research paper on relational model

Toward Greater Transparency and Inclusion in Manuscript Review Processes: A Relational Model

Peer review is widely accepted as critical to legitimating scholarly publication, and yet, it runs the risk of reproducing inequities in publishing processes and products. Acknowledging at once the historical need to legitimize SoTL publications, the current danger of reproducing exclusive practices, and the aspirational goal to “practice what we preach” as SoTL practitioners regarding effective feedback to students, we argue for rethinking “rigor,” developing more inclusive practices, and engaging in greater transparency in relation to peer review. To situate our discussion, we revisit foundational work in the development of SoTL and then offer an analytical framework informed by recent scholarship on redefining rigor and the emotional experience of receiving feedback. Using this framework, we propose a relational model of peer review and present two examples of efforts in which we have been involved as founding co-editors of the International Journal for Students as Partners to move toward greater transparency and inclusion in manuscript review processes.

Author Biographies

Alison cook-sather, bryn mawr college.

Alison Cook-Sather (USA) is Mary Katharine Woodworth professor of education at Bryn Mawr College. She has published 10 books and over 160 other works on student-faculty pedagogical partnership, and she has consulted in 13 countries on this practice.

Ruth L. Healey, University of Chester

Ruth L. Healey (GBR) is a professor of learning and teaching and a University Innovation Fellow at the University of Chester and a director of Healey HE Consultants. She was awarded a National Teaching Fellowship in 2017.  

Abbot, Sophia. 2024. “SoTL Citizen: A Memoir of Home and Exile in the Scholarship of Teaching and Learning.” In Becoming a SoTL Scholar, edited by Janice Miller-Young and Nancy Chick. Elon, North Carolina: Elon University Open Access Series. https://www.centerforengagedlearning.org/books/becoming-a-sotl-scholar/section-1/chapter-3/ .

Arumugam, Ashokan, Poonam Mehta, and G. David Baxter. 2020. “Doubleblind Peer Review of Manuscripts: Opportunities, Challenges, and Way Forward”, Physical Therapy Reviews 25 (1): 1–6. https://doi.org/10.1080/10833196.2019.1698161 .

Babin, Barry J., and Julie Guidry Moulard.2018. “To What is the Review Process Relevant? What’s Right and What’s Wrong with Peer Review for Academic Business Journals.” European Business Review 30 (2): 145–56. https://doi.org/10.1108/EBR-09-2017-0162 .

Brookfield, Stephen D. 2017. Becoming a Critically Reflective Teacher. New York: John Wiley and Sons.

Brooks, Jamiella, and Julie McGurk. 2022. “Rigor as Inclusive Practice.” Teaching + Learning Lab, October 6, 2022. https://tll.mit.edu/rigor-as-inclusive-practice/ .

Coalter, Jodi H. 2023. “Citation Power: Overcoming Marginalization One Citation at a Time.” In Perspectives on Justice, Equity, Diversity, and Inclusion in Libraries, edited by Nandita S. Mani, Michelle A. Cawley, Emily P. Jones, 62–77. Hershey, PA: IGI Global.

Chick, Nancy. 2024. “‘Dear Author’: A Transparent SoTL Peer Review.” Transformative Dialogues: Teaching and Learning Journal 17 (1). https://doi.org/10.26209/td2024vol17iss11800 .

Chun-Man Ho, Roger, Kwok-Kei Mak, Ren Tao, Yanxia Lu, Jeffrey R. Day, and Fang Pan. 2013. “Views on the Peer Review System of Biomedical Journals: An Online Survey of Academics from High-Ranking Universities.” BMC Medical Research Methodology 13 (74): 1–15. http://www.biomedcentral.com/1471-2288/13/74 .

Cliffe, Anthony, Alison Cook-Sather, Mick Healey, Ruth Healey, Beth Marquis, Kelly E. Matthews, Lucy Mercer-Mapstone, Anit Ntem, Varun Puri, and Cherie Woolmer. 2017. “Launching a Journal about and through Students as Partners.” International Journal for Students as Partners 1 (1). https://mulpress.mcmaster.ca/ijsap/issue/view/306 .

Cook-Sather, Alison. 2024. “Rethinking Time in Preparing for and Reflecting on Teaching: Pedagogical Partnership with Student Consultants as Empowering Educational Development.” To Improve the Academy 43 (1). https://doi.org/10.3998/tia.4793 .

Cook-Sather, Alison, Sophia Abbot, and Peter Felten. 2019. “Legitimizing Reflective Writing in SoTL: ‘Dysfunctional Illusions of Rigor’ Revisited.” Teaching & Learning Inquiry 7 (2): 14–27. https://doi.org/10.20343/teachlearninqu.7.2.2 .

Cook-Sather, Alison, Catherine Bovill, and Peter Felten. 2014. Engaging Students as Partners in Learning & Teaching: A Guide for Faculty. San Francisco: Jossey-Bass.

Cook-Sather, Alison, and Alice Lesnick. 2023. “Engaged Voices, Engaged Beings: Exploring Tensions, Transgressions, and Risks in Using Co-Created, Shared Google Docs to Build Classroom Community and to Co-Author.” The Journal of Interactive Technology and Pedagogy 23. https://cuny.manifoldapp.org/projects/jitp-23 .

Cook-Sather, Alison, Kelly E. Matthews, and Mick Healey. 2020. “Writing about Learning and Teaching in Precarious and Promising Times.” AAC&U Blog post. https://www.aacu.org/liberaleducation/articles/writing-about-learning-and-teaching-in-precarious-and-promising-times .

Cook-Sather, Alison, and Van Nguyen. 2023. “Using Shared Google Docs to Co-Create Life-Affirming Learning: A Case of Trauma-Informed Instructional Design at the Tertiary Level.” The Journal of Applied Instructional Design 12 (1). https://dx.doi.org/10.51869/121/acsvn .

Cruz, Laura, Eileen Grodziak, and Hillary H. Steiner. 2024. “Practice What We Preach?: A Review of Journal Publishing Practices Related to Reflective Writing in SoTL.” International Journal for the Scholarship of Teaching and Learning 18 (1). https://doi.org/10.20429/ijsotl.2024.180105 .

Enslin, Penny, and Nicki Hedge. 2018. “On Peer Review as the ‘Gold Standard’ in Measuring Research Excellence: From Secrecy to Openness?” Journal of Philosophy of Education 52 (3): 379–96. https://doi.org/10.1111/1467-9752.12312 .

Felten, Peter. 2013. “Principles of Good Practice in SoTL.” Teaching & Learning Inquiry: The ISSOTL Journal 1 (1): 121–25. https://doi.org/10.2979/teachlearninqu.1.1.121 .

Felten, Peter, Leigh Z. Gilchrist, and Alexa Darby. 2006. “Emotion and Learning: Feeling our Way Toward a New Theory of Reflection in Service-Learning.” Michigan Journal of Community Service Learning 12: 38–46. http://hdl.handle.net/2027/spo.3239521.0012.204 .

Friberg, Jennifer, Mandy Frake-Mistak, Ruth L. Healey, Shannon Sipes, Julie Mooney, Stephanie Sanchez, and Karena Waller. 2021. “A Developmental Framework for Mentorship in SoTL Illustrated by Three Examples of Unseen Opportunities for Mentoring.” Teaching & Learning Inquiry 9 (1): 395–413. https://doi.org/10.20343/teachlearninqu.9.1.26 .

Garrido-Gallego, Yeimy. 2018. “Open Peer Review for Evaluating Academic Legal Publications: The ‘Antidote’ to an ‘Ill’ Blind Peer Review?” Tilburg Law Review 23 (1): 77–90. https://doi.org/10.5334/tilr.128 .

George, Elizabeth. 2023. “Peer Review Basics: Who is Reviewer 2?” Researcher.Life, September 29, 2023. https://researcher.life/blog/article/peer-review-basics-who-is-reviewer-2/#:~:text=Peer%20reviewers%20are%20expected%20to,reviewer%20among%20a%20manuscript's%20evaluators .

Google Oxford Languages. 2024. “Rigour.” Accessed April 2, 2024. https://www.google.com/search?q=definition+of+rigour&rlz=1C1ONGR_en-GBGB991GB991&oq=definition+o&gs_lcrp=EgZjaHJvbWUqCAgAEEUYOBg7MggIABBFGDgYOzIGCAEQRRg5MgoIAhAAGLEDGIAEMgcIAxAAGIAEMgoIBBAAGLEDGIAEMgcIBRAAGIAEMgcIBhAAGIAEMgoIBxAAGLEDGIAEMg0ICBAAGIMBGLEDGIAEMgcICRAAGI8C0gEIMjIzMGowajSoAgCwAgA&sourceid=chrome&ie=UTF-8 .

Healey, Mick, Alison Cook-Sather, and Kelly E. Matthews. 2020. “Expanding the Conventional Writing Genres: A Matter of Equity and Inclusion.” Educational Developments 21 (3): 8–11.

Healey, Ruth L., Mick Healey, and Anthony Cliffe. 2018. “Engaging in Radical Work: Students as Partners in Academic Publishing.” Efficiency Exchange (Universities UK and Jisc in partnership with Hefce and the Leadership Foundation).

Hill, Jennifer, Kathy Berlin, Julia Choate, Lisa Cravens-Brown, Lisa McKendrick-Calder, and Susan Smith. 2021a. “Can Relational Feed-Forward Enhance Students’ Cognitive and Affective Responses to Assessment?” Teaching & Learning Inquiry 9 (2). https://doi.org/10.20343/teachlearninqu.9.2.18 .

Hill, Jennifer, Ruth L. Healey, Harry West, and Chantal Déry. 2021b. “Pedagogic Partnership in Higher Education: Encountering Emotion in Learning and Enhancing Student Wellbeing.” Journal of Geography in Higher Education 45 (2): 167–85. https://doi.org/10.1080/03098265.2019.1661366 .

Hill, Jennifer, Kathy Berlin, Julia Choate, Lisa Cravens-Brown, Lisa McKendrick-Calder, and Susan Smith. 2023. “Emotions Experienced by Instructors Delivering Written Feedback and Dialogic Feed-Forward.” Teaching & Learning Inquiry 11. https://doi.org/10.20343/teachlearninqu.11.6 .

Hutchings, Pat. 2000. Opening Lines: Approaching the Scholarship of Teaching and Learning. Carnegie Foundation for the Advancement of Teaching.

International Journal for Students as Partners. https://mulpress.mcmaster.ca/ijsap .

Irfanullah, Haseeb. 2021. “Let’s Talk About the Volunteers in Scholarly Publishing.” The Scholarly Kitchen, July 21, 2021. https://scholarlykitchen.sspnet.org/2021/07/21/lets-talk-about-the-volunteers-in-scholarly-publishing/ .

Kaur, Amrita. Forthcoming. “Developing an Open and Dialogic Review of Reflective Essays: An Empathic and Inclusive Approach to Advance Students-as-Partners Practices.” International Journal for Students as Partners.

Masika, Rachel, and Jennie Jones. 2016. “Building Student Belonging and Engagement: Insights into Higher Education Students’ Experiences of Participating and Learning Together.” Teaching in Higher Education 21(2): 138–50. https://doi.org/10.1080/13562517.2015.1122585 .

Matthews, Kelly, E., Catherine Sherwood, Eimear Enright, and Alison Cook-Sather. 2024. “What Do Students and Teachers Talk about When They Talk Together about Feedback and Assessment? Expanding Notions of Feedback Literacy through Pedagogical Partnership.” Assessment & Evaluation in Higher Education 49 (1): 26–38. https://doi.org/10.1080/02602938.2023.2170977 .

McDonald, Jacquie, and Aileen Cater-Steel, editors. 2017. Communities of Practice: Facilitating Social Learning in Higher Education. Singapore: Springer.

Nelson, Craig. 2010. “Dysfunctional Illusions of Rigor: Lessons from the Scholarship of Teaching and Learning.” To Improve the Academy: A Journal of Educational Development 28, 177–92.

Ollis, Lucie, and Karen Gravett. 2020. “The Emerging Landscape of Student–Staff Partnerships in Higher Education.” In Enhancing Student-Centred Teaching in Higher Education, edited by Karen Gravett, Nadya Yakovchuk, and Ian M. Kinchin, 11–27. Cham: Palgrave Macmillan. https://doi.org/10.1007/978-3-030-35396-4_2 .

Overall, Jeffrey. 2015. “Stop Drinking the Kool-Aid: The Academic Journal Review Process in the Social Sciences Is Broken, Let’s Fix It.” Journal of Academic Ethics 13: 277–89. https://doi.org/10.1007/s10805-015-9237-3 .

PAEditorial. 2021. “Double-Blind Peer Review: Pros, Cons and Everything In-Between.” Last modified October 29, 2021. https://paeditorial.co.uk/post/double-blind-peer-review-the-pros-the-cons-and-everything-in-between/#:~:text=The%20anonymity%20allows%20the%20reviewer,%2C%20gender%2C%20and%20personal%20bias .

Pekrun, Reinhard. 2019. “Inquiry on Emotions in Higher Education: Progress and Open Problems.” Studies in Higher Education 44 (10): 1806–11. https://doi.org/10.1080/03075079.2019.1665335 .

Rowe, Anna. D. 2017. “Feelings about Feedback: The Role of Emotions in Assessment for Learning.” In Scaling up Assessment for Learning in Higher Education. The Enabling Power of Assessment, edited by David Carless, Susan M. Bridges, Cecilia Ka Yuk Chan, and Rick Glofcheski, 159–172. Singapore: Springer.

Ryan, Tracii, and Michael Henderson. 2018. “Feeling Feedback: Students’ Emotional Responses to Educator Feedback.” Assessment & Evaluation in Higher Education 43 (6): 880–92. https://doi.org/10.1080/02602938.2017.1416456 .

Shulman, Lee. 2001. “From Minsk to Pinsk: Why a Scholarship of Teaching and Learning?” Journal of the Scholarship of Teaching and Learning 1(1): 48–53. https://scholarworks.iu.edu/journals/index.php/josotl/article/view/1582 .

Taylor and Francis. 2023. “What are the Different Types of Peer Review?” Accessed April 2, 2024: https://authorservices.taylorandfrancis.com/publishing-your-research/peer-review/types-peer-review/#:~:text=Single%2Danonymous%20peer%20review&text=This%20is%20the%20most%20common,will%20hold%20this%20against%20them .

Tennant, Jonathan P., and Tony Ross-Hellauer. 2020. “The Limitations to Our Understanding of Peer Review.” Research Integrity and Peer Review 5 (1): 6. https://researchintegrityjournal.biomedcentral.com/articles/10.1186/s41073-020-00092-1 .

West, Richard E, and Peter J. Rich. 2012. “Rigor, Impact and Prestige: A Proposed Framework for Evaluating Scholarly Publications.” Innovative Higher Education 37 (5): 359–71. https://link.springer.com/article/10.1007/s10755-012-9214-3 .

Winstone, Naomi E., Robert A. Nash, James Rowntree, and Michael Parker. 2017. “‘It’d Be Useful, but I Wouldn’t Use It’: Barriers to University Students’ Feedback Seeking and Recipience.” Studies in Higher Education 42 (11): 2026–41. https://doi.org/10.1080/03075079.2015.1130032 .

Yahlnaaw. 2019. “T’aats’iigang – Stuffing a Jar Full.” International Journal for Students as Partners 3 (2): 6–10. https://doi.org/10.15173/ijsap.v3i2.4081 .

How to Cite

• Endnote/Zotero/Mendeley (RIS)

Copyright (c) 2024 Alison Cook-Sather, Ruth L. Healey

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License .

OJS hosted by the University of Calgary .

Two-Bit History

Computing through the ages

Important Papers: Codd and the Relational Model

29 Dec 2017

It’s hard to believe today, but the relational database was once the cool new kid on the block. In 2017, the relational model competes with all sorts of cutting-edge NoSQL technologies that make relational database systems seem old-fashioned and boring. Yet, 50 years ago, none of the dominant database systems were relational. Nobody had thought to structure their data that way. When the relational model did come along, it was a radical new idea that revolutionized the database world and spawned a multi-billion dollar industry.

The relational model was introduced in 1970. Edgar F. Codd, a researcher at IBM, published a paper called “A Relational Model of Data for Large Shared Data Banks.” The paper was a rewrite of a paper he had circulated internally at IBM a year earlier. The paper is unassuming; Codd does not announce in his abstract that he has discovered a brilliant new approach to storing data. He only claims to have employed a novel tool (the mathematical notion of a “relation”) to address some of the inadequacies of the prevailing database models.

In 1970, there were two schools of thought about how to structure a database: the hierarchical model and the network model. The hierarchical model was used by IBM’s Information Management System (IMS), the dominant database system at the time. The network model had been specified by a standards committee called CODASYL (which also—random tidbit—specified COBOL) and implemented by several other database system vendors. The two models were not really that different; both could be called “navigational” models. They persisted tree or graph data structures to disk using pointers to preserve the links between the data. Retrieving a record stored toward the bottom of the tree would involve first navigating through all of its ancestor records. These databases were fast (IMS is still used by many financial institutions partly for this reason, see this excellent blog post ) but inflexible. Woe unto those database administrators who suddenly found themselves needing to query records from the bottom of the tree without having an obvious place to start at the top.

Codd saw this inflexibility as a symptom of a larger problem. Programs using a hierarchical or network database had to know about how the stored data was structured. Programs had to know this because they were responsible for navigating down this structure to find the information they needed. This was so true that when Charles Bachman, a major pioneer of the network model, received a Turing Award for his work in 1973, he gave a speech titled “ The Programmer as Navigator .” Of course, if programs were saddled with this responsibility, then they would immediately break if the structure of the database ever changed. In the introduction to his 1970 paper, Codd motivates the search for a better model by arguing that we need “data independence,” which he defines as “the independence of application programs and terminal activities from growth in data types and changes in data representation.” The relational model, he argues, “appears to be superior in several respects to the graph or network model presently in vogue,” partly because, among other benefits, the relational model “provides a means of describing data with its natural structure only.” By this he meant that programs could safely ignore any artificial structures (like trees) imposed upon the data for storage and retrieval purposes only.

To further illustrate the problem with the navigational models, Codd devotes the first section of his paper to an example data set involving machine parts and assembly projects. This dataset, he says, could be represented in existing systems in at least five different ways. Any program \(P\) that is developed assuming one of five structures will fail when run against at least three of the other structures. The program \(P\) could instead try to figure out ahead of time which of the structures it might be dealing with, but it would be difficult to do so in this specific case and practically impossible in the general case. So, as long as the program needs to know about how the data is structured, we cannot switch to an alternative structure without breaking the program. This is a real bummer because (and this is from the abstract) “changes in data representation will often be needed as a result of changes in query, update, and report traffic and natural growth in the types of stored information.”

Codd then introduces his relational model. This model would be refined and expanded in subsequent papers: In 1971, Codd wrote about ALPHA, a SQL-like query language he created; in another 1971 paper, he introduced the first three normal forms we know and love today; and in 1972, he further developed relational algebra and relational calculus, the mathematically rigorous underpinnings of the relational model. But Codd’s 1970 paper contains the kernel of the relational idea:

The term relation is used here in its accepted mathematical sense. Given sets \(S_1, S_i, ..., S_n\) (not necessarily distinct), \(R\) is a relation on these \(n\) sets if it is a set of \(n\)-tuples each of which has its first element from \(S_1\), its second element from \(S_2\), and so on. We shall refer to \(S_j\) as the \(j\)th domain of \(R\). As defined above, \(R\) is said to have degree \(n\). Relations of degree 1 are often called unary , degree 2 binary , degree 3 ternary , and degree \(n\) n-ary .

Today, we call a relation a table , and a domain an attribute or a column . The word “table” actually appears nowhere in the paper, though Codd’s visual representations of relations (which he calls “arrays”) do resemble tables. Codd defines several more terms, some of which we continue to use and others we have replaced. He explains primary and foreign keys, as well as what he calls the “active domain,” which is the set of all distinct values that actually appear in a given domain or column. He then spends some time distinguishing between a “simple” and a “nonsimple” domain. A simple domain contains “atomic” or “nondecomposable” values, like integers. A nonsimple domain has relations as elements. The example Codd gives here is that of an employee with a salary history. The salary history is not one salary but a collection of salaries each associated with a date. So a salary history cannot be represented by a single number or string.

It’s not obvious how one could store a nonsimple domain in a multi-dimensional array, AKA a table. The temptation might be to denote the nonsimple relationship using some kind of pointer, but then we would be repeating the mistakes of the navigational models. Instead. Codd introduces normalization, which at least in the 1970 paper involves nothing more than turning nonsimple domains into simple ones. This is done by expanding the child relation so that it includes the primary key of the parent. Each tuple of the child relation references its parent using simple domains, eliminating the need for a nonsimple domain in the parent. Normalization means no pointers, sidestepping all the problems they cause in the navigational models.

At this point, anyone reading Codd’s paper would have several questions, such as “Okay, how would I actually query such a system?” Codd mentions the possibility of creating a universal sublanguage for querying relational databases from other programs, but declines to define such a language in this particular paper. He does explain, in mathematical terms, many of the fundamental operations such a language would have to support, like joins, “projection” ( SELECT in SQL), and “restriction” ( WHERE ). The amazing thing about Codd’s 1970 paper is that, really, all the ideas are there—we’ve been writing SELECT statements and joins for almost half a century now.

Codd wraps up the paper by discussing ways in which a normalized relational database, on top of its other benefits, can reduce redundancy and improve consistency in data storage. Altogether, the paper is only 11 pages long and not that difficult of a read. I encourage you to look through it yourself. It would be another ten years before Codd’s ideas were properly implemented in a functioning system, but, when they finally were, those systems were so obviously better than previous systems that they took the world by storm.

If you enjoyed this post, more like it come out every four weeks! Follow @TwoBitHistory on Twitter or subscribe to the RSS feed to make sure you know when a new post is out.

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Relational models theory: Validation and replication for four fundamental relationships

Roles Conceptualization, Formal analysis, Software, Writing – original draft, Writing – review & editing

* E-mail: [email protected]

Affiliation Department of Psychology, University of Edinburgh, Edinburgh, Scotland, United Kingdom

Roles Conceptualization, Data curation, Project administration, Supervision, Writing – review & editing

• Michael Zakharin, 
• Timothy C. Bates

• Published: June 16, 2023
• https://doi.org/10.1371/journal.pone.0287391
• Peer Review
• Reader Comments

Relational models theory predicts that social relationships are formed from four underlying psychological models: communal sharing, authority ranking, equality matching, and market pricing. Here, in four studies, we test this four-factor model using the 33-item Modes of Relationships Questionnaire (MORQ). In Study 1, we administered the MORQ to N = 347 subjects. A parallel analysis supported the four-factor structure, but several items failed to load on their predicted target factors. In Study 2 (N = 617), we developed a well-fitting four-factor model of the MORQ with a total of 20 items (five items retained for each factor). This model replicated across multiple relationships reported by each subject. In Study 3, we replicated the model in an independent dataset (N = 615). A general factor associated with relationship type was required in both Study 2 and Study 3. In Study 4, we tested the nature of this general factor, finding that it was associated with the closeness of the relationship. The results support the Relational Models four-factor structure of social relationships. Given the mature theory and applications in a wide range of disciplines, from social to organisational psychology, we hope that this compact, valid, and interpretable instrument leads to increased usage of the scale.

Citation: Zakharin M, Bates TC (2023) Relational models theory: Validation and replication for four fundamental relationships. PLoS ONE 18(6): e0287391. https://doi.org/10.1371/journal.pone.0287391

Editor: Srebrenka Letina, University of Glasgow, UNITED KINGDOM

Received: December 19, 2022; Accepted: June 5, 2023; Published: June 16, 2023

Copyright: © 2023 Zakharin, Bates. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data and materials used in this paper as well as R code used to generate the results are openly available at the OSF site for this paper at https://osf.io/3ypxu/ , DOI 10.17605/OSF.IO/3YPXU .

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Relational models theory (RMT) offers a comprehensive model of interpersonal relationships [ 1 , 2 ]. The theory proposes that relationships are represented and processed within four underlying psychological models: Communal Sharing, Equality Matching, Authority Ranking and Market Pricing. A measure of these models–the Modes of Relationships Questionnaire (MORQ) [ 3 ]–has been developed, permitting testing of the theory. Analyses of the MORQ, however, have found a poor fit to the theorised four-factor model [ 3 – 5 ]. In the present paper, we set out to locate the cause of this poor fit, generate a well-fitting model, and establish the replicability of the newly proposed model. Below, we briefly introduce the RMT and the questionnaire associated with it.

Based on ethnographic fieldwork and a review of previous studies, Fiske [ 1 ] proposed four distinct relational systems constituting the structures of social relationships. These four models are theorised as fundamental and innate and serve as a comprehensive framework to describe all possible human relationships [ 6 ]. They depict how individuals evaluate their status in relation to others and elucidate appropriate or inappropriate behaviours in a given social context. In essence, they offer a framework for comprehending social interactions and the expected norms of behaviour in diverse social settings. The first of these models, Communal Sharing (CS), focuses on what people have in common and is exemplified in relationships where people share an identity with others, such as family, tribe, religion, or ethnic group, resulting in mutual recognition of social equivalence of individuals. This shared identity is reflected in helping others regardless of their past contributions, treating the property as communal, and making joint consensus-based decisions. The second relational model is Equality Matching (EM), in which individuals treat each other as distinct but equal partners. In EM relationships, work inputs and outputs are divided equally where possible. Where resources and work are not divisible equally, individuals keep a count of what they give and receive and equalise this over time. Examples of this relationship include mutual credit organisations and babysitting co-ops. EM also extends to vengeful behaviour, such as eye-for-an-eye justice [ 7 ]. The third model, Authority Ranking (AR), implements a hierarchy system in which social interactions are based on recognising and respecting different levels of authority. The distribution of resources in this model is expected to be unequal, with superiors feeling entitled to a larger share of resources and subordinates accepting this division as fair [ 8 ]. A range of factors can influence ranking in an AR, including age, gender, seniority, and achievement. One example of this model would be the relationship between employer and employee. The fourth and final relational model in RMT is Market Pricing (MP). The MP model suggests that people relate to each other based on the value they exchange in a relationship as if it were a market transaction. According to this model, individuals perceive their relationships as a means to obtain desired resources, assistance, or support from the other person. Examples of relationships that align with this model include commercial partnerships, where transactions are prominent, as well as cultural constructs like the concepts of price, wages, or dividends.

While these four models are conceptualised as distinct, RMT predicts that a given human relationship typically reflects combinations of two or more relational models. For example, relationships within a family context usually emphasise the CS relationship. However, children within a family may also be expected to respect their parents (AR relationship), do their fair share of chores (EM relationship) and, perhaps, to be paid for doing some of them (MP relationship).

A substantial amount of empirical research has demonstrated that relational models can accurately predict significant outcomes. For instance, Vodosek [ 9 ] found that horizontal collectivism was associated with equality matching and communal sharing relationships, whereas vertical individualism was related to a preference for authority ranking, and vertical collectivism was related to a preference for authority ranking and communal sharing. Biber et al. [ 10 ] investigated the relationship between relational models and universal human values [ 11 ]. They found that individuals who prioritise CS relationships place greater importance on benevolence and universalism values while placing less emphasis on power and achievement. Conversely, those who value AR or MP relationships tend to prioritise power and achievement values over benevolence and universalism. A disparity between anticipated and real relationship models resulted in a sense of inequity among employees at work [ 12 , 13 ], and they began to view their supervisors as lacking morals [ 8 ]. In clinical samples, different diagnoses were linked to either difficulties or extreme use of specific relational models [ 14 , 15 ]. For instance, dysthymia was found to be positively associated with high levels of AR relationships with close friends and family members, while hypomania was positively associated with high levels of CS and EM relationships with authority figures.

Updates and applications of RMT

Other models of social relations have been developed both before and since the innovations of RMT. Perhaps the key feature distinguishing RMT from theories of social relationships, such as interdependence theory [ 16 ], attachment theory [ 17 ] and social identity theory [ 18 ] is the emphasis RMT places on explaining the underlying structure of relationships. Rather than focusing on the role of interdependence within relationships, emotional bonds formed early in life or the sense of self derived from a social group membership, RMT provides a framework for understanding social interactions and the appropriate behaviours within them.

RMT has continued to evolve and expand its realm of application, with several changes being of particular relevance. First, a personality assessment tool–the Relationship Profile Scale [ 15 ], was developed to evaluate individual preferences for distinct relational models, measuring the perceived importance, satisfaction, challenges, and motivations associated with each of the four relational models. Together with the MORQ, the Relationship Profile Scale enables a comparison of individuals’ desired and actual relationship experiences. A significant theoretical advance known as Relationship Regulation Theory [ 19 ] extended RMT into the domain of moral psychology by associating each relational model with four distinct moral motives. For instance, the moral motive of hierarchy is based on the AR relationship and its focus on establishing and upholding a clear ranking in social groups. The motive of hierarchy motivates those in lower positions to show respect, obedience, and deference to those above them, including leaders, ancestors, or gods, and to punish those who go against them. Conversely, those in higher positions feel a moral responsibility to guide, direct, and safeguard those below them. This expansion links RMT to existing moral theories [ 20 , 21 ] but construes the nature of moral behaviour as relationship management and emphasises that the moral value of acts such as harming, unequal treatment, or being impure are dependent on the relationships and relational models within which they are deployed.

Most recently, RMT has undergone another significant enhancement by incorporating the well-established effects of incentives on behaviour into our understanding of relationships and relationship management. Known as Relational Incentives Theory [ 22 ], this extension posits that for incentives to be effective, they should align with relational models. For instance, incentives promoting communal sharing relations should be most effective when they align with the motive of unity, while proportional incentive schemes work best for market pricing relations. These recent advancements demonstrate the continued significance of RMT and highlight the crucial role of the four relational models in comprehending and predicting diverse behaviours, ranging from resolving moral disagreements to determining the efficacy of incentive schemes.

Measuring relational models

Realising the benefits of an instrument to test the predictions of RMT, Haslam and Fiske [ 3 ] developed the Modes of Relationships Questionnaire (MORQ), a 33-item instrument to measure the four social relationships specified in RMT. For each relationship, items were constructed to tap into each of eight classes of behaviour predicted to be influenced by social relationships: 1) distribution and use of resources, 2) work, 3) morals, 4) exchange, 5) decision-making, 6) social influence, and 7) identity, with an eighth “miscellaneous” category reserved for behaviours specific to each particular relationship mode. The EM relationship has two items in this miscellaneous category.

The MORQ has a slightly unusual administration process. Participants first generate a list of relationships they have with others, typically 40, from which 10 are selected randomly to avoid oversampling easier-to-recall relationships. They then rate each relationship on each of the 33 MORQ items. This creates data in which information from each participant generates correlated information on multiple target individuals. Between-participant variance is statistically removed to compensate for the dependency amongst responses so that the data analysed consisted of each participant’s deviation from their own mean rating across all items.

During the initial study, Haslam and Fiske [ 3 ] administered the questionnaire to 42 participants. Three theorised models were compared (see Fig 1 ): a) A two-factor model consisting of orthogonal bipolar dimensions, one running from EM to AR and one from CS to MP, thus capturing the equality-inequality and closeness-distance dimensions; b) A four-factor orthogonal and c) A four-factor oblique model. Confirmatory factor analyses preferred the four oblique factor model over the two other models [ 3 ]. However, the absolute fit of this four-factor model was well below the accepted criteria (GFI = .75, RMSEA = .243).

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

a) 2-factor Bipolar Model, b) 4-factor Orthogonal Model and c) 4-factor Oblique Model.

https://doi.org/10.1371/journal.pone.0287391.g001

Since this initial report, only a few studies have assessed the psychometric properties of the MORQ. Brito and colleagues [ 4 ] evaluated the four-factor structure of the MORQ in a sample of 63 Portuguese participants confirming that the fit metrics of the model were unsatisfactory (GFI = .74, RMSEA = .092). Vodosek [ 9 ] selected a different 22-item set from the 33-item MORQ, administered these to a US sample (N = 465), and tested the fit of a four-factor model. However, even this reduced set of items did not fit a four-factor model well (CFI = .80, GFI = .83, RMSEA = .07). In this study, however, participants were asked to indicate the degree to which they believed each MORQ statement should be true in an ideal working group, rather than rating their actual relationships. This difference in approach may account for the low level of fit observed. Finally, Bogodistov and Lizneva (5) used the MORQ with a Ukrainian sample of 99 metallurgical workers. They modelled three of the four scales, excluding EM (based on poor reliability) and also excluded 11 items from the remaining three scales based on low factor loadings, leaving a total of six items on the AR scale, four items on the CS scale and three items on the MP scale. A structural model of these items representing only 13 items and three factors showed a good model fit (CFI = .963, RMSEA = .058).

These previous attempts to model the MORQ suggest that some items have low validity (based on low factor loadings). They were also hampered by small samples, lacking the power to detect item structure reliably. Given the lack of fit for the simple four-factor models tested, it may also be that more complex structural models are needed to account for variance in the MORQ. Previous research, therefore, cannot be interpreted as rejecting the RMT but rather suggests the need for additional research and modelling. To advance the literature, we conducted four studies addressing these shortcomings.

In Study 1a, we used structural equation modelling to test the fit of the four-factor model (see Fig 1C ) proposed by Haslam and Fiske [ 3 ] in a large sample. Confirming that the model fits poorly, we then attempted to improve the model in a smaller set of items, retained based on high factor loadings suggested by factor analysis.

Participants.

A total of 347 United Kingdom residents (228 women, 118 men, 1 other; mean age 33.96 years, SD = 13.44) were recruited using Prolific Academic, an online research-recruiting system. The data were collected in April-May 2021. The study was approved by the University of Edinburgh PPLS Research Ethics Committee.

Relational models were assessed using the MORQ [ 3 ]. This instrument assesses the CS model (eight items, e.g., “ If either of you needs something , the other gives it without expecting anything in return ”); EM (nine items, e.g., “ If you have work to do , you usually split it evenly ”); AR (eight items, e.g., “ One of you is entitled to more than the other ”); and MP (eight items, e.g., “ What you get from this person is directly proportional to how much you give them ”).

Testing was done using the Qualtrics online survey platform. Before starting the study, participants received an explanation of the study and were asked to provide written consent by signing a consent form. After giving informed consent, each participant was asked to identify 40 people with whom they interacted at any closeness level, regardless of how superficially or infrequently, giving a memorable name for each. An automated branching logic in the questionnaire randomly selected one of these names, and the subject was prompted to complete the MORQ, rating this selected relationship. In order to increase the sample size and, consequently, the reliability of the study, each participant was asked to identify 40 relationships but to rate only one of these relationships instead of ten as in the original study [ 3 ]. Total testing took approximately 9 minutes per participant on average. All data were de-identified and collected using Prolific IDs to protect participants’ privacy. No personally identifying information was collected and the authors did not have access to information that could identify individual participants during or after data collection. For privacy, Prolific IDs have been anonymised and replaced with numerical IDs in the open data associated with this manuscript.

Model fit was assessed using the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and the root mean square error of approximation (RMSEA). The RMSEA evaluates the deviation of a hypothesised model from an ideal one. It ranges between 0 and 1, with values closer to zero indicating a better fit. In contrast, the CFI and TLI compare the fit of a hypothesised model to that of a baseline model, which assumes no correlation between any underlying continuous variables. Higher values, closer to 1.0, indicate a better fit for CFI and TLI. Following Hu & Bentler [ 23 ] and Yu [ 24 ], we adopted criteria of TLI and CFI > = .95 and RMSEA < = .06. The comparative fit of the models was assessed by the Akaike Information Criterion (AIC) [ 25 ], which penalises un-parsimonious models. All statistical analyses were completed in R [ 26 ] and umx [ 27 ].

Descriptive statistics and Cronbach’s alpha coefficients for the four relational models are given in Table 1 . Cronbach’s alphas ranged from 0.72 to 0.86 suggesting good internal consistency of the four scales.

https://doi.org/10.1371/journal.pone.0287391.t001

We first tested the best-fitting model of the MORQ presented by Haslam and Fiske [ 3 ], consisting of four factors with items loading only on their corresponding factor and the factors themselves permitted to correlate. This model had unsatisfactory fit, χ 2 (489) = 1563.59, p < 0.001; CFI = 0.738; TLI = 0.717; RMSEA = 0.08.

To explore the cause of this lack of fit, we conducted a parallel analysis [ 28 ] followed by an exploratory factor analysis using a promax (oblique) rotation. The parallel analysis supported a four-factor structure, with the first four factors accounting for 11.3%, 9.8%, 9.1% and 7.7% of the variance in MORQ scores, respectively. The exploratory factor analysis extracting four factors indicated some likely problems. Eight items had a larger loading on a factor other than that they intended to assess. Ten items had cross-loadings over .30, suggesting that they measured more than just one relational model.

Based on this factor analytic evidence and on previous studies indicating that some items in the MORQ loaded poorly on their target factor [ 5 , 9 ], we attempted to create an abbreviated 12-item scale (three items per factor to identify the model). Our selection criteria were high (> .50) loadings on their target factor and low off-factor loadings (< .20). Twelve items meeting these criteria were found which supported a well-fitting model, albeit in the same data set in which they had been discovered (χ 2 (48) = 107.53, p < 0.001; CFI = 0.943; TLI = 0.922; RMSEA = 0.06).

As the analyses of Study 1a were exploratory and therefore prone to yield unreplicable results [ 29 ], we attempted to replicate the final model in an independent dataset. In order to permit control of between-participant variance, we also asked participants in this new sample to rate ten individuals as in Haslam and Fiske [ 3 ], rather than just a single target individual, as we had done in Study 1a.

A total of 135 United Kingdom residents (100 women, 33 men, 2 other; mean age 35.74 years, SD = 14.27) were recruited using Prolific Academic, an online research-recruiting system.

The data were collected in May 2021. The study was approved by the University of Edinburgh PPLS Research Ethics Committee.

Measures and procedure.

Relationships were assessed using 12 MORQ items selected in Study 1a. Testing was done using Qualtrics online survey platform. Before starting the study, participants received an explanation of the study and were asked to provide written consent by signing a consent form. As in Study 1a, each participant identified 40 people with whom they interacted at any level, giving a memorable name to each. An automated branching logic in the questionnaire then randomly selected ten of these names. The subject was then prompted to complete the 12-item version of the MORQ, rating each of the selected relationships. Total testing took approximately 13 minutes per participant on average. All data were de-identified and collected using anonymous codes to protect participants’ privacy. No personal identifying information was collected, and the authors did not have access to any information that could identify individual participants during or after data collection.

Before conducting inferential analyses, following Haslam and Fiske [ 3 ], the impact of reporter-specific variance in the multiple target reports from each subject was controlled. Where Haslam and Fiske accomplished this by dummy coding the participant IDs and residualising the data for these dummy variables, we accomplished the same purpose in a multi-level analysis, with participant ID as a random variable, again retaining the unstandardised residuals.

We assessed the fit of the four correlated factor 12-item model developed in Study 1a. Unfortunately, the model fit poorly in this new sample (χ 2 (48) = 520.73, p < 0.001; CFI = 0.902; TLI = 0.865; RMSEA = 0.086), indicating a failure of replication.

Discussion of Study 1a and 1b

The aim of studies 1a and 1b was to test if a well-fitting model of the MORQ was possible and if this reliably supported the RMT. While a factor analysis supported evidence for four factors in the MORQ, it also showed that a substantial number of items either failed to load on their corresponding factor or showed large cross-loadings on other factors. While we could identify 12 items from this analysis such that three items were available for each predicted relationship model and fitted a 4-factor model, this model failed to replicate in an independent sample. Two possible accounts for this present themselves. First, the theoretical four-factor structure may be valid, but perhaps because of a small discovery sample, we were unable to select items which reliably assess this true structure, and instead, our item selection capitalised on sample-specific variance. Alternatively, the model replication may have failed because the four-factor structure itself is incorrect or incomplete. For example, it may be necessary to replace correlations between factors with a general relationship factor, representing a general tendency to initiate or avoid relationships with other people or to make some other model modifications. To address these possibilities, we conducted a second study with a larger discovery sample, tested a wider range of models in this sample, and requested five rather than one relationship from each participant, allowing us to validate the models across a range of participant responses.

Study 2: Alternate models and larger sample

Study 1 failed to find the well-fitting replicable structure of the MORQ. Although factor analysis indicated that four factors are needed to explain the variance in the scale, several items failed to load on the expected factors. Post-hoc 12-item model based on items that factor analysis suggested should be retained as relatively pure indicators of each of the four domains also failed to replicate. To address the Study 1 problems, in Study 2, we collected a larger sample and used a multi-trait multi-method approach to develop a well-fitting model of the MORQ and to test if this new model replicates well.

A total of 617 people (309 women, 304 men, 4 other; mean age 39.00, SD = 14.39) from the United Kingdom were recruited using Prolific Academic. The data were collected in January-February 2022. The study was approved by the University of Edinburgh PPLS Research Ethics Committee.

Participants’ endorsement of relational models was measured using the full 33-item Modes of Relationships Questionnaire (MORQ) [ 3 ]. The questionnaire was hosted on the Qualtrics survey platform. Before starting the study, participants received an explanation of the study and were asked to provide written consent by signing a consent form. After providing informed consent, each participant generated a list of 40 relationships. Qualtrics automation was then used to select five relationships at random, and for each of these, the subject was asked to complete the online MORQ with respect to this relationship. Total testing took approximately 19 minutes per participant on average. All data were de-identified and collected using anonymous codes to protect participants’ privacy. No personal identifying information was collected, and the authors did not have access to any information that could identify individual participants during or after data collection.

Descriptive statistics and Cronbach’s alpha coefficients for the four relational models are given in Table 2 . Cronbach’s alphas ranged from 0.74 to 0.87 suggesting good internal consistency of the four scales.

https://doi.org/10.1371/journal.pone.0287391.t002

To generate our model of the MORQ, we used only the first relationship (out of five reported by the participants), treating the remaining four relationships as internal hold-out replication datasets. Our initial model used a four-factor intercorrelated structure. To explore which, if any, sets of items would permit fit this structural model, we used a function designed to select the best items while keeping the factor structure intact. Procedurally, the function removed items one by one, starting from those that fit the model least well. This item removal process continued until the model reached a satisfactory model fit by at least two out of three criteria (CFI and TLI > = .95; RMSEA < = .06) [ 23 ]. The function is documented in the OSF site for this paper.

The automatic function removed four items from each of the EM, CS, and MP scales and three items from the AR scale, yielding a model which achieved a good fit in the test dataset but which did not replicate perfectly in the hold-out relationship datasets (see Table 3 ).

https://doi.org/10.1371/journal.pone.0287391.t003

While the drop-off was not substantial, we wished to investigate whether more complex models would reliably yield a good fit. Based on evidence that the four relational models are typically correlated [ 3 , 4 ], we tested the effect of removing the intercorrelations among the factors and instead modelling item covariance via a general factor loading on all items to the model. Examining the sources of a model misfit in earlier analyses also suggested a unique bivariate link between the CS and EM factors, which was added. Applying the same automated function to this new model supported this modification. A total of four EM and three CS, MP, and AR items were dropped, yielding a model with five indicators of each relational factor and resulting in a well-fitting model (χ 2 (149) = 990.79, p < 0.001; CFI = 0.956; TLI = 0.944; RMSEA = 0.044) which also replicated well across the holdout datasets. The RMSEA remained below the threshold in all tests while other fit indices improved or decreased in proportion as compared to the fit obtained in the initial data (see Table 4 ). Fig 2 shows the final model (using the dataset with all five relationships combined after controlling for between-subject variance). The final item set used in Study 2 is listed in the S1 Appendix .

https://doi.org/10.1371/journal.pone.0287391.g002

https://doi.org/10.1371/journal.pone.0287391.t004

Discussion of Study 2

In study two, we were able to create a well-fitting model of the MORQ retaining five items per relational mode. We also found that the model required a general factor loading on all items, which was not investigated in previous studies relying on factor correlations only. This model performed satisfactorily across holdout data from the same dataset, providing additional support for the validity of the four-factor MORQ model. One limitation of our study, however, is that the same participants generated both the discovery (the first reported relationship) and replication (the second to the fifth reported relationship) data. While the model replicated in the different target data provided by our subjects, we wished to replicate the model in a completely independent dataset to further corroborate the new model. We, therefore, conducted Study 3, testing the exact final model from Study 2 in a new dataset.

Our objective in Study 3 was to validate the 20-item model developed in Study 2 by replicating the results in an independent sample. We made no changes to the model, and the measures were identical to those used in Study 2. We expected to confirm the model structure using the same fit metrics as in Study 2 (TLI, CFI and RMSEA) and refine the model if needed.

A total of 615 people (307 women, 305 men, 3 other; mean age = 41.57, SD = 13.97) from the United Kingdom were recruited using Prolific Academic. Participants from studies 1 and 2 were excluded in order to ensure dataset independence. The data were collected in July 2022. The study was approved by the University of Edinburgh PPLS Research Ethics Committee.

In Study 3, we followed the same procedure and used the same materials (20-item MORQ with five items per each relational mode) as in Study 2. Total testing took approximately 15 minutes per participant on average. All data were de-identified and collected using anonymous codes to protect participants’ privacy. No personal identifying information was collected, and the authors did not have access to any information that could identify individual participants during or after data collection.

We fitted the exact model developed in Study 2 to the new dataset collected for Study 3 and examined its fit. The model replicated well showing excellent fit (χ 2 (149) = 990.79, p < 0.001; CFI = 0.955; TLI = 0.943; RMSEA = 0.043). In addition to a good fit, the factor loadings were also comparable to those found in Study 2. Moreover, the correlation between the CS and EM factors was also very similar (.85 vs .80 in Study 2). The replicated model is shown in Fig 3 ).

Full details of the model are tabulated on the OSF site for this paper.

https://doi.org/10.1371/journal.pone.0287391.g003

Discussion of Study 3

Study 3 successfully replicated the four-factor 20-item model of the MORQ developed in Study 2 in an independent dataset. Despite no structural changes to the model, the model fit metrics were excellent and comparable to those in Study 2. We can have confidence, therefore, that this is a reliable, well-fitting, and useful model of MORQ. Study 3 also confirmed the presence of a general factor in the MORQ, as shown in Study 2. This general factor had high loadings on CS, EM and MP relational model items and low (or negative) on AR relational model items. Given that the general factor was required in two independent datasets, we believe it is not a sampling error artefact and therefore requires further explanation. For instance, this factor may assess general commitment and devotion to form social relationships with others, with low scores representing a “null relationship” [ 2 ], an avoidant attitude towards forming relationships. However, traits such as socially desirable responding, differential emphasis on traditional social conventions and authorities across the four types, or artefacts of factors such as the closeness of the relationship to the respondent all could also account for some or all of the variance in the general factor. To explore these possibilities, we re-contacted the participants from studies 2 and 3 and tested associations with the general factor.

In Study 4, we investigate the nature of the general factor that emerged in both Study 2 and Study 3. We tested three possible explanations. First, several MORQ items describe socially desirable behaviours (e.g. “ If either of you needs something , the other gives it without expecting anything in return ”). For this reason, we hypothesised that the general factor might represent social desirability bias, exaggerating desirable traits due to honest self-deception or conscious impression management [ 30 ]. To test this hypothesis, we administered the Balanced Inventory of Desirable Responding (BIDR) [ 31 ] because it allows testing both types of bias: deliberate impression management and self-deceptive enhancement.

Second, given that in both studies 2 and 3, the general factor correlated negatively with AR items but positively with all items defining the other three MORQ models, we speculated that a simple authoritarian/non-authoritarian distinction could drive the general factor. To test this speculation, we asked participants to fill in the Right Wing Authoritarianism questionnaire (RWA) [ 32 ], which measures authoritarian personality traits such as submission to traditional authorities and social conventions.

Finally, we hypothesised that the general factor might reflect the specific relationship with the individual a respondent was rating. To test this, we asked participants to recall the relationships they reported during Study 2 and Study 3 data collection and to classify each of these individuals by type (e.g. “colleague” or “close family”). We coded this measure as a categorical variable with eight unordered levels. We expected higher general factor loadings for close relationship types (such as close family or close friend) and lower loadings for relationships that are typically less close (e.g. an employer or service personnel).

Study 2 and Study 3 participants were re-contacted on the Prolific academic platform 3–8 months after data collection from Study 2 and Study 3 was completed. A total of 447 participants agreed to participate in the follow-up study. The data were collected in September-October 2022. The study was approved by the University of Edinburgh PPLS Research Ethics Committee.

Before starting the study, participants received an explanation of the study and were asked to provide written consent by signing a consent form. After providing informed consent, the following three questionnaires were administered.

Right Wing Autoritarianism questionnaire (RWA) [ 32 ].

The RWA is a 22-item instrument measuring the tendency to defer to authorities, endorsement of traditional values, and support for aggression toward outgroups. Participants respond to a series of statements (e.g., “ What our country really needs is a strong , determined leader who will crush evil , and take us back to our true path ”) on a nine-point Likert scale ranging from 1 (Strongly disagree) to 9 (Strongly agree).

Balanced Inventory of Desirable Responding (BIDR) [ 31 ].

The BIDR is a 40-item instrument measuring the tendency to overstate one’s socially desirable behaviour and personality traits. The BIDR contains two separate 20-item measures, Impression Management, designed to test conscious self-presentation (e.g. “ I have never dropped litter on the street ”), and Self-Deceptive Enhancement (e.g. “ My first impressions of people usually turn out to be right ”). BIDR is scored on a seven-point Likert scale ranging from 1 (Not true) to 9 (Very true).

Relationship type.

We asked participants to recall the relationships they reported during the original data collection in studies 2 and 3 and to classify them by type. We coded the reported type of the relationship as a categorical variable with eight levels (“Your manager or employer”, “Your employee”, “Service personnel”, “Acquaintance”, “Colleague”, “Distant family”, “Close friend”, “Close family”.

The measures were hosted online on the Qualtrics survey platform. Total testing took approximately 8 minutes per participant on average.

First, we tested the hypothesis that authoritarianism explains the general factor scores extracted from the model using the umx function umxFactorScores(). Regression scoring was used to determine the factor scores, and potential confounding effects of authoritarianism were tested in each of the five relationships examined. The results showed no evidence of any association between the general factor and Right-Wing Authoritarianism (RWA). The correlation between the general factor and RWA was not significant in any of the five relationships, with correlations ranging from -.06 to .06 (e.g., in the relationship 1, r (437) = -.03, p = .504). Next, we tested whether the general factor was explained by social desirability, specifically self-deceptive enhancement and impression management scales of the BIDR. The results showed that the general factor was unrelated to both self-deceptive enhancement (e.g., in the relationship 1, r(527) = .03, p = .478) and impression management (r(527) = .04, p = .401) scales across all five relationships, contrary to our hypothesis.

Finally, we used regression to test the hypothesis that relationship type (dummy-coded with eight factors as indicated above) explains the general factor. We found that this was significant, explaining 6.2% of the variance (F(7, 2139) = 21.26, p < .001). The beta coefficients for each of the eight types of relationships can be seen in Table 5 . Fig 4 shows a boxplot depicting the relationship between the general factor and eight relationship categories.

https://doi.org/10.1371/journal.pone.0287391.g004

https://doi.org/10.1371/journal.pone.0287391.t005

Discussion of Study 4

Study 4 tested three possible causes of the general factor: authoritarianism, social desirability bias, and relationship type (e.g. colleague, close friend, etc.). While RWA (assessing authority) and BIDR measures (measuring socially desirable responding) were unrelated to general factor scores, relationship type did account for a portion of the variance in the general factor. High general factor loadings were associated with closer relationships (e.g. ‘close friend’). In contrast, low scores were associated with relationship types that are usually less close (e.g. ‘manager/employer’ or ‘service personnel’), suggesting that the general factor may represent the relationship’s closeness. The employee relationship had a much higher score on the general factor than the employer relationship, despite both representing the AR relationship. We believe this may reflect the paternalistic side of authoritarian leadership in which an employer’s role involves, to a degree at least, responsibility towards their employees. We discuss these findings in more depth in the general discussion.

General discussion

This paper aimed to test whether the MORQ measures four relationship models, as predicted by the RMT, and whether it accurately identifies the proposed structure of social relationship taxonomy. In the three studies reported above, we confirmed the existence of the original four factors, along with support for a general relationship factor. This new model of the MORQ has several implications for RMT and suggests additional directions for research. Each of these is discussed below.

Our main results (studies 2 and 3) supported the Haslam and Fiske [ 3 ] four-factor model of social relationships. The model demonstrated a good fit after eliminating items with significant cross-loadings and items that loaded on factors other than their intended ones. This refinement resulted in a model consisting of five items for each social relationship mode. The model also required some minor structural changes. Instead of four intercorrelated factors, the model required a general factor at the item level–in some ways, a more interpretable structure than the six factor intercorrelations it replaced. The general factor loaded positively on CS, EM and MP items but negatively on AR items.

We also found that EM and CS factors are highly correlated (r = .85 in Study 2 sample and r = .80 in Study 3 sample). This is consistent with the original Haslam and Fiske [ 3 ] findings, where these factors were positively correlated (r = .60). Despite the high correlation, combining these factors into one relationship worsened the model’s fit in our data, suggesting that these two relational models are distinct, but usually work together to define actual relationships. Our model was successfully replicated in an independent dataset, providing further support for the Relational Models four-factor structure of social relationships.

In Study 4, we tested the possible meaning of this general factor by undertaking three additional tests. First, we speculated that the general factor might reflect social desirability, overreporting desirable traits due to cognitive bias or conscious impression management. We tested this explanation by including two measures of social desirability [ 31 ], impression management, measuring conscious attempt to enhance self-presentation to others and self-deceptive enhancement, measuring an honest overestimation of one’s positive traits. Both these measures were not related to the general factor. This indicates that social desirability bias is not a major concern for the MORQ questionnaire.

As a second possible explanation, we tested whether the general factor measures a broad tendency to construe relationships based on a hierarchy. We theorised that CS, EM, and MP relational models describe relationships of individuals with the same status, whereas AR model explicitly implies unequal status. This would predict a significant negative relationship between the general factor and authoritarianism, but this was not the case; in our data, the correlation between the general factor and right-wing authoritarianism was not significant. Thus, we feel comfortable concluding that the general relational factor does not reflect hierarchical tendencies.

Finally, we tested whether the general factor represents the closeness of the relationship. We hypothesised that close relationships (those between close friends and family) should manifest as higher in CS and EM relational models than AR and MP, as is reflected in the general factor loadings. Our measure of reported relationship type correlated significantly and positively with the general factor. However, the strength of this relationship was weak, suggesting that relationship closeness is only a partial explanation of the general relational factor. As higher scores on this factor typically occurred for closer relationships, a useful direction for future work would be to study a tendency to invest in building connections. This would be consistent with the concept Fiske [ 2 ] termed general commitment and devotion to form social relationships with others, with low scores representing a “null relationship”, an avoidant attitude towards forming relationships.

Although our manuscript primarily aimed to enhance the MORQ’s psychometric properties and confirm its four-factor structure, our findings also provide insights into the underlying structure of social relationships. Our results suggest that relationships are structured around, at a minimum, these four models of interpersonal relations. Moreover, our findings refute the notion that these four models are merely consequences of a simpler, two-dimensional model (such as equality-inequality or close-distant) since these models did not adequately fit our data.

Limitations and future directions

We should keep in mind the limitations of the study. The present study supported a self-report measure of relational models with five instead of eight to nine items measuring each relational model. As the original 33 items were designed to cover a spectrum of behavioural domains, generating new items to replace the missing items may be of value to capture the complete spectrum of relationship models. That said, the scales developed here and scored by averaging responses for each scale should be valid for their intended purposes or for identifying relationship models. Of course, a further limitation is that we cannot rule out that other relational models may exist–seeking evidence for relationships that do not fit the four-model structure would be informative regarding the validity and generality of the broader theory. The present studies also were conducted thirty years after the initial study and in a different yet related culture (the UK compared to the US). This may partially account for the finding that some original items did not load on the expected factors. The finding that despite the three decades having elapsed and testing in a much changed and different culture in the UK, the model was validated is a testimony to the durability of the RMT model. However, international, cross-cultural replication of the model in non-western samples and further examination of the nature of the general relational factor are required. The future directions for this more compact, valid, and highly interpretable instrument appear wide. The four social models identified in this study represent universal building blocks of relationships and can be applied across a range of psychological sub-disciplines to gain insights into specific domains of social interactions. For example, within the context of parent-child relationships, one may explore how the four models manifest in parent-child interactions and their impact on child development. Similarly, within the field of education, the four models may be used to examine the dynamics of teacher-student relationships and their impact on student achievement and well-being. By mapping these universal models onto discipline-specific structures, researchers can gain a deeper understanding of the role of social interactions in various contexts and develop tailored interventions to promote positive relationship outcomes. For instance, workplace and organisational psychology is a particularly suitable discipline for the application of reliable, valid measures that can diagnose the current disposition of relationships among staff across different levels of business units or larger structures. These measures can test the alignment of these models with the intended and desired business strategy and assess the efficacy of interventions designed to incentivise relationship change where necessary. Surveying organisations using these measures can reveal if relations designed to primarily embody hierarchy and proportionality are functioning as intended. Additionally, these measures can test the association of incentives with the strength of reported models in a given relationship or modulate incentives to assess the effects predicted by the Relational Incentives Theory. Thus, these measures can add significant value to organisational research and practice.

Our study aimed to establish the validity and psychometric structure of the MORQ, resulting in a compact, reliable, and valid instrument suitable for use in various applied settings. By providing an efficient means of measuring individuals’ preferences for different relational models, the validated scale can be used to further explore and apply RMT. Our findings suggest that the MORQ can be a useful tool in both research and applied settings, facilitating a deeper understanding of the role of relational models in human behaviour and well-being.

Supporting information

S1 appendix. items retained in final relational models scale, validated in studies 2 and 3..

https://doi.org/10.1371/journal.pone.0287391.s001

• 1. Fiske AP. Structures of social life: The four elementary forms of human relations: Communal sharing, authority ranking, equality matching, market pricing: Free Press; 1991.
• View Article
• PubMed/NCBI
• Google Scholar
• 6. Fiske AP. Relational Models Theory 2.0. In: Haslam N, editor. Relational models’ theory: A contemporary overview. London: Lawrence Erlbaum Associates; 2004. p. 3–25.
• 7. Fiske AP, Rai TS. Virtuous violence: Hurting and killing to create, sustain, end, and honor social relationships.: Cambridge University Press.; 2014.
• 13. Arendt JFW, Kugler KG, Brodbeck FC. Being on the same page about social rules and norms: Effects of shared relational models on cooperation in work teams. Group Processes & Intergroup Relations. 2022.
• 16. Van Lange PAM, Balliet D. Interdependence theory. APA handbook of personality and social psychology, Volume 3: Interpersonal relations2015. p. 65–92.
• 18. Tajfel H, & Turner, J. C. The social identity theory of intergroup behavior. Political psychology: Psychology Press; 2004. p. 276–93.
• 21. Atari M, Haidt J, Graham J, Koleva S, Stevens ST, Dehghani M. Morality beyond the weird: How the nomological network of morality varies across cultures. 2022.
• 24. Yu C-Y. Evaluating cutoff criteria of model fit indices for latent variable models with binary and continuous outcomes. Los Angeles, CA: University of California, Los Angeles; 2002.
• 26. R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. 4.1.0 ed2021.
• 31. Paulhus DL. Measurement and control of response bias. In: Robinson JP, Shaver PR, Wrighsman LS, editors. Measures of personality and social psychological attitudes. San Diego, CA: Academic Press.; 1991. p. 17–59.
• 32. Altemeyer B. The authoritarians. Winnipeg: B. Altemeyer; 2007.

Relational Model: Structure and Architecture

• First Online: 05 December 2020

Cite this chapter

• Thomas Bressoud 3 &
• David White 3  

2410 Accesses

This chapter introduces the relational model and defines database management systems, data relations, database tables, fields, primary keys, and types of relationships between tables (i.e., one-to-one, one-to-many, and many-to-many). We describe how database systems fit within the framework of providers and clients, and we introduce the use of SQL.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info
• Durable hardcover edition

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

The “commands” of SQL, by convention, are written in all uppercase, to help clarify the use as an SQL syntactic reference.

For instance, under the tabular model, we might have a local directory structure that organizes a set of CSV files into an informal kind of database.

In the mathematical model, rows/records are called tuples and columns/fields are called attributes . We will avoid using these terms to avoid additional confusion.

When we use the term “key,” it is a singular noun. Even if a key is composite, the singular is still used, so do not get misled.

Using delimiters allows more complex table and field names, such as incorporating spaces into the name. For simplicity and readability, the tables and fields used in the examples of this book will be simple, and delimiters omitted.

A generic utility issuing SQL is possible because SQL is a language unto itself, agnostic to the programming language of the client, and using declarative commands to specify what data is desired.

Reliable communication over a network always requires bi-directional communication.

Codd, E.F.: A relational model of data for large shared data banks. Communications of the ACM 13 (6), 377–387 (1970)

Article   Google Scholar  

Merriam-Webster: database (2019). URL https://www.merriam-webster.com/dictionary/database . Accessed: 2019-09-27

Wikipedia: Relational database (2020). URL https://en.wikipedia.org/wiki/Relational_database . Accessed: 2020-05-31

Download references

Author information

Authors and affiliations.

Mathematics and Computer Science, Denison University, Granville, OH, USA

Thomas Bressoud & David White

You can also search for this author in PubMed   Google Scholar

Rights and permissions

Reprints and permissions

Copyright information

© 2020 Springer Nature Switzerland AG

About this chapter

Bressoud, T., White, D. (2020). Relational Model: Structure and Architecture. In: Introduction to Data Systems. Springer, Cham. https://doi.org/10.1007/978-3-030-54371-6_10

Download citation

DOI : https://doi.org/10.1007/978-3-030-54371-6_10

Published : 05 December 2020

Publisher Name : Springer, Cham

Print ISBN : 978-3-030-54370-9

Online ISBN : 978-3-030-54371-6

eBook Packages : Mathematics and Statistics Mathematics and Statistics (R0)

Share this chapter

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Adv Radiat Oncol
• v.5(6); Nov-Dec 2020

Artificial Intelligence Research: The Utility and Design of a Relational Database System

Although many researchers talk about a “patient database,” they typically are not referring to a database at all, but instead to a spreadsheet of curated facts about a cohort of patients. This article describes relational database systems and how they differ from spreadsheets. At their core, spreadsheets are only capable of describing one-to-one (1:1) relationships. However, this article demonstrates that clinical medical data encapsulate numerous one-to-many relationships. Consequently, spreadsheets are very inefficient relative to relational database systems, which gracefully manage such data. Databases provide other advantages, in that the data fields are “typed” (that is, they contain specific kinds of data). This prevents users from entering spurious data during data import. Because each record contains a “key,” it becomes impossible to add duplicate information (ie, add the same patient twice). Databases store data in very efficient ways, minimizing space and memory requirements on the host system. Likewise, databases can be queried or manipulated using a highly complex language called SQL. Consequently, it becomes trivial to cull large amounts of data from a vast number of data fields on very precise subsets of patients. Databases can be quite large (terabytes or more in size), yet still are highly efficient to query. Consequently, with the explosion of data available in electronic health records and other data sources, databases become increasingly important to contain or order these data. Ultimately, this will enable the clinical researcher to perform artificial intelligence analyses across vast amounts of clinical data in a way heretofore impossible. This article provides initial guidance in terms of creating a relational database system.

Introduction

This issue of Advances in Radiation Oncology presents a series of articles around applications of artificial intelligence (AI) in our field. One of the potential benefits of AI is that it can pore through large amounts of data to discover patterns not evident to clinicians. However, this vast volume of data cannot be accommodated within a single spreadsheet (which is how most clinical researchers work when conducting standard multivariable analyses). In fact, many researchers erroneously describe spreadsheets as databases. This article will demonstrate both what a relational database system is and how it is superior to a spreadsheet. It will also discuss considerations when implementing a relational database system (RDBS) for your own research purposes, using an actual lung cancer radiation therapy database as an example. I have also provided some excellent Wikipedia references that contain abundant additional information, beyond what can be encapsulated in a single article. (These, in turn, reference computer science literature for the very intrepid reader, but such references might extend beyond the level of understanding of all but the most technically inclined.)

One might question why a database system is necessary for AI research. This article will demonstrate that a database enables creation of a multidimensional structure to cleanly and accurately contain these data. To perform AI analysis requires efficient storage of hundreds or thousands of data points on a single patient or even on a single course of radiation therapy. There is a famous illustration of the “data science hierarchy of needs” ( Fig 1 ). To perform an AI analysis, one must first create an RDBS to serve as the storage mechanism. This creation of a system to store structured data entails a major part of the bottom row of the pyramid. To create a database, then, will set the reader down the path toward conducting AI research at their own institution.

The data science hierarchy of needs. Used with permission of Monica Rogati ( aipyramid.com ). For details, see text. Abbreviation : AI = artificial intelligence.

Origin of Relational Databases

The concept of a RDBS was first described in a seminal article in 1970. 1 The theoretic construct was that all data could be defined or represented as a series of relations with or to other data. The article was quantitative in that it used relational algebra and tuple relational calculus to prove its points. 2 IBM used this theoretic framework to design what became the initial SQL (pronounced “see-quell” or “ess-cue-ell”) language, which they used to manipulate and retrieve data from IBM’s original RDBS. 3 Since that time, the American National Standards Institute and the International Standards Organization have deemed SQL to be the standard language in relational database communication. 2 Today, there are a wide variety of commercial and open-source relational database systems available for use. These vary in their features and relative strengths or weaknesses, but, fundamentally, they all operate using the principles defined in the Codd article. 1 The SQL language is well defined and is used to write code to query (or update) the data within an RDBS.

Fundamental Disadvantage of Spreadsheets

Spreadsheets are designed to incorporate and analyze one-to-one (1:1) relationships ( Fig 2 a). Each patient has a single birth date and a single death date. However, medical records are rife with “one-to-many” relationships ( Fig 2 b). A patient might receive multiple different courses of radiation therapy treatment, as in the example provided, or might have multiple chemotherapy administrations. To accommodate these data, a spreadsheet quickly balloons in size ( Fig 2 c). Not only is this inefficient (duplication of data), but it also makes maintenance of the spreadsheet extremely cumbersome and prone to error. For instance, in this example, when patient “12345” passes away, the “DeathDate” needs to be updated in 5 rows of the spreadsheet (because she had 2 courses of radiation therapy and 4 cycles of chemotherapy). It is not difficult to imagine that a researcher could neglect to update the “DeathDate” in each place, introducing errors. To further expound upon the issue, imagine a patient who takes numerous medications or has variable numbers of comorbid illnesses; the rows required to encapsulate 1 patient explode. To use a data science term, the dimensionality of the data balloons. But, to reiterate the point, spreadsheets are only designed to encapsulate 1:1 relationships (2-dimensional data). But patient data are multidimensional.

(a) Spreadsheets are useful where there is a one-to-one correspondence of data. For instance, each unique medical record number (MRN) represents a single patient, with a single birth/death date and a single first and last name. (b) Spreadsheets “break down” when describing 1-to-many correspondences. In this example, 2 patients have a total of 5 courses of radiation therapy treatment between them. (c) To accommodate all the data in our simple example, a spreadsheet needs to store redundant data (colored in red). The data storage requirements quickly balloon. Furthermore, as additional traits and factors are added to the spreadsheet, it becomes impossible to follow, as one patient will require untold numbers of rows to capture all relevant data concepts. Stated another way, the data are multidimensional. Maintenance and updating of fields become error-prone (see text). Abbreviations : DOB = date of birth; Lname = last name; Fname = first name; LUL = upper lobe; MRN = medical record number; RLL, right lower lobe; RUL = right upper lobe; SBRT = stereotactic body radiotherapy.

Fundamental Advantage of Relational Databases

RDBS gracefully manage one-to-many relationships. They can do so because a database is created of numerous different tables, which are explicitly linked ( Fig 3 ). Every table must also contain a key, which is a unique, required identifier for each row of data. Relationships between the tables are defined when creating the database tables or fields. In the “demographics” table, medical record number, “MRN,” is the key. For the “TreatmentCourse” and “Chemotherapy” tables, the keys are “TreatmentCourse” and “ChemotherapyID,” respectively. Note that “TreatmentCourse 1” in the “TreatmentCourse” table pertains to breast radiation therapy treatment. This, in turn, is linked to 4 cycles of chemotherapy in the “Chemotherapy” table, each of which is uniquely identified in that table, in turn.

In a relational database, data are stored in multiple tables, which are joined via defined variables. In this fictitious example, note that the patient only has one “DeathDate” to update. Furthermore, each course of treatment (“TreatmentCourse”) can have multiple chemotherapy cycles associated with it. Note, too, that medical record number (MRN) only appears in 2 of the 3 tables (it is not needed in the “Chemotherapy” table). If the researcher wishes to retrieve the MRN, it can be obtained via a SQL query, linking back to one of the tables that contains it. Abbreviations : DOB = date of birth.

When comparing Fig 3 (a database) to Fig 2 c (a spreadsheet), note that Fig 3 contains the same information as Fig 2 c without the addition of repetitious information (colored red in Fig 2 c). Alternately, in a spreadsheet, the researcher could manually aggregate and summarize the chemotherapy delivered into a single cell in a single row of the spreadsheet (ie, “flatten” the data, to use a data science term), but then some data would be lost. Using the chemotherapy administrations as an example, if one were to “flatten” the data down to a single spreadsheet cell stating “4 cycles of Adriamycin/Cytoxan,” one loses the dates of administration. If one summarizes the data as “4 cycles of Adriamycin/Cytoxan: <date1>, <date2>, <date3>, <date4>,” the dates and the chemotherapy occupy the same cell and the data are retained but are no longer discrete; one loses the ability to filter the spreadsheet by chemotherapy kind or by date.

Conversely, a SQL database cleanly encapsulates these multidimensional data ( Fig 3 ). Each table is 2-dimensional in structure. But because it can contain multiple rows of data on 1 patient (chemotherapy administrations, to follow the same example), a multidimensional structure is created, as 4 chemotherapy cycles link to one of 2 courses of radiation therapy (“TreatmentCourse” table) for 1 patient (“Demographics” table). Now, imagine a clinical database with millions of rows of data spread across hundreds of tables, as in the real-life example described below. Clearly, a spreadsheet would not be adequate.

Additional Advantages of Relational Databases

• 1. Each row of data in a table has a unique identifier (a key). Consequently, one cannot accidentally add a row of data into a database table twice.

Note that each field in this database table is specifically designed. It has a “type” (kind) and a “size” (length). When importing data from numerous external sources, these definitions can prevent erroneous imports (for details, see text). Note that the field “MRN” is the key for this table. All the data in this table refer back to “MRN” via a one-to-one relationship. MRN can be used as the key because no 2 patients have the same MRN.  Abbreviations : DOB = date of birth; MRN = medical record number.

• 3. Not only must the data types correspond, but the data lengths must be observed. If the database design states that a field is a decimal with 3 places to the right of the decimal place, then a fourth decimal place would be truncated at import. Alternately, the database could declare an error, which might also imply that the field contains erroneous data.
• 4. A key from one table can be linked “backward” to a key from another table (termed a “foreign key”). As an example ( Fig 3 ), the database can be designed such that the MRN from “TreatmentCourse” must refer to an MRN already contained within the “Demographics” table. If one tried to import data into “TreatmentCourse” and it used an MRN not listed in “Demographics,” the import would fail. Such a situation might imply, for instance, that the MRN was incorrect in the external data source (or in the database). Or perhaps it relates to a patient who received prostate radiation therapy (but you have a breast cancer database).
• 5. Foreign key relationships also work in the opposite direction: If one realizes that a patient is represented in the database who should not be, one can delete the patient from the “Demographics” table and the database will delete all data about that patient from all the other data tables automatically.
• 6. RDBS are specifically optimized to manage vast amounts of data. Large spreadsheets (containing thousands of rows and hundreds of columns) are extremely slow and memory intensive. However, one can query across or manipulate many gigabytes of data in fractions of a second in many RDBS, as the data stores are highly optimized and efficient from both a computational and memory utilization perspective.
• 7. RDBS are much more secure than spreadsheets. An institution’s IT team might allow one to access some tables within an institution’s data warehouse, but not others. One’s access could be restricted to defined subsets of patients. One might have “read” access to these data, but not “write” access (or “write” access to only some subset of fields). Databases might likewise be set up such that only users from specific IP addresses or computers may access them. The login systems set up by IT departments for these purposes typically use state of the art encryption algorithms, 2-factor authentication, and the like. In contrast, an Excel spreadsheet can be “locked” such that only some fields are editable. But it is not possible to restrict data access by user. Furthermore, this restriction is to “write” access only, not to “read” access. It is true that one can “hide” columns in a spreadsheet and then lock it, to prevent a given user from viewing them, but the spreadsheet maintainer must do this manually before distributing the spreadsheet (time-consuming and prone to error).

Benefits of SQL

As described earlier, SQL is a defined, standardized language for composing queries within an RDBS, or to manipulate and update these data. Some database systems provide “extensions” to the SQL standard, to provide some additional and specific functionality (details available in the vendors’ literature). It is beyond the scope of this article to teach SQL coding. However, many excellent online resources are available for the interested reader. Functionally, SQL allows one to search for any number of variables or combinations of variables across any number of tables, simultaneously. This can be extremely powerful and useful, both for retrieving and for manipulating and updating data. Queries can be saved for reuse or modification later. As stated above, these queries typically produce output in fractions of a second, even across vast pools of data.

Our institution has a database of patients who have received radiation therapy to the lung, whether for primary cancer or metastatic disease to the lung. 4 The database and some of its details of implementation are described below, but first, some “real-life” examples of what such an RDBS system can do (not possible when using a spreadsheet):

• • Real-life example 1: Find patients who might be candidates for a certain lung cancer clinical trial. For this particular study, they must have previously received lung SBRT, have nonmetastatic disease, no evidence of recurrence, be alive (obviously), be at least 2 years out from the end of the prior SBRT treatment, and must have been seen in follow-up within the past 1.5 years. By constructing an appropriate SQL query, 135 patients were found (out of more than 4600 in the database) to pass along to the PI for closer inspection.
• • Real-life example 2: It takes only a few minutes to set up very complex queries. If one has a basic facility with SQL, one can design a query such as: “Find all patients with stage II or III lung cancer treated with concurrent chemoradiotherapy who developed neutropenia during treatment, who are female, 70 years of age or older, and who take any antihypertensive medication (defined in a certain list).” Ultimately, such queries are only limited by one’s imagination (and the richness or completeness of the data coming from the source systems).

It is true that one can “filter” data in Excel to rapidly find subsets. However, this filtering is limited to “true or false” matching. In this example, it would be impossible to discover the patients who developed neutropenia while undergoing radiation therapy unless one had a “neutropenia (yes or no)” column. But one cannot perform the arithmetic “where date of neutropenia > RadiationStartDate and < RadiationEndDate” to filter the data without writing code in Visual Basic, which is likely beyond the ability of most.

Disadvantage of SQL

With SQL, it is possible to create highly complex queries; it is a rich and powerful language. However, these can be quite complicated and obtuse to a nontechnical person. Some systems do provide graphical tools to help build SQL queries, but, even so, there are some users for whom all but the simplest SQL queries will be beyond their technical skills.

Database Implementation

Databases may, and often do, contain many thousands of tables and millions of rows of data. (In other words, they can contain data far in excess of the requirements of any one radiation oncologist or even any one radiation oncology department). In fact, some systems allow even single tables to contain terabytes or even petabytes of data. 5 Consequently, there are numerous systems available to accommodate any researcher’s needs. Some of the very best are open source (free). Software is available across a wide variety of operating systems. Wikipedia provides an outstanding overview of the topic. 5

To implement a database system, it is first necessary to have a discussion with the IT Department at your institution. There is no single solution for creating a data repository that holds true for researchers across all institutions. The solution can vary, depending on the resources at your institution and the level of access the IT Department has into the underlying patient data source systems (often defined in the institutional contracts signed with the individual vendors). Some large centers have elaborate data warehouse systems. Smaller centers might provide access to data from individual source systems but might not have compiled them into a single data warehouse. Some IT departments might have adequate resources to provide output data from their data warehouse to individual researchers, when needed. Others might not. Some might provide a dedicated research server on which the researcher can construct a database. Other researchers might need to rely instead upon existing servers within their department. I do not recommend that one set up a database system on a free-standing laptop or desktop machine, as there are Health Insurance Portability and Accountability Act concerns (the computer could be stolen). Data should be backed up across a secured network electronically.

Creation of a Lung Cancer Radiation Therapy Database

I began my own database several years ago. My need grew out of a sense of frustration regarding lack of access to clinical data. At the time, at my institution, it was a difficult (and somewhat mysterious) process to procure data from the data warehouse. However, data from Mosaiq (Elekta AB, Stockholm, Sweden), which is our department’s record or verify system, were available. These data formed the nucleus of the original database. Basic demographic information and radiation therapy prescriptions, dates of treatment, dose delivered, tumor stage, and the like, were exported, using custom software. Research IT provided a Linux server, on which I implemented the database. I chose to use MariaDB (MariaDB Foundation, DE), as it is a powerful, well-regarded, commercially supported, free, open-source database system whose SQL functions are congruent with those of Oracle (which is a database system I had used previously). Because my institution could not support an implementation of an Oracle, MariaDB was an excellent alternate option. MariaDB does include a Windows GUI database administration tool for administering its databases (creating tables, writing SQL code, importing or exporting data, and the like). I had previously used a similar commercial database administration product called Navicat (Premium Cybertech Ltd, Hong Kong), which provides similar functionality, so I elected to purchase that. Similarly, I imported all data I had captured in spreadsheets for previous research projects. More recently, I have gained access to data from our data warehouse and so have created numerous additional tables to store the information. At present, the database contains more than 3 million rows of data on approximately 4800 patients, spread across more than 170 tables.

Incorporation of Data from Other Institutional Data Systems

To import data from outside source systems requires a multistep process, referred to as “ETL” (“Extraction, Transformation, and Loading”) in the data science literature.” 6 The issues go far beyond the physical importation of data into the database; importing spreadsheets of data are a trivial task. There are numerous issues in ETL, which are critical to consider when designing a database and importing data into it. Furthermore, many of these issues are not inherently obvious. In fact, a large proportion of the time required to create a database and fill it with clinically useable data derives from the ETL involved. The oft-reproduced “Data Science Hierarchy of Needs” illustrates this fact ( Fig 1 ). Most of the discussion in this article addressed aspects of the bottom-most layer of the pyramid. ETL comprises the majority of the next 2 layers of the pyramid and is the topic of another article.

Research data are not available at this time.

Sources of support: none.

Disclosures: Dr Dilling reports personal fees and nonfinancial support from NCCN, personal fees from Varian, personal fees and nonfinancial support from Harborside Press, nonfinancial support from Astra Zeneca, all outside the submitted work.

Browse Books

January 1990

Menlo Park, California

• Addison-Wesley Longman Publishing Co., Inc.
• 75 Arlington Street, Suite 300 Boston, MA
• United States

Save to Binder

From the Preface (See Front Matter for full Preface)

An important adjunct to precision is a sound theoretical foundation. The relational model is solidly based on two parts of mathematics: firstorder predicate logic and the theory of relations. This book, however, does not dwell on the theoretical foundations, but rather on all the features of the relational model that I now perceive as important for database users, and therefore for DBMS vendors. My perceptions result from 20 years of practical experience in computing and data processing (chiefly, but not exclusively, with large-scale customers of IBM), followed by another 20 years of research.

I believe that this is the first book to deal exclusively with the relational approach. It does, however, include design principles in Chapters 21 and 22. It is also the first book on the relational model by the originator of that model. All the ideas in the relational model described in this book are mine, except in cases where I explicitly credit someone else.

In developing the relational model, I have tried to follow Einstein's advice, "Make it as simple as possible, but no simpler." I believe that in the last clause he was discouraging the pursuit of simplicity to the extent of distorting reality. So why does the book contain 30 chapters and two appendixes? To answer this question, it is necessary to look at the history of research and development of the relational model.

Book Downloads

• Kay M (2024). ggdist: Visualizations of Distributions and Uncertainty in the Grammar of Graphics, IEEE Transactions on Visualization and Computer Graphics , 30 :1 , (414-424), Online publication date: 1-Jan-2024 .
• Wolfengagen V, Ismailova L and Kosikov S (2024). Building a cognitive system based on process interaction, Cognitive Systems Research , 83 :C , Online publication date: 1-Jan-2024 .
• Li P, He Y, Yan C, Wang Y and Chaudhuri S (2023). Auto-Tables: Synthesizing Multi-Step Transformations to Relationalize Tables without Using Examples, Proceedings of the VLDB Endowment , 16 :11 , (3391-3403), Online publication date: 1-Jul-2023 .

• Robert A, Genest D and Loiseau S Cognitive Map Query Language for Temporal Domains Agents and Artificial Intelligence, (215-235)
• Grant J and Parisi F Measuring Inconsistency in a General Information Space Foundations of Information and Knowledge Systems, (140-156)
• Alattar M and Sali A Keys in Relational Databases with Nulls and Bounded Domains Advances in Databases and Information Systems, (33-50)
• Vasconcelos G, Kaster D and Cordeiro R On the Support of the Similarity-Aware Division Operator in a Commercial RDBMS Advances in Databases and Information Systems, (142-155)
• Chen L, Roggenbach M and Tucker J An Algebraic Theory for Data Linkage Recent Trends in Algebraic Development Techniques, (47-66)
• Auziņš A, Eiduks J, Vasiļevska A and Dzenis R (2018). Object-Relational Database Structure Model and Structure Optimisation, Applied Computer Systems , 23 :1 , (28-36), Online publication date: 1-May-2018 .
• Tuya J, Riva C, Suarez-Cabal M and Blanco R (2016). Coverage-Aware Test Database Reduction, IEEE Transactions on Software Engineering , 42 :10 , (941-959), Online publication date: 1-Oct-2016 .
• Köhler H, Leck U, Link S and Zhou X (2016). Possible and certain keys for SQL, The VLDB Journal — The International Journal on Very Large Data Bases , 25 :4 , (571-596), Online publication date: 1-Aug-2016 .
• Burger E, Henss J, Küster M, Kruse S and Happe L (2016). View-based model-driven software development with ModelJoin, Software and Systems Modeling (SoSyM) , 15 :2 , (473-496), Online publication date: 1-May-2016 .
• Rahmani H, Ranjbar-Sahraei B, Weiss G and Tuyls K (2016). Entity resolution in disjoint graphs, Intelligent Data Analysis , 20 :2 , (455-475), Online publication date: 1-Jan-2016 .
• Gao Z, Min H, Li X, Huang J, Jin Y, Lei A, Bourbonnais S, Zheng M and Fuh G Optimizing Inter-data-center Large-Scale Database Parallel Replication with Workload-Driven Partitioning Special Issue on Database- and Expert-Systems Applications on Transactions on Large-Scale Data- and Knowledge-Centered Systems XXIV - Volume 9510, (169-192)
• Köhler H, Link S and Zhou X (2015). Possible and certain SQL keys, Proceedings of the VLDB Endowment , 8 :11 , (1118-1129), Online publication date: 1-Jul-2015 .
• Köhler H, Link S and Zhou X (2018). Possible and certain SQL keys, Proceedings of the VLDB Endowment , 8 :11 , (1118-1129), Online publication date: 1-Jul-2015 .
• Palacios M, García-Fanjul J, Tuya J and Spanoudakis G (2015). Automatic test case generation for WS-Agreements using combinatorial testing, Computer Standards & Interfaces , 38 :C , (84-100), Online publication date: 1-Feb-2015 .
• McGinnes S and Kapros E (2015). Conceptual independence, Information Systems , 47 :C , (33-50), Online publication date: 1-Jan-2015 .
• Rodríguez-Jiménez J, Cordero P, Enciso M and Mora A Automated inference with fuzzy functional dependencies over graded data Proceedings of the 12th international conference on Artificial Neural Networks: advences in computational intelligence - Volume Part II, (254-265)
• Bermbach D and Kuhlenkamp J Consistency in Distributed Storage Systems Revised Selected Papers of the First International Conference on Networked Systems - Volume 7853, (175-189)
• Bautista-Ramos C, Guillén-Galván C and Rangel-Huerta A (2013). From orthogonal projections to a generalized quantum search, Quantum Information Processing , 12 :1 , (1-20), Online publication date: 1-Jan-2013 .
• Nambiar R, Chitor R and Joshi A Data Management --- A Look Back and a Look Ahead Revised Selected Papers of the First Workshop on Specifying Big Data Benchmarks - Volume 8163, (11-19)
• Budikova P, Batko M and Zezula P Query language for complex similarity queries Proceedings of the 16th East European conference on Advances in Databases and Information Systems, (85-98)
• Lesniak M Palovca Proceedings of the 14th international conference on Practical Aspects of Declarative Languages, (153-167)
• Xue Y, Ghenniwa H and Shen W (2012). Frame-based ontological view for semantic integration, Journal of Network and Computer Applications , 35 :1 , (121-131), Online publication date: 1-Jan-2012 .
• Osorio M, Torrijos T, Sánchez A and Arroyo O Preliminary analysis for an air quality management DSS in the metropolitan valley of Puebla, Mexico Proceedings of the 13th WSEAS international conference on mathematical methods, computational techniques and intelligent systems, and 10th WSEAS international conference on non-linear analysis, non-linear systems and chaos, and 7th WSEAS international conference on dynamical systems and control, and 11th WSEAS international conference on Wavelet analysis and multirate systems: recent researches in computational techniques, non-linear systems and control, (210-215)
• Kácha P Efficient structured log storage Proceedings of the 14th WSEAS international conference on Computers: part of the 14th WSEAS CSCC multiconference - Volume I, (368-374)
• Chiticariu L, Krishnamurthy R, Li Y, Raghavan S, Reiss F and Vaithyanathan S SystemT Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, (128-137)
• Germanus D, Khelil A and Suri N Increasing the resilience of critical SCADA systems using peer-to-peer overlays Proceedings of the First international conference on Architecting Critical Systems, (161-178)
• Johnson M and Kasangian S A relational model of incomplete data without nulls Proceedings of the Sixteenth Symposium on Computing: the Australasian Theory - Volume 109, (89-94)
• Chebotko A, Lu S and Fotouhi F (2009). Semantics preserving SPARQL-to-SQL translation, Data & Knowledge Engineering , 68 :10 , (973-1000), Online publication date: 1-Oct-2009 .
• Kharlamov E and Nutt W (2008). Incompleteness in information integration, Proceedings of the VLDB Endowment , 1 :2 , (1652-1658), Online publication date: 1-Aug-2008 .
• Matos V, Grasser R and Jalics P (2006). The case of the missing tuple, Journal of Computing Sciences in Colleges , 22 :1 , (23-32), Online publication date: 1-Oct-2006 .
• Polyviou S, Samaras G and Evripidou P A Relationally Complete Visual Query Language for Heterogeneous Data Sources and Pervasive Querying Proceedings of the 21st International Conference on Data Engineering, (471-482)
• Osei-Bryson K and Ngwenyama O (2004). Supporting Semantic Diversity in the Relational Data Model, Information Systems Frontiers , 6 :3 , (277-285), Online publication date: 1-Sep-2004 .
• Vieira M and Madeira H (2004). Joint evaluation of recovery and performance of a COTS DBMS in the presence of operator faults, Performance Evaluation , 56 :1-4 , (187-212), Online publication date: 1-Mar-2004 .
• Lammari N and Métais E (2004). Building and maintaining ontologies, Data & Knowledge Engineering , 48 :2 , (155-176), Online publication date: 1-Feb-2004 .
• Siau K The psychology of information modeling Advanced topics in database research vol. 1, (106-118)
• Prinz A and Thalheim B Operational semantics of transactions Proceedings of the 14th Australasian database conference - Volume 17, (169-179)
• Darwen H Relational database Encyclopedia of Computer Science, (1519-1524)
• Hsu C and Knoblock C (2000). Semantic Query Optimization for Query Plans of Heterogeneous Multidatabase Systems, IEEE Transactions on Knowledge and Data Engineering , 12 :6 , (959-978), Online publication date: 1-Nov-2000 .
• Colburn T (1998). Information Modeling Aspects of Software Development, Minds and Machines , 8 :3 , (375-393), Online publication date: 1-Aug-1998 .
• Lewis P What Clinical Database Management System does the NHS need? (Position Paper) Proceedings of the Eighth International Conference on Scientific and Statistical Database Management, (162-166)
• Niemi T and Järvelin K (1991). Prolog-Based Meta-rules for Relational Database Representation and Manipulation, IEEE Transactions on Software Engineering , 17 :8 , (762-788), Online publication date: 1-Aug-1991 .

• Publication Years 1952 - 2007
• Publication counts 31
• Citation count 8,172
• Available for Download 26
• Downloads (cumulative) 160,449
• Downloads (12 months) 22,085
• Downloads (6 weeks) 2,256
• Average Downloads per Article 6,171
• Average Citation per Article 264

Index Terms

Information systems

Data management systems

Database design and models

Database management system engines

Parallel and distributed DBMSs

Query languages

Theory of computation

Theory and algorithms for application domains

Database theory

Database query languages (principles)

Reviewer: John M. Artz

In 1970 E. F. Codd published a paper in the Communications of the ACM [1] that introduced the relational data model and made an indelible mark on the evolution of database management systems. For the last 20 years, Codd has staunchly defended the relational data model against the pragmatic compromises of vendors who wished to call their products relational without delivering on the full promise of a relational database. In a 1985 Computerworld interview Codd announced his 12 criteria for determining whether a database management system could or should be called relational [2,3]. Since none of the products on the market met all 12 criteria, bitter debate ensued. Vendors tried to protect their turf (the database products) and Codd fought to protect his (the relational model), Codd's premises are certainly straightforward. He wishes to keep the relational model simple and abstract: simple so the semantics of the model do not become unwieldy for casual users, and abstract so integrity can be insured by the database management system rather than by application programmers. His rationale for these premises is also simple. Codd believes that data should be shared and accessible by a wide variety of people who may or may not be familiar with the physical peculiarities of the database. He wishes to deliver simplicity and reliability to the database user at the expense of the database product developer. This book is a feature-by-feature description of the Relational Model Version 2 (RM/V2). The original relational model, now designated RM/V1, had 12 specific requirements as stated in the Computerworld interview. Version 2 has over 300 features, only 50 of which were implicit in RM/V1. Once the industry has absorbed the shock of this new treatise, the controversy will begin again. This time, however, the debate will include a new group—those champions of the relational model who disagree with Codd. The dissension will occur in three main areas. First, Codd addresses his readers in a heavy-handed prescriptive tone that unequivocally declares the conditions for being fully relational, leaving no room for alternative views or further debate. Second, Codd has extended the model in certain ways that run counter to conventional wisdom, some of which are in direct conflict with more fundamental aspects of the model. Third, he has failed to extend the relational model in other ways that would almost certainly have gained wide acceptance while maintaining the integrity of the model. Codd states that in order to be fully relational in the 1990s a database management system must meet more than 300 stringent criteria. I have three problems with this position. First, while the industry is grateful to Codd for introducing the relational model in 1970, it is not at all clear that he alone should maintain it in 1990. Many brilliant minds have addressed relational issues in the last 20 years, and many more opinions remain to be gathered. Second, many of the features are ill-defined, so it would probably not be possible to implement all of them even if a vendor were so inclined. Finally, some of the features are counter to the premises of simplicity and abstraction and should not be implemented. Instead of dictating the features as requirements, it would have been more appropriate for Codd to offer them as his opinions and as a basis for further debate. The debate will ensue regardless, but Codd's extreme position on these features puts him at great risk for lost credibility if the industry does not ultimately agree. Another area of potential discord is the extensions to the model that are not widely accepted as appropriate directions. For example, Codd wishes to enforce closure on relational operators, so any relational operators must return a valid relational table. The rationale for this is that relational operators can be composed—the results of one operation can always be fed into another operation—but the implications are far-reaching. Relational closure requires that no relational operator return a table with duplicate rows, since a table with duplicate rows is not a valid relational table. A second implication of closure is the need for a nomenclature that will ensure that columns in any derived table are uniquely named. With certain operations this nomenclature gets very complicated. In general, I think closure within the relational operators is a good idea because it will greatly increase the expressive power of a relational language. This power does not come without a cost, and I expect many different opinions on its value. Another questionable extension is the inclusion of four-valued logic. RM/V2 allows two types of nulls: missing but applicable, and missing and not applicable. Granting that RM/V1 does not have systematic treatment for nulls, this extension is complicated, confusing, and counterintuitive. If simplicity and abstraction were goals of RM/V2, they have been badly violated by the four-valued logic. I do not agree with this extension and do not believe many other people will either. Since the conception of the relational data model, other useful data models have been identified, notably the deductive and object-oriented data models. Whether or not they are “data” models in any absolute sense, they both embody powerful ideas. The deductive data model allows the inclusion of inferential rules to extend the information in the database to include those facts that can be derived from existing facts. The object-oriented model allows data objects to have behavior as well as models. Both these concepts could make the relational model much more powerful without sacrificing simplicity or abstraction. Indeed, these ideas increase simplicity and abstraction. Unfortunately, Codd appears to be in a competitive rather than a cooperative mode and criticizes the ideas of others rather than attempting to see their value. This book has several shortcomings. While it appears to be highly systematic, with more than 300 features and 18 classes, it is really quite disorganized. The features are neither orthogonal nor independent. A feature may have implications in other features throughout the book. The discussion of a single concept may be carried from place to place throughout the book, rehashed, and even redefined; it is thus difficult for the reader to pin a concept down well enough to understand it fully. You have to read the book over and over, and then speculate as to what Codd really had in mind. Codd frequently rambles on about an idea rather than just stating it. For example, the concept of duplicate rows arises 13 times in the first chapter alone and many more times in the rest of the book. It would have been better if Codd had made his case and left it at that. I often felt that he was anticipating an argument and kept coming back to certain points in order to strengthen them. I have very mixed feelings about this book. Some of the new features are very good ideas, and some are very bad ideas. The book is badly in need of a good technical editor to organize and clarify the ideas. The book is sometimes lofty and sometimes petty. Nonetheless, I am glad Codd wrote it. With all its shortcomings, it is a very important book. It is not the last word on relational databases, and I do not believe that it will be the standard against which relational databases are compared. I do, however, believe that it will become the standard against which opposing views are compared. Whether you agree with the contents or not, this book is absolutely must reading for anyone seriously interested in relational databases in the 1990s.

Access critical reviews of Computing literature here

Become a reviewer for Computing Reviews.

Recommendations

Model transformation from object relational database to nosql document database.

With the high increase of data growing, NoSQL databases play a key role in storing large amount of data, to gain flexibility and scalability many industries are now replacing their relational and object relational databases by adopting NoSQL database ...

Relational Database Management Systems: The Business Explosion

This special issue (part 2 of a series began with the special issue in October-December 2012) tells the history of how IBM and several new, independent software companies built companies that supplanted the database management system companies and their ...

Model Transformation From Object Relational Database to NoSQL Column Based Database

NoSQL databases play an important role in saving a huge amount of data. To benefit from the advantages of horizontal scalability and flexibility and with the fast data growing many companies are now replacing their traditional database management ...

Export Citations

• Please download or close your previous search result export first before starting a new bulk export. Preview is not available. By clicking download, a status dialog will open to start the export process. The process may take a few minutes but once it finishes a file will be downloadable from your browser. You may continue to browse the DL while the export process is in progress. Download
• Download citation
• Copy citation

We are preparing your search results for download ...

We will inform you here when the file is ready.

Your file of search results citations is now ready.

Your search export query has expired. Please try again.

Browse Course Material

Course info, instructors.

• Prof. Samuel Madden
• Prof. Robert Morris
• Prof. Michael Stonebraker
• Dr. Carlo Curino

Departments

• Electrical Engineering and Computer Science

As Taught In

• Information Technology
• Algorithms and Data Structures
• Data Mining
• Software Design and Engineering

Learning Resource Types

Database systems, lecture 2: the relational model.

In this lecture, we will continue our discussion of data models and database system architecture, looking in more detail at the relational model.

There is a lot of reading for this lecture. You should start early and try to digest it all, as it lays the foundation for much of what is to come. The papers are:

• Stonebraker, Michael, and Joseph Hellerstein. “What Goes Around Comes Around.” In Readings in Database Systems (aka the Red Book), or online here ( PDF ). Read sections 1-4 (if you know something about XML, you may also enjoy reading sections 10 and 11).
• Codd, E. F. “A Relational Model of Data for Large Shared Data Banks.” Communications of the ACM 13, no. 6 (1970): 377-387. (Focus on sections 1.3 and all of section 2.)

You may also find in useful to read pages 57-63 of Ramakrishnan and Gehrke for a brief overview of the relational model.

As you read these papers, think about and be prepared to answer the following questions in Lecture:

• What is the notion of data independence? Why is it important?
• Codd spends a fair amount of time talking about “Normal forms”. Why is it important that a database be stored in a normal form?
• What are the key ideas behind the relational model? Why are they an improvement over what came before? In what ways is the relational model restrictive?
• What, according to Codd, are the most important differences between the “hierarchical” model (as exemplified by systems like IMS) and the relational model that Codd proposes? Make sure you understand what Codd means by “Data Dependencies”.

You are leaving MIT OpenCourseWare