Preliminary research on the effect of spatial layout on peer academic support relationships in first-year university students: a case study of the school of architecture at SCUT

ABSTRACT Contemporary university curricula increasingly encourage students to develop peer academic support relationships, especially first-year students who are facing the transition from teacher-oriented to self-directed learning. Previous studies have identified spatial layout as a factor impacting peer relationships by comparing the effects of different spatial layout categories. More insights into the quantitative correlation between spatial layout and peer academic support relationships will meaningfully complement the existing findings. This paper therefore uses spatial configuration and spatial proximity as quantitative measures and adopts a quasi-experimental approach to investigate the effects of spatial layout on peer academic support relationships in first-year university students. Data were collected from a sample of first-year students enrolled in the School of Architecture at South China University of Technology. We use longitudinal data to measure and compare the peer academic support relationships as well as the corresponding spatial layout. The results support the importance of the spatial layout of the learning space for the development of peer academic support relationships. In addition, suggestions for university administrators and architects are proposed.


Introduction
Contemporary university curricula increasingly encourage students to develop peer academic support relationships (Brouwer et al. 2018;Celant 2013), especially first-year students who are facing the transition from teacher-oriented to self-directed learning (Chow et al., 2008;Brouwer et al. 2016).Building peer relationships is particularly challenging but significant for first-year students.First, since most students move to university cities that are physically distant from their families, classmates who live and study together may become the most important source of support to help them adjust to university life (Chow and Healey 2008;Buote et al. 2007).Building peer relationships during the transition period contributes to first-year students' engagement and study success (Brouwer et al. 2016).Second, the self-directed learning mode in a university depends on the peer academic support network more than other motivations for self-directed study.Academic communication and collaboration play a significant role in the learning process.
Previous studies have identified several factors affecting support networks.Studies linking learning space with learning process and outcomes mainly focus on indoor environment quality, such as ventilation, lighting, and acoustics (Castilla et al. 2018;Lewinski 2015;Marchand et al. 2014;Lansdale et al. 2011;Montello 1988), which relate to the comfort dimension.Sociopsychological factors (López-Chao et al., 2021) and spatial layout typologies (Park et al., 2014;Shernoff et al. 2017) are discussed as well, revealing the social dimension of learning space.Existing studies on spatial layout have proven the effect on social relationships by comparing different categories of learning space.More insights into the effect of spatial layout on peer relationships derived by quantitative methods are meaningful for complementing the research findings.
The quantitative correlation between spatial layout and social relationships has been examined in the workplace (Peponis et al. 2007;Sailer et al., 2012;Wineman et al. 2014).Spatial configuration as a global measure and spatial proximity as a local measure show how spatial layout affects social interaction.However, research on university learning space remains scarce.The difference in occupant behaviour and spatial layout calls for specialized empirical studies on the spatial effect on university students.This paper therefore sets out to explore how the spatial layout of the learning space affects peer academic support relationships in first-year university students.While there are different dimensions of peer relationships (social and academic), we have focused on academic support relationships, as they are more attached to the learning space and directly benefit academic performance.In applying spatial configuration and spatial proximity as quantitative measures of spatial layout, this paper adopts a quasi-experimental approach to investigate how spatial layout influences freshman peer academic support relationships on both the organizational scale and personal scale.Data were collected from a sample of first-year students enrolled in the school of architecture at South China University of Technology who experienced a change of classrooms for major courses.We use longitudinal data to measure and compare the effect of spatial layout on peer academic support relationships among the same group of students in two semesters.The results support the importance of spatial layout for building peer academic support relationships.

Peer academic support relationships
Identifying the factors that affect academic performance is an essential part of educational research.Previous studies have documented the importance of intelligence, motivation, personality traits, and personal behaviours (e.g., class attendance) (Busato et al. 2000;Farsides et al., 2003;Kassarnig et al. 2018).There is a growing interest in the influence of social ties, which have been argued to be positively associated with academic performance (Brouwer et al. 2016;Tomás-Miquel, Expósito-Langa, and Nicolau-Juliá 2016;Eggens, Van der Werf, and Bosker 2008;Rizzuto, Ledoux, and Hatala 2009).In higher education, social ties consist of peer relationships, family relationships, and faculty relationships (Brouwer et al. 2016).Peer relationships have been argued to be the most influential form of support since university students stay together most frequently and provide psychological support and exchange knowledge with each other (Vignery et al., 2020).For first-year university students who are facing the transition from teacheroriented to self-directed learning (Chow and Healey 2008;Brouwer et al. 2016), peer relationships are particularly important for helping them adjust to university life and obtain engagement (Tani, Gheith, and Papaluca 2021;Thiele, Sauer, and Kauffeld 2018;Brooman and Darwent 2014) and study success (Brouwer et al. 2016;Tomás-Miquel, Expósito-Langa, and Nicolau-Juliá 2016;Eggens, Van der Werf, and Bosker 2008).Among various peer relationships, academic support relationships that are related to knowledge exchange may have a stronger influence on academic performance than those more related to friendship (Tomás-Miquel, Expósito-Langa, and Nicolau-Juliá 2016).A student's peer academic support network consists of those whom he or she has frequent academic communication and collaboration with.Work-related communication and collaboration are necessary for a work process to culminate in success (Hillier et al., 1991;Penn andHillier 1992 &, 1999;Allen et al., 2007;Wineman et al. 2014;Sailer et al., 2009).In previous studies, work-related relationships were evaluated and measured by face-to-face interaction frequency or network properties (Allen and Henn 2007;Sailer and McCulloh 2012;Wineman, Kabo, and Davis 2009).

Learning space in higher education
Learning space in higher education has been proven to be an essential element for student academic outcomes.It has specific conditions related to teaching methodologies, including promoting academic interaction in the classroom (López-Chao and López-Pena 2021).Existing studies linking learning space with learning process and outcomes mainly focus on indoor environment quality, such as ventilation, lighting, acoustics, etc. (Castilla et al. 2018;Lewinski 2015;Marchand et al. 2014;Lansdale et al. 2011;Montello 1988).The findings often relate ventilation and lighting to the design of windows and light shelves (Kruger et al., 2004), while acoustics is related to the relative position of recreation space and material choices (Zannin et al., 2009).Whether and how much these factors affect comfort during learning processes are the focus of research.There are also studies focusing on the spatial effect on social interactions.López-Chao and López-Pena (2021) considered sociopsychological factors in their study and identified satisfaction, functionality, the possibilities of social interaction, and place attachment as part of the social dimension of learning space.They argued that flexible spaces, which relate social relationships to the spatial layout typology, are more appropriate for interaction.There are other similar studies linking the spatial layout of learning space with academic interaction.For instance, the educational effects of traditional classrooms and active learning classrooms have been compared based on questionnaire surveys (Park and Choi 2014).The results show that traditional classrooms may disadvantage student learning experiences for those in certain seat positions and that the gap in learning attitudes was offset in active learning classrooms.The better performance of active learning classrooms is due to their promotion of interactive and collaborative learning through spatial layout.The influence of students' seating location on student engagement, classroom experience, and academic performance is also examined in the setting of a large university lecture hall (Shernoff et al. 2017).Such studies prove the effect of spatial layout on academic relationships by comparing different categories of learning space.Since a multimethod approach would be beneficial to complement the research approaches already used, more insight into the quantitative correlation between spatial layout and peer academic support relationships could be meaningfully extracted from mixed method studies.

Effect of spatial layout on social relationships
Spatial layout as a physical arrangement enables space to acquire its social logic through the probabilities of encounters by their frequency and type (Hillier et al., 1984).Existing research on the measurement of the spatial layout effect on performance has mainly focused on working spaces within office buildings.Office spatial layout is argued to be an enabling factor of social relations by affecting the way workers move and interact in the workplace (Sugiyama et al. 2021;Haapakangas et al. 2019;Wineman, Kabo, and Davis 2009).The measures studied in office spatial layouts include the classification of office type (closed-or open-plan office), the amount and location of shared space (Hua et al. 2011), spatial configuration (Hillier and Penn 1991;Penn et al., 1992Penn et al., &, 1999;;Allen and Henn 2007;Peponis et al. 2007;Wineman, Kabo, andDavis 2009 &, 2014;Sailer andMcCulloh 2012 &, 2019;Deng, Liu, and Ji 2021), distances between coworkers, and visibility of coworkers (Sugiyama et al. 2021).The correlations between these spatial measures and workrelated communication or collaboration are examined to explain the effect of spatial layout.
Spatial configuration is a core concept of space syntax (Hillier and Hanson 1984), which is a sophisticated analytic technique widely applied for measuring spatial layout.Spatial configuration describes how space is organized and reveals the overall topological relationships of each space.Figure 1 is an illustration of different spatial configurations.Different spatial configurations result in different levels of accessibility, which impacts actual movement that may trigger encounters and social interactions.Integration and choice are representative syntactic properties used to quantify spatial configuration in terms of to-movement and through-movement potentials on a global scale.Although spatial configuration cannot reflect real movement precisely because of the multiple factors involved in movement, to some extent it reflects the spatial potential for movement according to accessibility.Previous empirical studies have proven strong correlations between real movement and spatial configuration (e.g., Serrato et al., 1999;Penn, Desyllas, and Vaughan 1999;Wineman, Kabo, and Davis 2009;Liu et al. 2021), indicating the applicability and reliability of spatial configuration as a global predictor in spatial layout studies.
Spatial proximity is another important measure of spatial layout, especially on the local scale, which measures interpersonal distances.Kabo (2017) argues that the dyad is the most basic unit of social interaction and implies that a dyadic level of analysis is necessary.Previous studies have found that spatial proximity can lower the barriers to encounters and dyadic communication, thus facilitating dyadic relationship formations (Allen 1977;Allen and Fustfeld 1975;Allen and Henn 2007;Festinger, Schachter, and Back 1950;Sailer and McCulloh 2012;Wineman, Kabo, and Davis 2009;Kabo 2017).Except for physical distance, visibility is a measurement of spatial proximity as well.Existing studies have proven the existence of stronger relationships between those who have interpersonal visibility and face-to-face interaction in the workplace (Stryker, 2011;Penn et al., 1992;Sailer and McCulloh 2012;Peponis et al. 2007;Toker et al., 2008).
The correlation between spatial layout and social relationships has been examined in the workplace.In Peponis's (2007) study of a communication design firm, the social network properties (density and centrality) were used to measure the effect of spatial layout.Sailer (2012) used exponential random graph models (ERGMs) as network probability models to analyse whether and how much spatial configuration contributed to social network construction in four organizations.Wineman (2014) argued that spatial layout structures patterns of circulation, proximity, awareness of others, and encounters in organizations, which are fundamental to social networks.In her study of three organizations, the spatial layout measures of proximity and movement choice and the social network measures of betweenness and degree are examined to understand their influence on innovation.(Hillier 2007, 22).
Although the research framework of the quantitative correlation between spatial layout and social networks has been developed in the workplace, existing empirical studies remain scarce in learning spaces for university students.Unlike employees, modern university students are supposed to take responsibility for their own learning process.Except for lecture time, student use of classrooms is not constrained by time, while employee attendance should conform with organizational regulations.Moreover, there is no hierarchical difference (such as a manager and subordinates in offices), and students share the learning space more equally.Therefore, the occupant behaviour in learning spaces differs from workers in offices in terms of time and styles.The difference in occupant behaviour and spatial layout calls for specialized empirical studies on the effect of spatial layout on university students.

Conceptual model and research scope
The scope of the present study aims to examine the relationships between the spatial layout of the learning space and peer academic support relationships in first-year university students.The conceptual model is summarized in Figure 2, which shows the measures of spatial layout and academic support network on personal and organizational scales.This study sets out to examine whether and how overall spatial configuration and dyadic spatial proximity (calculated by Depthmap) matter for personal and organizational academic support relationships (calculated by UCINET).The following two research questions are derived from the theoretical background related to the spatial layout effect on performance: 1) How does spatial configuration affect academic support networks in first-year university students?2) How does dyadic spatial proximity affect personal academic support relationships in first-year university students?

Research design and procedure
The research was conducted using an exploratory quasi-experimental approach that sought to examine the relationship between peer academic support relationships and spatial layout in terms of spatial configuration and spatial proximity.To avoid other impact factors on academic support relationships, the experiment was designed as pre-and posttreatments that compare peer academic support relationships within the same organization in two spatial layout settings.
First, the floor plans and seat allocation of the selected classrooms were obtained as the original spatial layout data.An online survey was conducted to collect original data for peer academic support relationships, including attendance, academic communication relationships, academic collaboration relationships, and organizational proximity.Attendance enables the comparison of the frequency of face-to-face academic communication across the two semesters.The data of academic communication relationships, academic collaboration relationships, and organizational proximity are obtained at the dyadic dyad level to construct interpersonal relationship matrices.The organizational proximity was used as a reference to discuss spatial layout effects.
Then, based on the original data, spatial configuration and proximity as layout measures are calculated by Depthmap (Turner 2006).Frequency, density, and centrality as network measures are calculated by UCINET (Bugatti et al., 2002).
Data was analysed at the organizational scale and the personal scale.On the organizational scale, the preand postdifference of spatial configuration and academic support network properties were compared by t test.On a personal scale, the correlations between spatial proximity matrices and academic support relationship matrices were examined by the QAP correlation and regression analysis with a randomization test.

Sample
The experiment was conducted at South China University of Technology (SCUT).Data were collected from five classes of first-year students enrolled in the school of architecture.They experienced a move at the beginning of the second semester, which enables us to measure the effect of spatial layout in a pre-and postexperiment.The studied classrooms consist of personal desks and shared discussion tables.Because of the disciplinary requirement for manual drawing and modelling, each student owns a fixed desk and seat for work.The discussion tables are used for class with teachers approximately twice a week.Except for class time, students can use these classrooms at any time.Due to the flexibility and autonomy of use, the occupant behaviour of the learning space is less affected by teaching methodology.Academic interaction in classrooms occurs more spontaneously.Although the students belong to five classes with three overarching majors (two in architectural design: AD1, AD2; two in urban planning: UP1, UP2; and one in landscape architecture: LA), they have the same courses in the first year.In other words, they had the same lecture time and learning tasks.In addition, most spatial characteristics (e.g., building, interiors, furniture) are the same except for the spatial layout.Therefore, these five classes are comparable regarding use purposes and spatial quality (except spatial layout).
We conducted an online survey to collect peer academic support network data across two waves: at the end of the first semester and the end of the second semester in the 2020-2021 academic year.In the first semester, out of 207 students, 177 respondents completed the questionnaire (response rate of 85.5%).In the second semester, one international student dropped out due to the COVID-19 pandemic, and the response rate increased to 97.6%.To ensure the credibility of the pre-and postresearch, we use only the data from respondents of both semesters (response rate of 84.1%) for analysis.The composition of respondents is represented in Table 1.We informed students about the study's aims, procedure, and ethical aspects.The completion time was 10-15 min, and participation was voluntary.

Measures for peer academic support network
Attendance, organizational proximity, academic communication relationships, and academic collaboration relationships were obtained by an online survey.
• Attendance Attendance is measured on a five-point scale by a questionnaire with all students in both semesters: 5 = daily, 4 = several times per week; 3 = weekly, 2 = monthly, 1 = less than monthly.
• Organizational proximity Organizational proximity is investigated as a reference to demonstrate the spatial layout effect on academic support relationships.The question lists all the names of classmates in a complete list.The respondent makes ranked choices (0 = no organizational relation, 1 = one or more organizational relations) according to his or her relations with each classmate.The surveyed organizational relations include roommates, former acquaintances, members of the same student union, classmates in elective courses, members of the same major course groups, etc.The data collected from all the respondents will be combined into a symmetrical binary matrix.Only if both individuals choose "1" (one or more organizational relations) will the relationship between the dyad be recorded as "1".
• Academic communication relationships Similar to the organizational network questions, respondents rate the frequency of their face-to-face academic communication (taking place in classrooms; 5 = daily, 4 = several times per week, 3 = weekly, 2 = monthly, 1 = less than monthly) with each classmate whose name is listed.Only if both individuals choose greater than "4" (several times a week) will the dyad be recorded with the mean value of the two sides in a symmetrical matrix, representing the interpersonal academic communication relationship.In the analysis related to frequency, we use this matrix with multiple values.Regarding the social network analysis, we transform the matrix into binary values.

• Academic collaboration relationships
As collaboration chances depend more on the task requirements for first-year university students, we survey whether they have opportunities for academic collaboration instead of the frequency of academic collaboration.The obtained data are used for analysis on a personal scale (to analyse who they collaborate with) but not on an organizational scale (to compare the change in frequency pre-and postexperience).The question of academic collaboration is set with the options (0 = none, 1 = have once or more times).Only if both sides choose "1" (once or more times) will the dyad be recorded as "1".The data collected are combined into a binary matrix that represents the interpersonal academic collaboration relationships within the class.
Based on the above data, social network analysis by UCINET 6 (version 6.212) is conducted to evaluate academic support networks.The attributes that we employ for network evaluation are as follows.
• Density Density measures the number of directed relationships divided by the number of possible directed relationships (Wasserman et al., 1994), which is used to evaluate the academic communication network on an organizational scale.In this study, density by groups is calculated as well.The results help demonstrate the influence of separated classrooms.
• Normalized degree centrality Degree is simply the number of connections a student has.The normalized degree centrality is the existing degrees divided by the maximum possible degrees expressed as a percentage (Borgatti, Everett, and Freeman 2002), which can reflect a student's level of popularity (Wasserman and Faust 1994) within the network.

Measures of spatial layout
The measures of spatial layout are calculated in Depthmap.We conduct angular segment analysis on an organizational scale to calculate the spatial configuration properties.Angular segment analysis is a powerful tool for measuring accessibility in linear networks and thus predicting social activities (Al-Sayed 2014).This analysis is on the level of segments constructed according to the movement space, considering their topological, metric, and angular connections.The distance between two segments is the least angular cost within the network.The linear network in Figure 3 is an example of angular segment analysis of a classroom.Different colours represent different accessibility values, such as integration and choice, which are global attributes of the spatial network.Integration measures the accessibility (calculated by angular cost) of a segment as the destination from all other segments within the network.Choice measures the probability of a segment on the shortest path (calculated by angular cost) between every two segments within the network.The specific syntactic properties that we employ by angular segment analysis are as follows.
• NAIN (R = n) Integration measures the to-movement potential of a space, which is one of the most powerful syntactic properties in both analysis and movement prediction.It shows how deep or shallow space is in relation to all other spaces.Using integration, spaces are ranked from the most integrated to the most segregated.Integration is usually indicative of how accessible space is and is thought to correspond to rates of social encounter (Hillier 1996).In this study, we conduct angular segment analysis to calculate the integration value of each segment within a class on a global scale (R = n). 1 The results are normalized to permit comparison of different models.NAIN represents the normalized integration.
• NACH: (R = n) Choice measures the through-movement potential of a space.Segments that record high global choice are located on the shortest paths from all origins to all destinations.Compared to integration, choice is descriptive of movement rather than occupation.In this study, the choice of each segment within a class is calculated by angular segment analysis on a global scale (R = n).NACH represents the normalized choice.
We conduct step depth analysis on a personal scale to calculate the spatial proximity of dyads (metric step shortest-path length, visual step depth).Step depth analysis is based on a graph of a plan, which is divided into grids, in which each grid is a unit for analysis.The spatial relationship of each unit can be calculated in terms of interaccessibility or intervisibility.In our study, the analysis grid is set as 250 mm*250 mm, which can locate each student's seat precisely.Figure 4 shows an example of step depth analysis of a seat within a classroom.The syntactic properties that we employ in this analysis are as follows.
• Dyad physical proximity: Metric step shortestpath length The metric step shortest-path length measures physical proximity by the metric distance between every two students' seat locations.It is calculated by step depth analysis in Depthmap.The results are presented as a symmetrical matrix with metric distance values.

Spatial configuration of pre-and posttreatment
The spatial configuration properties of each class are compared between pre-and posttreatment (Table 2).
The results of NACH (R = n) did not show a significant change, while the NAIN (R = N) of all five classes showed a significant change: AD1 and UP2 increased while AD2, UP1, and LA decreased.The pre-and postspatial layout of each class are presented in Figures 5  and 6.Students of AD1 are divided into two classrooms in both semesters.The increase in NAIN (R = n) is due to the absence of a floor difference between the two classrooms in the second semester.Students of AD2 and LA were in a single classroom in the first semester, while they were divided into three and two classrooms on the same floor in the second semester.This results in a significant decrease in NAIN (R = n).Although UP1 and UP2 did not experience separated classrooms, the NAIN (R = n) changed due to the layout of the furniture.The appearance of cul-de-sacs in the post plan of UP1 reduces the NAIN (R = n), while the seat location is adjusted directly along the main path that increases accessibility in the post plan of UP2.

Peer academic support relationships pre-and posttreatment
Table 3 shows the peer academic communication frequency before and after the change in classrooms.Among the five classes, the average attendance in the four of the classes changes.To avoid the influence of attendance, we use the ratio of academic communication frequency per session (academic communication frequency divided by attendance) to evaluate the change in academic communication behaviour between pre-and posttreatment.The results show that the values of AD2, UP1, and LA significantly decreased, which corresponds with the NAIN(R = n) results.Although the trends of AD1 and UP2 (increase) correspond to the NAIN (R = n) (increase), the p values by t test do not show significance at the 0.05 level.The corresponding trends of the five classes demonstrate the effect of spatial layout on peer academic communication to some extent.The density comparison of academic communication networks in pre-and posttreatment is conducted in UCINET 6, which runs the bootstrap paired sample t test.The results of density and degree centrality (Tables 4 and 5) are similar, with a significant decrease in AD2, UP1, and LA.The corresponding trend of the network following spatial change indicates the effect of spatial layout on peer academic support networks within a class.In particular, the more segregated the learning space, the worse the peer academic support network will be.The spatial configuration of the learning space affects the possibility of peer academic communication.
To further explore the effect of spatial layout by separated classrooms of an organization, density by groups is conducted.The results of four cases (AD1pre, AD1-post, AD2-post, La-post) report on the obvious preference for choosing academic communication with peers in the same classroom.As Table 6 shows, density values within the same classrooms (in    grey blocks) are much higher than interclassroom density, even if they belong to the same class organization.Therefore, dividing first-year university students who belong to a class into separate classrooms is not conducive to building peer academic support relationships.

Personal scale
The analysis on a personal scale aims to test whether dyadic spatial proximity affects peer academic support relationships.Organizational proximity is analysed as a reference.The analysis is conducted by the QAP correlation analysis in UCINET 6, which runs the correlation analysis of different relational matrices with a randomization test.The results are summarized in Table 7.

Dyadic physical proximity
The results show that nine of the ten cases report a negative and statistically significant correlation between physical proximity and academic communication relationships of which the correlation coefficients range from −0.494 to −0.131.The results of organizational proximity perform better.All ten cases show a positive and statistically significant correlation with academic communication relationships of which the correlation coefficients range from 0.130 to 0.345.By comparing the correlation coefficient of pre and posttreatment (Figure 7), it is found that the absolute values of physical proximity in four of five classes (AD1, AD2, UP2, LA) increase, but that of organizational proximity shows a decreasing trend.In contrast, the absolute values of physical proximity in UP1 decrease significantly and those of organizational proximity increase slightly.The opposite trend in physical proximity and organizational proximity indicates that physical proximity can affect peer academic communication relationships by reducing the dependence on organizational proximity to some extent, which is beneficial for the diversity of peer academic support relationships.
Regarding academic collaboration, the results report a negative and statistically significant correlation with physical proximity (correlation coefficients range from −0.088 to −0.293) and a positive and statistically significant correlation with organizational proximity (correlation coefficients range from 0.062 to 0.674).The correlation effects of organizational proximity in all five classes are significantly higher than those of physical proximity in the first semester.However, the effect of  organizational proximity decreases dramatically after the move, and the absolute values become even lower than physical proximity in two classes (AD2 and LA).The students in these two classes are divided into separate classrooms, and the effect of physical proximity increases (Figure 8).In other words, the distance brought by separated classrooms segregates academic collaboration.In sum, physical proximity affects peer academic collaboration in first-year university students to some extent.However, their academic collaboration depends more on organizational proximity.The effect of physical proximity is especially obvious in the context of separated classrooms.
The comparison between academic support relationships (communication and academic collaboration) and physical or organizational proximity is further discussed.As Figures 9 and 10 show, academic communication has a stronger correlation with physical proximity, while academic collaboration depends more on organizational proximity.

Dyadic visual proximity
The results show that eight of the ten cases report a negative and statistically significant correlation between visual proximity and academic  communication of which the correlation coefficients range from −0.073 to −0.416.Regarding academic collaboration, only six of the ten cases report a negative and statistically significant correlation with visual proximity of which the correlation coefficients range from −0.087 to −0.270.
Visual proximity shows high similarity with physical proximity but has poorer performance as a spatial proximity predictor for academic support relationships (Figures 11 and 12).
Based on the results of the QAP correlation analysis, we further conduct the QAP regression analysis on the academic communication relationships by metric distance and organizational proximity (Table 8).Eight of the ten cases prove metric distance to be a significant independent variable of which the standard coefficient runs from −0.112363 to −0.465666.The comparison with organizational proximity is similar to the above correlation analysis.

Discussion
The current study explored whether and how the spatial layout of the learning space is relevant for building peer academic support relationships in first-year university students.Spatial configuration as a global measure and spatial proximity as a local measure have been examined as predictive measures of the spatial layout effect on social relationships in the workplace.Previous studies on learning space have proven the effect of spatial layout on academic performance by comparing traditional classrooms and active learning classrooms, which detail different spatial layouts by typology.Since a multimethod approach would be beneficial to complement the research approaches already used, this research focused on quantitative methods.This study therefore adopts spatial configuration and spatial proximity as quantitative measures to further explore the spatial layout effect on peer academic support relationships.An exploratory quasiexperimental approach was conducted with a case using the learning spaces of the school of architecture at SCUT.The research is conducted by answering how spatial configuration affects peer academic support networks on an organizational scale and how dyadic spatial proximity affects personal academic support relationships on a personal scale.
The findings of the case study on an organizational scale show that the spatial configuration of the learning space affects peer academic communication relationships in terms of frequency (a measure of how often communication happens) and density (a measure of the number of communicating students).A segmented spatial layout (lower NAIN) leads to lower frequencies of peer academic communication and network density.Different floors, separated classrooms, and new furniture layouts are factors that affect spatial integration.In addition, the organization of space within the classroom to facilitate circulation should reduce cul-de-sacs, and students' seats have a more positive effect when they are directly placed along paths of major movement, which increases spatial accessibility and thus improves communication possibilities.
The results of the case study on the personal scale report a statistically significant correlation between physical proximity and face-to-face academic communication.The increasing coefficient of physical proximity and decreasing coefficient of organizational proximity of the pre-and posttreatment groups indicate that physical proximity can affect peer relationships based on face-to-face academic communication and reduce the dependence on organizational proximity, which is beneficial for the diversity of peer academic support relationships.The findings suggest that physical proximity also affects peer academic collaboration in first-year university students to some extent.However, their academic collaboration depends more on organizational proximity.The effect of physical proximity on peer academic collaboration is especially obvious in the context of separated classrooms.In other words, the distance brought by separated classrooms significantly segregates dyads who might otherwise have engaged in academic collaboration.Comparing the peer relationships in face-to-face academic communication and academic collaboration, it is found that face-to-face academic communication has a stronger correlation with physical proximity, while academic collaboration depends more on organizational proximity.Therefore, to build a stronger peer academic support network, seat allocation should make use of the promoting effect of physical proximity to bridge students without established organizational relationships.The effect of dyadic visual proximity is examined as well.The results show high similarity with physical proximity but have poorer performance when judged by statistical significance, indicating physical proximity is a better spatial predictor for academic support relationships.

Conclusion
This research bridges disparate disciplines to explore peer academic support relationships that are embedded in a specific spatial milieu.It complements the research on spatial factors in educational settings and on the learning space in the architectural literature.In general, the findings of this study correspond to those of previous research that compares different spatial layout categories and align with previous findings that the spatial layout of learning space affects academic relationships.The findings contribute to a better understanding of how the spatial layout of the learning space affects peer academic support relationships in first-year university students on both the organizational scale and personal scale, using a case study of the learning space at the school of architecture at SCUT.The case study proved that the more integrated the spatial configuration, the higher density of the peer academic support networks.Moreover, spatial proximity was proved as a positive factor for peer academic support networks, which can even reduce the dependence on organizational proximity.As such, the findings of this research can be a preliminary attempt to build a correlation between spatial layout and peer academic support relationships, which may support administrators and architects in planning and managing learning spaces and other issues.
The conclusions of this study remain limited to the case context as a preliminary attempt to quantify the effect of spatial layout on peer academic support relationships in first-year university students.Further research is necessary to extend the research scope to other samples from different schools and cover longer periods to test the findings on their robustness.

Figure 3 .
Figure 3.An example of angular segment analysis of a classroom.The linear network is constructed according to the movement space; the colour is according to the values of different measures.The above figure is an example of angular segment analysis coloured by NAIN (R = n).

Figure 4 .
Figure 4.An example of step depth analysis of a seat within a classroom.The upper figure represents the result of the metric distance (measured by metric step shortest-path length) from a seat (the highlighted yellow grid) to all other grids (including other seats).The lower figure represents the result of the visual distance (measured by number of turns) from a seat (the highlighted yellow grid) to all other grids (including other seats).Different colours represent different values, and the values of each grid can be exported for further statistical analysis.

Figure 5 .
Figure 5. Spatial layout of the five classes in the first semester (distributed on two floors within a building).

Figure 6 .
Figure 6.Spatial layout of the five classes in the second semester (on one floor within a building).

Figure 7 .Figure 8 .
Figure 7. Correlation between pre-and posttreatment academic communication relationships and physical/organizational proximity.

Table 1 .
The composition of respondents.

Table 2 .
Spatial layout properties of pre-and posttreatment (t test).

Table 3 .
Peer academic communication frequency of pre-and posttreatment (paired t test).

Table 4 .
Academic communication network density pre-and posttreatment (bootstrap paired t test).

Table 5 .
Degree centrality of the academic communication network pre-and posttreatment (paired t test).

Table 6 .
Organizational density by separated classrooms.

Table 7 .
Results of QAP correlation.