Thought Action Fusion

Introduction

Over the past decade, the study of underlying cognitive processes in Obsessive Compulsive Disorder (OCD) has led to the introduction of a new construct into the field. When Thought- Action Fusion (TAF) first received empirical attention (Rachman, Thordarson, Shafran, & Woody, 1995), it was studied in terms of its relation to perceived responsibility in individuals with OCD; subsequent research studied TAF and its relation to other concepts like neutralization and thought suppression. To date, TAF has not been investigated as a phenomenon in its own right. Thus, the present research aims to fulfill this void in the field.

Thought- Action Fusion, as its name suggests, refers to the tendency to see a "thought" and an "action" to be inextricably linked (i.e., "fused together"). Specifically, a person who has TAF believes that an internal thought increases the probability that an event will occur in the external world. This cognitive bias consists of three aspects: 1) TAF Likelihood- Self refers to the belief that a particular thought increases the probability that an aversive event will happen to oneself (e.g., "if I think about dying, I'll be more likely to die"), 2) TAF Likelihood- Others refers to the belief that a thought increases the probability that the event will happen to others (e.g., "if I think about dying, my beloved friend will be more likely to die"), and 3) TAF Moral refers to the belief that unacceptable thoughts are morally equivalent to overt actions (e.g., "I must be a bad person if I think about murdering my own mother because it is the same as actually murdering her).

The goal of the proposed research is to investigate whether TAF beliefs can be created in an experimental setting using the illusory correlation (IC) paradigm. The following discussion will first illuminate the motivations for studying TAF development in the laboratory. This will be followed by an explanation on the rationale for combining TAF and IC in this proposal.

Discussion

Why study Thought- Action Fusion?

In a recent review, Berle and Starcevic (2005) noted that there is an overall dearth of research on factors which may contribute to the development of TAF. This is justified in light of the fact that this subfield of research in cognitive biases is extremely young. Furthermore, TAF is also most likely to develop from a complex multitude of subtle and interacting factors occurring over many years (Salkovskis, Shafran, Rachman, & Freeston, 1999), which makes it difficult to isolate and tease apart all factors involved.

While it is critical to note what is currently lacking in TAF research, past attempts should not be dismissed. The few studies which did look into the development and maintenance of TAF investigated related factors such as neutralization (Rachman, Shafran, Mitchell, Trant, & Teachman, 1996; van den Hout, Kindt, Weiland, & Peters, 2002; van den Hout, van Pol, & Peters, 2001; Zucker, Craske, Barrios, & Holguin, 2002), inflated responsibility (Shafran, Thordarson, & Rachman, 1996; Sma´ri, & Ho´lmsteinsson, 2001; Zucker et al., 2002), and thought suppression (Rassin, 2001; Sma´ri, J., & Ho´lmsteinsson, 2001). In addition, the hypothesis that TAF may arise from strict belief systems and behavioural codes is supported by correlations between TAF and religiosity (Rassin & Koster, 2003) and between religiosity and OCD symptoms (Hutchinson, Patock-Peckham, Cheong, & Nagoshi, 1998; Sica, Norvara, & Sanavio, 2002). Finally, Salkovskis et al. (1999) have postulated that TAF may develop following a specific event. For instance, the coincidental occurrence of a thought and a significant event may lead an individual to perceive an erroneous causal link between the two. Indeed, Tallis (1994) described two individuals who developed OCD symptoms following episodes in which each wished ill upon a relative who coincidentally died soon after. Similarly, Rheaume, Freeston, Leger, and Ladouceur (1998) also described a case where a powerful coincidental event precipitated TAF- like beliefs. It is important to note that such events, in themselves, are not always sufficient to promote TAF. However, the latter is perhaps possible with the presentation of multiple coincidences, or when combined with preexisting attitudes or beliefs (Berle & Starcevic, 2005).

In sum, although previous research has outlined several related factors of TAF, a dearth of research clearly exists. The question of whether TAF can be created using the IC paradigm then becomes especially important because of the advantages it confers. Firstly, it potentially informs if TAF can be learnt via the same mechanism as IC, which contributes to research on the factors and mechanisms in TAF development. Secondly, by isolating the TAF phenomenon in the laboratory, researchers are provided with a potentially valid method of inducing TAF. This is beneficial to those who aim to study TAF and its related factors in a controlled setting.

Apart from the lack of sufficient research on the causative conditions of TAF, another motivation for this proposed study stems from flaws in past studies. In their review, Shafran and Rachman (2004) provided a comprehensive summary of 23 TAF- related studies conducted between 1995 and 2004. Of the 23 studies, one was an illustration of 2 case studies, 16 were correlational, and the remaining 6 involved experimental manipulations of TAF in laboratory settings. It is important to point out that TAF studies to date have relied mostly on correlational data. Thus, there is a pressing need to devote more resources to study TAF in well- controlled experiments. Also, while previous attempts to evoke TAF in experimental situations should be applauded, it is critical that the flaws inherent in these methods be recognized. Two of these methodological flaws are germane to the rationale behind this proposed study, and each flaw will be given separate treatment in the following discussion.

All except one of the six experimental studies attempted to induce TAF in the laboratory using a task called the "Sentence Paradigm" (Rachman et al., 1996). Before the task is implemented, baseline anxiety levels of the participants are established and efforts are taken to ensure that they are sufficiently relaxed (i.e., the level of anxiety is below a predetermined level). They are instructed to keep in mind a close friend or relative, and then write the name of the friend/ relative on a paper containing the following sentence: "I hope _________ is in a car accident". Following the writing, they are told to close their eyes and think about the scenario for a few seconds. After this task is completed, their anxiety levels are measured again. Most studies replicated Rachman et al.'s (1996) finding that TAF, as assumed to be induced by the sentence paradigm, led to an increase in anxiety and also the tendency to generate an obssessional- like thought.

The most critical problem with these studies is that there are usually no control groups. For example, in the original study (Rachman et al., 1996), the authors explicitly stated that they predict participants high in TAF to be "more likely to respond to a manipulation designed to generate an obssessional- like thought" compared to those low in TAF. However, the participants in this study consisted only of individuals who scored high on a scale measuring TAF. Since they did not compare the results with a control group of low TAF scorers, their argument breaks down completely. Further, an increase in anxiety level after the task is not a good indicator of whether TAF was actually elicited in the participants. This increase in anxiety can be explained by the fact that the sentence paradigm had successfully induced a distressful thought. It does not necessarily indicate that the participants actually believed that their thought increased the probability that a friend/ relative will be in an accident shortly (i.e., TAF).

The one study which did not use the sentence paradigm attempted to evoke TAF in a very

different manner (Rassin, Merckelbach, Muris, & Spaan, 1999). Participants were each strapped to a bogus EEG machine. In an effort to evoke the belief that their thoughts can cause an event to occur in the real- world, the experimental group was told that the machine was able to read simple thoughts like "apple" and then administer a shock to a "co- participant" of the experiment in another room. The control group was simply told that the researchers aimed to find out if the machine was sensitive to read simple thoughts like "apple". As predicted, the experimental group showed more obssessional- like thoughts. As innovative as this procedure is, there are several critical flaws in this study. Firstly, because the sample consisted of naïve high- school students and most of them were female, the results cannot be extended to people who suffer from OCD in the real world. It is probably much more difficult to convince the general population that there exists a machine that can read thoughts. The pattern of cognition in females and that of males are also likely to be different. The most critical flaw in this study comes from the way TAF was implicitly presented. As defined in the literature and as observed in actual clinical practice, TAF is essentially a cognitive bias. This means that despite what reality is (i.e., thinking about death will not increase the probability of death in oneself or another), people who suffer from TAF and OCD hold onto such beliefs with great conviction. This study, however, designed the apple thought- shock contingency in a manner in which an apple thought most likely led to a shock. The researchers were actually evoking an objective belief and not a cognitive bias.

In light these flaws, the investigation of whether TAF can truly be evoked experimentally is then extremely significant. Past researchers have attempted to do so, and they have built their arguments about the roles of several important factors (e.g., neutralization, inflated responsibility etc) in maintaining TAF/OCD based on the assumption that they have indeed elicited TAF in their studies. However, the aforementioned flaws cause a breakdown in this assumption. Hence, any subsequent theory, however fancy, becomes invalid. In sum, if the proposed study is able to present a truly valid way of inducing TAF experimentally, future researchers can capitalize on this to further study the concepts which have been postulated to relate significantly to TAF.

Besides elucidating more on the causes and mechanisms of TAF development and also providing researchers with a plausible way to induce TAF in the laboratory, the study of TAF in general already holds great significance. This is because while TAF has been found to occur in OCD, there is increasing evidence that it may also be prevalent in other disorders like Generalized Anxiety Disorder (GAD), social anxiety disorder, depression, eating disorders, and psychotic disorders (Berle & Starcevic, 2005). The significance of these disorders in clinical practice firmly justifies why TAF research should be given greater attention.

In conclusion, the rationale for this proposal stems from three sources. Firstly, although TAF is a very new area of interest, its apparent role in many disorders makes it a worthwhile candidate for experimental investigation. Secondly, the lack of research in such an important area is, by itself, already an incentive to motivate more research. Lastly, no study in the existing literature presented earlier has met up to the rigour and stringency of scientific examination. The design of the present study aims to address all three concerns.

Why the link between TAF and IC?

The aim of this research is to offer a technique to elicit TAF experimentally. The proposal is to use the Illusory Correlation (IC) paradigm to induce TAF.

IC is a form of cognitive bias first identified by Chapman (1967). It is a misperceived correlation between two events which are not correlated or are correlated to a much lesser extent in the real world. He demonstrated that the co- occurrence of two stimuli with strong semantic

association (e.g., bacon- eggs) or the co- occurrence of two statistically infrequent stimuli leads

to an overestimation of the frequency of co- occurrence (i.e. IC).

The rationale for bringing TAF and IC together in this proposal is based on the following three observations. Firstly, both concepts deal with the idea of correlation as TAF involves a causative thought- action relationship and IC involves a relationship between two events. Secondly, TAF is a cognitive bias in that the causative relationship between thought and action has no basis in reality; IC is also a cognitive bias in that the perception that two events always occur together is not realistic. Finally, as mentioned earlier, TAF appears to play a contributive role in anxiety disorders; IC has also been found to play a role in many anxiety disorders (de Jong, Merckelbach, Bogels, & Kindt, 1998; de Jong, Merckelbach, & Nijman, 1995; Hermann, Ofer, & Flor, 2004; Kennedy, Rapee, & Mazurski, 1997; Pauli, Montoya, & Martz,1996; Tomarken, Mineka, & Cook, 1989; Tomarken, Sutton, & Mineka, 1995). In light of the similarities between these two concepts, the proposed study postulates that the causal relationship which exists between a thought and an action (i.e., TAF) can be learnt via the same mechanism in which individuals learn a relationship between two events (i.e., IC). Essentially, TAF is a specific elaboration of IC, and thus a paradigm which elicits an IC belief has the potential to elicit a TAF belief.

The link between TAF and IC has never been drawn explicitly, but existing research shows some indicative support. As mentioned earlier, the coincidental occurrence of a thought and a significant event may lead to misperception of a strong causal link between the two (Rheaume et al., 1998; Salkovskis et al., 1999; Tallis, 1994). It is likely that this is even more probable if the thought and action are both statistically infrequent and semantically related. Lieberman (1999) also demonstrated that people are more likely to overestimate of the co- occurrence of two stimuli during a threatening or fearful state of mind. This suggests that people might be more likely to wrongly conclude that a thought always leads to an event if that thought itself is fear- inducing or threatening. Indeed, this seems to be the case for TAF because the thoughts involved are anxiety- provoking in nature.

As stated above, Chapman (1967) identified IC in pairs of semantically associated or statistically infrequent stimuli. The idea of statistical infrequency as a basis for forming ICs has been extended by Hamilton and Gifford (1976). Since the methodology of this proposal is modelled after their work, it is necessary to include a description of their original study.

In this heavily cited study, participants were presented with a series of statements which described socially desirable or undesirable behaviours performed by members of Group A or Group B. Since two- thirds (26/39) of the statements described Group A and the rest (13/39) described Group B, Group B was the minority group. In addition, two- thirds of the statements depicted desirable behaviours while the remaining one- third depicted undesirable ones. This was true for both groups, which means that the proportion of desirable to undesirable statements was the same for both groups. Hence, there was no relationship between group membership and desirability of behaviour. After reading the statements, participants were assessed on three measures- trait rating (rate members as high or low on desirable traits), attribution of group membership (indicate the membership of person who has performed a behaviour), and frequency estimate (estimate how many of the statements which described Group A and B members depicted undesirable behaviour).

The results revealed that the trait rating and attribution measures were consistent with the hypothesis that the minority, Group B, was evaluated less positively than the majority group. In other words, two statistically rare events (i.e., Group B and undesirable behaviour) were misperceived to co- occur more often than it objectively does. Conversely, when desirable behaviours were made infrequent, the minority group was judged more positively than the majority group. Hence, the researchers concluded that the co- occurrence of two rare events leads to the formation of a biased, illusory correlation.

The methodology for the present proposal will be described in the following section. Although it will employ the same paradigm as that used in Hamilton and Gifford (1976), there is a change to the dependent variables because several concerns regarding the validity of the original measures have been raised. McGahan and Wight (1989) replaced them with propositional statements and responses to such statements have been demonstrated to contain high face validity as a dependent measure (Yates, McGahan, & Williamson, 2000). Thus, the proposed method described below will employ propositional statements modelled after the work of Yates et al. (2000).

Method

Procedure

Participants will be tested individually. Upon entering the laboratory, each participant will be acquainted with another "co- participant" and is told that both of them are paired in this experiment. The "co- participant" is, of course, a confederate of the experimenter.

Both the participant and the confederate will be informed that they are part of a pilot project to test the ability of a new machine to administer random shocks because this particular function is critical in the post- pilot project. The rationale for this is to lead the participant to believe that the machine is originally pre- programmed to administer shocks randomly. The experimenter will go on to explain that the participant has been randomly assigned to be the person who performs a simple cognitive task while the confederate will be the one who receives the shocks in another room. After the confederate leaves the room, the experimenter then provides the instructions of the task as shown below:

"In this experiment, you will be shown a sequence of individual words which represent objects. At the presentation of each word, your task is to try your best to conjure up a vivid image of the object that the word represents. Upon attaining the most vivid image that you possibly can, respond by pressing the "enter" key on the keyboard.

With each of response that you make, a shock may or may not be administered to the other person in this experiment. If a shock is administered, you will hear some static- like noise through the headphones; if no shock is administered, you will not hear anything. Recall that our purpose is to investigate whether the machine is able to randomize shocks. As you will be asked to report the frequency of shock administration after this task, it is necessary for you to keep the headphones on at all times."

The participant is also told that there is nothing inherently special about this task as it is just one of the many cognitive tasks which will be used in the experiment. He or she will then be led into a booth and asked to put the headphones on before the actual task begins.

Stimulus materials

The presentation stimuli will consist of a total of 60 words from two categories- plant and furniture. Since a between- group design is proposed, there will be two conditions in this experiment. In condition- F, participants will be presented with 40 plant words and 20 furniture words; in condition- P, participants will be presented with 40 furniture words and 20 plant words. In other words, the furniture category will be the infrequent category in condition- F and the plant category will be the infrequent category in condition- P.

In condition- F, 32 plant words will be paired with no shock and 8 plant words will be paired with shock; 16 furniture words will be paired with no shock and 4 furniture words will be paired with shock. In condition- P, 32 furniture words will be paired with no shock and 8 furniture words will be paired with shock; 16 plant words will be paired with no shock and 4 plant words will be paired with shock. Since the ratio of no shock to shock is 4:1 across both conditions and across both categories of words, the shock outcome is always a statistically infrequent event.

Dependent Variables

At the end of the task, the participant will be administered a set of propositional statements and is asked to respond true or false to these statements. As recommended by Yates et al. (2000), the propositional statements should fall into three categories: 1) IC- consistent (correlation is found to be consistent with the hypothesized correlation), 2) IC- inconsistent (correlation is opposite to that of the hypothesized correlation), and 3) Non- contingency (no correlation). Furthermore, it is also necessary to frame the questions around both variables equally. In considering these two factors in the development of the dependent measure, the following propositional statements can be used in this experiment.

Frame 1: Outcome given Image

1a. A (plant) is more likely than a (furniture) to result in a shock

1b. A (plant) is less likely than a (furniture) to result in a shock

1c. A (plant) just as likely as a (furniture) to result in a shock

1d. A (plant) is more likely than a (furniture) to result in no shock

1e. A (plant) is less likely than a (furniture) to result in no shock

1f. A (plant) is just as likely as (furniture) to result in no shock

Frame 2: Image given Outcome

2a. A shock is more likely than no shock after (plant)

2b. A shock is less likely than no shock after (plant)

2c. A shock is just as likely as no shock after (plant)

2d. A shock is more likely than no shock after (furniture)

2e. A shock is less likely than no shock after (furniture)

2f. A shock is just as likely as no shock after (furniture)

Note: (plant) and (furniture) refer to entirely new images which will not be presented during the task. Since there are 12 statements altogether, 12 new images from each category will be required.

"True" responses to (1b, 1d, 2b, 2d) and "False" responses to all others are consistent with the belief that the furniture category is correlated with a shock. Thus, these responses will be labelled as F- responses. The total number of F- responses will be scored out of 12, and this will be termed the F- acceptance rate. Similarly, "True" responses to (1a, 1e, 2a, 2e) and "False" responses to others are consistent with the belief that the plant category is correlated with shock and thus they are labelled P- responses. The number of P- responses will also be scored out of 12 to obtain a P- acceptance rate. Lastly, "True" responses to (1c, 1f, 2c, 2f) and "False" responses to all others are consistent with the belief that there is no relationship between category and shock. The number of such responses will be scored out of 12 to obtain a no- relation acceptance rate.

Predicted results

Recall that a shock is always less frequent than no shock. For condition- F, the prediction is that the occurrence of the shock will be linked more strongly to the furniture images because both are presented infrequently. Hence, it is hypothesized that the overall F- acceptance rate will be higher than the other two rates in condition- F. For condition- P, the prediction is that the occurrence of the shock will be linked more strongly to the plant images. Thus, the P- acceptance rate will be higher than the other two rates in condition- P.

The above pattern of results is consistent with the hypothesis that the IC paradigm is successful in eliciting a TAF- like belief. That is, a participant from condition- P probably believes that thinking about a plant- related image is likely to lead to an aversive shock outcome even though there is no semantic or statistical association consistent with this belief. With regards to TAF development in the real world, this implies that the coincidence of a relatively infrequent thought and an infrequent event can lead some to believe that both are related or one can lead to the other.

Alternatively, it is also possible that the participants from both conditions report higher no- relation acceptance rates than either F- or P- acceptance rates. This pattern of results holds significance because it means that participants did not succumb to the usual cognitive bias of illusory correlation. One reason for this may be that the presentation stimuli are neutral and not meaningfully related to shock outcomes. This contrasts with Hamilton and Gifford (1976), where group membership and traits have a socially meaningful relationship. Hence, it may be useful to replicate this experiment using relatively more negative stimuli and this makes sense because thoughts and anticipated outcomes in TAF beliefs are usually associated. In addition, it is also interesting to investigate whether the same results will occur if the aversive outcome of shock is experienced personally rather than by a stranger.

Conclusion

The purpose of this proposal is to examine if the IC paradigm can be used to elicit TAF beliefs in the laboratory.

The proposed experiment is designed in response to the shortcomings of the present literature on TAF- an increasingly important construct that has been found to play a role in anxiety disorders like OCD and social phobia. Besides an overall lack of research in general, past studies have not been able to isolate the phenomenon of TAF in the laboratory to study it in its own right. This is extremely critical because before researchers can even begin to examine the link between TAF and other concepts like neutralization and thought suppression, TAF has to be elicited properly. Hence, it is necessary to design a plausible paradigm to induce TAF in the laboratory, where stringent control can be exercised.

In addition, this experiment also has important treatment implications. For instance, if it is known that TAF beliefs are learnt in the same way that ICs are learnt, is there a possibility of "unlearning" via the same mechanism? Even if TAF cannot be unlearnt, can we at least present the objective statistical relations in the IC paradigm as a form of educational intervention to dispute the distorted "logic" held by people with OCD?

In conclusion, the experiment as proposed above is strongly motivated by a pressing need to advance research in a new but valuable construct TAF. Not only does it bridge the all- too- common gap between research in clinical and cognitive science, it also has potentially major implications on treatment methods.

References

Berle, D., & Starcevic, V. (2005). Thought-action fusion: Review of the literature and

future directions. Clinical Psychology Review, 25, 263- 284.

Chapman, L. J. (1967). Illusory Correlation in observational report. Journal of Verbal Learning and Verbal Behaviour, 6, 151- 155.

de Jong, P. J., Merckelbach, H., Bogels, S., & Kindt, M. (1998). Illusory correlation and social anxiety. Behaviour Research and Therapy, 36, 1063- 1073.

de Jong, P. J., Merckelbach, H., & Nijman, H. (1995). Hemisphere preference, anxiety and covariation bias. Personality and Individual Differences, 18, 363- 371.

Hamilton, D. L., & Gifford, R. K. (1976). Illusory correlation in interpersonal perception: A cognitive bias of stereotypic judgments. Journal of Experimental Social Psychology, 12, 392- 407.

Hermann, C., Ofer, J., & Flor, H. (2004). Covariation Bias for Ambiguous Social Stimuli

in Generalized Social Phobia. Journal of Abnormal Psychology, 113(4), 646- 653.

Hutchinson, G. T., Patock-Peckham, J. A., Cheong, J. W., & Nagoshi, G. T. (1998). Personality predictors of religious orientation among protestant, catholic and non-religious college students. Personality and Individual Differences, 24, 145-151.

Kennedy, S. J., Rapee, R. M., & Mazurski, E. J. (1997). Covariation bias for phylogenetic versus ontogenetic fear-relevant stimuli. Behaviour Research and Therapy, 35, 415- 422.

Lieberman, J. D. (1999). Terror management, illusory correlation, and perceptions of minority groups. Basic and Applied Social Psychology, 21(1), 13- 23.

McGahan, J. R., & Wight, R. (1989). Intuitive correlational reasoning with two binary variables: The influence of subtle linguistic variations. Paper presented at the annual meeting of the Southeastern Psychological Association, Washington, DC.

Pauli, P., Montoya, P., & Martz, G. E. (1996). Covariation bias in panic-prone individuals. Journal of Abnormal Psychology, 105, 658- 662.

Rachman, S., Shafran, R., Mitchell, D., Trant, J., & Teachman, B. (1996). How to remain neutral: An experimental analysis of neutralization. Behaviour Research and Therapy, 34, 889-898.

Rachman, S., Thordarson, D. S., Shafran, R., & Woody, S. R. (1995). Perceived responsibility: Structure and significance. Behaviour Research and Therapy, 33, 779-784.

Rassin, E. (2001). The contribution of thought-action fusion and thought suppression in the development of obsession-like intrusions in normal participants. Behaviour Research and Therapy, 39, 1023- 1032.

Rassin, E., & Koster, E. (2003). The correlation between thought-action fusion and religiosity in a normal sample. Behaviour Research and Therapy, 41, 361-368.

Rassin E., Merckelbach, H., Muris, P., & Spaan, V. (1999).Thought- action fusion as a causal factor in the development of intrusions. Behaviour Research and Therapy, 37, 231- 237.

Rheaume, J., Freeston, M. H., Leger, E., & Ladouceur, R. (1998). Bad luck: An underestimated factor in the development of obsessive-compulsive disorder. Clinical Psychology and Psychotherapy, 5, 1- 12.

Salkovskis, P., Shafran, R., Rachman, S., & Freeston, M. H. (1999). Multiple pathways to inflated responsibility beliefs in obsessional problems: Possible origins and implications for therapy and research. Behaviour Research and Therapy, 37, 1055- 1072.

Shafran, R., & Rachman, S. (2004). Thought-action fusion: a review. Journal of Behavior Therapy and Experimental Psychiatry, 35, 87- 107.

Shafran, R., Thordarson, D. S., & Rachman, S. (1996). Thought-Action Fusion in Obsessive Compulsive Disorder. Journal of Anxiety Disorders, 10(5), 379- 391.

Sica, C., Norvara, C., & Sanavio, E. (2002). Religiousness and obsessive-compulsive cognitions and symptoms in an Italian population. Behaviour Research and Therapy, 40, 813- 823.

Sma´ri, J., & Ho´lmsteinsson, H. E. (2001). Intrusive thoughts, responsibility attitudes, thought- action fusion, and chronic thought suppression in relation to cbsessive- compulsive symptoms. Behavioural and Cognitive Psychotherapy, 29, 13- 20.

Tallis, F. (1994). Obsessions, responsibility and guilt: Two case reports suggesting a common and specific aetiology. Behaviour Research and Therapy, 32, 143-145.

Tomarken, A. J., Mineka, S., & Cook, M. (1989). Fear-relevant selective associations and covariation bias. Journal of Abnormal Psychology, 98, 381- 394.

Tomarken, A. J., Sutton, S. K., & Mineka, S. (1995). Fear-relevant illusory correlations: what types of associations promote judgmental bias? Journal of Abnormal Psychology, 104, 312- 326.

van den Hout, M., Kindt, M., Weiland, T., & Peters, M. (2002). Instructed neutralization, spontaneous neutralization and prevented neutralization after an obsession-like thought. Journal of Behavior Therapy and Experimental Psychiatry, 33, 177- 189.

van den Hout, M., van Pol, M., & Peters, M. (2001). On becoming neutral: Effects of experimental neutralizing reconsidered. Behaviour Research and Therapy, 39, 1439- 1448.

Yates, M. C., McGahan, J. R., & Williamson, J. D. (2000). Intuitive covariation assessment of the illusory correlation.The Journal of General Psychology, 127(4), 397- 411.

Zucker, B. G., Craske, M. G., Barrios, V., & Holguin, M. (2002). Thought action fusion: Can it be corrected? Behaviour Research and Therapy, 40, 653- 664.