Final Report Summary - GEC (A model of generalised evaluative conditioning)
(i). A summary of the project objects
Evaluative learning has been implicated in the development of preferences and aversions. It has been suggested that evaluative learning is also important in the development of many clinical problems (e.g. depression). In these latter areas, the generalisation of negative evaluative, and particularly self-evaluative, judgements across many aspects of a person's life may be a serious problem for everyday functioning. Given the applied relevance of this work, it is essential that the basic mechanisms of self-evaluative learning, as well as how learned self-evaluations become generalised across a range of situations, is investigated comprehensively. It has previously been shown that exposure to certain schedule conditions will lead to particular self-efficacy ratings, and that stimuli can also gain a particular valence through their relation/ association with such situations leading to generalisation of the ratings. Previous work examined how psychological functions can be changed through sameness (equivalence) relations. The current proposal will extend this work by: 1) examining the potential to develop a model of generalised evaluative functions resulting from a combination of the literatures and techniques from evaluative learning and derived relational responding; 2) investigating the impact of the structure of such classes on generalisation, including whether such transfer of function reliably emerges across non-sameness relations. The latter is a process that potentially allows much greater, and, perhaps, more varied routes of generalisation of evaluative learning, than transfer through relations of sameness alone. A coherent understanding of how self evaluations, and judgements of efficacy, are learned, maintained, and generalised, would be both informative at a theoretical level, as well as for the applied domain, providing a useful model of the behaviours that need to be targeted in order to circumvent the generalisation of negative self evaluations.
(ii). A description of the work performed since the beginning of the project
Five experiments investigating the transformation of functions via derived stimulus relations have been completed. Experiment 1 involved comparing a standard match-to-sample procedure was compared with an errorless procedure. Experiment 2 looked at alternating the presentation and feedback of stimulus pairs in respondent training. Broadly speaking, the findings from Experiments 1 and 2 support previous literature that stimuli that were associated with schedules of reinforcement that produced related either good or bad causal evaluations were later categorised with stimuli that had previously been established as having the same function (‘good’ or ‘bad’) through stimulus equivalence classes. These findings have the potential to account for the processes involved in disorders such as depression in which the negative evaluations spread to many areas (e.g. people and events) of an individual’s life. Just as causal evaluations can transfer through equivalence classes, it is possible that so too can feelings of exclusion. Such findings would have important implications for our understanding of how exclusion from one activity (e.g. an online game) may generalise to other contexts or activities in an individual’s life without the person having direct exclusion from these activities (e.g. to similar sports or activities such as basketball or tennis).
Experiments 3 and 4 aimed to determine whether the functions of ostracism (feelings of exclusion and inclusion) would transfer across equivalence classes. That is, would participants expect to feel excluded from (or included in) a new game based on the fact that this game was in an equivalence class with a game from which they were directly excluded.
To that end, Experiments 3 and 4 explored the transfer of feelings of exclusion and inclusion through stimulus equivalence classes. Previous studies have found that evaluations and social exclusion can cause distress. One method used to study social exclusion is through a virtual ball-toss game known as Cyberball. In this game, participants may be excluded from or included in the ball-toss game and typically report lower feelings of self-esteem, control, belonging and meaningful existence following exclusion. In both experiments, participants were trained to form two three-member equivalence classes (e.g. A1-B1, B1-C1; A2-B2, B2-C2) and tested with novel stimulus combinations (A1-C1, C1-A1, A2-C2, C2-A2). Thereafter, participants were exposed to the Cyberball exclusion and inclusion games. In this game, one stimulus (C1) from one equivalence class was assigned as the Cyberball inclusion game name, while one stimulus (C2) from the other equivalence class was assigned as the Cyberball exclusion game name. In Experiment 4, participants were only exposed to the Cyberball exclusion game. During a subsequent transfer test, participants were asked to rate how included or excluded they thought they would be in other online games, corresponding to members of both equivalence classes. Experiments 3 and 4 above have been accepted for publication in Learning and Behavior (see Munnelly, Martin, Dack, Zedginidze, McHugh, 2014 attached).
Experiment 5 aimed to extend Munnelly et al., (2014) by assessing for generalization including transformation of exclusion functions via same and opposite relations respectively. Recent research has provided support for the Relational Completion Procedure (RCP) as an efficient alternative to the more widely used match-to-sample procedure for establishing Same and Opposite relations with adult humans. Therefore, Experiment 5 employed a non arbitrary relational training and testing procedure to establish contextual cue functions in arbitrary shape stimuli. Next the established contextual cues were employed to train and test two separate arbitrary relational networks, both involving arbitrary nonsense syllables in both trained and derived same and opposite relations. Thereafter, using a procedure similar to that employed in Munnelly et al (2014), participants were exposed to the cyberball game where one of the members of the relational network taught in the RCP training will be the cyberball game name. In the final phase of the procedure, participants were required to rate how excluded they would be on a scale of 1 to 7 from a number of different games. It was predicted that participants would rate themselves as likely to be excluded from game names related to the cyberball game name. This effect should not emerge for game names that are trained during the RCP as opposite to the cyberball game.
(iii). A description of the main results achieved so far
Results of Experiment 1 indicate that the errors made during training led to the stimuli acquiring aversive properties which interfered with positive transfer and facilitated negative transfer demonstrating an important role for negative thoughts in the generation of spread of negative evaluations.
Results of Experiment 2 suggested that lower ratings of causal efficacy were and more negative transfer emerged for a match to sample rather than a positive only respondent group.
Results from Experiments 3 & 4 indicated that participants felt they would be excluded from online games if they were members of the same equivalence class as C2. In contrast, participants reported that they felt they would be included in online games if they were members of the same equivalence class as C1. Results indicated the transfer of feelings of inclusion (Experiment 3) and feelings of exclusion (Experiments 3 and 4) through equivalence classes.
The results from Experiment 5 showed the derived generalisation of exclusion functions from the game name to words related as the same as the game name and derived inclusion to words related as opposite to the game name.
(iv). The expected final results and their potential impact and use (including the socio-economic impact and the wider societal implications of the project so far).
Findings from Experiments 1-4 suggest that stimulus equivalence provides a useful theoretical framework for understanding the derived generalization of negative evaluative judgements (e.g. self efficacy and exclusion). This has important implications for the understanding and remediation of clinical disorders. In addition to promoting basic understanding of the derived generalization of negative evaluative judgements and exclusion in a new paradigm the current work also provides insight into the factors (e.g. perceived lack of control) in the environment that affect social exclusion. Experiments 3 & 4 explored the effects of exclusion that are potentially important to understanding clinical disorders such as depression. For instance, the current findings may bear relevance to the literature on learned helplessness in which an individual learns to behave helplessly due to a perceived lack of control over the outcome of a situation (Seligman, 1975). That is, following repeated exposure to an aversive situation from which an individual cannot escape, feelings of helplessness may govern behaviour. Later, when the individual is presented with the opportunity to escape the aversive situation, they are unable to do so due to this perceived lack of control. With respect to the current findings, following exposure to the Cyberball exclusion game, participants reported a decreased sense of “control”. The current findings may therefore be important in the sense that the feelings of lack of control reported following exclusion on the Cyberball game may, generalize to other situations (e.g. work and personal life) in an individual’s life. The current results may also inform us about the problems experienced by people that can occur without direct experience (e.g. a fear of spiders without ever being in contact with one). Furthermore, and as demonstrated in the current experimental series, exclusion attempts can transfer to stimuli that are closely related to the targeted stimulus, such generalisation could cause a negative cascade that would aggravate disorders such as depression (Walther, Nagengast & Trasselli, 2005).
Experiment 5 examined the effects of varying the instructions given to the participants when they are rating their level of exclusion functions. In complex human performance, there are many rules that may be derived, and which could impact on the generalisation of exclusion functions, especially in clinically-relevant situations. Experiment 5 involved examining whether different patterns of relational responding resulted in ‘derived exclusion’. For example, previous research has demonstrated derived relational responding in accordance with multiple stimulus relations such as ‘distinction,’ ‘hierarchy,’ ‘conditionality,’ ‘causality,’ and ‘opposition’ (e.g. Dymond, & Barnes, 1995; Dymond, Roche, Forsyth, Whelan, & Rhoden, 2008; Gil, Luciano, Ruiz, & Valdivia-Salas, 2012; Steele & Hayes, 1991). One example that may be particularly interesting in respect to the current phenomenon are the relations of ‘opposition’. Expanding the model from equivalence relations to multiple stimulus relations bolsters the derived stimulus relations’ account of the generalisation of exclusion. In turn, it provides additional dimensions to a model of how exclusion (e.g. how ostracised an individual is socially) are produced by certain contingencies, and how they can generalise to other stimuli related to the initial exclusion episode.
The results from the Experiment 5 provided the first demonstration of derived exclusion across relations of opposition. If a person rates themselves as excluded at one task, and views this task as opposite to another task, then they derive that they will also be excluded from the second task. The generalisation of ratings of exclusion across multiple stimulus relations is particularly interesting with the rapid increase of opportunities for exclusion even when physically alone via the internet. Modelling such generalisation and developing a bottom up account of such generalisation will help inform the development of interventions that can both prevent and remediate the negative effects of generalised exclusion. The current research supplements previous work on transformation of function from Relational Frame theorists. At this point there is a substantial quantity of work demonstrating the derivation of multiple different relations and the transformation of multiple different functions via that multiplicity of different relations. The current work supplements previous work on the transformation of functions via same and opposite (e.g. Barnes-Holmes, Barnes-Holmes & Smeets, 2004; Dymond & Barnes, 1995; Dymond, Roche, Forsyth, Whelan & Rhoden, 2007).
Evaluative learning has been implicated in the development of preferences and aversions. It has been suggested that evaluative learning is also important in the development of many clinical problems (e.g. depression). In these latter areas, the generalisation of negative evaluative, and particularly self-evaluative, judgements across many aspects of a person's life may be a serious problem for everyday functioning. Given the applied relevance of this work, it is essential that the basic mechanisms of self-evaluative learning, as well as how learned self-evaluations become generalised across a range of situations, is investigated comprehensively. It has previously been shown that exposure to certain schedule conditions will lead to particular self-efficacy ratings, and that stimuli can also gain a particular valence through their relation/ association with such situations leading to generalisation of the ratings. Previous work examined how psychological functions can be changed through sameness (equivalence) relations. The current proposal will extend this work by: 1) examining the potential to develop a model of generalised evaluative functions resulting from a combination of the literatures and techniques from evaluative learning and derived relational responding; 2) investigating the impact of the structure of such classes on generalisation, including whether such transfer of function reliably emerges across non-sameness relations. The latter is a process that potentially allows much greater, and, perhaps, more varied routes of generalisation of evaluative learning, than transfer through relations of sameness alone. A coherent understanding of how self evaluations, and judgements of efficacy, are learned, maintained, and generalised, would be both informative at a theoretical level, as well as for the applied domain, providing a useful model of the behaviours that need to be targeted in order to circumvent the generalisation of negative self evaluations.
(ii). A description of the work performed since the beginning of the project
Five experiments investigating the transformation of functions via derived stimulus relations have been completed. Experiment 1 involved comparing a standard match-to-sample procedure was compared with an errorless procedure. Experiment 2 looked at alternating the presentation and feedback of stimulus pairs in respondent training. Broadly speaking, the findings from Experiments 1 and 2 support previous literature that stimuli that were associated with schedules of reinforcement that produced related either good or bad causal evaluations were later categorised with stimuli that had previously been established as having the same function (‘good’ or ‘bad’) through stimulus equivalence classes. These findings have the potential to account for the processes involved in disorders such as depression in which the negative evaluations spread to many areas (e.g. people and events) of an individual’s life. Just as causal evaluations can transfer through equivalence classes, it is possible that so too can feelings of exclusion. Such findings would have important implications for our understanding of how exclusion from one activity (e.g. an online game) may generalise to other contexts or activities in an individual’s life without the person having direct exclusion from these activities (e.g. to similar sports or activities such as basketball or tennis).
Experiments 3 and 4 aimed to determine whether the functions of ostracism (feelings of exclusion and inclusion) would transfer across equivalence classes. That is, would participants expect to feel excluded from (or included in) a new game based on the fact that this game was in an equivalence class with a game from which they were directly excluded.
To that end, Experiments 3 and 4 explored the transfer of feelings of exclusion and inclusion through stimulus equivalence classes. Previous studies have found that evaluations and social exclusion can cause distress. One method used to study social exclusion is through a virtual ball-toss game known as Cyberball. In this game, participants may be excluded from or included in the ball-toss game and typically report lower feelings of self-esteem, control, belonging and meaningful existence following exclusion. In both experiments, participants were trained to form two three-member equivalence classes (e.g. A1-B1, B1-C1; A2-B2, B2-C2) and tested with novel stimulus combinations (A1-C1, C1-A1, A2-C2, C2-A2). Thereafter, participants were exposed to the Cyberball exclusion and inclusion games. In this game, one stimulus (C1) from one equivalence class was assigned as the Cyberball inclusion game name, while one stimulus (C2) from the other equivalence class was assigned as the Cyberball exclusion game name. In Experiment 4, participants were only exposed to the Cyberball exclusion game. During a subsequent transfer test, participants were asked to rate how included or excluded they thought they would be in other online games, corresponding to members of both equivalence classes. Experiments 3 and 4 above have been accepted for publication in Learning and Behavior (see Munnelly, Martin, Dack, Zedginidze, McHugh, 2014 attached).
Experiment 5 aimed to extend Munnelly et al., (2014) by assessing for generalization including transformation of exclusion functions via same and opposite relations respectively. Recent research has provided support for the Relational Completion Procedure (RCP) as an efficient alternative to the more widely used match-to-sample procedure for establishing Same and Opposite relations with adult humans. Therefore, Experiment 5 employed a non arbitrary relational training and testing procedure to establish contextual cue functions in arbitrary shape stimuli. Next the established contextual cues were employed to train and test two separate arbitrary relational networks, both involving arbitrary nonsense syllables in both trained and derived same and opposite relations. Thereafter, using a procedure similar to that employed in Munnelly et al (2014), participants were exposed to the cyberball game where one of the members of the relational network taught in the RCP training will be the cyberball game name. In the final phase of the procedure, participants were required to rate how excluded they would be on a scale of 1 to 7 from a number of different games. It was predicted that participants would rate themselves as likely to be excluded from game names related to the cyberball game name. This effect should not emerge for game names that are trained during the RCP as opposite to the cyberball game.
(iii). A description of the main results achieved so far
Results of Experiment 1 indicate that the errors made during training led to the stimuli acquiring aversive properties which interfered with positive transfer and facilitated negative transfer demonstrating an important role for negative thoughts in the generation of spread of negative evaluations.
Results of Experiment 2 suggested that lower ratings of causal efficacy were and more negative transfer emerged for a match to sample rather than a positive only respondent group.
Results from Experiments 3 & 4 indicated that participants felt they would be excluded from online games if they were members of the same equivalence class as C2. In contrast, participants reported that they felt they would be included in online games if they were members of the same equivalence class as C1. Results indicated the transfer of feelings of inclusion (Experiment 3) and feelings of exclusion (Experiments 3 and 4) through equivalence classes.
The results from Experiment 5 showed the derived generalisation of exclusion functions from the game name to words related as the same as the game name and derived inclusion to words related as opposite to the game name.
(iv). The expected final results and their potential impact and use (including the socio-economic impact and the wider societal implications of the project so far).
Findings from Experiments 1-4 suggest that stimulus equivalence provides a useful theoretical framework for understanding the derived generalization of negative evaluative judgements (e.g. self efficacy and exclusion). This has important implications for the understanding and remediation of clinical disorders. In addition to promoting basic understanding of the derived generalization of negative evaluative judgements and exclusion in a new paradigm the current work also provides insight into the factors (e.g. perceived lack of control) in the environment that affect social exclusion. Experiments 3 & 4 explored the effects of exclusion that are potentially important to understanding clinical disorders such as depression. For instance, the current findings may bear relevance to the literature on learned helplessness in which an individual learns to behave helplessly due to a perceived lack of control over the outcome of a situation (Seligman, 1975). That is, following repeated exposure to an aversive situation from which an individual cannot escape, feelings of helplessness may govern behaviour. Later, when the individual is presented with the opportunity to escape the aversive situation, they are unable to do so due to this perceived lack of control. With respect to the current findings, following exposure to the Cyberball exclusion game, participants reported a decreased sense of “control”. The current findings may therefore be important in the sense that the feelings of lack of control reported following exclusion on the Cyberball game may, generalize to other situations (e.g. work and personal life) in an individual’s life. The current results may also inform us about the problems experienced by people that can occur without direct experience (e.g. a fear of spiders without ever being in contact with one). Furthermore, and as demonstrated in the current experimental series, exclusion attempts can transfer to stimuli that are closely related to the targeted stimulus, such generalisation could cause a negative cascade that would aggravate disorders such as depression (Walther, Nagengast & Trasselli, 2005).
Experiment 5 examined the effects of varying the instructions given to the participants when they are rating their level of exclusion functions. In complex human performance, there are many rules that may be derived, and which could impact on the generalisation of exclusion functions, especially in clinically-relevant situations. Experiment 5 involved examining whether different patterns of relational responding resulted in ‘derived exclusion’. For example, previous research has demonstrated derived relational responding in accordance with multiple stimulus relations such as ‘distinction,’ ‘hierarchy,’ ‘conditionality,’ ‘causality,’ and ‘opposition’ (e.g. Dymond, & Barnes, 1995; Dymond, Roche, Forsyth, Whelan, & Rhoden, 2008; Gil, Luciano, Ruiz, & Valdivia-Salas, 2012; Steele & Hayes, 1991). One example that may be particularly interesting in respect to the current phenomenon are the relations of ‘opposition’. Expanding the model from equivalence relations to multiple stimulus relations bolsters the derived stimulus relations’ account of the generalisation of exclusion. In turn, it provides additional dimensions to a model of how exclusion (e.g. how ostracised an individual is socially) are produced by certain contingencies, and how they can generalise to other stimuli related to the initial exclusion episode.
The results from the Experiment 5 provided the first demonstration of derived exclusion across relations of opposition. If a person rates themselves as excluded at one task, and views this task as opposite to another task, then they derive that they will also be excluded from the second task. The generalisation of ratings of exclusion across multiple stimulus relations is particularly interesting with the rapid increase of opportunities for exclusion even when physically alone via the internet. Modelling such generalisation and developing a bottom up account of such generalisation will help inform the development of interventions that can both prevent and remediate the negative effects of generalised exclusion. The current research supplements previous work on transformation of function from Relational Frame theorists. At this point there is a substantial quantity of work demonstrating the derivation of multiple different relations and the transformation of multiple different functions via that multiplicity of different relations. The current work supplements previous work on the transformation of functions via same and opposite (e.g. Barnes-Holmes, Barnes-Holmes & Smeets, 2004; Dymond & Barnes, 1995; Dymond, Roche, Forsyth, Whelan & Rhoden, 2007).