1
How can human learning be measured and quantified using constancy and
contingency as parameters in a human foraging experiment?
Author: Nada Alberts
Student number: 11885637
Institute of Interdisciplinary Science, University of Amsterdam
Supervisors: Dr. Emiel van loon & Dr. Renske Hoondert
Credits: 18EC
Word Count: 3500
Date: 30th of May 2021
2
Abstract
Recognizing spatial patterns in foraging strategies in humans is a poorly studied academic sub-field of psychology and ecology. The aim of the study therefore is to understand and define the relationship between constancy and contingency in travel pathways and cognition in a human foraging
experiment. Multiple levels of predictability, search trials and spatial agreements will be distinguished to test not solely the relationship, but also the development of constancy and contingency over the scope of the experiment. The results indicate that there is no structured relationship between different runs, the data is therefore completely stochastic. It did indicate however, that in an environment with uniform distributed resources there is a significant improvement in search efficiency (measured by calculating the relative distance) between the first day of the experiment and the last day when performing tests with corrected data. The latter improvement is not observed within the afternoon group, whether corrected or not. Lastly there is a significant difference in retrieval of ribbons that were present every single day between the uniform and clustered groups. The largest complication of this study consisted out of the sample size, meaning that when reproducing this pioneering experiment with a larger sample size, values will significantly perform better at explaining possible relationships between time and search efficiency.
3
Contents
Abstract ... 2 Introduction ... 4 Methods ... 7 Study area ... 7 Experimental design... 7 Participants ... 8 Data Analysis ... 8 Results ... 11 Runs test ... 11 Linear regression ... 11 Group Clusters ... 14 Discussion... 15 Conclusion ... 17 Acknowledgements ... 18 References ... 19 Appendix ... 20A. Protocol experimental design ... 20
B. Protocol search team ... 22
C. Protocol hide team ... 25
D1. Metadata experiment Track Data ... 28
D2. Metadata experiment analysis data ... 28
4
Introduction
Animal psychology and ethology were studied independently for most of the 20th century (Shettleworth, 2001). Ethologists focused on behavior in nature, whereas animal psychologists focused predominantly on learned behavior in captivity (Shettleworth, 2001). The combination of both fields yields the possibility of the new field of cognitive ecology. Herein, cognition relating to all means of gathering, processing and acting upon information (Shettleworth, 2001) and ecology which is related to the interacting between organisms, including humans, and the physical environment. This field can merge resource availability with species presence due to animals recognizing patterns in resource availability. Research from Wolfe, Cain & Alaoui-soce (2018) stated that observers would collect more of a constant resource after previous retrieval due to a preference for the most common targets.
How efficient humans are at spatial learning differs between every individual in its unique search behaviors. There are, however different aspects that can impact spatial learning in humans, such as hydration levels (Grandjean & Grandjean, jaartal) , the distribution of the resources (Kalff, Hills & Wiener, 2010) and whether or not humans have a partner with whom they can discuss possible patterns in spatial presence of different resources (Tsay & Brady, 2010).
Research of Silva & Hare (2020) states that humans mostly learn through model-free learning, which is a strategy that strengthens or weakens associations between stimuli and actions, with or without reward. According to this strategy, participants are more likely to repeat a first stage action if there is a stimulus, such as finding a particular resource in a park. For humans foraging in space, Area Restricted Search is one of the most studied behavioral patterns in animal foraging (Kalff, Hills & Wiener, 2010). It states that whenever a resource has been found, the turning angle would have been larger, whereas it would have been smaller if no resource was found. Contributing to the theory that collecting a reward changes search tactic by transitioning into a more goal oriented strategy. The use of Area Restricted Search also depends on the distribution of the resource, since it is most effective in clustered environments, rather than uniform distributed environments (Kalff, Hills & Wiener, 2010).
5 Research by Kalff et al. (2010) consisted out of creating a virtual environment in which participants were requested to forage for different resources. The current study at the University of Amsterdam brought this foraging experiment into practice on a physical test site. This may provide insights in currently unknown search motives and patterns.
In order to state our research questions and hypothesis, several constructs needed to be clarified such as such as contingency, constancy, experimental runs and trials. Contingency and constancy aid the scientific community in describing patterns of fluctuation in travelled distance (Colwell, 1974). The definition of the two parameters is as follows: constancy is how uniformly an event occurs over a time period and contingency measures the repeatability of patterns over time periods (Colwell, 1974). The repeatability of patterns over time periods is crucial in investigating learning behaviors regarding foraging experiments. Measuring contingency is possible by defining runs as relative distances between two resource discoveries, meaning that the path from resource A to resource B accounts for one run. A run is the distance travelled between resource A and resource B and a search trial on one day consists out of 19 runs.
This study will investigate the connection between cognition and movement in humans, in particular their memory of patterns over the course of multiple days in different resource distributions and what effect that has on spatial learning. This will be studied by answering the research question: What is the
effect of constancy and contingency on travel pathways and cognition in a human foraging
experiment?
The sub questions used to answer the main research question are as follows: 1) How are constancy
and contingency measured in the study?, 2) How does the stochasticity of the experimental trials
impact pattern memorizing?, 3) How is this pattern memorizing observable in the retrieving of a
constant resource?
It is hypothesized that the stochasticity of experimental trials impact pattern memorizing by showing whether a clear relationship between time and experimental trials exist. In addition pattern
6 memorizing in the retrieving of a constant resource is more observable if there was a clear
relationship between time and experimental trials.
The data and knowledge on human foraging experiments is very scarce, even though it is more easily measured than for example the foraging experiments on primates. Knowing the relationship between movement and cognition in humans may aid the research of other primates, such as chimpanzees, since these can be harder to track and foraging behaviors may be similar between certain primates and humans. Knowing how quickly humans learn from patterns and how this impacts their decision making may also aid in programming different robots, targeted at retrieving or placing resources in a clustered or uniformly distributed environment. Therefore, The objective of the proposed research is to analyse animal thought processes regarding movement patterns by setting up a foraging experiment with two different resource distributions and connecting this to existing literature. Lastly, the data will be analyzed as well as visualized using statistical software.
7
Methods
Study area
The area chosen for this particular study was Flevopark, located in East-Amsterdam. The park its perimeter is approximately 1406 meters and provides a number of different functions, such as recreation, sports and religion burying (Gemeente Amsterdam, n.d.).
Experimental design
The experiment conducted from this study consisted out of different search trials over the course of five days (April 12th to April 16th of 2021). These trials were carried out twice a day, one in the morning and one in the afternoon. Both trials had a different distribution of the resource that needed to be retrieved. The morning resources were randomly distributed without the presence of significant clusters, whereas the afternoon group mostly encountered resources located in clusters. The resources that needed to be collected were ribbons, attached to different kind of trees that were selected prior to the search trials. See appendix A for further information.
Every day 40 ribbons were attached to a selection of 40 out of 200 trees by the hide team. The
strategy of placement was completely unknown to the participants. Of these 40 ribbons 25% belonged to either group cluster one, two, three or four which respectively mean ribboned every day, ribboned every other day, ribboned two days in a row and ribboned only once. In figure 2 the placement of all ribbons is illustrated as well as the placement of solely group cluster 1 ribbons
Figure 1: Overview of the experiment site with the red line indicating the travel routes of the participants with the blue triangle indicating the starting point
8 The participants, one in the morning and two in the afternoon had to collect 20 out of 40 ribbons every day for the duration of five consecutive days. The movements of the search parties were mapped by using both GPS trackers and testimonies of the participants. The testimonies of the participants consisted out of time of ribbon retrieving, time of ribbon sighting and the number located on the particular ribbon, so that every resource collection could be tied to one of the 200 chosen trees. For a more detailed explanation of the protocol for both search and hide teams, see appendices B & C.
Participants
All three participants fell within the age range of 20-25. The participants were chosen through social contacts of the researchers.
Data Analysis
After the experiment was finished, the data from the GPS tracker was exported into an excel file containing time, latitude, longitude, pace, speed and cumulative distance. For this study, the data that was most valuable consisted out of the cumulative distance travelled, coordinates and the time between two ribbon collections, of which the latter was noted on the participants search form.
To measure search efficiency, also referred to as relative distance in this study, the actual distance between two ribbon collections was divided by the shortest distance, measured directly. Search efficiency was measured with the following formula:
Figure 2: The placement of all trees (left) and the placement of trees belonging to group cluster 1 in the study site Flevopark (right)
9 𝑆𝑒𝑎𝑟𝑐ℎ 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 = 𝐴𝑐𝑡𝑢𝑎𝑙 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑡𝑟𝑎𝑣𝑒𝑙𝑙𝑒𝑑
𝑆ℎ𝑜𝑟𝑡𝑒𝑠𝑡 𝑝𝑜𝑠𝑠𝑖𝑏𝑙𝑒 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒
Actual distance travelled was calculated by subtracting the second cumulative distance travelled in meters from the first cumulative distance travelled, and so forth. The distance, directly, was measured by the formula (Voráčová, n.d.) :
𝐷𝑖𝑟𝑒𝑐𝑡 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 𝑅𝑎𝑑𝑖𝑢𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑎𝑟𝑡ℎ 𝑖𝑛 𝑚𝑒𝑡𝑒𝑟𝑠 ∗ arccos ((sin(𝑙𝑎𝑡𝑖𝑡𝑢𝑑𝑒1) ∗ sin(𝑙𝑎𝑡𝑖𝑡𝑢𝑑𝑒2)) + cos(𝑙𝑎𝑡𝑖𝑡𝑢𝑑𝑒1) ∗ cos(𝑙𝑎𝑡𝑖𝑡𝑢𝑑𝑒2) ∗ cos(𝑙𝑜𝑛𝑔𝑖𝑡𝑢𝑑𝑒2 – 𝑙𝑜𝑛𝑔𝑖𝑡𝑢𝑑𝑒1)
The next paragraphs will explain the statistical test preformed.
Firstly the relative distance (search efficiency), days and group cluster data were imported into R Studio. This was done by running the import syntax in R studio. The relative distance showed some abnormalities due to GPS inaccuracies and possible errors of participant testimonials, which lead to the decision to correct the acquired data. The latter was done by rounding all relative distances smaller than one to one, namely the highest possible search efficiency and to omit some extremely outliers.
Secondly, a runs test was conducted, located in the ‘snpar’ package in R studio version 4.5. The test determines whether the differences in relative distance are stochastic, measuring contingency in the search efficiency.
After acquiring the runs data, a simple linear regression analysis was done in order to compare the relative distance data over the course of five days. The resulting p value showed whether or not a pattern increased search efficiency was observed within the morning and afternoon group, respectively with an uniform and clustered resource distribution.
After the linear regression, a Wilcoxon test was conducted to compare the relative distances of Monday to Friday, separately for both the morning and the afternoon group. The resulting p-value indicated if the search efficiency of the participants increased on Friday compared to the first day of the experiment.
10 Lastly, the values of the group clusters were analyzed. A Wilcoxon test was conducted between the morning and afternoon group. The data used were the proportions of ribbons retrieved that belonged to group cluster one. This value demonstrated whether or not a significant difference could be observed within the morning group with an uniform distribution and the afternoon group with a clustered distribution. The results of this latter test were connected to the results of the multiple regression. The exact commands for the analysis are situated in appendix E.
11
Results
Runs test
In the conducted runs test, different values of relative distance were observed for the different days. Note that for all values, the maximum amount of runs was 19. The results described below in table 1 and two indicated no p-values smaller than 0.05, meaning all data is stochastically, there is no determinism.
Day P-values Runs
1 0.8032 11
2 0.4698 12
3 0.8032 11
4 0.2316 13
5 0.2316 13
Day P-values Runs
1 0.8225 10 2 0.4698 12 3 0.4853 9 4 0.2415 8 5 0.4698 12
Linear regression
For the linear regression test, a division was made between the morning group and the afternoon group and both linear regressions between days and relative distance were repeated with corrected data. For the corrected data the morning linear regression (see figure 3) resulted in a p value of 0.01113 and an R2 of 0.0673, meaning that 6.73% of the variance in relative distance is due to differences in days, corresponding to a small effect size according to Miles (2014). The linear relationship is significant in the morning since the p value < 0.05. For the afternoon group (see figure
Table 1: The results of the runs test of the morning group over the course of five days, no p-values were significant
12 4) the p value is 0.997 and the R2 number so low it was considered neglectable. The linear
relationship in the afternoon is therefore not significant since the p value > 0.05.
For the uncorrected data the morning linear regression (see figure 5) showed a p-value of 0.1735 and an R2 of 0.01983, meaning that only 1.9% of the variance in relative distance is explained by the difference in days. For the afternoon group the linear regression (see figure 6) showed a p-value of 0.882 and an R2 of 0.0002381 meaning only 0.02381% of the variation in relative distance is explained by the difference in days. Both p-values were > 0.05 meaning that the results of the linear regression with uncorrected data is not significant.
Figure 3: Linear regression analysis of the morning group (Uniform distribution & Corrected)
Figure 4: Linear regression analysis of the afternoon group (Clustered distribution & Corrected)
Figure 5: Linear regression analysis of the morning group (Uniform distribution & Uncorrected)
Figure 6: Linear regression analysis of the morning group (Uniform distribution & Uncorrected)
13 In order to investigate whether or not a significant difference was observed regarding relative distance between the first and the last day of the experiment, an non-parametric t-test, a Wilcoxon rank test was performed with solely the corrected data . There was a significant effect for the morning group (uniformly distributed resources) , p=0.01015 however there was no significant effect for the afternoon group (clustered distributed resources), p=0.2582.
14
Group Clusters
For this next analysis, the amount of ribbons collected from different group clusters were investigated. There were at all days 10 ribbons present in the field belonging to group cluster 1, which entails that these 10 ribbons were attached to 10 same trees for the duration of the study. In table 3, the proportion of retrieved ribbons from group cluster one can be observed.
Day Percentage of ribbons retrieved from group cluster 1 in the morning
Percentage of ribbons retrieved from group cluster 1 in the afternoon
1 60% 50%
2 70% 50%
3 70% 40%
4 80% 60%
5 80% 50%
Visual inspection of the values in table 1 showed that the morning group, which had a uniform distribution collected more resources tied to group cluster 1 than the afternoon group. A non-parametric test, namely the Wilcoxon test showed that indeed the morning group did significantly performed better than the afternoon group in locating trees of group cluster 1 (p=0.01409).
15
Discussion
The uncorrected results show no significant relationship between the day and the relative distance. This implies that the participants did not learn the pattern of chosen trees over the course of five days, as would be expected when looking at repeated exposure to patterns. This could be due to the fact that all participants were participating in Ramadan, meaning they don’t consume any drinks or foods from sunrise to sunset, which could lead to dehydration and therefore less cognitive abilities. Research from Grandjean & Grandjean (2007) found that mental performance, such as alertness and decision making decreases when dehydration levels reached values of 2% of the individuals body weight. Research from Lieberman (2007) already shows an significant decrease in reported concentration after 13 hours of dehydration, corresponding to a weight loss of 1% bodyweight in water.
Important to note is that the corrected dataset however did show a significant improvement of search efficiency when laid out against time. This data however is heavily modified, meaning any
conclusions derived from it should be critically handled.
In the results it is shown that even though the relationship between learning and search efficiency is not significant in the uncorrected data from the morning group, it’s p-value is still significantly lower than in the afternoon (p=0.1735 and p=0.882 in the uncorrected dataset). This could be due to the fact that learning is more optimal whenever resources are distributed uniformly instead of clustered. Another reason could be that the afternoon group experiences signs of dehydration due to longer fasting times than the morning group. The first seems rather unlikely since research has shown that clustered distributions are more beneficial to search efficiency (Kalffs, Hill & Wiener, 2010), leading to believe increased dehydration played a crucial role in the results.
Another aspect to note is that the afternoon group performed better at the first day (mean relative distance=1.321136) than the morning group performed at the last day (mean relative distance= 1.421105), meaning that the afternoon group possibly showed no clear pattern of improvement since they already optimally preformed since day one.
16 Research from Wolfe, Cain & Alaoui-Soce (2017) stated that some observers prefer to collect more easy accessible targets that were kept constant throughout the experiment. This is what was observed within the morning group, since they increasingly collected more of ribbons belonging to group cluster one. This inherently also states that the participants collect less of the ‘harder’ to obtain resources that differs within days.
The study conducted did contain some areas of improvement for further studies (using the
experimental design as laid out in appendices A & B), starting with the participants involved. Both the morning and afternoon group were supposed to consist out of two people, but due to a corona
suspicion in the morning group, one participants had to unfortunately withdraw from the research. The latter could have had significant effects on the measured search efficiency, since doing the research in duo’s could not only cause some distraction, it can also optimize ribbon collections due to discussing and dividing tasks and area to map. Research from Tsay & Brady (2010) even states that success is greater whenever preforming tasks in duo’s. Future research can increase this participant size in order to obtain more data to be able to potentially make statements about differences in search behaviors between males and females and/or possible positive interactions.
Furthermore, participants were instructed beforehand to not enter bushes in the research field due to possible interference with other ongoing studies, which could have significantly impacted ribbon collections. Verbal testimonials of the afternoon group stated that they had sometimes retrieved a ribbon by chance, but that they would have never went off path as far as was needed for some ribbon collections.
In conclusion, the results obtained from the human foraging experiment show only a pattern of spatial learning in the corrected dataset of the morning group. This is tied with the results of the collection of resources belonging to group cluster 1, meaning that an increase in search efficiency provides a small sense of prediction in the collection of constant resources.
17
Conclusion
Learning is not a stationary, but a developing part of the human brain and depends on multiple different factors. Including, but not limited to water intake, distribution of resources and team size. The results of the research show no determinism in the different runs of the experiment but a completely stochastic nature. Relative distance however did demonstrate significant improvements over time in the corrected dataset for solely the morning group. In addition, the morning group performed significantly better at retrieved resources belonging to group cluster one throughout the five days of the experiment. The afternoon group showed no significant values for everything stated above. The reason as to why these results were experienced range from dehydration to effects of different placement strategies and search party size. A lot is still unknown about foraging patterns in humans, however our research provides an insight in foraging patterns by fasting humans and a well formulated experimental design that can be repeated all over the globe in different populations in order to make cognitive ecology as prevalent in the scientific community as it deserves.
18
Acknowledgements
Conducting this study has been a huge challenge in covid-19 times. Physical contact with the supervisors and fellow students was very little, with the exception of the week of fieldwork, which has been a very pleasant experience. Because of this I would like to thank Emiel van Loon & Renske Hoondert for providing quality guidance in the forms of meetings and feedback in these hard times.
Secondly I would like to thank Precious Held for all her feedback and emotional support throughout the experiment and Liam Adam & Joris Oud for their much needed help during unforeseen
circumstances in the fieldwork experiments. The help of my dear friends Moonya Amro & Yassine Salamat and Tamara Stoof has been crucial since they all volunteered to spend a few hours in either the afternoon or morning to act as participants in order to help us with the research. Without them it would not have been possible.
Lastly, I would like to thank my family, partner and friends for supporting me through this thesis in corona times.
19
References
Colwell, R. K. (1974). Predictability, Constancy, and Contingency of Periodic Phenomena. Ecology,
55(5), 1148–1153. doi:10.2307/1940366
Feher da Silva, C., & Hare, T. A. (2020). Humans primarily use model-based inference in the two-stage task. Nature Human Behaviour. doi:10.1038/s41562-020-0905-y
Gemeente Amsterdam. (n.d.). Flevopark. Retrieved April 30, 2021, from https://www.amsterdam.nl/toerisme-vrije-tijd/parken/flevopark/
Grandjean, A. C., & Grandjean, N. R. (2007). Dehydration and Cognitive Performance. Journal of the
American College of Nutrition, 26(05), 549S-554S. doi:10.1080/07315724.2007.10719657
Kalff, C., Hills, T., & Wiener, J. M. (2010). Human foraging behavior: A virtual reality investigation on area restricted search in humans. Proceedings of the Annual Meeting of the Cognitive Science
Society, 32. Retrieved from https://escholarship.org/uc/item/74m6d4qr
Lieberman, H. R. (2007). Hydration and Cognition: A Critical Review and Recommendations for Future Research. Journal of the American College of Nutrition, 26(05), 555S-561S.
doi:10.1080/07315724.2007.10719658
Miles, J. (2014). R Squared, Adjusted R Squared. Wiley StatsRef: Statistics Reference
Online. doi:10.1002/9781118445112.stat06627
Shettleworth, S. J. (2001). Animal cognition and animal behaviour. Animal Behaviour, 61(2), 277– 286. doi:10.1006/anbe.2000.1606
Tsay, M., & Brady, M. (2010). A case study of cooperative learning and communication pedagogy: Does working in teams make a difference? Journal of the Scholarship of Teaching and Learning,
10(2), 78–89. Retrieved from https://files.eric.ed.gov/fulltext/EJ890724.pdf
Voráčová, S. (2017, November 14). Orthodromic distance. Retrieved May 1, 2021, from https://www.geogebra.org/m/qjdD3fak
Wolfe, J. M., Cain, M. S., & Alaoui-Soce, A. (2017). Hybrid value foraging: How the value of targets shapes human foraging behavior. Attention, Perception, & Psychophysics, 80(3), 609–621.
20
Appendix
A. Protocol experimental design
version 2 – 10 April 2021 Emiel van Loon
Introduction
In this experiment, which takes place in the spring of 2021, we will investigate the relationship between movement and cognition (hence the choice for ‘MoCo2021’ as an acronym).
The overarching questions behind this experiment are coming from animal ecology:
a. Will spatio-temporal regularity as well as co-location in resource availability influence search strategy and search efficiency?
b. How (well) are clues about resource location (such as regularity and co-location) learned or memoirzed?
c. Can this learning be observed in features (derived variables) of movement tracks? These questions are currently under discussion and actively being research. It is also an important topic of research at the Theoretical and Computational Ecology department at the Institute for Biodiversity and Ecosystem Dynamics at UvA. And this experiment forms part of this research effort. The MoCo2021 experiment has several goals:
1. piloting and design
This experiment is the first of its kind, so both the experimental treatments and protocol need to be tested and will be improved in due course. Of course, the experiment will be designed and
implemented as best as possible.
2. provide suitable material to answer substantive questions for BSc thesis research in 2021
We aim to investigate whether:
• constancy and contingency of relative distance traveled are related to levels of cognition (understanding the best search strategy or memorizing the location of
resources), and/or search efficiency
• search order or efficiency are related to the constancy and contingency of resource availability
3. provide suitable material for developing methodologies at a later stage,
e.g. to analyse additional aspects of movement (e.g. track shape, path formation etc.) in relation to levels of cognition, search efficiency and other variables of relevance.
21 Over the coming years we aim to repeat and enhance the experiment. Subsequently we plan
to extend the experiment (e.g. with different treatments). The experiments will be conducted in the context of BSc projects, MSc projects, assignments in specific courses and possibly also a targeted research project.
Page Break
Overall rationale and expected experimental outcomes
The overall logic of the experiment is as follows.
a. Several easily recognizable ribbons are available in a search arena. The ribbons are placed according to an underlying logic but also under influence of randomness. b. The experimental subjects (a team consisting of two persons, henceforth called ‘search
team’) will search and retrieve the ribbons.
c. The logic by which the ribbons are placed is hidden for the search team.
d. Information about the movement tracks by the search team during the experiment is automatically recorded (by GPS loggers) and additional observations are recorded as well (notes by search-team).
e. The search trials are repeated (using the same underlying logic) several times by the same search team, which will allow to evaluate of learning or change in search efficiency. This experiment has two treatments: 1) each set of rules and 2) the number of repeated that has been done by the search team.
Potentially, additional treatments can be added by placing subsets of ribbons with a specific logic.
The expected experimental outcome is that:
a. There is an overall effect of resource availability pattern, with as general rule that spatial clustering of resources leads to a more efficient search efficiency than a more uniformly spread resource distribution.
b. Search efficiency (as well as several other movement features) increases with experience (repeated visits).
c. The way by which efficiency (and other features) change, depends on the complexity of the resource availability pattern (both in space and time).
Outcomes a, b and c relate respectively to treatments 1, 2 and the interaction among treatments 1
and 2.
The MOCO2021 experiment is a first version of this general experiment. It will be conducted by applying 3 treatments:
• Spatial arrangement, with two levels: one with a regular distribution of ribbons and one
with a clustered distribution (three clusters).
22
• Predictability, with four levels: of all locations with a ribbon, ¼th gets a ribbon at every repetition, ¼th in a strictly alternating way, ¼th of in an irregular way but more than once and ¼th only once.
It will take place in Flevopark during 5 days in the period April 6 – April 10. At every day, one trial takes place during the morning and one during the afternoon.
Hiding of ribbons will be done by two ‘hide teams’: one team per spatial arrangement treatment. In the mornings the hide team for the regular distribution will place ribbons, and in the afternoon the hide team for the clustered distribution will place ribbons.
Similar to the hiding, the searching will be done by two search teams: one team per spatial arrangement treatment and only mornings or afternoons.
Ribbons will be placed in relatively large trees between 1 m and 2 m above the ground. The trees to which ribbons are attached have been selected prior to field visits.
B. Protocol search team
Prior information for participating search team
We inform persons who might be interested /suitable to take part in the experiment as a ‘searcher’ with the following information, when asking them to participate:
• you are asked to search together with 1 fellow-student to search for 20 ribbons, which are hidden in Flevopark on 5 consecutive days
• the experiment takes place from 12 to 16 April during the morning/afternoon
• the search will take 60 minutes at maximum, but will probably be finished within 60 minutes
• the experiment is not dangerous and participating is just fun; when it is finished you will of course be informed about the results
• the experiment is serious (i.e. input for real research) where we investigate search behaviour and possible change of search efficiency over time
•
Information for participating search team, after they agree to join
• Approximately one week before the experiment starts, Emiel will send an email to each search team (2 persons) with the following info:
o the information above
o introduction of the two members in the search team (if they don’t know each other)
o the dates & times for the searches (5 consecutive days, either morning or afternoon)
o place and time to meet on april 6 (first experimental trial)
Instruction for search team right at the start of the experiment (before the 1st trial)
• 15 minutes before the start time of the 1st trial the search team meets at the reception of Science Park 904 with Emiel
• The team gets paper forms with clipboard & pencil to write down observations.
• One of the participants of the search team also gets a rucksack with GPS trackers mounted on it.
• The clock of the search team (on mobile phone or watch) is checked (so that it is accurate).
23
• From Science Park 904, we walk in the direction of Flevopark. The route is indicated in Figure 2. It is explained to the search team that they have to search for ribbons, an example of such a ribbon is shown.
• While walking in the direction it is explained that:
o these ribbons are attached to trees at 1 m to 2 m height
o three of these ribbons are attached to trees along the walking route to Flevopark
o they should try to spot them
o once they spot a ribbon they should write this down at field form 1 with an accurate time
o they should then go to that ribbon, take it from the tree and write down the ribbon-ID with time of collection – if the original observation turns out to be wrong (there appeared to be no ribbon after all) this is also noted at the form
o if they make any other observation or decision with respect to search, they should note it (with time) at field form 2
• If the search team misses one or more of the three ribbons, they are pointed at the omission and this ribbon is also collected. So that eventually three trial observations + entries on the form are made.
• The search team stops at the entrance of Flevopark at the valentijnskade next to Jeugdland (Figure 2, blue triangle). This is the start of the trial.
At this point, right before the start of the trial the essentials of the search protocol are explained (these are also listed below and handed over to the search team on a printed sheet):
• All ribbons are visible from the paths and lawns, and that they may move over both paths and the lawns to move around efficiently.
• They may not enter or cross the ‘bushes.’ Only to retrieve a ribbon they might need to enter a bush a little bit at the border, but not further than 2 m.
• It is not about being as quick as possible (no running, just keep walking – stopping in between, e.g. to make a note or to look around is no problem) but it is the aim to walk the shortest possible route.
• the search team should continue until they have retrieved 20 ribbons or the 60 minutes have passed.
Instruction for search team to recover the ribbons
The search team should search for ribbons in Flevopark within the search area (see red line in Figure 2).
The start of the experiment is at the location indicated in Figure 2 (blue triangle) and the start time is noted at field form 1. The experiment ends after 60 minutes.
The goal of the search team is to recover all 20 ribbons, while covering a minimum distance. Only move around at walking speed (no running).
The search team should stay on the trails within the search area (see Figure 2) and may cross the lawns, but not move through the bushes. They may only enter the bushes/tree areas at the borders to recover a ribbon.
After the search team has finished the search (20 ribbons recovered or 60 minutes have passed). They send an sms message to Emiel and return to the starting point. There they hand-over the ribbons they have found, the field forms and the backpack to Emiel. Emiel prepares field form 2 for the hide team. This form lists which ribbons are still in the field, and at which location IDs these can be found.
24
Figure 2. Map of Science Park and Flevopark, with route from Science Park 904 (start of experiment
before first trial; purple line), delimitation of the search area (red line) and starting point of experiment (blue triangle).
25
C. Protocol hide team
The experimental area is in the southern area of Flevopark (indicated by the red line in the map of Figure
1)
Figure 1. Map of Flevopark, with delimiteation of the search area (red line) and starting point of
experiment (blue triangle).
There are two experimental trials per day: a ‘morning’ and and ‘afternoon’ trial. The time schedule for each of these is given in Tables 1 and 2.
26
Table 1. Time schedule for Morning trial
Time: Activity: Who:
8:30 Hide team gets field form 1 from Emiel and starts the placement of ribbons.
When finished (not later than 10:15), hide team sends a message to Emiel.
Liam, Joris & Emiel
10:30 Search team gets field forms 1 and 2, backpack with GPS & explanation from Emiel. Thereafter the search team starts the search.
Search team 1 & Emiel
12:00 (at latest)
Hide team gets field form 2 from Emiel and starts retrieval of ribbons that were left in the field.
Liam, Joris & Emiel
13:00 Ribbons retrieved by hide team and handed over to Emiel. The material is ready for afternoon-trial.
Table 2. Time schedule for Afternoon trial.
Time: Activity: Who:
13:30 Hide team gets field form 1 from Emiel and starts the placement of ribbons.
When finished (not later than 15:15), hide team sends a message to Emiel.
Nada, Precious & Emiel
15:30 Search team gets field forms 1 and 2, backpack with GPS & explanation from Emiel. Thereafter the search team starts the search.
Search team 2 & Emiel
17:00 (at latest)
Hide team gets field form 2 from Emiel and starts retrieval of ribbons that were left in the field.
Nada, Precious & Emiel
18:00 Ribbons retrieved by hide team and handed over to Emiel. The material is ready for the next day.
Instruction for hide team to place the ribbons
At the time indicated in Tables 1 and 2 (8:30 and 13:30 for the morning and afternoon trials respectively), the hide teams meet with Emiel at the starting point of the experiment (blue triangle in Figure 1).
The hide team get 40 ribbons (ropes with a numbered steel ring attached to it) and receives information about 40 coordinates where to place the ribbons. This information consists of:
27
• a printed map of flevopark with the IDs indicated
The hide team and Emiel go jointly to the first tree from the list to place the ribbon. After placing it, the number of the ribbon that was placed at the respective location is noted on field form 1 (the ribbons are be placed in random order). Next a photograph is made of the tree with the ribbon attached at a distance such that both the ribbon can be seen and the tree can be identified in its surroundings.
Instruction for hide team to recover the ribbons (right after the search team finished)
After the search team has finished their search. They hand in the ribbons they have found, the field forms and the backpack to Emiel. Emiel prepares field form 2 for the hide team. This form lists which ribbons are still in the field, and at which location IDs these can be found.
The Hide team gets this from to recover the remaining ribbons in Flevopark. When ribbons are not found back, this is noted on the field form .
When ready, the ribbons are handed over to Emiel. If some ribbons were not recovered, these are replaced by new ones.
28
D1. Metadata experiment Track Data
Variable Description
Time Time noted down in GMT times
Latitude Latitude noted down as in the WGS 84 coordinate system
Longitude Longitude noted down as in the WGS 84 coordinate system
Distance Cumulative distance travelled by the participants in the experiment
Found A value of 1-20, meaning in which sequence ribbons have been collected
The exact values of the meta data, in order words the track data can be obtained by following this link: https://github.com/NadaAlberts/Thesis-human-foraging-experiment
D2. Metadata experiment analysis data
Latitude Determined by the GPS tracker in the WGS 84 coordinate system
Longitude Determined by the GPS tracker in the WGS 84 coordinate system
Group cluster Ranging from group cluster 1-4, which can be observed in appendix A
Found The sequence of collected ribbons, ranging form 1-20
Shortest_distance Measured by measuring the distance directly between point A and point B (formula explained in the method section)
Actual_distance Distance between point A and point B as measured by the GPS trackers on the participants
Relative_distance Actual distance/shortest distance
The data can be retrieved following the link: https://github.com/NadaAlberts/Thesis-human-foraging-experiment
29
E. R-Script
The R-Script can be obtained through this link:
