprisingly, resilience did not appear to suffer with the addition of a concurrent, central tracking task. Experiments 1a and 1b failed to find a clear difference in processing efficiency for redundant visual targets between single- and dual-task conditions.
However, it is possible that the between-participants design of Experiment 1 simply was not sensitive enough to detect differences between the single- and dual-task conditions. To address this issue, Experiment 2 used a within-participants design to replicate Experiments 1a and 1b, providing a second test of the relationship between task load and target processing efficiency. No participants had performed any of the previous experiments. Participants were all fluent in English, with normal color vision, and normal or corrected-to-normal visual acuity.
The tracking and discrimination tasks were performed in the same way as in Experiment 1. However, each participant completed two blocks of trials. In one block, the participant performed the target-detection task alone, following the same procedure as Experiment 1a single-task condition. In the other block, the participant performed both tasks simultaneously, as in Experiment 1b dual-task condition. Block order was counterbalanced across participants.
At the beginning of each block, participants were given a s practice session, before completing 20 s intervals. Participants were given a short break between blocks. As in the previous experiments, participants finished the testing session by completing the FLANDERS questionnaire and recording their driving experience.
Analysis was as in Experiment 1, but was adapted to account for the within-participant manipulation of task load. Analysis of RTs now included additive effects of task load and the interaction of target condition by load Kruschke, ,. Preliminary inspection found no effect of block order on any of the measures. As such, all analyses were carried out collapsed across block order.
For support, please contact
Analyses excluded data from three participants with excessive error rates ranging from 0. These exclusions left data from 27 participants for analysis. These results replicate the findings of Experiment 1, showing no credible effect of task load on dual-channel processing efficiency. As in Experiments 1a and b, resilience was highly limited, but was not credibly smaller when participants performed a concurrent manual tracking task. Thus, the tracking and detection tasks did not appear to compete for processing resources Wickens, , , producing no performance tradeoff between the tasks.
The previous experiments found that processing efficiency for redundant visual targets, as measured by resilience, was similar across single- and dual-task conditions. In both cases, resilience was limited, producing mean Rz scores decisively below 0.
- Business Vocabulary In Use Elementary-Answers!
- The Claiming of Sleeping Beauty (Sleeping Beauty Trilogy, Book 1)?
- Parallel Systems: Redundancy in Government by Jonathan B. Bendor, Hardcover | Barnes & Noble®.
- An Introduction to the Mental Model of Redundancy (with examples).
Experiment 3 sought to generalize the results of Experiments 1 and 2 by testing the effects of dual-task load on Rz under conditions in which the baseline, single-task resilience scores were not highly limited. Although neither of the first two experiments included a manipulation to diagnose system architecture, the observed resilience scores suggest that the left and right channels in the target detection task were processed in parallel. As noted above, a serial processing architecture can produce limited-resilience processing. This type of processing only occurs when the time needed to process a distractor is significantly lower than the time needed to process a target Little et al.
Schloss Dagstuhl : Seminar Homepage
There is little reason to expect that this would have been the case in Experiments 1 and 2. Experiment 3 measured resilience under single- and dual-task conditions using target and distractor stimuli designed to force serial processing and push resilience above the levels observed in the first two experiments.
This design meant that stimulus onsets could still be detected in the retinal periphery. However, to identify targets and distractors, participants had to foveate the stimuli, with little peripheral information to guide participants preferentially toward the target on single-target trials. Assuming that target and distractors required roughly the same amount of time to process, resilience should have reached super-capacity levels Little et al.
We planned for a sample size of 30 participants who met the performance criteria for both the target detection and tracking tasks. No participants had participated in any of the previous experiments. Participants all exhibited normal or corrected-to-normal visual acuity, normal color vision, and English fluency. The apparatus was the same as above. Stimuli were similar except for the following changes. First, stimuli for the target detection task were reduced to 4-point font, with each letter subtending approximately 0.
Second, targets and distractors were embedded within five-item letter arrays. Target letters appeared in the same upper left and upper right locations as in Experiments 1—2, but were flanked on both sides by two letters randomly and independently selected from the set F, H, K, M, N, V, W, X, and Z. For an illustration of a dual-task single-target trial from Experiment 3, please return to Fig. Data from one participant were removed from analysis for high false alarm rates in both the single- 0.
Furthermore, data from two participants who produced roughly equal tracking error in both the single- and dual-task conditions e. Experiment 3 assessed target processing efficiency within a forced serial process paradigm. As expected, serial scanning produced super-capacity processing of redundant targets Little et al.
But, despite the difference in processing architecture between experiments, none of the experiments found an effect of task load on processing efficiency.
In This Section
Resilience remained largely unaffected by variations in task load, despite large variations in baseline resilience values and changes to processing architecture. The current studies examined redundant-target processing within a dual-task paradigm. As expected, a concurrent manual tracking task increased RTs for target detection in the periphery. But, despite this difference in baseline target detection times, the efficiency with which redundant targets were processed was invariant with task load.
In other words, a central task slowed responses to peripheral targets, but did not change the rate at which multiple targets were processed relative to single targets. This effect was true regardless of whether targets were processed in parallel with limited resilience Experiments 1—2 , or in serial with super-capacity resilience Experiment 3.
One interpretation is that the central manual tracking task and the peripheral target detection task tapped into partially independent pools of information-processing resources Wickens, Although multiple resource theory includes visual attention as one form of processing resource, it posits separate pools of processing resources for both focal and ambient vision, linking focal processing to the central visual field and ambient to the peripheral visual field. The theory thus allows that the task-load manipulation might not have affected processing efficiency because the tracking task engaged central resources and the target detection task engaged ambient resources.
Contrary to this hypothesis, though, mean single-target RTs for dual-task conditions were credibly longer than those for single-task conditions in across all experiments. These results suggest the central tracking and peripheral detection tasks likely tapped common processing resources, presumably at the stages that Wickens labels perception or cognition; the target and distractor stimuli of Experiment 3 were in fact designed to be indiscriminable in ambient vision, ensuring competition for focal attention between the central and peripheral tasks.
Moreover, the Wickens model proposes that focal processes are specialized for detailed object perception and recognition, whereas ambient processes are specialized for spatial processing. Assuming that participants fixated near the display center to perform the tracking task, the central and peripheral processing demands in the current experiments, therefore, would not have aligned well with the attentional pools hypothesized by multiple resource theory.
To optimize the distribution of resources under the model, participants would have had to fixate near the boundary of the display while tracking the moving target with peripheral vision. Eye movement data might test whether any participants adopted such a strategy, or to estimate more generally how often eye movements occurred between the central and peripheral tasks.
At best, though, the data indicate that distributing task load over different resource pools would have attenuated dual-tasks costs, not eliminated them. By this account, participants would have performed the central tracking task while using a diffuse attentional window to monitor the display periphery for targets and distractors Steelman et al. In Experiments 1—2, attention in this interval would have been spread broadly over the left and right stimuli, processing them in parallel.
By contrast, the design of the stimuli in Experiment 3 would have demanded that attention focus on the stimuli in serial, through a series of saccadic eye movements. In both cases, after detecting a target or confirming that both peripheral items were distractors, attention would have returned to the tracking task. Resilience would have been similar across the single- and dual-task blocks, because, in both cases, attention would have been disengaged from the central task while peripheral items were being processed.
One caveat of this attention-switching account is that such a theory predicts a positive association between tracking error and resilience for the redundant targets, whereas our data trended in the opposite direction. The tendency toward better tracking among participants with higher resilience hints at individual differences in effort or ability, differences that might have masked any tradeoffs between tracking and resilience. In application, our results indicate that redundant visual signals are likely to be as effective at aiding visual detection under multi-task conditions as under single-task conditions.
This means both that redundant coding will be useful within multi-task workspaces, and that the results of single-task pilot testing can be used to predict the magnitude of RT gain that redundant signals will purchase in a multi-task environment. Thus, design guidelines for complex visual workspaces, such as pilot cockpits or vehicle dashboards, should encourage the use of redundant coding of visual alerts for enhancing detection. The data also imply a tradeoff between redundancy gains and display complexity.
We find that redundant visual targets in peripheral visual displays are of greatest value for low-salience stimuli, those that demand focused attention for detection or recognition, such as when monitoring a large set of gauges or meters. Stimuli of higher salience, discriminable enough to be processed in parallel, are more likely to be processed with limited resilience and with far more modest redundancy gains.
This pattern suggests that as a general guideline, display designers might trade redundant target presentation against target salience, reserving highly salient display modes for the most critical signals and presenting information that is less urgent but still time-sensitive in lower salience, redundant signals. By using redundant presentation as a substitute for high conspicuity, this strategy would reduce the risk of a salience-saturated environment in which high-contrast signals compete for attention or overwhelm the operator. Notably, our findings only consider identical redundant visual signals.
Within a peripheral redundant-target paradigm, data give no evidence for poorer target processing efficiency while under the load of a secondary tracking task. As expected Little et al. Findings suggest there is a modest benefit to employing redundant targets in peripheral visual displays e. However, we find that redundant displays have more substantial benefits for target items that demand serial processing. Many thanks to Paul Douglas for assistance with programming, and to Kate Coden and Silas Ellery for assistance with data collection.
SAM collected the data. All authors read and approved the final manuscript. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Stephanie A. Morey, Email: ua. Nicole A. Thomas, Email: ua. Jason S. McCarley, Email: ude. National Center for Biotechnology Information , U. Cogn Res Princ Implic. Published online Feb Morey , 1 Nicole A. Thomas , 1, 2 and Jason S. McCarley 3. Author information Article notes Copyright and License information Disclaimer. Corresponding author.
Received Mar 16; Accepted Dec Abstract Monitoring visual displays while performing other tasks is commonplace in many operational environments.
Keywords: Capacity coefficient, Limited capacity, Multi-task, Redundancy gain, Redundant signals effect, Super-capacity, Target detection, Workload capacity, Workload resilience. Significance statement High-workload environments often mean dividing attention between multiple visual tasks or displays. Background Operators in high-stress domains often need to divide attention between the central and peripheral visual fields.
Measuring the efficiency of redundant-target processing In a standard redundant-target task, participants make a speeded response to a target presented in either of two channels e. General method Here, we describe methods of stimuli and procedure common to all of the experiments that follow. Open in a separate window. Procedure Participants completed the task in a well-lit room. Experiment 1a Experiment 1a provided a baseline estimate of resilience for a parafoveal target detection task performed alone i. Apparatus and stimuli In Experiment 1a, stimuli for the target detection task were a black capital T target and L distractor presented in point Arial font 1.
Results Error rates Detection error rates were analyzed to ensure that participants had correctly followed instructions. RTs In all experiments, RTs were only analyzed for correct target-present trials i. Tracking performance In Experiment 1a, the participant-controlled cursor was invisible, and participants were told to ignore the movements of the red dot of the tracking task. Discussion The goal of Experiment 1a was to provide a baseline measure of processing efficiency before any secondary task load was added. Experiment 1b Experiment 1b replicated the procedure of Experiment 1a but with the addition of a central manual tracking task, to test whether concurrent task load reduced processing resilience.
Method Participants As we aimed to match sample size from Experiment 1a, we ran participants until we had data for 25 participants who met the inclusion criteria for detection error rates. Apparatus and stimuli The apparatus and stimuli were identical to those used in Experiment 1a, except that the cursor in the pursuit tracking task was made visible. Procedure In Experiment 1b, participants performed the peripheral target detection and manual tracking tasks concurrently. Analysis Analysis was identical to that of Experiment 1a.
Results Error rates As with Experiment 1a, participants with false alarm or miss rates greater than 0. Discussion Experiments 1a and 1b tested whether a manual tracking task impairs processing efficiency for redundant visual targets. Experiment 2 Experiments 1a and 1b failed to find a clear difference in processing efficiency for redundant visual targets between single- and dual-task conditions.
Apparatus and stimuli Apparatus and stimuli were the same as in Experiment 1. Procedure The tracking and discrimination tasks were performed in the same way as in Experiment 1. Analysis Analysis was as in Experiment 1, but was adapted to account for the within-participant manipulation of task load. Results Preliminary inspection found no effect of block order on any of the measures.
Discussion As in Experiments 1a and b, resilience was highly limited, but was not credibly smaller when participants performed a concurrent manual tracking task. Experiment 3 The previous experiments found that processing efficiency for redundant visual targets, as measured by resilience, was similar across single- and dual-task conditions. Method Participants We planned for a sample size of 30 participants who met the performance criteria for both the target detection and tracking tasks. Apparatus and stimuli The apparatus was the same as above.
Procedure and analysis Procedure and data analysis were identical to Experiment 2. Results Data from one participant were removed from analysis for high false alarm rates in both the single- 0. Discussion Experiment 3 assessed target processing efficiency within a forced serial process paradigm.
- Moral Philosophy After 9 11!
- Getting to Know the World Bank: A Guide for Young People;
- Product details.
- Introduction to Genetic Algorithms?
- Series Approximation Methods in Statistics: 88 (Lecture Notes in Statistics)!
General discussion The current studies examined redundant-target processing within a dual-task paradigm. Conclusions Within a peripheral redundant-target paradigm, data give no evidence for poorer target processing efficiency while under the load of a secondary tracking task. Acknowledgements Many thanks to Paul Douglas for assistance with programming, and to Kate Coden and Silas Ellery for assistance with data collection.
Not applicable The authors declare that they have no competing interests. Contributor Information Stephanie A. Impact of age, redundancy, and perceptual noise on visual search. Journal of Gerontology. Look here but ignore what you see: Effects of distractors at fixation. Species of redundancy in visual target detection. Effects of aging and distractors on detection of redundant visual targets and capacity: Do older adults integrate visual targets differently than younger adults?
PLoS One. Processing redundant information. Journal of Experimental Psychology. Interaction effects in parafoveal letter recognition. The eccentricity effect: Target eccentricity affects performance on conjunction searches. The contribution of covert attention to the set-size and eccentricity effects in visual search. Distribution inequalities for parallel models with unlimited capacity. Journal of Mathematical Psychology. Bimodal and trimodal multisensory enhancement: Effects of stimulus onset and intensity on reaction time.
Evaluating perceptual integration: Uniting response-time- and accuracy-based methodologies. Evaluation of cognitive processing in redundant audio-visual signals. Bello, M. Guarini, M.