Letter-colour consistency but not conscious awareness of synesthetic colours is
necessary when defining grapheme-colour synesthesia
Jip Gudden
under supervision of dr. R. Rouw and dr. N. Root July 2020
Institute for Interdisciplinary Studies, University of Amsterdam, Science Park 904, Amsterdam, The Netherlands
Abstract
Grapheme-colour synesthesia (GCS) is a condition where a specific grapheme (letter or number) leads to a conscious experience of a specific colour. However, studies find contradictory results on what characterizes GCS and what distinguishes grapheme colour (GC) synesthetes from non-synesthetes. Inevitably, a not yet clearly defined definition of GCS makes it difficult to correctly distinguish between different types of GC synesthetes and between GC synesthetes and non-synesthetes. Namely, to be classified as a GC synesthete, one also needs to consistently experience the same specific colour for a grapheme and be consciously aware of having synesthetic experiences. We are interested to know whether grapheme-colour consistency is a necessary trait of GCS and if someone explicitly needs to state to experience colours with letters. Or can we see behavioural differences when only one of those two criteria is met? To answer this question, we compared self-claimed synesthetes (who explicitly state to have GCS but lack consistent letter-colour associations) and know-associators (who deny to experience GCS but have consistent letter-colour associations) with definite synesthetes (meet both criteria) and non-synesthetes (fail to meet both criteria). Specifically, we compared these four groups on tasks in which GC synesthetes have shown to perform differently than non-synesthetes in previous studies, namely a synesthetic Stroop task, a colour memory task and a colour perception task. We found that know-associators performed similarly as definite synesthetes on all tasks, whereas self-claimed synesthetes performed similarly as non-synesthetes on all tasks. Furthermore, greater letter-colour consistency was correlated with synesthetic-like performance in every task, whereas greater conscious awareness of having GCS was only correlated with synesthetic-like behaviour in tasks where perception plays an important role. In sum, we conclude that GC consistency is a necessary trait of GCS. We also raise the question whether synesthetic experiences are responsible for the behavioural differences, or solely the presence of strong letter-colour associations.
© University of Amsterdam, All rights reserved.
Keywords: Grapheme-colour synesthesia, grapheme-colour consistency, conscious awareness, know-associators
Introduction
Multisensory integration refers to the formation of a rich, integrated and coherent percept when information from different sensory modalities are combined (Stein & Stanford, 2008). Multisensory integration is something that affects all people, as can be seen in cross-modal illusions like the McGurk effect where conflicting sensory information changes perception (McGurk & MacDonald, 1974; O'Callaghan, 2008). But there are also forms of multisensory integration that only affect a small subset of the population, named synesthesia. Synesthesia is a condition where an experience in one sensory modality leads to a conscious second experience in another sensory modality. This second sensory
experience is evoked exclusively internally and is a remarkable way of perceiving the world, very different from what most people experience (Ward, 2013). Therefore, synesthesia can give us an extraordinary insight into cognition and perception (Chiou & Rich, 2014). Additionally, the experience of synesthesia is an altered conscious experience of similar external stimuli (van Leeuwen, Singer, & Nikolić, 2015) and therefore it could possibly give insights into the “hard” problem of consciousness; explaining the neural correlates of “what something feels like” (Chalmers, 1995). Although the neural basis of synesthesia is not yet fully understood, general consensus is that the neurological mechanisms that give rise to synesthesia are reflected by some type of hyper-association between brain regions (Hupé & Dojat, 2015), be it functional or structural in nature (Bargary & Mitchell, 2008). The most studied
form of synesthesia is grapheme-colour synesthesia (GCS), a form of synesthesia where achromatic letters or numbers evoke the sensory experience of colour (Simner & Hubbard, 2013), with a prevalence of around 1 percent (Carmichael et al., 2015; Watson et al., 2017). To gain an understanding of what GCS is, neuronal differences (for a review see Rouw, Scholte, & Colizoli, 2011) and cognitive differences (for a review see Rothen, Meier, & Ward, 2012) between grapheme-colour (GC) synesthetes and non-synesthetes have been reported. Behaviourally, a synesthetic version of the Stroop task has been shown to reliably detect differences between GC synesthetes and non-synesthetes (Dixon et al., 2000; Ward, 2013). The synesthetic Stroop task is an adapted version of the Stroop task in which letters are printed in the same colour as elicited by the synesthetic experience (congruent) or in a different colour (incongruent). When GC synesthetes are asked to name the printed colour of a letter, they take longer to do so for incongruent coloured letters than for congruent coloured letters, leading to an interference effect (Mattingley et al., 2001). The longer reaction time can be explained by the automatically evoked synesthetic colour that interferes with the printed colour (Wollen & Ruggiero, 1983). Interestingly, non-synesthetes can experience similar interference effects as some GC synesthetes when they are trained to form grapheme-colour associations (Colizoli, Murre, & Rouw, 2014; Meier & Rothen, 2009). This implies that semantic associations are sufficient to create an interference effect in the synesthetic Stroop task, even for non-synesthetes. The question therefore arises whether differences between GC synesthetes and non-synesthetes on the synesthetic Stroop task reflect synesthetic experiences or solely the presence of strong grapheme-colour associations.
To answer this question, Nikolić, Lichti, and Singer (2007) found that GC synesthetes experience a colour opponency effect; namely they experience more interference when the incongruent colour of a letter is opponent in colour space to the synesthetic colour (a “synesthetic red” A printed in green) than when a letter is printed in arbitrary or independent colours (a “synesthetic red”
A printed in blue). In contrast, non-synesthetes who were trained to form grapheme-colour associations do not show a colour opponency effect (Nikolić et al., 2007) and do not report to perceptually experience colours with letters (Colizoli et al., 2014). Thus the existence of a colour opponency effect proves that the presence of synesthetic experiences in the synesthetic Stroop task leads to different behaviour for GC synesthetes and non-synesthetes. Moreover, the study of Nikolić et al. (2007) showed that synesthetic colour perception in the synesthetic Stroop task closely resembles real colour perception. Other studies also found examples where synesthetic colours act as real colours, such as visual search (Ramachandran & Hubbard, 2001; Palmeri et al., 2002), apparent motion paradigms (Kim, Blake, & Palmeri, 2003; Ramachandran & Azoulai, 2006) and adaptation effects (Blake,
Palmeri, Marois, & Kim, 2005). However, the perceptual nature of GCS cannot fully explain the difference between GC synesthetes and non-synesthetes as studies found that synesthetic experiences can also be induced without perceptual input. For instance, two studies (Dixon, Smilek, Cudahy, & Merikle, 2000; Jansari, Spiller, & Redfern, 2006) found that a sum (e.g. 5 + 2) is sufficient enough to trigger the synesthetic colour associated with the number 7. Moreover, synesthetic colour matching is evenly variable when colours are retrieved from memory and when they are visually observed (Arnold, Wegener, Brown, & Mattingley, 2012) and synesthetic associations could even be formed by childhood memories (Witthoft & Winawer, 2013). To sum up, studies find different and contradictory results on the perceptual or cognitive nature of GCS and thus what distinguishes GC synesthetes from non-synesthetes. A possible explanation for this ongoing discussion is that there is not yet a clear understanding of what GCS is and who should be considered as a GC synesthete - and just as important - who should not (Ward, 2013).
Heterogeneity in the synesthetic population
The contradictory results might be caused by a heterogeneity in the synesthetic population which – when ignored - leads to studies measuring different groups (Simner, 2012). Dixon, Smilek and Merikle (2004) proposed to differentiate GC synesthetes into two subgroups; projectors that project colours onto the inducing graphemes and associators who see colours internally or in the “mind’s eye”. While performing a synesthetic Stroop task, subjects in the Dixon et al. (2004) study were given two conditions; they were asked to name the printed colour of a grapheme or they were asked to name the evoked synesthetic colour of a grapheme. Dixon and colleagues (2004) found that associators were faster in naming the printed colour of a grapheme relative to naming the evoked synesthetic colour whereas projectors showed the reverse profile. This finding has been replicated by Ward, Li, Salih, and Sagiv (2007) but even to this day, the dichotomy between projectors and associators evokes resistance as there are studies that failed to find differences (Sagiv, Heer, & Robertson, 2006; Ward, Jonas, Dienes, & Seth, 2010; Nijboer, Satris, & van der Stigchel, 2011) or report GC synesthetes experiencing colours both projected and in the “mind’s eye” (Edquist, Rich, Brinkman, & Mattingley 2006). However, these studies have very limited sample sizes (Edquist et al., 2006; Sagiv et al., 2006; Nijboer et al., 2011) or primarily focus on visual search (Edquist et al., 2006; Sagiv et al., 2006; Ward et al., 2010) which could be influenced by attention rather than synesthetic experiences (Mattingley et al., 2001). When the examined synesthetic sample differs per study, this automatically alters the outcomes of studies that try to characterize GCS. But a not yet clearly defined definition of GCS entails another problem, namely that it might be difficult to correctly distinguish between GC
synesthetes and non-synesthetes (i.e. a GC synesthete could be regarded as non-synesthetic and vice versa). An incorrect differentiation between GC synesthetes and non-synesthetes inevitably leads to studies examining different synesthetic samples and therefore finding different results.
Questionnaires and consistency
By definition, GCS is a condition in which graphemes automatically and consistently induce the conscious and perceptual experience of colours (Mattingley, 2009; Rothen & Meier, 2014; Colizoli et al., 2014). In other words, GC synesthetes have to subjectively report that graphemes automatically evoke a certain colour. This raises the question whether we should make the distinction on the basis of subjective reports. Questionnaires are able to distinguish GC synesthetes from non-synesthetes (Eagleman et al., 2007) and are able to distinguish between different types of GC synesthetes (Rouw & Scholte, 2007; Rothen, Seth, Witzel, & Ward, 2013). However questionnaires pose the problem that distinction is solely based on one’s own subjective idea of having synesthetic experiences or not, which can vary per person and therefore making the definition of GCS vary. This means in its extremity that malingers or people who misunderstood the questions could be included as well. So this led to a need for an objective measurement to distinguish GC synesthetes from non-synesthetes.
GC synesthetes tend to experience a consistent synesthetic colour with a specific grapheme, so grapheme-colour consistency proved to be a reliable way of distinguishing GC synesthetes from non-synesthetes. In a typical consistency test, subjects are examined in a single session where they are asked to provide grapheme-colour associations for every letter three times after which their consistency of the three measurements can be calculated (Rothen et al., 2013). Alternatively, subjects are asked to provide the colours that they associate with certain graphemes and then they are retested after a certain period of time (Ward & Simner, 2003). GC Synesthetes generally outperform non-synesthetes as non-non-synesthetes entirely depend on memory strategies and not on perception (Simner, 2012). Consistency tests are therefore seen as the “gold standard” to reliably distinguish GC synesthetes from non-synesthetes (Cytowic, 1997; Eagleman et al., 2007). But from a theoretical point of view, the necessity of having consistent associations to explain a perceptual phenomenon is remarkable as “normal” perception is also susceptible to change over time, as top-down influences like expectations consistently shape and change perception (de Lange, Heilbron, & Kok, 2018). This raises the question why synesthetic experiences have to be consistent over time. Especially because it is known that cognition plays a role in GCS (Simner, 2012), although to what degree and when is under debate (Weiss, Greenlee, & Volberg, 2018). GCS is defined as a condition with perceptual qualities and when
consistency is the only objectively measurable way to distinguish GC synesthetes from non-synesthetes, this constrains the definition of GCS (Lynall & Blakemore, 2013).
Circular definition
Simner (2012) raised the problem of circularity when GC synesthetes are solely distinguished based on their consistency. Namely, when GC synesthetes are selected based on having high consistency, literature will self-bias the belief that consistency is a necessity for GCS because every GC synesthete fits this profile. In other words, every examined GC synesthete has consistent letter-colour associations because only GC synesthetes that do are selected (Eagleman, 2012). Large sample studies report that even if you remove malingers and subjects who misunderstood the nature of the condition, self-claimed GC synesthetes who fail to have consistent associations will remain (Simner et al., 2006). This implies that individuals who claim to have GCS but who lack consistent grapheme-colour associations are currently being disregarded as GC synesthetes, which has a major impact on all questions relating to GCS because some GC synesthetes are constantly being left out of the synesthetic samples. However this also means that - when selection is only based on consistency - individuals who deny having synesthetic experiences but have strong and robust associations could be seen as GC synesthetes.
There is a subgroup of GC synesthetes, called know-associators (Ward et al., 2007) or implicit sensory know-associators (Lynall & Blakemore, 2013), who consistently and confidently know that a colour belongs to a grapheme without actually seeing colours. When subjects are asked to associate colours with letters, they show remarkably similar trends across different languages and cultures, which holds true for both GC synesthetes (Root et al., 2018) and even for non-synesthetes (Rouw, Case, Gosavi, & Ramachandran, 2014). Therefore it would be realistic to expect that at least some people will show consistent associations in the absence of synesthetic experiences. A study by Steven (2004) provided evidence for this when she asked 13 self-reported GC synesthetes and 29 self-reported non-synesthetes to give the first colour that comes to mind when a grapheme was given. All GC synesthetes were consistent in their colour associations, but interestingly an apparently distinct group of eight non-synesthetes were also surprisingly consistent, showing “synesthetic-like” grapheme-colour associations. This raises the question whether know-associators, who know the colour of a grapheme but explicitly state that they do not see colours or have synesthetic experiences, should be considered as GC synesthetes (Lynall & Blakemore, 2013). From the eight non-synesthetes of Steven’s study, five were initially unaware of their colour-associations. This calls into question whether they have now turned into GC synesthetes because they have become aware of their letter-colour associations (i.e. show high consistency and subjectively report to
be synesthetic on questionnaires). Or does this mean that they were already GC synesthetes and that the conscious awareness of letter-colour associations is not a necessary trait of GCS.
The present study
Put into different words, is grapheme-colour consistency a necessary trait of GCS and does someone needs to explicitly state to experience colours with letters? Or can we see behavioural differences when only one of those two criteria is met? To answer this question, we will classify subjects in this study on the basis of their letter-colour consistency and on their subjective idea of having GCS and if so, whether this feels like seeing or knowing (see Table 1). A subject’s consistency will be assessed in a single session through a colour picker (adapted from Eagleman et al., 2007). The experiences choosing colours for letter will be assessed in a colour picker questionnaire (similar to Simner et al., 2006; identical to Root, unpublished) in which we ask about the subjective feeling of seeing or knowing colours with letters. Furthermore, we will give subjects a definition of GCS and we will ask them if they have similar experiences whereby letters always evoke a certain colour (based on the synesthesia battery; Eagleman et al., 2007). By doing so, we are able to compare four groups; self-claimed synesthetes, know-associators, definite synesthetes and non-synesthetes. Self-claimed synesthetes explicitly state that to have GCS but lack consistent letter-colour associations whereas know-associators show the reverse profile and show strong letter-colour consistency in the absence of experienced GCS. Definite synesthetes meet both criteria whereas non-synesthetes meet neither of the two. In addition, subjects who deny having GCS, but who do have strong letter-colour consistency will also be counted as know-associators. From a theoretical perspective, the conscious awareness of their consistent letter-colour associations would turn these subjects into GC synesthetes. But the absence of this awareness would not change behavioural performance, hence we believe that these subjects would perform similarly as the previously defined know-associators and therefore should be classified similarly.
We will test these four groups on several tasks that measure colour automaticity, colour perception and colour memory. We specifically chose tasks that measure these cognitive and perceptual skills because previous studies found that GC synesthetes show better performance on these tasks than
non-synesthetes (Dixon et al., 2000; Ward, Hovard, Jones, & Rothen 2013; Banissy, Walsh, & Ward, 2009).
Synesthetic Stroop task
To test for the automaticity by which colour associations are evoked, subjects will perform a synesthetic Stroop task. In the synesthetic Stroop task we will test for colour interference (the difference in response times for incongruent versus congruent trials) by showing participants four letters which can be printed in four different personalized colours. We know a subject’s associated colour for every letter because we asked this in the colour picker when we assessed a subject’s consistency. In this study, all subjects (including the non-synesthetes) will see two letters for which they have the most consistent colour associations (i.e. the two most invariable letter-colour associations). Additionally, all subjects will see two letters for which their associated colours are most dissimilar (i.e. distant in colour space like red-green). All letters can be printed in all four colours, so the two “consistent” letters can be printed in a subject’s congruent synesthetic or associated colour, or in a colour that is incongruent with a subject’s synesthetic or associated colour (see Figure 1). The two “distant” letters can also be shown in a subject’s congruent synesthetic or associated colour, or in a colour that is incongruent with the synesthetic or associated colour. But different than the consistent letters, the colour of the distant letters can also be printed in a colour that is opponent in CIELUV colour space to the congruent colour of the letter to test for a colour opponency effect.
Figure 1.
An example of a subject’s letter-colour associations in this study.
The two letters for which the associated colours are most distant in colour space (“A” and “T”) can be printed in a subject’s congruent synesthetic or associated colour for those letters (red and green), in colours opponent to the associated colours (green and red), or in arbitrary colours independent of the associated colours (grey and orange). The two most consistent letters (“O” and “G”) can be printed in a subject’s congruent synesthetic or associated colours for those letters (grey and orange) or in colours independent of a subject’s associated colours (red, green and orange/grey). Table 1
Summary of the measures used to differentiate between the different groups. High consistency means strong letter-colour associations whereas low consistency means no or weak letter-colour consistency. Conscious awareness is the subjective report of experienced GCS whereby letters always cause a certain colour experience. Knowing or seeing is the subjective report of a subject’s experience when choosing colours with letters.
Consistency Conscious awareness Knowing or seeing
Definite synesthetes High Yes. Seeing.
Know-associators High Yes and No. Knowing.
Self-Claimed synesthetes Low Yes. Seeing.
The consistent condition is the difference in reaction times between the congruent and incongruent trials when subjects view consistent letters. The distant opponent condition is the difference in reaction times between congruent trials and incongruent opponent trials (i.e. the incongruent colour is opponent in colour space to the synesthetic/associated colour) when subjects view distant letters. The distant independent condition is the difference between congruent and incongruent independent trials (i.e. the incongruent colour is independent of the synesthetic/associated colour) when subjects view the distant letters. By looking at the difference in interference effects between the distant opponent condition and the distant independent condition, we can directly assess if there is a colour opponency effect and thus if the interference effect is perceptually evoked.
We expect that definite synesthetes and know-associators have strong semantic colour associations for letters and therefore we expect that they will show interference effects in the consistent condition and the distant conditions (see Table 2). But we expect that know-associators show smaller interference effects in all conditions than definite synesthetes as they only have a semantic mismatch and not a perceptual mismatch. Since self-claimed synesthetes have a perceptual and a semantic mismatch in the consistent condition, we expect them to show an interference effect, which we expect to be of similar size as definite synesthetes. Self-claimed synesthetes do not necessarily have letter-colour associations for the distant letters because of their relatively weak overall letter-colour consistency, therefore we expect that they will not show any interference effects in the distant conditions and that they perform similarly as non-synesthetes. Finally, we do not expect that know-associators show a colour opponency effect; a larger interference effect when distant letters are printed in the opponent colour relative to them being printed in an independent colour, in contrast to definite synesthetes.
Colour memory
We will test colour memory with two memory tasks; one task with meaningless stimuli (squares) where subjects are asked to match the colour of a memorized square and one task with meaningful visual stimuli (natural scenes; identical to a task of Ward et al., 2013) where subjects have to discriminate between old and new images, measuring both working memory and long-term
memory respectively. We expect that definite synesthetes and know-associators show better colour memory than non-synesthetes, measured by accuracy in the task with meaningful visual stimuli and by colour distance in the task with meaningless visual stimuli. Since the letter-colour associations of self-claimed synesthetes are less consistent than definite synesthetes, we think that they rely less on memory processes to experience synesthesia and self-claimed synesthetes are therefore expected to perform worse than definite synesthetes in the colour memory tasks.
Colour perception
To test for the sensitivity of colour perception, we created a colour matching task where subjects were asked to match the colour of a square with the colour of another on-screen visible square (similar to the colour matching task of Arnold et al., 2012). As it is found that GC synesthetes show enhanced perceptual colour sensitivity (Banissy et al., 2009), we can examine whether self-claimed synesthetes and know-associators show a similar trend. We expect to find that self-claimed synesthetes have equal perceptual colour sensitivity as definite synesthetes and therefore these two groups perform equally well on the colour matching task and both perform better than non-synesthetes, measured by colour distance. Since know-associators state to only know the colour of a letter without seeing it, we expect them to have a similar perceptual colour sensitivity as non-synesthetes and worse than definite synesthetes.
Research aims
By testing subjects on colour interference-, colour memory- and colour perception tasks, we can examine whether self-claimed synesthetes and know-associators show equal “synesthetic-like” behaviour as definite synesthetes, thus all should be considered as GC synesthetes. But more importantly, these tasks give us the opportunity to see whether consistency scores and the conscious awareness of having GCS are necessary when defining GCS. Namely, we think that a possible explanation for the variety of results from studies that try to explain GCS is that these studies fail to acknowledge the different subtypes of GC synesthetes and count these all as a single group of GC synesthetes or non-synesthetes nevertheless.
Table 2
Hypotheses of the synesthetic Stroop task per group.
Interference effect in the consistent condition
Interference effect in the distant conditions
Colour opponency effect in the distant condition
Definite synesthetes Yes. Yes. Yes.
Know-associators Yes (but weaker). Yes (but weaker). No.
Self-claimed synesthetes Yes. No. No.
Methods
ParticipantsA hundred twenty two Dutch speaking subjects (mean age ± SD: 32.8 ± 14.3 years; range 18 - 73 years) of which ninety-nine were women, participated in this study. Participants were recruited through various means, including social media and adverts on internet websites in which we explicitly stated that we were interested in participants who associate colours with letters. Furthermore participants could be recruited via the University of Amsterdam lab website and these participants were naive about the nature of this study. We grouped subjects into four groups; definite synesthetes (N = 59), know-associators (N = 13), self-claimed synesthetes (N = 25) and non-synesthetes (N = 25) based on a subject’s consistency score and on their subjective idea of having GCS and if so, whether this feels like seeing or knowing. A subject’s consistency was assessed through a colour picker where subjects were asked to give their associated colour for each individual letter three times in a single session. By looking at the invariance of the three repeated measurements of each letter-colour association in CIELUV letter-colour space, we distinguished between high and low consistency using a cut-off score of 135 based on the recommendations of Rothen et al. (2013). A lower score (<135) indicates strong letter-colour consistency whereas a higher score (>135) indicates no or weak letter-colour consistency. Furthermore, subjects were asked whether their experiences when choosing colours with letters felt like knowing or seeing (asked positively and negatively). Subjects with a score similar to, or lower than 1.5 on a Likert scale ranging from 1 (knowing) to 5 (seeing), which is more than 2 SD’s lower than the average of definite synesthetes (<1.62, see Figure 2), were classified as know-associators.
Figure 2.
Experienced feeling of seeing or knowing colours with letters.
Amount of subjects that responded whether the colour associations felt more like knowing or seeing on two questions of the CPQ. A score of 1.0 indicates knowing whereas a score of 5.0 indicates seeing. Definite synesthetes are visualized in blue and know-associators are visualized in orange.
In addition, we gave subjects a definition of GCS and we asked them if they have similar experiences whereby letters always cause a certain colour experience (based on the synesthesia battery; Eagleman et al., 2007). It has been found that GC synesthetes can differ from each other on certain tasks, for instance projectors and associators on the synesthetic Stroop task (Dixon et al., 2004). Projectors and associators were distinguished from each other by scores on the PA (projector-associator) questionnaire (Rouw & Scholte, 2010) where scores above zero indicate projector-like GCS and scores below zero indicate associator-like GCS. So before every analysis, we checked whether projectors perform similarly as associators and whether they can both be counted as a single group of definite synesthetes. If projectors performed differently than associators, we conducted the analyses with projectors and associators as different groups.
Fisher’s exact test revealed that there was no difference in gender, or education between the formed groups. However Fisher’s exact test revealed that there was a difference in the age between the formed groups, as the non-synesthetes (M = 21.8) are significantly younger than the other groups (M = 36.0). Therefore age was a covariate in all analyses conducted in this study.
Procedure
The experiment consists of 6 smaller tasks (or 7, which we will explain later), all build in Qualtrics survey software, version 072020 (Qualtrics, Provo, UT) to enable online participation. Subjects could have stopped before all tasks were finished, hence we mention how many subjects participated in every task. At the beginning of the experiment, subjects were instructed to perform the test on a desktop or laptop, put the brightness of their screen to the highest level and turn off any colour adjustment (e.g. night mode). Furthermore subjects were asked to perform the experiment in a quiet place without any distractions.
Colour picker
The subjects started the experiment with a colour picker (adapted from Eagleman et al., 2007) and provided their associated colour for each individual letter three times (i.e. 78 trials total). An example of the colour picker is shown in Figure 4. The first 26 trials started with asking participants if they experience a colour with a letter and if so, whether they experience one or more colours and if there is something special about the letter except the colour. Then participants were asked to give the “best” colour for each letter without overthinking the decision and/or using cognitive strategies. Thereafter participants were asked how certain they were about their choice by using a slider that could be moved from 0 (“I felt like I was guessing”) to a 100 (“I am certain that the letter is in the right colour”). After the first 26 trials, all letters appeared for the second and third time, but in these trials participants only were asked to give the best colour for a letter. Between the first
block (trial 1-26) and the middle block (trial 27-52), and between the middle block and the last block (trial 53-78), participants were asked to give best “pure” example of the colours: red, green, blue, yellow, pink and cyan to ensure that subjects had accurate vision. The colour selection in RGB colour space was recorded for the three repeated measurements of each colour association for a particular letter (e.g. A, A and A). These were converted into CIELUV colour space, which enables us to calculate the colour distance between each letter-colour association (see Figure 3; Rothen et al., 2013). Letter-colour associations that are perfectly similar to each other have an Euclidian colour distance of 0, whereas prototypical red (#ff0000) and prototypical green (#008000) have a difference of 220 in CIELUV colour space.
Figure 3.
Formula to calculate Euclidian distances in CIELUV colour space.
Formula to calculate Euclidian distances in colour space for the three repeated measurements of each letter-colour association. i refers to all possible combinations over the three trials (i.e. trial 1-2, trial 1-3 and trial 2-3). Retrieved from Rothen et al., 2013.
Individual trials were excluded when the three associated colours were the brightest form of black (#000000) or the brightest form of white (#ffffff), due to its major impact on consistency scores and the ease by which participants could choose these two colours. From this initial dataset, we excluded participants that did not change the colour more than 80% of the time (suggesting that they did not take the task seriously) or participants who chose black more than 80% of the time (indicating that they misunderstood the task and chose the printed colour of the letter).
Questionnaires
After this task, participants were given the colour picker questionnaire (CPQ; see Appendix A.1) that asked about their experiences choosing colours with letters (12 questions) with possible answers on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). These questions are subdivided into two blocks so that similar questions (asked positively and negatively) are not in the same block and participants cannot compare their previous answers. A block of demographic questions (gender, age, handedness, educational attainment, visual acuity, hearing and medication use) was presented between these blocks. Furthermore, the order of these blocks was counterbalanced between participants. These questions were designed so that all participants, whether they have consistent colour associations or not, could be asked about the nature of their experiences choosing colours for letters. The questions measure the process of choosing colours for letters in terms of the 1) certainty of choice, 2) specificity, 3) interference, 4)
automaticity, 5) deliberate use and 6) whether it felt like seeing or knowing.
Next, the participants were given a definition of GCS and were asked if letters cause them to have a colour experience, based on a similar question from the synesthesia battery (Eagleman et al., 2007; see appendix A.2). If they responded “yes”, subjects were sent through to the PA questionnaire of Rouw and Scholte (2007; see appendix A.3), the Coloured Letters and Numbers (CLaN) questionnaire of Rothen et al. (2013; see appendix A.6) and two self-designed questionnaires (see appendix A.4 & A.5). The factors measured in the CLaN questionnaire are localisation (6 questions), automaticity (4 questions), deliberate use (3 questions) and longitudinal changes (3 questions). The PA questionnaire consisted of six “projector” type questions and six “associator” type questions from which a PAscore can be calculated (ranging from -4 to -4) by subtracting the “associator” type questions from the “projector” type questions. The PA questionnaire and the CLaN questionnaire were semi-randomized so that similar questions (in terms of positively/negatively asked and questions measuring the same factors) did not appear more than twice in a row. On both questionnaires, subjects were asked to rate how strongly they agreed with each statement on a Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).
Furthermore, subjects were given four questions in which they were asked to state how much they “confuse” synesthetic colours with real colours (see appendix A.4), with two type of questions asked positively and negatively. Two of these questions were presented at the start of the PA questionnaire and two were given at the end of the PA questionnaire. Subjects could give answers on a five-point Likert where 1 referred to “strongly disagree” and 5 to “strongly agree”. A separate questionnaire consisting of three questions was presented where subjects were asked to compare the intensity of synesthetic colours with real colours (see appendix A.5) on a five-point Likert scale ranging from 1 (clearly the real colour) to 5 (clearly the synesthetic colour). At last subjects were asked whether they experience different forms of synesthesia (see appendix A.7) where we presented eight statements about different forms of synesthesia on which subjects could respond “yes”, “maybe”, or “no”.
Synesthetic stroop task
In the synesthetic Stroop task we showed subjects four letters which could be printed in four different idiosyncratic colours (see Figure 2). We could measure the colour distance of each letter-colour association based on CIELUV letter-colour distance in the letter-colour picker task. A JavaScript code was written so that it transferred the two letters with a subject’s most consistent associated colour from the colour picker task (e.g. the letters “O” and “G” and their associated colours “grey” and “orange”; see Figure 2). Additionally, the two letters for which a subject’s associated colours were most
distant in colour space (e.g. the letters “A” and “T” and their colours “red” and “green”) were transferred from the colour picker. The consistent letters could be presented in the congruent associated synesthetic colour, or in the three other incongruent not-associated colours. The distant letters could be printed in the associated synesthetic colour (congruent) or in an arbitrary colour, independent of the synesthetic associated colour (incongruent independent). But it could also be printed in the colour that is opponent in colour space to the associated synesthetic colour (incongruent opponent). All letter-colour combinations were presented in 10 trials, adding up to 160 trials in total. A JavaScript code was written so that the letters are at least 40 units apart in CIELUV colour space to ensure that colours were dissimilar enough. Three subjects automatically skipped the Stroop task as they chose consistent colours that were within 40 units in CIELUV colour space. A blank screen was presented for 500 ms, followed by a fixation cross for 1000 ms. The letters and the fixation point were presented centrally on the screen (font = 80px) on a light purplish background (RGB = #ede6f5) which was chosen because synesthetic and associated (non-synesthetic) colours tend to be (almost) absent in this part of the colour spectrum (Rouw & Root, 2019). The subjects could submit an answer by pressing one of the four key buttons (1, 2, 9, and 0) that corresponded to the four colours. Which key button corresponded to what colour (i.e. a consistent or distant colour) was intermixed between subjects to prevent any possible effects of handedness on the reaction time. The coloured letter remained on screen until the participant responded. Before the Stroop task began, participants were trained on the colour-button responses in order to accurately record reaction times during the Stroop task. Participants had to learn which of the four buttons corresponded to the four colours used. Participants completed one round of 32 trials with an aid on the bottom of the screen (that helped them to remember what key button corresponds to what colour) and one round of 32 trials without the aid. After the training session, the subjects were instructed that this would be the last opportunity to memorize which key button corresponds to what colour. There was no extra opportunity for subjects to practice. Results revealed that subjects had a correct response 95.3% of the time, indicating that two practice rounds of 64 trials in total are sufficient to learn which button corresponds to what colour.
All trials were presented in random order in a single block with a total of 160 trials (40 congruent trials, 120 incongruent trials). Participants were instructed to be as fast and accurate as possible. Only the correct answers were included in the reaction time analysis. Reaction time outliers were removed when they were more than 1.5 times the interquartile range (IQR) of that subject (Ratcliff, 1993), which resulted in 5.5% of the amount of correct trials being removed (congruent trials = 4.5%, incongruent opponent trials = 7.6% and incongruent independent trials = 5.4%).
When the cut-off was set to 2.5 times the IQR, the amount of removed incongruent opponent trials remained to be 1.6 - 1.7 as large as the number of removed congruent trials, hence we chose to keep the IQR at 1.5. No subjects had an average reaction time more than three times the IQR and therefore no subjects were removed from analysis.
Colour sensitivity (matching task)
The next task participants were asked to do was a colour matching task, measuring perceptual colour sensitivity (see Figure 4). In this task, subjects were instructed to match the colour of a square on the right of the screen with a square on the left of the screen. Subjects could adjust the colour of the square with an identical colour picker as when they were asked to give colour associations with letters. The squares (100px x 100px) were presented in front of a colour mask (300px x 100px) consisting of 150 colours representing the RGB colour palette. The colour hue and luminance of the left square were set so that subjects were presented 24 unique combinations. These were presented in 24 trials; 6 primary colours (red, yellow, green, blue, cyan and magenta), 6 fully-saturated non-primary colours (halfway in between each of those), 6 darker versions of the primaries (half of their luminance) and 6 darker versions of the non-primaries. Performance could be measured by calculating the difference in CIELUV colour space between the colour of the presented left square and the colour of the adjusted right square. CIELUV colour space was calculated in the same way as when we calculated letter-colour consistency (see figure 3), but there is only one set of combinations in this task (i.e. trial 1-2).
Figure 4.
Image of the colour sensitivity matching task.
Subjects were asked to match the colour of the square on the right to the colour of the square on the left (the image above). Subjects could do so by using a colour picker (the image below) which is similar to the colour picker when we assessed a subject’s consistency.
The colour space distances of all 24 trials were averaged whereby lower scores indicate better perceptual colour sensitivity. The best performance was an average colour distance of 0, meaning the colour of the adjusted squares are all identical to the presented squares. In contrast, an average colour distance of 454 indicates that the adjusted squares were most dissimilar in colour to the presented squares. Trials in which the colour space distance was more than 1.5 the IQR of an individual subject were removed which resulted in 1.9% of the trials being removed. Three subjects were removed because their average colour distance was more than tenfold the IQR and their reaction times were all in the 95th percentile of the slowest reaction times of all subjects, indicating that they did not take the task seriously.
Colour memory (natural scenes)
The colour memory task is identical to a task used by Ward et al. (2013) where subjects see thirty images and are asked to pay attention to the colours used in the images. A blank screen was presented for 1000 ms, followed by a fixation cross for 1000 ms and then followed by the image for 1000 ms. Hereafter subjects were shown 30 images; 15 of them were images which the participants already saw and 15 of them were new (see Figure 5). Participants were instructed that - although the images look familiar - the colour of an object in the scene could have changed. Subjects were asked to press the key button “9” when they thought the image was old and press “0” when they thought the image was new. Performance is measured by the number of correct answers, where higher scores indicate better colour memory. Possible scores range from 0 to 30. The subject that performed the task the worst had a score of 15. In contrast, the subject that performed the task the best had a score of 29. No subjects performed three times the IQR but one subject was removed because her median response was 64 ms. Three subjects had to be removed because their answers were not saved due to a technical error.
Figure 5.
Example of two images of the colour memory task with natural scenes. Subjects were shown one of the two images, where after they were shown the same or the different image. Subjects had to state whether they already saw the image, or whether an object in the image changed colour and they had not seen the image before.
Colour memory (colour matching task)
Next, the subjects were asked to do a colour matching task, similar to the colour sensitivity matching task described earlier. However, in this task subjects saw a centrally presented square (100px x 100px) for 5000 ms and were instructed to memorize the colour of the square. After the square disappeared, a mask (similar to the mask described in the colour sensitivity task, but now with size 100px x 100px) was presented for 1000 ms. After the mask disappeared, a centrally presented square appeared again and subjects had to match the colour of the memorized square with the square presented on-screen (the procedure to do so is similar as described in the colour sensitivity task). There were 24 trials and the colours of the squares that needed to be memorized were similar to the colours used in the colour sensitivity task. Performance could be measured by calculating the difference in CIELUV colour space between the colour of the memorized square and the colour of the adjusted square presented on-screen. Colour space distance of all 24 trials were averaged whereby lower scores indicate better colour memory. Trials in which the colour space distance was more than 1.5 the IQR of an individual subject were removed which resulted in 2.1% of the trials being removed. Three subjects showed a performance more than tenfold the IQR and showed the lowest average reaction time of all subjects, hence we chose to remove them (two of whom were also excluded in the colour perception matching task). One subject had an average colour distance 3.04 times the IQR. But this subject appeared to execute the task seriously, seen in both the absence of individual outliers and an average RT that is similar to the mean of other subjects, hence we chose to leave the subject in.
Mental imagery
Halfway through when the experiment was running, a last seventh task was added. This task consists of the SUIS (Reisberg, Pearson, & Kosslyn, 2003) and the VVIQ (Marks 1973). The SUIS is a self-report scale measuring the ability to use mental imagery during daily life and the VVIQ is a self-report scale measuring the ability to form clear mental images. These tasks will not be discussed in this paper.
Statistics and analyses
A subject’s group membership was based on the features mentioned in Table 1. Group membership was the independent variable in all analyses, which were analysed using SPSS (IBM SPSS Statistics for Windows, Version 26.0). In all tests, skewness, kurtosis and Shapiro-Wilk statistics were inspected for each group. Homogeneity of the variance was tested with Levene’s statistic and post-hoc analyses were executed to detect differences between groups using Bonferroni when equal variance was assumed (with an α of .05). Before every analysis, projectors and associators were compared to see if they performed the task similarly and thus could
both be analysed as a single group of definite synesthetes, or whether they performed the task differently and had to be analysed as separate groups.
To analyse if a group experienced an interference effect in the synesthetic Stroop task, a paired sample t-test was executed to compare mean reaction times between congruent and incongruent independent or incongruent opponent trials. A 2x4 (or 2x5) one-way ANCOVA was used to examine if the possible interference effect (in the consistent condition and the distant conditions) differed between groups, after accounting for the effects of age. To examine if there was a colour opponency effect, an one-way ANCOVA was used in which the interference effects between the distant independent and distant opponent conditions was compared for the distant letters, after controlling for the effect of age. The two matching tasks (measuring colour perception and colour memory) were both analysed using an one-way ANCOVA with group membership as the independent variable, age as covariate and performance on the two tasks, measured in CIELUV colour distance, as the dependent variable. The colour memory task with natural scenes was analysed using a Kruskal-Wallis one-way ANOVA with score as dependent variable. To assess in what proportion letter-colour consistency and conscious awareness of having GCS were related to the performance on all tasks, Pearson’s correlations were performed on the synesthetic Stroop task and the colour matching tasks, and Kendall’s tau-b was performed on the colour memory task for natural scenes. To determine statistical significance, the alpha level was set to .05 (two-tailed) for all tests.
Results
Synesthetic Stroop task
Before the analysis, we first examined whether projectors and associators show similar interference effects in all the three conditions. Three one-way ANOVA’s revealed that projectors and associators performed similarly in the consistent condition, F(1, 56) = .17, p = .734, the distant opponent condition, F(1, 56) = .35, p = .557 and the distant independent condition, F(1, 56) = .08, p = .779, hence we performed all analyses with projectors and associators as a single group; definite synesthetes. Inspection of the data revealed that four definite synesthetes, one know-associator and one self-claimed synesthete had a reaction time in the consistent condition which was more than three times the IQR. We chose to winsorize these values to three times the IQR (Hoo et al., 2002) since they all represent slower reaction times for incongruent trials compared to congruent trials. 58 Definite synesthetes, 13 know-associators, 21 self-claimed synesthetes and 22 non-synesthetes participated in this task. The assumption of normality was violated for the data of definite synesthetes in the consistent condition, W(58) = .885, p < .001, the distant opponent condition, W(58) = .893, p < .001 and the distant independent condition, W(58) = .853,
p < .001. Know-associators also showed deviations from normality in the consistent condition, W(13) = .831, p = .016, the distant opponent condition, W(13) = .866, p = .046 and the distant independent condition, W(13) = .825, p = .014. According to the central limit theorem, a sample larger than 50 subjects will behave like a normally distributed sample even when the distribution is skewed (Nixon, Wonderling, & Grieve, 2010), a sample size which definite synesthetes exceed (N = 58). In addition, small to moderate deviations from normality are not extremely worrisome, as Mena et al. (2017) found that an ANOVA is still robust when the data is not normally distributed but likely represents real data. Violation of normality was only slightly violated for know-associators, hence we decided that an ANCOVA is the appropriate statistical test for this task.
The consistent letters could be shown in a subject’s congruent associated/synesthetic colour, or in a colour that is incongruent with the associated/synesthetic colour. The difference in reaction time between the congruent and incongruent trials is the interference effect in the consistent condition. A paired sample t-test revealed that definite synesthetes had a statistically significant interference effect of 88.1 ms (SD = 161.5 ms) in the consistent condition, t(57) = -4.15, p < .001 (see Figure 6). Know-associators had an interference effect of 70.1 ms (SD = 117.3 ms) in the consistent condition, which was not statistically significant but appeared to be a trend, t(12) = -2.15, p = .052. Self-claimed synesthetes had an interference effect of 1.2 ms (SD = 67.8 ms) and non-synesthetes had an interference effect of 30.3 ms (SD = 93.6 ms) in the consistent condition, both not reaching statistical significance.
Figure 6. Results from the consistent condition in the synesthetic Stroop task.
Definite synesthetes showed a significant delay (interference effect) between trials when the colours of the consistent letters are congruent with their synesthetic colours (in blue) and when the colours of the consistent letters are incongruent with their synesthetic colours (in orange). Know-associators showed a similar trend, but failed to show a significant difference. Error bars represent standarderrors of the means. Three Asterix’s means p < .001, and a trend (tr.) means .05 < p < .10.
An one-way ANCOVA revealed that there was no effect of age, F(2, 89) = .50, p = .483, or group membership, F(2, 89) = .53, p = 591, on the difference in reaction time in the consistent condition when know-associators were compared to non-synesthetes and definite non-synesthetes. The variance of the data of self-claimed synesthetes was smaller than other groups, F(2, 98) = 4.31, p = .016, hence a Welch test showed that self-claimed synesthetes did not statistically differ from non-synesthetes, but they had significantly smaller interference effects in the consistent condition than definite synesthetes, Welch’s F(2, 56.05) = 6.77, p = .004. Age did not affect this relationship, F(2, 96) = .72, p = .397.
The distant letters could be shown in a subject’s congruent associated/synesthetic colour, or in a colour that is incongruent with the associated/synesthetic colour (distant independent condition). But different than in the consistent condition, the colour of the distant letters could also be printed in a colour that is opponent in CIELUV colour space to the congruent colour of the letter (distant opponent condition). A paired t-test revealed that definite synesthetes had a statistically significant interference effect of 93.1 ms (SD = 150.7 ms) in the distant independent condition, t(57) = -4.15, p < .001, and that they had a statistically significant interference effect of 95.1 ms (SD = 133.8 ms) in the distant opponent condition, t(57) = -5.39, p < .001 (see Figure 7). Know-associators had an interference effect of 86.6 ms (SD = 160.8 ms) in the incongruent independent condition, which appeared to be a trend, t(12) = -1.94, p = .076, and they showed an interference effect of 114.9 ms (SD = 148.0 ms) in the incongruent opponent condition, which reached statistical significance t(12) = -2.80, p = .016. Self-claimed synesthetes had a statistically significant interference effect of 33.8 ms (SD = 60.7 ms) in the distant independent condition, t(20) = -2.56, p = .019, but they did not show a statistical interference effect (28.8 ms, SD = 80.6 ms) in the incongruent opponent condition, t(20) = -1.62, p = .116. Non-synesthetes showed an interference effect of 30.6 ms (SD = 79.8 ms) in the incongruent independent condition and 8.6 ms (SD = 111.1 ms) in the incongruent opponent condition, which were both not statistically significant.
Two one-way ANCOVA’s revealed that, after controlling for the effects of age, F(1, 32) = 4.38, p = .044, know-associators had a significantly larger interference effect than non-synesthetes in the distant independent condition, F(1, 32) = 6.26, p = .018 and in the distant opponent condition, F(1, 32) = 7.21, p = .011. Age did not affect the second ANCOVA when we compared know-associators with non-synesthetes, F(1, 76) = 1.82, p = .187. Furthermore, two one-way ANCOVA’s revealed that self-claimed synesthetes showed a significantly smaller interference effect in the distant opponent condition than definite synesthetes, F(1, 76) = 4.20, p = .044, which appeared to be a trend in the distant independent condition F(1, 76) = 2.89, p = .093. Age did not affect these relationships, F(1, 76) = 1.55, p = .217; F(1, 76) = .69, p = .408.
No other significant relationships were observed for age or group membership on interference effects in the distant consistent condition and the distant opponent condition. An one-way ANCOVA was conducted to examine the existence of an opponency effect by measuring reaction time differences in the distant independent condition and the distant opponent condition, with age as a covariate. No difference was found between know-associators (opponency effect of 28.3 ms, SD = 56.0) and definite synesthetes (opponency effect of 1.3 ms, SD = 74.2), F(1, 68) = 1.60, p = .211.
Figure 7. Results from the distant conditions in the synesthetic Stroop task.
Self-claimed synesthetes and definite synesthetes showed a significant delay (interference effect) between trials when the colours of the distant letters are congruent with their synesthetic colours (in green) and when the colours of the distant letters are incongruent with their synesthetic colours (in blue). Know-associators showed a similar trend, but failed to show a significant difference. Know-associators and definite synesthetes showed a significant interference effect between trials when the colours of the distant letters are congruent with their synesthetic colours (in green) and trials in which the colours of the distant letters are opponent in CIELUV colour space to the congruent colours of the distant letters (in yellow). One Asterix means p <.05, three Asterix’s means p < .001, and a trend (tr.) means .05 < p < .10.
Colour perception (colour matching task)
Similar to the synesthetic Stroop task, we first examined whether projectors and associators showed equal perceptual colour sensitivity. An one-way ANOVA indicated that projectors and associators performed similarly on the colour matching task, therefore we treated these two groups as a single group of definite synesthetes. 58 Definite synesthetes, 23 self-claimed synesthetes, 13 know-associators and 23 non-synesthetes participated in this task. Inspection of the data revealed that the distribution of definite synesthetes, W(58) = .903, p < .001, and self-claimed synesthetes, W(23) = .905, p = .032, was skewed to the right. But as we have described earlier, we believe that an ANCOVA is the appropriate statistical test for this task based on the central limit theorem and the robustness of ANOVA’s (Nixon et al., 2010; Mena et al., 2017).
An one-way ANCOVA indicated that there was no effect of age on colour sensitivity when we compared definite synesthetes, know-associators and non-synesthetes, F(2, 90) = .15, p = .699. But there was an effect of group membership on perceptual colour sensitivity, measured in CIELUV colour distance, F(2, 90) = 3.69, p = .029 (see Figure 8A). Further inspection with post-hoc pairwise comparisons indicated that definite synesthetes (M = 9.48, SD = 3.71) had better perceptual colour sensitivity than non-synesthetes (M = 12.13, SD = 4.68), p = .014. Although know-associators showed even better perceptual colour sensitivity than definite synesthetes (M = 9.41, SD = 2.91), they did not statistically differ from non-synesthetes (due to their small sample size) but showed a clear trend, p = .057. A second one-way ANCOVA in which we compared self-claimed synesthetes with definite synesthetes and non-synesthetes revealed that there was no effect of age, F(1, 100) = 1.04, p = .309, but also that after accounting for the effects of age, there was a significant effect of group membership on colour sensitivity, F(2, 90) = 5.23, p = .007. Post-hoc testing revealed that the effects of group membership were caused by self-claimed synesthetes (M = 12.20, SD = 5.70) having significantly worse perceptual colour sensitivity than definite synesthetes, p = .040.
Figure 8. Results from the colour perception (A) and colour memory matching tasks (B).
Definite synesthetes showed significantly better colour sensitivity than self-claimed synesthetes and non-synesthetes in the colour perception task. Know-associators showed a similar trend, but failed to show a significant difference. In the colour memory task, definite synesthetes and know-associators were better in matching colours retrieved from memory than non-synesthetes. Self-claimed synesthetes did not show any statistical differences, but they did show a trend when compared to definite synesthetes and non-synesthetes. Error bars represent standard errors of the means. One Asterix means p <.05, three Asterix’s means p < .001, and a trend (tr.) means .05 < p < .10.
Colour memory (colour matching task)
Again, we first examined whether projectors and associators showed equal colour memory on the colour matching task. There was no difference between projectors and associators, hence we treated them as a single group. 58 Definite synesthetes,
22 self-claimed synesthetes, 13 know-associators and 24 non-synesthetes participated in this task.
An one-way ANCOVA indicated that age was significantly related to colour memory when we compared know-associators with definite synesthetes and non-synesthetes, F(2, 95) = 7.23, p = .009. But also, after accounting for the effects of age, group membership was significantly related to colour memory, measured in CIELUV colour distance, F(2, 91) = 5.79, p = .004 (see Figure 8B). Post-hoc testing revealed that definite synesthetes (M = 19.62, SD = 5.37) and know-associators (M = 20.53, SD = 5.72) remembered colours significantly better than non-synesthetes (M = 23.12, SD = 7.76), p = .001; p = .014. A second one-way ANCOVA revealed that age, F(1, 99) = 10.35, p = .002, and group membership after accounting for the effects of age, were statistically significant when we compared self-claimed synesthetes with definite synesthetes and non-synesthetes, F(2, 99) = 6.75, p = .002. Post-hoc comparisons revealed that the significant effect of group membership was driven by the difference between definite synesthetes and non-synesthetes, p = .002. However self-claimed synesthetes (M = 22.09, SD = 5.64) did show a trend when they were compared with definite synesthetes, p = .065, and with non-synesthetes, p = .058, with the effect of age being controlled for.
Colour memory (natural scenes)
Projectors and associators performed similarly on the colour memory for natural scenes task, as there was no significant difference between the two groups on the amount of images correctly memorized, as indicated by a Mann-Whitney U test. Therefore we treated these two groups as a single group of definite synesthetes. 57 Definite synesthetes, 13 know-associators, 22 self-claimed synesthetes and 24 non-synesthetes participated in this task. The data for all the groups were normally distributed and there were no outliers. There were no significant results between definite synesthetes (Mean Rank = 65.35), know-associators (Mean Rank = 54.00), self-claimed synesthetes (Mean Rank = 54.28) and non-synesthetes (Mean Rank = 51.15), although it appeared as definite synesthetes have a better long-term memory for colours than non-synesthetes, p = .090, indicated by a Kruskal-Wallis test.
Letter-colour consistency and conscious awareness
When groups showed different performances on our tasks, this could be caused by the presence or absence of letter-colour consistency and/or the conscious awareness of having GCS. Furthermore, it could also be that group differences can be explained by other factors, such as the localisation of the synesthetic experiences (PA questionnaire; Rouw & Scholte, 2007). For instance, know-associators (M = -2.22, SD = 1.59) had a significantly lower PA score than definite synesthetes (M = -1.12, SD = 1.34), as indicated by an one-way ANOVA, F(1, 67) = 4.26, p = .043. Table 3 describes the characteristics of the formed groups.
