Jumping to conclusions bias in early psychosis: Relationships with symptom dimensions, neurocognition and social cognition

Abstract

Cognitive biases, such as “jumping to conclusions” (JTC), play an important role in the development and maintenance of delusions in psychotic disorders. However, their associations with symptom dimensions in early psychosis (EP) remain unclear. We aimed to investigate whether patients with EP who tend to jump to conclusions differ from those who do not in terms of symptom dimensions. We also examined relationships among symptoms, JTC, neurocognition, facial emotion recognition (FER) and theory of mind (ToM).

Seventy-five patients attending an EP intervention programme were assessed using the Positive and Negative Syndrome Scale and cognitive tasks. A multivariate analysis of covariance (MANCOVA) and a permutation-based nonparametric MANOVA (PERMANOVA), adjusted for demographic factors, were conducted to examine differences between patients with and without JTC bias across symptom dimensions and depressive symptoms. Bivariate correlations were performed to explore associations between variables.

A significant multivariate effect was found (Pillai’s trace = 0.291; F_{5, 65} = 5.330; p < .001) with large effect sizes. JTC patients scored higher than non-JTC patients on all symptom dimensions. At the symptom level, JTC bias was significantly correlated with delusions and hallucinations, as well as with ToM and FER.

Concluding, patients with EP who jump to conclusions exhibit a distinct symptom dimension pattern characterised by more severe symptoms. Furthermore, ToM and FER are related to JTC and symptom dimensions. These findings are clinically relevant, as psychological interventions targeting cognitive biases and social cognition appear to be effective in improving psychotic symptoms. Further studies are needed to replicate these findings.

Highlights

• •EP patients with JTC bias show more severe symptoms and poorer functioning.
• •ToM and FER are related to JTC and symptom dimensions.
• •At the symptom level, JTC bias is associated with delusions and hallucinations.
• •ToM strongly correlates with neurocognition and is related to delusions.
• •Adding strategies to reduce JTC in early treatments may improve clinical outcomes.

1. Introduction

The Australian Clinical Guidelines for Early Psychosis (2016) defined early psychosis (EP) as “the early course of a psychotic disorder, specifically the prodrome and the period up to five years from first entry into treatment for a psychotic episode (i.e., first episode psychosis, or FEP)”. This period has been demonstrated to be critical for prognosis and recovery (Gouveia et al., 2023).

On the other hand, research mostly supports the relationship between cognitive biases (CBs) and delusions in psychosis (De Rossi and Georgiades, 2022; Sheffield et al., 2023). Prike et al. (2018) defined CBs as “a deviant factor in judgement, where the inferences we make about other persons and situations may be illogical”. Various CBs have been identified in psychosis and other psychiatric disorders (e.g., Pena-Garijo et al., 2022; Peters et al., 2014; Sanford and Woodward, 2017), including the “jumping to conclusions” bias (JTC), belief inflexibility and attributional biases. Hostility attribution and bias against disconfirmatory evidence also contribute to symptom maintenance, but with less empirical support (Samson et al., 2024). Medium-to-large effect sizes were found for JTC when clinical samples were compared with controls using experimental tasks (Samson et al., 2024). More recently, other CBs, such as aberrant salience, have been demonstrated to be a core mechanism in the early onset of psychosis (Marano et al., 2025). This bias is highly interconnected with JTC and delusion severity in individuals with psychosis (Jagtap and Best, 2024). Furthermore, distinct and shared neural mechanisms could underpin different CBs (Kowalski et al., 2021). For example, Miyata et al. (2024) revealed the association between JTC bias, aberrant salience and the default brain mode. Neurocognitive models have also stated that dopamine activity, aberrant salience and JTC are empirically associated with delusions (Broyd et al., 2017).

The JTC bias is the most widely studied cognitive bias. It is the tendency to make premature judgments without accounting for alternative explanations, and its influence on the formation and maintenance of delusions has been documented (Dudley et al., 2016; McLean et al., 2017; So et al., 2016). JTC bias is central to psychological theories of delusion, as it is thought to lead to the rapid acceptance of implausible ideas and prevent consideration of more realistic alternative explanations of events (Freeman and Garety, 2014). However, Doherty et al. (2025), based on their systematic review and meta-analysis across the psychosis continuum, critically questioned whether JTC is indeed related to delusions. Nevertheless, as the same authors state, their conclusions cannot be easily generalised to symptom severity in clinical populations.

JTC and other CBs have been described in individuals at high risk for psychosis (HR), FEP and established schizophrenia (Gawęda et al., 2024; Livet et al., 2020). Furthermore, Catalan et al. (2022) found that the rate of JTC bias was significantly higher among HR individuals who transitioned to psychosis when they were assessed two years after transition. Those who developed psychosis showed a more considerable JTC bias than those who did not.

Moreover, JTC is present even in the early stages of psychosis, is associated with delusions, and is related to poorer neuropsychological functioning (Falcone et al., 2015; González et al., 2018; Rodriguez et al., 2019). Patients with neuropsychological deficits tended to jump to conclusions from early to chronic stages of the illness (Takeda et al., 2018; Tripoli et al., 2021).

Additionally, distinct neurocognitive profiles have been identified in FEP patients (Espinosa et al., 2025) and are associated with social cognition and metacognition. The most frequent profile involved individuals who were more prone to JTC bias, and those with lower social cognition showed worse neuropsychological and clinical features, with the worst functioning (Ferrer-Quintero et al., 2021).

In this sense, impairments in social cognition domains have been reported in FEP (Healey et al., 2016), including theory of mind (ToM) and facial emotion recognition (FER). FER deficiencies have been described from their early manifestations in individuals with HR to advanced phases of the disease (Pena-Garijo et al., 2023). Moreover, impairments in recognising anger and fear have been suggested as vulnerability markers for psychotic disorders (Pena-Garijo et al., 2023; Tripoli et al., 2022). Similarly, deficits in recognising negative emotions may predict the transition from HR to full-blown psychosis (Corcoran et al., 2015). Likewise, patients with FEP who jump to conclusions are impaired in processing social information (Díaz-Cutraro et al., 2022). Those individuals who tend to interpret situations with a higher level of confidence fail to correctly “read” the mental states of others (Bora and Pantelis, 2013; Takeda et al., 2018). These deficits in ToM can be observed in EP and appear relatively independent of other cognitive functions (Catalan et al., 2018; Lindgren et al., 2018).

As previously outlined, existing literature has extensively investigated the relationships between JTC bias and positive symptoms, particularly delusions. However, evidence for associations with other symptom dimensions is almost non-existent, except for some prospective data on interactions with affective dysregulation in trajectories of psychotic experiences (Dudley et al., 2016; Rauschenberg et al., 2021). Only evidence for associations with negative symptoms has been previously reported (Hayashi et al., 2022).

Beyond investigations of associations between CBs and symptom dimensions, new approaches to psychopathology, such as network analysis (Borsboom, 2017), have been applied to psychosis, including EP samples, demonstrating meaningful symptom clusters and interrelation patterns linked to clinical and functional outcomes (e.g., Brasso et al., 2023). To date, psychotic symptoms are typically measured using the Positive and Negative Syndrome Scale (PANSS; Kay et al., 1987). While several studies aimed to establish the best PANSS factor structure for assessing patients with EP (e.g., Langeveld et al., 2013), recent studies have addressed its dimensional structure (Dal Santo et al., 2024; Griffiths et al., 2021; Petruzzelli et al., 2018). Using novel network analysis methods, Dal Santo et al. (2024) proposed a four-dimensional PANSS version comprising 21 key items that captured the core symptoms of FEP.

Finally, cognitive interventions targeting specific CBs, such as metacognitive training (MCT), have demonstrated their efficacy in treating psychotic symptoms (Goncalves et al., 2025; Penney et al., 2022; Sauvé et al., 2020). Moreover, MCT may be a promising intervention for patients with EP (Hua and Kanas, 2020; Ochoa et al., 2017).

1.1. Aims and rationale

To the best of our knowledge, no previous studies have explored the associations between JTC bias and symptom dimensions in EP.

Moreover, although JTC and other CBs have demonstrated their influence on the development and maintenance of delusions in psychotic disorders (De Rossi and Georgiades, 2022), some contemporary studies have questioned whether JTC is actually related to delusions (Doherty et al., 2025). Likewise, fewer studies have been conducted in individuals with EP, with mixed results (Xenaki et al., 2022).

Furthermore, although relationships between JTC and social cognition have been investigated in EP, further research is needed (Díaz-Cutraro et al., 2022).

Finally, in line with dimensional approaches to psychopathology, a novel dimensional PANSS version has recently been proposed to assess the key symptoms of FEP (Dal Santo et al., 2024).

In order to elucidate the aforementioned issues, our study planned the following objectives:

(a) To determine whether patients with EP who tend to jump to conclusions differ from those who do not in terms of symptom dimensions.

(b) To examine whether JTC was associated with cognitive domains, such as neurocognition, ToM or FER.

(c) To verify the relationships between JTC and cognitive domains with positive symptoms, such as delusions, hallucinations, suspiciousness and unusual thought content.

(c) Additionally, given its novel development, we aimed to evaluate the reliability of the PANSS four-dimensional version (Dal Santo et al., 2024) used in this study for assessing symptom dimensions.

We hypothesised that patients who tend to jump to conclusions will exhibit a distinct presentation characterised by more severe symptoms. Moreover, significant relationships between JTC, neurocognition and social cognition would be identified. Lastly, based on previous studies, we expected that JTC bias and cognitive domains would be related to symptom dimensions, specifically to positive symptoms.

2. Materials and methods

2.1. Participants and procedure

A cross-sectional study was conducted based on data from patients attending the Early Intervention for Psychosis (EIP) programme at the University Hospital Doctor Peset in Valencia, Spain.

Seventy-five outpatients (age ranging from 16 to 45) diagnosed with schizophrenia-spectrum disorders according to DSM-5 criteria (American Psychiatric Association, 2013), within the previous five years, were included in the study. An experienced psychiatrist confirmed the diagnosis through a clinical interview upon entry into the programme. Assessment tools were administered once the patients were considered sufficiently stable to undergo evaluation. Those with neurological conditions, premorbid IQ < 70 or current substance dependence (except tobacco and caffeine) were excluded. The evaluation was carried out as part of the programme's assessment protocol, which included four in-person sessions to administer the clinical and cognitive assessment instruments and complete a set of sociodemographic questionnaires and online assessment tools, both under the supervision of a clinician.

Sixty-eight patients (90.67 %) were taking antipsychotics at the time of evaluation. Premorbid cannabis use six months prior to evaluation was present in 32 patients (42.7 %). The mean duration of untreated psychosis (DUP) was 33.6 weeks (SD = 79.4). Data on DUP were missing for 13 participants.

This study is part of a broader research project that investigates the impact of an EIP programme. The study adhered to the ethical standards outlined in the Declaration of Helsinki and was approved by the ethics committee of the University Hospital Doctor Peset (reference: CEIm 24/21). All participants were informed verbally and in writing, and their written informed consent was obtained.

2.2. Measures

2.2.1. Symptoms

The PANSS (Kay et al., 1987; Peralta and Cuesta, 1994) is the most widely used tool to assess psychotic symptoms in patients with schizophrenia-spectrum disorders. The four-dimensional solution proposed by Dal Santo et al. (2024) was used. It comprises 21 items from the original 30-item scale, divided into four dimensions (positive, cognitive/disorganised, excited/aggressive and negative). All variables showed item stability values above 0.75, indicating good reliability.

The Calgary Depression Scale for Schizophrenia (CDSS; Addington et al., 1990) was used for assessing depressive symptoms. The internal consistency of the Spanish version was 0.83 (Sarró et al., 2004).

2.2.2. JTC bias

The “Beads task” is a well-established experimental task for assessing JTC bias. It is a probabilistic reasoning task in which the subject sees an image of two jars filled with balls of two different colours in different proportions (Dudley and Over, 2003). A computerised version with a balls' ratio of 60:40 was used. Participants are informed that a jar has been randomly selected and the balls will be extracted and displayed one by one on the screen. The task consists of deciding which jar the balls come from, seeing as many balls as necessary until the subjects are sure of their decision. We computed the number of “draws to decide” (DTD) (i.e., the number of balls the subject needed to see to reach a decision) to assess the data-gathering bias (the lower the DTD, the greater the JTC bias). DTD ≤ 2 served as a threshold for considering a JTC bias (Garety et al., 2011).

2.2.3. Neurocognition

The Screen for Cognitive Impairment Psychiatric Scale (SCIP-S; Pino et al., 2008) is a neuropsychological test designed to screen for cognitive impairments in psychiatry. We used the overall percentile score to summarise neurocognitive performance. The Spanish version achieved good reliability (Cronbach's alpha = 0.72).

The TAP-30, a Spanish version of the National Adult Reading Test (NART; Russell et al., 2000), was used to assess premorbid IQ. The NART has been widely used to estimate premorbid IQ in patients with schizophrenia (Gomar et al., 2011). The TAP-30 requires the pronunciation of 30 low-frequency Spanish words with removed accents (Gomar et al., 2011).

2.2.4. Theory of mind

The Hinting Task was designed to assess ToM in patients diagnosed with schizophrenia and provides an overall score for 10 stories. Each story ends with one character dropping a hint, and participants are asked to indicate what the character truly means (Klein et al., 2020). The Spanish version demonstrated acceptable internal consistency, with Cronbach's alpha coefficients ranging from 0.64 to 0.69 (Gil et al., 2012).

2.2.5. Facial emotion recognition

The PERE (Prueba de Evaluación de Reconocimiento de Emociones, from Spanish) comprises 56 pictures to assess the perception of the six basic emotions. The instrument was developed and validated in the Spanish population (Gil-Sanz et al., 2017). All pictures had an accuracy higher than 89 % and test-retest reliability over 0.80 for controls and 0.61 for psychotic outpatients. The instrument was administered through an online computerised version. The total number of adequately recognised emotional expressions was used to estimate FER.

2.2.6. Global functioning

The Global Assessment of Functioning (GAF) was used to assess clinical and social functioning (Pedersen et al., 2018).

2.3. Statistical analyses

Statistical analyses were conducted using SPSS version 23.0 (IBM Corp., 2015), JASP version 0.19.1 (JASP Team, 2024), G*Power version 3.1.9.4 (Faul et al., 2007), and R Studio Pro version 2025.05.1 (Posit team, 2025).

Demographic and clinical data were analysed using appropriate statistics. The significance level was set at 0.05 for all tests.

The reliability of the PANSS 21-item version used in the current study was evaluated using Cronbach's alpha.

To assess the data-gathering bias (continuous variable), we computed the number of “draws to decide” (DTD) (the lower the DTD, the greater the JTC bias). The presence of a JTC bias was defined as DTD ≤ 2 on the Beads task. This threshold was used to construct the categorical variable and allocate participants into either the JTC group (n = 15) or the non-JTC group (n = 60).

We tested whether participants classified by groups differed on the PANSS-21 symptom dimensions (positive, cognitive/disorganised, excited/aggressive and negative) and depression (CDSS).

Two sequential approaches were used: (1) a conventional multivariate analysis of covariance (MANCOVA) as a first line of evidence, and (2) given the expected violations of multivariate normality and equality of variances, a robust permutation multivariate analysis of variance (PERMANOVA) as a sensitivity check. This last method was explicitly developed for unbalanced and/or asymmetrical designs (Anderson, 2017).

First, the conventional MANCOVA was performed with the four PANSS dimensions and the CDSS scores entered together as dependent variables and JTC bias as the primary between-subjects factor. To account for potential confounding variables, we adjusted the analysis to include gender, age, educational level and premorbid IQ as covariates. Univariate tests were conducted based on the linearly independent pairwise comparisons among the estimated marginal means (means corrected for covariates).

As the assumptions for parametric analysis checks indicated potential violations of multivariate normality and equality of variances, Pillai's trace (V) was preferred as the main statistic because it is the most robust to such violations, even in small and unbalanced samples (Ateş et al., 2019). For the MANCOVA, the effect size was estimated by using Cohen's f², which is associated with Pillai's V. For the univariate tests, partial eta squared (partial ŋ²) was used.

Second, group differences across multiple symptom dimensions were examined using PERMANOVA. PERMANOVA partitions the variation into a distance matrix according to an ANOVA-like design and assesses the significance of a pseudo-F statistic by repeatedly permuting the data labels. This approach does not assume multivariate normality and is robust to modest sample sizes (Anderson, 2017). Euclidean distances were computed from the scaled outcome variables (PANSS symptom dimensions and CDSS scores). The model included JTC bias as the main factor, with gender, age, educational level, and premorbid IQ entered as covariates. Significance was assessed using 4999 permutations, with Type II (marginal) tests reported for each predictor. Homogeneity of dispersion was checked using betadisper and permutest to ensure that observed differences reflected group centroid separation rather than unequal variance.

To further explore which symptom dimensions contributed to the multivariate effect, we conducted univariate permutation tests for each dependent variable separately. Each outcome was analysed with the same covariates and 4999 permutations, using Euclidean distances and the same ANOVA-like framework. These follow-up tests provided permutation-based F statistics, R² values and p-values for each predictor, allowing identification of specific symptom dimensions associated with JTC bias while controlling for covariates. The R code to run these analyses is provided in the Supplementary Materials.

Sample size was determined a priori based on power analysis, with Pillai's trace as the test statistic. Assuming a moderate-to-large effect size (Cohen's f² = 0.09–0.16), an α level of 0.05 and a desired power of 0.80, the required total sample size ranged from 62 to 92 participants, depending on the exact effect size and group allocation. To account for the expected violation of multivariate normality, we applied a 15 % inflation factor, resulting in a final target sample size of 76–92 participants (with approximately 15–18 in the smaller group and 60–74 in the larger group). An additional 10 % buffer was considered to compensate for potential attrition. Our final sample size (n = 75) was close to the minimum required, and we considered it appropriate.

A sensitivity analysis was conducted to determine the minimum detectable effect size for the multivariate tests (Pillai's trace). At an α level of 0.05 and a chosen power of 0.80, the minimum detectable effects were estimated to be Cohen's f² = 0.185.

Bivariate correlations (Spearman's rho) were calculated to examine relationships among the study variables within the whole sample (n = 75). Given the small size of the JTC group (n = 15), separate group-level correlation analyses were considered unfeasible.

3. Results

3.1. Demographic and clinical characteristics

Fifteen (20 %) patients exhibited a JTC bias. Patients with JTC were more frequently males (χ² = 3.77; p < .05). The JTC group scored lower in FER (F = 4.67; p < .05) and ToM (F = 6.55; p < .05). Global functioning was also lower in the JTC group (F = 5.06; p < .05). No between-group differences were found in age, educational level, premorbid cannabis use or antipsychotic dose. Within the entire sample, males scored higher in positive symptoms (F = 4.81; p < .05) and premorbid cannabis use (Z = 1.48; p < .05). The PANSS overall score (Mean = 56.76; SD = 14.71) indicated mild-to-moderate symptom severity. Table 1 presents the demographic and clinical data.

Table 1.

Demographic and clinical data.

Variables		N / mean	% / SD
Gender (male)		49	65.3 %
Age (in years)		24.35	6.14
Educational level	Primary	15	21.13 %
	Secondary	39	54.93 %
	Superior	17	23.94 %
Premorbid cannabis use	No use	36	48.0 %
	Occasional	7	9.3 %
	Monthly	3	4.0 %
	Weekly	10	13.3 %
	Daily	19	25.3 %
JTC bias	Jump	15	20.0 %
	No jump	60	80.0 %
Beads Task (DTD)		6.31	3.76
Positive		8.45	4.06
Negative		13.55	5.59
Disorganised/cognitive		12.52	3.90
Excitement/aggression		4.28	2.04
Depression (CDSS)		3.52	4.03
Facial emotion recognition (PERE)		47.63	3.75
Theory of Mind (Hinting task)		16.36	2.39
Neurocognition (SCIP-S)		57.60	23.93
Premorbid IQ (TAP-30)		102.97	6.37
Global functioning (GAF)		67.33	15.16
Antipsychotics		11.92	11.61

SD: Standard deviation. CDSS: Calgary Depression Scale for Schizophrenia. DTD: draws to decide. GAF: Global Assessment of Functioning. Symptom dimension scores were based on the 21-item PANSS version (Dal Santo et al., 2024). Antipsychotic doses were expressed as the olanzapine equivalent dose (mg/day). A formula was used to convert TAP scores into an estimated IQ (Gomar et al., 2011).

3.2. Reliability of the PANSS version

Cronbach's alphas for each dimension were: 0.75 for Positive, 0.69 for Disorganised/cognitive, 0.72 for Excited/aggressive and 0.88 for Negative. For the full scale, Cronbach's alpha = 0.88. Table 2 shows the selected items and their corresponding dimensions.

Table 2.

The PANSS-21 selected items and their corresponding dimensions.

Dimension	Item/symptom
Positive	P1, Delusions; P3, Hallucinatory Behaviour; P6, Suspiciousness/Persecution; G9, Unusual Thought Content.
Cognitive/disorganised	P2, Conceptual Disorganisation; N5, Difficulty in Abstract Thinking; N7, Stereotyped Thinking; G5, Mannerisms and Posturing; G10, Disorientation; G11, Poor Attention; G13, Disturbance of Volition; G15, Preoccupation.
Excited/aggressive	P7, Hostility; G8, Uncooperativeness; G14, Poor Impulse Control.
Negative	N1, Blunted Affect; N2, Emotional Withdrawal; N3, Poor Rapport; N4, Passive/Apathetic Social Withdrawal; N6, Lack of Spontaneity and Flow of Conversation; G7, Motor Retardation.

PANSS 21-item, four dimensions version (Dal Santo et al., 2024).

3.3. Differences in symptom dimensions

The MANCOVA showed a significant multivariate effect, with Pillai's trace (V) = 0.291; F_{5, 65} = 5.330; p < .001; Cohen's f² = 0.410, indicating that symptom dimensions differed significantly between the groups, with a large effect size. The observed statistical power (1-β = 0.980) indicated high sensitivity and sufficient power to detect multivariate effects as small as f² = 0.185 at 80 % power. JTC patients scored significantly higher than non-JTC patients on symptom dimensions. No differences in depressive symptoms were found. Table 3 describes the results of the univariate tests.

Table 3.

Differences between JTC and non-JTC patients in symptom dimensions.

Symptom dimensions	JTC (N = 15)		non-JTC (N = 60)		F	p	Partial ŋ2
	Mean	SE	Mean	SE
Positive	11.915	0.958	7.588	0.468	16.083	0.000	0.189
Negative	17.225	1.340	12.627	0.654	9.277	0.003	0.119
Disorganised	15.639	0.918	11.740	0.448	14.229	0.000	0.171
Excited	5.555	0.527	3.961	0.258	7.198	0.009	0.094
Depression (CDSS)	3.170	1.028	3.607	0.502	0.142	0.707	0.002

Means adjusted for gender, age, educational level and premorbid IQ.

SE: Standard error. Partial ŋ²: partial eta squared (effect size). CDSS: Calgary Depression Scale for Schizophrenia. JTC: jumping to conclusions bias.

NOTE. Interpretation of effect sizes (partial ŋ²): 0.01, small; 0.06, medium; 0.14 or above, large.

The PERMANOVA revealed a significant multivariate effect of JTC bias on symptom dimensions (pseudo-F = 9.006; R² = 0.103; p < .001). Covariates (age and premorbid IQ) also contributed to the model, with slight significance (p < .05). Homogeneity of dispersion tests (betadisper and permutest) indicated that group differences reflected centroid separation rather than unequal variances. Fig. S1 (Supplementary materials) illustrates the multivariate distribution of the two groups.

The results of univariate permutation tests are shown in Table 4, which displays the permutation-based F statistics, R² values and p-values for JTC bias across each outcome. JTC bias was significantly associated with higher scores on all symptom dimensions, but not with depressive symptoms. Given the observed heteroscedasticity and nonnormality, the PERMANOVA provides the more robust p-values.

Table 4.

Univariate permutation tests for JTC bias.

Symptom dimensions	R2	F	Perm. p
Positive	0.170	16.083	0.0004
Negative	0.102	9.277	0.0034
Disorganised	0.150	14.229	0.0012
Excited	0.091	7.198	0.0098
Depression (CDSS)	0.002	0.142	0.7096

CDSS: Calgary Depression Scale for Schizophrenia. JTC: Jumping to conclusions bias. Perm: Permutation.

NOTE: Number of permutations = 4999.

3.4. Correlations between symptoms, JTC and cognitive variables

Regarding demographics, age was correlated with negative (r = −0.26; p < .05) and disorganised/cognitive symptoms (r = −0.23; p < .05). Educational level also correlated with negative (r = −0.26; p < .05) and disorganised/cognitive symptoms (r = −0.31; p < .01). No significant correlations were found between cognitive domains and demographic variables.

The “correlation heatmap” among the studied variables is represented in Fig. 1.

Fig. 1.

Correlation heatmap: symptom dimensions and cognitive variables.

POS21: Positive. NEG21: Negative. COG21: Cognitive/disorganised. EXC21: Excited/aggressive. Depression: Calgary depression scale (CDSS). Neurocognition: Screen for Cognitive Impairment Psychiatric Scale (SCIP-S). Premorbid IQ: TAP-30 (from Spanish, Test de Acentuación de Palabras). JTC bias: Jumping to conclusions bias. Beads Task: DTD (number of draws to decide). FER: Facial emotion recognition. Hinting: Hinting task (ToM). ^⁎p < .05; ^⁎⁎p < .01; ^⁎⁎⁎p < .001.

When analysing positive symptoms at the separate symptom level, delusions were significantly correlated with JTC bias (r = 0.39; p < .001) and ToM (r = −0.27; p < .05). Hallucinatory behaviour correlated with JTC (r = 0.35; p < .01) and FER (r = −0.28; p < .05). Finally, unusual thought content correlated with JTC (r = 0.29; p < .05) and FER (r = −0.33; p < .01).

4. Discussion

To the best of our knowledge, this is the first study to investigate whether patients with EP who tend to jump to conclusions differ from those who do not in terms of symptom dimensions. Although the existing literature has extensively investigated the relationship between JTC bias and positive symptoms, no prior work has examined how JTC relates to a broader range of symptom dimensions.

Additionally, we explored the associations between positive symptoms, JTC bias and cognitive domains.

4.1. Symptoms

The PANSS version used in this study (Dal Santo et al., 2024) obtained satisfactory reliability indices. This fact could be an important feature because this version was developed to capture the core symptoms of FEP.

The results demonstrated that the groups differ significantly across symptom dimensions, with a large effect size. Patients in the JTC group scored higher than those in the non-JTC group on symptom dimensions. The observed statistical power of the multivariate analyses was robust and sensitive. The parametric MANCOVA and the nonparametric PERMANOVA converged on the same substantive conclusion: JTC bias is associated with a multivariate difference in symptom dimensions, and the effect is driven primarily by positive, negative and disorganised/cognitive symptoms.

JTC is a well-established CB in schizophrenia and has been related to positive symptoms, being present from the early phases of illness (Gawęda et al., 2024; González et al., 2018; Samson et al., 2024). Likewise, our results showed that the JTC group also scored higher on the negative and cognitive/disorganised dimensions, in line with Hayashi et al. (2022), who found that JTC correlates with negative symptoms. Other studies have reported significant associations between cognitive and negative symptoms (Engen et al., 2019) and have demonstrated the central role of negative symptoms and their link to psychosocial functioning in FEP samples (Chang et al., 2020).

Furthermore, at the symptom level, delusions, hallucinatory behaviour and unusual thought content were correlated with JTC. Delusions have been associated with a tendency to jump to conclusions (Dudley et al., 2016; McLean et al., 2017; So et al., 2016). Additionally, JTC bias was more prevalent in deluded than in non-deluded patients (Garety and Freeman, 2013) and appeared to be specific for emotional and cognitive dimensions of delusions (Gawęda et al., 2017). Moreover, JTC is present in FEP and is associated with delusions (Falcone et al., 2015).

CBs are not simply epiphenomena of psychosis but active elements in the maintenance of symptoms and functional impairment (Ahuir et al., 2021). In this sense, JTC bias is central to psychological theories of delusion (Freeman and Garety, 2014). Nevertheless, there is no consensus among contemporary studies on the relationship between JTC bias and delusions. Contrary to our results, Xenaki et al. (2022) reported that the severity of positive symptoms is not associated with JTC bias in patients with FEP. The same authors proposed possible reasons for their discrepancies from previous literature, including differences in study samples, evaluation timelines, instruments used to assess symptoms (e.g., the PANSS), delusion content, or tasks used to assess data-gathering bias (e.g., the Beads task). Nonetheless, other explanations could be proposed. Samson et al. (2024) found a medium-to-large effect size when examining the association between “extreme responding” in JTC and the severity of delusions. In the current study, we assumed that JTC bias is present when a patient requires one or two DTD on the Beads task as defined by Garety et al. (2011), thereby creating a binary variable to categorise patients as JTC or non-JTC. In contrast, other studies, such as that of Xenaki et al. (2022), operationalised JTC bias as a continuous variable, using DTD as an indicator of data-gathering bias (the lower the DTD, the greater the JTC bias). Furthermore, studies employing variants of the Beads task (with distinct levels of difficulty or content) to assess data-gathering bias have obtained mixed results regarding its presumed relationship with delusions (Moritz et al., 2020; Pytlik et al., 2020; Romero-Ferreiro et al., 2022). We used the most challenging version of the traditional Beads task (ratio of 60:40), as it is more sensitive to detecting effects related to psychotic symptoms (Catalan et al., 2022). Within our sample, only 20 % of participants exhibited a JTC bias. This proportion is somewhat lower than that typically reported in meta-analyses and multicentre studies (Dudley et al., 2016; Tripoli et al., 2021), but comparable to recent studies in FEP. For example, Díaz-Cutraro et al. (2022) reported that only 14 % of individuals showed a JTC bias when comparing a sample of FEP patients under the same parameters as us (ratio of 60:40; DTD ≤ 2 in the Beads task). These disparities may be due to sample attributes, such as relatively preserved cognitive functioning or less severe delusional symptoms, or to methodological issues in task administration. Notably, recent research has questioned whether JTC bias is genuinely related to delusions and whether the Beads task is a valid tool to assess data-gathering bias (Doherty et al., 2025). Previous meta-analyses revealed only weak associations between JTC and delusional severity (Dudley et al., 2016). In contrast, Bayesian modelling suggests that task parameters exert more influential effects on DTD than delusional ideation itself (Tan et al., 2024). Moreover, investigations of clinical HR samples show stronger links between JTC and cognitive factors, such as working memory and intolerance of uncertainty, than with delusional severity (Broome et al., 2007). Our findings, however, align with traditional cognitive models, supporting the association between JTC bias and psychotic symptoms. In any case, we should consider the view that JTC is not a universal marker of delusional reasoning but rather a variable cognitive bias shaped by task design and sample context.

Notably, studies assessing JTC by means of self-report instruments have also revealed significant correlations with psychotic-like experiences in healthy or HR individuals, as well as with positive symptoms in FEP or schizophrenia (e.g. Livet et al., 2020; Nguyen et al., 2025; Pena-Garijo et al., 2022).

4.2. Neurocognition and social cognition

Significant between-group differences were found, with JTC patients scoring lower on ToM and FER, but not on neurocognition. ToM and FER impairments have been demonstrated to be important predictors of clinical outcomes in schizophrenia-spectrum disorders (Green et al., 2019). JTC has been associated with FER and ToM deficits, suggesting that patients who jump to conclusions are impaired in processing social information, which is linked to worse FER (Díaz-Cutraro et al., 2022). FER impairment, especially fear and anger recognition, has been proposed as a potential vulnerability marker of psychosis (Pena-Garijo et al., 2023; Tripoli et al., 2022). Furthermore, we found that FER was related to hallucinatory behaviour and unusual thought content. FER deficits have been linked to positive symptoms across the psychosis continuum (Bosnjak Kuharic et al., 2019; Pena-Garijo et al., 2023) and to more severe delusional distress (Mehl et al., 2020). However, Larsson et al. (2022) found that FER is impaired in FEP but is not associated with psychotic symptoms.

Moreover, ToM was related to delusions. Previous research has demonstrated that ToM is more impaired in patients with FEP than in healthy controls (Bora and Pantelis, 2013; Lindgren et al., 2018; Takeda et al., 2018) and is the strongest predictor of functionality among social cognitive domains (Vass et al., 2023). We also found that ToM was related to JTC. However, Díaz-Cutraro et al. (2022) did not find a relationship between JTC and ToM in FEP. It should be noted that, in their study, ToM was assessed using only three stories from the Hinting Task, whereas we used the complete ten-story version. Although the Hinting task is a reliable measure of social cognition in FEP, some psychometric problems (e.g., ceiling effects) have been observed (Lindgren et al., 2018). These could be possible explanations for these inconsistencies.

In addition, our study revealed a strong correlation between ToM and neurocognition, as well as with the cognitive/disorganised dimension, consistent with a recent meta-analysis that highlighted the close association between ToM and the cognitive/disorganised dimension in psychotic disorders (Thibaudeau et al., 2023). Furthermore, cognitive tests have been observed to be strongly connected with ToM in FEP (Catalan et al., 2018).

No significant relationships were observed between neurocognition and JTC bias. Cognitive impairments affect up to 80 % of young people with EP and are present in FEP and chronic psychoses (Rodriguez-Jimenez et al., 2019; Tschentscher et al., 2023). A greater tendency to jump to conclusions has been observed in EP patients with neuropsychological deficits (Tripoli et al., 2021). It should be noted that although neurocognitive deficits are marked in FEP patients, their severity varies among individuals (Catalan et al., 2024). Nevertheless, our study identified a correlation between the number of DTD and both neurocognition and premorbid IQ, supporting the presumed link between data-gathering bias and neuropsychological performance (Takeda et al., 2018; Tripoli et al., 2021).

Finally, patients with JTC were more frequently males, in line with recent studies that observed sex-specific patterns in JTC among FEP patients (Díaz-Cutraro et al., 2025). Although no between-group differences were found in age, education, premorbid cannabis use or antipsychotic dose, within the entire sample, age and education were negatively related to negative and disorganised/cognitive symptoms. No significant associations were found between social cognition and demographic variables. However, FER and ToM were related to premorbid IQ, partially agreeing with Casado-Ortega et al. (2021), who reported that IQ and age are influential variables for social cognition in FEP samples.

4.3. Clinical implications

Research has widely investigated the relationships between JTC bias and positive symptoms, particularly delusions, but these associations with other symptom dimensions in EP remain understudied. The principal novelty of the current study was to explore whether differences across a broader range of symptom dimensions are observed in individuals with EP who tend to jump to conclusions, compared to those who do not. Our study showed that patients with JTC are characterised by greater severity across all symptom dimensions and poorer functioning than those without this bias. Additionally, few studies have examined how JTC relates to social cognition in EP. We found significant associations between JTC bias and impairments in social cognition, such as ToM and FER. Moreover, an increased JTC bias has been demonstrated among HR individuals who transitioned to psychosis (Catalan et al., 2022). Likewise, recent investigations involving non-clinical adolescent samples support that emotional processes and CBs are strongly interconnected in psychosis vulnerability (Xi and Wang, 2025).

Furthermore, recent studies suggest that JTC bias may serve as a predictive marker for the trajectories of delusional ideation in the general population, indicating its potential utility in early identification and intervention strategies (Kuhn et al., 2023). Assessing JTC bias and cognitive skills could help identify individuals at higher risk of psychosis and provide early, personalised interventions. In this regard, Díaz-Cutraro et al. (2022) recommend early interventions to modify CBs and improve FER skills for patients with FEP who jump to conclusions.

Since psychological therapies addressing CBs, such as MCT, have been shown to improve psychotic symptoms, even in EP (Hua and Kanas, 2020), integrating strategies to reduce JTC into early treatment plans could improve clinical outcomes in this population.

4.4. Limitations, strengths and future research

Some limitations should be pointed out. Firstly, the cross-sectional design prevents us from establishing causal relationships between JTC bias and symptom dimensions, affecting our understanding of their temporal interaction. Likewise, the relatively modest sample size and participants recruited from a single geographical area suggest that the results should be interpreted cautiously when generalising. Second, excluding individuals with substance dependence could also limit the generalisability of the results, given that it is a common comorbidity. Third, the imbalance in group sizes and violations of some parametric normality assumptions may have compromised the stability of the multivariate model. Next, most patients were taking antipsychotics at the time of evaluation, which is a confounding factor that should be taken into consideration. In addition, we utilised the SCIP-S to assess neurocognition. Although a detailed exploration of cognitive deficits is beyond the scope of this study, we acknowledge it as a limitation. As well, other well-established CBs in psychosis research were not evaluated. Lastly, while network analysis is increasingly recommended for the study of symptom dimensions, its application in the present work was not feasible due to sample size constraints. Consequently, we opted for more parsimonious multivariate methods that provide greater reliability under these conditions.

On the other hand, some strengths of our study can be highlighted. The major strength lies in the novelty of identifying a distinct presentation of symptom dimensions in individuals with EP who tend to jump to conclusions, compared to those who do not, with large effect sizes. Furthermore, all participants were recruited and evaluated in a routine clinical setting, thereby strengthening the external validity of the results. Next, we minimised potential methodological limitations by using robust statistical procedures, combining parametric and nonparametric analyses, and conducting sensitivity analysis. These methods converged on the same substantive conclusions, demonstrating the statistical power and enabling the results to be contrasted, thereby improving their statistical validity. Finally, we sought assessment tools that were validated in Spanish populations and had previously been used in studies of patients with EP. Notably, in our study, we employed a recently proposed version of the PANSS (Dal Santo et al., 2024) specifically developed to capture the core symptoms of FEP. It showed satisfactory reliability, indicating its suitability for use in EP samples. This fact could represent an important feature. Moreover, the choice of the SCIP-S for assessing neurocognition is justified because it appeared valuable in routine assessments (Schmid et al., 2021) and has been validated in EP settings (Zbukvic et al., 2025), allowing clinicians to save time and resources, ensuring that cognition is routinely assessed (Stainton et al., 2025).

Future research should focus on recruiting larger, more diverse samples, including participants from different sociodemographic backgrounds, and on employing multicentre designs to enhance the generalisability of results across various clinical and cultural settings. It would help determine whether JTC bias differs across these conditions. Likewise, longitudinal designs could support understanding how JTC manifests over the course of the disease. Moreover, more comprehensive cognitive assessment tools should be used to describe better cognitive skills and their associations with CBs and social cognition. Furthermore, instruments commonly used to evaluate social cognition, including ToM and FER, should be explicitly validated in EP samples. More naturalistic approaches, such as mobile health tools, could clarify how these variables work in real-world contexts. Importantly, future studies should examine whether the Beads task is a valid measure of JTC bias. Testing alternative task versions and paradigms would help to determine whether it reflects a consistent cognitive tendency or shifts with context. In addition, traditional clinical instruments (e.g., the PANSS) should be adapted to better capture key symptoms in EP, ideally utilising dimensional rather than categorical frameworks. Finally, future research should employ advanced analytical methods to explore the heterogeneity among patients with EP (Cuesta et al., 2025). They should include data-driven procedures, such as cluster analysis, to identify distinct profiles, as well as dimensional approaches (e.g., network analysis) and novel IA-based techniques (e.g., machine learning).

Refining study designs, assessment tools and analysis methods could be essential for connecting basic cognitive knowledge with clinical practice in EP. This guideline could finally lead to personalised interventions that improve clinical and functional outcomes.

5. Conclusions

This study is pioneering in investigating whether symptom dimensions differ between patients with EP who tend to jump to conclusions and those who do not. No prior work has investigated how JTC relates to a broader range of symptom dimensions. The results demonstrate that patients who exhibit a JTC bias present with a different clinical picture, characterised by greater severity across all symptom dimensions, mainly the positive, negative and disorganised dimensions, compared with those who do not. Moreover, JTC is associated with worse social cognition, including ToM and FER. Furthermore, at the symptom level, JTC bias is related to positive symptoms, involving delusions, hallucinations and unusual thoughts.

Beyond symptoms, addressing CBs and social cognition in the assessment and treatment of patients with EP may be crucial, as psychological interventions targeting these features have been shown to improve clinical and functional outcomes. Further research is needed to address the limitations of this study and replicate its findings.

CRediT authorship contribution statement

Josep Pena-Garijo: Writing – original draft, Validation, Supervision, Resources, Investigation, Formal analysis, Data curation, Conceptualization. María José Masanet: Writing – original draft, Supervision, Resources, Project administration, Investigation. Ana Palop-Grau: Writing – original draft, Validation, Supervision, Resources, Project administration, Investigation, Data curation. María Lacruz: Writing – original draft, Validation, Supervision, Resources, Project administration, Investigation, Data curation.

Declaration of competing interest

The authors declare no conflicts of interest.

Appendix A. Supplementary data

Supplementary material