MATERIALS AND METHODS

The measurement items in this study are 86 questionnaires were adopted and some were adapted from the existing literature to suit the current study (Table 1 in Appendix A). First, the items for measuring behaviour factors contained three constructs: Perceived Behaviour, Attitude and Subjective Norms and were measured with 30 items. The TPB scales adopted from Heirman et al.³² and Al-Debei et al.³³. The researcher, thus, created five items in Attitude, three items in Perceived Behaviour and finally eight items in Subjective Norms (Appendix A). Second, the communication factors contained three constructs: Visual, Textual and Duration. The researcher adapted eight items from Bailey³⁶ as well as Elmasry et al.³⁷ five-items. The researcher generated twelve items in these three factors (Table 1 in Appendix A). Third, the measurement for online social interaction contained two constructs: Online Friendships and Online Relationships, which were measured with 17-items using Facebook Relational Maintenance Measurement (FRMM) adopted from McEwan et al.³⁸.

Appendix A Table 1:	Adapted/adopted questionnaire items

The FRMM was valid and reliable for this study because many researchers in the field used it in their study to measure different aspects of friendship and relationship, either online and offline^39-41. The FRMM was been marked by Walther³⁰ as a great task and function of social media. The researchers McEwan et al.³⁸ developed the FRMM scale to suit the changes that occur in social media in order to measure the interpersonal relationship with high validity, as it includes all the new and recent technics, posts, private messages, likes, tags, status and profile.

The research was conducted through a self-administered survey with a total of 86 questions. A five-point Likert scale was used as the measure for the participants’ responses on the questionnaire. The questionnaire was validated through a pre-test and a pilot test.

Appendix A Table 2:	Reliability for pilot test (N = 30)
Cronbach's alpha: Measured the correlation of two items that measure the same construct

Appendix A Table 3:	Number of indicators in each construct
Interval: Type of measurement that used numbers in order to gives result of hypothesis

For the pre-test, a questionnaire was given to three (3) experts in the field, who verified its logical consistency, ease of understanding, wording and the appropriateness of the instruments. The pilot test was conducted using 30 Saudi Facebook users (Table 2, 3 in Appendix A).

Survey administration: The research model was tested using data collected from Facebook users in Saudi Arabia (18th May-20th August, 2017). Facebook was chosen, as it is the second most popular site in the world. In addition, the usage of Facebook amongst the Saudi population is increasing yearly. This conforms to a report released in 2014, which revealed that the number of Facebook users in Saudi Arabia had risen from six million in 2012 to 7.8 million in 2013. In 2015, this number had grown to 12 million active users⁴².

Data was collected via a self-administered survey using a stratified random sampling method. The sample was composed of 500 participants (450 after data screening) with the study targeting all Saudi Facebook users. The questionnaire had a cover page that presented the purpose of the study, the freedom to withdraw at any time and an assurance of privacy. The participants were instructed to answer all the questions based on their experience using Facebook. After they finished the data-collection process, small rewards were given to all the participants in the submission. The demographic information of the respondents was listed in Table 1.

Data analysis: The study used SPSS and AMOS software by applying the Structural Equation Model (SEM) like many recent and close earlier studies in the field^{10,31,34,39,43}. The first step was CFA, followed by a measurement model and the last step tested the hypothesis through structural modelling (hypothesis model). Both the CFA and the measurement models tested the reliability and validity of the measures before assessing the structural model. The CFA, measurement and hypothesis models were also tested for the goodness-of-fit and met the cut-point as shown in Table 4-6 in Appendix A.

Appendix A Table 4:	CFA Goodness of fit index (n = 450)

Appendix A Table 5:	Fit of measurement model

Appendix A Table 6:	Fit of hypothesized model (structural model)

Table 1:	Information about the participants of the study (n = 450)

Therefore, the first step involved evaluating the CFA and the measurement model on the criteria of reliability, convergent validity and discriminant validity. The purpose was to test the relationship between the variables in the structural model^31,44. The, structural model was applied to test the twelve hypothesis.

RESULTS AND DISCUSSION

Tests for reliability and validity: In regards to composite reliability (CR) values, all of the CR values were well above the 0.9 thresholds and all of the AVE values ranged from 0.92-0.99 as shown in Table 2. A threshold of above 0.70 was used to establish convergent validity. Table 3 showed that all the indicators' loadings on their corresponding constructs exceeded 0.70. For the evaluation of convergent validity, the threshold used was 0.5 or greater for the average variance extracted (AVE) for each construct. The cross-loadings and the square root of the AVEs were used to establish the discriminant validity. The square roots of the AVE on the diagonal of the correlation matrix were greater than the corresponding off-diagonal inter-construct correlations as was presented in Table 4. The results indicated that the first and the second steps fulfilled the criteria of composite reliability, convergent validity and discriminant validity.

Structural-path analysis was the third and last step in SEM. The results of the model were as presented in Table 5. Six of the hypotheses displayed a p-value of less than 0.05 and therefore, they were accepted. The other six that had a p-value of above 0.5 were rejected. Overall, the base model as presented in (Fig. 2) explained 0.154 of the variance in the online friendships and 0.015 of the variance in the online relationship. This was supported by the results of the descriptive statistics (Table 7, 8 in Appendix A) that showed that the 450 participants of the study had online friendships while only a small number, 11.6%, had online relationships. This explained the rejection of all online relationship hypotheses as well as the lower number of R² on the hypothesis.

Table 2:	Reliability and average variance extracted of the constructs

Table 3:	Convergent validity for constructs
Factor loading2: Factor loading is basically the correlation coefficient for the variable and factor. Factor loading shows the variance explained by the variable on that particular factor

Table 4:	Discriminant validity (the square roots of the average variance extracted)
**Correlation is significant at the 0.01 level (2-tailed). *Correlation is significant at the 0.05 level, BEA: Attitude, BEB: Perceived Behavior, BEN: Subjective norms, COMV: Visual, COMT: Textual, COMD: Duration, SIF: Friendship, SIR: Relationship

Appendix A Table 7:	Percentage of online friendship

Appendix A Table 8:	Percentage of online relationship

Fig. 2:	Structural model

Table 5:	Direct hypotheses testing result
BEA: Attitude, BEB: Perceived behavior, BEN: Subjective norms, COMV: Visual, COMT: Textual, COMD: Duration, SIF: Friendship, SIR: Relationship

Structural model also called (Hypothesis model) which gives the result of the 12 Hypothesis (the relation between the independent and dependent variables) which mentioned above in Fig. 1.

This Hypothesis model gives the result (Estimate, Beta, C.R. and P) which presented in Table 5.

DISCUSSION

Behaviourial factors: As TPB theory explains, Facebook users react to and act on specific behaviour. In this study, the behaviour factors represented how Saudi Arabian Facebook users respond to online friendships and relationships and were tested using H₁-H₆. The results from the hypothesis model revealed that behaviour factors had a significant influence on online friendships and a very limited impact on online relationships. H₁ and H₂ examined the influence of attitude on online friendships and relationships. Attitude presents the extent to which a person has a positive or negative evaluation towards a specific behaviour⁹. H₁ was accepted indicating that the attitude factor had a positive influence on online friendships for the Saudis. Therefore, the study revealed that Saudi Facebook users have a positive opinion about using Facebook and about Facebook friendships. On the contrary, H₂ was rejected indicating that attitude did not impact online relationships for the Saudis. This was attributed to the fact that the positive perception of online relationships was rejected and found unexciting in the Saudi culture and religion^17,19. In fact, most of the Saudis recognise that to be a harmful and dangerous behaviour that could yield detrimental outcomes in form of divorces, honour killings or extortions²⁰.

H₃ and H₄ tested the impact of perceived behaviour on online friendships and relationships. According to TPB, the perceived behaviour is considered as the reason for the situation where a person has less control over a certain behaviour. According to Ajzen¹⁰ this component influences the attitude that one has towards a behaviour⁴⁵. The H₃ was accepted indicating that perceived behaviour influences online friendships, while H₄ was rejected indicating that perceived behaviour has no significant influence on online relationships.

H₅ and H6 tested the impact of subjective norms on the online friendships and relationships. According to Ajzen¹⁰ subjective norm is comprised of the beliefs about the normative expectation of others. H₅ was accepted indicating that subjective norms have a significant impact on online friendships, while H6 was rejected indicating that subjective norms have a limited impact on online relationships. The participants exhibited a high inclination towards meeting the expectations of the society and of the people around them. They also drew influence from the imperative and influential people, who displayed a similar behaviour such as accepting friendship requests from people that they had never met and accepting friendship requests from people of the opposite gender.

The study showed attitude to be the highest behavioural factor (β = 0.180) in influencing online friendships, followed by subjective norms (β = 0.172) and the least was perceived behaviour (β = 0.101). The study took a different and unique approach in examining TPB in the online social interaction across both genders, which cannot be compared with the past studies as they assessed online friendships and relationships on different perspectives and in different cultures^32,34. In entirety, the results of the current study were consistent with the TPB theory, as it explained the behaviour of Saudi Facebook users with respect to online social interaction.

A study by Al-Saggaf and Begg¹³ found that gender segregation influenced Saudi Arabian’s attitude towards their online identity presentation. The study also revealed that Saudi females found difficulties in making Facebook friendships and relationships because they were expected to minimize communications with the opposite sex as expected from them. On the other hand, Saudi males had more freedom regarding Facebook friendship and relationship. This present study was conducted seven years later and found some changes in the Saudis’ attitude as well as norms due to the introduction of new technology. The current study found that Saudi Arabians represent a good level of online friendships across genders as the males to females was 36.9% while 28% was from females to males. This was higher than the same-sex online friendships where males to males was 14.9% and females to females was 19.8% (Table 7 and 8 in Appendix A). The findings of the study were also in line with past study by Wang et al.⁴⁶ which indicated that both male and female subjects were more willing to initiate friendships with opposite-sex as the cross-gender relationships controlled online friendships.

Communication factors: In regards to the communication factors, the model showed that all the three communication factors have significant positive influence on online friendship and no great impact on online relationships. The factors were tested using H₇-H₁₂. H₇ and H₈ tested the impact of the visual factor on the online friendships and relationships. H₇ was accepted indicating that visual factors had a significant impact on online friendships, while H₈ was rejected indicating that the visual factors did not have a significant impact on the online relationships. The findings were consistent with past studies that highlighted the importance of the visual factors in online interactions^{1,5,8,10,25,28,46-48}.

H₉ and H₁₀ tested the impact of the textual factors on the online friendships and relationships. H₉ was accepted indicating that textual factors had a significant impact on online friendships, while H₁₀ was rejected indicating that the textual factors did not have a significant impact on the online relationships. The findings were consistent with past researches that mention that the online users reveal personal and sensitive information with the other user with whom they are exchanging personal information^{3,11,21,27,31}.

H₁₁ and H₁₂ assessed the impact of the duration factors on the online friendships and relationships. H₉ was accepted indicating that duration factors had a significant impact on online friendships, while H₁₀ was rejected indicating that the duration factors did not have a significant impact on the online relationships. The results were also consistent with previous studies that highlighted that time is considered as an important factor in the development of online relations^12,31. Also, previous scholars asserted that when people spend more time interacting online, the chances of developing close ties increased. Besides, they also showed that when people frequently communicate online, they tend to become at ease with each other and develop close ties⁴⁹.

The study revealed the textual factors to be the most influential communication factor (β = 0.148) in influencing online friendships, followed by the visual factors (β = 0.121) and the least was duration (β = 0.101). The study revealed that the Saudis were honest in the texts more than the visual aspects due to the cultural restrictions. The Saudi Arabian Facebook users dealt more with the textual aspects (breadth of information-brief statement, honesty and confidence in self-presentation) than the visual (breadth of information-avatar, fake images, other image and part of the body image). This was consistent with a past study by Guta and Karolak⁵⁰ which revealed that Saudi females behave more carefully when they present themselves online because they understand what their society expects from them. Thus, they had a greater reluctance in posting and presenting their photos online. Elmasry et al.³⁷ also indicated that the Saudis had a great inclination to use fake, avatar and part of their body (selective self) photos instead of a real photos, which further supported the research findings on the supremacy of the textual over the visual factors in influencing the Saudis online interactions.

The order of influence was inconsistent with the previous studies in the Western and far Asia that argued that the visual factors had the greatest impact on the online interactions. The researcher referred the difference in the order of importance due to the variances in the cultural orientations, whereby the Saudi culture is more limited in the interactions between both the genders. The current study was consistent with past studies on the positive and significant influence of the communication factors on online friendship. However, the study was inconsistent with the findings from previous studies that indicated that communication factors influence online relationships. This was also attributed to the conservative nature of the Saudis’ culture.

Strengths and limitations: The study engaged a high number of female participants, 48.2%, surpassing the challenge of low female participation due to gender segregation and females’ identity in Saudi. Nevertheless, there were some limitations to the study. The Saudi’s culture was identified as a limiting factor to the freedom of interaction among the individuals, particularly the females. Also, the model was only assessed on Facebook whereas there are other social platforms like Snapshot, Twitter and Instagram that influence online social interactions. The study also limited its assessment on posts, private messages, likes, tags, status and profile and did not assess all the elements on Facebook.

SIGNIFICANCE AND STATEMENT

The present study evaluated the influence of both behaviour and communication factors on online social interaction in Saudi. The study findings also provide the Saudi government and the Al-Ma'arouf Commission with a better understanding of the online behaviour of the Saudi Arabian. The study also provides insights to future researchers because they will build on the theory derived from the study for further explorations on online social interactions.

CONCLUSION

The study revealed the existence of both online friendships and relationships amongst the Saudi Arabians, but with only a small number of online relationships due to the Saudi’s conservative culture. The results indicated that both behavioural factors and communication factors significantly influence online friendship across gender in Saudi and explains 0.154 of the online friendship. Dissimilarly, both behavioural factors and communication factors explained only 0.015 of the online relationships.

ACKNOWLEDGMENTS

I would like to acknowledge the inputs of past researchers in trying to create an understanding of core issues that surround the use of Facebook and related demographics. I also acknowledge the inputs of Umm Al-Qura University, Mecca, for offering their database and resources to ensuring that this project is a success.