In 2014, author and speaker Tom Asacker presented the popular TED talk entitled “Why TED Talks don't change people's behaviors”—arguing that lasting behavioural change is only truly achievable by appealing to an individual’s identity and self-concept. His message, that “who we think we are, is why we do what we do” is fundamental to Social Identity Theory (Tajfel, 1978; Tajfel & Turner, 1979, 1986)—which posits that we derive an essential part of our self-concept from our group memberships. Rather than being external to the self, our group memberships and their content—including the consensual perceptions of what is normative for a given group—are internalised within the self-concept in the form of social identities (Tajfel, 1981; Turner, 1982, 1985). Properties of the social group, such as group norms, become subjectively interchangeable with personal norms and stereotypes—influencing thought and guiding action. Our social identities therefore provide some insight into who we are, what we think, and what we do (Abrams et al. 1990; Oakes et al. 1994; Reicher, 2001). Social identities are not only important in describing where we sit within our perceived social reality, but also in predicting what we will become; identities have the power to shape and change our future behaviour (Reicher, 2001).

Implicit in both this messaging and social identity theorising is the suggestion that this relationship between social identity and behaviour may not be so straight forward; while it is true who we think we are contributes to how we behave, equally, the way we behave—and the extent to which it is seen to be normative for a given group—guides the groups we identify with. Therefore, our behaviour has implications for our self-concept. The acquisition of self-knowledge in this fashion is fundamental to theories of cognitive consistency—such as self-perception theory (Bem, 1972)—which posit that individuals infer their own attitudes, beliefs, and other internal states from observing their own behaviours and the circumstances under which they occur (Bem, 1972). For example, when a university student repeatedly engages in behaviours seen to be normative of the student identity—such as drinking (Davoren et al. 2016; John & Alwyn, 2014)—they may begin to reflect on this congruency between their own behaviour and the behaviour of other students. Over time, this individual may begin to identify more strongly as a student, as they perceive their behaviour to align with what it means to be a student. Failing to engage in such behaviours might (over time) even lead to disidentification from the group. This reciprocal causality is consistent with theories and research that examine the wider associative network of the self-concept (Greenwald & Pratkanis, 1984; Koffka, 1935), and has been modelled in relation to other social psychological constructs such a attitudes and self-esteem (e.g. the balanced identity design; Greenwald, Banaji, Rudman, et al. 2002, Dunham, 2013).

While traditional and contemporary theories of social identity have focused primarily on different bivariate relationships among social identity, group norms and behaviour, these relationships are yet to be unified in a single, dynamic model that theorises how the interrelations among these self-relevant constructs emerge and are maintained. This thesis presents, and comprehensively tests, a Social Identity Model of Behavioural Associations (SIMBA)—a theoretical and methodological integration of both social (Tajfel, 1978; Tajfel & Turner, 1979; Turner et al. 1987) and balanced identity theorising (Greenwald, Banaji, Rudman, et al. 2002)—which models the relationships among social identification, group norms, and individual-level behaviour as reciprocal associative links in a triangular constellation.

Throughout the thesis, the SIMBA is used explore two overarching research questions. First, we determine whether identity, norms, and behaviour can be measured—not only directly (i.e. via self-report), but also indirectly (i.e. via implicit measures) as cognitive associations among self-group, group-behaviour, and self-behaviour concepts. We also establish whether the three constructs demonstrate a mutually interactive relationship; that is, whether any one of the constructs may be explained by the interaction between the remaining two. Second, we aim to demonstrate reciprocal causality; beyond describing interactive relations among concepts, the SIMBA assumes that each construct is not only explained, but predicted by the remaining two. Therefore, we explore the cognitive processes through which identity, norms and behaviour are created, and determine whether cognitive balance emerges among them. Ultimately, the thesis tests the utility of the SIMBA as a theoretical framework for understanding the relation between, and the formation of, social identity, group norms, and individual-level behaviour.

In Chapter 1, we review the relevant theories that underpin the theoretical and methodological assumptions of the SIMBA, that is, the theories relevant to the social identity approach (i.e. social identity theory and self-categorisation theory) and cognitive consistency (i.e. balance theory and balanced identity theory). These theories are discussed in terms of their origins and fundamental principles, while also highlighting the ways in which they have inspired and contributed to the development of contemporary theories in their respective fields. With regards to the social identity approach, we discuss how traditional social identity theorising has advanced through the development of models that consider social identity and group norms to be mutually interactive and influential in the formation of one another. Regarding theories of cognitive consistency, we discuss the way in which these theories have become increasingly broad in their applicability and explanatory potential—explaining cognitive balance in a variety of contexts such as interpersonal relations, social identity, and intergroup attitudes.

In Chapter 2, we provide a theoretical overview of the SIMBA. The theory’s key predictions regarding measuring associations, creating associations, and changing associations are outlined, alongside the supporting literature that aided in their development. This chapter also discusses the ways in which the SIMBA theoretically and methodologically advances both SIT and BIT by drawing direct comparisons between the theories. This chapter ends with an overview of the thesis, which highlights the specific research questions that will be addressed in each empirical chapter, and the progression of ideas across chapters.

Chapters 3 and 4 focus on testing the utility of the SIMBA for the measurement of self-group-behaviour associations. Across these chapters, we investigate the primary assumption of the SIMBA (i.e. the balance-congruity principle; that the strength of any one association in the SIMBA can be predicted by the interactive strength of the remaining two) in relation to a variety of group memberships and behaviours—establishing the generalisability of the SIMBA.

In Chapter 3, we tested the SIMBA in the context of drinking behaviour in relation to student (Studies 1 and 2) and British national (Study 3) identities. These studies identified good support for the balance-congruity principle on implicit measures of social identity, group norms, and behaviour. Evidence for this prediction was weaker on explicit scale measures; we argue that this difference may be dependent on the explicit measures possessing theoretically meaningful zero-points.

In Chapter 4, we tested the SIMBA in the context of social distancing behaviour in relation to the British national identity. Study 4 included two different types of explicit measure (e.g. traditional Likert-type scales and visual analogue scales); confirming that support on explicit measures was strong, and equal to that on implicit measures, when using visual analogue scales (i.e. measures that possess a zero-point indicative of associative indifference). We also assessed balance-congruity across two timepoints during the COVID-19 pandemic—finding cognitive balance to be stable over time.

In Chapter 5, we step back from directly testing the assumptions of the SIMBA and adopt a network analytic approach to understanding the different social identities found to co-occur within individuals, and the behaviours thought to correspond with specific identities. This research was initially developed to help identify associated groups and behaviours to feature in future tests of the SIMBA—particularly for the measurement of associations, following on from Chapters 3 and 4—but the co-occurrences identified also serve as a meaningful resource for those conducting research into identities and their associated behaviours more generally. Therefore, while the chapter may appear to divert from the SIMBA narrative, it has an important function in progressing the theory (e.g. establishing its wider generalisability to multiple groups and behaviours), and social identity research more broadly. In Study 5, we generated three different network-types—identity-by-identity, behaviour-by-identity, and identity-by-behaviour—where we identified several identities and behaviours to significantly co-occur at a rate both higher and lower than chance. Although the overall structure of the networks was found to be significantly different than expected by chance, networks were low in modularity (i.e. there was no evidence of clustering within the data).

In the final empirical chapter (i.e. Chapter 6) we focus on testing the utility of the SIMBA for the creation of novel self-group-behaviour associations. In two pilot studies, we established the initial efficacy of both the partial-IAT and probabilistic learning tasks for generating associations among concepts. After refining these methodologies, Study 6 directly compared them, and conducted an experimental test of the balance-congruity principle. While being equivalent to the probabilistic learning methodology in its ability to create self-group-behaviour associations, and generate balance-congruity among explicit associations, the partial-IAT was less effective in generating balance-congruity among implicit associations. We therefore find the probabilistic learning methodology to be the most appropriate of the two. Study 7 then examines changes in associative strength across three timepoints—finding associations to be stable across time, irrespective of whether participants were exposed to a ‘booster session’ of the probabilistic learning manipulation.

In Chapter 7, we discuss all of the empirical results taken together. We synthesise, and evaluate, the strength of evidence in relation to the four key predictions of the SIMBA outlined in Chapter 2. We also discuss the key theoretical and methodological contribution of our research, while acknowledging some of the limitations that persist across empirical chapters. Finally, we discuss avenues for future research regarding the use of the SIMBA not only for the measurement and creation of self-group-behaviour associations, but also as a framework for changing associations. In doing so, we highlight some of the practical implications of the model.