2.1. Participants

Participants were recruited via convenience sampling. First year psychology students were recruited and awarded course credit for participation. The participants had to have self-reported normal or corrected-to-normal vision. No other inclusion and exclusion criteria were employed. The sample was made up of 54 participants (22 males, 31 females, and 1 participant preferred not to say). Age ranged from 18 to 54, M = 23, SD = 7.38. No missing data was recorded. The recruitment period begun on the 13^th of May 2022 and ended on the 11^th of August 2022. Informed (written) consent was obtained from all participants prior to the beginning of the online study.

As there is no agreed-upon sample size estimation for generalized linear mixed models (GLMM) [68], we used a sample size estimation for a repeated measures ANOVA to decide the required power, using G*Power [69]. Since GLMMs are considered more powerful than ANOVA’s [70], our estimate was a conservative one. Firstly, we used a partial eta-squared of 0.06 from the McEwen et al. [32] study, which used a similar design, investigating the interaction between ADHD group and horizontal and vertical biases in MIB disappearances. We calculated partial eta-squared from the raw data from that study (https://osf.io/m8zxc/files/osfstorage), when performing a mixed ANOVA with the factors: ADHD group (high/low; median split on Conners Adult ADHD Rating Scales: < 52 low, ≥ 52 high) and location (same 4 locations as in the current study). Critical alpha was set at 0.05, and a desired power at 0.80. The calculations estimated a required sample size of 34. We exceeded this with our sample of 54 participants. This increase sample size was obtained because we had set out a data collection window, and finished earlier collecting the sample size that the power analysis indicated. We decided to continue collecting, so that we could potentially find less strong effects.

2.2. Materials

2.3. Stimulus

The experiment was conducted online, using ‘Gorilla Experiment Builder’ (www.gorilla.sc). Descriptions below describe how the animation was constructed, but for example, stimulus size would have depended on the participants’ set-up. However, to compensate for any potential variability that may have been introduced depending on the participants’ set-up, it was encouraged that participants sat approximately 30 centimetres from the screen. Additionally, the experiment was limited to only laptops and desktops; therefore, ruling out use of smaller screens such as phones or tablets, that may have added more variability.

The MIB visual trials included all possible 15 different combinations of where yellow dots appeared (i.e., upper left [UL], upper right [UR], lower right [LR], and lower left [LL]), allowing for trials with 1,2, 3, or 4 yellow dots. Stimuli were presented on a black background. The moving stimulus included 192 blue squares (5x5 pixels; 0.1 cm x 0.1 cm) rotating as if placed with a uniform random distribution on a virtual sphere (8 cm x 7.5 cm). Yellow dots (10-pixel diameter; 0.2 cm x 0.2 cm) were equally spaced across each quadrant and were placed just within visual projection of the sphere. Blue and yellow are conventional colours [17] and work well within the paradigm. Brightness of the yellow dots (opacity) were set through pilot testing, to ensure about half of the dots disappeared during a trial. This was to avoid floor and ceiling effects as much as possible. A static white cross (0.7 cm x 0.7 cm) in the centre of the screen served as a fixation point for participants. Fig 1 demonstrates an example frame of an MIB stimulus with two targets. The experiment was run in full-screen mode, with the stimulus taking up 100% of the screen.

2.4. Procedure

The ethics application for this study was approved by the University of Canberra Human Research Ethics Committee-11597. Students who chose to enrol in the study were taken to the external site ‘Gorilla Experiment Builder’ (www.gorilla.sc), where the experiment was hosted [73]. From here, participants were provided with general information outlining the project, confidentiality, data storage, risks, and consent. Once participation was agreed upon by the participant, they were asked demographic questions such as gender and age. Next, instructions were provided to explain how the MIB trials were to be completed. Participants were instructed to fixate on the white cross. Each trial began once the spacebar was pressed. There were a total of 200 randomized trials, each lasting four seconds. Once a trial finished, participants were asked to select the yellow dots that were visible at the end of each trial by clicking on small markers on a response screen at the four possible locations of the yellow dots (i.e., UL, UR, LL, LR). Furthermore, to prevent fatigue and possible eye strain, participants were encouraged to take self-paced breaks in between trials when necessary. Participants completed eight practice trials to become familiar with the process before beginning the recorded trials. After the MIB task, participants completed the SPQ and the PAS. To limit any response bias (i.e., social desirability) in the questionnaires, their true nature was disguised, and renamed to the ‘perception of self’ questionnaire and the ‘body sensation’ questionnaire, respectively.

2.5. Data analysis

Because participants indicated which locations they perceived a dot, we could calculate which dots had perceptually disappeared. For example, the dots that were physically present, but not reported as visible. We also checked that participants did not have too many “false alarms” (i.e., reporting a dot as visible where no dot was presented). Several incorrect responses would suggest the individual did not sincerely respond, and this action would be more evident towards the end of the trials where fatigue/boredom may have set in. The second half of participants’ responses were checked to ensure there were no more than 3 incorrect responses for dots that were not presented in the preceding trial. No participants were excluded based on this criterion.

Participants were excluded if they had completed the entire experiment in under 15 minutes, as it was calculated that the minimum time required for the MIB trials was almost 13 minutes. Three participants were excluded based on this criterion. The reported number of participants in this study is that after these exclusions were implemented.

2.6. Statistical analysis

The data was analysed using the program jamovi version 2.3.2 [74]. We performed a generalized linear mixed model (GLMM) using the GAMLj module. We used a binomial distribution and logit link function (i.e., a logistic regression), which is most appropriate for our binary dependent variable ‘disappearance’ of a dot (0 visible, 1 invisible). Two independent variables for locations were created, for which we coded the locations of the dots (UL, UR, LL, LR) into Horizonal (left/right), and Vertical (top/bottom). Another, continuous independent variable was SPQ (or PAS in other analyses). The model included a random intercept to account for individual differences (coded as the identifier, ID, of a participant). The final model was: Disappearances ~ 1 + Vertical + Horizontal + SPQ + Vertical:Horizontal + Vertical:SPQ + Horizontal:SPQ + Vertical:Horizontal:SPQ + (1 | ID), where colons indicated interactions between terms.

The GLMM therefore analysed the effect of schizophrenia traits (measured by the SPQ and PAS) and target location (lower/upper/left/right) on the strength of the MIB illusion (i.e., the disappearance of a dot). The GLMMs included both horizontal and vertical factors, but are written up in separate sections for ease of discussion.

3.1. SPQ as a main effect against disappearances

We tested whether SPQ predicted MIB disappearances. Firstly, beta estimates (β = −0.023 [CI = −0.053 to 0.007], SE = 0.016) and a visual inspection of a scatterplot (Fig 2) suggested higher SPQ scores resulted in less perceptual disappearances. However, counter to our hypothesis, the GLMM revealed that SPQ scores were not a significant predictor for MIB disappearances (χ²(1) = 2.18, p = .139).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Scatter Plot of SPQ Against Average of MIB Disappearances.

The grey dots represent participant averages.

https://doi.org/10.1371/journal.pone.0325609.g002

3.2. The interaction between SPQ scores and horizontal locations against disappearances

Our second hypothesis predicted that an interaction between SPQ scores and horizontal locations, in which higher SPQ scores would demonstrate a decreased left field bias (less pseudoneglect). This hypothesis was not supported by the data. The GLMM indicated that the interaction between SPQ and Horizontal location was not significant (χ²(1) = 0.49, p = .485). Therefore, less pseudoneglect amongst higher SPQ scores could not be concluded.

3.3. The interaction between SPQ scores and vertical locations against disappearances

Our final hypothesis stated an interaction between SPQ scores and vertical locations would be a predictor for MIB disappearances, in which higher SPQ scores would be correlated with a top field bias. The interaction was significant (χ²(1) = 12.61, p < .001). Hence, the number of MIB disappearances was dependent on this interaction (β = −0.005 [CI = −0.008 to 0.002], SE = 0.001). Additionally, a visual inspection of the data implied a top field bias amongst higher scoring participants (Fig 3). This evidence together, supported our third hypothesis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Scatter Plot of SPQ and Top Field vs Bottom Field Against Average of MIB Disappearances.

A blue and red dot pair represents each participant.

https://doi.org/10.1371/journal.pone.0325609.g003

4.1 PAS scores

A GLMM was run for PAS scores against MIB disappearances with horizontal and vertical locations included. The results of the GLMM showed a significant interaction between PAS and vertical locations (χ²(1) = 4.48, p = .034), demonstrating a similar top field bias as the SPQ interaction. Additionally, a significant interaction between PAS and horizontal locations (χ²(1) = 7.31, p = .007) was observed, which was not found in the SPQ analyses. Those individuals with lower PAS scores demonstrated a left field bias on average, but as PAS scores increased, a right field bias emerged.

4.2 SPQ factors

The SPQ and PAS are suggested in the literature to be correlated [75], and this was so also in our data (r = .65, p < .001). To understand why the SPQ and PAS show different effects in our data, SPQ factors (cognitive-perceptual, interpersonal, and disorganized [16]) were explored separately. A correlation matrix revealed that cognitive-perceptual (r = .64), interpersonal (r = .60), and disorganized (r = .61) shared a significant moderate relationship with the PAS.

A GLMM for each factor demonstrated cognitive-perceptual (χ²(1) = 9.86, p = .002) and interpersonal (χ²(1) = 11.36, p < .001) factors had significant interactions with horizontal locations. Interestingly, cognitive-perceptual demonstrated a negative beta estimate (β = −0.024 [CI = −0.040 to –0.009, SE = 0.008), while interpersonal had a positive beta estimate (β = 0.032 [CI = 0.013 to 0.050], SE = 0.009). These results imply that the cognitive-perceptual factor had a stronger right field bias at lower scores, and a stronger left field bias at higher scores (Fig 4). On the contrary, the interpersonal factor had a stronger left field bias at lower scores, and a stronger right field bias at higher scores (Fig 4). Meanwhile, the disorganized factor was not significant (χ²(1) = 0.73, p = .394; Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Effects Plot.

Cognitive-Perceptual Factor (left), Interpersonal Factor (middle) and Disorganized Factor (right), and Left Field vs Right Field Against MIB Disappearances. The shaded area refers to 95% confidence intervals of the two lines (group means).

https://doi.org/10.1371/journal.pone.0325609.g004

To conclude, although SPQ had a moderate significant relationship with its factors and the PAS, SPQ overall did not show a significant interaction with horizontal location. Although, two of its factors (cognitive-perceptual and interpersonal) did, but in opposite directions.

5.1. SPQ and its factors

Although studies including the SPQ and MIB are limited, there is a moderate amount of literature supporting weaker MIB in schizophrenia [20,35,65]. Although a significant effect did not exist in our data, the trend was in the same direction, with higher SPQ scores resulting in weaker MIB. Perhaps, because we researched schizophrenia traits, and not a sample of individuals diagnosed with schizophrenia, our effect is less strong.

We investigated the interaction between traits and the vertical spatial biases in disappearances, and therefore attention. Our results showed that as SPQ scores increased, so did a bias towards the top field. Since a bias to the bottom visual field is associated with intact dorsal stream function, our results are consistent with previous literature that hypothesizes that schizophrenia is associated with deficits in dorsal stream [39–41,43,44].

We also predicted higher traits would show less pseudoneglect (left field bias) on MIB trials, based on literature suggesting a right hemispheric dysfunction in schizophrenia [36,42,51,53,54,56]. However, our results did not reflect a significant effect. These results do not replicate literature findings, in which the majority of studies were able to demonstrate limited pseudoneglect in schizophrenia.

This unusual finding warranted further exploration. We therefore looked at the three factors of the SPQ: cognitive-perceptual, interpersonal, and disorganized [15]. Our rationale was that possibly one or more of the factors showed opposite effects on disappearance. Significant links between both cognitive-perceptual and interpersonal factors and SPQ were found, but not with the disorganized factor. Interestingly, the cognitive-perceptual factor showed that higher scores were associated with an increase left field bias. This goes against the reduced left field bias that is generally observed in schizophrenia [36,42,53,54,56]. However, the link to the cognitive-perceptual factor was not investigated in those studies. There does not appear to be existing literature to explain the positive correlation between the cognitive-perceptual factor and left field bias, which is therefore an interesting topic for future study. The interaction between the interpersonal factor and horizontal locations was significant, and suggested that there was less pseudoneglect as scores increased. This was consistent with what we expected to see between overall SPQ scores and horizontal locations. Though speculative at the moment, the correlation between interpersonal traits and pseudoneglect may exist because the right hemisphere has been found to be important for social interaction (i.e., recognizing faces [52]). Our significant results that show less pseudoneglect when considering the interpersonal factor, could be an indirect reflection of right hemisphere dysfunction in schizophrenia. If such a finding could be rigorously replicated, it could suggest that the interpersonal factor may help identify social deficits and right hemispheric dysfunction in schizophrenia; while also being a good measure of spatial biases in general. This last point is supported by Chen et al. [10] who found that the interpersonal factor is associated with poorer attention, while the cognitive-perceptual factor was not. Our results highlight that considering the cognitive-perceptual, interpersonal, and disorganized factors, in addition to the overall SPQ score, could be beneficial in perception studies.

5.2. PAS

Unlike the SPQ, the PAS had significant interactions with both horizontal and vertical locations. This difference was surprising because SPQ and PAS share a moderate correlation as per our study and the literature [75]. Therefore, it was speculated the PAS would show similar results to the SPQ analysis. Our speculative interpretation of these findings is the PAS and the interpersonal factor of the SPQ share similar characteristics of self-perception and how others perceive oneself, because the origins of the PAS include a strong emphasis on perception of body-image [8]. Indeed, the PAS and interpersonal factor share a strong significant correlation in our data.

5.3. Spatial biases

Aside from our focus on schizophrenia, our results also complement studies that have demonstrated differences in attention measured by MIB. For instance, our analysis confirmed that there are differences in MIB disappearances between the top and bottom visual field. The exploratory analyses highlighted that perceptual differences between the left and right visual field also exist. These findings are in line with increased perceptual disappearances in different patterns in spatial attention [21,27,32]. From implications in driving and distraction [21,57], our ability to attend to stimuli [27], possible links to mental disorders such as schizophrenia and ADHD [32], and even explanations for medical conditions such as hemispatial neglect [42]; MIB has proven that it can be empirically and practically reliable in measuring spatial biases. Our findings demonstrate the possible usefulness of MIB illusions as a tool to measure mechanisms and concepts including spatial biases, and atypical perception and attention. Based on our results, MIB perceptual disappearances, and specifically vertical biases, have potential to be good measures for atypical perception, and be possibly reflective of positive symptoms and dorsal stream dysfunction in schizophrenia. Our findings also raise some speculative possibilities. For example, horizontal field biases could be measures for deficits in interpersonal and social traits, as reflected in the interactions between the PAS and the SPQ interpersonal factor. This finding may reflect negative symptoms found in schizophrenia, such as social anxiety, inability to recognize faces and emotions, and difficulty socializing [1].

A possible limitation of our study is that we had a non-clinical sample, using schizophrenia-like traits measures. This may limit our ability to generalize to the schizophrenia population. However, using traits is also an advantage. This is because any of the effects are likely to be smaller, and thus our findings are likely to be stronger in a schizophrenia cohort. Relatedly, in our study we correlated MIB to SPQ scores, with SPQ score being used as a proxy for schizophrenia. The additional value of MIB over SPQ in schizophrenia diagnosis would still need to be thoroughly established in samples with schizophrenia diagnosis. Additionally, future directions should also consider measures including Symptoms of Schizophrenia Inventory [76] and Symptom Self-rating Scale for Schizophrenia [77].

The nature of the online setting of our experiment may too be a limitation, and most likely increases the chances of variability in the data and decreases the chances of finding significant effects. However, several post-pandemic studies have shown little to no differences in data replicability between in-lab and online experiments [78–84], and indeed our findings align well with the current literature. Furthermore, we increased the sample number to help combat the potentially lower-quality online data.