In this section the results of the Principal Component Analysis (PCA) are shown. The PCA was been interpreted using Eigen value scree plots and analysis of the structure matrices. While “the number of positive eigenvalues determines the number of dimensions needed to represent a set of scores without any loss of information”[Rietveld and Van Hout, 1993] the main method for interpreting the number of relevant factors in this study consisted of looking for the factors before the breaking point or ‘elbow’ on the scree plots [Rietveld and Van Hout, 1993, Field, 1993]. For example the Figure 4.5.2 below shows the scree plot for the entire data set, with a breaking point around the 6 factor mark. When analysing the factors, the structural matrix was used as a guide. There appears to be much debate throughout the statistics literature about whether to use the pattern or structure matrix, but ultimately this study followed the conclusions of Everitt [1993], who state “when undertaking an oblique rotation, the factor structure matrix should be the focus of factor identification and interpretation”.
Tables 4.28 and 4.29 show the PCA results of both the competitive and cooperative sections of the dataset, when looking for two factors. While there was a split of sorts the results were not overly convincing. Tables 4.30 and 4.31 however show the same PCA results but when looking for three factors, interestingly this showed a clear split between the cooperative and competitive elements of the questionnaire. In this table there was also some interesting crossloading, for example the items ‘The game was a battle of skill’ and
‘My opponent was challenging’ loaded across both their own competitive factor and onto the cooperative factor. This may suggest that these two items, which focus on the competitive challenge of the opponents, contribute towards cooperative social presence. These results, though not predicted were unsurprising, as
it is logical that a greater level of competitive challenge in a team based environment would likely increase the awareness of a player’s team performance, and perhaps increase the need for teamwork to overcome greater challenge.
Figure 4.4: All Data. Rotation
Section Item Factor 1 Factor 2
1 Iactedwithmyopponentinmind 0.183 0.703
1 Ireactedtomyopponentsactions 0.327 0.651
1 Myopponentsplayedasignificantroleinmyexperienceofthegame 0.145 0.368 1 Itseemedasthoughmyopponentwasactingwithawarenessofmyactions 0.42 0.441
1 Iknewwhatmyopponentwastryingtoachieve 0.184 0.561
1 Iwasawarethatmyopponentmightworkoutmygoals 0.36 0.535
1 TheactionsofmyopponentsaffectedthewayIplayed 0.314 0.46
1 IfeltIaffectedmyopponentsactions 0.165 0.401
1 Myopponentwaschallenging 0.559 0.236
1 Thegamewasabattleofskill 0.489 0.296
1 Thegamewasabattleofwits 0.418 0.387
1 Ifelttensewhileplayingmyopponent 0.38 0.405
1 Myopponentcreatedasenseofurgency 0.442 0.311
1 Thepresenceofmyopponentmotivatedme 0.411 0.54
Table 4.28: Loadings over 0.4 highlighted. Section 1 = Competitive, Section 2 = Cooperative
Section Item Factor 1 Factor 2
2 Iwasawareofmyteam 0.525 0.542
2 Iactedwithmyteammatesinmind 0.587 0.487
2 Iconsideredmyteammatespossibleplansthoughts 0.593 0.4
2 IfeltlikeIwaspartofateam 0.808 0.422
2 Ifeltasocialconnectiontomyteammatescamaraderie 0.784 0.211
2 Ifeltmyteammateswerelookingoutforme 0.77 0.272
2 IfeltItookmorerisksingamewhenIhadmyteammateswithme 0.189 0.175
2 IfeltIcontributedtotheteam 0.137 0.48
2 Ifelttheteamhelpedme 0.705 0.276
2 Ifeltmyactionsmadeadifferencetomyteammates 0.329 0.611 2 Theactionsofmyteammatesaffectedmythoughtsandactions 0.657 0.47 2 Myteammatesplayedasignificantroleinmyexperienceofthe 0.69 0.447
2 Myteamcommunicatedwell 0.744 0.165
2 Theteamhadamutualunderstanding 0.735 0.227
2 Iputtheperformanceoftheteamovermypersonalperformance 0.591 0.265 2 Myactionsweredeterminedbytheobjectivesoftheteam 0.665 0.426
2 Iwantedmyteamtovalueme 0.497 0.484
2 Beingpartofateammotivatedme 0.803 0.375
2 Ifeltresponsibleforachievingtheteamsobjectives 0.456 0.569
2 IdidnotwantmyteamtothinkIhadletthemdown 0.538 0.424
2 Ifeltmyteamwascommittedtoworkingtogether 0.833 0.186
2 Imadeanefforttoworkwithmyteammates 0.684 0.346
2 Ifeltmyteamsharedacommonoverallaim 0.676 0.317
2 Ifeltmyteamsharedcommonshorttermgoals 0.721 0.466
2 Itwasasmuchabouttheteamasaboutmyowngame 0.728 0.321
2 Myteammateswereuseful 0.799 0.241
Table 4.29: Loadings over 0.4 highlighted. Section 1 = Competitive, Section 2 = Cooperative
Section Item Factor 1 Factor 2 Factor 3
1 Iactedwithmyopponentinmind 0.129 0.588 0.497
1 Ireactedtomyopponentsactions 0.308 0.633 0.330
1 Myopponentsplayedasignificantroleinmyexperie 0.085 0.174 0.474 1 Itseemedasthoughmyopponentwasactingwithawaren 0.343 0.144 0.682
1 Iknewwhatmyopponentwastryingtoachieve 0.168 0.567 0.244
1 Iwasawarethatmyopponentmightworkoutmygoals 0.307 0.358 0.532 1 TheactionsofmyopponentsaffectedthewayIplayed 0.265 0.294 0.479
1 IfeltIaffectedmyopponentsactions 0.148 0.375 0.224
1 Myopponentwaschallenging 0.495 -0.082 0.634
1 Thegamewasabattleofskill 0.442 0.069 0.514
1 Thegamewasabattleofwits 0.372 0.194 0.494
1 Ifelttensewhileplayingmyopponent 0.305 0.114 0.655
1 Myopponentcreatedasenseofurgency 0.370 0.001 0.647
1 Thepresenceofmyopponentmotivatedme 0.354 0.340 0.572
Table 4.30: Loadings over 0.4 highlighted. Section 1 = Competitive
Section Item Factor 1 Factor 2 Factor 3
Table 4.31: Loadings over 0.4 highlighted. Section 1 = Competitive, Section 2 = Cooperative
Competitive
In the competitive section of the CCPIGv1 there were two modules consisting of 14 items. These two modules represented the two factors of Behavioural & Cognitive Involvement (including such concepts as Theory of Mind), and Competitive Engagement. The PCA results from this section confirmed that two factors was a reasonable interpretation of the items. The scree plot for the competitive section (Fig. 4.5.2) showed a breaking point of 2 components, and the structure matrix showed a clear split in components between the two pre-defined modules (Table 4.32). Table 4.33 shows a PCA attempting to find three factors, which added little new insight, showing that two factors was the more sensible interpretation. Two items in the competitive section did not load as expected were ‘My opponents played a significant role in my experience of the game’ and ‘It seemed as though my opponent was acting with awareness of my actions’. These two items were originally in Module 1.1, designed to measure how the interplay between the player and their opponents affected their thoughts and actions, while they loaded onto Module 1.2,
designed to measure the competitive feelings of the player. Conceptually these items were potentially applicable to either module, ‘It seemed as though my opponent was acting with awareness of my actions’
was created to draw on the Theory of Mind, but this concept of an opponent being aware of one’s actions may, in the mind of the participant, be more closely related to the level of challenge an opponent presents.
The concept of ‘experience’ in the item ‘My opponents played a significant role in my experience of the game ’ , may also be conceptually closer to Module 1.2, as Module 1.1 is phrased in a more reflective way, Module 1.2 refers to the sensations of competitive play. The results of the PCA lead to a revaluation of where these items fitted conceptually, and it was decided that they would be moved from Module 1.1 to Module 1.2.
Figure 4.5: Competitive Data. Rotation
Module Item Factor 1 Factor 2
1 Iactedwithmyopponentinmind .294 .699
1 Ireactedtomyopponentsactions .261 .626 1 Myopponentsplayedasignifican... .400 .298 1 Itseemedasthoughmyopponentwa... .634 .392 1 Iknewwhatmyopponentwastrying... .078 .697 1 Iwasawarethatmyopponentmight... .442 .573 1 Theactionsofmyopponentsaffec... .418 .474 1 IfeltIaffectedmyopponentsact... .138 .577
2 Myopponentwaschallenging .756 .112
2 Thegamewasabattleofskill .639 .123
2 Thegamewasabattleofwits .534 .335
2 Ifelttensewhileplayingmyoppo... .685 .223 2 Myopponentcreatedasenseofurg... .700 .121 2 Thepresenceofmyopponentmotiv... .555 .442 Table 4.32: Two component PCA, loadings over 0.4 highlighted
Module Item Factor 1 Factor 2 Factor 3
1 Iactedwithmyopponentinmind 0.254 0.683 0.298
1 Ireactedtomyopponentsactions 0.229 0.717 0.021 1 Myopponentsplayedasignifican... 0.385 0.209 0.349 1 Itseemedasthoughmyopponentwa... 0.62 0.367 0.254 1 Iknewwhatmyopponentwastrying... 0.034 0.704 0.226 1 Iwasawarethatmyopponentmight... 0.415 0.608 0.152 1 Theactionsofmyopponentsaffec... 0.387 0.251 0.723 1 IfeltIaffectedmyopponentsact... 0.098 0.42 0.577
2 Myopponentwaschallenging 0.764 0.122 0.089
2 Thegamewasabattleofskill 0.65 0.306 -0.323
2 Thegamewasabattleofwits 0.527 0.447 -0.096
2 Ifelttensewhileplayingmyoppo... 0.684 0.241 0.102 2 Myopponentcreatedasenseofurg... 0.704 0.074 0.221 2 Thepresenceofmyopponentmotiv... 0.533 0.333 0.46
Table 4.33: Three component PCA, loadings over 0.4 highlighted
Cooperative
The PCA results for the cooperative section were less definitive than those of the competitive. The scree plot for the section suggested between three and five components which suited the presumed four modules quite well, however in general the structure matrix showed a great amount of cross loading and no convincing 4 component split. In short the cooperative section seemed to consist of 1 single component.
While this may not be as predicted the single component showed a lot of coherence, with the majority of the items loading strongly, the cooperative section as a whole having a KMO score of 0.94, and a Cronbach’s α of 0.94. However this did not mean that the modules would be abandoned in favour of a single huge section. While the PCA showed the modules cannot be statistically separated from the overall concept of cooperative social presence, it was considered that they would show the breakdown of difference aspects of the concept. The subscales in the cooperative section also scored high KMO and Cronbach’s α which suggests they did work well as subscales to the main section. This single component with sub-scales which are used to breakdown the concept they are measuring is similar to the IEQ [Jennett et al., 2008]
and GEngQ [Brockmyer et al., 2009] questionnaires. The final trimming of the CCPIGv1 was done by removing the item ‘I felt I took more risks in game when I had my teammates support’. This item did not load convincingly onto any component when considering the cooperative section as a whole and was considered conceptually too context dependant. There are games in which people cooperate and in which there are no risks to be taken, therefore the item was removed from the CCPIG on both statistical and conceptual terms.
Figure 4.6: Cooperative Data. Rotation
Module Item Factor 1 Factor 2 Factor 3 Factor 4
3 Iwasawareofmyteam .408 .003 -.697 -.322
3 Iactedwithmyteamm... .482 .253 -.705 -.174
3 Iconsideredmyteam... .565 .082 -.532 -.315
3 IfeltlikeIwaspart... .786 .199 -.642 -.245
3 Ifeltasocialconne... .819 .130 -.501 -.089
4 Ifeltmyteammatesw... .786 .144 -.556 -.135
4 IfeltItookmoreris... .188 .638 -.075 .158
4 IfeltIcontributed... .086 .744 -.219 -.388
4 Ifelttheteamhelpe... .735 .281 -.372 -.129
4 Ifeltmyactionsmad... .296 .706 -.380 -.308
4 Theactionsofmytea... .552 .074 -.701 -.354
4 Myteammatesplayed... .592 .054 -.678 -.370
4 Myteamcommunicate... .782 .064 -.477 -.062
4 Theteamhadamutual... .768 .087 -.421 -.327
5 Iputtheperformanc... .537 -.100 -.515 -.193
5 Myactionsweredete... .605 .073 -.458 -.677
5 Iwantedmyteamtova... .377 .427 -.749 -.037
5 Beingpartofateamm... .755 .053 -.638 -.383
5 Ifeltresponsiblef... .342 .284 -.439 -.782
5 Ididnotwantmyteam... .378 .109 -.705 -.271
6 Ifeltmyteamwascom... .867 .017 -.476 -.238
6 Imadeanefforttowo... .620 .015 -.643 -.282
6 Ifeltmyteamshared... .680 .234 -.386 -.518
6 Ifeltmyteamshared... .697 .174 -.506 -.467
6 Itwasasmuchaboutt... .707 .135 -.518 -.359
6 Myteammateswereus... .826 .092 -.394 -.254
Table 4.34: Four component PCA, loadings over 0.4 highlighted
Module Item Factor 1 Factor 2
3 Iwasawareofmyteam 0.545 0.41
3 Iactedwithmyteamm... 0.582 0.51 3 Iconsideredmyteam... 0.62 0.325 3 IfeltlikeIwaspart... 0.814 0.382 3 Ifeltasocialconne... 0.79 0.188 4 Ifeltmyteammatesw... 0.782 0.256 4 IfeltItookmoreris... 0.132 0.366 4 IfeltIcontributed... 0.138 0.748 4 Ifelttheteamhelpe... 0.686 0.254 4 Ifeltmyactionsmad... 0.342 0.729 4 Theactionsofmytea... 0.66 0.434 4 Myteammatesplayed... 0.687 0.403 4 Myteamcommunicate... 0.754 0.131 4 Theteamhadamutual... 0.751 0.216 5 Iputtheperformanc... 0.588 0.155 5 Myactionsweredete... 0.665 0.399 5 Iwantedmyteamtova... 0.493 0.63 5 Beingpartofateamm... 0.807 0.341 5 Ifeltresponsiblef... 0.454 0.64 5 Ididnotwantmyteam... 0.514 0.476 6 Ifeltmyteamwascom... 0.839 0.14 6 Imadeanefforttowo... 0.693 0.316 6 Ifeltmyteamshared... 0.684 0.388 6 Ifeltmyteamshared... 0.728 0.394 6 Itwasasmuchaboutt... 0.73 0.331 6 Myteammateswereus... 0.782 0.161 Table 4.35: Two component PCA, loadings over 0.4 highlighted
Module Item Factor 1 Factor 2 Factor 3
3 Iwasawareofmyteam 0.428 0.058 -0.709
3 Iactedwithmyteamm... 0.496 0.257 -0.659 3 Iconsideredmyteam... 0.576 0.145 -0.539 3 IfeltlikeIwaspart... 0.792 0.235 -0.603 3 Ifeltasocialconne... 0.815 0.137 -0.433 4 Ifeltmyteammatesw... 0.786 0.159 -0.497
4 IfeltItookmoreris... 0.183 0.564 0.019
4 IfeltIcontributed... 0.108 0.802 -0.257 4 Ifelttheteamhelpe... 0.732 0.297 -0.319 4 Ifeltmyactionsmad... 0.313 0.739 -0.376 4 Theactionsofmytea... 0.569 0.137 -0.709 4 Myteammatesplayed... 0.608 0.124 -0.69 4 Myteamcommunicate... 0.777 0.068 -0.41 4 Theteamhadamutual... 0.771 0.165 -0.426 5 Iputtheperformanc... 0.543 -0.061 -0.504 5 Myactionsweredete... 0.626 0.239 -0.562
5 Iwantedmyteamtova... 0.391 0.38 -0.662
5 Beingpartofateamm... 0.766 0.134 -0.646 5 Ifeltresponsiblef... 0.374 0.464 -0.576 5 Ididnotwantmyteam... 0.397 0.144 -0.699 6 Ifeltmyteamwascom... 0.866 0.073 -0.452 6 Imadeanefforttowo... 0.631 0.067 -0.635 6 Ifeltmyteamshared... 0.692 0.356 -0.439
6 Ifeltmyteamshared... 0.71 0.279 -0.541
6 Itwasasmuchaboutt... 0.716 0.212 -0.526 6 Myteammateswereus... 0.824 0.153 -0.378 Table 4.36: Three component PCA, loadings over 0.4 highlighted
4.5.3 Discussion
Overall then there was evidence that the CCPIGv1 was able to make a clear distinction between competitive and cooperative game play. The competitive section also seemed reasonably interpreted as two components of the awareness of other humans (previously called Behavioural Involvement & Theory of Mind) and actual engagement, the sensations and feelings of competitive play. In contrast, the cooperative section seemed to be measuring more holistic sensation, the overall cooperative experience of social engagement.
The CCPIGv1.1 (found in full in Appendix 7.2.11) the modules with unwieldy titles were re-named:
Final structure CCPIGv1.1 Section 1: Competitive
Module 1.1: Awareness
Module 1.2: Engagement Section 2: Cooperative
Module 2.1: Team Identification Module 2.2: Social Action Module 2.3: Motivation Module 2.4: Team Value
Section 1 is designed to measure ‘competitive social presence’, the social presence felt towards one’s opponents in a digital game. Module 1.1 measures competitive involvement, how the interplay between the player and their opponents affected the respondent’s thoughts and actions. The module is phrased in a fairly reflective way, it aims to measure the extent to which a respondent is using the Theory of Mind, and how aware they were of the behavioural and cognitive interplay between them and their opponent(s).
Module 1.2 measures competitive engagement, and the sensations of competitive play with another human.
Section 2 is designed to measure ‘cooperative social presence’, the social presence felt towards team-mates in cooperative digital games. As the PCA results from this study suggested this section functioned as a single component, and a number of concepts cut across the section, including Theory of Mind, and social joint commitments[Clark, 2006].
4.6 Chapter Summary
Questionnaire Development
This chapter documented the development of the Competitive and Cooperative Presence in Gaming (CCPIG) questionnaire, a measure for social presence specifically designed for use in team-based digi-tal games. The development refined a sprawling list of 116 items into a 39 item tool which successfully measures the two distinct concepts of competitive and cooperative social presence. The development also helped hone the online community survey methodology, a method of deploying a questionnaire to gain both quantitative data, and qualitative user feedback to contextualize the results. Questionnaires are useful for gathering quantitative data and can reveal much about user experience, but it a pragmatic view of the data must be taken to gain valuable insights. In complex scenarios such as those found in team-based online games it is beneficial to combine the results of a questionnaire with observations of gameplay, and a knowledge of the games and game communities involved in the studies.
Social Presence
The studies in this chapter not only contributed to the development of the CCPIG, but enlightened more of the nature of social presence, and the user perceptions of social presence in team-based digital games. The development of the questionnaire confirmed that the experiential vignettes of the previous chapter were revealing reliable insights into social presence in team-based digital games. The items of the questionnaire were based upon the results of the vignettes, and the these coherence and reliability of these items shows that the results were valid. The CCPIG was further validated in Chapter 5 and the single factor structure of the cooperative section was rearranged and split into two factors following a principal component analysis with a larger data set.
Further Work
While the development of the CCPIG could be considered successful, in that it produced a functioning questionnaire, the structure of the cooperative section was still of some concern. It was expected that the cooperative section would split into a number of other factors, and the results which suggested it was in fact one single factor made the section seem large and ungainly compared to the competitive section.
Therefore, rather than simply accept the results of the PCA analysis in this chapter, further factor analysis was be conducted using the far larger dataset gathered for the team trust study in the following chapter.
This analysis would confirm that the cooperative section was not one single factor, but could be scored as two conceptually coherent factors, providing a far more useful and fine grain view of cooperative social presence in team-based games. The user feedback from the various studies also suggested a number of elements of gameplay context which might affect levels of social presence. Respondents suggested that playing with friends or strangers greatly affects the levels of social presence in team-based games, a concept which would be investigated in the final study of this thesis.