Delusion‐proneness or miscomprehension? A re‐examination of the jumping‐to‐conclusions bias

Abstract

Previous research has consistently shown that individuals with delusions typically exhibit a jumping‐to‐conclusions (JTC) bias when administrated the probabilistic reasoning ‘beads task’ (i.e., decisions made on limited evidence and/or decisions over‐adjusted in light of disconfirming evidence). However, recent work in this area has indicated that a lack of comprehension of the task may be confounding this finding. The purpose of the present study was to evaluate the influence of task administration, delusion‐proneness, and miscomprehension on the elucidation of the JTC bias. A total of 92 undergraduate university students were divided into one of two task conditions (i.e., non‐computerised and computerised) and were further identified as either delusion‐prone or non‐delusion‐prone and as comprehending or non‐comprehending the task. Overall, 25% of the sample demonstrated a JTC bias, and just over half made illogical responses consistent with a failure to comprehend the task. Qualitative evidence of comprehension revealed that these ‘illogical responses’ were being driven by a misunderstanding of task instructions. The way the task was administrated and levels of delusion‐proneness did not significantly influence JTC. However, miscomprehending participants were significantly more likely to exhibit the bias than those who did comprehend. These results suggest that miscomprehension rather than delusion‐proneness may be driving the JTC bias, and that future research should include measures of miscomprehension.

Key words:

• delusion‐proneness
• jumping to conclusions
• miscomprehension
• probabilistic reasoning

The role that cognitive‐reasoning biases may play in the development and maintenance of delusions has been the focus of much research in recent years. The biases described in people with delusions include attribution biases, where individuals are prone to exhibiting an externalising bias for negative events, with a particular inclination to blame others rather than the situation or chance (Bentall, Kinderman, & Kaney, 1994; Garety & Freeman, 1999); attentional biases, in which individuals are more selectively attentive towards threat‐related stimuli than neutral stimuli (e.g., Bentall & Kaney, 1989); high need for closure, which is a bias to seek definite answers to avoid ambiguity but at the cost of accuracy (e.g., McKay, Langdon, & Coltheart, 2007); and a theory of mind impairment, which leads to biases in assessing the intentions of others (e.g., Frith, 1994).

However, perhaps the most documented and robust cognitive bias identified by researchers in this area is the jumping‐to‐conclusions (JTC) bias. The JTC bias occurs when decisions are made on the basis of limited evidence. The task most commonly used to elicit the JTC bias is the ‘beads task’, first adapted by Huq, Garety, and Hemsley (1988). This task requires two containers each filled with coloured beads. Participants are shown that one jar has, for example, green and red beads in the proportion of 85:15, while the other has the reverse proportion of beads. The containers are then removed from sight, and the experimenter randomly draws out a bead and asks the participant which container a particular bead was drawn from, before placing the bead back and removing another, supposedly from the same container. However, the sequence of beads presented to participants is predetermined. The experiment is normally administered in one of two ways: the ‘draws to decision’ procedure, where participants take as many trials as needed to reach a definite decision (i.e., which container bead sequence is coming from); or the ‘graded‐estimates’ procedure, where the number of trials is fixed, and for each trial, participants must provide a probability estimate that a particular bead is from one of the two containers (Garety & Freeman, 1999; Moritz & Woodward, 2005).

The most commonly reported finding is that participants with delusions typically reach a decision and are more confident about that decision on less evidence than controls, whether the task is administrated in its original non‐computerised form (e.g., Garety, Hemsley, & Wessely, 1991; Huq et al., 1988; Peters & Garety, 2006; Peters, Thornton, Siksou, Linney, & MacCabe, 2008; So, Freeman, & Garety, 2008) or in its computerised version (e.g., Dudley, John, Young, & Over, 1997; Ellett, Freeman, & Garety, 2008; Fear & Healy, 1997; Moritz & Woodward, 2005).

At the same time, graded‐estimates beads task experiments have repeatedly revealed that when deluded individuals are faced with potentially disconfirmatory evidence (e.g., a green bead when the majority have been red), they are more likely and quicker than controls to change their hypothesis in regard to which container a bead came from (e.g., Garety et al., 1991; Moritz & Woodward, 2005; Peters, Day, & Garety, 1997; Young & Bentall, 1997). In other words, deluded participants over‐adjust for conflicting disconfirmatory evidence and readily jump to a different conclusion. This aspect of the JTC bias has been dubbed as a ‘bias towards disconfirmatory evidence’ or an ‘over‐adjustment’ bias.

Researchers employing the task have also found that the JTC bias can be elucidated in deluded patients using bead ratios as low as 60:40 (Garety & Freeman, 1999); is stable over time (Menon, Mizrahi, & Kapur, 2008; Peters & Garety, 2006); and is related to other cognitive biases such as the ‘theory of mind’ impairment (Langdon, Ward, & Coltheart, 2008). There is also growing evidence that the JTC bias can be found in non‐clinical individuals identified as ‘delusion‐prone’ (e.g., Broome et al., 2007; Colbert & Peters, 2002; Ellett et al., 2008; Van Dael et al., 2006), but not in non‐delusional patients with schizophrenia (Peters et al., 2008), which suggests the bias is specific to delusions rather than to a diagnosis of schizophrenia.

Despite the apparent robustness of the beads task at elucidating the JTC bias, recent work has exposed a potential confounding factor. In a computerised replication of the graded‐estimates procedure, Moritz and Woodward (2005) included a measure of non‐comprehension (referred to hereafter as miscomprehension), defined as ‘extreme over‐adjustment’ by selecting the opposite container to the one that was expected (e.g., judging beads were coming from the container with 90% green beads/10% red beads, when the sequence of ten beads only contained one green bead and nine red beads). It was proposed that participants may simply have misinterpreted or forgotten the basic principle of the task, which was that beads were only coming from one container and not both. Consequently, participants may incorrectly assume that containers swap throughout the task (e.g., ‘red beads must always come from the mostly red container, while all green beads must come from the mostly green container’). Although the ‘miscomprehension’ construct most directly relates to the ‘over‐adjustment’ aspect of the JTC bias, it may also potentially account for ‘premature decisions’, as participants may simply be responding to the current bead (‘red beads = red container’) and not to the bead sequence.

The results of this study indicated that 52% of the schizophrenia sample and 23% of the healthy controls made illogical responses congruent with the miscomprehension style of responding described above. Moreover, participants exhibiting a JTC bias were significantly more likely to fail apparently to comprehend the task. Once the miscomprehending participants were removed, the deluded group still exhibited a stronger JTC ‘premature decisions’ bias than controls, but these results nonetheless highlight the confounding nature of miscomprehension.

With the exception of Warman, Lysaker, Martin, Davis, and Haudenschield (2007), no other study using the beads task to assess the JTC bias in participants with delusions has formally observed or even recognised miscomprehension as a possible confound, despite the common finding that participants with a JTC style of responding usually have a lower intelligence quotient (IQ) ( e.g., Garety et al., 1991; Van Dael et al., 2006), and consequently might be finding it more difficult to grasp the task instructions. Furthermore, there is an extensive literature demonstrating cognitive deficits in clinical populations, such as people with schizophrenia (see Szöke et al. 2008 for a meta‐analytic review), hence, if participants are diagnosed with these conditions, then it would be expected that they would have more difficulty understanding and remembering instructions, compared with healthy controls. The original non‐computerised versions of the beads task have yet to be tested for levels of miscomprehension, but it is conceivable that comprehension for this version may be easier considering that there is more interaction between the participant and the experimenter.

Consequently, the present study pursued three major aims. The first aim was to determine if the style of task administration (i.e., non‐computerised/computerised) and/or delusion‐proneness (see ‘Participants’ in Methods) affected JTC (i.e., premature decisions and over‐adjustment), and/or comprehension levels. The second aim was to examine the specific influence of miscomprehension itself on JTC. Finally, the study assessed qualitative evidence of comprehension to determine if the ‘illogical responses’ made by participants were really being driven by a misunderstanding of task instructions.

METHODS

Participants

Undergraduate students given the beads task have shown to display styles of responding consistent with the JTC bias (Warman et al., 2007; Warman & Martin, 2006). For this reason, a sample of 72 first‐year undergraduate psychology students (60 female; 12 male; mean age = 21.15, standard deviation (SD) = 7.01) was employed for the study on the assumption that their higher‐than‐average IQ might negate the influence of miscomprehension. The two ‘task administration’ groups (i.e., non‐computerised and computerised) revealed no significant differences in age (t(70) = 0.65, p > .05) nor gender (p = .75, Fisher's exact test (two‐sided)).

All participants were administrated the Peters et al. Delusions Inventory (PDI; Peters, Joseph, & Garety, 1999) at the beginning of the experiment. The PDI is designed to assess ‘delusion‐proneness’ (or delusional ideation in the absence of active delusions) in the general population. The total score (including an affirmation of the unusual belief as well as the distress, preoccupation, and conviction subscales, each with subtotal of 200) can range from 0 to 640. In line with other research (e.g., Linney, Peters, & Ayton, 1998; Warman et al., 2007), participants were classified as delusion‐prone if their PDI score fell above the median (median = 75.5) and non‐delusion‐prone if their score fell below the median. The PDI scores of the current sample were consistent with other studies using the scale and are summarised in the Table-1. Delusion‐proneness revealed no significant differences in age (t(70) = 0.92, p > .05) nor gender (p = .75, Fisher's exact test (two‐sided)).

Table 1 Mean (SD) PDI scores, including subtotal PDI and subscales of distress, preoccupation, and conviction (across both samples)

–	PDI (SD)	Distress (SD)	Preoccupation (SD)	Conviction (SD)	Total PDI (SD)	Median total PDI
Quantitative sample	10.49 (5.74)	21.82 (15.08)	22.36 (15.88)	27.88 (17.55)	82.65 (52.48)	75.5
Qualitative sample	11.90 (6.47)	26.50 (21.03)	23.85 (16.59)	29.20 (17.64)	90.35 (60.32)	84.5

N: Quantitative = 72; qualitative = 20.

PDI = Peters et al. Delusions Inventory; SD = standard deviation.

A further 20 undergraduate participants (12 female; 8 male; mean age = 20.85, SD = 6.29) were recruited separately for the qualitative analyses (see Table-1 for PDI scores). As with the former sample, no differences arose between groups or delusion‐proneness for age (t(18) = 1.59, p > .05; t(18) = 0.67, p > .05, respectively) or for gender (p = .17, Fisher's exact test (two‐sided); p = 1.00, Fisher's exact test (two‐sided), respectively).

Materials and procedure

Apart from the nature of administration, both the computerised and non‐computerised versions of the task were kept exactly the same, each consisting of two ‘graded‐estimates’ beads tasks, in replication of the Moritz and Woodward (2005) experimental procedure. In task 1, participants were presented with two containers (or with a picture of two containers for the computerised version) full of red and green beads (90% red (R) and 10% green (G) for one and vice versa for the other) and were told that the experimenter/computer will randomly select beads from same container for the duration of the task (i.e., only one container). The sequence of ten beads, however, was predetermined and was presented in the following order:

R‐R‐R‐R‐G‐R‐R‐R‐R‐R.

After each trial, participants were asked to select from one of the following seven options (by pressing keys 1–7 for computerised version or by stating aloud their response for the non‐computerised version): 1 = beads are definitely from container A; 2 = beads very likely from container A; 3 = beads probably from container A; 4 = no estimate possible yet; 5 = beads probably from container B; 6 = beads very likely from container B; and 7 = beads definitely from container B. This rating scale was displayed for the duration of the experiment, as was the explicit instruction that estimates/decisions should be carried out while considering all beads being drawn. To ensure participants remembered the proportion of beads in each container, the containers themselves (non‐computerised version) or the pictures of the containers (computerised version) also remained displayed for the duration of the task. Participants were then shown a demonstration of a trial and were given the opportunity to clarify any questions before the task begun (only for task 1). Participants were said to have demonstrated a JTC bias if they gave a definite rating (i.e., 1 or 7) when presented with only one bead.

Task 2 followed a similar procedure with exactly the same instructions using blue and yellow beads; however, this task increased the number of trials, which represented potentially disconfirmatory evidence (i.e., a change from yellow‐to‐blue beads and vice versa). Consequently, 20 beads were presented as coming from two containers (80% yellow (Y), 20% blue (B) and vice versa) to the following order:

Y‐Y‐Y‐B‐Y‐Y‐Y‐Y‐B‐Y‐B‐B‐B‐Y‐B‐B‐B‐B‐Y‐B.

Participants were informed that there was no time limit to complete the tasks, and that they had as long as they wished before making a decision. This instruction was included to reduce the chances of participants making rash decisions on the basis of a perceived time limit.

Measures

The various measures that were employed in the study included

1	JTC‐premature decisions—a participant was identified as displaying this aspect of the JTC bias if they made a definite decision after only one bead on at least one task.
2	JTC‐over‐adjustment—represented the amount of change in the judgement ratings between trials with potentially disconfirmatory evidence (i.e., when beads changed from red to green/vice versa in task 1 and from yellow to blue/vice versa for task 2).
3	Miscomprehension—or ‘extreme over‐adjustment’ was defined as selecting the opposite container to the one that was being suggested, that is, if participants made an estimate that beads were coming from container B (i.e., ratings 5–7) within the first ten trials for task 1 where the sequence was clearly indicating container A; and/or made an estimate that beads were coming from container A (i.e., ratings 1–3) within the first ten trials for task 2 where the sequence was clearly indicating container B (Moritz & Woodward, 2005). Participants were identified as miscomprehending based on ‘illogical responses’ on at least one task.

Qualitative measures

Twenty participants were randomly selected for a qualitative analysis of the miscomprehension measure. They completed the experiment as described above, but they were asked to state aloud why they made their response for each bead (e.g., based on a hunch or probability) and, if applicable, what was influencing a change in confidence. This was recorded for the duration of the experiment. Additionally, all participants were asked at the end of the experiment to reveal their strategies throughout both tasks or to state what rules they thought governed the tasks. Because of the slightly different methods used within this group of participants, they were not included in the quantitative analysis.

RESULTS

The results are divided into three sections. The first examines the influence of task administration and delusion‐proneness on JTC and miscomprehension. The second examines the association miscomprehension has on the JTC bias. A third section presents the qualitative evidence relating to miscomprehending behaviour.

Task administration and delusion‐proneness

JTC‐premature decisions

As indicated in Table-2, ‘JTC‐premature decisions’ (i.e., a definite rating after only one bead on at least one task) was relatively high considering the sample consisted of healthy undergraduate students. Levels of this measure of JTC were identical for both the computerised and the non‐computerised versions of the task (i.e., 25% for each version).

Table 2 JTC‐premature decisions (%), JTC‐over‐adjustment, and non‐comprehension (%) by task administration and delusion‐proneness (across tasks)

–	JTC‐premature decisions (n)	JTC‐over‐adjustment (SD)	Non‐comprehension (n)
Overall	25.0(18)	1.33(1.18)	51.4(37)
Task administration
Non‐computerised	25.0(9)	1.41(1.06)	58.3(21)
Computerised	25.0(9)	1.25(1.30)	44.4(16)
Delusion‐proneness
Delusion‐prone	33.3(12)	1.37(1.20)	58.3(21)
Non‐delusion‐prone	16.7(6)	1.28(1.18)	44.4(16)

Notes. JTC‐premature decisions: relative percentage of participants who gave a definite rating after only one bead (for at least one task). JTC‐over‐adjustment: mean difference in judgement ratings between trials with potentially disconfirmatory evidence. Non‐comprehension: relative percentage of participants displaying non‐comprehension of at least one task.

N for all conditions = 72 (task administration: n = 36 per group; delusion‐proneness: n = 36 per group).

JTC = jumping to conclusions; SD = standard deviation.

Although not significant (χ²(1, N = 72) = 2.68, p > .05), participants identified as delusion‐prone exhibited higher ‘JTC‐premature decisions’ levels than non‐delusion‐prone individuals, with around a third of the delusion‐prone participants displaying the bias (Table-2). This trend is consistent with previous findings that have shown that individuals higher in delusion‐proneness demonstrate significantly higher JTC‐premature decisions compared with non‐delusion‐prone individuals (e.g., Warman & Martin, 2006).

JTC‐over‐adjustment

There were no significant differences for task type (t(70) = 0.58, p > .05) nor for delusion‐proneness (t(70) = 0.34, p > .05) in relation to over‐adjustment across either task 1 or 2, although there was a slight trend for the non‐computerised group and the delusion‐prone participants to display slightly higher levels of over‐adjustment (Table-2).

Miscomprehension

As seen in Table-2, just over half of the sample appeared to lack comprehension of the task as evidenced by responses that indicated that beads were coming from container B (i.e., ratings 5–7) within the first ten trials for task 1, even though sequence was clearly indicating container A and/or an estimate that beads were coming from container A (i.e., ratings 1–3) within the first ten trials for task 2, where the sequence was clearly indicating container B. Although the non‐computerised condition appeared to show higher levels of miscomprehension (Table-2), the results were not significant, χ²(1, N = 72) = 1.39, p > .05. Delusion‐prone individuals had a greater tendency to not comprehend (Table-2), but this was likewise non‐significant, χ²(1, N = 72) = 1.39, p > .05.

Task comprehension

JTC‐premature decisions

As seen in Table-3, miscomprehending participants were more likely to demonstrate premature decisions compared with comprehending participants (37.8% and 11.4%, respectively). These proportional differences were found to be significant, χ²(1, N = 72) = 6.69, p < .05.

Table 3 JTC‐premature decisions (%) and JTC‐over‐adjustment by comprehension (across tasks)

–	JTC‐premature decisions (n)	JTC‐over‐adjustment (SD)
Task comprehension
Non‐comprehension	37.8(14)*	2.19(1.03)*
Comprehension	11.4(4)	0.41(0.32)

N: Non‐comprehension = 37; comprehension = 35.

JTC = jumping to conclusion; SD = standard deviation.

* p < .05.

JTC‐over‐adjustment

Miscomprehending participants were significantly more likely to over‐adjust their responses in the face of potentially disconfirmatory evidence across both task 1 and task 2, t(70) = 9.77, p < .001, as can be clearly seen in Table-3.

Note: All analyses were repeated with task 2 removed from the miscomprehension variable. This more conservative approach ensured that the variable was not contaminated from the fact that ‘miscomprehending behaviour’ in task 2 (i.e., selecting a bead from a non‐suggestive container) may have arisen as a result of the proportions being more balanced (i.e., 80% vs 20% instead of 90% vs 10% in task 1). However, the results as presented above remained unaltered.

Qualitative measures of miscomprehension

A further 20 participants were randomised into the computerised/non‐computerised conditions in an effort to assess qualitatively whether participants who demonstrated ‘miscomprehending behaviour’ actually failed to understand the task instructions. They were asked to state throughout each task what was influencing their decisions/confidence and what strategies they used and/or what rules governed each task. In line with the previous sample, 55% of all participants displayed ‘miscomprehending’ behaviour, and 20% demonstrated both forms of the JTC bias, all of whom were miscomprehending.

The most interesting observation was that over 90% of all miscomprehending participants thought that containers were swapping throughout the task (i.e., beads were coming from both rather than one container). This is despite being explicitly told that beads are only coming from the one container throughout the task. Examples of what these participants said include: ‘I think that this is probably from Container B because it has been Container A for a while now, so it's got to switch over’; ‘I thought that if there was a series of red [beads] and then it suddenly swapped to a green [bead], it could be picking a green bead out of Container A [mostly red], but I thought it would be more likely that it would be switching it up’; and ‘I assumed the program was designed to randomly select containers throughout the task’. This left some participants rather confused and unconfident in their decisions, even by the end of task 1 that was clearly indicating container A, ‘My confidence is decreasing, I don't know where beads are coming from now, there doesn't seem to be a system. So I think I'll choose 4 [no estimate possible]’.

Additionally, just under half of these miscomprehending participants revealed a misunderstanding of the laws of probability that governed the tasks, ‘My strategy was based on the probability that beads would come from the other container eventually. They could not have come from one container’; ‘I thought if [a bead] had been drawn a few times it would have to eventually come from another container, so I just tried to use my knowledge of probability more than hunches’; ‘If there are too many red beads coming up, I would say that eventually the computer is going to choose a red bead from Container B, with mostly green beads’; and ‘Since there has been so many red beads, and its only a 90% chance that it would be Container A, the likelihood of it being from Container B is becoming ever increasing’.

Finally, some miscomprehending participants in the non‐computerised conditions claimed that they actually tried to listen to where beads were picked from and stated that the task involved some level of deception, for instance, ‘In Task 1, at first I was thinking you've picked beads from Container A, then you pick a green bead to try and trick me’.

In contrast to the comments made by miscomprehending participants, those who comprehended the task generally confirmed the task rules that only one container was used throughout, with one participant going so far to say ‘made it easier knowing that only one container was being used for the whole task, otherwise it would have been confusing’.

DISCUSSION

The present study aimed to investigate the potentially confounding influences of task administration, delusion‐proneness, and miscomprehension on the JTC bias using the ‘beads’ task. Neither task administration nor delusion‐proneness could account for the bias, although there was a non‐significant trend for delusion‐prone individuals to display higher levels of JTC and miscomprehension. The results suggest instead that illogical responses due to miscomprehension of the task instructions may be driving the effect.

Overall, levels of both JTC and miscomprehension were relatively high compared with previous findings. Most studies report JTC in healthy controls to be around 10–20% (Freeman, 2007). Interestingly, the Moritz and Woodward (2005) study did not detect the JTC bias at all among healthy controls, whereas the present findings revealed that up to 25% of all participants demonstrated a JTC bias on at least one task (this was also using a conservative threshold of JTC behaviour – i.e., making a definite decision after only one bead). Furthermore, the Moritz and Woodward study found that only 23% of healthy controls made at least one illogical response consistent with miscomprehension, compared with the present findings where just over half the total healthy sample made at least one illogical response, which is actually more consistent with Moritz and Woodward's finding that 52% of the schizophrenia sample did not comprehend.

Task administration and delusion‐proneness

When the sample was analysed for task administration type (i.e., computerised and non‐computerised), levels of ‘JTC‐premature decisions’ were identical, yet there was a non‐significant trend for the computerised condition to display lower rates of ‘JTC‐over‐adjustment’ and miscomprehension. These findings may help explain why the Moritz and Woodward study, which only employed the computerised condition, found lower levels of JTC and miscomprehension among the healthy controls. This also has implications for much of the ‘beads task’ literature, both past and present, which have and continue to employ non‐computerised versions of the task (e.g., Peters et al., 2008; So et al., 2008). It is highly conceivable that clinical samples may have even higher rates of miscomprehension than the present findings, based on the fact that 52% of Moritz and Woodward's (2005) schizophrenia sample were in a computerised condition. While the exact mechanisms behind this task type trend are not fully understood, and any suggestion here remains speculative, it is possible that computerised versions allow participants to work at their own pace, and re‐read instructions, without having to consult the experimenter.

Delusion‐proneness similarly yielded some interesting trends in concordance with the previous literature, with individuals identified as delusion‐prone consistently scoring higher levels of JTC and miscomprehension. However, these trends failed to reach significance, making it difficult to draw any conclusions about the role delusion‐proneness plays in explaining the bias. It is also worth pointing out here that although the study had low power because of the sample size, the magnitude of the effect was quite small.

Task comprehension

As stated above, levels of miscomprehension were particularly high, and these results suggest that they may be at least partially responsible for the ‘premature decisions’ and ‘over‐adjustment’ findings. Nearly 40% of miscomprehending participants exhibited ‘JTC‐premature decisions’, or alternatively, nearly 80% of all participants show that the JTC bias were simultaneously miscomprehending the task instructions. Similarly, miscomprehending participants were also significantly more likely to ‘over‐adjust’ in the face of potentially disconfirmatory information, which in part confirms the definition of miscomprehension as ‘extreme over‐adjustment’. The qualitative comments made by participants confirm that ‘miscomprehension’ is a valid measure, as the majority of the miscomprehending participants stated that the containers swapped throughout the tasks. The notion that containers were swapping is of theoretical importance, as it can alternatively explain both aspects of the bias (i.e., premature decisions and over‐adjustment), as participants may have been responding to the current bead (e.g., ‘red beads come from the red container and green beads come from the green container’) rather than the bead sequence (cf. Young & Bentall's (1997) ‘recency effect’, which claims that people with delusions may simply be responding to the most recent information being presented). This style of reasoning was not simply due to the nature of task 2 (first ten trials indicate container B and second ten indicate container A), as these participants implied containers swapped every time the beads changed colour, and the comments were also observed for task 1, where the entire sequence was strongly indicating one container. Such a belief may have been generated simply because these participants somehow missed or had forgotten the crucial instruction that only one container is picked. This could also explain why some of these participants demonstrated a misunderstanding of the probabilities that governed the tasks; yet it is equally conceivable that their knowledge of probabilities was limited from the outset.

Taken together, the results suggest that at least for healthy controls, task administration and delusion‐proneness contribute little towards the bias, while there is a somewhat greater association between JTC and miscomprehension, which may undermine the validity of the effect.

Nonetheless, it can be pointed out that the ‘JTC‐premature decisions’ effect has been shown to be significantly higher among participants with delusions even when those who did not comprehend were removed from the analysis (Moritz & Woodward, 2005). A recent study by Speechley, Whitman, and Woodward (2010) further explored this issue by attempting to reduce the risk of participants not comprehending the task. This was done by employing a more realistic ‘fish/lakes’ stimuli set rather than the typical abstract beads/containers set. Additionally, four of the six series of tasks incorporated ten fish of the same colour (‘uniform’ condition) rather than the usual alternating pattern (‘alternate’ condition), which helped participants understand that only one lake was being drawn from. Moreover, in contrast to previous designs, the study included two 10‐point rating scales (from very unlikely to very likely) for each of the two ‘lakes’. It was argued that a single rating scale results in a loss of information, such that it is impossible to know whether a movement in one direction implies a downward rating adjustment for one option with a simultaneous upward rating adjustment for the other or vice versa. The results were hence able to show how people reacted to ‘matching lakes’ (i.e., a lake with a ratio of fish consistent with the colour of the current fish) and ‘non‐matching lakes’ at the same time.

The results showed that even within the four ‘uniform colour’ lakes, where miscomprehension could effectively be ruled out, participants with active delusions rated ‘matching lakes’ significantly higher (i.e., more likely) than all non‐delusion groups, consistent with the JTC ‘premature decisions’ effect. Moreover, participants with delusions seemingly over‐adjusted to the ‘disconfirmatory evidence’ in the ‘alternating lakes’ conditions, suggesting that they still misunderstood the instructions for these ‘lakes’ despite efforts to reduce miscomprehension.

However, the most interesting finding was there were no differences between delusional and non‐delusional groups for ‘non‐matching lakes’, whether the trials of fish were of ‘uniform’ colour or were ‘alternating’. Hence, the authors concluded that in the absence of miscomprehension, ‘JTC‐premature decisions’ is better thought of as a hyper‐salience of positive matches between the evidence and the hypothesis, as the non‐matching lakes were not rated any lower by the delusional participants compared with controls. This finding also affirms the notion that ‘JTC‐over‐adjustment’ findings are not over‐reactions to disconfirming evidence in the ‘alternating’ condition, as ratings for the ‘original’ matching lake were not any lower than those reported by the healthy controls, as would be expected if decisions were ultimately changed. Rather, the ‘over‐adjustment’ may represent a hyper‐salience of positive matches between the evidence and the hypothesis when instructions are misunderstood (i.e., containers/lakes ‘swapping’).

The present study is not without some limitations. Most notably, delusion‐proneness is conceptually different from clinical delusions, thus a replication would benefit from the inclusion of a clinical subsample of individuals with a diagnosis of active psychosis. Replications of the study might also include a qualitative analysis of miscomprehension for all participants, rather than a subsample, to better account for any potential variability. To determine the potential influence of IQ on miscomprehension, future research should also include a measure of IQ, particularly verbal IQ. An additional experimental condition could also be added, whereby participants are explicitly instructed that ‘beads always come from the same container’, and that ‘containers do not swap at any point’, outlining that it is more logical to down‐rate confidence for a particular container in the face of ‘disconfirming evidence’ rather than jump between containers. This instruction could effectively remove miscomprehension entirely (even for ‘alternating’ containers/lakes), and if miscomprehension and ‘over‐adjustment’ are indeed the same construct as suggested by the current study, this should likewise remove any evidence of over‐adjustment.

Despite these caveats, the present study has nonetheless demonstrated the confounding nature of miscomprehension in the beads task. The findings question the validity of the JTC bias previous ‘beads’ studies have revealed and further question its unmodified use in future research. In sum, the present study highlights the importance for experimenters in this area to check that participants actually understand the task before proceeding to ‘jump to conclusions’ about the potential influence a bias may have on the formation and/or maintenance of delusions.

ACKNOWLEDGEMENT

The authors wish to thank Todd Woodward for his insightful comments on earlier versions of the manuscript.