This project explores how the brain processes speech in individuals with schizophrenia, with a focus on the role of cognitive ability. Using functional MRI data, we examine whether differences in intelligence help explain variability in brain responses during speech perception.
About 1 in 345 people worldwide live with schizophrenia, and many experience difficulty understanding spoken language, even during everyday conversations [1]. Speech perception relies on several brain regions that process sounds and meaning. The auditory cortex helps detect and interpret sounds, while areas such as the right cerebellar Crus I support higher-level cognitive processes involved in language comprehension.
Neuroimaging studies have shown that people with schizophrenia often show altered activity in these speech-related brain regions compared to healthy controls. Previous research has mainly tried to explain these differences by separating patients based on whether they experience auditory hallucinations, but this does not fully explain why brain activity during speech processing varies so widely across individuals [2].
This raises an important question: could cognitive ability, particularly IQ, help explain variability in speech-related brain activity in schizophrenia?
In this project, we examined how brain activation during different speech conditions varies with IQ. We focused on two regions involved in speech and cognitive processing: the auditory cortex and the right cerebellar Crus I.
To investigate this question, we analyzed a publicly available fMRI dataset from OpenNeuro, which includes individuals with schizophrenia and matched healthy controls. The dataset contains information from 77 participants, including sex, age, IQ score, and diagnostic group. Because this project primarily focuses on IQ, we imputed missing IQ values for some participants using mean imputation (an IQ score of 100). The diagnostic groups consist of healthy controls (HC), individuals with schizophrenia who experience auditory verbal hallucinations (AVH+), and individuals with schizophrenia who do not experience hallucinations (AVH-). An additional variable, PSYRATS-H (hallucination severity, is included for participants with schizophrenia who experience hallucinations.
Figure 1. Participant's information that includes age, sex, IQ, and group
Figure 2. Three auditory tasks: words, sentences, and reversed speech
During MRI scanning, participants completed three auditory tasks:
A block design was used, with each listening condition presented in randomized 26-second blocks. White noise served as a baseline between task blocks. The experiment lasted approximately 11 minutes. Audio was delivered through MRI-compatible headphones, and participants viewed a neutral gray screen throughout the task. Functional brain images were collected using a 3-Tesla MRI scanner.
Before analysis, the brain images went through standard cleanup steps to reduce noise and improve accuracy. This included correcting for small head movements, smoothing the images to reduce random variation, and removing slow signal drifts over time. Structural brain images were also processed to isolate brain tissue and align all participants’ data to a common brain template, making comparisons across people possible.
We analyzed each participant’s brain data separately using the General Linear Model (GLM) in FSL. We modeled three speech conditions: sentences, words, and reversed speech, and used a gamma hemodynamic response function (HRF) to reflect the delayed brain blood-flow response. White noise served as the baseline, and head movement was included to reduce noise. We then created activation maps for each condition (e.g., sentences > baseline), which were used later for group analysis.
After analyzing each person individually, we combined results to study group patterns using mixed-effects modeling in FSL, which supports conclusions about a broader population. We first estimated average activation within each group (Healthy Controls, AVH+, AVH−).
Next, we tested whether IQ was related to activation by adding IQ as a continuous variable. We also controlled for age and sex to reduce confounding.
We focused on two speech-related regions: the Auditory Cortex (sound processing) and Right Cerebellar Crus I (higher-level language and cognition). For each participant, we measured activation during sentences, words, and reversed speech, then used linear regression to test whether activation was associated with IQ within each group.
In the sections below, you’ll first see the main whole-brain findings, including group mean activation patterns and covariate effects. Then, you’ll interact with two visualizations that focus on ROI-level relationships and broader variable comparisons.
As you explore, consider how overall activation patterns, covariate effects, and group differences might support or challenge the study’s main hypothesis.
We first examined whole-brain average activation patterns in participants with schizophrenia in the Auditory cortex region.
Figure 3. Whole-brain activation maps for the reversed speech contrast in participants with schizophrenia. Warm colors indicate significant activation (Z > 2.3, cluster-corrected p < 0.05), with yellow showing stronger effects than red.
Next, we tested whether activation changed as a function of IQ while controlling for age and sex. The map below shows the Words contrast in the AVH− group in the Right cerebellar Crus I, which was the only group and condition where a positive association with IQ survived correction.
Figure 4. Whole-brain activation maps showing the effect of IQ during the words contrast in the AVH− group. Highlighted regions indicate voxels where activation increases with IQ (Z > 2.3, cluster-corrected p < 0.05).
Quick guide
Quick guide
In addition to the auditory cortex, we examined the right cerebellar Crus I, a region in the posterior cerebellum that has been linked to higher-level cognitive and language-related processes. Prior research suggests that this region may play an important role in syntactic and cognitive aspects of speech processing, making it a relevant target for understanding variability in schizophrenia [3].
Unlike the auditory cortex, the right cerebellar Crus I showed different patterns of association with IQ across participant groups. In the healthy control group, IQ was positively associated with activation across all three speech conditions. Among participants with schizophrenia who experienced auditory verbal hallucinations (AVH+), the relationship between IQ and activation was negative across all contrasts. In contrast, participants with schizophrenia who did not experience hallucinations (AVH−) showed positive associations across all contrasts.
Figure 3. Right cerebellar Crus I region of interest.
Table 1. Linear regression results for IQ and ROI activation (COPE values) across groups in words contrast. Significant results (p < 0.05) are highlighted in red.
The clearest finding emerged in the Words condition, where the relationship between IQ and right cerebellar Crus I activation was statistically significant in both schizophrenia groups. For the AVH+ group, higher IQ was associated with lower activation in right Crus I, which alligns with previous researches that talks about this idea as well [4, 5]. Meanwhile, for the AVH− group, higher IQ was associated with higher activation. This suggests that cognitive ability may relate to speech processing differently depending on hallucination status.
Overall, these findings suggest that the right cerebellar Crus I may capture important cognitive differences in how speech is processed in schizophrenia. Rather than reflecting basic auditory perception alone, this region may be more sensitive to the higher-level language and cognitive processes that vary across individuals.
This study examined how IQ relates to brain activation during speech processing. All groups showed consistent activation in the auditory cortex, highlighting its key role in speech perception. However, IQ was not significantly related to activation in this region. In contrast, IQ-related effects appeared in right cerebellar Crus I, suggesting that cerebellar regions may reflect cognitive differences in language processing.
The results of this analysis should be interpreted within the scope of a limited dataset. The available data included only sex, age, IQ, and participant group for each participant, which restricted our ability to account for other potentially important confounding factors that may influence brain activation. In addition, the relatively small number of participants in each group may limit how well these findings generalize to broader populations.
Future work could include larger and more diverse datasets with additional demographic and clinical variables to better understand factors that shape brain activation patterns. Including other cognitive measures and tasks could also help identify whether these effects extend beyond auditory and language processing to other brain systems.
Interested in the project or want to collaborate? Feel free to reach out to any of us. We’re always happy to discuss the research, methods, or future directions.
