Introduction

An individual’s cognitive style refers to his or her preferred method for acquiring and processing information and this is regarded as relatively stable over time. Cognitive styles can encompass attitudes, preferences or strategies used by individuals that influence functions such as perceiving, remembering, thinking, and problem solving [1] (see [2], for review). We aimed to develop a new questionnaire—the Sussex Cognitive Styles Questionnaire (SCSQ)–to investigate the interrelationships between visual and verbal cognitive styles and to incorporate recent findings on global/local processing biases and systemising. To preface our study, we describe below the different facets of cognitive styles that one might wish to capture in a comprehensive cognitive styles questionnaire. We then describe the design of our own questionnaire and the way we have validated it in different populations of typical and atypical individuals.

One traditional view of visual and verbal cognitive styles places those who prefer visual processing strategies and those who prefer verbal processing strategies as on opposite ends of a visual-verbal continuum [3,4]. That is, visualisers who prefer to use visual mental imagery to process information are placed at one end of the continuum and verbalisers, who prefer verbal techniques, are on the other. The self-report Verbaliser-Visualiser Questionnaire (VVQ; [4]) was originally designed to classify people along this dimension. Recent neuroimaging results support the distinction between visualisers and verbalisers. For example Kraemer, Rosenberg and Thompson-Schill [5], demonstrated that participants who score highly on verbal preference exhibit greater activation in phonological regions when presented with pictorial stimuli, and those who score highly on visual preference exhibit greater activation in visual cortex when presented with written descriptions of visual stimuli (see also [6,7]). However, research has also demonstrated independence between visual and verbal dimensions on self-report measures [8,9], and it is therefore theoretically possible for an individual to score high (or low) on both visualising and verbalising [10].

Recent research has shown visual style can be further subdivided into two relatively independent imagery factors—object imagery and spatial imagery [11,12]. Object imagery refers to the subjective quality of cognitive representations in terms of resemblance to an object’s form (encompassing, for example, size, shape, colour, brightness), whereas spatial imagery refers to the quality of representations in terms of spatial relationships between objects (encompassing, for example, location, movement, spatial transformation). Blajenkova et al. [11] developed the two-factor self-report Object-Spatial-Imagery Questionnaire (OSIQ) to capture dissociations between object and spatial imagery, with items such as “My images are very colourful and bright” loading on to the former factor and items such as “I was very good in 3D geometry as a student” loading on to the latter. Object imagery scores correlated with performance on a degraded pictures task (identifying a picture from partial information) and spatial imagery correlated with a performance on paper folding and mental rotation tasks, thereby demonstrating predictive validity of the measure. Blazhenkova and Kozhevnikov [13] subsequently developed the Object-Spatial Imagery and Verbal Questionnaire (OSIVQ) to include a third factor representing verbal style. This factor predicted performance on a task of generating sentences using a set of four nouns.

Spatial mental imagery, in particular, has been shown to correlate with another cognitive style, namely systemising [14–16]. Systemising refers to the drive to understand rules, regularities or variables that govern how a system works, or the tendency to construct systems to understand the world [17–19]. Systemising can be gauged via the self-report questionnaire called the Systemising Quotient (SQ; [19]). Behavioural studies have shown that systemising is correlated with the ability to mentally rotate 3D shapes [15], and other aspects of mental rotation such as stimulus encoding and same-different comparison processes [14]. Furthermore, scores on the short-form SQ [15] correlate well with the spatial imagery subscale of the OSIVQ (r = .61; [16]). Spatial abilities such as mental rotation, or everyday activities such as map reading or solving the Rubik’s cube, are thought to involve systemising as input features must be transformed via the application of rules in order to arrive at the correct outcome [17]. The present research considers whether systemising and spatial imagery reflect a unitary cognitive style or two separate, but related, processes.

Systemising also presupposes excellent attention to detail [20] and as such bears at least a superficial similarity to a local processing bias. When attending to a complex stimulus, such as a scene, one can focus on it holistically or focus on its constituent parts. These can be thought of as global (holistic) or local (featural or analytic) processing biases respectively. Systemising is presumably dependent on sufficient local processing for relevant task details to be attended to and then acted upon. Indeed, Billington, Baron-Cohen and Bor [21] demonstrated that scores on the Revised Systemising Quotient (SQ-R; [22]) correlate with local bias in the Navon [23] letters task (r = .45).

Global/local biases can influence visual processing more generally. For example, a global bias is positively correlated with performance on judgments of line orientation [24] and inappropriate local processing orientations can disrupt recognition of complex visual stimuli such as faces and cars [25–27]. Additionally, Bouvet, Rousset, Valdois and Donnadieu [28] found correlations between local and global processing across different modalities (e.g., a participant exhibiting a faster response time in identifying global patterns in visual stimuli was likely to exhibit a similar response time advantage for global audio stimuli, r = .55), and they suggest that global/local biases may be indicative of a particular cognitive style (see also [29] for a similar suggestion). Thus there is some support for the notion that global/local biases represent somewhat stable traits. The key difference between local processing and systemising is that systemising requires understanding the operations involved in a task and noting the consequences of actions [18]. A local processing bias does not in and of itself presuppose a drive to understand the relations between elements in a problem, but does suggest a predisposition towards attention to detail. Thus it is not clear whether systemising tendencies and global/local bias would be driven by the same latent factor, or whether they are largely independent. Despite this, we are aware of no published self-report measure which gauges whether a person thinks of themselves as naturally inclined towards local or global processing, which we address here. Furthermore, the relationship between systemising and local processing, on the one hand, and visual-verbal measures of cognitive style, on the other, is essentially unknown.

We sought to design a comprehensive new questionnaire measure—the SCSQ—with items designed to capture self-reported imagery, systemising and global/local bias. We aimed to replicate the finding that visual and verbal styles can be divided, and that certain types of self-reported mental imagery can be further subdivided. We also aimed to extend the scope of such measures by additionally considering systemising and global/local bias. Specifically, we aimed to determine whether systemising and global/local bias items would or would not cluster together (cf [21]), and how these variables would relate to different forms of visual imagery (cf [13]), as it was unclear a priori whether (some) systemising and global/local bias items would cluster with aspects of visual imagery.

Study 1 describes the new SCSQ, and presents factor and reliability analyses as well as the correlations between factor scores. Gender differences are also considered. Following the factor analysis and subscale establishment in Study 1, we aimed to validate the SCSQ in two further studies. Study 2 sought to validate relevant subscales of the SCSQ via correlation with performance on visual and verbal memory tasks in a subset of participants from Study 1.

Study 3 aimed to validate the SCSQ via testing on a group of unusual individuals. We asked whether the presence of certain forms of synaesthesia—which have been linked to particularly vivid imagery (e.g., [10,30])–would predict scores on SCSQ subscales. Individuals with synaesthesia have additional perceptual-like experiences (‘concurrents’) in the presence of certain materials (‘inducers’), such as experiencing colours when reading or listening to music [31], and we might therefore expect the presence of certain forms of synaesthesia to be predictive of visual, verbal, systemising and global/local processing preferences in particular ways compared to non-synaesthetes. We focus on two variants of synaesthesia. Firstly grapheme-colour synaesthesia, where letters and/or numbers result in additional colour experiences (e.g., the letter A might be crimson red; for review see [32]), and secondly sequence-space synaesthesia, where particular sequences (e.g., numbers, months, years) are consciously visualised as having spatial arrays (e.g., the months of the year might be experienced in an ellipse in front of the body, with January, say, to the right; for review see [33]). Following the results of the factor analysis in Study 1, we were able to make predictions about how synaesthetes’ cognitive styles may differ from those of non-synaesthetes, and thereby validate the SCSQ subscales. We address this in Study 3.

Study 1

The aim of Study 1 was to design a new cognitive styles questionnaire (the SCSQ). We then sought to assess its factor structure and reliability, explore the correlations between the factors, and investigate gender differences on the obtained factors.

Results and Discussion

Of the 1542 participants who completed the measure, 87 (5.6%) reported they thought they may experience some form of synaesthesia and these participants were excluded from further consideration. The analyses were conducted on the remaining 1455 participants (907 females, 548 males). Age data were available for 1437 of these participants (range 16–90 years; M = 27; SD = 13). Data are included in S1 File.

Exploratory factor analysis.

Exploratory factor analysis (EFA) with weighted least squares estimation was conducted on all 84 items to determine the model that best described the data. Spearman correlations were used in the analysis due to the ordinal nature of the items. Initially, we determined the number of factors to extract using Velicer’s [36] Minimum Average Partial (MAP) method. This suggested a nine factor solution at this stage. Thus an EFA was specified with a nine factor solution, with an oblique rotation (Oblimin-Quartimin) to allow for correlations between factors. Oblimin rotations parametrise the obliqueness of the rotated factors. Quartimin rotations are a special case, where the solution is most oblique by minimising the product of the squared loadings on each factor. Items with factor loadings > .30 were considered worthy of retention. Seventeen items did not load at .30 at this stage and were removed from further analysis (Q03, Q12, Q13, Q14, Q34, Q35, Q37, Q39, Q49, Q50, Q68, Q70, Q74, Q75, Q76, Q79, Q83). Five further items were removed for cross-loading on more than one factor (Q26, Q54, Q63, Q66, Q82).

A second MAP analysis was conducted, suggesting a six factor solution to the remaining 62 items. This resulted in a clear and interpretable factor structure. Regarding this second EFA, the Kaiser-Meyer-Olkin measure of sampling adequacy suggested a factorable solution (KMO = .87). At this stage, two remaining items did not load at the .30 level on to the obtained factors and were removed from further consideration (Q30, Q80). Table 1 presents the results of the EFA. The reduced version of the questionnaire contained sixty items, which explained 32% of the total variance. See Tables A-F in S2 Table for the inter-item correlation matrices for the six obtained factors.

Download:

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

Table 1. Rotated factor matrix of factor loadings for the final sixty items of the SCSQ.

https://doi.org/10.1371/journal.pone.0155483.t001

Factor 1 contained items mostly related to the use of visual imagery in everyday life (e.g., “I often use mental images or pictures to help me remember things” from the IDQ), with some items which gauge the subjective quality of visual imagery (e.g., “My mental images are very vivid and photographic” from the OSIVQ), hence this factor was termed ‘Imagery Ability’. Factor 2 contained items relating to spatial mental imagery (e.g., “I can easily imagine and mentally rotate three-dimensional geometric figures” from the OSIVQ), but also items relating to mathematics abilities (e.g., “I am fascinated by numbers” from the AQ) and items relating to technical interests (e.g., “If I were buying a computer, I would want to know exact details about its hard drive capacity and processor speed” from the SQ), thus this factor was termed ‘Technical/Spatial’. All items on Factor 3 related to aspects of language. In particular most of these items indicate an interest in the visual appearance of written language as opposed to oral verbal ability (e.g., “When I hear a new word, I am curious to know how it is spelled”, a new item; “When I read something, I always notice whether it is grammatically correct”, from the SQ). These items clearly formed their own factor separate from visual imagery. Therefore this factor was termed ‘Language & Word Forms’.

Factor 4 contained items relating to organisation (e.g., “If I had a collection (e.g., CDs, coins, stamps), it would be highly organised” from the SQ) and order (e.g., “Order is important to me”, a new item) and was termed ‘Need for Organisation’. Negatively loaded items on Factor 5 indicate attention to detail or a local processing preference (e.g., “I tend to focus on details in a scene rather than the whole picture”, a new item), whereas positively loaded items indicate a more holistic preference (e.g., “I usually concentrate on the whole picture, rather than the small details” from the AQ), therefore this factor was termed ‘Global Bias’. Finally, items on Factor 6 refer to the drive to categorise (e.g., “When I look at an animal, I like to know the precise species it belongs to” from the SQ), or interest in systems (e.g., “I am fascinated by dates” from the AQ) and was termed ‘Systemising Tendency’. One item (“When I look at a tree I focus on its features such as branches and leaves rather than the whole”) cross-loaded onto two factors: Factors 5 and 6. This item better captures attention to detail or local bias and was therefore included on Global bias (Factor 5). Interestingly, as both the globally and locally phrased items loaded on to the same factor as opposed to two different factors, it seems that these preferences are at opposite ends of a spectrum in terms of self-report.

Reliability analysis.

Next, we assessed the reliability of the obtained factors via ordinal α (see Table 1). All factors exhibited good levels of reliability (all α ≥ .73). Next we consider improvements to α with the removal of items. The reliability of Imagery Ability would not improve with the removal of any item. The reliability of Technical/Spatial would increase from .885 to .887 with the removal of Q48. The reliability of Language & Word Forms would increase from .798 to .804 with the removal of Q1. The reliability of Need for Organisation would increase from .770 to .776 with the removal of Q38. The reliability of Global Bias would increase from .736 to .746 with the removal of Q27. The reliability of Systemising Tendency would not benefit from the removal of any item. As the removal of any item would only result in minor improvement in ordinal α (≤ .01) all items were deemed worthy of retention.

Next we calculated the mean scores for each factor (negatively loading items were reverse coded prior to calculation; see Table 1 for mean factor scores). Table 2 displays the correlation coefficients between the six obtained factors. Although most factors tended to correlate the effect sizes were small (rs < .30) in all but one case (based on the categorisation of effect sizes by Cohen [37]). Global Bias and Systemising Tendency correlated negatively, and one might expect sufficient local processing to systemise effectively. Systemising Tendency correlated positively with Technical/Spatial, showing that spatial skills and systemising are related. Technical/Spatial also correlated negatively with Global Bias. It is not surprising that Systemising Tendency and Technical/Spatial were correlated; one might expect that sufficient local detail is needed for spatial transformation and attention to detail in systemising domains. Imagery Ability also correlated negatively with Global Bias.

Download:

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

Table 2. Correlation coefficients (Pearson’s r) between scores on the scales of the SCSQ.

https://doi.org/10.1371/journal.pone.0155483.t002

Gender differences.

Next we assessed gender differences on SCSQ factor scores. Mean scores for females and males were calculated and compared via independent samples t-tests (Table 3).

Download:

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

Table 3. Mean factor scores as a function of gender.

https://doi.org/10.1371/journal.pone.0155483.t003

Consistent with Blazhenkova and Kozhevnikov [13], females scored higher on Imagery Ability (similar to their object imagery factor) and males scored higher on Technical/Spatial (similar to their spatial factor, with additional mathematics and technical interest items). Similar results for imagery according to gender are also presented by Campos [38]. Males also scored higher than females on Systemising Tendency, consistent with Baron-Cohen et al. [19], Ling et al. [15] and Wheelwright et al [22].

Study 2

As the SCSQ used items pertaining to the everyday use of dimensions on the measure, the goal of Study 2 was to validate the visual and verbal subscales of the new measure by correlating mean factor scores with performance on modified versions of tasks taken from a standard memory battery, the Wechsler Memory Scale Revised (WMS-R; [39]). Specifically, we adapted the visual paired associates and verbal paired associates tasks. Both tasks included additional trials to prevent ceiling effects in the current student sample (the original versions of these tasks are used in clinical samples), and both tasks were computerised using E-Prime 2 software. We also sought to validate the measure with an additional visual recognition task using fractal images as stimuli. The fractal image recognition task has previously been used by Ward, Hovard, Jones and Rothen [40]. The task was chosen here because the visual stimuli are hard to verbalise and performance on the task tends to be overall lower and more variable than those used in the WMS-R (which was designed to measure memory impairment).

Results and Discussion

Data are included in S2 File. Firstly, we consider performance on the memory tasks before investigating the relationships between memory task performance and scores on the SCSQ subscales (note that we do not consider gender in the following analyses due to the small number of males in the sample). The proportions of correct responses for the verbal and visual paired associates tasks were calculated (see Table 4 for descriptive statistics). Considering performance on the visual paired associates task, there was a significant linear trend over the learning blocks, F(1, 43) = 66.78, p < .001, η²_p = .61, indicating performance improved over the blocks, and performance in each block differed significantly from performance in every other block, ps ≤ .001. Considering performance on the verbal paired associates task, there was a significant linear trend over the learning blocks, F(1, 43) = 80.15, p < .001, η²_p = .65, indicating performance improved over the blocks, and performance in each block differed significantly from performance in every other block, ps ≤ .002. Thus we can be confident that learning occurred during both the visual and verbal paired associates tasks.

Download:

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

Table 4. Mean proportion of correct responses over blocks for the visual and verbal paired associates tasks.

https://doi.org/10.1371/journal.pone.0155483.t004

Performance in the fractal recognition task was measured via the signal detection measure, d’. This statistic represents the difference between standardised hit rate (correctly endorsing test stimuli present during learning) and false alarm rate (erroneously endorsing test stimuli not present during learning). Above chance recognition is shown if d’ is greater than zero. Mean d’ was 0.92 (SE = .09), which was significantly greater than zero, t(43) = 10.33, p < .001. Thus sufficient discrimination between targets and lures was demonstrated.

We then correlated task performance, averaging across the initial learning blocks in the paired associates tasks, with scores on the questionnaire subscales (see Table 5). Technical/Spatial scores correlated significantly with average visual associates performance, visual associates delayed performance, and fractal recognition d’. Language & Word Forms correlated significantly with average verbal associates performance and verbal associates delayed performance. Thus these factors demonstrated convergent validity. Perhaps surprisingly, Imagery Ability did not correlate significantly with performance on these tasks. In the visual associates task, participants were required to associate abstract line drawings with colours and in the fractal recognition task were required to recognise abstract images, thus performance on these tasks may be plausibly related to more abstract representational styles (captured by Technical/Spatial scores), rather than concrete representational style (captured by Imagery Ability scores). Convergent validity of the Imagery Ability factor could potentially be demonstrated using more naturalistic stimuli (such as natural scenes). Global Bias correlated negatively with performance in the learning blocks of the Visual Paired Associates task, indicating better performance was related to local bias. However, this factor did not correlate with performance in the delayed block. This could indicate that local bias confers a benefit to memory encoding when associating complex, abstract shapes with other visual materials. It may also reflect a Type I error (given the number of analyses performed we would predict 1.5 significant results by chance). Although most results would not survive a Bonferroni correction for multiple comparisons, the significant results represent medium effect sizes [37], generally form a stable pattern across measures, and are in line with theory.

Download:

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

Table 5. Correlation coefficients (Spearman’s rho) between SCSQ factor scores and memory task performance.

https://doi.org/10.1371/journal.pone.0155483.t005

Study 3

In Study 3 we sought to validate the SCSQ subscales identified in Study 1 by issuing it to a sample of participants with atypical experiences, namely grapheme-colour and sequence-space synaesthetes. It has been suggested that synaesthetes’ cognitive preferences may reflect affinities for materials associated with their synaesthesia, either in the domain that induces the experience, or the domain of the experience itself [10]. Thus, in the case of grapheme-colour synaesthesia we would expect a higher rating than that of non-synaesthetes on verbal and visual styles (here, Language & Word Forms and Imagery Ability respectively) because the inducer is verbal and the concurrent is visual.

Although the bulk of research on imagery, cognitive styles and individual differences in synaesthesia has focused on grapheme-colour variants (e.g., [10,31,41]), recent work has begun to address imagery in sequence-space synaesthesia. For example, since sequence-space synaesthetes appear to be superior in tasks of mental rotation (e.g., [42,43]; but see [44]) we might predict they would score highly in our questionnaire on spatial imagery. Interestingly, these synaesthetes had not scored highly on previous questionnaires of spatial imagery [45], which conflicts with their performance in the equivalent behavioural task (e.g., rotation; [42]). It may therefore be possible that previous self-report questionnaires were failing to detect spatial skills in this group. This may be because spatial imagery forms part of constellation of abilities, as shown by the Technical/Spatial factor, and only by taking into account related abilities would we observe the posited advantage. Additionally, as spatial abilities and systemising are correlated [15], we sought to determine whether sequence-space synaesthetes in particular would also score highly on Systemising Tendency. We therefore look with interest at how sequence-space synaesthetes might perform in our own questionnaire, particularly in relation to spatial imagery and systemising. Such results would help verify these subscales of the SCSQ.

Results and Discussion

Data are included in S3 File. The results are summarised in Table 6. The three groups of synaesthete are defined by the presence of either or both types of synaesthesia and the controls are defined by the absence of both.

Download:

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

Table 6. Mean factor scores as a function of presence/absence of grapheme-colour and sequence-space synaesthesia.

https://doi.org/10.1371/journal.pone.0155483.t006

We conducted separate multiple regression analyses to determine whether the presence of different types of synaesthesia predicted scores on each of the SCSQ subscales. It is important to note that individuals with synaesthesia often have multiple forms (e.g., [49]). By specifically controlling for this in the regression analyses, we can address which forms of synaesthesia contribute most to different forms of cognitive styles. The presence/absence of grapheme-colour and sequence-space synaesthesia were coded as 0 (absent) and 1 (present) and entered into regression models as binary predictors for each SCSQ subscale. Importantly, Study 1 revealed gender differences on the Imagery Ability, Technical/Spatial, and Systemising Tendency subscales. Therefore we entered gender as a third predictor into each regression to assess whether effects of synaesthesia were detectible when accounting for gender differences. Females were coded as 1 and males as 0.

Table 7 shows the results of the regression analyses. Both grapheme-colour synaesthesia and sequence-space were linked to increased Imagery Ability and increased Language and Word forms scores. Sequence-space synaesthesia, but not grapheme-colour synaesthesia, was linked to increased Technical/Spatial, increased Systemising Tendency and reduced Global Bias (i.e. a greater local bias) scores. The differences observed between grapheme-colour and sequence-space synaesthetes on SCSQ scales shows that different forms of synaesthesia may predict different aspects of cognition, and should be accounted for in future research.

Download:

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

Table 7. Regression weights (unstandardised b, standard error, and standardised β) for each subscale of the SCSQ, as a function of the presence/absence of grapheme-colour synaesthesia, sequence-space synaesthesia, and gender.

https://doi.org/10.1371/journal.pone.0155483.t007

The relationship between gender and SCSQ scores reflects that observed in Study 1, even when accounting for synaesthesia. Females scored higher on Imagery Ability, and males scored higher on both Technical/Spatial, and Systemising Tendency. Thus gender and the presence of different forms of synaesthesia make independent contributions to SCSQ scores.

General Discussion

The current research aimed to produce a new cognitive styles questionnaire—the SCSQ—that links previously unconnected domains: the verbaliser-visualiser tradition alongside global/local processing and systemising. To achieve this, we issued the SCSQ to a large number of participants and assessed its factor structure and reliability. We found six factors: Imagery Ability, Technical/Spatial, Language & Word Forms, Need for Organisation, Global Bias, and Systemising Tendency which all exhibited good levels of internal consistency. We also found generally low to modest, but statistically significant, correlations between almost all of the factors.

We replicated the finding that visual and verbal styles are separable [8,9] and that visual imagery is not a unitary construct, but can be subdivided further [11,13]. Here we found Imagery Ability and Technical/Spatial factors, when new items were included in the measure. We also found that global/local bias items formed their own factor, as did systemising items. Notably, each factor contained items from more than one source (i.e. the factor analysis did not simply cluster items according to their original measure). Furthermore, all factors (with the exception of Need for Organisation) demonstrated construct validity. Firstly Study 1 revealed gender differences on SCSQ subscales. Secondly SCSQ scores predicted performance on visual and verbal memory tasks in Study 2, in-line with previous findings that these cognitive styles predict objective task performance [5,6,7]. Thirdly, SCSQ scores differed according to the presence and absence of synaesthesia in Study 3.

The Imagery Ability factor is conceptually similar to the object visual factor from the OSIVQ containing a number of items related to detail in object imagery. Indeed, this factor contained six items from this subscale (and one negatively loading item from the spatial subscale of the OSIVQ). The Imagery Ability factor also contained items from a number of other sources—four items from the IDQ and six new items designed to capture the use of imagery in everyday life (e.g., “When I can’t find something I’m looking for, I automatically visualise the last place I saw it”). Thus it may be the case that self-rated object imagery is tightly linked with its use in everyday life.

Our Technical/Spatial factor focuses on spatial imagery and other activities based on abstract cognitive skills such as mathematical abilities and technical interests. This subscale contained seven items from the spatial subscale of the OSIVQ, but also contained six items from the SQ and two from the AQ, thus this latent factor captures more than spatial mental imagery. That mathematical items clustered with spatial imagery items perhaps is not surprising; mathematical problem solving has previously been shown to correlate positively with spatial imagery and negatively with pictorial imagery (e.g., [50]; see [51], for review) and latent factors driving performance on spatial imagery tasks could covary with mathematical ability (see [51] for the neuro-anatomical basis of numerical and spatial processes). Interestingly, previous factor analyses of the SQ show that items relating to interest in technology form their own component of systemising [15,52]. That items relating to an interest or concern with technical details also loaded on the Technical/Spatial factor also indicates a more general capacity for dealing with abstract information.

In terms of scores for our imagery factors, females rated themselves higher on Imagery Ability than males, and males rated themselves higher on Technical/Spatial than females. The findings with respect to gender and different forms of imagery are in good agreement with other studies [11,13,38], and thus validate the imagery subscales of the SCSQ.

Both the presence of grapheme-colour and sequence-space synaesthesia significantly predicted scores on the Imagery Ability factor. This supports previous research showing these groups rate themselves highly on object imagery (e.g., [10,42,43]). Additionally, sequence-space synaesthesia predicted scores on the Technical/Spatial factor where previous research failed to find an advantage for this group in terms of self-report spatial imagery [44,45], despite others’ sometimes showing greater objective spatial imagery skills (e.g., [42]). Our finding of superior scores in the Technical/Spatial factor is more in-line with most behavioural findings in this domain. Additionally, this may reflect the factor’s capture of abilities correlated with spatial processing, such as mathematics, on which synaesthetes may also differ. For example, 77% of a sample of sequence-space synaesthetes reported consciously using their number forms in calculations [53].

Language & Word Forms contained one item (pertaining to language) from the SQ and five new items. Most of the items on this subscale relate to the appearance of words and sentences (such as spelling and grammar structure), thus this factor emphasises orthographic qualities. That all language items formed their own factor supports the tripartite model of visual imagery and verbal skills (e.g., [13]), even though most of the verbal items used in the SCSQ we used were novel and de-emphasised oral aspects of verbal skills. In addition to Imagery Ability scores, the presence of grapheme-colour synaesthesia predicted scores on Language & Word Forms. This supports the hypothesis that synaesthetes’ cognitive styles relate to inducing- or concurrent-related information, i.e. verbal and visual information in the case of grapheme-colour synaesthesia, replicating findings by Meier and Rothen [10] who issued the VVQ to grapheme-colour synaesthetes. That the SCSQ performed similarly to the VVQ offers further validation. Interestingly, the presence of sequence-space synaesthesia also predicted scores on Language & Word Forms. Notably, letters of the alphabet are often cited as inducing spatial forms in sequence-space synaesthetes. This also fits well with Meier and Rothen’s [10] suggestion that synaesthetes’ cognitive styles reflect a particular affinity for their inducing materials.

Need for Organisation emerged as a factor, which contained one item from the SQ and five new items, and was not predicted a priori. Neither gender, nor the presence of either form of considered synaesthesia predicted scores on this factor. Future validation work could explore this factor’s relationship to orderliness factors on other questionnaire measures (e.g. Comrey Personality Scales, [54]).

Global bias contained three items from the AQ, four new items and one item from the OSIVQ object factor (which focused on attention to specific details). As globally oriented items loaded positively on this factor and locally oriented items negatively, it appears that self-reported global/local bias can be conceptualised as opposite ends of a single dimension. Systemising Tendency contained four items from the SQ, one item from the AQ and one new item. All items related to systemising, and in particular precision in categorisation (of dates, animals, plants etc.).

Although Systemising Tendency and Global Bias emerged as separate factors, there was a modest negative correlation between them which indicates that those who exhibit higher systemising are also likely to exhibit a local bias (cf. [21]). The correlation between Global Bias and Systemising Tendency supported the notion that these traits are related, but not identical, presumably as one needs to pay sufficient attention to detail to systemise effectively [17]. Males also scored more highly than females on Systemising Tendency, replicating work by others [19,15,22].

Bouvet et al. [28] and Dale and Arnell [29] suggest there may be a cognitive style that underpins performance on global/local tasks and this is the first attempt to create a self-report subscale that attempts to gauge this style. Future work should aim to further validate the Global Bias scale by assessing its relationship with objective tests of local and global processing, such as the Navon [23] task. It could be the case that scores on this subscale predict performance on this task, or predict the extent to which individuals are affected by inappropriate processing orientations in a more direct manner than the SQ-R [52]. This subscale may also provide a useful proxy for global/local orientations when laboratory-based testing is impractical (e.g. online data collection).

In terms of the relationship between processing orientation and imagery, we observed that increased systemising and local processing were associated with increased Imagery Ability and Technical/Spatial scores. Kozhevnikov et al. [12] suggest that spatial imagers encode and process images in an analytic manner to arrange and analyse components. Additionally, Kozhevnikov, Hegarty and Mayer [55] show that, although object visualisers tend to encode images as a single unit (global processing), they can also apply sequential, local reasoning in the verbal domains. This suggests that they may be adept at switching between processing orientations but self-report a local bias. The only non-significant correlation was between Technical/Spatial and Language & Word Forms.

The presence of sequence-space synaesthesia, but not grapheme-colour synaesthesia was associated with a lower Global Bias score (i.e. a local bias) and a higher Systemising Tendency score. Synaesthetic experiences are by their nature systematic (e.g., consistency between inducer and concurrent or application of rules to the spatial form to arrive at some outcome), and a proclivity to understand or construct systems is a hallmark of systemising. Being able to locally inspect, manipulate and analyse forms may well depend on local processing. For instance, Simner et al. [36] (also [56]) showed that synaesthetes with time-space forms have superior recall of dates and events in time and that most reported that target events were retrieved from their spatial forms. (See also [57], for a detailed phenomenological report of sequence-space synaesthesia.). Additionally, sequence-space synaesthesia is often linked to, amongst others, calendar and numerical systems, that themselves are often subject to systemising [20].

Finally, past research into cognitive styles has had trouble connecting theory with measurement [2,13]. We consolidated research from seemingly disparate areas (object, spatial and verbal processing, systemising, global/local bias, and synaesthesia) into a novel cognitive style measure that can be applied to many domains of research (e.g. in typical and atypical populations). In Study 3 we predicted how the subscales of the SCSQ would behave with respect to a special population of synaesthetes, and these predictions were largely borne out through validation with external measures.

Supporting Information