This study examined the effects of verbal interference on web content localization and
recall. Data from 70 participants was collected through a custom web application that was built specifically for this experiment. The application presented participants with a randomly generated web page and then asked them to recall the location of the content under several different conditions. Their responses were then analyzed using SPSS with special attention given to the accuracy of their localizations. It was hypothesized that participants would have lower accuracy when recalling content location if asked to describe the location verbally, and would have higher accuracy if asked to describe the location graphically. Our findings did not support this hypothesis, but it did support prior research suggesting that reading spatial texts can interfere with the recall of spatial information.
This study examined the effects of verbal interference on web content localization and
recall. Data from 70 participants was collected through a custom web application that was built specifically for this experiment. The application presented participants with a randomly generated web page and then asked them to recall the location of the content under several different conditions. Their responses were then analyzed using SPSS with special attention given to the accuracy of their localizations. It was hypothesized that participants would have lower accuracy when recalling content location if asked to describe the location verbally, and would have higher accuracy if asked to describe the location graphically. Our findings did not support this hypothesis, but it did support prior research suggesting that reading spatial texts can interfere with the recall of spatial information.
This study examined the effects of verbal interference on web content localization and
recall. Data from 70 participants was collected through a custom web application that was built specifically for this experiment. The application presented participants with a randomly generated web page and then asked them to recall the location of the content under several different conditions. Their responses were then analyzed using SPSS with special attention given to the accuracy of their localizations. It was hypothesized that participants would have lower accuracy when recalling content location if asked to describe the location verbally, and would have higher accuracy if asked to describe the location graphically. Our findings did not support this hypothesis, but it did support prior research suggesting that reading spatial texts can interfere with the recall of spatial information.
Micah H. Fenner Carnegie Mellon University VERBAL INTERFERENCE 2 Abstract This study examined the effects of verbal interference on web content localization and recall. Data from 70 participants was collected through a custom web application that was built specifcally for this experiment. The application presented participants with a randomly generated web page and then asked them to recall the location of the content under several different conditions. Their responses were then analyzed using SPSS with special attention given to the accuracy of their localizations. It was hypothesized that participants would have lower accuracy when recalling content location if asked to describe the location verbally, and would have higher accuracy if asked to describe the location graphically. Our fndings did not support this hypothesis, but it did support prior research suggesting that reading spatial texts can interfere with the recall of spatial information. VERBAL INTERFERENCE 3 Verbal Interference in Content Localization Internal representations in working memory have long been a topic of study within the feld of cognitive psychology, as it is safe to assume that much of the information that we process every day is visual. Looking at the the amount of visual content that is consumed on the internet confrms this assertion. Much of the internet relies on our visual processing abilities to encode and maintain representations of the content that we are consuming. Unfortunately, the impact on psychology on the feld of web design is still budding, and thus many of these representations are either presented in a way that is not conducive to proper encoding and recall, or the recall environment interferes with the processing of the representation by working memory. The recall environment is of particular interest because it relies on accurate decoding and interpretation of spatial relationships. If web content recall is inaccurate, it is due to interference with the decoding of the visual and verbal representations of content location within the brain. Thus, this research focuses on whether the particular manner of recall has an impact on accuracy. Specifcally, we were interested in whether asking participants to describe web content location verbally or graphically would yield better performance in a content localization task. Research in this area shows that not only is mental imagery crucial to maintaining internal representations in working memory, but that the propositional model of internal representations is inadequate at explaining many common working memory tasks (Kosslyn & Pomerantz, 1977). The propositional model states that knowledge representations are related using propositions: structures expressing relations between entities (e.g. your [hand] is at the end of your [arm]). The fact that this model of information storage does not adequately address behavior in many working memory tasks indicates that the primary alternative, mental images, may more accurately describe the mental operations required for working memory tasks. Later work by Postle, DEsposito, and Corkin (2005) shows that mnemonic representations of featural characteristics of objects in memory are commonly verbal while the representations of a objects location are not. They discerned this by using interference tasks in order to attempt to ascertain what areas of the brain were being used during a task and how VERBAL INTERFERENCE 4 verbal and spatial interference would affect performance on the task. This indicates that recalling the location of an object verbally could be more diffcult than recalling it graphically as verbal recall would require additional encoding. In fact, it appears that not only are these attributes stored differently in working memory, but certain types of interference can disrupt the process of recalling visual representations from working memory. In one study, participants were presented with two types of texts: spatial texts and visual texts. Spatial texts contained information about the locations of object features and visual texts contained information about visual features of the object (colors, form, etc). Learners who were presented with spatial texts and an image were consistently worse at recalling that information than learners presented with visual texts and an image (Schler, Scheiter, & Gerjets, 2012). This indicates that the spatial texts are interfering with the working memory process for image recall. Additionally, it may be the case that these two types of representations are very much isolated from each other and do not interact. Research by Munnich, Landau, Dosher (2001) indicates that though spatial language and spatial memory have similar properties and foundations, spatial properties are not preserved across languages or memory tasks. In light of the research mentioned previously, this would indicate that the results from the interference experiments above are certainly explainable by events occurring during the recall process. What the above research fails to address is if recall of spatial information in working memory is performed better in a verbal or graphical context. To this point, it is clear that the decoding of working memory is affected by context and that this is particularly relevant when dealing with spatial relationships. Furthermore, it is generally understood that spatial and verbal processing occurs in different areas of the brain. Thus, this experiment stands to demonstrate that if spatial relations are encoded into working memory, they are most effectively retrieved and decoded graphically rather than verbally. As such, we hypothesized that participants would perform better in a visuospatial recall condition, as opposed to a verbal one because the representations and spatial relationships being VERBAL INTERFERENCE 5 accessed were encoded spatially and those relationships cannot be adequately described verbally. Additionally, applied this principle to the feld of web design as the consumption of web content is hugely important to information intake. To do this, presented spatial information in the context of a website and asked participants to recall those relationships either verbally or graphically. Overall, we expected to see participants perform better when asked to recall this information graphically. Additionally, we expected to see academic major and gender have some sort of interaction with the treatment condition, as both of those covariates have been shown in the past to affect spatial cognition. Methods Participants We observed 70 participants of equal gender distribution between the ages of 17 and 24. Our primary group of participants were students at Carnegie Mellon University, with around 20 of those students being drawn from the course 85-310 Research Methods in Cognitive Psychology. Design In this case, the independent variable was the type of quiz that the participant was presented with after being exposed to the stimulus. This variable was between-subjects Participants were randomly assigned to receive a quiz requiring them to drag-and-drop content to the location it was on the page, or they were presented with a series of verbal descriptions of content locations and will be asked to match those with the content that was there. We measured the accuracy of these responses and compared them between the two groups in order to identify any signifcant differences. Our primary covariates were gender and academic major, as both of these have been shown in the past to impact verbal and visual processing. Materials The experiment consisted of a Python web app running on a Heroku web server. The site had three pages: a pre-test survey to collect demographic information, a randomly generated webpage consisting of several text areas and two images, and the randomly selected post-test quiz. Upon loading the website and signing the consent form, participants were presented with a VERBAL INTERFERENCE 6 short survey asking for what college or university they attend, their age, gender, academic major, and hours spent online per week. Additionally, at the top of the page were the following instructions: After completing this form, you will have 30 seconds to examine a randomly generated webpage. Please try and remember as much as possible about it's design. After the survey submission, the web app would randomly select a page layout from one of fve possible confgurations. Each confguration had four possible text locations and two possible image locations. The images were always placed side-by-side. The images and texts to fll these spots were randomly chosen from a set of 10 paragraph-long texts and 13 color images. The pool of possible stimuli was selected by the experimenters in advance. After being selected, the stimuli were inserted into the page and the page was rendered in the participants browser. After 30 seconds of viewing time, the participants were redirected to the post-test quiz page. Based on the group that was randomly assigned to the participant upon submitting the pre- test survey, they were presented with one of two types of quizzes. In the graphical quiz condition, they were presented with a set of randomly ordered images and text next to a smaller version of the blank page layout and asked to Drag and drop the content to the correct section. In the verbal condition, participants were presented with all of texts and images in a random order. Under each text/image was the phrase, Select a location followed by a drop down menu allowing them to choose from First on the page, Second on the page, etc. In the case of images, they were also allowed to select whether the image appeared on the left or on the right. Once they submitted their responses, they were shown their score and given a brief overview of the purpose of the study. Procedure Participants were approached by the co-experimenters either in person or over the internet and given the link to the experiment webpage. Upon visiting the page, they were assigned a subject ID in the database of results, randomly assigned to a treatment group, and asked to answer demographics questions including their gender, age, school, major, and hours spent on the internet per week. After answering all of these questions, all participants were VERBAL INTERFERENCE 7 presented with a webpage populated with randomly generated content for 30 seconds. The website then redirected them to a quiz based on their treatment group. The submitted responses which were then automatically entered into the database of results. Coding The primary dependent variable percentage of questions answered correctly on the post-quiz was not coded. However, we also considered gender, major as covariates and these were be coded numerically. We also considered age and hours per week spent online as covariates but they did not yield any signifcant results. Results Data Manipulations We recoded the gender and major covariates numerically to aid our analysis. We also removed all data points that appeared to be fraudulent, such as one participant who claimed to spend 12,500 hours on the internet per week. Lastly, we summed the quiz response values and divided by the total number of questions (6) to determine each participants overall accuracy. Data Exploration Before analyzing our data, we explored our pool of 70 valid participants. The majority of them were humanities majors (N = 25), followed closely by engineering (N = 18) and sciences (N = 17). Arts (N = 3) and other majors (N = 7) comprised the minority of our participants. We then calculated descriptive statistics of the numerical covariates and raw accuracies for each question. See Table 1 below for these descriptive statistics: VERBAL INTERFERENCE 8
Table 1: Descriptive statistics for the population including individual question accuracies
Analyses Given the question of how verbal interference would affect recall of a webpage layout, our hypothesis was that participants who were asked questions about the layout of the webpage in a verbal format would recall the layout less accurately than participants who were asked to recall the same webpage but in a graphical format. The graphical quiz group had a mean accuracy of 0.6852 (SD = 0.3132) and the verbal quiz group had a mean accuracy of 0.6667 (SD = 0.2752). These means are illustrated in Figure 1 below: N Mean Standard Dev. Internet Experience (hours) 69 32.23 20.057 Question 1 70 0.8 0.403 Question 2 70 0.61 0.490 Question 3 70 0.51 0.503 Question 4 70 0.61 0.490 Question 5 70 0.69 0.468 Image Area 70 0.61 0.490 Image Side 70 0.83 0.380 VERBAL INTERFERENCE 9 Figure 1: Comparison of means between the two treatment groups
We used one way analysis of variance to compare the mean accuracy between the two treatment groups. The ANOVA revealed no signifcant main effect between the two treatment groups, F(1, 70) = 0.069, p = 0.794, indicating that our initial hypothesis does not hold. This means that quiz format had no signifcant effect on recall of the web page layout. We conducted further analysis through several analysis of covariance tests to see if any of the demographic covariates (gender, major, internet experience) caused any signifcant differences in accuracy. A summary of the results of this ANCOVA test are shown in Table 2: Table 2: Demographic ANCOVA results summary with no signifcant results As shown above, none of the covariates in this experiment had a signifcant on accuracy. df F N p Gender 1 0.039 70 0.844 Major 1 1.515 70 0.223 Internet Experience 1 2.375 69 0.128 VERBAL INTERFERENCE 10 Finally, we performed a repeated measures ANOVA test using quiz format as the between subjects variable and accuracy on questions about the images on the page and accuracy on questions about the texts on the page as within subjects variables. Mean accuracy on image questions was 0.7124 (SD = 0.3472) while mean accuracy on text questions was 0.6457 (SD = 0.31973). The ANOVA results confrmed this difference as signifcant (F(1, 70) = 5.087, p = 0.027). This means that participants were more accurate in recalling image positioning on the web page than they were in recalling text positioning on the same page. Discussion The purpose of this study was to investigate the effect of recall conditions on accuracy in reconstructing the layout of a randomly generated webpage. Our hypothesis was that recalling the layout with a verbal description would yield lower accuracy than recalling the layout graphically. Though no signifcant difference was found between the accuracies on the two different types of quizzes, we did fnd a signifcant difference in accuracy between questions about images and questions about text. In this case, participants were more accurate in recalling the position of images on a webpage than they were in recalling the position of text. Because the questions that participants in the verbal condition had were spatial in nature, these results confrm previous research by Schler, et al. (2012) indicating that reading spatial texts can interfere with the recall of spatial information. Our results were not as clearly signifcant as theirs, however this is something that could be remedied in future iterations of the experiment by using a larger sample size. Our study had some limitations that inhibited its effectiveness. Sample size is one limitation that could have been a determining factor in limiting the anticipated effect of the experiment. Our population of participants (N = 70), while not insignifcant, was mostly limited to Carnegie Mellon students who are well-versed in technology and likely have above-average mental abilities. Thus, we see a ceiling effect in our data where many of the participants in both conditions had perfect scores. One solution to this problem given the limited audience of the VERBAL INTERFERENCE 11 study could be to perform it again with more texts and images on each page or with a more complex layout. In conclusion, we have shown that the spatial properties of images are recalled correctly signifcantly more often than the spatial properties of text in the context of web pages. These results are in line with prior research (Schler, et al., 2012) and make sense given that the visual system is geared towards recognizing imagery rather than text. This alone carries implications for the feld of web and interface design, as it implies that effective use of imagery could yield more memorable and usable interfaces. If our initial hypothesis holds true in the future, then interfaces and interactions relying on verbal descriptions of spatial relationships would fall to the wayside in favor of more user- friendly interfaces that harness the innate spatial processing power of the brain. For example, the use of a graphical timeline as a method of displaying events in a certain order is likely to be recalled much more accurately than a standard list with dates before each item. In some ways, this is an abstraction of spatial and verbal descriptions into the temporal realm, but when it comes to web interfaces, these types of elements often overlap. In all, there is much more work to be done in this area. Spatial perception has yet to be fully understood, but such understanding will undoubtedly lead to enormous leaps of development in other realms such as Human-Computer Interaction that rely heavily on the processes that underly our understanding of the world around us. VERBAL INTERFERENCE 12 References Griffn, I. C., & Nobre, A. C. (2003). Orienting attention to locations in internal representations. Journal of Cognitive Neuroscience, 15(8), 1176-1194.
Kosslyn, S. M., & Pomerantz, J. R. (1977). Imagery, propositions, and the form of internal representations. Cognitive Psychology, 9, 52-76.
Munnich, E., Landau, B., & Dosher, B. A. (2001). Spatial language and spatial representation: a cross-linguistic comparison. Cognition, 81, 171-207.
Postle, B. R., D'Esposito, M., & Corkin, S. (2005, March). Effects of verbal and nonverbal interference on spatial and object visual working memory. Memory & Cognition, 22(2), 203-212.
Schler, A., Scheiter, K., & Gerjets, P. (2012). Verbal descriptions of spatial information can interfere with picture processing. Memory, 20(7), 682-699. VERBAL INTERFERENCE 13 Appendix A
Figure 1
Pre-experiment survey form VERBAL INTERFERENCE 14 Figure 2
Sample web page layout VERBAL INTERFERENCE 15
Figure 3
Post-stimulus quiz: verbal format Note: Some text may be cut off due to the screenshot
VERBAL INTERFERENCE 16 Figure 4
Post-stimulus quiz: graphical format Note: Some felds and images are cut out due to size constraints VERBAL INTERFERENCE 17 Appendix B
Possible texts for experiment
Text 1
The western coastal area of Ecuador borders the Pacifc Ocean to the west, encompasses a broad coastal plain, and then rises to the foothills of the Andes Mountains to the east. It is estimated that 98% of the native forest of coastal Ecuador has been eliminated in favor of cattle ranching and other agricultural production, including banana, cacao and coffee plantations. The forest fragments that still survive are primarily found along the coastal mountain ranges of Mache- Chindul, Jama-Coaque, and Chongon-Colonche, and include tropical dry forest, tropical wet forest, tropical moist evergreen forest, premontane cloud forest, and mangrove forest.
Text 2
The club was founded in 1948 by a group of students (the "Estudiantes") of a public preparatory school (the Instituto Ramiro de Maeztu, IRM) in Madrid. By the time when the frst Spanish-wide season-long championship was organized in 1955, by the Spanish Basketball Association (FEB), it was one of the six clubs participating in that tournament, as the second best team from the Province of Madrid (the frst being Real Madrid. Since then, it has always participated in the premier Spanish basketball league. It is one of only three of such clubs, together with Real Madrid and Joventut. It is also one of two only Spanish basketball clubs with teams both at the top male and female Spanish championships.
Text 3
Prime Minister Harper said he would move forward with his top fve priorities from the campaign. At least four of these would require legislative action: the passage of a Federal Accountability Act in response to the sponsorship scandal; setting longer mandatory sentences; lowering the Goods & Services Tax to 6% (and eventually to 5%); giving $1,200 for parents per child under the age of 6; and negotiating with the provinces to shorten wait-times (this priority was replaced, post- election, with combating crime by creating more police offcers). The child allowance and frst GST were in place by July 1, 2006. On December 6, 2006, another issue many expected to arise in the frst session of parliament did, in fact, come to the fore, when the government introduced a motion calling "on the government to introduce legislation to restore the traditional defnition of marriage without affecting civil unions and while respecting existing same-sex marriages."
Text 4
Upon his return to France, he joined CNRS (19491955). During all these years, he pursued research on the relationship between Egyptians and pre-Hellenes, providing some frm conclusions on the relationship between these two great civilizations and the history of the ancient Aegean world.
VERBAL INTERFERENCE 18 Text 5
Egyptology's modern history begins with the invasion of Egypt by Napoleon Bonaparte. The subsequent publication of Description de l'Egypte between 1809 and 1829 made numerous ancient Egyptian source materials available to Europeans for the frst time. Jean Francois Champollion, Thomas Young and Ippolito Rosellini were some of the frst Egyptologists of wide acclaim. The German Karl Richard Lepsius was an early participant in the investigations of Egypt; mapping, excavating, and recording several sites. Champollion announced his general decipherment of the system of Egyptian hieroglyphics for the frst time, employing the Rosetta Stone as his primary aid. The Stone's decipherment was a very important development of Egyptology. With subsequently ever-increasing knowledge of Egyptian writing and language, the study of Ancient Egyptian civilization was able to proceed with greater academic rigour and with all the added impetus that comprehension of the written sources was able to engender. Egyptology became more professional via work of William Matthew Flinders Petrie, among others. Petrie introduced techniques of feld preservation, recording, and excavating. Howard Carter's expedition brought much acclaim to the feld of Egyptology.
Text 6
Reviews were overwhelmingly negative, with flm critic David Nusair calling it "An utterly worthless piece of work" and giving it zero out of four stars. It received a score of 3.1 out of 10 on IMDb. Evangelical publications were much kinder. Christian Ted Baehr of MOVIEGUIDE described it as "An entertaining movie and a useful tool for leading people to consider the claims of the Gospel." Christian author John Hagee said of the flm: "The power of the gospel to transform the lives of the characters is shown with deeply moving reality. So real is the presentation of the plan of salvation in this movie, the viewer will be left without excuse." Evangelist Jack Van Impe was quoted as saying: "Both Rexella and I believe this to be the greatest religious release we have ever watched.
Text 7
After Ya'qob, Ahadabui. He was given this name because of his striking similarity to his father. Shortly before he died Ya'qob instructed two of his disciples, Ahadabui and Qamisho', to go to Antioch, so that the patriarch of Antioch could consecrate whichever of them he chose and send him back. They did so, and when they both arrived in Antioch they were lodged in the house of a certain believer.
Text 8
Since May 2006, more than 10,000 people in the Porong District have been displaced by the hot mud fowing from a natural gas well being drilled by Lapindo Brantas, an oil well company that is part of a conglomerate owned by Coordinating Minister for the People's Welfare Aburizal Bakrie. Gas and hot mud started spewing from the well on May 28, when the drill penetrated a layer of liquid sediment. Attempts to pump concrete down the well did not stop the fow. While some scientists have speculated that the earthquakethat struck Yogyakarta on May 27, the day before the well erupted, may have cracked the ground, creating potential pathways for the mud VERBAL INTERFERENCE 19 to reach the surface, others have suggested that the drilling procedure was faulty by not using a casing.
Text 9
Automatic transmission fuids have many performance-enhancing chemicals added to the fuid to meet the demands of each transmission. Some ATF specifcations are open to competing brands, such as the common DEXRON specifcation, where different manufacturers use different chemicals to meet the same performance specifcation. These products are sold under license from the OEM responsible for establishing the specifcation. Some vehicle manufacturers will require "genuine" or Original Equipment Manufacturer (OEM) ATF.
Text 10
Derived from the Persian khwaja, a term of honour, the word Khoja referred to those converted to Nizari Ismaili Islam in the Indian sub-continent from about the 13th century onward. More particularly, it included certain groups, predominantly from Gujarat and Kutch, who retained strong Indian ethnic roots and caste customs while sustaining their Muslim religious identity under continual threats of persecution. In the 19th century, the Ismaili Imamat (offce of the Imam) became established in India and a programme of consolidation and reorganisation of the community and its institutions began. These changes led to differences of opinion among Khojas. While the majority of Khojas remained Ismaili, one group became Ithna' ashari and a smaller group adopted Sunnism.
Chase, W. G., & Clark, H. H. (1972) - Mental Operations in The Comparison of Sentences and Pictures. in L. W. Gregg, Cognition in Learning and Memory. John Wiley & Sons.